Binding of the beta2 adrenergic receptor to N-ethylmaleimide-sensitive factor regulates receptor recycling.

Following agonist stimulation, most G protein-coupled receptors become desensitized and are internalized, either to be degraded or recycled back to the cell surface. What determines the fate of a specific receptor type after it is internalized is poorly understood. Here we show that the rapidly recycling beta2 adrenergic receptor (beta2AR) binds via a determinant including the last three amino acids in its carboxyl-terminal tail to the membrane fusion regulatory protein, N-ethylmaleimide-sensitive factor (NSF). This is documented by in vitro overlay assays and by cellular coimmunoprecipitations. Receptors bearing mutations in any of the last three residues fail to interact with NSF. After stimulation with the agonist isoproterenol, a green fluorescent protein fusion of NSF colocalizes with the wild type beta2AR but not with a tail-mutated beta2AR. The beta2AR-NSF interaction is required for efficient internalization of the receptors and for their recycling to the cell surface. Mutations in the beta2AR tail that ablate NSF binding reduce the efficiency of receptor internalization upon agonist stimulation. Upon subsequent treatment of cells with the antagonist propranolol, wild type receptors return to the cell surface, while tail-mutated receptors remain sequestered. Thus, the direct binding of the beta2AR to NSF demonstrates how, after internalization, the fate of a receptor is reliant on a specific interaction with a component of the cellular membrane-trafficking machinery.

G protein-coupled receptors (GPCRs) 1 are a family of integral plasma membrane proteins that transduce signals into cells from diverse extracellular ligands. An almost universal feature of these receptors is their ability to be desensitized in response to prolonged exposure to ligand (1). This is achieved mainly by the action of GPCR kinases that only phosphorylate agonist-occupied receptor molecules, allowing ␤arrestin molecules to then bind and physically interdict further coupling of the receptor to heterotrimeric G proteins (2).
Receptor desensitization is usually accompanied by the rapid internalization of the receptor molecules. This serves two main purposes. Firstly, it allows a cell to resensitize its responsiveness to a ligand by trafficking the receptors through a series of intracellular vesicular compartments in which they are returned to a naïve state (i.e. dephosphorylated) before recycling them back to the cell surface (3)(4)(5)(6). Secondly, by trafficking the receptors to alternative vesicular structures in which they are degraded, a cell can permanently reduce the receptor density in the plasma membrane and thus diminish its responsiveness to subsequent exposures to the ligand (7)(8)(9).
The proportion of a GPCR that is targeted for recycling or degradation varies greatly between receptor types, as do the rates with which these two processes occur. Some GPCRs, including the ␤2 adrenergic receptor (␤2AR), are rapidly recycled back to the cell surface within minutes of being internalized (3, 10 -12). Other receptors are detained within the cell for much longer periods of time before they are recycled (13)(14)(15)(16)(17), while still others are mostly degraded (18,19). The postendocytic fate of some receptors appears to be determined by the mechanism by which they are internalized (20,21). For others it is an intrinsic property of the receptor molecule itself. This is demonstrated most clearly in situations where two types of receptors internalizing by the same mechanism subsequently suffer different postendocytic fates. For example, in HEK293 cells both the ␤2AR and ␦ opioid receptor are internalized through clathrin-coated pits, but while almost all the ␤2AR is then rapidly recycled, the ␦ opioid receptor is mainly targeted to lysosomes for degradation (18). Furthermore, studies utilizing receptor chimeras between the ␤2AR and vasopressin V2 receptor (V2R) (16) and between the substance P receptor and the thrombin PAR1 receptor (22) have shown that interchanging the carboxyl termini of receptors is sufficient to bestow the recycling or degradative sorting properties of one receptor onto the other. From such studies it has been hypothesized that there exist domains within the carboxyl termini of receptors that act as determinants of receptor fate. These domains are presumably amino acid sequences that act as specific binding sites for protein factors involved in sorting, trafficking, and recycling or degrading the receptor. One such domain, a serine cluster in the tail of the V2 receptor, has implicated the ␤arrestins as members of this group of factors (15)(16)(17). Upon agonist binding, phosphorylation of the V2R within the cluster promotes the formation of a stable complex between the recep-tor and ␤arrestin. The complex is subsequently internalized and targeted to a recycling pathway in which the receptor is detained within endosomal compartments for long periods of time before being returned to the cell surface. Furthermore, removal of such a serine cluster from, or its addition to, the tail of a receptor is sufficient to bestow on it fast or slow recycling kinetics.
A number of other proteins have been identified that interact with the tails of specific receptors (23)(24)(25). While in some cases such proteins have been shown to alter the rate of receptor internalization, their effects on receptor fate are largely unknown. One exception is the Na ϩ /H ϩ exchanger regulatory factor (NHERF), which binds to the tails of a small number of GPCRs including the ␤2AR (26 -28). Disruption of the interaction between the ␤2AR and NHERF is accompanied by a shift in the postendocytic sorting of the receptor from being rapidly recycled to instead being mostly targeted for degradation (28). It is likely that NHERF represents just one member of a large group of factors that influence the fate of specific sets of GPCRs. Identification of these factors is an important step in furthering our understanding of the complex mechanisms that regulate receptor fate. Here we describe the identification of such a fate-determining factor for the ␤2AR, the N-ethylmaleimide sensitive factor (NSF), and describe its ability to regulate the sorting of the internalized receptor.
Construction of GST Fusion Protein Expression Clones-Fragments and point mutations of the ␤2ARct flanked by EcoRI and SalI restriction sites were generated by PCR using the FLAG-␤2AR/pcDNA3 clone as template DNA. The PCR products were purified and digested with EcoRI and SalI and ligated into the pGEX-4T vector. The sequences and orientations of the clones were confirmed by automated DNA sequence analysis.
Yeast Two-hybrid Screening-A DNA fragment encoding amino acids 328 -413 of the human ␤2AR and flanked by EcoRI and SalI sites was amplified by PCR and ligated into the EcoRI and SalI restriction sites of the pAS2-1 vector DNA. This construct (pAS2-1/␤2ARct) was used as bait to screen a rat brain cDNA library in pGAD10 (CLONTECH). Plasmid pAS2-1/␤2ARct and the rat brain cDNA library were cotransformed into yeast strain PJ69 -4A using standard yeast transformation protocols from the manufacturer's instructions. Yeast were plated on selective medium (SD-Leu-Trp-His, ϩ 2 mM 3-aminotriazole) and allowed to grow for 4 -6 days at 30°C. Yeast colonies capable of growth were restreaked on selective medium plates (SD-Leu-Trp-Adenine or SD-Leu-Trp-His). Plasmid DNA was rescued from clones that exhibited growth on both selective media plates and were then transformed into DH5␣ Escherichia coli cells. Plasmid DNAs were purified, sequenced, and then identified using the BLAST search program at the National Center for Biotechnology Information web site.
Receptor Sequestration and Recycling Assay-HEK293 cells overexpressing FLAG-␤2AR, FLAG-␤2AR-412A, or FLAG-␤2AR-410A alone, or FLAG-␤2AR and FLAG-␤2AR-412A with NSF were stimulated with 10 M isoproterenol for various times. Agonist-induced receptor internalization was measured as the loss of cell surface FLAG epitopes available for M2 antibody binding by detection of a fluorescently labeled secondary antibody as previously described (31). For recycling experiments, 10 M of the antagonist propranolol was added to displace isoproterenol and return the receptors to an inactive state.
Confocal Microscopy-HEK293 cells were transiently transfected with FLAG-␤2AR or FLAG-␤2AR-412A and with NSF-GFP. Under serum free conditions, transfected cells were treated with or without isoproterenol at 37°C for various times. To look at receptor recycling, 10 M propranolol was added for an additional 20 min after 30 min of isoproterenol treatment. Cells were fixed for 20 min in 4% formaldehyde in PBS. After washing three times with PBS, cells were incubated with M2 anti-FLAG antibody (1:250) in 0.1% Triton X100 solution for 45 min. Receptors were then labeled with Texas Red-conjugated rabbit anti-mouse IgG (1:250) in 0.1% Triton X100 solution. NSF-GFP and Texas Red-labeled receptor were visualized with a Zeiss LSM-510 laser confocal microscope.
Fusion Protein Overlays and Western Blotting-4 g of GST-fusion proteins were resolved on 4 -20% SDS-PAGE gels and transferred to nitrocellulose filters. Filters were blocked with 5% w/v fat-free milk powder in Tris-buffered saline with Tween 20 (TTBS: 20 mM Tris, pH 7.4, 500 mM NaCl, 0.1% v/v Tween 20) and incubated overnight at 4°C in a solution containing 100 nM purified NSF. Blots were then washed three times with TTBS buffer and incubated with anti-NSF monoclonal antibody for 1 h at room temperature. After three washes with TTBS, filters were incubated for 1 h with horseradish peroxidase-conjugated anti-mouse secondary antibody (Amersham Pharmacia Biotech), washed again with TTBS, immersed in ECL reagent (Amersham Pharmacia Biotech), and exposed to x-ray film.

RESULTS
To identify potential binding partners for the tail of the ␤2AR, a yeast two-hybrid screen of a rat brain cDNA library was performed using as bait the GAL4 binding domain fused to residues 328 -413 (the entire carboxyl terminus) of the ␤2AR. In this manner three clones encoding the rat NSF (32,33) were isolated, which allow growth on selective media plates of yeast coexpressing both fusion proteins (data not shown). The interaction is detectable in vitro using an overlay assay where recombinant NSF binds to a GST fusion protein of the ␤2AR tail but not to a GST fusion of the ␤1 adrenergic receptor (␤1AR) tail or to GST alone (Fig. 1A). Furthermore, NSF coex-pressed in COS7 cells with epitope-tagged receptors is detected specifically in immunoprecipitates of the ␤2AR and not in those of ␤1AR or of control cells lacking a tagged receptor (Fig. 1B). Moreover, NSF coexpressed in cells with epitope-tagged receptors is only detected in immunoprecipitates of the wild type ␤2AR and not in those of ␤2AR mutants carrying deletions of 53 and 28 residues from the distal region of the tail (Fig. 2A). These data suggest that the carboxyl-terminal tail of the ␤2AR contains all the determinants required for the receptor to interact with NSF.
This hypothesis was tested further by comparing the NSF binding ability of ␤2AR and vasopressin V2 receptor chimeras, whose tails had been interchanged (Fig. 2B). NSF is detectable in immunoprecipitates of the V2 receptor chimera carrying the ␤2AR tail but is absent or only weakly detectable in immunoprecipitates of the wild type V2 receptor or of the ␤2AR chimera carrying the V2 receptor tail. These data demonstrate that the presence of the ␤2AR tail is sufficient to confer on a receptor the ability to bind NSF.
The site in the tail of the ␤2AR responsible for binding to NSF was mapped further using a series of GST fusion proteins of fragments of the ␤2AR tail in overlay assays with recombinant NSF (Fig. 2, C and D). Deletions of large sections of the proximal region of the tail do not affect its ability to bind NSF (residues 329 -359 or 329 -384 deleted), indicating that NSF interacts at a site within the last 28 residues of the tail. This finding is in agreement with the coimmunoprecipitation data from the ␤2AR truncation mutants, where deletion of the distal 28 residues of the tail also ablates NSF binding ( Fig. 2A). Indeed, a GST fusion consisting of just this distal region of the tail retains the ability to bind NSF (GST 385-413 in Fig. 2, C and D). Furthermore, a GST fusion protein of the ␤2AR tail lacking just the final 10 amino acids (GST 329 -403) no longer binds NSF. These data indicate that all the components of the NSF binding domain in the ␤2AR reside within the distal 10 residues of the tail, thus making it possible to precisely define the key residues responsible for this interaction. This was done utilizing the NSF overlay approach with GST-␤2AR tail fusions in which each of the last 10 residues was sequentially mutated to alanine. While mutations at the seven proximal sites have no effect on the ability of the tail to interact with NSF, mutations at any of the last three residues (S411A, L412A, or L413A) ablate NSF binding (Fig. 3A). Introduction of the equivalent mutations in to the tail of the ␤2AR greatly reduces the quantity of NSF detectable in immunoprecipitates of the ␤2AR from cells overexpressing both the proteins (Fig. 3B). Furthermore, another GPCR that has three residues at the end of its tail that are similar to those of the ␤2AR, the vasoactive intestinal peptide receptor (SLV rather than SLL), also binds NSF (data not shown).
It has been reported previously that the NHERF binds via a PDZ domain to the same region of the ␤2AR tail as is demonstrated here for NSF (27,28). NSF does not contain a PDZ domain, and therefore it must be interacting with the tip of the ␤2AR tail through a wholly different mechanism than that utilized by NHERF. In support of this, sequential mutation analysis of the ␤2AR tail has revealed different residue requirements for NHERF and NSF binding; in particular, an alanine at position 412 does not affect NHERF binding while it decreases NSF binding, and an alanine at position 410 ablates NHERF binding but does not affect NSF binding (shown here by cellular coimmunoprecipitations in Fig. 3, B and C and by in vitro overlay in Ref. 27). If the sites of interaction of NHERF and NSF do indeed overlap, we would expect that the binding of one protein to the ␤2AR tail should inhibit the binding of the other. We tested this by measuring what effect overexpression of NHERF has on the ability of NSF to coimmunoprecipitate with the ␤2AR (Fig. 4). Coexpression of both NHERF and NSF with the ␤2AR leads to a complete loss of NSF from receptor immunoprecipitates, while NHERF is now detected. This is supportive of the concept that NSF and NHERF share an overlapping binding domain within the tail of the ␤2AR and that NHERF is capable of competing with NSF for binding to this site.
Having demonstrated that NSF can bind to the tail of the ␤2AR, we then investigated the sites at which the interaction occurs in cells using a GFP-tagged NSF molecule and fluorescent-labeled antibodies for detecting the epitope-tagged ␤2AR. In unstimulated HEK293 cells, the majority of the ␤2AR is located at the cell surface in the plasma membrane, while NSF is evenly distributed throughout the cytosol with a small amount colocalizing with the ␤2AR at the plasma membranecytosol interface (Fig. 5). After agonist stimulation with (-) isoproterenol, the ␤2AR undergoes endocytosis into the cell to collect in small puncta, presumably endocytic vesicular structures that enlarge with prolonged exposure of the cells to agonist. The initial weak colocalization of the receptor with NSF becomes more pronounced with continuing exposure to agonist and occurs mainly at the vesicular structures within the cytosol (Fig. 5, lower panels). As would be expected for a mutant receptor that is unable to interact with NSF, no colocalization of the ␤2AR-412A and NSF is observed (Fig. 6). Prior to stimulation with agonist, the same pattern of receptor distribution is seen for ␤2AR-412A as for the wild type receptor; ␤2AR-412A is located primarily at the cell surface, and NSF is uniformly distributed in the cytosol. Upon agonist stimulation, ␤2AR-412A also undergoes endocytosis and collects in small vesicular structures; even after prolonged exposure to agonist, however, NSF fails to colocalize with the receptor at these sites.
The primary function ascribed to NSF is as an ATPase that acts as a key regulator of membrane fusion events occurring during vesicle trafficking and exocytosis (33)(34)(35). The interaction of NSF with the cytoplasmic tail of the ␤2AR therefore suggests that it may also play a more specific role in the trafficking of those vesicles containing the ␤2AR. Ablation of NSF binding to the tail of the ␤2AR might then be expected to lead to alterations in the rates of internalization or recycling of the receptor. To test this, the wild type ␤2AR and the mutant ␤2AR-412A, which shows decreased binding to NSF, were expressed in the absence or presence of overexpressed NSF, and the rate and extent of agonist-induced receptor internalization was measured. Subsequent treatment with the membranepermeable antagonist propranolol to displace agonist from the internalized receptors then allowed the rate of recycling to be determined (Fig. 7). The cells rapidly sequester the wild type ␤2AR following isoproterenol treatment, reaching 24% of the cell-surface receptor being internalized after 30 min. Overexpression of NSF enhances both the rate and the total amount of internalized receptors (to 38%) after 30 min of stimulation (Fig.  7A). Subsequent treatment of the cells with propranolol leads to rapid recycling of internalized receptors, with 90% of the receptors being detectable on the cell surface within 30 min (Fig. 7A). In the presence of overexpressed NSF, both the rate and extent of recycling are enhanced; 90% of the receptors are detectable on the cell surface just 10 min after antagonist treatment, and all the receptors are detectable on the cell surface after 30 min (Fig. 7A). This is accompanied by the complete loss of all the sites of colocalization of NSF and receptor at intracellular vesicles (Fig. 7C). A smaller fraction of the mutant receptor ␤2AR-412A is internalized following isoproterenol treatment, reaching only 63% of the level of internalization of the wild type ␤2AR after 30 min (Fig. 7A). Overexpression of NSF with ␤2AR-412A does not enhance the rate of receptor internalization, and nor does it lead to colocalization of the receptor and NSF (Fig. 7, A and D and previously in Fig.  6B). It should also be noted that mutation of the ␤2AR at this site does not affect the level of receptor expression at the plasma membrane prior to agonist stimulation. However, the most striking effect of perturbing the NSF-␤2AR interaction is seen after the addition of propranolol to the cells; the mutant ␤2AR-412A is completely inhibited in its recycling, with NSF overexpression having no discernable ability to rescue this effect (Fig. 7, A and D).
It has been reported that NHERF binding to the tail of the ␤2AR plays a role in determining the rate and amount of receptor that is recycled to the cell surface following internalization (28). The data presented here in Fig. 7A indicates that the binding of NSF to the tail of the ␤2AR is also important for receptor recycling. In fact, the recycling-deficient mutant ␤2AR-412A can still bind to NHERF (see Ref. 27 and Fig. 3, B and C), suggesting that NHERF binding may not be involved in ␤2AR recycling. To test further whether both NHERF and NSF are involved in receptor recycling, we compared the recycling capabilities of the wild type ␤2AR and the two mutant receptors that bind to only one of these two proteins: ␤2AR-412A that binds NHERF but not NSF, and ␤2AR-410A that binds NSF but not NHERF (Fig. 7B). The wild type receptor and NSF-binding D410A receptor both show equivalent levels of internalization after 30 min of agonist treatment, and both recycle back to the cell surface with the subsequent addition of the antagonist propranolol. However, ␤2AR-412A poorly internalizes and is unable to recycle following propranolol treatment. These data imply that the direct binding of the ␤2AR to NSF, but not to NHERF, is a required event for correct trafficking of the receptor during agonist-induced internalization and recycling. DISCUSSION The balance struck between the resensitization (recycling) and the down-regulation of desensitized GPCRs determines the magnitude and endurance of a cell's response to subsequent exposures to ligand. Studies have indicated that this balance varies greatly between receptors and is determined, in part, by signals within the receptor molecules themselves (15)(16)(17)22). These signals are presumed to be domains through which receptors interact with protein factors whose functions are to direct the sorting of internalized receptors for either recycling or degradation. Thus, through understanding the interactions that any given receptor makes with these factors it may be possible to determine its fate following internalization.
Here, we set out to identify proteins that, through direct interactions with the carboxyl terminus of the ␤2AR, could be serving such fate-determining functions. We identified one such candidate, the membrane fusion regulatory ATPase, NSF.
Through its interaction with ␣SNAP (soluble NSF attachment protein), NSF binds vesicular membrane and target membrane-specific proteins (␣SNAP receptors, SNARE's) to form a complex called the "20S particle" (33,34). This complex is a critical component of the machinery necessary to perform membrane fusion events during the processes of intracellular membrane trafficking and exocytosis.
It has been shown in a number of studies that following agonist stimulation the ␤2AR is one of the most swiftly internalized and recycled of the GPCRs (3, 10 -12). The ability of NSF to directly bind the ␤2AR may provide the basis of the mechanism underlying this rapidity; vesicles containing the internalized ␤2AR would have associated with them large quantities of NSF and the other 20S particle proteins, and thus they would preferentially be targeted for rapid trafficking. This is supported here by the demonstration that overexpression of NSF enhances the rates of internalization and recycling of the wild type ␤2AR but not of a mutant receptor that no longer binds NSF (Fig. 7). In fact, the failure of the mutated ␤2AR to recycle at all suggests that the binding of NSF to the receptor is essential for this process.
NSF also interacts with the ␤arrestins, two proteins that mediate GPCR desensitization by binding to agonist-occupied phosphorylated GPCRs and thereby inhibiting their coupling to G proteins (36). A further function of the ␤arrestins is as adaptors/scaffolds for recruiting a wide range of accessory proteins to GPCRs (37)(38)(39)(40)(41)(42)(43). While NSF is likely to directly interact with only a small number of GPCRs, its interaction with ␤arrestins may represent a mechanism by which it can influence the trafficking of all members of this receptor class. Whether recruitment of NSF to a receptor through direct binding or via a ␤arrestin scaffold results in differences in receptor fate awaits further investigation.
We also define a site at the tip of the ␤2AR tail that mediates binding of the receptor to NSF. This is overlapping with the site previously reported to mediate receptor binding to NHERF (27). However, while NHERF binds to the ␤2AR via a PDZ domain, NSF lacks such a domain so it must bind by an alternative mechanism. This is supported here by the demonstration that a mutation at position L412A in the receptor, which is known not to affect NHERF binding, ablates NSF binding. In addition, a mutation at position D410A that ablates NHERF binding does not affect NSF binding (27). The ␤2AR-NHERF interaction has been reported to affect postendocytic sorting of the receptor (28). Perturbation of the interaction is reported to prevent the receptor from recycling and to increase the proportion that is degraded after it is internalized. Here we show that the mutant ␤2AR-412A, which does not bind NSF but does still bind NHERF, is also defective in receptor recycling. Furthermore, we demonstrate that the mutant ␤2AR-410A, which still can bind NSF but cannot bind NHERF, recycles normally. Thus, NSF might in fact mediate some of the effects on receptor sorting that the ␤2AR-NHERF interaction has been previously reported to have (28).
It also has been recently reported that, like the ␤2AR, the human leutinizing hormone receptor (hLHR) undergoes rapid recycling (44). This process was shown to be dependent on the presence of a specific stretch of residues found near the tip of the receptor tail (GTALL), a sequence missing from the tail of the poorly recycling rat LH receptor (rLHR). Addition of this sequence, or the last four residues found at the tip of the ␤2AR tail (DSLL), to the rLHR bestowed on it hLHR recycling characteristics. Furthermore, this study demonstrated that NHERF does not play a role in hLHR recycling as manipulations that have been reported to disrupt NHERF function in ␤2AR recycling fail to affect hLHR recycling. It is interesting to note the sequence similarity that exists between the hLHR and the ␤2AR recycling determinants (GTALL and SLL, respectively). Both sequences contain a leucine dimer and a hydroxylcontaining residue (Thr or Ser) and may represent a conserved sequence through which NSF might bind and mediate receptor recycling. However, it remains to be shown whether NSF can bind the hLHR and mediate recycling.
The effect of NSF on ␤2AR recycling that we describe here parallels a recently revealed and hitherto unknown function for NSF in neurons where it acts as a chaperone during trafficking of the ␣-amino-3-hydroxy-5-methyl-isoxazolepropionate (AMPA) glutamate receptor (45)(46)(47). Rapid and continuous cycling of AMPA receptors to and from the postsynaptic membranes regulates the responsiveness of neurons to the presynaptic release of glutamate. Like for the ␤2AR, the recycling of the AMPA receptor is dependent on the direct interaction of the GluR2 subunits of the receptor with NSF (48,49). Manipulations that inhibit exocytosis or that specifically block the interaction between the GluR2 subunits and NSF prevent AMPA receptor recycling. This leads to a rapid reduction in receptor density in the postsynaptic membrane and a subsequent reduction in glutamate responsiveness of the neuron. The direct binding of NSF to both the AMPA receptor and the ␤2AR, therefore, is critical to allow these receptors to undergo rapid recycling, a process that is important for sustaining the signaling potential of the cells in which they are expressed. There is also evidence suggesting that the mechanism of membrane fusion underlying AMPA receptor recycling is somewhat divergent from that utilized by the usual 20Smediated pathway. While both NSF and ␣SNAP can be copurified with the AMPA receptor, SNAREs do not appear to be present in the same complex (47). Such an alternative SNAREindependent mechanism may only be employed for the trafficking of membrane proteins where rapidity of recycling is crucial for their correct function as is the case for both the AMPA receptor (50 -52) and the ␤2AR (3, 11, 12).
We have described here the identification of NSF as a ␤2ARinteracting protein and present evidence for a critical role of this interaction in the recycling of the receptor. Thus, NSF represents a new member of a class of proteins whose binding to receptors determines their fate after agonist-induced internalization.