Evolving Concepts in G Protein-Coupled Receptor Endocytosis: The Role in Receptor Desensitization and Signaling

1. Second Messenger-Dependent Protein Kinases.

The second messenger-dependent protein kinases, PKA and PKC, are phosphotransferases that catalyze the transfer of the γ-phosphate group of ATP to serine and threonine residues contained within specific amino acid consensus sequences of proteins. Second messenger-dependent protein kinase are activated in response to GPCR-stimulated increases in intracellular second messengers such as cAMP, Ca²⁺, and diacyglycerol and participate in GPCR signaling by mediating the phosphorylation of downstream target proteins. However, these kinases also feedback phosphorylate GPCRs at phosphorylation consensus sites within their intracellular loops and carboxyl-terminal tail domains. For example, β₂AR-activated PKA activity leads to receptor desensitization in response to PKA-mediated phosphorylation of at least one of two PKA consensus sites within the receptor (Bouvier et al., 1988; Yuan et al., 1994; Moffet et al., 1996). One site is found within the G protein-binding domain of the third intracellular loop of the β₂AR and the other site is found within the proximal region of the β₂AR carboxyl-terminal tail (Bouvier et al., 1988; Yuan et al., 1994; Moffett et al., 1996). It is proposed that covalent modification of the β₂AR at only the PKA site within the third intracellular loop domain contributes to receptor/G protein coupling (Yuan et al., 1994). However, phosphorylation of the PKA site within the β₂AR carboxyl-terminal tail occurs subsequent to the depalmitoylation of cysteine residue 341, suggesting that this site contributes to the agonist-dependent desensitization of β₂AR responsiveness (Moffet et al., 1996). Two recent studies suggest that PKA-mediated phosphorylation of the β₂AR involves the direct and constitutive association of the AKAP (AKAP79/150 and AKAP250) with the receptor (Fraser et al., 2000; Lin et al., 2000). PKC activation leads to the phosphorylation and desensitization of many G_i- and G_q-linked GPCRs (e.g., Diviani et al., 1997;Liang et al., 1998; Tang et al., 1998). Nonetheless, second messenger-dependent protein kinase-mediated mechanisms of receptor desensitization have received less attention than GRK-mediated mechanisms of receptor desensitization.

2. G Protein-Coupled Receptor Kinases.

a. The G Protein-Coupled Receptor Kinase Family.

The GRK family of kinases is comprised of seven family members that share significant sequence homology (reviewed by Premont et al., 1995;Stoffel et al., 1997) (Table1). Each of the GRKs share a similar functional organization with a central catalytic domain, an amino-terminal domain that is thought to be important for substrate recognition and that contains an RGS-like domain, and a carboxyl-terminal domain that contributes to the plasma membrane targeting of the kinase (Fig.1). The members of the GRK family can be subdivided into three groups based on sequence and functional homology: 1) GRK1 (rhodopsin kinase) (Shichi and Somers, 1978) and GRK7 (a new candidate cone opsin kinase) (Weiss et al., 1998); 2) GRK2 (β-adrenergic receptor kinase 1, βARK1) (Benovic et al., 1986) and GRK3 (β-adrenergic receptor kinase 2, βARK2) (Benovic et al., 1991); and 3) GRK4 (Premont et al., 1994; Sallese et al., 1994), GRK5 (Kunapuli and Benovic, 1993), and GRK6 (Benovic and Gomez, 1993).

View this table:

• View inline
• View popup

Table 1

Characteristics of GRK family members

View this table:

• View inline
• View popup

Table 2

Characteristics of arrestin family members

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Schematic representation of the domain architecture for GRK1–GRK7. The amino-terminal GPCR-binding domain of GRK1–GRK7 contains a conserved RGS domain (Carman et al., 1999). The plasma membrane targeting of each of the GRKs is mediated by distinct mechanisms that involves their carboxyl-terminal domains. GRK1 and GRK7 are farnesylated at CAAX motifs in their carboxyl termini (Inglese et al., 1992). The carboxyl-terminal domains of GRK2 and GRK3 contain a βγ-subunit binding domain that exhibits sequence homology to a pleckstrin homology domain (Pitcher et al., 1992; Touhara et al., 1994). The GRK5 carboxyl-terminal domain contains a stretch of 46 basic amino acids that mediate plasma membrane phospholipid interactions (Kunapuli et al., 1994; Premont et al., 1994). GRK4 and GRK6 are palmitoylated at cysteine residues (Stoffel et al., 1994, 1998). Figure adapted from Stoffel et al., 1997.

3. Other Kinases.

In addition to serving as substrates for PKA, PKC, and GRK phosphorylation, GPCRs have been shown to serve as substrates for phosphorylation by other protein kinases. Casein kinase 1a-mediated phosphorylation of the third intracellular loop domain of the m3 mAChR occurs in response to agonist activation of the receptor (Tobin et al., 1997; Budd et al., 2000). Moreover, casein kinase 1a-mediated phosphorylation of the m3 mAChR was blocked by either the expression of a catalytically inactive casein kinase 1a mutant or a peptide corresponding to the third intracellular loop domain of the m3 mAChR (Budd et al., 2000). Nonetheless, the functional consequence of casein kinase 1a receptor phosphorylation remains to be fully elucidated (Budd et al., 2000). The presence of casein kinase phosphorylation consensus motifs within the intracellular loop and carboxyl-terminal tail domains of many GPCRs makes these observations particularly intriguing.

There is evidence that tyrosine kinase-mediated GPCR phosphorylation may influence the activity of some GPCRs. Mutagenesis of tyrosine residues in the carboxyl-terminal tail of the μ opioid receptor reduced the agonist stimulated down-regulation of the receptor (Pak et al., 1999). This effect could be mimicked using the tyrosine kinase inhibitor genistein (Pak et al., 1999). Insulin-stimulated tyrosine phosphorylation of the β₂AR has also been reported (Valiquette et al., 1995). However, whereas carboxyl-terminal tyrosine residues contributed to β₂AR down-regulation, these residues were not the substrates for insulin-promoted tyrosine phosphorylation (Valiquette et al., 1990,1995). Tyrosine phosphorylation of the bradykinin B₂ receptor in response to agonist can be prevented by genistein treatment and seems to contribute to receptor signaling leading to arachidonic acid release (Jong et al., 1993). The identity of the tyrosine kinase(s) mediating the phosphorylation of these receptors is unknown.

1. The Arrestin Family.

To date, four arrestin family members have been identified (Table 1). The members of the arrestin family can be divided into two groups based on sequence homology, function, and tissue distribution: 1) visual arrestin (S antigen) (Shinohara et al., 1987; Yamaki et al., 1987) and cone arrestin (X-arrestin or C-arrestin) (Murakami et al., 1993; Craft et al., 1994) and 2) β-arrestins (β-arrestin1 and β-arrestin2) (Lohse et al., 1990a; Attramadal et al., 1992). Visual arrestin is a major protein constituent of rod outer segments and is localized primarily to the retina with low expression in the pineal gland (Smith et al., 1994). C-Arrestin is highly enriched in both retina and pineal gland, but is localized primarily within cone photoreceptors in the retina (Craft et al., 1994). The β-arrestins are ubiquitously expressed outside the retina, but are predominantly localized in neuronal tissues and in the spleen (Attramadal et al., 1992). In the rat central nervous system β-arrestin2 is more abundant than β-arrestin1 (Attramadal et al., 1992). The evaluation of β-arrestin1 and β-arrestin2 protein expression in the brain reveals extensive, heterogenous neuronal labeling (Attramadal et al., 1992). β-Arrestin protein is found in several neuronal pathways and immunoelectron microscopy reveals that β-arrestins are concentrated at neuronal synapses along with GRKs (Arriza et al., 1992; Attramadal et al., 1992). Thus, these proteins are ideally localized to modulate neuronal function. A third family of arrestin proteins might exist, since partial cDNA clones for D- and E-arrestin have been reported (Craft et al., 1994). However, although the mRNAs for D- and E-arrestin are expressed in a broad range of tissues, there is still question whether full-length D- and E-arrestin proteins truly exist (Craft et al., 1994).

Alternative splice variants have been identified for visual arrestin, β-arrestin1, and β-arrestin2. Bovine visual arrestin is expressed as a 404 amino acid residue protein, as well as two polypetide variants, one for which the last 35 amino acid residues are replaced by an alanine residue (p44) and another that lacks residues 338–345 encoded by exon 13 (Yamaki et al., 1987, 1990; Smith et al., 1994). The p44 visual arrestin variant is specifically localized to the rod outer segment and is severalfold more potent an inhibitor of rhodopsin signal transduction than the long form (Palczewski et al., 1994). Thus, the carboxyl-terminal domain of visual arrestin does not seem to be important for binding to rhodopsin. This observation has been confirmed using β-arrestin truncation mutants (Gurevich, 1998). However, it is the β-arrestin carboxyl-terminal domain that distinguishes these arrestin isoforms from visual arrestins (Fig.2, A and B). Similar to visual arrestin, at least two alternatively spliced forms of β-arrestin1 and β-arrestin2 are expressed (Parruti et al., 1993b; Sterne-Marr et al., 1993). The variant form of β-arrestin1 involves the insertion of eight amino acid residues between amino acids 333 and 334, and the variant form of β-arrestin2 involves the insertion of 11 amino acid residues between amino acids 361–362 (Parruti et al., 1993b;Sterne-Marr et al., 1993). Although the existence of alternatively spliced variant β-arrestin isoforms increases the potential number of functionally distinct β-arrestins, there are no reported differences in the functional activity of the β-arrestin variants. However, considering the limited number of arrestins and the preponderance of GPCR subtypes, it is likely that receptor specificity is governed by discrete differences in receptor structure and tissue-specific arrestin protein expression patterns rather than by a multitude of different arrestin isoforms.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Molecular architecture of arrestins. Panel A, the arrestin regulatory domains identified by the solution of the visual arrestin crystal structure (Hirsch et al., 1999) and mutagenesis studies (Gurevich et al., 1995). R1, amino terminal regulatory domain (residues 1–24); A, receptor activation domain (residues 24–180); P, phosphate sensor domain (residues 163–182); S, secondary receptor-binding domain (residues 180–330); and R2, carboxyl-terminal regulatory domain (residues 330–404). Underlined is the proline-rich region conserved in mammalian β-arrestin1 and β-arrestin2 but not visual arrestins (Luttrell et al., 1999). The arrow points to serine residue 412, which is phosphorylated by MAPK (Lin et al., 1999). The black box highlights the clathrin- and β-adaptin-binding domains that are conserved among nonvisual arrestins. Panel B, sequence alignment of visual and nonvisual arrestins from different species. The alignment highlights the functional differences between visual versus nonvisual arrestins that arise as the consequence of the clathrin (bold) and β2-adaptin (bold and asterisked) binding domains among nonvisual arrestins (Krupnick et al. 1997; Laporte et al., 2000).

2. Receptor Binding.

Arrestins preferentially bind to agonist-activated and GRK-phosphorylated GPCRs as opposed to second messenger protein kinase-phosphorylated or nonphosphorylated receptors (Lohse et al., 1990a, 1992). In vitro, the affinity of β-arrestin binding to the β₂AR is increased 10- to 30-fold by GRK phosphorylation (Lohse et al., 1992), and this selectivity is even more pronounced for visual arrestin binding to rhodopsin (Gurevich et al., 1995). GRK phosphorylation occurs within either the third intracellular loop domain (e.g., m2 mAChR and α_2AAR) or the carboxyl-terminal tails of receptors (e.g., rhodopsin and β₂AR) (Bouvier et al., 1988; Nakata et al., 1994; Eason et al., 1995; Brannock et al., 1999). Thus, to bind and interdict the signaling of multiple distinct GPCR subtypes, arrestins must exhibit the capacity to recognize and bind multiple receptor domains and conformations. This idea is supported by the following observations: 1) the interaction of visual arrestin with rhodopsin can be blocked by synthetic peptides representing the first and third intracellular loops of rhodopsin (Krupnick et al., 1994); 2) β-arrestin can be coimmunoprecipitated with the third intracellular loop domains of the m3 mAChR, α_2AAR, and 5-hydroxytryptamine2A receptor (Wu et al., 1997; Gelber et al., 1999); and 3) the β₂AR carboxyl-terminal tail is not absolutely required for β-arrestin binding (Ferguson et al., 1996b). Gurevich et al. (1995) examined the ability of several arrestins to bind to various functional forms of rhodopsin, β₂AR, and m2 mAChR. Although each of the arrestin isoforms demonstrated preference for binding to the GRK-phosphorylated agonist-activated form of the receptors, there was also substantial binding to phosphorylated nonactivated receptors, as well as agonist-activated nonphosphorylated receptors (Gurevich et al., 1995). This suggests that, depending on the GPCR isoform studied, agonist-independent β-arrestin association may be observed (Anborgh et al., 2000). The clear exception is visual arrestin, which binds selectively to only GRK-phosphorylated and light-activated rhodopsin (Gurevich et al., 1995). In addition, the site of phosphorylation within the rhodopsin carboxyl-terminal tail appears critical for visual arrestin-dependent quenching of rhodopsin activity (Brannock et al., 1999).

The recent solution of the crystal structure for arrestin, together with mutagenesis studies, has provided further insight into the molecular events involved in arrestin binding to phosphorylated light-activated rhodopsin (Gurevich et al., 1995; Granzin et al., 1998;Vishnivetskiy et al., 1999; Hirsch et al., 1999). With respect to receptor binding, mutagenesis studies revealed that the molecular structure of visual arrestin can be divided into three functional and two regulatory domains (Gurevich et al., 1995). The functional domains include: a receptor activation recognition domain (amino acid residues 24–180), a secondary receptor binding domain (amino acids 180–330), and a phosphate sensor domain (amino acid residues 163–182). The regulatory domains are comprised of an amino-terminal regulatory domain (amino acid residues 1–24) and a carboxyl-terminal regulatory domain (amino acid residues 330–404) (Fig. 2A). The solution of the visual arrestin crystal structure is consistent with these observations and reveals that visual arrestin is comprised of two major functional domains that are each constructed from a seven-stranded β sandwich (Granzin et al., 1998; Hirsch et al., 1999). The two domains, the N domain (amino acid residues 8–180) and C domain (amino acid residues 188–362), respectively, comprise the activation recognition and secondary receptor binding domains originally identified by mutational analysis (Fig. 2A). The carboxyl-terminal tail of visual arrestin (amino acids 372–404) is connected to the C domain by a flexible linker, and the carboxyl-terminal tail forms various interactions with parts of the arrestin N and C domains to regulate their structure. The phosphate sensor domain, identified by mutagenesis, constitutes a polar core that in the basal state is embedded between the N and C domains and forms the fulcrum of the arrestin molecule. It is likely that this core structure is highly conserved among arrestin isoforms. Residues from both the amino-terminal and carboxyl-terminal regulatory domains (Asp-30 and Arg-382) also contribute to the polar core of visual arrestin. It is predicted from both mutagenesis studies and the crystal structure for visual arrestin that the interaction of the carboxyl-terminal tail with the polar core stabilizes the basal state structure of visual arrestin (Gurevich et al., 1995; Hirsch et al., 1999). However, in response to receptor binding, the phosphorylated receptor tail invades the polar core, thereby disrupting polar residues and releasing the arrestin carboxyl-terminal tail. This leads to the reorientation of the N and C domains along the fulcrum formed by the polar core facilitating the formation of a receptor-arrestin complex (Hirsch et al., 1999). This model is consistent with the observation that the p44 visual arrestin isoform exhibits greater affinity for rhodopsin and that the mutation of polar residues within the polar core of the visual arrestin protein results in arrestin mutants able to bind nonphosphorylated rhodopsin (Palczewski et al., 1994; Vishnivetskiy et al., 1999). Thus, it can be concluded that the conformation of free arrestin has evolved to resist agonist- and phosphorylation-independent interactions with receptors. This may be further ensured by the formation of arrestin oligomers (Schubert et al., 1999).

1. Role of G Protein-Coupled Receptor Kinase and β-Arrestin Proteins.

Sibley et al. (1986) were the first to suggest that receptor phosphorylation might be involved in GPCR endocytosis. However, when this hypothesis was tested, using β₂AR mutants lacking sites for both second messenger-dependent protein kinase- and GRK-mediated receptor phosphorylation, no significant differences were observed between the internalization of wild-type and mutant β₂ARs (Bouvier et al., 1988; Hausdorff et al., 1989). A similar result was obtained when permeablized A431 cells were treated with PKA and GRK inhibitors (Lohse et al., 1990b). This led to a commonly held view that receptor phosphorylation did not contribute to GPCR endocytosis.

Despite the fact that phosphorylation was originally not considered to play a role in the internalization of the β₂AR, there was growing evidence that phosphorylation might be involved in the endocytosis of other GPCRs, such as the m2 mAChR (Moro et al., 1993). Early experiments demonstrated that the internalization of the m2 mAChR was reduced by the mutation of serine and threonine residues within the third intracellular loop domain of the receptor (Moro et al., 1993). Subsequently, Tsuga et al. (1994) demonstrated that the overexpression of wild-type GRK2 enhanced both the rate and maximal extent of m2 mAChR internalization, whereas a dominant-negative GRK2 mutant (K220W) impaired both the phosphorylation and the internalization of the receptor in COS7 cells. However, whereas catalytically inactive GRK2 dominant-negative mutants blocked m2 mAChR in COS7 cells, they had no effect on m2 mAChR internalization in BHK-21 and HEK 293 cells (Tsuga et al., 1994; Pals-Rylaarsdam et al., 1995). Thus, these experiments provided the first indication that differences in the cellular context in which dominant-negative mutants are used could result in discordant observations. Subsequently, it was determined that GRK2 protein expression levels vary from cell type to cell type, with lowest levels of GRK2 protein expression found in COS7 cells and substantially higher levels found in cells derived from the hematopoetic system (e.g., COS7 < HEK 293 cells < RBL-2H3 cells) (Aramori et al., 1997; Menard et al., 1997; Barlic et al., 1999). Therefore, the effectiveness of dominant-negative proteins may be dependent on the level of GRK protein expressed in the particular cell line utilized for experimentation.

The observation that GRK2 phosphorylation played a role in the internalization of the m2 mAChR, but that β₂AR mutants lacking sites for GRK2 phosphorylation internalized normally, led to speculation that GRK2-dependent internalization was peculiar to G_i-coupled receptors (Tsuga et al., 1994). However, a role for GRK-mediated phosphorylation in the internalization of the G_s-coupled β₂AR was eventually demonstrated using an internalization-defective β₂AR-Y326A mutant (Ferguson et al., 1995). The β₂AR-Y326A mutant was not only internalization-defective, but also did not serve as a substrate for GRK-mediated phosphorylation. The overexpression of GRK2 not only promoted the internalization of this receptor mutant, but re-established GRK-mediated phosphorylation of the mutant receptor. The GRK expression-dependent rescue of β₂AR-Y326A internalization required intact sites for GRK2 phosphorylation, indicating that receptor phosphorylation rather than GRK2 association per se was required to allow β₂AR internalization in HEK 293 cells. Wild-type β₂AR phosphorylation and internalization were reduced by the overexpression of a catalytically inactive GRK2 mutant. The role of GRK2-mediated phosphorylation in facilitating GPCR internalization has now been confirmed for multiple other receptors, e.g., the AT_1AR (Smith et al., 1998), endothelin A receptor (Bremmes et al., 2000), D2 dopamine receptor (Itokawa et al., 1996), follitropin receptor (Lazari et al., 1999), and monocyte chemoattractant protein-1 receptors (Franci et al., 1996) CCR-5 (Aramori et al., 1997) and CXCR1 (Barlic et al., 1999).

The contribution of other GRK family members to GPCR internalization has not been the subject of intense investigation and their role in promoting GPCR internalization remains less clear. For example, the agonist-promoted internalization of the β₂AR-Y326A mutant was facilitated by GRK2, GRK3, GRK5, and GRK6, but not by GRK4, which increased receptor internalization in the absence of agonist (Menard et al., 1996). The GRK-mediated phosphorylation of the follitropin receptor was blocked by the expression of dominant-negative mutants of both GRK2 and GRK6 (Lazari et al., 1999). However, the internalization of the follitropin receptor was blocked only by the dominant-negative GRK2 mutant (Lazari et al., 1999). The role of different GRK isoforms in facilitating the internalization of mAChR subtypes seemed to be even more complicated (Tsuga et al., 1998a). The internalization of human m2-m5, but not m1 mAChR, was increased by GRK2 overexpression (Tsuga et al., 1998a). However, the ability of GRK4, GRK5, and GRK6 to promote the internalization of mAChR subtypes differed for each receptor subtype tested and the cell line used (Tsuga et al., 1998a).

Depending on the GPCR studied, GRK-mediated phosphorylation is not absolutely required for internalization (Bouvier et al., 1988;Hausdorff et al., 1989; Ferguson et al., 1995). Rather, it seems that phosphorylation stabilizes a conformation state required to promote the interaction of GPCRs with some other cellular element that directly promotes the internalization of the receptor. Consequently, β₂AR mutants lacking sites for GRK-mediated phosphorylation must be able to interact with an endocytic adaptor protein, even in the absence of phosphorylation. In fact, GRK-phosphorylation increases the affinity of the β₂AR to bind β-arrestins, which in addition to uncoupling receptors from heterotrimeric G proteins, act as endocytic adaptor proteins targeting GPCRs for internalization via clathrin-coated vesicles (Ferguson et al., 1996b; Zhang et al., 1996) (Fig. 3). When overexpressed, both β-arrestin1 and β-arrestin2 ameliorate the internalization defect of the β₂AR-Y326A even in the absence of GRK-mediated phosphorylation (Ferguson et al., 1996b). Furthermore, β-arrestins facilitate the endocytosis of β₂ARs lacking either carboxyl-terminal tails or putative sites for GRK phosphorylation. Thus, depending on the level of β-arrestin protein expression and β-arrestin binding affinity, GRK-mediated phosphorylation may be dispensable for some GPCR subtypes (Menard et al., 1997). However, for most receptors, including the β₂AR, a synergistic relationship exists between GRK-mediated phosphorylation and β-arrestin binding (Ferguson et al., 1996b).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Molecular mechanisms involved in the GRK- and β-arrestin-dependent desensitization and internalization of GPCRs. GPCR activation leads to GRK-dependent phosphorylation of intracellular serine and threonine residues that facilitate the translocation and binding of β-arrestin proteins to the receptor. β-arrestins, via their association with the β2-adaptin subunit of the AP-2 heterotetrameric adaptor complex, target GPCRs to clathrin-coated pits (Ferguson et al., 1996b; Zhang et al., 1996; Barak et al., 1997b; Laporte et al., 1999). In addition to their association with β2-adaptins, β-arrestins also bind clathrin (Goodman et al., 1996). The GPCR is subsequently internalized via clathrin-coated vesicles. AP-2, AP-2 heterotetrameric adaptor complex; βArr, β-arrestin; H, hormone; P, phosphate group.

As mentioned previously, the relationship between GRK-mediated phosphorylation and β-arrestin binding is likely different for each GPCR subtype. This may explain the myriad of related, but dissimilar, observations that have been reported in the literature regarding the relative importance of GRKs and β-arrestins in the internalization of different GPCR subtypes. For example, the internalization of two chemokine receptors, CCR-5 and CXCR1, in HEK 293 cells requires the overexpression of both GRK and β-arrestin proteins (Aramori et al., 1997; Barlic et al., 1999). In contrast, whereas the internalization of the m2 mAChR seems dependent on GRK-mediated phosphorylation, the internalization of this receptor subtype does not seem to require β-arrestin, depending on the cellular system in which the receptor is expressed (Tsuga et al., 1994; Schlador and Nathanson, 1997;Vogler et al., 1999; Werbonat et al., 2000). There are also examples of receptors that do not serve as substrates for GRK and β-arrestin proteins and do not internalize in response to agonist activation (Jockers et al., 1996). Thus, differences in GPCR structure-activity relationships likely play an equally important role in regulating differences in the patterns of GPCR endocytosis as they do for regulating differences in agonist and G protein coupling specificity.

In addition to apparent differences in GPCR structure-activity relationships with regard to GRK and β-arrestin binding, GPCR endocytosis is also regulated by both the context of agonist activation and the cellular milieu in which a receptor is expressed. There are several excellent examples that illustrate this point. 1) Etorphine, but not morphine, stimulates μ-opioid receptor phosphorylation and internalization in HEK 293 cells (Zhang et al., 1998). However, GRK2, but not β-arrestin overexpression, allows both phosphorylation and internalization of the μ-opioid receptor in response to morphine. This observation indicates that different receptor agonists stabilize distinct receptor conformations that are able to discriminate between G protein coupling and GRK phosphorylation. This may explain why some peptide receptor antagonists are observed to stimulate receptor internalization (Roettger et al., 1997; Bhowmick et al., 1998). 2) The extent of agonist-promoted β₂AR internalization was different, depending on the cell line in which it was tested (Menard et al., 1997). For example, the maximal extent of β₂AR internalization in different cell lines correlates nicely with the levels of GRK and β-arrestin protein expression (Menard et al., 1997). It is possible that GPCR structure may have evolved to match the levels of GRK and β-arrestin protein expression in the cells in which they are normally expressed in vivo. For example, CXCR1 is effectively internalized in neutrophil-like RBL-2H3 cells, but does not internalize in HEK 293 cells (Barlic et al., 1999). The difference between these cell lines is that RBL-2H3 cells express substantially high levels of GRK2 and β-arrestin2 protein than HEK 293 cells (Barlic et al., 1999). 3) The endocytosis of GPCR subtypes differs depending on the complement of β-arrestin protein isoforms expressed by a particular cell. It was recently reported that there are striking differences in the ability of different arrestin isoforms to bind different GPCRs at the plasma membrane (Oakley et al., 2000). Therefore, the ability of a specific GPCR to internalize in a particular cell type may be dictated by the complement of β-arrestin isoforms expressed in the cell. Consequently, the endogenous complement of GRK and β-arrestin proteins expressed in immortalized cell culture systems, such as HEK 293 cells, may not accurately reflect protein expression levels that will be observed in the diverse physiological environments in which a particular GPCR subtype may be expressed such as in hematopoetic, cardiac or neuronal cells. Therefore, potentially important differences in the physiological regulation of different GPCR subtypes that occur in vivo will require the examination of endocytosis patterns under conditions of varied GRK and β-arrestin protein expression levels.

2. Clathrin and β-Adaptin Interactions.

The first indication that β-arrestins specifically target GPCRs for endocytosis via clathrin-coated vesicles came from experiments testing the effects of β-arrestin and dynamin dominant-negative mutants on the internalization of the β₂AR and AT_1AR (Zhang et al., 1996). Dynamin is a large GTPase that is involved in the pinching off of clathrin-coated vesicles from the plasma membrane (Damke et al., 1994). The expression of a dynamin mutant (K44A) lacking GTPase activity effectively blocked both β₂AR internalization and β-arrestin-stimulated AT_1AR internalization (Zhang et al., 1996). Furthermore, Goodman et al. (1996) demonstrated that both β₂ARs and β-arrestins were colocalized with clathrin in clathrin-coated pits. The idea that β-arrestins specifically target GPCRs for endocytosis via clathrin coated vesicles has been corroborated by recent studies showing that β-arrestins interact directly with components of the endocytic machinery involved in the formation of clathrin-coated pits (Goodman et al., 1996; Laporte et al., 1999, 2000). β-Arrestins bind to both the clathrin heavy chain and the β2-adaptin subunit of the heterotetromeric AP-2 adaptor complex (Goodman et al., 1997; Laporte et al., 1999, 2000).

β-Arrestins bind with high affinity and stoichiometry to purified clathrin in vitro (Goodman et al., 1996). β-Arrestin2 binds clathrin with approximately 6-fold higher affinity than β-arrestin1 (Goodman et al., 1996). Visual arrestin, while structurally related to the β-arrestins, does not promote β₂AR internalization and does not bind to clathrin (Goodman et al., 1996). The clathrin β-arrestin binding domain is localized to residues 89–100 of the amino-terminal globular region in the terminal domain of the clathrin heavy chain that lies at the distal end of each clathrin triskelion (Goodman et al., 1997). The β-arrestin domain involved in clathrin binding is localized to amino acid residues 373–377 in the carboxyl terminus of β-arrestin2 (Krupnick et al., 1997) (Fig. 2). Mutation of the residues within this region of β-arrestin2 substantially reduced clathrin cage binding without altering binding to phosphorylated rhodopsin (Krupnick et al., 1997). However, the mutation of the amino residues within this domain to alanine residues did not affect the ability of the mutant β-arrestin2 to facilitate β₂AR internalization in COS7 cells (Laporte et al., 1999). The ability of β-arrestins to bind to clathrin led to the proposal that β-arrestins, rather than the AP-2 adaptor complex, had evolved as GPCR-specific clathrin adaptors allowing GPCR endocytosis via clathrin-coated vesicles (Goodman et al., 1996, 1997; Krupnick et al., 1997). However, unlike observed for the AP-2 adaptor complex, β-arrestins do not promote clathrin coat assembly (Goodman et al., 1997). More recently, Laporte et al. (1999, 2000) demonstrated that, in addition to binding clathrin, β-arrestins also bind to the β2-adaptin subunit of the heterotetrameric AP-2 adaptor complex and that this interaction is essential for β₂AR internalization.

The heterotetrameric AP-2 adaptor complex consists of four subunits: two large 100 kDa subunits (α-adaptin and β2-adaptin), one medium size 50 kDa subunit (μ2), and one small 17 kDa subunit (ς2) (reviewed by Kirchhausen, 1999). When AP-2 complexes are examined by electron microscopy they appear as two smaller lobes (ears) attached to a brick-like structure. The two ears of the AP-2 adaptor complex are formed by the α- and β2-subunits. Each of the AP-2 adaptor complex subunits is ascribed a distinct function. The α-adaptin subunit binds to clathrin, dynamin, and Eps15 (Goodman and Keen, 1995; Wang et al., 1995; Benmerah et al., 1996). The β2-adaptin subunit is essential for clathrin coat formation and interacts with both clathrin and dileucine motifs (Kirchhausen, 1999). The μ2-subunit recognizes tyrosine-based internalization signals such as the NPXY (where N, P, and Y are asparagine, proline and tyrosine residues and X can be any amino acid) and Yppφ (where Y is a tyrosine residue, φ is a residue with a bulky hydrophobic side chain, and p is a polar residue) motifs. Yppφ was first identified in the carboxyl terminus of the transferrin receptor and is used more broadly as an endocytic motif than the NPXY motif (Kirchhausen, 1999).

The β-arrestin domain involved in binding to the β2-adaptin subunit of the AP-2 adaptor complex is also localized to the carboxyl termini of β-arrestin1 and β-arrestin2 (Laporte et al., 1999) (Fig. 2, A and B). In particular, two arginine residues (R394 and R396) in β-arrestin2 mediate binding to β2-adaptin in vitro (Laporte et al., 2000) (Fig. 2B). The binding of β-arrestins to β2-adaptin is independent of clathrin binding (Laporte et al., 2000). In cells, the association of β2-adaptin with β-arrestins is stimulated by receptor activation and β2-adaptin-GFP is recruited to the plasma membrane in response to β₂AR activation (Laporte et al., 1999). Accordingly, it is possible that GPCR activation may lead to the nucleation of clathrin-coated pits (Laporte et al., 1999). Moreover, whereas β₂AR/β-arrestin complexes lacking the β-arrestin clathrin binding motif redistributed to coated pits, receptor/β-arrestin complexes lacking the β2-adaptin binding site did not (Laporte et al., 2000). Thus, β-arrestin interactions with the AP-2 complex, rather than with clathrin, are necessary for the initial targeting of receptors to coated pits (Laporte et al., 2000) (Fig. 3). Consequently, it seems that GPCR endocytosis is mediated by the same molecular intermediates that have been previously described to involved in the endocytosis of receptor tyrosine kinases (reviewed byKirchhausen, 1999), except that GPCRs seems to use a common proximal intermediate, β-arrestin. The role of β-arrestin in GPCR endocytosis may be analogous to the function of Eps15 in coated pit formation and the endocytosis of both transferrin and the EGF receptor (Benmerah et al., 1998, 1999). Although there is no doubt that β-arrestins bind clathrin directly, the consequence of the association between β-arrestin and clathrin with regards to regulating GPCR internalization requires additional investigation.

3. β-Arrestin Regulation.

In response to GPCR activation, cytosolic β-arrestin proteins translocate to the plasma membrane and then subsequently redistribute to clathrin-coated pits bound to receptors (Barak et al., 1997b). The extent of β-arrestin translocation recapitulates the dose-response curves for agonist-stimulated receptor-G protein coupling and internalization (Zhang et al., 1999). However, the mechanism underlying this receptor-mediated response remains unclear. It is not known whether β-arrestin translocation is a passive process involving diffusion-dependent interactions with agonist-activated phosphorylated receptors or whether β-arrestin translocation is signal-driven. However, feedback regulation of β-arrestin activity may occur, particularly in the case of β-arrestin1. β₂AR activation leads to the feedback phosphorylation of β-arrestin1 on serine residue 412 by extracellular signal-regulated kinases (ERKs) (Lin et al., 1998, 1999). The endocytic function of β-arrestin1 is reported to depend on the phosphorylation/dephosphorylation state of β-arrestin1 (Lin et al., 1998). β-Arrestin1 recruited to the plasma membrane-bound receptor is apparently rapidly dephosphorylated, whereas cytoplasmic β-arrestin1 is primarily phosphorylated (Lin et al., 1998). Consequently, the mutation of serine residue 412 to an aspartic acid residue in β-arrestin1 creates a β-arrestin mutant that functions as a dominant-negative with respect to β₂AR endocytosis (Lin et al., 1998). However, this mutation has no effect on β₂AR desensitization (Lin et al., 1998). Moreover, β-arrestin1 dephosphorylation does not seem to be required for β-arrestin1 translocation to the plasma membrane (Oakley et al., 2000). Nonetheless, β-arrestin1 dephosphorylation seems to increase clathrin association (Lin et al., 1999). It remains unknown whether ERK-mediated phosphorylation contributes to the regulation of β-arrestin1/β-adaptin interactions. Interestingly, serine residue 412 is not conserved in β-arrestin2. Therefore, ERK-mediated phosphorylation either occurs at a different site, and/or β-arrestin2 activity is regulated by an alternative mechanism. The answer to this question will be important, because β-arrestin2 is considered to play a predominant role in the endocytosis of most GPCRs (Oakley et al., 2000; Santini et al., 2000).

β-Arrestin activity seems to be regulated by phosphoinositides, in particular IP₆ (Gaidarov and Keen, 1999; Gaidarov et al., 1999). The β-arrestin phosphoinositide-binding site is localized to amino acid residues 233–251 of β-arrestin2. The mutation of basic residues within this domain significantly reduces the phosphoinositide binding to β-arrestin2 and produces a β-arrestin mutant that does not support β₂AR internalization in COS1 cells. Furthermore, the β-arrestin mutant defective in phosphoinositide binding did not concentrate at clathrin-coated pits, but was still recruited to the plasma membrane in response to β₂AR activation. Taken together, these observations suggest that phosphoinositide binding may contribute to the recruitment of receptor/β-arrestin complexes to clathrin-coated pits. However, the mechanism by which this is achieved remains to be determined.

4. Alternative G Protein-Coupled Receptor Endocytic Pathways.

The precise mechanism(s) by which all GPCRs internalize remains a controversial topic. It is now thought that not all GPCRs necessarily internalize in a β-arrestin- and clathrin-dependent manner. This idea has arisen from two initial experimental observations. First, when expressed in different cell types, the internalization profiles for some, but not all GPCRs, is different (e.g., Zhang et al., 1996). In COS7 cells, which express relatively little GRK and β-arrestin protein, the maximal extent of AT_1AR internalization is virtually indistinguishable from that observed in HEK 293 cells (Zhang et al., 1996; Menard et al., 1997). In contrast, the maximal extent of β₂AR internalization in response to agonist activation is markedly lower in COS7 cells (Zhang et al., 1996; Menard et al., 1997). Second, the internalization of some GPCRs is less sensitive to the effects of dominant-negative β-arrestin and dynamin mutants (Zhang et al., 1996; e.g., Vogler et al., 1999). In the case of the AT_1AR and m2 mAChR, the coexpression of either dominant-negative β-arrestin or dynamin mutants does not block agonist-stimulated internalization (Zhang et al., 1996; Vogler et al., 1999), whereas β-arrestin stimulated AT_1AR internalization was blocked completely by dominant-negative dynamin (Zhang et al., 1996). Furthermore, GIT-1, a GTPase-activating protein for the ADP ribosylation factor family of small GTP-binding proteins, only affects the function of G protein-coupled receptors that are internalized through the clathrin-coated pit pathway in a β-arrestin- and dynamin-sensitive manner (Claing et al., 2000). Together, these observations have led to the suggestion that an alternative endocytic mechanism may account for the internalization of some GPCRs, at least in the presence of dominant-negative inhibitors of clathrin-mediated endocytosis. However, the molecular identity of these alternative pathway(s) continues to elude characterization.

Although the internalization of the β₂AR is reported to be caveolin-dependent in A431 cells, the putative β₂AR caveolin-binding motif is found within the extracellular portion of the seventh transmembrane spanning domain of the β₂AR (Raposo et al., 1989; Watson and Arkinstall, 1994; Couet et al., 1997). The putative caveolin binding motifs found within other members of both the β-adrenergic receptor and mAChR families are also found within the extracellular portion of the seventh transmembrane spanning receptor domain (Watson and Arkinstall, 1994). The caveolin binding motif in the endothelin A receptor, which internalizes in a β-arrestin-dependent manner, is localized to the first extracellular loop (Watson and Arkinstall, 1994;Bremmes et al., 2000). Therefore, these putative caveolin-binding motifs are not appropriately localized to receptor domains that would facilitate interactions with caveolin.

Although both the AT_1AR and m2 mAChR seem to internalize normally in the presence of dominant-negative β-arrestin and dynamin mutants, the internalization of both these GPCR subtypes seem to be dependent on receptor phosphorylation (Smith et al., 1998;Tsuga et al., 1998b). Furthermore, the internalization of both GPCRs is facilitated by the coexpression of GRK and β-arrestin proteins (Zhang et al., 1996; Schlador and Nathanson, 1997). Thus, these receptors retain the capacity to internalize in a β-arrestin- and clathrin-dependent manner, suggesting the possibility that, in the absence of dominant-negative inhibitors, the receptors normally internalize via clathrin-coated vesicles. This is a particularly important question since the overexpression of dynamin mutants is reported to induce the up-regulation of alternative endocytic pathways, such as pinocytosis (Damke et al., 1995).

The idea that the AT_1AR and m2 mAChR normally internalize via clathrin-coated vesicles is supported by two observations. First, the mutation of all three dynamin GTPase domains creates a dynamin dominant-negative mutant that blocks the internalization of both the AT_1AR and m2 mAChR (Werbonat et al., 2000). Second, the dynamin I-K44A mutant traps β-arrestin bound AT_1ARs in coated pits and blocks the trafficking of β-arrestin with the receptor to endosomes (Anborgh et al., 2000). A potential explanation for these apparently incongruous observations is provided by the observation that functionally and biochemically distinct subpopulations of clathrin-coated pits might exist (Cao et al., 1998). As a consequence, GPCRs may be targeted to compositionally distinct membrane vesicles that exhibit differences in their sensitivity to the limited battery of dominant-negative mutants currently used by researchers in the field. Regardless, these observations do not rule out the existence of an alternative GPCR endocytic pathway. However, they do suggest that, under physiological conditions, the predominant pathway for GPCR endocytosis is β-arrestin- and dynamin-dependent.

5. Receptor Determinants for Endocytosis.

The identification of receptor-specific determinants for GPCR internalization has been the subject of intense investigation but has not yielded the identity of a conserved GPCR-specific endocytosis consensus motif. Rather it seems that multiple receptor domains contribute to GPCR-specific differences in the rate and extent of receptor internalization. In fact, many of the receptor domains that are necessary for G protein coupling are also important for GPCR internalization. Although the mediators of GPCR endocytosis (e.g., β-arrestin) and G proteins likely interact at similar sites at the intracellular face of the receptor, these processes are functionally distinguishable (Mahan et al., 1985; e.g.,Moro et al., 1993; Barak et al., 1994; Hunyady et al., 1995). Perhaps the best evidence that G protein coupling is not required for receptor endocytosis is provided by the observation that in S49 murine lymphoma cell lines, which either lack G_sα or have point mutations preventing receptor/G protein interactions, β₂AR internalization in response to agonist is normal (Mahan et al., 1985).

Several intracellular receptor domains and conserved motifs have been identified as potential determinants for receptor endocytosis. Although each of these domains and motifs are involved in the internalization of some receptors, they are not required for the internalization of all GPCRs. For example, the NP(X)_2,3Y motif found at the intracellular face of the seventh transmembrane domain of many GPCRs was originally considered to serve as a GPCR-specific endocytosis motif due the similarity between this motif and the tyrosine-based NPXY internalization signal (Barak et al., 1994; Kirchhausen, 1999). However, whereas the mutation of the tyrosine residue in the NP(X)_2,3Y motif reduced the internalization of the β₂AR and neurokinin 1 receptor, the same mutation did not reduce the internalization of either the AT_1AR or the gastrin-releasing peptide receptor (Barak et al., 1994; Hunyady et al., 1995; Slice et al., 1994; Bohm et al., 1997b). Further characterization of the NP(X)_2,3Y motif revealed that it serves as a critical determinant of receptor conformations required for normal receptor interactions with agonist, G proteins, and GRKs (Barak et al., 1995; Ferguson et al., 1995).

Depending on the GPCR tested, regions of the second and third intracellular loop domains contribute to GPCR internalization. In the case of the m2 mAChR, the determinants for internalization are localized to a serine/threonine rich region within the large third intracellular loop of the receptor (Moro et al., 1993). These amino acid residues presumably represent the GRK phosphorylation sites that are involved in promoting the endocytosis of this GPCR subtype (Tsuga et al., 1998b). In addition to the third intracellular loop, the conserved DRYXXV/IXXPL sequence found within the second intracellular loop domain also serves as an internalization determinant for some GPCRs. In particular, mutation of the leucine residue within this motif reduces the internalization of both the m1 mAChR and gonadotropin-releasing hormone (GnRH) receptor (Moro et al., 1994,Arora et al., 1995). Mutation of the DRS sequence in the GnRH receptor to DRY substantially increases the relative internalization of the GnRH receptor (Arora et al., 1995). However, the second intracellular loop also contributes to the regulation of ligand binding affinity, G protein coupling, and receptor stability (e.g., Arora et al., 1995;Alewijnse et al., 2000). Therefore, similar to what is observed for the NP(X)_2,3Y motif, the effects of mutating this motif on receptor sequestration are not easily distinguishable from effects on other receptor properties. Furthermore, although the DRY motif within the N-formyl peptide receptor contributes to β-arrestin binding, mutants defective in β-arrestin binding are still able to internalize in response to agonist (Bennett et al., 2000).

Numerous studies have examined the role of GPCR carboxyl-terminal tails and putative sites for GRK phosphorylation in regulating agonist-stimulated GPCR internalization. These studies have provided a broad spectrum of different results, highlighting the complex diversity in receptor structure-function relationships as they pertain to receptor endocytosis. For example, despite the fact that the internalization of the β₂AR is β-arrestin-dependent, neither the truncation of the β₂AR carboxyl-terminal tail nor the mutation of all potential GRK phosphorylation sites prevents β₂AR internalization (Hausdorff et al., 1989;Ferguson et al., 1995, 1996b). In contrast, whereas the internalization of the AT_1AR is not absolutely dependent on β-arrestin, truncation of the carboxyl-terminal tail or the mutation of potential sites for GRK-dependent phosphorylation prevents AT_1AR internalization (Thomas et al., 1995; Zhang et al., 1996; Smith et al., 1998). In other cases (e.g., CCR5 and GnRH receptor), GPCR internalization is β-arrestin-dependent and requires an intact carboxyl-terminal tail domain (e.g., Aramori et al., 1997;Blomenrohr et al., 1999). Furthermore, both positive and negative regulators of agonist-stimulated internalization are found within the carboxyl-terminal tail and third intracellular loop domains of the parathyroid hormone receptor and m2 mAChR, respectively (Huang et al., 1995; Lee et al., 2000). A dileucine motif within the transmembrane proximal domain of the β₂AR carboxyl-terminal tail is also important for β₂AR internalization (Gabilondo et al., 1997). In contrast, a carboxyl-terminal tail dileucine motif negatively regulates lutropin/choriogonadotropin receptor internalization because the mutation of the dileucine motif increased the agonist-stimulated internalization of the receptor (Nakamura and Ascoli, 1999). Therefore, the mutation of similar motifs in different GPCRs can result in diametrically opposed effects.

Taken together, the structure-activity studies and mutational analyses of intracellular GPCR domains and motifs indicate that GPCR internalization is regulated by the interaction of multiple intracellular domains and residues with either β-arrestins and/or other endocytic adaptor proteins. This point is best illustrated by the study of Jockers et al. (1996), where all of the intracellular domains of the β₃AR had to be substituted with the equivalent domains from the β₂AR to establish a β₂AR-like internalization phenotype. Consequently, it is becoming increasingly apparent that the agonist-stimulated GPCR conformation allowing the association of endocytic adaptor proteins can be achieved by multiple mechanisms that are GPCR subtype-specific. Furthermore, the relative requirement for GRK-mediated phosphorylation to allow GPCR internalization varies with each individual receptor. Thus, from a structure-function standpoint, GPCRs should be considered as bipartite integral membrane proteins that consist of an extracellular ligand-binding domain and an intracellular protein-binding domain. Similar to what is observed for ligand binding, the substrate affinity and specificity of the protein-binding domains will be GPCR subtype-specific and will involve multiple intracellular receptor domains.

1. Endocytosis and G Protein-Coupled Receptor Desensitization.

GPCR internalization was originally considered to be a primary mechanism of receptor desensitization due to the spatial uncoupling of the receptor from its effector system (Sibley and Lefkowitz, 1985). However, this hypothesis subsequently fell out of favor for three reasons. 1) Since receptor desensitization proceeds more rapidly than receptor endocytosis, it is expected that the majority of internalized GPCRs will be predesensitized as the consequence of protein kinase phosphorylation. 2) Pharmacological treatments that blocked GPCR internalization, such as hypertonic sucrose and concanavalin A, did not alter the β₂AR desensitization profile (Yu et al., 1993;Pippig et al., 1995). 3) It was observed that internalization played a role in the resensitization of GPCR responsiveness (Yu et al., 1993;Pippig et al., 1995; Zhang et al., 1997). Recently, with the advent of new techniques to examine GPCR intracellular trafficking patterns, protein-protein interactions and recycling, the field has come full circle and has now begun to recognize the important role of GPCR internalization in regulating differences in the patterns of GPCR desensitization.

In response to agonist-stimulation, many GPCRs are internalized but are not recycled back to the cell surface (Hermans et al., 1997; Trejo and Coughlin, 1999; Groarke et al., 1999; Oakley et al., 1999; Zhang et al., 1999; Anborgh et al., 2000; Bremnes et al., 2000). In some cases, internalized GPCRs, such as the protease-activated receptors (PAR) and endothelin B receptor, are predominantly targeted to lysosomes for degradation (Trejo and Coughlin, 1999; Bremnes et al., 2000). In the case of protease-activated receptors, the carboxyl-terminal tail regulates targeting to lysosomes (Trejo and Coughlin, 1999). For other GPCRs, such as the AT_1AR, internalized receptors are not necessarily sorted to lysosomes, but may be retained within the endosomal compartment (Zhang et al., 1999; Anborgh et al., 2000). As a consequence, by virtue of the fact that some receptors do not recycle, they will mediate transient responses to agonist. For example, endothelin A receptors are efficiently recycled and mediate persistent responses to endothelin, whereas the endothelin B receptors are targeted to lysosomes and mediate only transient responses to endothelin (Bremnes et al., 2000).

Several recent studies have used chimeric receptor constructs to examine differences in the intracellular sorting patterns of GPCRs (Trejo and Coughlin, 1999; Oakley et al., 1999; Anborgh et al., 2000). These studies have revealed that specific determinants within the cytoplasmic tails of GPCRs determine whether GPCRs are either recycled back to the plasma membrane or are retained within the intracellular compartment of the cell and/or targeted to lysosomes (Innamorati et al., 1998; Trejo and Coughlin, 1999; Oakley et al., 1999; Anborgh et al., 2000). In the case of the V2 vasopressin receptor, a carboxyl-terminal serine cluster was demonstrated to prevent receptor recycling (Innamorati et al., 1998). Further characterization of V2 vasopressin receptor trafficking revealed that, like many peptide receptors (e.g., AT_1AR, neurokinin 1 receptor, thyrothropin-releasing hormone receptor, PAR2, and neurotensin receptor), the V2 vasopressin receptor internalized as a complex with β-arrestin in endocytic vesicles (Dery et al., 1999; Groarke et al., 1999; McConalogue et al., 1999; Oakley et al., 1999; Zhang et al., 1999; Anborgh et al., 2000). The capacity of both the V2 vasopressin receptor and AT_1AR to internalize with β-arrestin bound was conferred by their carboxyl-terminal tail domains (Oakley et al., 1999; Zhang et al., 1999; Anborgh et al., 2000). The same cluster of serine residues that prevented V2 vasopressin receptor recycling was required for the internalization of β-arrestin with the receptor (Innamorati et al., 1998; Oakley et al., 1999). GPCRs that do not internalize with β-arrestin bound, such as the β₂AR, lack this cluster of serine residues in the appropriate context (Oakley et al., 1999). This has led to the hypothesis that β-arrestin interactions with the carboxyl-terminal tails of GPCRs dictate the rate of receptor recycling and resensitization (Oakley et al., 1999).

Although β-arrestin internalization is well correlated with the inhibition of GPCR recycling, exceptions exist. For example, the neurokinin 1 receptor, which internalizes complexed with β-arrestin, is efficiently recycled back to the cell surface and is effectively resensitized (Grady et al., 1995; McConalogue et al., 1999). Therefore, multiple patterns of GPCR desensitization and resensitization exist. These include: 1) GPCRs that do not internalize in response to agonist [e.g., β₃AR (Jockers et al., 1996)]. 2) GPCRs that are internalized without β-arrestin bound, and that are dephosphorylated and recycled back to the cell surface [e.g., β₂AR (Oakley et al., 1999;Anborgh et al., 2000)]. 3) GPCRs that are internalized with β-arrestin bound, but that are dephosphorylated and rapidly recycle back to the cell surface after the dissociation of β-arrestins in the intracellular compartment of the cell [e.g., neurokinin 1 receptor (Grady et al., 1996; McConalogue et al., 1999)]. 4) GPCRs that are internalized with β-arrestin bound and are either retained in endosomes and/or targeted to lysosomes [e.g., AT_1AR and PAR (Trejo and Coughlin, 1999; Dery et al., 1999; Anborgh et al., 2000)] (Fig.4A). Although β-arrestin seems to play a central role in regulating the intracellular trafficking properties of GPCR, it is likely that the manifestation of distinct intracellular trafficking properties also involves other components of the endocytic machinery.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

β-Arrestin-dependent regulation of GPCR trafficking and the mechanisms involved in GPCR resensitization. Panel A, mechanism I: GPCR resensitization is achieved by the clathrin-dependent internalization of GPCRs to endosomes, and the receptors are subsequently dephosphorylated and recycled back to the cell surface (1). Alternatively, GPCRs that internalize with β-arrestin either dissociate from β-arrestin in the intracellular compartment, and are both dephosphorylated and recycled (2) or are retained within large core vesicles and/or are targeted for degradation in lysosomes (3) (Krueger et al., 1997; Trejo and Coughlin, 1998; Oakley et al., 1999; Zhang et al., 1999; Anborgh et al., 2000). Panel B, mechanism II: resensitization of GPCRs that are internalized with β-arrestin bound and do not recycle is mediated by the mobilization of a reserve pool of intracellular receptors and/or de novo receptor synthesis in the Golgi apparatus (Shapiro et al., 1996;Shapiro and Coughlin, 1998; Anborgh et al., 2000). A, agonist; AP-2, AP-2 heterotetrameric adaptor complex; βArr, β-arrestin; P, phosphate group.

2. Endocytosis and G Protein-Coupled Receptor Resensitization.

The physiological importance of receptor resensitization in the maintenance of normal tissue homeostasis is obvious since prolonged or irreversible receptor desensitization would leave a cell unable to respond appropriately to extracellular stimuli. Just as GPCR desensitization provides a mechanism protecting cells against receptor overstimulation, GPCR resensitization protects cells against prolonged receptor desensitization. The mechanism by which the resensitization of many GPCRs is achieved is thought to be the agonist-stimulated internalization of receptors to an intracellular membrane compartment (endosomes) enriched in a GPCR-specific phosphatase activity (Fig. 4A). GPCRs dephosphorylated in endosomes are subsequently recycled back to the cell surface in the preligand-exposed state (Fig. 4A). The rate at which GPCRs traffic from the plasma membrane to the endosomal compartment and back again varies, depending on the GPCR subtype tested, but at a minimum takes several minutes. In contrast, GPCR desensitization occurs within seconds to minutes. Thus, it is clear that the processes involved in GPCR resensitization are less efficient than GPCR desensitization. This accounts for the observation that many GPCR agonists have limited long-term therapeutic usefulness and that GPCR resensitization does not adequately compensate for increased GRK-mediated receptor desensitization observed in both congestive heart failure and hypertension (Ungerer et al., 1993; Gros et al., 1997).

The concept that internalization is important for receptor resensitization arose from several early observations. 1) β₂ARs isolated from “light vesicular” membrane fractions (endosomes) exhibited lower stoichiometry of phosphorylation than β₂ARs isolated from the plasma membrane (Sibley et al., 1986). 2) The “light vesicular” membrane fractions were enriched in a GPCR-specific phosphatase activity (Sibley et al., 1986; Pitcher et al., 1995b). 3) Pharmacological treatments, such as concanavalin A and hypertonic sucrose, that blocked receptor internalization also blocked resensitization without affecting either receptor G protein-coupling or desensitization (Pippig et al., 1995; Garland et al., 1996;Hasbi et al., 2000). 4) Internalization-defective mutants were identified that signaled and desensitized, but were not resensitized (Barak et al., 1994). The requirement of internalization for GPCR resensitization has been demonstrated for the β₂AR (Pippig et al., 1995), m3 mAChR (Edwardson and Szekeres, 1999), neurokinin 1 receptor (Garland et al., 1996), delta opioid receptor (Hasbi et al., 2000), mu opioid receptor (Wolf et al., 1999), C5a anaphylatoxin receptor (Giannini and Boulay, 1995), cholecystokinin receptor (Lutz et al., 1993), and endothelin A receptor (Bremnes et al., 2000).

As a consequence of their role in GPCR endocytosis, β-arrestins not only play an important role in regulating GPCR desensitization, but also resensitization. For example, in COS7 cells that express relatively low levels of β-arrestin protein when compared with other cell lines (Menard et al., 1997), β₂AR resensitization does not occur in COS7 cells unless β-arrestins are overexpressed (Zhang et al., 1997). These observations indicate that, due to differences in GRK and β-arrestin protein expression levels, the relative capacity of an individual GPCR to be resensitized might differ depending on the cell type and tissue in which it is expressed. Furthermore, the relative capacity of a GPCR to become resensitized may be depend upon the relative protein expression levels of other regulators of intracellular trafficking.

The internalization of GPCRs to an intracellular membrane compartment is considered to be required for dephosphorylation by a GPCR-specific phosphatase (Pitcher et al., 1995b). GPCR resensitization not only requires the β-arrestin-dependent targeting of receptors to clathrin-coated pits and their subsequent endocytosis to endosomes, but also requires that the receptors release bound β-arrestin, become dephosphorylated, sorted, and recycled back to the cell surface. The majority of the studies examining the molecular mechanisms underlying the trafficking and dephosphorylation of GPCRs have focused on the β₂AR as a model GPCR. These studies have been facilitated by the use of GFP-tagged β₂AR, β-arrestin, and RabGTPases. In addition to β-arrestins, the internalization of the β₂AR, as well as the D2 dopamine receptor, is dependent on the small GTPase Rab5 (Iwata et al., 1999; Seachrist et al., 2000). Rab5 not only regulates the movement and fusion of β₂AR-bearing vesicles to/with early endosomes, but also regulates the formation of β₂AR-bearing vesicles at the cell surface (Seachrist et al., 2000) (Fig. 5). Furthermore, although β-arrestins are required for β₂AR internalization and colocalize with the receptor in clathrin-coated pits (Zhang et al., 1999), β-arrestin does not internalize with the β₂AR in endocytic vesicles (Zhang et al., 1999; Anborgh et al., 2000). The exclusion of β-arrestin from endocytic vesicles likely facilitates β₂AR dephosphorylation by allowing phosphorylated receptors to be exposed to the GPCR-specific phosphatase (Anborgh et al., 2000). Although rhodopsin is not internalized, visual arrestin dissociation is also required for rhodopsin dephosphorylation in rod outer segments (Palczewski et al., 1989). Furthermore, the association of the phosphatase with the β₂AR requires the acidification of the receptor in the endosomal compartment (Krueger et al., 1997). The dephosphorylation of the β₂AR occurs as the receptor transits between Rab5- and Rab4-positive endosomal compartments (Seachrist et al., 2000) (Fig. 5). Rab4 also regulates the recycling of the β₂AR back to the plasma membrane and is required for β₂AR resensitization (Seachrist et al., 2000) (Fig. 5). Although the endothelin A receptor has also been localized to Rab5-positive endosomes (Bremnes et al., 2000), little is known about the role of Rab5 and other RabGTPases in the intracellular trafficking of other GPCRs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Model depicting the regulation of β₂AR internalization, trafficking, dephosphorylation, and recycling. Agonist activation of the β₂AR results in β-arrestin (βArr)-mediated desensitization and targeting to clathrin-coated pits. Rab5 contributes to the formation of endocytic vesicles, the trafficking of vesicles to early endosomes, and the fusion of endocytic vesicles with early endosomes. The β₂AR is dephosphorylated by a G protein-coupled receptor specific phosphatase (GRP) as the receptor transits from the Rab5- to Rab4 positive compartments of early endosomes. The budding and/or recycling of β₂AR-bearing recycling vesicles is regulated by Rab4 (Seachrist et al., 2000). A, agonist; P, phosphate group. Figure reprinted with permission from Seachrist et al. (2000).

GPCR internalization and dephosphorylation in endosomes is not the only mechanism by which GPCR resensitization is achieved. Protease-activated receptors are unique members of the GPCR superfamily that are irreversibly activated by a proteolytic mechanism, then internalized and degraded in lysosomes (Coughlin, 1999). Consequently, resensitization of PAR responses requires an alternative mechanism of resensitization (Fig. 4B). The resensitization of PAR-mediated cellular responses is mediated in two ways: 1) the de novo synthesis of new receptors and 2) the maintenance of an intracellular reserve of naı̈ve receptors that can be mobilized to replace degraded receptors (Shapiro et al., 1996; Shapiro and Coughlin, 1998). This intracellular PAR pool is maintained by the agonist-independent movement of receptors between the cell surface and an intracellular membrane compartment (Shapiro et al., 1996; Shapiro and Coughlin, 1998). Although agonist-independent GPCR internalization is a phenomenon that is not well accepted, it is also observed for the AT_1AR and cholecystokinin receptor A (Hein et al., 1997; Tarasova, 1997; Anborgh et al., 2000). In the case of the AT_1AR, β-arrestin association with the receptor in the absence of agonist seems to be responsible for the agonist-independent loss of cell surface receptor (Anborgh et al., 2000). Therefore, β-arrestin-dependent GPCR internalization may regulate GPCR resensitization by two distinct mechanisms.

3. Endocytosis and G Protein-Coupled Receptor Signaling.

MAPKs, such as ERK1 and ERK2, are activated by a wide variety of GPCRs (reviewed by van Biesen et al., 1996; Luttrell et al., 1999a). GPCR-mediated activation of MAPKs via G_i-, G_q-, and G_o-mediated pathways has been studied extensively (van Biesen et al., 1996; Luttrell et al., 1999a). In the case of G_i-mediated signals, the Gβγ-subunits are involved in the activation of a Src family tyrosine kinase followed by the subsequent tyrosine phosphorylation of the same downstream adaptor proteins used by receptor tyrosine kinases (Scheme 1) (van Biesen et al., 1995) and seems to involve transactivation of tyrosine kinase receptors such as the EGF receptor (Prenzel et al., 1999; Pierce et al., 2000).GPCR→Gβγ→Tyr kinase→Shc→Grb2­mSos  →Ras→Raf→MEK→MAPK Scheme 1 The activation of MAPK by the G_s-coupled β₂AR involves the PKA phosphorylation-dependent switch of β₂AR G protein coupling to the activation of G_i rather than G_s (Daaka et al., 1997).

Recently, it was observed that one additional mechanism by which MAPKs were activated by GPCRs involved both endocytosis and β-arrestin proteins (Daaka et al., 1998; Luttrell et al., 1999b; Vogler et al., 1999; Barlic et al., 2000). Using β-arrestin and dynamin dominant-negative inhibitors of internalization, both β₂AR- and m1 mAChR-mediated activation of MAPK was demonstrated to require clathrin-coated vesicle-mediated endocytosis (Daaka et al., 1998; Vogler et al., 1999). The plasma membrane-delineated event involved in this process was the tyrosine phosphorylation of Shc and Raf kinase activation by Ras (Daaka et al., 1998). The internalization of this complex was required for subsequent MAPK activation (Daaka et al., 1998).

Consistent with the concept that GPCRs activate MAPKs by multiple converging mechanisms, many GPCRs were observed to mediate MAPK activation in manner that was independent of their internalization (Blaukat et al., 1999; Budd et al., 1999; DeGraff et al., 1999; Li et al., 1999; Schramm and Limbird, 1999; Whistler and von Zastrow, 1999). In addition, Whistler and von Zastrow (1999) reported that MAPK activation by an internalization-deficient mu opioid receptor was inhibited by a dynamin dominant-negative mutant. These data were interpreted to suggest that dynamin played a unique role in signal transduction that was independent of its function in clathrin-coated vesicles-mediated endocytosis (Whistler and von Zastrow, 1999). However, a recent study indicated that both α₂AR and β₂AR activation of MAPK required the transactivation and internalization of the epidermal growth factor receptor (EGFR) (Pierce et al., 2000). Whereas EGFR internalization was required for MAPK activation, neither EGFR internalization nor MAPK activation required GPCR internalization. Consequently, the requirement of clathrin-mediated endocytosis for MAPK activation was separated from the need for GPCR internalization. Taken together, these observations indicate that GPCR-stimulated activation of MAPK is mediated by multiple overlapping mechanisms, all of which may involve endocytosis.

In addition to contributing to the regulation of GPCR desensitization, internalization and resensitization, β-arrestins contribute to the assembly of signaling protein complexes (Luttrell et al., 1999b; Barlic et al., 2000; DeFea et al., 2000). β-Arrestins are essential for the agonist-stimulated formation of a protein complex containing the β₂AR, β-arrestin, and c-Src (Luttrell et al., 1999b). The formation of this complex is dependent on the ability of β-arrestin to directly associate with c-Src. The β-arrestin/c-Src interactions involve the association of the β-arrestin amino terminus with the Src homology 1 domain of c-Src (Miller et al., 2000). β-Arrestin mutants defective in their ability to interact with c-Src also effectively blocked β₂AR-mediated activation of MAPK (Luttrell et al., 1999b). As a consequence, β-arrestins can couple the termination of receptor-G protein coupling with the initiation of alternative signal transduction cascades in which the desensitized receptor and β-arrestin act as scaffolds.

An indication of the physiological role of β-arrestin signaling via Src family tyrosine kinases comes from a recent study examining the role of tyrosine kinase activation in the degranulation of neutrophils (Barlic et al., 2000). The internalization of the chemokine receptor CXCR1 is not only β-arrestin-dependent (Barlic et al., 1999), but CXCR1-mediated granule release and tyrosine kinase activation is inhibited by a β-arrestin dominant-negative mutant (P91G-P121E) that does not bind to c-Src kinase (Luttrell et al., 1999b; Barlic et al., 2000). Furthermore, the activation of the c-Src family tyrosine kinases, Hck and c-Fgr, in primary human neutrophils was dependent on the formation of a complex with endogenously expressed β-arrestins in neutrophils (Barlic et al., 2000) (Fig.6A). β-Arrestin mediated tyrosine kinase activation, receptor internalization, and the subsequent redistribution of β-arrestin/Hck complexes to granules were required for chemokine-induced granule exocytosis. These studies provide evidence that β-arrestin-mediated signaling is physiologically relevant to normal GPCR function in a physiological setting.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

β-Arrestin-dependent scaffolding of signaling complexes. A, role of β-arrestins in the regulation of c-Src family tyrosine kinase activity in response to chemokine receptor activation in neutrophils. CXCR1 activation leads to the β-arrestin-dependent formation of a complex involving the receptor, β-arrestin, and tyrosine kinase (e.g., Hck). The complex is redistributed to endocytic vesicles and may or may not lead to the activation of Ras, Raf, and MAPK pathway in neutrophils. Regardless, β-arrestin-mediated tyrosine kinase activation and redistribution to granules is required for the tyrosine kinase-dependent exocytosis of granules (Barlic et al., 2000). B, β-Arrestin acts as a scaffold for the formation of GPCR complexes with Raf-1 kinase and ERK (MAPK). The β-arrestin-dependent recruitment and internalization of these complexes in endocytic vesicles lead to ERK activation. However, the compartmentalization of the β-arrestin/Raf/ERK complex in vesicles prevents the translocation of activated ERK to the nucleus. As a consequence, ERK activity is limited to cytosolic targets (DeFea et al., 2000). A, agonist; βArr, β-arrestin; P, phosphate group.

In addition to interacting with Src family tyrosine kinases, β-arrestins were recently shown by DeFea et al. (2000) to interact directly with both Raf-1 kinase and MAPK in response to the activation of the G_q-coupled PAR2. Gel filtration experiments demonstrated that PAR2 activation resulted in the formation of a multiprotein complex containing the receptor, β-arrestin, Raf, and MAPK (Fig. 6B). As a consequence, the activation of wild-type PAR2 prevented the translocation of MAPK to the nucleus and by virtue of the fact that MAPK was retained within the cytosol prevented cell proliferation. In contrast, a PAR2 mutant defective in β-arrestin binding stimulated MAPK activation and cell proliferation. Under these conditions, the receptor failed to form a complex with both Raf and MAPK via β-arrestin, and MAPK was able to translocate to the nucleus in response to receptor activation. These observations suggest that, due to their role as endocytic adaptor proteins, β-arrestins not only regulate the formation of signal transduction complexes, but also regulate the intracellular compartmentalization of these complexes. Consequently, similar to what is observed for GPCR desensitization, differences in β-arrestin-mediated internalization and trafficking of GPCRs may determine distinct patterns of MAPK activation and the mitogenic potential of GPCR activation. However, the relative role for β-arrestins as endocytic versus scaffolding proteins in GPCR-mediated activation of MAPK signaling remains to be determined.