The structural basis of arrestin-mediated regulation of G-protein-coupled receptors

Abstract

The 4 mammalian arrestins serve as almost universal regulators of the largest known family of signaling proteins, G-protein-coupled receptors (GPCRs). Arrestins terminate receptor interactions with G proteins, redirect the signaling to a variety of alternative pathways, and orchestrate receptor internalization and subsequent intracellular trafficking. The elucidation of the structural basis and fine molecular mechanisms of the arrestin-receptor interaction paved the way to the targeted manipulation of this interaction from both sides to produce very stable or extremely transient complexes that helped to understand the regulation of many biologically important processes initiated by active GPCRs. The elucidation of the structural basis of arrestin interactions with numerous non-receptor-binding partners is long overdue. It will allow the construction of fully functional arrestins in which the ability to interact with individual partners is specifically disrupted or enhanced by targeted mutagenesis. These "custom-designed" arrestin mutants will be valuable tools in defining the role of various interactions in the intricate interplay of multiple signaling pathways in the living cell. The identification of arrestin-binding sites for various signaling molecules will also set the stage for designing molecular tools for therapeutic intervention that may prove useful in numerous disorders associated with congenital or acquired disregulation of GPCR signaling.

Fig. 1

The “classical” model of arrestin-mediated GPCR desensitization. The agonist-activated receptor activates cognate heterotrimeric G proteins that subsequently stimulate various signaling cascades increasing the activity of protein kinases PKA, PKC, etc. Active receptor is specifically phosphorylated by GRKs. Arrestin binds the active phosphoreceptor with high affinity, precluding further G protein activation. Arrestin serves as an adaptor linking the receptor to the internalization machinery of the coated pit (clathrin, adaptor complex AP-2), facilitating receptor internalization. Low pH in the endosome promotes agonist dissociation, which facilitates the release of arrestin, whereupon the receptor can be dephosphorylated and recycled back to the plasma membrane (resensitization). Alternatively, the receptor can be transported to lysosomes and destroyed (down-regulation).

Fig. 2

The sequential multisite binding mechanism ensures high arrestin selectivity for the phosphorylated active receptor. (A) Direct arrestin binding to the 4 functional forms of rhodopsin (inactive phosphorylated, P-Rh; light-activated phosphorylated, P-Rh*; inactive, Rh; light-activated, Rh*). P-Rh* is the preferred arrestin target. (B) Model of the arrestin–receptor interaction. First, arrestin binds via its activation sensor to receptor elements that change conformation upon activation or via the phosphate sensor to receptor-attached phosphates, respectively. If the receptor is active and phosphorylated, simultaneous engagement of both sensors promotes arrestin transition into the active state with concomitant engagement of additional binding sites, stabilizing the arrestin–receptor complex. Eventual loss of the active receptor conformation reverses this sequence of events and induces arrestin dissociation (adapted from Gurevich & Benovic, 1993).

Fig. 3

The function of the arrestin C-domain and C-tail. The deletion of the arrestin C-tail yields truncated arrestin(1–367) that binds P-Rh* essentially as well as full-length (WT) arrestin. This deletion dramatically decreases arrestin selectivity for P-Rh*, enhancing the binding to dark P-Rh and unpho-sphorylated Rh*, suggesting that the C-tail is a regulatory element “suppressing” the interactions with non-preferred forms of rhodopsin. The deletion of the whole C-terminal half of the molecule yields arrestin(1–191). This “mini-arrestin” demonstrates essentially the same binding to dark P-Rh and unphosphorylated Rh* as WT arrestin, but its binding to P-Rh* is many times lower. In the case of this mutant (in sharp contrast to WT arrestin), P-Rh* binding roughly equals the sum of the binding to dark P-Rh and Rh*. Thus, arrestin(1–191) does not have an additional binding site that can be mobilized to ensure arrestin selectivity (Gurevich & Benovic, 1992).

Fig. 4

Identification of the main phosphate-binding element of visual arrestin. (A) The N-domain element 161–179 contains a cluster of positively charged residues that are conserved in arrestin family (shown in bold). (B) The neutralization of several of these charges reduces the binding to P-Rh*, identifying these residues as phosphate-binding elements. In contrast, the R175N mutation enhances the binding to P-Rh* and dramatically increases the binding to unphosphorylated Rh* (C). (D) The same mutation in the context of the “mini-arrestin” (1–191) reduces P-Rh* binding without affecting arrestin interaction with Rh* (E). These data identify K166, R171, R175, and K176 as phosphate-binding elements. R175 also serves as the main “phosphate sensor”. Apparently, the neutralization of its charge by receptor-attached phosphates (or by the R175N mutation) is necessary to make the high-affinity arrestin binding possible (Gurevich & Benovic, 1995).

Fig. 5

Polar core is the phosphate sensor in arrestin. (A) Five solvent-excluded interacting charged residues are localized at the center of the arrestin molecule (hence the term “polar core”). These include Asp30 and Arg175 of the N-domain, Asp296 and Asp303 of the C-domain, and Arg382 of the C-tail. (B) The disruption of the salt bridge between Arg175 and Asp296 from either side by R175E or D296R charge reversal mutations dramatically increases arrestin binding to Rh*. Simultaneous reversal of both charges restores the salt bridge. Functionally the double-reversal mutation restores high arrestin selectivity for P-Rh*. These data identify the Arg175–Asp296 salt bridge as the phosphate sensor and demonstrate that receptor-attached phosphates simply break it by neutralizing the positive charge of Arg175.

Fig. 6

How do the phosphates get to the shielded Arg175? (A) In the basal conformation arrestin’s main phosphate sensor Arg175 (highlighted with the atoms shown) in the polar core is shielded, whereas numerous other positively charged residues in the N-domain (highlighted) are highly exposed. (B) Lysines 14 and15 (highlighted with the atoms shown in panel A) in β-strand I interact with receptor-attached phosphates, as evidenced by the progressive decrease of P-Rh* binding with neutralization and reversal of their charges. (C) The K15A mutation dramatically reduces WT arrestin binding to P-Rh*. However, in the context of arrestin mutants in which the polar core is already disrupted (R175E or D296R), the effects of the same K15A mutation on P-Rh* binding are mild (K15A+R175E and K15A+D296R). Importantly, the K15A mutation suppresses arrestin binding to dark P-Rh (which is mediated solely by phosphate interactions) in any context, supporting the identification of Lys15 as one of the residues directly binding phosphates. Thus, the presence of Lys15 is required for arrestin binding to P-Rh* only when the polar core is intact, suggesting that its function is to “meet” the phosphates first and then “guide” them to the polar core (adapted from Vishnivetskiy et al., 2000).

Fig. 7

Three sets of intramolecular interactions hold arrestin in its basal (inactive) conformation. Side chains of participating residues are shown in CPK: (1) TE, the residues participating in the 3-element interaction; (2) polar core; (3) DI, hydrophobic residues participating in the extensive interaction between the bodies of the two domains are shown in lighter (N-domain residues) or darker (C-domain residues) pattern. The inter-domain hinge is also highlighted.

Fig. 8

The 3-element interaction: another “clasp” holding arrestin in its basal state. (A) The 3-element interaction between β-strand I and α-helix I of the N-domain and β-strand XX of the C-tail involves triplets of bulky hydrophobic residues in each element (Val11+Ile12+Phe13, Leu103+Leu107+Leu111, and Phe375+Val376+Phe377, respectively). Disrupting the 3-element interaction by replacing the hydrophobic residues with alanines in β-strand XX (3A), β-strand I (N3A), or α-helix I (h3A) yields constitutively active mutants, suggesting that it is disrupted in WT arrestin by P-Rh*. Two highly conserved lysines (Lys14 and Lys15) are present immediately downstream of the participating hydrophobic residues in β-strand I. The movements accompanying phosphate binding to Lys15 and Lys14 likely melt the short β-strand I, disrupting the hydrophobic interaction of adjacent residues with the arrestin C-tail and α-helix (adapted from Vishnivetskiy et al., 2000).

Fig. 9

Current model of arrestin–receptor interaction. Receptor-attached phosphates (blue circles with yellow stripe) bind Lys14 and Lys15 (blue), forcing one of them to flip over. The resulting distortion of β-strand I disrupts the 3-element interaction, allowing Lys14 and Lys15 with bound phosphates to move towards phosphate-binding residues in β-strand X (Lys166, Arg171, Arg175, and Lys176; blue). Ultimately the phosphates neutralize the charge on Arg175, thereby breaking its salt bridge with Asp296 and destabilizing the polar core. The breakup of the 3-element interaction also releases the arrestin C-tail (green), removing the remaining positive charge (Arg382; orange) from the polar core. Having lost both interaction partners, the amphipathic α-helix I (magenta) swings out and participates in receptor binding. After the constraints holding the 2 arrestin domains in their basal orientation are released, the N-domain and C-domain move relative to each other. This movement brings all receptor-binding elements into contact with the receptor, so that arrestin encloses its cytoplasmic tip. In the bound form the “patch” of the phosphate-binding residues is in contact with receptor-attached phosphates and an unidentified arrestin elements (possibly one of the flexible loops; red) occupies the inter-helical cavity that opens upon receptor transition into its active state.

Fig. 10

The selectivity of visual arrestin is unrivaled in the family. Comparative binding of visual (rod) arrestin and its 2 non-visual “cousins” (arrestin2 and arrestin3) to the 4 functional forms of rhodopsin (A), m2 muscrarinic cholinergic receptor (B), and β2-adrenergic receptor (C). Visual arrestin binding to the active phosphorylated form (P-Rh*) of its cognate receptor, rhodopsin, is many times greater than its binding to inactive P-Rh or unphosphorylated Rh*. In contrast, the difference in binding of both non-visual arrestins to inactive and active phosphoreceptors is about 2-fold or less. The difference in their binding to phosphorylated and unphosphorylated forms of the receptors is also much less impressive. Visual arrestin also shows much stronger preference for its cognate receptor, rhodopsin, over the m2 mAChR and the b2AR, whereas non-visual arrestins bind all 3 receptors comparably. Both arrestin2 and 3 show a similar selectivity profile for rhodopsin binding, demonstrating that the difference lies in the functional characteristics of the arrestin rather than the receptor (adapted from Gurevich et al., 1995).

Fig. 11

Phosphate-dependent activation mechanism is conserved in the arrestin family. Comparative binding of WT and mutant forms of visual, arrestin2, and arrestin3 to phosphorylated and unphosphorylated light-activated rhodopsin (A) and b2AR (B). The mutations destabilizing the polar core and 3-element interaction produce a similar phenotype in all 3 arrestins: enhanced binding to the unphosphorylated active cognate and to the phosphorylated active non-cognate receptor, but not to the unphosphorylated active non-cognate receptor. The mutations in visual, arrestin2 and 3, respectively, are designated, as follows: RE: R175E, R169E, and R170E; 3A, triple alanine substitution in the C-tail: FVF→AAA (375–377), IVF→AAA (386–388), and IVF→AAA (386–388); T, C-terminal truncation yielding 1–378, 1–382, and 1–392 mutants.

Fig. 12

Arg18: an additional phosphate-binding residue in visual arrestin. (A) The neutralization (R18A) and reversal (R18E) of the charge in position 18 progressively reduces the binding of visual arrestin to P-Rh* and dramatically impedes its binding to dark P-Rh, which is mediated solely by phosphate interactions. These data identify Arg18 as yet another phosphate-binding residue. (B) Arg18 in visual arrestin is located in the loop between β-strands I and II, where no other arrestin has a positive charge. Instead, a proline is present in the equivalent position in non-visual arrestin2 (P14) and cone arrestin (P13) (adapted from Sutton et al., 2005). Non-visual and cone arrestins often have additional prolines and/or glycines in this loop. Higher rigidity of this loop and the presence of the additional phosphate-binding residue in visual arrestin likely contribute to its unparalleled selectivity for the active phosphorylated receptor.

Fig. 13

Enhanced stability of the N-domain β-strand sandwich contributes to the receptor specificity of visual arrestin. (A) The substitution of Val90 with serine in visual arrestin greatly reduces its receptor specificity, allowing the mutant to bind the P-m2 mAChR* almost as well as arestin2. However, the introduction of valine in the homologous position in arrestin2 (S86V) does not appreciably change its binding selectivity, suggesting that there are other structural features in non-visual arrestins contributing to their broad receptor specificity. (B) The interactions between the 2 “layers” of the β-strands in visual arrestin are stabilized by the presence of Val90 (highlighted) that interacts with several hydrophobic partners (Val45, Val57, Val59, and Phe118; highlighted), whereas arrestins 2 and 3 have Ser86 and Ala87, respectively, in the equivalent positions (adapted from Han et al., 2001). Note that arrestin in panel B is shown with its N-domain on the right-hand side (in contrast to Figs. 6, 7, 9, and 14) to show the interactions of Val90 better.

Fig. 14

Receptor-binding elements are localized on the concave sides of both arrestin domains. Upper panel: “side” view of the arrestin molecule. Lower panel: view down the cavities of both domains on the receptor-binding surface. Receptor-binding elements are pattern-coded, as indicated. Element identification is based on truncation (Gurevich & Benovic, 1992, 1993) and site-directed mutagenesis (Gurevich & Benovic, 1995, 1997; Sutton et al., 2005; Vishnivetskiy et al., 2000), chimera construction (Gurevich et al., 1995; Vishnivetskiy et al., 2004), chemical modification and H/D-exchange (Ohguro et al., 1994), peptide inhibition (Pulvermuller et al., 2000), and epitope insertion (Dinculescu et al., 2002).

Fig. 15

The structural basis for the formation of different arrestin–receptor complexes. Arrestin binding to the receptor is mediated by multiple elements in both proteins. Different relative positioning of the arrestin-binding parts of the receptor results in the variation of arrestin orientation on the receptor, in some cases with profound functional consequences. For the sake of simplicity, the example shows only 2 elementary interactions. One of the flexible loops of arrestin (filled pink circle) moves into the inter-helical cavity that opens upon GPCR activation (red oval) and interacts with receptor elements that become accessible upon activation (red circle within the red oval). At the same time, phosphate-binding arrestin residues (2 yellow elements on β-strand X and one on β-strand I) interact with receptor-attached phosphates (filled red circles). The difference in the positions of receptor-attached phosphates forces arrestin to orient itself in different ways on the receptor to achieve the best fit. The views of the receptor from its cytoplasmic side and of the arrestin with its receptor-binding side facing the receptor (away from the reader) are shown.

Fig. 16

The formation of a transient arrestin–receptor complex prevents receptor down-regulation. In contrast to WT arrestin2, R169E mutant binds with high affinity to unphosphorylated b2AR* (A), thereby competing with GRK2 and inhibiting receptor phosphorylation (B). The expression of the R169E mutant in HEK cells at levels sufficiently high to out-compete GRKs and endogenous arrestins and to form complexes with unphosphorylated b2AR* prevents the loss of the receptor even after very long agonist exposure, in contrast to the overexpression of WT arrestin2 (C). However, concomitant overexpression of GRK2 (that ensures that the receptor is phosphorylated faster, so that both WT and R169E arrestins bind phosphorylated receptor) “rescues” receptor down-regulation (C). The R169E mutant facilitates receptor cycling in two ways. First, deactivation of P-b2AR (due to agonist loss in the endosome) does not dramatically reduce arrestin binding, whereas deactivation of unphosphorylated b2AR sharply reduces arrestin binding, so that upon internalization the mutant dissociates from the receptor faster. Second, after the release of WT arrestin the emerging multi-phosphorylated receptor must be fully dephosphorylated before it can be recycled, whereas the dissociation of R169E immediately releases the unphosphorylated recycling-competent receptor. Thus, the decrease in stability of the arrestin–receptor complex facilitates receptor cycling, and rapid recycling, in its turn, prevents receptor down-regulation (Pan et al., 2003).

Fig. 17

Relative sizes of the receptor, arrestin, and its non-receptor-binding partners. Crystal structures of rhodopsin (Okada et al., 2004), arrestin (Hirsch et al., 1999) (shown from the “back”, that is, from the side that does not interact with receptors), the arrestin-binding N-terminal domain of clathrin heavy chain (ter Haar et al., 1998), the arrestin-binding appendage of β2-subunit of the AP-2 clathrin adaptor complex (Owen et al., 2000; Traub, 2003), ARF6 (Shiba et al., 2003), c-Src (Xu et al., 1997b), JNK3 (Xie et al., 1998), MEK1 (Ohren et al., 2004), Mdm2 (Grasberger et al., 2005), ERK2 (Canagarajah et al., 1997), the kinase domain of Raf (Wan et al., 2004), the catalytic domain of phosphodiesterase PDE4 (Huai et al., 2003), and the sec7 (exchange activity) domain of ARNO (Cherfils et al., 1998). The structures are shown to scale. The relative sizes of arrestin and its partners suggest that only a few of them can be bound to a single receptor-associated arrestin simultaneously.