Each rhodopsin molecule binds its own arrestin

Abstract

Arrestins (Arrs) are ubiquitous regulators of the most numerous family of signaling proteins, G protein-coupled receptors. Two models of the Arr–receptor interaction have been proposed: the binding of one Arr to an individual receptor or to two receptors in a dimer. To determine the binding stoichiometry in vivo, we used rod photoreceptors where rhodopsin (Rh) and Arr are expressed at comparably high levels and where Arr localization in the light is determined by its binding to activated Rh. Genetic manipulation of the expression of both proteins shows that the maximum amount of Arr that moves to the Rh-containing compartment exceeds 80%, but not 100%, of the molar amount of Rh present. In vitro experiments with purified proteins confirm that Arr “saturates” Rh at a 1:1 ratio. Thus, a single Rh molecule is necessary and sufficient to bind Arr. Remarkable structural conservation among receptors and Arrs strongly suggests that all Arr subtypes bind individual molecules of their cognate receptors.

G protein-coupled receptors (GPCRs) are the most numerous family of signaling proteins. Arrestins (Arrs) bind to phosphorylated activated receptors terminating G protein activation and redirecting the signaling to alternative pathways, in which Arr serves as an adaptor for protein kinase Src, ubiquitin ligase Mdm2, phosphodiesterase PDE4, or as a scaffold for MAP kinase cascades (1, 2). The dimerization of some GPCRs (3, 4) led to the idea that one Arr needs both receptors in a dimer to bind (5). The stoichiometry of the Arr–receptor complex has profound implications for the mechanism of Arr-mediated desensitization of GPCRs and receptor trafficking. The composition of the complex, i.e., the size and structure of the “signalosome,” also determines its scaffolding potential and the ability to initiate the “second round” of signaling by organizing downstream proteins, such as Src, MAP kinases, ubiquitin ligase, etc.

Rod photoreceptors provide a unique model where the stoichiometry of the Arr–receptor interaction can be determined in vivo because they express comparable amounts of the receptor (Rh) and Arr at very high levels unparalleled in any other cell type (6–8). In dark-adapted rods, Arr is predominantly localized in the inner segment, where it is held by its low-affinity binding to microtubules that are abundant in this compartment (9). Light induces Arr translocation to the outer segment (OS), where it is retained in bright light through binding to light-activated Rh (8, 9). The expression of Rh and Arr can be genetically manipulated independently, and the extent of Arr translocation can be used to determine the stoichiometry of their interaction.

Results

Hemizygous Rh (Rh^+/−) and Arr (Arr^+/−) mice express about half of these respective proteins as compared with wild-type animals (10, 11). To obtain mice expressing different ratios of Arr and Rh, we bred Rh^+/−, Arr^+/−, and Rh^+/−/Arr^+/− animals and compared the content of Rh and Arr in their retinas with that of wild-type mice (Table 1). Both proteins were measured by quantitative Western blot in the homogenates of whole eyecups by using the corresponding purified proteins to construct calibration curves (Fig. 1). The results indicate that in both cases the elimination of one allele reduces the expression by half, so that the Arr/Rh ratio in wild-type and doubly hemizygous animals is similar (0.8:1 and 0.94:1, respectively), whereas in Arr^+/− and Rh^+/− mice it is significantly shifted in the expected direction, to 0.38:1 and 1.74:1, respectively (Table 1). The absolute levels of Rh and Arr expression in wild-type retinas obtained by these measurements are in good agreement with previous reports (6–8).

Table 1.

Light-dependent translocation of arrestin to the outer segment as a function of arrestin and rhodopsin expression

Genotype	Rh content, nmol per retina	Arr content, nmol per retina	Arr/Rh molar ratio	Arr, % in OS (dark)	Arr, % in OS (light)	Arr in OS (light), nmol	Rh occupied by Arr, %
WT	0.40 ± 0.05 (7)	0.32 ± 0.04 (7)	0.80	1.8 ± 0.8	81.4 ± 1.3	0.26	65
A+/−	0.40 ± 0.03 (5)	0.15 ± 0.02 (5)	0.38	3.1 ± 1.1	88.8 ± 0.5	0.13	33
A+/−Rh+/−	0.18 ± 0.02 (4)	0.17 ± 0.02 (4)	0.94	1.4 ± 0.5	89.1 ± 1.0	0.15	83
Rh+/−	0.19 ± 0.01 (4)	0.33 ± 0.02 (4)	1.74	1.1 ± 0.4	43.5 ± 6.5	0.14	75
tg+ A−/−	0.32 ± 0.02 (4)	0.77 ± 0.04 (4)	2.41	1.2 ± 0.2	29.3 ± 6.2	0.23	72
tg+ A+/+	0.37 ± 0.05 (5)	1.01 ± 0.03 (5)	2.70	1.3 ± 0.7	23.4 ± 3.4	0.24	65

Mice with the indicated genotypes were dark-adapted overnight (dark) or exposed to 2,700 lux for 1 h (light). One eyecup from each mouse was fixed and processed for immunohistochemistry (9), whereas the other was homogenized for Rh and Arr quantification by Western blot as described (7, 37), by using the corresponding purified proteins to construct calibration curves. The proportion of Arr localized in the OS was quantified by the intensity of Arr immunostaining in 10 images per animal from 3–5 animals per genotype per light condition. Means ± SD are shown. The data were analyzed by one-way ANOVA with genotype as a main factor. The Rh content of A^+/−Rh^+/− and Rh^+/− and the arrestin content of A^+/− and A^+/−Rh^+/− were statistically different from all other genotypes (P < 0.0001) but were not different from each other. Arr expression in tg⁺ A^−/− and tg⁺ A^+/+ animals was statistically different from all other genotypes (P < 0.0001). Statistical significance of the differences in the percent of Arr in the OS in the light are indicated in Fig.1. The amount of Arr in the OS (nanomoles) was calculated by multiplying the Arr content by the percentage of Arr in the OS in the light. The percentage of Rh occupied was determined by dividing this value by the total Rh content.

Fig. 1.

The extent of light-dependent Arr translocation is determined by the Arr/Rh expression ratio. (A) Mice with the indicated genotypes were dark-adapted overnight (DARK) or exposed to 2,700 lux for 1 h (LIGHT). One eyecup was fixed and processed for Arr immunohistochemistry (9). The positions of the outer (OS) and inner segments (IS) and of the outer nuclear layer (ONL) are indicated. (B) The proportion of Arr localized in the OS was quantified by the intensity of Arr immunostaining (green) in 10 images per animal from 3–5 animals per genotype per light condition. Means ± SD are shown. The data for light- and dark-adapted mice were analyzed separately by one-way ANOVA with genotype as a main factor. Statistical significance of the differences is indicated above the corresponding bars: ∗, P < 0.001; ∗∗, P < 0.0001. (C) Typical Western blots for Arr and Rh (Rh) in mice with the indicated genotypes and corresponding standards.

If two Rh molecules bind one Arr, virtually complete Arr translocation to the OS in the light would only be expected in Arr^+/− animals but not in wild-type mice. In contrast, in the case of a 1:1 interaction, the translocation would be incomplete only in Rh^+/− animals that express Arr in excess of Rh. We invariably observed virtually complete Arr translocation in the light-adapted retinas of mouse lines expressing more Rh than Arr (Fig. 1). Quantitative image analysis shows that in wild-type mice and the other two lines that express excess Rh, 81–89% of Arr translocates to the OS in the light (Table 1). Not surprisingly, in Rh^+/− mice, only about half of that amount of Arr moves to the OS (Table 1). These data are consistent with 1:1 binding and cannot be reconciled with the model of Rh dimer interacting with just one Arr molecule. Based on the 1:1 model, the extent of Arr translocation, and the absolute expression levels of both proteins, we calculated that in the light-adapted retinas of these mice, 65–83% of Rh is occupied at steady state by bound Arr, with the exception of Arr^+/− mice, where Arr is limiting and Rh occupancy is about half that in wild type (33%) (Table 1). The Rh occupancy in both Rh^+/− lines (75–83%; Table 1) exceeds the theoretical maximum for a 1:2 model up to 1.66-fold.

To further extend the range of Arr/Rh ratios in vivo we also analyzed transgenic mice expressing mutant mouse Arr(L374A,V375A,F376A) under the control of the Rh promoter. Earlier. we showed that this mutant demonstrates light-dependent translocation that is qualitatively similar to wild-type Arr but proceeds more slowly (9). However, after 60 min, the translocation of both WT and mutant Arr reaches its maximum (9). Here, we used the transgenic line expressing this Arr at 240% of wild type and compared Arr translocation in tg⁺Arr^−/− and tg⁺Arr^+/+ mice that have Arr/Rh ratios of 2.4 and 2.7, respectively (Table 1). As expected, in animals with a large excess of Arr, the extent of its light-dependent translocation to the OS was much lower (29% and 23%, respectively, which is equivalent to Rh occupancy of 65–72%) (Fig. 1). Importantly, the percentage of Arr that moves to the OS in the light progressively decreases with an increasing excess of Arr over Rh, but the absolute amount of Arr in the OS is proportional to Rh expression and remains essentially the same: 0.23–0.26 nmol in Rh^+/+ mice, and 0.13–0.15 nmol in both Rh^+/− lines. These numbers correspond to 0.65–0.83 mol of Arr per 1 mol of Rh. Thus, Arr translocation in every genetic background is only consistent with the 1:1 model of Arr–Rh binding in vivo.

To further test this model and ascertain that other proteins present in live photoreceptors do not affect the translocation of Arr to the OS, we reproduced similar Arr/Rh ratios (0.5, 1.1, 2.2, and 3.3) in vitro with purified proteins carefully quantified by amino acid analysis. In these experiments, we mixed recombinant bovine Arr with bovine Rh in native disk membranes that was phosphorylated with endogenous Rh kinase (12) and fully regenerated with 11-cis-retinal, illuminated the samples for 5 min at 37°C, and then pelleted Rh along with bound Arr. Control samples contained the same amounts of Arr and no Rh. Equal aliquots of the original samples, pellets, and supernatants were resolved by SDS/PAGE and stained with Coomassie blue (Fig. 2). In addition, we measured the amounts of Rh and Arr in the original samples, pellets, and supernatants by quantitative Western blot with appropriate standards (Fig. 2). We found that Rh pellets quantitatively, whereas the amount of Arr pelleted in the absence of Rh does not exceed 4.5%. Under these conditions we observed clear saturation of Rh by increasing concentrations of Arr, reaching up to 0.9 mol of specifically bound Arr per 1 mol of Rh (Fig. 2). These results are in perfect agreement with our in vivo data (Fig. 1) and can only be rationalized in the context of the model of one Arr molecule binding one Rh molecule.

Fig. 2.

Rh “saturation” with Arr is achieved at equimolar binding. (A) Purified bovine Arr (142, 286, 572, and 858 pmol) was incubated for 5 min in the light with (+P-Rh*) or without (−P-Rh*) 10 μg (261 pmol) of phosphoRh in 25 μl of 50 mM Mops-Na, pH 7.2, 100 mM NaCl. Rh with bound Arr was pelleted through a 50-μl cushion of the same buffer with 0.2 M sucrose. Equal aliquots of the sample before centrifugation (Load) (1/14), Pellet (1/4), and supernatant (Sup) (1/4) were subjected to SDS/PAGE. Arr was visualized by Coomassie blue staining. (B) The absolute amount of Arr and Rh in the pellet was measured by quantitative Western blot. Arr binding is plotted as a function of the Arr/Rh molar ratio in the sample. Means ± SD of two experiments are shown. The analysis of the binding data (by using GraphPad Prizm) yields Bmax (saturation) at 0.99 ± 0.08 mol/mol.

Discussion

Because of their mechanism of action, a molecule of agonist or a photon of light can activate only a single GPCR or visual pigment, respectively. Therefore, until recently, it was implicitly assumed that a single GPCR is a signaling unit that activates its cognate G protein and becomes subsequently inactivated by Arr binding (recently reviewed in ref. 13). The observations by indirect and direct methods that many GPCRs (14), including Rh (4), dimerize, along with the reports that GPCR heterodimerization affects the pharmacological profile of their signaling (14, 15), challenged this view. The diameter of the cytoplasmic tip of the only GPCR for which the crystal structure has been determined, dark (inactive) Rh (16, 17), is ≈40 Å, which is about half the length of the long axis of the two proteins that tightly bind to it, heterotrimeric G protein (18) and Arr (19–21). Arr is an elongated molecule, which consists of two cup-like domains. The individual domains are a≈35 Å in diameter and contain receptor-binding elements on the concave sides (22). This suggestive geometry led to the hypothesis that one Arr may bind a Rh dimer, with each domain interacting with an individual receptor molecule (5, 23). This model predicts that two Rhs are necessary to bind one Arr, which is clearly at odds with the stoichiometry of this interaction determined in vivo (Fig. 1; Table 1) and in vitro (Fig. 2).

Visual Arr is known to self-associate, forming dimers and tetramers in solution (24–26). Theoretically, the binding of an Arr dimer to a Rh dimer could yield the same 1:1 stoichiometry as monomer–monomer interaction. In a recent study of the structure and functional capabilities of Arr oligomers in solution under physiological conditions (26), we used long-range distance measurements between spin-labeled Arr units to follow its oligomerization state. We found that, upon addition of phosphorylated activated Rh sufficient to bind all Arr present in the assay, all distances characteristic for Arr oligomers completely disappear, clearly demonstrating that only Arr monomer is competent to bind Rh. Interestingly, using the same approach we found that the Arr tetramer binds microtubules without dissociating (26), ruling out nonspecific effects of added protein on the state of Arr self-association.

It has been shown that the reduction of Rh concentration in discs in Rh^+/− mice makes the kinetics of photoresponse faster, possibly by increasing the lateral diffusion of Rh that facilitates its interactions with transducin (10). Interestingly, the “saturation” of Rh with Arr tends to be higher (0.75–0.83 mol/mol) in the two Rh^+/− lines than in the Rh^+/+ lines (0.65–0.72 mol/mol) (Table 1), suggesting that the reduction of Rh concentration in the disk membrane favors Arr binding as well. Thus, it appears that the relief of Rh “overcrowding” facilitates its interactions with all signaling proteins. In the dark, Arr is retained in the inner segment and cell bodies because of its low-affinity binding to the microtubules, which are abundant in these compartments (9, 27). Apparently, microtubules in the rod have ample spare capacity for Arr binding, because the proportion of Arr in the OS remains virtually the same in dark-adapted rods when Arr expression ranges from 50% to 340% of the wild-type level (Table 1). The ability of microtubules to bind Arr tetramer (26) may explain their enormous spare Arr-binding capacity revealed in our experiments with mice expressing more than three times as much Arr as wild type (Fig. 1; Table 1). Interestingly, even at a very high Arr/Rh expression ratio, the proportion of Arr-occupied Rh “saturates” at 70–80%, which almost exactly corresponds to the relative expression of these proteins in wild-type animals (Table 1 and ref. 8). This indicates that mice express just enough Arr to occupy all Rh that is binding-competent [i.e., light-activated, light-activated phosphorylated, dark phosphorylated, and phosphoopsin (9, 28)] at equilibrium in fully light-adapted rods. Some of the Rh in light-adapted OS exists as unphosphorylated opsin that does not bind Arr and does not support Arr translocation when the equilibrium is shifted toward this form by hydroxylamine treatment (9). It should be noted that this decreases the effective amount of Rh competent to bind Arr, further strengthening our conclusion that only the binding of Arr by each individual Rh can account for the observed extent of Arr translocation in vivo.

In summary, our data in live photoreceptors in genetically modified mice expressing Arr and Rh at molar ratios ranging from 0.4 to 2.7 and in vitro with purified bovine Arr and Rh mixed at 0.5–3.3 molar ratios clearly demonstrate that Arr binding to Rh saturates at 0.8–0.9 mol/mol. These data are obviously inconsistent with the proposed model of one Arr binding two Rh molecules in a dimer (5). In conjunction with our recent finding that only an Arr monomer can bind Rh (26), these results lend strong support to the 1:1 binding model (29). Rod photoreceptors demonstrate single-photon sensitivity in various species (reviewed in ref. 30). Because an individual photon can induce isomerization of just one chromophore, it can activate only a single Rh molecule. Rh phosphorylation and subsequent Arr binding is absolutely necessary to maintain normal kinetics of photoresponse recovery (11, 31, 32). Therefore, Arr binding to a single active phosphoRh makes perfect sense from the biological viewpoint. Importantly, our experiments do not answer the question whether Rh or other GPCRs exist as monomers, dimers, or higher-order oligomers or whether Arr has a preference for any of these forms of the receptor. However, our data demonstrate that, regardless of the oligomerization state, each Rh molecule binds its own Arr.

Based on the remarkable structural and functional conservation in the Arr family (reviewed in refs. 2 and 33), it is most likely that the stoichiometry of the interaction of nonvisual Arrs with other GPCRs is the same. The Arr–receptor complex often serves as a nucleus of the “signalosome” (reviewed in refs. 1 and 34), where receptor-bound Arr recruits and activates a variety of signaling proteins. Although receptor-bound Arrs were shown to interact with >20 different proteins (1), a unitary complex is too small to accommodate more than three to five nonreceptor partners simultaneously (2). Therefore, it is tempting to speculate that receptor oligomers with multiple bound Arr molecules solve this problem, suggesting that self-association of GPCRs may be the mechanism enabling post-G protein Arr-dependent signaling. The recent finding that rod Arr also interacts with several nonreceptor binding partners (35) indicates that Rh dimerization may serve the same function.

Methods

Animals and Tissue Preparation.

The generation of Arr (11) and Rh (24) knockout mice, as well as transgenic mice expressing Arr mutant [under the control of the Rh promoter (36)] at different levels has been described (9). Animal research was conducted in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the institutional Animal Care and Use Committee. Animals were dark-adapted for 12 h or light-adapted for 1 h at 2,700 lux. The mice were anesthetized with isoflurane and killed by cervical dislocation, and the eyes were enucleated. Light and dark adaptation, tissue processing for Arr detection by immunohistochemistry, and quantitative image analysis were performed as described (9) [see supporting information (SI) Methods]. One eyecup from each mouse was fixed and processed for immunohistochemistry, whereas the other was homogenized for protein quantification as described (7).

Quantification of Arr and Rh.

Arr and Rh content was quantified by Western blot as described (37) using the corresponding purified proteins carefully quantified by amino acid analysis to construct calibration curves (SI Methods and SI Fig. 3) and visualized with monoclonal anti-Arr F4C1 (38) and anti-Rh 4D2 (39) antibodies. To increase the reliability of the measurements, we added an equal amount of total retinal protein from Arr^−/− and Rh^−/− mice to the Arr and Rh standards, respectively. To control for the possible variability of the yield and to avoid errors associated with variable contamination with vitreous fluid, we used tubulin (rather than total protein) for normalization. Tubulin content was measured by quantitative Western blot in all retina homogenates, with calibration curves constructed by using known amounts of purified tubulin (Cytoskeleton, Denver, CO). The calculations were based on the assumption that the highest tubulin content (22.6 ± 2.4 μg per retina; n = 12) represents 100% yield of retinal material. The samples where tubulin content was <70% of this level were disqualified.

Protein Expression/Purification and in Vitro Binding Assay.

Mouse and bovine rod Arr was subcloned between NcoI and HindIII sites into pTrcHisB vector (Invitrogen, Carlsbad, CA), expressed in Escherichia coli and purified, following the procedure described for bovine rod Arr (40) (with minor modifications for the mouse protein). Bovine Rh was phosphorylated by endogenous Rh kinase in purified rod OSs and then purified and fully regenerated by 11-cis-retinal, as described (28), with some modifications. Because our previous studies demonstrated that only Rh containing two or more phosphates binds Arr with high affinity (41, 42), we phosphorylated Rh for 3 h in bright light in the presence of 3 mM ATP and 1 mM GTP (the latter was added to prevent Rh “shielding” by bound transducin to reduce the proportion of unphosphorylated Rh). Mass-spectrometric analysis of this Rh preparation (kindly performed by Drs. M. Kennedy and J. B. Hurley, as described in refs. 9 and 43) confirmed that >93% of Rh molecules had two or more phosphates. The quantification of bovine and mouse Arr and bovine Rh was confirmed by amino acid analysis (as described in SI Methods). In the in vitro binding assay, indicated amounts of purified bovine Arr were mixed in the dark with 10 μg (261 pmol) of bovine Rh and illuminated (2,000 lux) for 5 min in 25 μl of 50 mM Mops-Na, pH 7.2, and 100 mM NaCl. The samples were cooled on ice, and Rh-containing disk membranes along with bound Arr were pelleted through a 50-μl cushion of the same buffer with 0.2 M sucrose. Aliqouts of the sample, pellet, and supernatant were analyzed by SDS/PAGE (Coomassie blue staining) and quantitative Western blot, as described above.

Abbreviations

GPCR

G protein-coupled receptor

OS

outer segment

Arr

arrestin

Rh

rhodopsin.