The hydrodynamic properties of dark- and light-activated states of n-dodecyl beta-D-maltoside-solubilized bovine rhodopsin support the dimeric structure of both conformations.

Rhodopsin (Rho) has been extracted in n-dodecyl beta-D-maltoside (DM) from bovine retinal rod outer segments and purified to homogeneity by affinity chromatography on concanavalin A-Sepharose. Because chemical cross-linking of Rho and photoactivated Rho (Rho*) provided initial evidence for the oligomeric nature of the photoreceptor protein, we carried out a hydrodynamic characterization of the native and activated conformations of detergent-solubilized Rho. The molecular weights of the complexes between dark and photoexcited states of Rho and DM were determined by gel filtration chromatography on Sephacryl S-300, in the presence of 0.1% DM. Subtracting the size of the corresponding detergent micelles resulted in molecular masses of 78 kDa for native Rho and 76 kDa for Rho*. The measured content of 0.97 g of detergent/g of protein resulted in a calculated partial specific volume of 0.765 cm(3)/g for the protein-detergent complex and a molar mass of 64-65 kDa for the protein moiety. The sizes of Rho.DM and Rho*.DM complexes were also evaluated by sedimentation on 10-30% sucrose gradients, in the presence of 0.1% DM, and molecular masses of about 60 kDa were estimated for both the dark- and light-activated states of the photoreceptor protein. The size of Rho was determined to be 65,300 and 69,800 Da, respectively, when the purified Rho.DM complex was either chromatographed on Sephacryl S-300 or ultracentrifuged on sucrose gradients in the absence of DM. All these results were consistent with a dimeric quaternary structure for both conformations of Rho. Additionally, the functional integrity of the purified photoreceptor protein following gel filtration chromatography and ultracentrifugation was demonstrated by three criteria as follows: (i) its characteristic UV-visible absorption spectra, (ii) its capability to photoactivate transducin, and (iii) its ability to serve as a substrate for rhodopsin kinase.

G protein-coupled receptors (GPCRs) 1 are a large group of integral membrane proteins that respond to environmental signals and initiate transduction pathways that activate cellular processes. In general, the activation of the receptor by binding of an extracellular signal or by light absorption triggers a conformational change in its structure, which then activates a peripherally membrane-associated heterotrimeric G protein. One of the most important unanswered questions is how these receptors operate and couple to their cognate G proteins. A growing body of recent pharmacological, biochemical, and biophysical data strongly suggests that GPCRs are organized as functional homo-and heterodimers as well as higher order oligomers (1,2). Oligomerization of GPCRs may cluster these receptors in particular regions of the membrane. This process could be critical for the proper kinetics of GPCR signaling, selectivity, desensitization, and internalization. Receptor maturation during biosynthesis and translocation to the plasma membrane could also benefit from oligomerization (3). Additionally, the formation of heteromers may expand the repertoire of GPCRs and their physiological responses.
The photoreceptor protein rhodopsin (Rho) is a prototypical GPCR, which is involved in the molecular transformation of light energy into a neuronal signal transmitted to the secondary neurons of the retina, and ultimately to the brain, during scotopic vision. Rho is composed of the protein opsin covalently linked to 11-cis-retinal. The light-induced isomerization of 11cis-retinal to its all-trans configuration leads to a conformational change in Rho that triggers the signal transduction cascade via reactions of the heterotrimeric G protein transducin (T). T, which is arranged as two units, the ␣ subunit (T ␣ ), and the ␤␥-complex (T ␤␥ ) transduces the visual stimuli by activating a cGMP phosphodiesterase. Fast depletion of cGMP in the rod outer segments results in the closure of cGMP-gated channels located in the plasma membrane of the ROS and the blockage of the inward flux of Na ϩ and Ca 2ϩ ions. The reduction in the circulating electrical current leads to the hyperpolarization of the membrane and to the generation of a neuronal signal.
Electron microscopy, low angle x-ray diffraction, and neutron diffraction analyses have indicated that Rho is a monomer randomly distributed in the plane of the membrane without any special ordering (4 -8). Additionally, biophysical measurements using high speed flash photometry and microspectrophotometry have shown that Rho undergoes rapid rotational and lateral diffusion (9,10). Cross-linking studies have also suggested a monomeric organization for rhodopsin (11,12). Accordingly, the concept of how Rho functions in the disk membrane of retinal ROS has been dominated by the hypothesis that Rho rapidly diffuses as a monomeric unit in the fluid membranes, which are mostly composed of highly unsaturated phospholipids, to encounter T. However, Fotiadis et al. (13,14) and Liang et al. (15) have recently demonstrated by atomic force microscopy that both Rho and opsin molecules are packed as dimers in isolated murine disk membranes at both low and room temperatures. To contribute to the clarification of this controversy, we have analyzed here the hydrodynamic properties of n-dodecyl ␤-D-maltoside (DM)-solubilized bovine Rho and photoactivated Rho (Rho*), with the purpose of elucidating their native quaternary structures.
Preparation of ROS and Washed Membranes-ROS membranes were isolated from frozen bovine retinas as described previously (16). Darkdepleted ROS membranes were prepared by washing ROS with 5 mM Tris-HCl (pH 7.4), 2 mM EDTA, and 5 mM ␤-mercaptoethanol until no significant amount of peripheral proteins was released with the wash buffer. ROS membranes and dark-depleted ROS membranes were stored in the dark at Ϫ70°C.
Preparation of an Enriched-fraction of Rhodopsin Kinase-Freshly prepared ROS membranes were washed three times with a buffer containing 70 mM potassium phosphate (pH 6.8), 5 mM magnesium acetate, 5 mM ␤-mercaptoethanol, and 0.1 mM phenylmethylsulfonyl fluoride. Following centrifugation, the isotonically washed ROS pellet was hypotonically extracted with 5 mM Tris-HCl (pH 7.4), 5 mM magnesium acetate, 5 mM ␤-mercaptoethanol, and 0.1 mM phenylmethylsulfonyl fluoride. An enriched fraction of rhodopsin kinase was obtained in the supernatant produced after centrifugation. The whole procedure was carried out at 4°C, in the dark under red light.
Purification of Rho and T-Rho was extracted from the ROS membranes, under dim red light, with 1% DM in Rho buffer (50 mM Hepes (pH 6.6), 140 mM sodium chloride, 3 mM magnesium chloride, 20% glycerol, 2 mM calcium chloride). The sample was then diluted 10 times with Rho buffer to reduce the concentration of DM to 0.1% and centrifuged at 100,000 ϫ g, for 30 min, at 4°C. The resulting supernatant was transferred to a different tube, and Rho was purified by batchwise affinity chromatography on concanavalin A-Sepharose (17), using DM instead of n-octyl-␤-D-glucopyranoside as the detergent.
T was isolated from ROS membranes prepared under room light, at 4°C, following the affinity procedure carried out by Kü hn (18). GTP (ϳ100 M) was used to elute T from the washed illuminated ROS membranes, and T was further purified to homogeneity by anion exchange chromatography on a diethylaminoethylcellulose DE52 column, as described elsewhere (19,20).
Cross-linking of Rho and Rho*-Samples of washed ROS membranes or purified Rho (1.28 M) were incubated in the dark or in the presence of light with sulfo-SMCC (5 mM), MBS (5 mM), o-PDM (2-8 mM), or p-PDM (2-8 mM) for 1 h at room temperature. Stock solutions of MBS, o-PDM, and p-PDM were freshly prepared in dimethyl sulfoxide, and sulfo-SMCC was dissolved in water. The reactions with sulfo-SMCC and MBS were carried out in 10 mM sodium phosphate (pH 7.2), 5 mM magnesium acetate, whereas the reactions with both phenylenedimaleimides were performed in 50 mM Tris-HCl (pH 7.5), 5 mM magnesium acetate. As controls, samples of washed ROS membranes or purified Rho were incubated with the corresponding vehicles. Additionally, the time course of cross-linking was determined by incubating purified Rho (1.28 M) with 5 mM sulfo-SMCC, MBS, or o-PDM, as described above. At designated time intervals (0 -60 min), the reactions were terminated by the addition of 20 mM ␤-mercaptoethanol. Various concentrations of purified Rho (1.28 -20.5 M) were also incubated for 1 h, at room temperature, with 5 mM sulfo-SMCC. In all cases, the samples were separated by SDS-PAGE, and the cross-linked products were stained with Coomassie Blue or Silver.
Gel Filtration Chromatography of Purified Rho and Rho* in DM-Because Rho was purified in the presence of 0.1% (1.96 mM) DM, which is above its critical micelle concentration (0.18 mM), the formation of detergent micelles was eminent, and Rho⅐DM complexes were formed. Purified Rho samples (0.5-0.8 mg) were applied to a Sephacryl S-300 size exclusion column (total volume (V t ) ϭ 50.3 ml) previously equilibrated with 50 mM Hepes (pH 6.6), 150 mM sodium chloride, 3 mM magnesium chloride, 2 mM calcium chloride, 5 mM ␤-mercaptoethanol, and 0.1% DM. Protein standards were used to calibrate the column and were chromatographed together with the photoreceptor samples, under dim red light, at 4°C. The excluded (V o ) and included volumes were determined by chromatographing blue dextran and potassium dichromate, respectively. Parallel separations using the same Sephacryl S-300 column were also carried out with purified Rho samples in the absence of protein standards. The column was run at a flow rate of 150 l/min, and the eluting proteins were simultaneously monitored at 280, 380, and 498.5 nm and subsequently separated by SDS-PAGE. The elution of Rho was also immunologically monitored by Western blot using the monoclonal antibody 1D4. The elution volume (V e ) was measured for each protein, and the corresponding K av was calculated from Equation 1, The molecular weight of the complex between Rho and DM was empirically determined by plotting the K av value of each standard versus the logarithm of its molecular weight. Additionally, a linear relationship was obtained by plotting (Ϫlog K av ) 1/2 against each Stokes radius (21). An identical methodology was employed to determine the size of the complex between purified Rho* and DM, with the exception that the whole procedure was performed under illumination. Sucrose Gradient Ultracentrifugation of Purified Rho and Rho* in DM-Linear 10 -30% sucrose gradients (4.6 ml) were prepared in 50 mM Tris-HCl (pH 8.0), 0.1 mM EDTA, 5 mM magnesium chloride, 0.15 M ammonium chloride, 0.2 mM dithiothreitol, and 0.1% DM. Marker proteins with known sedimentation coefficients were employed to calibrate the gradients, and the soluble form of a variant surface glycoprotein purified from the TEVA1 Trypanosoma evansi Venezuelan isolate, which sedimentation coefficient was recently reported (22), was also included as a standard. Samples containing Rho (0.3 mg) and all protein markers (0.3 mg) were carefully layered on top of the sucrose gradients, under dim red light. The gradients were spun at 200,000 ϫ g, for 18 h, at 4°C, in a Beckman SW 50Ti rotor. Fractions were collected from the bottom of the tubes, and aliquots were analyzed by SDS-PAGE. The migration of the Rho⅐DM complex was determined by monitoring its absorption at 280, 380, and 498.5 nm and by Western blot analyses using the monoclonal antibody 1D4. The volume of migration of each standard was plotted against its corresponding sedimentation coefficient, and the resulting linear curve was utilized to calculate the sedimentation coefficient of the Rho⅐DM complex (23). The molecular mass of the complex between Rho and DM was estimated from a calibration curve of the protein standards molecular masses versus their Stokes radii multiplied by their sedimentation coefficients. A similar procedure was employed to determine the sedimentation coefficient and size of the complex between purified Rho* and DM.
Other Procedures-The light-dependent guanine nucleotide binding activity of T was measured by Millipore filtration using [ 3 H]GMPpNp (24,25). T GTP hydrolysis assays were performed in the presence of 0.0075% phosphatidylcholine (16). Protein concentration was determined according to Bradford (26), using bovine serum albumin as protein standard. SDS-PAGE was carried out on 1.5-mm thick slab gels containing 10 or 12% polyacrylamide (27). Because heat induces the formation of high-molecular weight Rho aggregates, Rho-containing samples were not boiled prior to SDS-PAGE. For Western blot analyses, the proteins were electrotransferred from the gels to nitrocellulose filters (28). For immunodetection, the filters were incubated with the monoclonal antibody 1D4 (dilution 1:15,000). The membranes were then treated with alkaline phosphatase-conjugated secondary antibodies against mouse IgG, at a dilution of 1:2,000, and the immunoreactive bands were visualized with bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium. The phosphorylation of Rho* samples was carried out in 50 mM Tris-HCl (pH 8.0), 5 mM magnesium acetate, 20 mM potassium fluoride, and 50 M [␥-32 P]ATP (specific activity Ϸ4,500 cpm/pmol), in the presence of a 50-l aliquot of an enriched fraction of rhodopsin kinase. Following incubation for 1 h at room temperature, the kinase reactions were terminated with sample buffer for SDS-PAGE (27) and completely loaded on 12% polyacrylamide slab gels. The resulting 32 P-labeled phosphopolypeptides were separated by electrophoresis and electrotransferred to polyvinylidene difluoride membranes (28). The membranes were exposed to Kodak X-Omat x-ray films for 24 h, at Ϫ80°C, using intensifying screens, and the phosphorylated bands were qualitatively analyzed by autoradiography. The concentration of DM was estimated by the anthrone method (29). Briefly, a 200-l aliquot of the appropriate DM solution was mixed with 800 l of the anthrone reagent (2 g/liter in 17 M H 2 SO 4 ) and heated for 15 min at 100°C. After cooling, the mixture was diluted with 10 ml of 17 M H 2 SO 4 , and the absorbance was measured at 620 nm. A standard linear curve for the relation of A 620 nm against the concentration of DM was determined in the range from 10 to 100 g of DM. Because Rho is a glycoprotein containing two sites of oligosaccharide attachment, at Asn 2 and Asn 15 (30), and these two sites were found to contain predominantly the uniquely small GlcNAc 3 Man 3 , with smaller amounts of chains containing Man 4 and Man 5 (31, 32), we prepared control samples containing an excess of 20 mol of ␣-methylmannoside/mol of Rho used in the determinations. Under our conditions, this amount of carbohydrate contributed less than 1% to the absorbance at 620 nm and did not influence DM estimations. N-Acetylglucosamine was not included in the controls because it has been reported that hexosamines do not give any color in this reaction (33). Because it is known that tryptophan can influence the absorption yield at 620 nm by reaction with anthrone (33), we have added an amount of bovine serum albumin known to contain five times more tryptophan than Rho. Bovine serum albumin also contributed less than 1% to the absorbance at 620 nm. The partial specific volume of Rho was calculated from its amino acid and carbohydrate composition using the known partial specific volumes of the amino acids (34) and the corresponding carbohydrates (35).

Covalent Cross-linking of Rho and Rho*-Washed
ROS membranes were incubated with sulfo-SMCC and MBS, two bifunctional reagents capable of forming bridges between Cys and Lys spatially located about 11.6 and 9.9 Å apart, respectively (36 -38). As expected, the migration of Rho in SDS-PAGE was not modified in the control sample (Fig. 1A, top). On the other hand, treatments of washed ROS membranes with 5 mM sulfo-SMCC or MBS, at room temperature, resulted in a decrease of the ϳ35-kDa species corresponding to the monomer of Rho (R1), with a concomitant covalent formation of dimers (R2), trimers (R3), and oligomeric forms (Rn) (Fig. 1A, top). The Rn species were not able to enter the staking gel and probably consisted of several multimeric arrays of Rho. As also seen in Fig. 1A (top), the formation of Rho cross-linking products with either sulfo-SMCC or MBS was incomplete, and a clear prevalence of cross-linked dimers was obtained.
The application of chemical cross-linking to the membrane systems is, in general, complicated by the fact that membranebound proteins exist at sufficiently high densities that accidental collisional cross-linking between monomeric proteins or between oligomeric complexes cannot be ruled out. Therefore, covalent cross-links formed do not necessarily reflect the stable, naturally occurring association of proteins. Dissolution of the membrane significantly lowers the effective protein concentration, with a consequent diminution in the frequency of random collisions between protein molecules. Then Rho was extracted in DM from ROS and purified to homogeneity by affinity chromatography on concanavalin A-Sepharose. Fig. 1A (bottom) resumes the results of incubating purified Rho with 5 mM sulfo-SMCC or MBS. Similar to Rho in washed ROS membranes, the appearance of species migrating at twice the apparent molecular mass of the monomeric protein (R2, ϳ75 kDa), as well as other supramolecular arrangements (R3 and Rn), following treatment of purified Rho with these cross-linking agents was also observed. Yet, partial formation of R2, R3, and Rn species was also obtained when DM-solubilized Rho was incubated with either sulfo-SMCC or MBS, and again the majority of the resulting cross-linked products consisted of dimers (Fig. 1A, bottom).
o-PDM and p-PDM are specific homobifunctional agents that cross-link Cys residues located at 9 or 12 Å, respectively (36,39). Fig. 1B shows the reaction of purified Rho, in its dark (Ϫ) and photolyzed (ϩ) states, with increasing concentrations of o-PDM (2-8 mM). Similar to sulfo-SMCC and MBS, o-PDM was also capable of cross-linking Rho producing R2 cross-linked species in the dark. Most interestingly, a significant reduction of Rho monomers (R1) with a concomitant enhancement of dimeric (R2), trimeric (R3), and multimeric (Rn) species were obtained when Rho* was incubated with increasing concentrations of the cross-linking agent (Fig. 1B). An almost complete disappearance of the R1 species was attained when 8 mM o-PDM was employed (Fig. 1B). These cross-linking data corroborate the conformational changes caused in Rho upon illumination. Identical results were achieved when Rho and Rho* were incubated with p-PDM (data not shown).
In an attempt to further distinguish transient from stable interactions, time courses of the cross-linking of DM-solubilized Rho were carried out with the various cross-linking compounds. Study of the kinetics of the sulfo-SMCC, MBS, and o-PDM reactions followed by SDS-PAGE and stained by silver showed that the ϳ75-kDa species (R2) was formed in proportion to the disappearance of the ϳ35-kDa species corresponding to the monomer of Rho (R1) (Fig. 2A). Nevertheless, in all cases the reactions were not stoichiometric, and limited formation of the R2 species was obtained even after 1 h of incubation with the cross-linking reagents. These results may indicate a shortage of suitable target residues that are required to be located at the right distances in Rho. Monofunctional modification by only one end of the bifunctional reagent is an additional possibility because a high percentage of the target residues may end up with a defunct reagent after some have become modified. Although the formation of trimeric (R3) and multimeric (Rn) Rho species was not evident in the results shown in Fig. 2A, primarily due to the amount of protein loaded (1.5 g of Rho/lane), these higher molecular weight species clearly appeared when parallel experiments were analyzed by Western blotting and revealed with the anti-Rho monoclonal antibody 1D4 (data not included). Most interestingly, as the time courses of the various cross-linking reactions proceeded, the resulting Rho species were not efficiently colored by silver ( Fig. 2A), suggesting that the incorporation of these bifunctional compounds hindered the appropriate staining of the protein.
Slightly greater apparent molecular masses were observed for the ϳ35-kDa Rho species through the time course of the various reactions ( Figs. 1 and 2A), indicating either the formation of intramolecular cross-linked products within the Rho monomeric unit or an extensive monofunctional incorporation of the cross-linking compounds into the protein by only one end of the reagents. Alternatively, cross-links introduced by the various chemical agents used here may reduce binding of SDS and, consequently, decrease migration. The R2 region may also contain cross-linked dimers formed from mixed combinations between native Rho and/or intramolecularly cross-linked Rho. Fig. 2B, I, shows the cross-linked products obtained when increasing concentrations of Rho⅐DM (1.28 -20.5 M) were incubated with a fixed concentration of sulfo-SMCC (5 mM). An increase in the amount of the cross-linked R2 species was proportionally obtained when a 13-l aliquot of each reaction was loaded onto the gel. However, when aliquots of each sample containing the same amount of Rho (1.5 g) were electrophoresed, it was clearly observed that the different concentra-tions of Rho did not affect the ratios of R2 to R1 species obtained following incubation with the cross-linking compound (Fig. 2B, II). Because the same cross-linked products were formed even at concentrations of 20.5 M Rho, they probably reflect a stable, naturally occurring association between native Rho molecules rather than an accidental and transient collisional interaction.
Gel Filtration Chromatography in the Presence of DM-As illustrated in Fig. 3A, chromatography of Rho⅐DM on Sephacryl S-300 resulted in elution of a single sharp peak absorbing at 498.5 nm, which did not show detectable absorbance at 380 nm. On the contrary, chromatography of detergent-solubilized Rho* on the same column produced a single symmetric peak absorbing at 380 nm, with no perceptible absorbance at 498.5 nm (Fig.  3B). In both cases, single sharp peaks absorbing at 280 nm were obtained that overlapped with the corresponding absorbing peaks at 498.5 or 380 nm for Rho⅐DM and Rho*⅐DM, respectively (data not included). Moreover, the ratios of absorbance at 280 to 498.5 nm for Rho⅐DM and absorbance at 280 to 380 nm for Rho*⅐DM were constant throughout the peaks, further indicating the presence of a single species. Additionally, equivalent separations of Rho⅐DM and Rho*⅐DM were also performed in the presence of various protein standards, and the purified receptors were eluted in the same fractions as in Fig.  3. The figure also illustrates the elution peaks for the molecular weight markers employed (Fig. 3). A yield of ϳ85-90% of the initial amount of Rho⅐DM and Rho*⅐DM that was loaded onto the resin was recovered following gel filtration chromatography. After separating Rho in the dark or in the light, a reddish or yellowish layer, respectively, was always observed on the top of the resin. These results suggested the occurrence of a small proportion of oligomers of DM-solubilized Rho and Rho* that were permanently adsorbed by the resin, and that may account for the 10 -15% final loss of protein.
The sizes of the native Rho⅐detergent and illuminated Rho⅐detergent complexes were empirically determined to be 128,000 and 126,000 Da, respectively (Fig. 4A). Given that the molecular weight of DM micelles has been calculated to be about 50,000 (40), the molecular masses of native Rho and Rho* were calculated by subtracting this value from the total size of the protein-detergent complexes, yielding 78,000 and 76,000 Da, respectively. These results predicted that both conformations of Rho are dimeric.
A calibration curve of (Ϫlog K av ) 1/2 versus the Stokes radii of each standard was also obtained (Fig. 4B). The measured elution volumes yielded the (Ϫlog K av ) 1/2 for Rho⅐DM and Rho*⅐DM complexes, and Stokes radii of 4.18 and 4.15 nm, respectively, were determined by interpolation (Fig. 4B).
Estimation of the Partial Specific Volume of Rho⅐DM Complexes-The partial specific volume 2 is mainly dependent on the chemical composition of the protein-detergent complex and a good approximation of its value is given by Equation 2, where p is the partial specific volume of the protein; d is the partial specific volume of the detergent, and ␦ is the binding ratio of detergent to the protein. Accordingly, the p value was calculated to be 0.709 cm 3 /g for Rho.
The detergent concentration of the Rho⅐DM fraction after gel filtration on the Sephacryl S-300 column was measured by the anthrone method. The free detergent concentration was determined in a fraction eluted outside the Rho⅐detergent elution volume. The value for ␦ was calculated to be 0.97 g of detergent/g of protein.
Then we calculated the molecular mass of the protein moiety contained in the protein-detergent complex as 65,000 Da for Rho and 64,000 Da for Rho*, using the measured ␦. Combined with the molar mass of ϳ40,000 Da determined from the primary structure and carbohydrate composition of Rho, we estimated that the protein-detergent complex contained 1,63 copies of Rho/mol. A similar analysis estimated that the Rho*⅐DM complex contained 1.6 copies of Rho*/mol. These results were also consistent with native Rho and Rho* being dimeric oligomers. From the determined p and ␦ values, and using the d for DM of 0.824 cm 3 /g, which was previously reported by Møller and le Maire (41), a of 0.765 cm 3 /g was calculated for the Rho⅐detergent complex.
Sucrose Gradient Ultracentrifugation in the Presence of DM-The sizes of Rho⅐DM and Rho*⅐DM complexes were also evaluated by subjecting purified Rho and illuminated Rho to velocity sedimentation on a 10 -30% sucrose gradient prepared in a buffer containing 0.1% DM. Proteins with known sedimentation coefficients were included as standards, and the migration of Rho was identified by SDS-PAGE and Western blot analysis using the monoclonal antibody 1D4 (data not shown). About 10% of the original Rho⅐DM and Rho*⅐DM samples possessed a very high isopycnic point and sedimented at the bottom of the corresponding centrifuge tubes, indicating that some higher order oligomers of Rho were preserved even after detergent solubilization. However, the bulk of purified Rho and Rho* fractionated traveling to their respective buoyant densities on the sucrose gradient, at which point they ceased to move. The sedimentation coefficients for both Rho⅐DM and Rho*⅐DM complexes were determined to be 5.78 S (Fig. 4C).
For spherical molecules, the molecular mass of a species can be calculated from a combination of the measured Stokes radius and sedimentation coefficient using Equation 3 (21), where M is the molecular mass; a is the Stokes radius; s is the sedimentation coefficient, is the partial specific volume; is the viscosity of the medium; is the density of the medium, and N is Avogadro's number. By using this approximation, a calibration curve of M versus the Stokes radius multiplied by the sedimentation coefficient was prepared using the values reported for the protein markers, and a molecular mass of 107,000 Da was estimated for the photoreceptor protein-DM complexes, under dark and light conditions (data not included). Once subtracted the molecular mass of the detergent micelle, a size of about 57,000 -61,000 Da, was obtained for both conformations of Rho. Again, these results were consistent with a dimeric quaternary structure for the photoreceptor protein.

Determination of the Frictional Coefficient f/f 0 for Rho⅐DM and Rho*⅐DM Complexes-
The frictional coefficient f/f 0 can also be determined from the molecular mass and the Stokes radius as seen in Equation 4, When the molecular weights obtained by gel filtration were employed, the calculated frictional ratios for the Rho⅐DM complexes were 1.56 and 1.58 for the dark and illuminated states, respectively. However, frictional ratios of 1.4 for both native Rho⅐DM and illuminated Rho⅐DM complexes were found when the sizes determined by ultracentrifugation were used. Similar to most detergent-membrane protein complexes, which have frictional ratios in the range of 1.4 (42), the f/f 0 values attained here suggested some asymmetry in the Rho⅐DM and Rho*⅐DM complexes and indicated that their native-like conformations lie in the boundary between globular and moderately expanded, when compared with compact spheres (43,44).

Molecular Exclusion Chromatography and Sedimentation of Purified Rho in the Absence of DM-Purified
Rho was chromatographed on a Sephacryl S-300 column in the absence of DM to prevent the formation of detergent micelles. The peak of Rho eluted as a unique species with a molecular weight of 65,300 (data not included). Purified Rho was also ultracentrifuged on a 10 -30% sucrose gradient in the absence of DM. Following sedimentation, two species of Rho were found with sizes of 122,600 and 69,800 Da, respectively (data not shown). The high molecular weight fraction probably corresponded to some remaining Rho⅐DM complex, which was expected to persist as the final detergent concentration was slightly below its critical micelle concentration. Both the 69,800-and 65,300-Da species attained by gel filtration chromatography and sedimentation, respectively, must represent the native Rho dimer. Approximately 10% of the total protein persisted as Rho oligomers because it was adsorbed on top of the Sephacryl S-300 matrix and sedimented in the first fraction after isopycnic centrifugation. Fig. 5A, the detergent-solubilized Rho maintains its characteristic absorption spectrum after gel filtration (GF) or sedimentation (S). These samples of Rho have the ability of catalyzing the light-dependent GMPpNp binding activity of transducin (Fig. 5B), up to 60 -75% of the level induced by washed ROS membranes or concanavalin A-Sepharose affinity-purified Rho. In addition, both samples of Rho were capable of stimulating the GTPase activity of transducin under illumination (Fig. 5C). The ability of both Rho samples to serve as substrates for rhodopsin kinase was also evaluated. Fig. 5D shows that an enriched fraction of rhodopsin kinase was capable of phosphorylating these samples in a light-dependent manner. However, no Rho phosphorylation was attained when the reaction was carried out in the dark. All these results demonstrated that Rho conserved its native-like structural integrity and functional features following gel filtration chromatography and sedimentation on sucrose gradients. DISCUSSION ROS disk membranes contain densely packed Rho molecules for optimal light absorption and subsequent amplification by  (24,25). Experiments with concanavalin A-Sepharose affinity-purified Rho (Rho-DM (Con A)) and washed ROS membranes (W-ROS) were also included as controls. C, light-dependent stimulation of the T GTPase activity. The Rho⅐DM complex after gel filtration (Ⅺ) or sedimentation (ࡗ) was used to induce the light-dependent [␥-32 P]GTP hydrolytic activity of T. Assays were also performed in the dark as controls. D, autoradiography showing the light-induced in vitro phosphorylation of Rho⅐DM by rhodopsin kinase (RK). I, intact ROS membranes incubated with [␥-32 P]ATP under dark (Ϫ) or light (ϩ) conditions. II, enriched-fraction of RK incubated with [␥-32 P]ATP in the presence of light (ϩ); identical results were obtained in the dark (data not included). III, Rho⅐DM following gel filtration (GF) or sedimentation (S) was incubated with the enriched fraction of RK and [␥-32 P]ATP in the dark (Ϫ) or light (ϩ). The arrow indicates the migration of phosphorylated Rho*. the visual signaling cascade (45). Low angle x-ray diffraction studies have suggested that Rho is monomeric on the basis of the occurrence of particles 40 -50 Å in diameter in frog retinal receptor disk membranes that were immunologically identified as the photopigment molecules (4). The nature of the diffraction was not consistent with a planar crystalline lattice of the particles within the disk membranes but with a planar liquidlike arrangement of the particles (5). These measurements have been questioned (6) because information from x-ray diffraction studies is limited by the imperfect stacking of the membranes, the low contrasts of electron densities among the components of the lipid-protein-water structure, and by the difficulty of placing electron density profiles on an absolute scale. However, neutron diffraction studies of retinal ROS also suggested a random distribution and monomeric organization of Rho in the membrane (7). Fast, transient, flash-induced photodichroism showed the rapid rotational diffusion of Rho in situ (9), and the kinetics of flash-bleaching recovery indicated that Rho undergoes rapid lateral diffusion in intact rods (10). Yet, transient photodichroism cannot reliably distinguish between freely rotating dimers and monomers. Most interestingly, no detectable change was observed in the rotational diffusion of Rho upon illumination indicating that oligomers of Rho do not form during excitation (46). Electron microscope images of snap-frozen, freeze-etched frog rods (8) were also able to resolve Rho monomers in a random array, with no evidence of dimers. From all these early experiments a regular distribution of Rho in disk membranes would be expected. However, recent studies (47)(48)(49) have demonstrated the existence of detergent-resistant membrane microdomains or lipid rafts in ROS and therefore a nonuniform distribution of lipid and protein. In particular, Rho is found in raft and nonraft portions of the membrane, and its distribution does not change in the light or dark (47,48). In addition, virtually all the key components of the phototransduction cascade are either permanently associated with the ROS lipid rafts or translocate there in a light-dependent manner. Thus, alternative interpretations of the early biophysical results may be appropriate. Finally, atomic force microscopy experiments have revealed distinct rows of Rho dimers and paracrystalline arrays in native murine optic disk membranes (13,15). Two different types of Rho-containing domains were identified: 1) large uniform paracrystals and 2) rafts of smaller Rho paracrystals separated by lipid (14,15). Topographs recorded at higher magnification unveiled rows of Rho dimers forming the paracrystal in both domains 1 and 2, identifying the Rho dimers as the building blocks of the paracrystals (14,15). This supramolecular arrangement was also found for the apoprotein, opsin (15). Occasionally, single receptor monomers were detected on such topographs, but the presence of Rho monomers was relatively rare in these images. Dimerization and higher order organization of Rho were also observed when electron micrographs and atomic force microscopy topographs were measured on native disk membranes prepared at room temperature (14), indicating that the observed packing arrangement of Rho and opsin was not artificially induced by the segregation of protein and lipid at low temperatures. Also, freeze-fracture electron microscopy has revealed paracrystalline Rho arrays in Drosophila photoreceptive membranes (50) and in the plasma membrane of bovine ROS (51). Consequently, the sum of all these early and recent results has led to the emergence of an interesting controversy (52,53).

Functional Integrity of the Rho⅐Detergent Complex Following Molecular Exclusion Chromatography and Ultracentrifugation-As illustrated in
By analyzing the hydrodynamic properties of DM-solubilized bovine Rho and Rho*, we have elucidated here the native quaternary structures of Rho and Rho*. Our results are consistent with a dimeric structure for both conformations of the photoreceptor protein and agree with the results reported by Liang et al. (15) and Fotiadis et al. (13,14). Molecular exclusion chromatography demonstrated that Rho and Rho* have molecular weights of 78,000 and 76,000, respectively, which is approximately twice the size observed by SDS-PAGE under denaturing conditions. A Stokes radius of 4.18 and 4.15 nm for Rho and Rho* was also determined, which again indicated the dimeric structure of the photoreceptor protein. In addition, both conformations of Rho showed sedimentation coefficients of 5.78 S and frictional ratios of about 1.4 -1.6 were calculated for the Rho⅐DM and Rho*⅐DM complexes. By assuming a globular and compact shape for the protein-detergent complexes, a slightly lower molecular weight (ϳ60,000) was estimated for Rho and Rho*. However, most biological macromolecules are not spheres, and ellipsoids of revolution (prolate or oblate ellipsoids) are more realistic models than a sphere. The incorrect assumption that Rho is globular may account for the small discrepancy obtained when its size was determined by ultracentrifugation. Ellipsoids have larger frictional coefficients than equivalent spheres. Because the volume of a molecule is proportional to the molecular weight, it has been reported that the more a molecule deviates from a sphere, the larger its frictional coefficient will become. The f/f 0 value determined for the dark and light states of the Rho⅐DM complex corresponds to a macromolecule having either a prolate ellipsoid shape with an axial ratio of 10:1 or an oblate ellipsoid shape with an axial ratio of about 12:1 (54). In fact, the crystal structure of Rho (55)(56)(57) shows that the protein has an ellipsoidal shape. The dimensions of the ellipsoid are ϳ48 Å wide and ϳ35 Å thick in the plane of the membrane and ϳ75 Å perpendicular to the membrane. Analysis by electron microscopy of preparations containing crystalline arrays has also shown that Rho molecules have planar dimensions of about 28 ϫ 39 -40 Å and are ϳ63-64 Å in height (58,59). All these results are consistent with the asymmetric shape deduced here for the protein photoreceptor.
The pattern of cross-linking for the various preparations of Rho showed predominance of formation of cross-linked dimers with a progressively diminishing yield of cross-linked products from dimer to trimer and higher oligomers. This pattern could result from a native oligomeric assembly of Rho or alternatively by cross-linking between random collision complexes of monomeric Rho molecules. Given that Rho is embedded in the disk of the ROS membranes, the protein is free to rotate and diffuse in the plane of the membrane, allowing an estimated collision frequency between molecules of 10 5 and 10 6 /s (9, 10). However, it has been reported that at concentrations of protein below 10 M, no significant accidental intermolecular cross-linking between separate molecules of protein occurs when a solution of protein is mixed with a cross-linking agent (60). Instead, what is observed are the products that result from intra-oligomeric cross-linking among the fixed number of polypeptides of which the protein is composed. Additionally, solubilization of the membrane with detergents significantly lowers the effective protein concentration, with a concomitant reduction in the frequency of transient collisions between Rho molecules. Initially, all of our cross-linking experiments with Rho in washed ROS membranes or DM-solubilized Rho were carried out at protein concentrations of 1.28 M, hindering the probability of accidental collisions between Rho molecules. Moreover, when increasing concentrations of Rho (1.28 -20.5 M) were incubated with fixed concentrations of bifunctional reagents, the cross-linking compound did not alter the proportion of the resulting Rho cross-linked products. These facts strongly imply that the cross-linked products probably reflect a stable association between native Rho molecules rather than a random interaction and suggest a dimeric/oligomeric structure for the photoreceptor protein. Semi-empirical models for the packing arrangement of Rho molecules derived from atomic force microscopy topographical data (14,15) and the crystal structure (55,61) suggest that the intradimer interface comprises contacts between helices H4 and H5. Additionally, most of the interacting residues are located on the cytoplasmic loop between helices H3 and H4, and on the carboxyl-terminal region. Moreover, other interaction sites are also located within the membrane. Cross-linking occurring on any pair of residues located at these intradimeric interfaces will account for the formation of Rho dimeric cross-linked products.
Some cross-linked Rho trimers and higher order oligomers were also attained using the various bifunctional reagents. These results were a little surprising because if Rho forms specific dimers then the interface of one molecule should already be occupied hindering the formation of oligomeric crosslinked products. However, evidence from atomic force microscopy, supported by electron microscopy, revealed distinct rows of Rho dimers and paracrystalline arrays in native disk membranes (13,15). Contacts between dimers are created entirely by the intracellular loop between helices H5 and H6 from one monomer in a dimer with the loop between helices H1 and H2 and the carboxyl-terminal residues from the same monomer at the adjacent dimer. Contacts between rows of dimers are maintained through hydrophobic residues from helix H1 close to the extracellular side. Then formation of cross-linked Rho trimers and oligomers may be easily explained by these various interfaces between dimers and rows of dimers. Most interestingly, a small portion of Rho⅐DM and Rho*⅐DM was strongly adsorbed on the top of the Sephacryl S-300 gel filtration resin. Additionally, a minor fraction of the original Rho⅐DM and Rho*⅐DM samples (ϳ10%) sedimented at the bottom of the centrifuge tube during isopycnic ultracentrifugation on 10 -30% sucrose gradients. These results suggested the occurrence of some Rho oligomers even in the presence of DM, which probably account for the high molecular weight products obtained following cross-linking of Rho⅐DM and Rho*⅐DM. In the case of DMsolubilized Rho*, an enhancement of cross-linked dimeric, trimeric, and multimeric Rho species was apparent, which was consistent with the additional sulfhydryl reactivity previously reported for Rho* (62) and with the conformational changes produced in Rho upon illumination (63)(64)(65)(66).
The formation of Rho cross-linked dimers was always incomplete rather than stoichiometric. However, several factors influence the formation of a cross-linked product. These include the availability of the appropriate amino acid residues in the proteins, the chemical specificity of the bifunctional crosslinker, and the reaction conditions. Negative results in chemical cross-linking experiments do not conclusively demonstrate that two protein components are not close to each other. A paucity of cross-linked products may be the result of a lack either of spatial proximity or of appropriate reactive groups on the adjacent polypeptide chains. In addition, the reaction of each of the chemical ends of the bifunctional reagents with their target residues in the protein, in aqueous solution, is a competition between formation of the desired products and the possibility of hydrolysis of the reagents. For example, it has been reported that the two reactive maleimide rings of the bismaleimides, such as o-PDM and p-PDM, are hydrolyzed much more rapidly than the single maleimide ring of the monofunctional analogue N-ethylmaleimide (67). Because it renders the maleimide ring unreactive toward cysteine, this rapid hydrolysis can limit the extent of cross-linking of protein by the bismaleimide. In consequence, any of these factors either indi-vidually or in combination could generate an incomplete formation of Rho cross-linked products in the dimeric Rho unit.
The concept of oligomerization in the presence or absence of ligands is generally accepted for many GPCRs. This oligomerization has been reported to affect GPCR trafficking, signaling, and pharmacology. Based on certain key sequences, GPCRs can be grouped into several distinct families. Rho belongs to the class A or family 1 of GPCRs (also known as Rho-like GPCRs), and several of its members have been shown to homo-oligomerize (68). Rho does not seem to be an exception, as indicated by our results that provide strong evidence for its dimeric state. Furthermore, many pairs of family 1 GPCRs have been shown to form heteromers as well (68), exhibiting novel functional characteristics distinct from the individual homomeric receptors. When GPCRs are activated, the oligomers rearrange and cluster, and a novel mechanism by oligomer intercommunication assisted by components of the plasma membrane and by scaffolding proteins is possible (69).
A simple model of a 1:1 Rho-T interaction is not compatible with the size of the cytoplasmic surface of Rho, which is too small to anchor both T ␣ and T ␤␥ , and with the reported cooperativity for this interaction, which exhibits a Hill coefficient of ϳ2 (70). The packing of Rho molecules as dimers provides a platform that can easily accommodate both T functional units (71,72) and is consistent with the kinetic studies of the Rhocatalyzed guanine nucleotide exchange (70), as well as with binding studies between Rho and T (73,74), which demonstrated allosteric regulation of the interaction of T with Rho*. In addition, it has been speculated that one molecule of Rho in the dimer is needed for productive coupling with T, whereas the second one provides a partial scaffold to dock subunits of T (75). The application of the evolutionary trace method to 113 aligned G protein ␣ subunit sequences resulted in the identification of two functional sites (76). One large, well defined site was clearly identified with the binding of ␤␥-complexes, regulators of G protein signaling (RGS), and effector proteins like adenylyl cyclase. The other functional site, which extends from the raslike or GTPase domain onto the helical domain, had the correct size and electrostatic properties for GPCR dimer binding (76). These theoretical predictions can be extrapolated to T ␣ and are consistent with the dimeric quaternary structure reported here for Rho. Recently, Filipek et al. (75) also modeled how T docks onto oligomeric Rho and described structural details of this critical interface in the signal transduction process. Visual arrestin, another Rho-binding protein, has a bipartite structure of two structurally homologous seven-stranded ␤-sandwiches, forming two putative Rho binding grooves (77,78). The positive charge arrangement on the surface of the Rho dimer matches the negative charges on arrestin (15). Thus, one arrestin monomer is likely to also bind one Rho dimer. Finally, the crystal structure of G protein-coupled receptor kinase, GRK2, is also in structural agreement with the oligomeric structures of GPCRs (79).
Results from nondenaturing gel electrophoresis and analytical ultracentrifugation have also suggested the presence of oligomeric states of T and its subunits (80). T oligomers have been trapped by using bifunctional maleimides (81) and by 1-ethyl 3-(3-dimethylaminopropyl)carbodiimide-induced crosslinking, providing physical evidence for the existence of these oligomers under native conditions (82). Moreover, T ␣ has been reported to spontaneously form disulfide linkages in the absence of reducing agents (83), a condition that produced the total inactivation of the holoenzyme once reconstituted with native T ␤␥ (84). Additionally, oligomeric forms of T ␣ were the predominant species when highly specific photoactivated crosslinking reagents were employed (85). Compatible with these findings is the cooperativity reported for the interaction of Rho with T (72,73,86). Mixon et al. (87) have described the threedimensional structure of G i ␣ 1 , an ␣ subunit isotype of G i . They have shown that the ␣ subunits form extensive quaternary contacts with neighboring ␣ subunits in the crystal lattice. Specifically, the ␣ subunits are organized in a "head-to-tail" oligomer such that each subunit is related to the next by a 2-fold screw rotation that positions the NH 2 terminus from the ras-like or GTPase domain of one subunit into the ␣-helical domain of the adjacent subunit (87). In a membrane environment where the concentration of macromolecules is high, the kinetics of interactions between receptor and G protein is likely to be diffusion-limited. Evidently, the formation of Rho dimers and oligomers would overcome the limitation by allowing T molecules to interact with locally concentrated pools of Rho dimers. Moreover, the formation of multimeric complexes of T would also permit the interaction of Rho dimers with clusters of T ␣ subunits. Both phenomena will facilitate the amplification of the light response.