Gβγ Affinity for Bovine Rhodopsin Is Determined by the Carboxyl-terminal Sequences of the γ Subunit*

Two native βγ dimers, β1γ1 and β1γ2, display very different affinities for receptors. Since these γ subunits differ in both primary structure and isoprenoid modification, we examined the relative contributions of each to Gβγ interaction with receptors. We constructed baculoviruses encoding γ1 and γ2 subunits with altered CAAX (where A is an aliphatic amino acid) motifs to direct alternate or no prenylation of the γ chains and a set of γ1 and γ2 chimeras with the γ2 CAAX motif at the carboxyl terminus. All the γ constructs coexpressed with β1 in Sf9 cells yielded β1γ dimers, which were purified to near homogeneity, and their affinities for receptors and Gα were quantitatively determined. Whereas alteration of the isoprenoid of γ1 from farnesyl to geranylgeranyl and of γ2 from geranylgeranyl to farnesyl had no impact on the affinities of β1γ dimers for Gαt, the non-prenylated β1γ2 dimer had significantly diminished affinity. Altered prenylation resulted in a <2-fold decrease in affinity of the β1γ2dimer for rhodopsin and a <3-fold change for the β1γ1 dimer. In each case with identical isoprenylation, the β1γ2 dimer displayed significantly greater affinity for rhodopsin compared with the β1γ1 dimer. Furthermore, dimers containing chimeric Gγ chains with identical geranylgeranyl modification displayed rhodopsin affinities largely determined by the carboxyl-terminal one-third of the protein. These results indicate that isoprenoid modification of the Gγ subunit is essential for binding to both Gα and receptors. The isoprenoid type influences the binding affinity for receptors, but not for Gα. Finally, the primary structure of the Gγ subunit provides a major contribution to receptor binding of Gβγ, with the carboxyl-terminal sequence conferring receptor selectivity.

Heterotrimeric G proteins 1 transduce a wide variety of extracellular signals recognized by seven-transmembrane receptors, initiating signaling through a diverse array of intracellu-lar effectors (1). G proteins are composed of three polypeptides: a GTP-binding ␣ subunit and a dimer of ␤ and ␥ subunits that functions as a monomer. Ligand-activated G protein-coupled receptors catalyze the exchange of GTP for GDP tightly bound to the inactive G␣ subunit, resulting in dissociation of the GTP-activated ␣ subunit from both its cognate G␤␥ dimer and the receptor. The GTP-activated ␣ subunit and the dissociated G␤␥ dimer in turn regulate intracellular effectors (1,2). At least 20 different ␣ subunits, 5 ␤ subunits, and 12 ␥ subunits (3) have been identified to date. Such a diversity of structures is believed to contribute importantly to the specificity of signaling through these pathways.
It is well established that the selective interaction between receptors and the G␣ subunits provides a major determinant of signaling specificity. There are numerous examples of the selectivity of receptors for G␣ subunits. For example, the ␤-adrenergic receptor couples primarily to members of the G␣ s family (4 -6), whereas the ␣ 2A -adrenergic receptor couples to members of the G␣ i family (7)(8)(9)(10). G protein-coupled receptors can even display remarkable selectivity among closely related G␣ structures. The bombesin receptor subtypes selectively couple with different subtypes of G␣ q (11). A growing body of evidence also points to the contribution of ␤␥ subunits in determining receptor-G protein coupling selectivity. The role of ␤␥ diversity in the specificity of G protein signaling is supported by both in vivo and in vitro studies. Antisense RNA constructs for G␣, G␤, and G␥ selectively disrupt receptor signaling in rat pituitary GH 3 cells (12,13). These studies have demonstrated a specific requirement of ␤ 3 and ␥ 4 subunits for muscarinic receptor signaling, whereas ␤ 1 and ␥ 3 subunits mediate somatostatin receptor signaling. Also, ribozyme-mediated suppression of ␥ 7 subunit expression has led to a specific attenuation of ␤-adrenergic receptor signaling in HEK293 cells (14). Moreover, in vitro reconstitution with dimers of differing ␤␥ composition demonstrates that both the ␤ subunit (15)(16)(17) and the ␥ subunit (16,18,19) provide coupling specificity for various receptors. We initially interpreted the synergy and ␤␥ selectivity of rhodopsin-catalyzed GTP binding to G␣ t to mean that rhodopsin made independent binding contacts with the ␣ and ␤␥ subunits (20). Direct interaction of transducin ␤␥ with rhodopsin has been demonstrated separately with fluorescence spectroscopy (21) and with surface plasmon resonance measurements (22). A receptor contact site on the ␤ subunit was identified in crosslinking studies in which a synthetic peptide derived from the third intracellular loop of the ␣ 2A -adrenergic receptor could be cross-linked to the carboxyl-terminal region of the ␤ subunit (23).
A series of post-translational modifications are required for the functions of G proteins. In the case of G␥ subunits, these modifications include the thioether linkage of an isoprenoid * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  group to the conserved cysteine side chain within a carboxylterminal CAAX motif (where C is cysteine, A is an aliphatic amino acid, and X is any amino acid), proteolytic cleavage of the AAX tripeptide, and methylation of the resulting carboxyl terminus. The terminating residue in the CAAX motif appears to determine the identity of the isoprenoid. G protein subunits ␥ 1 , ␥ 8 (cone) and ␥ 11 with serine at this position are modified with a 15-carbon farnesyl, whereas the other ␥ subunits terminating in leucine are modified with a 20-carbon geranylgeranyl. The lipid modification of the ␥ subunits has been shown to be responsible for attaching G proteins to membranes (24,25). There is also ample evidence suggesting the involvement of the prenyl groups in protein-protein interactions (26 -29). Two ␤␥ dimers, ␤ 1 ␥ 1 and ␤ 1 ␥ 2 , display a Ͼ10-fold difference in their affinity for bovine rhodopsin (20,30), but no apparent difference in interaction with retinal G␣ t . Since these two ␥ subunits differ in both primary amino acid sequence and isoprenoid modification, we designed this study to examine the quantitative contributions of isoprenoid modification and protein structure to the ␤␥ dimer affinity for rhodopsin. Our data indicate that although the lipid modification is essential for a competent ␤␥ dimer, the type of prenyl group on the ␥ subunits influences the binding affinity for rhodopsin, but not for G␣ t . In addition, the primary structure of the ␥ subunit provides a major contribution to rhodopsin binding of ␤␥ dimers, with the carboxylterminal sequences conferring receptor selectivity.

EXPERIMENTAL PROCEDURES
Construction of ␥ Prenylation Mutants and Chimeras-Mutant and chimeric ␥ cDNA clones were made by PCR amplification. Bovine ␥ 1 and ␥ 2 were the original templates. A recombinant baculovirus encoding ␥ 1 -CVIL (31) was kindly provided by Dr. James Garrison (University of Virginia), and recombinant baculoviruses encoding ␥ 2 -CVIS and the cysteine-terminated construct "␥ 2 -A" were gifts from Dr. Nick Ryba (NIDCR, National Institutes of Health). Plasmids encoding chimeric ␥ 1 /␥ 2 structures as previously reported (32) were used as PCR templates for generating the current chimeric ␥ cDNA constructs, in which all the 3Ј-reverse primers contained nucleotides encoding CAIL prior to the stop codon. The resulting chimeric ␥ structures therefore all contain a carboxyl-terminal sequence that is known to direct geranylgeranylation. The PCR products encoding the chimeric ␥ structures were cloned into the transfer vector pBacPAK8 (CLONTECH) for production of recombinant baculoviruses. Insect cell culture, transfection, plaque purification, and virus amplification were carried out according to the manufacturer's recommended procedures (CLONTECH).
Purification of G Protein Subunits-G proteins were isolated from bovine retina and baculovirus-infected Sf9 cells expressing recombinant ␤ 1 ␥ dimers. Bovine retinal transducin was isolated from rod outer segment discs prepared by discontinuous gradient sedimentation (33). G␣ t and ␤ 1 ␥ 1 were purified using previously published procedures (20,34,35). Sf9 cell-expressed ␤ 1 ␥ dimers were purified as described for ␤ 1 ␥ 2 (30) with minor modifications. Sf9 cells were co-infected with ␤ 1 -encoding and the appropriate ␥-encoding baculoviruses at multiplicities of infection of 1 for the ␤ 1 virus and 3 for the ␥ virus. The cells were harvested 60 h after infection; the cells were lysed; and a P2 membrane fraction was isolated. All the dimers were purified from Sf9 cell membranes, except ␤ 1 ␥ 2 -A, which was purified from the cytosolic fraction. After sequential chromatography over DEAE-Sephacel (Amersham Pharmacia Biotech) and Ultrogel AcA44 (IBF), the fractions containing the ␤ 1 ␥ dimers were pooled and concentrated and then further purified on a Superdex HR-75 FPLC column (Amersham Pharmacia Biotech) with the final storage solution (10 mM MOPS, pH 7.5, 100 mM NaCl, and 8 mM CHAPS). Two of the samples, ␤ 1 ␥ 221 and ␤ 1 ␥ 2 -A, required an additional step of purification through a Mono-Q FPLC column (Amersham Pharmacia Biotech) before the final Superdex HR-75 FPLC sizeexclusion chromatography.
Electrophoretic Methods-Pre-cast Tricine-16% acrylamide gels (Novex) were run according to the manufacturer's protocol. Samples were electrophoresed at 125 V for ϳ90 min at room temperature. Protein bands were visualized by Coomassie Blue staining.
Mass Spectrometry-Liquid chromatography/mass spectrometry was performed by directly coupling a Shimadzu LC10VP HPLC system to the electrospray ion source of an LCQ ion trap mass spectrometer. Reversed-phase HPLC was performed at a flow rate of 10 l/min using a 0.3 ϫ 100-mm Betabasic 4 column (Keystone Scientific Inc., Bellefonte, PA). Solution A was H 2 O/acetonitrile (19:1) and 0.1% formic acid; Solution B was acetonitrile/1-propanol (4:1) and 0.1% formic acid. The column was equilibrated at 80% Solution A and 20% Solution B, and the chromatograph was developed using a linear gradient to 20% Solution A and 80% Solution B in 20 min. Full scan mass spectra over the m/z 300 -2000 range were acquired continuously for the duration of the chromatography.
␤␥ dimers were prepared for mass spectrometry by acetone precipitation. Samples (60 -201 pmol in 20 l) in 8 mM CHAPS were mixed with 9 volumes of ice-cold acetone in a Vortex mixer and incubated for 20 min on ice, and precipitates were collected by sedimentation at 12,000 ϫ g for 15 min at 4°C. After aspirating the supernatant liquid, pellets were allowed to air-dry and then were suspended in 25 l of Solution A, and 5 l of the sample was injected for liquid chromatography/mass spectrometry analysis.
Protein Concentration Determination-G␣ t concentration was determined by rhodopsin-catalyzed GTP␥S binding (20). G␤ 1 ␥ 1 was determined by the Amido Black binding assay (36) using bovine serum albumin as a standard. The protein concentrations of all other ␤ 1 ␥ dimers were determined from the staining intensity of the ␤ chain calibrated with different amounts of ␤ 1 ␥ 1 on a Coomassie Blue-stained gel by densitometry, with all the samples and ␤ 1 ␥ 1 standards within linear range.
Reconstitution Assays-The activities of the ␤ 1 ␥ dimers were quantified by rhodopsin-catalyzed GTP␥S binding to G␣ t and pertussis toxincatalyzed ADP-ribosylation of G␣ t as described previously (20). For both assays, the detergent CHAPS was adjusted to a final concentration of 1.6 mM. Urea-washed rod outer segment disc membranes, the source of rhodopsin, were prepared as described (37). Initial rate estimates for the catalyzed GTP␥S binding to G␣ t and pertussis toxin-catalyzed ADPribosylation of G␣ t were determined by single time point reactions, which consumed Ͻ20% of the G␣ t substrate.

RESULTS
The goal of this study was to delineate the relative contributions of isoprenoid modification and protein structure of the G␥ subunit to the interactions of the ␤␥ dimer with G␣ subunits and receptors. We have examined these questions using the retinal G protein and rhodopsin because affinity differences between ␤ 1 ␥ 1 and ␤ 1 ␥ 2 dimers are well characterized for this receptor, and these two dimers differ in both the primary structure and isoprenoid modification of the ␥ chain. Data from several laboratories have suggested that the isoprenoid modification provides the major determinant of differences between these two dimers, so we initially tested the effects of altering the prenylation. To examine this question, we tested three mutant ␥ constructs: the ␥ 1 mutant ␥ 1 -CVIL was expected to change ␥ 1 from farnesyl to geranylgeranyl; the ␥ 2 mutant ␥ 2 -CVIS was expected to change ␥ 2 from geranylgeranyl to farnesyl; and ␥ 2 -A, a truncated form of ␥ 2 that terminates at the cysteine residue of the CAAX motif, was expected to result in a mature protein lacking any prenyl group. Co-infection of Sf9 cells with viruses encoding these mutant ␥ constructs and a ␤ 1 -encoding virus resulted in the expression of a ␤ 1 ␥ dimer for all constructs. Fig. 1 shows the ␤ 1 ␥ dimers resulting from such co-infections purified to near homogeneity and separated on a Tricine gel stained with Coomassie Blue. The electrophoretic mobility differences among the ␥ chains are noticeable in this gel system. None of the samples contained any detectable G␣ contamination measured in the rhodopsin-catalyzed GTP␥S exchange assay or pertussis toxin-catalyzed ADP-ribosylation assay (data not shown).
To confirm the nature of the post-translational modifications, we analyzed purified ␤ 1 ␥ dimers using electrospray ionization mass spectrometry. Fig. 2 shows the deconvoluted mass spectra of ␥ 1 , ␥ 1 -CVIL, ␥ 2 , and ␥ 2 -CVIS. The observed molecular mass for each of the ␥ constructs is summarized in Table I. As predicted, the ␥ 1 subunit showed a molecular mass of 8332 Da, in agreement with the calculated molecular mass of the polypeptide chain deduced from the DNA sequence plus all known post-translational modifications. They include removal of the amino-terminal methionine, attachment of a farnesyl group to cysteine 71, removal of the carboxyl-terminal tripeptide, and methylation of the new carboxyl terminus. The ␥ 1 -CVIL sequence is predicted to direct geranylgeranyl modification. The observed mass of 8401 Da is consistent with exchange of the farnesyl group of wild-type ␥ 1 for a geranylgeranyl. Consistent with the findings of Lindorfer et al. (31), the mass spectra also showed a set of signals corresponding to the fully processed molecule minus the prenyl group. These authors have demonstrated that the loss of the farnesyl group occurs during the ionization process and is not due to heterogeneity of modification. Our data are consistent with their findings since the HPLC elution resolves ␥ 1 chains bearing differing isoprenoid modification prior to the ionization for the mass spectrometry. The ␥ 2 and ␥ 2 -CVIS subunits also showed the observed masses consistent with the amino acid composition deduced from the DNA sequence plus the predicted isoprenoid modifi-cations (farnesyl for ␥ 2 -CVIS and geranylgeranyl for ␥ 2 ), endoproteolysis, carboxyl methylation, the removal of the aminoterminal methionine, and amino-terminal acetylation. The ␥ 2 -A chain showed an observed mass of 7463 Da, consistent with that of the predicted amino acid sequence of ␥ 2 terminating at cysteine of the CAIL carboxyl terminus, lacking prenylation or methylation, with its amino-terminal methionine removed and the new amino terminus acetylated. Minor amounts of inappropriately prenylated species were observed for the ␥ 2 products (Fig. 2, C and D). The relative abundance of farnesylated versus geranylgeranylated ␥ subunits is summarized in Table I as well.
b Fractional isoprenoid modification was calculated from the integrated peaks of the farnesylated and gernylgeranylated species in the deconvoluted mass spectra. c ND, not determined.
dopsin decreased Ͻ2-fold (Fig. 3B), distinct from the lower affinity of ␤ 1 ␥ 1 for rhodopsin (K1 ⁄2 ϭ 31 versus 227 nM). However, although the alteration of ␥ 2 prenylation from geranylgeranyl to farnesyl diminished affinity for rhodopsin only modestly, the absence of a prenyl group diminished the interaction of the ␤ 1 ␥ 2 -A dimer to undetectable levels, with no enhancement of GTP␥S binding at all (Fig. 3A). These data suggest an absolute requirement for the isoprenoid modification of the ␥ subunit for the ␤␥ interaction with receptors.
To determine whether the differences in affinity observed among ␤␥ subunit forms were due to differing affinities for interaction with G␣ t , we analyzed the ␤␥ dependence of pertussis toxin-catalyzed ADP-ribosylation of G␣ t . Fig. 4 shows the saturation of ADP-ribosylation of G␣ t by increasing concentrations of different forms of ␤ 1 ␥ dimers. In contrast with the rhodopsin-catalyzed GTP␥S binding to G␣ t , the pertussis toxincatalyzed ADP-ribosylation of G␣ t displayed no distinction between ␤ 1 ␥ 1 and ␤ 1 ␥ 2 with either form of prenylation (Table II). The non-prenylated ␥ 2 dimer (␤ 1 ␥ 2 -A) displayed a measurable enhancement of the ADP-ribosylation of G␣ t , but with an affinity for G␣ t that was Ͼ30-fold lower than that for the prenylated ␤ 1 ␥ 1 or ␤ 1 ␥ 2 dimer. These data suggest that for the ␤␥ interaction with the G␣ subunit, the isoprenoid modification of the ␥ subunit, not the exact identity of the prenyl group, is important. Since the difference in prenyl group did not account for the majority of the affinity difference between ␤ 1 ␥ 1 and ␤ 1 ␥ 2 for rhodopsin, we investigated the contribution of primary structures. Fig. 5 depicts the set of chimeras we employed containing heterologous sequence at approximately one-third intervals based upon regions of sequence identity between the ␥ 1 and ␥ 2 proteins. These constructs were based upon the previously published chimeras (32), except that all of the chimeric constructs were altered to code for CAIL at the carboxyl terminus so that they would all have the same geranylgeranyl modification. Co-infection of Sf9 cells with baculoviruses encoding these chimeric ␥ chains and a ␤ 1 -encoding virus led to the expression of ␤ 1 ␥ dimers for all constructs. Because the yields of ␤ 1 ␥ 111 protein were quite low, we used the ␤ 1 ␥ 1 -CVIL product for this protein structure. ␤ 1 ␥ 222 is the wild-type ␤ 1 ␥ 2 construct. Fig. 6 shows the ␤ 1 ␥ chimeras along with the parent structure ␤ 1 ␥ 1 -CVIL and ␤ 1 ␥ 2 purified to near homogeneity and separated on FIG. 3. Saturation of rhodopsin-catalyzed GTP␥S binding to G␣ t by ␤ 1 ␥ subunits with altered isoprenylation. The indicated concentrations of ␤ 1 ␥ 1 (f), ␤ 1 ␥ 1 -CVIL (OE), and ␤ 1 ␥ 2 -A (q) in A and ␤ 1 ␥ 2 () and ␤ 1 ␥ 2 -CVIS (ࡗ) in B were incubated in reactions containing 500 nM G␣ t and 30 nM rhodopsin as described under "Experimental Procedures." The curves drawn are the best fit for a single site binding model using GraphPAD Prism.

TABLE II
Activity assessment of ␤ 1 ␥ dimers with altered isoprenylation a Tricine-16% acrylamide gel stained with Coomassie Blue. As found for the prenylation mutant ␥ chains, this gel system clearly resolves the chimeric ␥ chains. HPLC electrospray ionization mass spectrometry confirmed that the isoprenoid modification of the ␥ chains was predominantly the geranylgeranyl directed by the CAIL sequence. Geranylgeranyl-modified chain accounted for Ͼ74% of the mass of each purified ␥ subunit (Table III). Fig. 7 shows the saturation of rhodopsin-catalyzed GTP␥S binding to G␣ t for the purified ␤ 1 ␥ chimeras. They displayed clear differences in both affinity and maximum catalytic rate, as seen for the dimers containing the parent ␥ 1 or ␥ 2 protein. These differences could be sorted based upon the identity of the sequence for the carboxyl-terminal one-third of the chimera. As summarized in Table IV, ␤ 1 ␥ 122 and ␤ 1 ␥ 112 had similar high affinity for rhodopsin compared with ␤ 1 ␥ 2 (K1 ⁄2 ϭ 24, 23, and 18 nM, respectively). The other two chimeric dimers, ␤ 1 ␥ 221 and ␤ 1 ␥ 211 , had lower apparent affinity (K1 ⁄2 ϭ 39 and 68 nM, respectively), with the value for the latter close to the value for ␤ 1 ␥ 111 (i.e. ␤ 1 ␥ 1 -CVIL, 78 nM). This result suggests that the primary structure in the carboxyl-terminal region of the ␥ subunit determines the affinity of the interaction with rhodopsin. The amino acid sequence in the middle and amino-terminal regions of the ␥ 2 subunit also contributed to the high affinity interaction, although to a lesser degree.
As expected from the similar affinities of wild-type ␤ 1 ␥ 1 and ␤ 1 ␥ 2 , the ␤ 1 ␥ chimeras, ␤ 1 ␥ 1 -CVIL, and ␤ 1 ␥ 2 showed essentially identical affinity for G␣ t in the pertussis toxin-catalyzed ADPribosylation assay (Fig. 8 and Table IV). These data imply that the affinity differences found for dimers with distinct ␥ subunit primary structures are due to differences in rhodopsin-␤␥ interaction rather than differences in ␤␥-G␣ t interaction.

DISCUSSION
With the emerging evidence suggesting that in addition to the G␣ subunit, both the ␤ and ␥ subunits of G proteins also affect receptor-G protein interactions (15)(16)(17)(18)(19), it has become an increasingly accepted notion that ␤␥ dimers contribute to the selectivity of the receptor interaction. This study examined quantitatively the role of the prenyl modification and amino acid sequence of the G␥ subunit in determining signaling specificity. We confirmed that the apparent affinity for rhodopsin was Ͼ12 times higher for ␤ 1 ␥ 2 than for ␤ 1 ␥ 1 for the wild-type structures, whereas the two ␤ 1 ␥ dimers displayed no difference in affinity for retinal G␣ t . Although one study using rhodopsin to catalyze GDP/GTP exchange on the G␣ t subunit showed that the ␤ 1 ␥ 1 dimer supported a faster rate of exchange than did ␤ 1 ␥ 2 or ␤ 1 ␥ 3 (38), others have found a much lower apparent affinity for ␤ 1 ␥ 1 than for other ␤␥ dimers (20,27,30,39). The direct measurement of rhodopsin-G protein interactions using surface plasmon resonance also revealed a higher affinity of ␤ 1 ␥ 2 for rhodopsin compared with ␤ 1 ␥ 1 and showed that this difference was due to a lower dissociation rate for ␤ 1 ␥ 2 (22). We have capitalized upon this significant quantitative difference in the binding interaction with rhodopsin to assess the independent contributions of the ␥ subunit primary structure and posttranslational modification to this difference. Our results clearly implicate the primary structure of the ␥ chain as a major determinant of the affinity of ␤ 1 ␥ dimers for rhodopsin. The set of ␥ chimeras with uniform geranylgeranyl modification further pinpointed the carboxyl-terminal sequence of the ␥ subunit as the region responsible for receptor selectivity. Underscoring the importance of the amino acid sequence of the ␥ subunit in receptor-G protein interactions, a recent study showed that ␤ 1 ␥ 1 supported only a very low level of coupling to the ␣ 2A -adrenergic receptors, whereas the ␤ 1 ␥ 11 dimer produced a high level of coupling (16). Since both the ␥ 1 and ␥ 11 subtypes are farnesylated, the difference in coupling level observed for these dimers must also be attributed to differences in the amino acid sequences.
Consistent with previous findings on 5-hydroxytryptamine type 1A receptors (40), adenosine receptors (28), and bombesin receptors (11), we found that bovine rhodopsin prefers G␣␤␥ complexes containing the geranylgeranylated ␥ subunit. Our results obtained with prenylation mutants confirmed the findings of several laboratories that dimers containing geranylgeranylated ␥ subunits have higher affinities for receptors than those with farnesylated ␥ chains (11,27,28,40). In our study, given the same amino acid sequence of the ␥ subunit, the   FIG. 5. Construction of chimeric ␥ subunits. A, the amino acid sequences of the ␥ 1 and ␥ 2 subunits served as parent structures for the chimeric ␥ subunits as previously described (32). Numbers on the right indicate the amino acid residue positions. Before C-terminal processing, ␥ 1 contains 74 residues, and ␥ 2 contains 71 residues. In the current constructs, the C-terminal CAAX motif in ␥ 1 has been changed from its native CVIS to CAIL so that all the resulting chimeric ␥ subunits contain the same C-terminal CAIL as ␥ 2 that directs geranylgeranylation. Boldface residues (QLK and EDPL) correspond to sequence motifs used as junctions between chimeric segments. B, shown is a schematic overview of the ␥ 1 /␥ 2 chimeras. Open bars indicate polypeptide sequence derived from ␥ 1 , and closed bars indicate that derived from ␥ 2 . Vertical lines indicate junctions between chimeric segments. ␤ 1 ␥ dimer was more effective in coupling the G␣ t subunit to rhodopsin when the ␥ subunit was modified with geranylgeranyl compared with farnesyl. Furthermore, a prenylation-deficient construct was found to be incompetent to participate in these protein-protein interactions. Although this has suggested that the affinity differences found between retinal ␤ 1 ␥ 1 and other ␤␥ dimers could be attributed solely to the isoprenylation difference (27,28), we found that identically isoprenylated ␤ 1 ␥ 1 and ␤ 1 ␥ 2 retained significant affinity differences for bovine rhodopsin for both geranylgeranylated and farnesylated ␥ chains. Moreover, the effect of the prenyl type was more profound with the ␥ 1 primary structure than with the ␥ 2 structure, which implies the importance of the amino acid sequence of the ␥ 2 subunit in conferring a high affinity rhodopsin interaction.
The apparent affinity differences that we measured for bovine rhodopsin for this set of ␤ 1 ␥ dimers with varying ␥ compositions clearly do not originate from distinct G␣ t interaction since all of the dimers displayed essentially identical apparent affinity for G␣ t , as others have also noted for the retinal G protein subunits (20,41). This may not be the case for ␤␥ dimers interacting with other G␣ proteins. A significant difference in apparent affinity for G␣ s , but not for G␣ o or G␣ i , has been reported for ␤␥ dimers of differing compositions (42). Whereas ADP-ribosylation by pertussis toxin did not reveal a difference among brain ␤␥ dimers for G␣ o , inhibition of GDP release did (41). Since the primary contacts between G␣ and G␤␥ occur within the amino-terminal sequences of ␣ subunits, and these are distinct for the various G␣ species, it is likely that ␣ subunit selectivity may contribute importantly to differences among ␤␥ dimers in receptor signaling. This will be important to clarify in future studies.
The differences that we observed among the ␤␥ dimers in rhodopsin interaction are also not likely to originate from the different purities of the samples. We have previously shown that substances interfering with the quantitative assay of ␤␥ dimers in cholate extracts from Y1 adrenal cortical tumor cells were completely removed by chromatography over Ultrogel AcA44 (43). Similarly, we found that chromatography over DEAE-Sephacel was sufficient to produce partially purified recombinant ␤ 1 ␥ 2 dimers expressed in Sf9 cells using baculoviral vectors that showed quantitatively identical apparent rhodopsin affinity as homogenous preparations of the dimer. 2 All ␤␥ dimers utilized in the present study were purified by both DEAE-Sephacel and Ultrogel AcA44 chromatography with additional FPLC size-exclusion chromatography to assure hydrodynamic ␤␥ dimers with defined detergent concentration, and they were assessed to contain no measurable contaminating GTP-binding proteins or pertussis toxin substrates. We also note that the ␤ 1 ␥ 211 dimer, which had equal apparent purity compared with the ␤ 1 ␥ 2 dimer sample, displayed reduced apparent affinity for rhodopsin, but not for G␣ t . Furthermore, although the measured affinity of the ␤␥ dimers containing chimeric ␥ chains varied systematically with the carboxyl-terminal sequence of the ␥ chain, it did not vary systematically with sample purity, and the affinity of all samples for G␣ t was indistinguishable. We think it unlikely, then, that either specific contaminants or sample purity accounts for the differences in activities found in our recombinant ␤␥ dimers.
Similar to the previously reported results from Sf9 expression systems (27,31,44), we found heterogeneity in isoprenoid modification of several of the G␥ subunit constructs. There is no doubt that in the presence of a high concentration of virally encoded ␤␥ substrate, both farnesyl-and geranylgeranyltransferases in Sf9 cells misprenylate a fraction of the ␥ protein.
Whether such misprenylation is caused by excessive substrate, as suggested by Lindorfer et al. (31), or prenyltransferase specificity is also encoded by primary structures other than the CAAX motif, as suggested by Kalman et al. (44), remains to be tested. Neither our nor the previous data have directly assessed the prenyltransferase activities; rather, we have measured the long-term accumulation of recombinant protein products. Our data appear to suggest the lack of a consensus 2 D. E. Wildman and J. K. Northup, unpublished data.  7. Saturation of rhodopsin-catalyzed GTP␥S binding to G␣ t by ␤ 1 ␥ subunits with chimeric ␥ chains. The indicated concentrations of ␤ 1 ␥ 122 (f), ␤ 1 ␥ 112 (OE), ␤ 1 ␥ 221 (), and ␤ 1 ␥ 211 (q) were incubated in reactions containing 500 nM G␣ t and 30 nM rhodopsin as described under "Experimental Procedures." The curves drawn are the best fit for a single site binding model using GraphPAD Prism. sequence upstream of the CAAX motif that directs farnesylation or geranylgeranylation since the proportion of prenylation did not vary systematically in our ␥ 1 /␥ 2 chimeras with the presence of ␥ 1 or ␥ 2 sequence. In addition to the lipid modification, G␥ subunits are also methylated at the carboxyl-terminal cysteine after the proteolytic cleavage of the AAX tripeptide. The importance of the ␥ subunit carboxyl methylation to the ␤␥ interaction with G␣ subunits has been reported (45,46); its effect on ␤␥ interaction with receptors has yet to be fully examined. Whatever the effect is, it should not complicate the interpretation of our data for the following reasons. First, our mass spectra showed that except for retinal ␤ 1 ␥ 1 , which showed Ͼ50% methylation, all the Sf9-expressed ␤ 1 ␥ dimers were predominantly methylated (Ͼ90%). Second, we did not observe any difference among all the tested ␤ 1 ␥ dimers in their interaction with G␣ t , including retinal ␤ 1 ␥ 1 . Moreover, fully methylated ␤ 1 ␥ 1 purified from heterologously expressed Sf9 cells showed low receptor coupling with adenosine receptors (28,31), just as we have observed for retinal ␤ 1 ␥ 1 with rhodopsin.
The data presented in this study demonstrate that the prenyl group on the ␥ subunit is necessary for ␤␥ interaction with the ␣ subunit and rhodopsin. However, the type of prenyl group affects the binding only with rhodopsin, not with G␣. The primary structure of the ␥ subunit, particularly the carboxylterminal sequence, is of equal or greater importance for receptor-G protein interaction.  8. Saturation of pertussis toxin-catalyzed ADP-ribosylation of G␣ t by ␤ 1 ␥ subunits with chimeric ␥ chains. The indicated concentrations of ␤ 1 ␥ 111 (i.e. ␤ 1 ␥ 1 -CVIL; q), ␤ 1 ␥ 122 (f), ␤ 1 ␥ 211 (OE), ␤ 1 ␥ 112 (), ␤ 1 ␥ 221 (ࡗ), and ␤ 1 ␥ 222 (i.e. wild-type ␤ 1 ␥ 2 ; E) were incubated in reactions containing 500 nM G␣ t and 5 g/ml pertussis toxin as described under "Experimental Procedures." The curves drawn are the best fit for a single site binding model using GraphPAD Prism.