INTRODUCTION

Heterotrimeric GTP-binding regulatory proteins (G proteins) contain one each of subunits , , and  and transduce signals from activated plasma membrane receptors to intracellular second messenger-generating effector proteins (1-3). The  and  subunits exist as a tightly bound complex, which can only be dissociated under denaturing conditions and which functions as an entity throughout the signaling cycle. Both the G protein  subunit and the complex transmit signals to effector molecules (4-6). The recent determination of the three-dimensional structure of G protein heterotrimers (7, 8) and a free heterodimer (9) revealed that, in addition to a seven-bladed -propeller, the complex contains a two-stranded parallel -helical coiled-coil structure involving N-terminal segments of  and .

Coiled coils consist of two or more -helices that wind around each other to form a left-handed supercoil (10, 11). They form a key structural element in many proteins, imparting an extended rodlike structure to -fibrous proteins such as myosin, and comprising the dimerization domain in certain regulatory proteins such as the bZIP transcriptional factors c-Fos and c-Jun (12, 13) as well as GCN4 in yeast (14). Proteins that form coiled coils contain a repeating pattern of seven amino acids (a heptad repeat, with residues within a heptad designated abcdefg), in which the first (a) and fourth (d) position of the heptad are occupied by hydrophobic residues, and other positions are occupied by polar residues, resulting in a hydrophobic surface along one face of each -helix that is buried within the core of the assembled complex. The complex may be further stabilized by interhelical salt bridges flanking the hydrophobic interface (Fig. 1, A and B) (10, 15).

G protein  heptad repeats and the G coiled-coil region.

The presence of a coiled coil in the G protein complex had been anticipated, based on analysis of primary sequence data by computer algorithm (16) and on the results of computer modeling and site-directed mutagenesis (17, 18). Residues of the yeast G homolog STE4p in a region homologous to the coiled-coil region of G₁ have been implicated in effector signaling (Fig. 1A) (19, 20). To better understand the role of the coiled coil in dimerization and effector signaling, systematic mutagenesis of the coiled-coil region of ₁ was performed and the properties of point mutants examined in transient transfection assays.

EXPERIMENTAL PROCEDURES

Constructs encoding wild-type ₁, ₁, ₂, and nonisoprenylated G mutants ₁-C71S (₁*) and ₂-C68S (₂*) in the vector pCDM8.1 (21) were described previously (22-24). Point mutants of ₁ in pCDM8.1 were generated by the polymerase chain reaction (25) by overlap extension using either Pfu or Pwo thermostable DNA polymerases. Sequences of the mutagenic primers employed are available upon request. The cDNA for human phospholipase C-₂ (26) (in pMT2) was a gift from Dr. S. G. Rhee (National Institutes of Health, Bethesda, MD). The expression construct for hemagglutinin epitope-tagged (HA)¹-JNK in pcDNA3 was a gift from Drs. O. Coso and S. Gutkind (National Institutes of Health, Bethesda, MD) (27).

The finished G₁ point mutant constructs all contained the sequence GAATTCAAGATG at their 5 ends (starting methionine codon underlined), were followed after the stop codon by an XbaI site at their 3 end, and were ligated between EcoRI and XbaI sites of pCDM8.1. Constructs in pCDM8.1 were amplified in Escherichia coli MC1061/P3 (Invitrogen). The resulting plasmid preparations were purified by column chromatography (Qiagen Maxiprep kits). The DNA sequence of all inserts was verified by the chain termination method (28) using Sequenase 2.0 (U. S. Biochemical Corp.).

Growth, maintenance, transfection (29), and fractionation of COS-7 cells were as described previously (22). Protein was determined by the method of Bradford (30) using bovine serum albumin as a standard. Crude particulate or cytosolic fractions, or detergent lysates of whole cells, were separated on 11% slab gels by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (31), and electrotransferred onto polyvinylidene difluoride membranes in Dunn's buffer (32). For analysis of G subunit expression, the Tricine gel system of Schägger and von Jagow (33) was employed as indicated. Detection of G₁ subunits employed primary antibodies generated in rabbits, antibody SW against a peptide corresponding to residues 330-340 at the C terminus of ₁, or antibody RA against an internal sequence corresponding to residues 256-265 as described previously (34). Detection of G subunits employed the primary antibodies LVKG and EDPL, generated in rabbits against synthetic peptides corresponding to residues 53-70 near the C terminus of ₁ and residues 47-64 near the C terminus of ₂, respectively, as described previously (24). Detection of HA epitope-tagged JNK in COS cell lysates utilized mouse monoclonal HA.11 (Berkeley Antibody Co.). Secondary detection utilized ¹²⁵I-Protein A for rabbit polyclonal antibodies and ¹²⁵I-labeled goat anti-mouse IgG for mouse monoclonal antibodies, followed by autoradiography on film or a storage phosphor screen (Molecular Dynamics PhosphorImager). Quantitation of band intensity and digital imaging on storage phosphor screens employed ImageQuant software (version 1.11, Molecular Dynamics).

The PI-PLC activity of transfected cells was estimated by a modification of the procedure of Berridge et al. (35) as described previously (36, 37).

The assays for JNK activity were essentially as described by Coso et al. (27). Approximately 2.5 × 10⁶ COS-7 cells were plated into 75-cm² flasks and incubated at 37 °C overnight. On the following day, the cells were transfected by the DEAE-dextran method (29), using a total of 15 µg of DNA/cotransfection, typically including 5 µg of HA-JNK (27), 5 µg of G, and 5 µg of G or point mutant. Vector DNA was added where necessary to keep the total amount of plasmid DNA per flask constant. The remainder of the assay was as described (27), except that mouse monoclonal HA.11 was used for the immunoprecipitations (3 µl of HA.11 ascites fluid/900 µl of detergent lysate). GST-c-Jun-(1-79) substrate was from Stratagene.

The schematic graphical representation of the G and G coiled-coil regions employed the HelicalWheel program of the University of Wisconsin Genetics Computer Group (GCG) (38) with a specification of 102° rotation/residue (3.5 residues/turn). The Pileup algorithm was also from GCG (38). Analysis of primary protein sequences for regions of predicted coiled coil was performed online, and employed both the COILS algorithm of Lupas et al. (16)² and the PAIRCOILS algorithm of Berger and Kim (39).³

RESULTS

Conservation of the Coiled Coil in G Complexes

X-ray crystallographic determination of the structure of two mammalian G protein heterotrimers (an _i1·₁·₂ complex (7) and a complex between an _t-_i1 chimera and ₁·₁ (8)) and free G₁₁ complex (9) reveals the presence of a two-stranded parallel coiled coil involving N-terminal portions of G and G. Analysis of the protein sequences encoded by all vertebrate, invertebrate and plant G cDNAs identified to date employing the COILS (16) or PAIRCOILS (39) computer algorithms predicts an N-terminal coiled coil, with the exception of the Schizosaccharomyces pombe G homolog Gpb1 (40) (Fig. 1A and data not shown). This indicates the G coiled coil is a highly conserved structure, which is likely to be of fundamental importance to the function of heterodimers.

The Role of Glu-10 in G Assembly

Previous computer-assisted modeling of the ₁₁ coiled coil (18) predicted an electrostatic interhelical interaction between residue Glu-10 of G₁ and Lys-23 of ₁, an interaction that was subsequently confirmed in the crystal structure (9). An acidic residue in this position in G subunits and a basic residue in G corresponding to Lys-23 in ₁ are highly conserved among phylogenetically diverse species. Site-directed mutagenesis of the G₁ cDNA to produce the ₁-E10K mutant resulted in a construct that failed to assemble with ₂* in a cotransfection assay, suggesting this electrostatic interaction was critical for heterodimer formation (18). To further explore the effect of mutations at this position, point mutants ₁-E10R and E10A were constructed. Like Glu and Lys, Arg and Ala residues at this position would be expected to support the -helical secondary structure at this site known from the crystal structure (7, 9). For comparison, a ₁-E12K mutant was generated and its properties studied in parallel. Unlike the Glu at position 10, which occupies the g position of the heptad repeat, Glu-12 maps to a heptad b position (Fig. 1, A and B).

To assess the ability of the mutants to assemble with G, a cotransfection assay in COS cells employing the nonisoprenylated ₁-C71S mutant, ₁*, was used (Fig. 2). As shown previously (18, 22-24), cotransfection of wild-type ₁ with ₁* produced a large increase in cytosolic  and  immunoreactivity above that seen with transfection of either construct alone (Fig. 2, cf. lanes 2 and 3 with lane 4). This results both from the failure of complexes containing nonisoprenylated  subunits to undergo membrane targeting (22, 41) and from the mutual stabilizing effect of heterodimerization on G and G (42-44). This assay can be thus used to assess the ability of G and G mutants and chimeras to interact (18, 23, 24). As seen in Fig. 2, although the ₁-E12K mutant demonstrates -dependent expression nearly to the level of wild-type ₁ and comparably promotes ₁* expression in the soluble fraction (Fig. 2, cf. lanes 4 and 12; see also Figs. 3 and 4), neither ₁-E10K nor ₁-E10R exhibits significant -dependent expression or enhances the steady-state expression of ₁* (Fig. 2, lanes 5-8). The ₁-E10A mutant showed an intermediate level of expression, suggesting neutralization of negative charge was less deleterious to assembly than charge reversal at this position (Fig. 2, lanes 9 and 10). As can be seen in Fig. 2, analysis of the particulate fractions in these experiments is much less informative due to the presence of endogenous G immunoreactivity and the predominant localization of G* complexes to the cytosolic fraction (22). Such analyses were therefore not included in evaluation of the G* compatibility of the G coiled-coil point mutants described below.

Expression of wild-type G or G Glu-10 and Glu-12 point mutants and G₁* in crude particulate and soluble fractions of cotransfected COS cells.

Expression of wild-type G or G coiled-coil point mutants and G₁* in the soluble fraction of cotransfected COS cells.

Summary of relative expression levels of G coiled-coil point mutants and G₁* in the soluble fraction of cotransfected COS cells.

Scanning Mutagenesis of the Coiled-coil Region of G₁

To more thoroughly evaluate the role of individual residues in the coiled-coil region of G in the interaction with G, scanning mutagenesis was performed on 16 consecutive residues from the N-terminal -helix of G₁ forming the coiled coil with  (7, 9) (Gln-6 through Ala-21; see Fig. 1C). Neutral and basic residues were mutated to Glu, whereas acidic residues were mutated to Lys, both amino acids compatible with -helices. It was felt that these charged substituents might be potentially more disruptive to the G coiled coil than Ala, which is widely represented among all seven positions of repeats in known coiled coils (16, 45).

The ability of the 16 point mutants to assemble with ₁* was assessed in COS cells by the cotransfection assay. Blots of the soluble fractions of COS cells cotransfected with wild-type or mutant G cDNAs and ₁* were probed with G or ₁-directed antibodies, and further developed with ¹²⁵I-Protein A (Fig. 3). This allowed quantitation of the immunoreactive bands and estimation of the -dependent G expression levels and the -dependent increment in  expression for the point mutants relative to wild-type G₁ (Fig. 4). The strong correlation between the steady-state  and  expression associated with any particular G mutant-₁* combination in this cotransfection assay is readily seen from the data in Figs. 3 and 4. This likely reflects the much greater stability of the two subunits when complexed in heterodimeric form (42-44).

In addition to the ₁-E10K mutant, three other G₁ point mutations were found to largely abrogate interaction with ₁* in this assay: A11E, L14E, and I18E (Figs. 3, 4). This trio of G₁ point mutants also interacted poorly with the ₂* isoform in parallel experiments (data not shown). These three residues map to heptad a and d positions, where Glu is uncommon in coiled coils (16, 45), tending to disrupt the interhelical hydrophobic interface. Indeed, in the G₁₁ crystal structure, all three residues are seen to form hydrophobic interactions with residues in ₁ (9). Interestingly, two other heptad d position mutants in this series, L7E and A21E, were found to be more compatible with  assembly. Among the mutations involving residues in the heptad e and g positions of the G₁ coil, the E10K (g) mutation had a profound effect on  assembly as noted previously (18), whereas the R8E (e) and K15E (e) mutants showed intermediate expression levels. The ₁-Q17E mutant (g) had a phenotype indistinguishable from wild-type G₁ in this assay (Fig. 3 and 4). Three other G₁ point mutants employed in functional studies described below (A26E, D27N, and A28E), from a short non-helical region immediately following the coiled coil (Fig. 1C) (9), demonstrated -dependent expression levels in the range of 70-100% of wild-type G₁ in the same assay (data not shown).

JNK Effector Signaling of -competent G Coiled-coil Mutants

Recently subunits were found to activate the MAPK/ERK (46-48) and JNK (49) pathways in vertebrates, signaling cascades with many homologies to the -driven pheromone response pathway in the yeast Saccharomyces cerevisiae (50). Genetic screening in yeast has identified several pheromone signaling-defective G(STE4) mutants with point mutations in the vicinity of the putative coiled-coil region (Fig. 1A) (19, 20). Because the yeast G coiled-coil region has been implicated in effector signaling, the ability of the -competent G₁ coiled-coil point mutants to activate the JNK pathway was evaluated in a cotransfection assay in COS cells involving an HA epitope-tagged JNK reporter (49). The -competent mutations from the G₁ N-terminal -helical segment included two corresponding to yeast residues mutated in dominant negative alleles. The ₁ D20K mutation maps to STE4 residue Lys-55, where a K55E mutation produces a signaling defect (19); and the G₁ A21E mutation involves the residue homologous to Ala-56 in STE4, where an A56P point mutation abrogates effector signaling (20) (Fig. 1A). In addition, three -competent mutants were assayed that involved residues from a short interhelical region (9) of the G chain immediately following the coiled coil: A26E, D27N, and A28E (Fig. 1A). The D27N mutation corresponds to the dominant-negative D62N STE4 mutation (19, 20) and Ala-28 in G₁ corresponds to Ala-63 in STE4, another site of point mutation in yeast causing defective effector signaling (20). The A26E mutation was made as well for control purposes.

Cotransfection with G₂ was used instead of G₁ for these assays, as the former supported higher levels of G-dependent JNK activity in pilot experiments. Previous work has shown that G₁ interacts readily with both G₁ and G₂ (51-53) and has excluded the coiled-coil region of G₁·₂ chimeras as a major site of discrimination between these two G isoforms (23, 51, 54).

All of the -competent G point mutants were able to activate the JNK signaling pathway in the COS cell cotransfection assay to at least 50% of the wild-type level (Fig. 5). This was in contrast to a negative control construct encoding wild-type G₅, recently found to lack MAPK/ERK and JNK effector signaling ability (37) (Fig. 5). The A21E, D27N, and A28E mutants were found to be JNK-competent (Fig. 5). Taken together with evidence of their G*-dependent expression in the cytosol (Figs. 3 and 4 and data not shown), the ability of these G₁ mutants to activate the JNK effector pathway provides evidence of their ability to form functional G complexes. An unexpected finding was that the D20K point mutant produced consistently greater JNK activation than wild-type G under these assay conditions (Fig. 5). This was true despite the slightly lower expression level of the D20K mutant relative to wild-type G₁ in the G* cotransfection assay (Figs. 3 and 4). This phenomenon was further explored in experiments described below.

c-Jun N-terminal kinase activity in COS cells cotransfected with wild-type or -competent G coiled-coil point mutants and wild-type G₂.

Comparison of the JNK and PLC Signaling Ability of the ₁-D20K Mutant

To ascertain whether the ₁-D20K mutant was enhanced in its ability to signal via more than one effector pathway, its JNK and PLC signaling ability was compared with wild-type G at different doses of transfected cDNA (Fig. 6). Whereas the JNK stimulation of the ₁-D20K mutant was greater than that of wild-type over a range of doses of transfected DNA (Fig. 6A, left panel), its PLC-stimulating ability was at a level near or below that of the wild-type under two different PLC assay conditions (Fig. 6B). The relative PLC-stimulating ability of the ₁-D20K mutant compared with wild-type paralleled their relative G-dependent expression levels (Fig. 6A, right panel; see also Figs. 3 and 4). The enhanced JNK-stimulatory activity of the ₁-D20K mutant relative to wild-type G₁ was not due to increased expression of cotransfected HA-JNK reporter or G₂ (Fig. 6A, right panel). This finding implied that the enhanced signaling of the D20K mutant was pathway-selective and that the G coiled-coil region was involved in JNK signaling.

Comparison of wild-type and G-D20K c-Jun N-terminal kinase and phospholipase C- stimulatory activity in COS cells cotransfected with G₂.

DISCUSSION

Determination of the three-dimensional crystal structure of G protein heterotrimers (7, 8) and free G₁₁ complex (9) revealed that the heterodimer consists of two distinct structural domains: a seven-bladed -propeller encompassing the GH-WD repeats of G (55) and a coiled-coil involving N-terminal -helical segments of both G and G (16-18). Although some other GH-WD repeat-containing proteins may also contain N-terminal coiled-coils (e.g. the LIS-1 gene product (56, 57), residues 49-80 (data not shown and Refs. 16 and 39)), many GH-WD repeat and other -propeller proteins do not (55, 58), including, apparently, the fission yeast G homolog (40) (Fig. 1A). Thus, the presence of an N-terminal coiled coil in many G heterodimers may impart a functionality not possible with the -propeller core alone.

This study was undertaken, therefore, to better understand the protein-protein interactions of the G coiled coil by the use of point mutations in this region. The ₁*-compatibility of the series of G₁ Glu-10 point mutants characterized in the present work is compatible with the G₁₁ crystal structure demonstrating a salt bridge between this Glu and Lys-23 in ₁ (9); the mutations inducing charge reversal, E10K and E10R, were tolerated more poorly than the E10A substitution. It is nevertheless of interest that charge reversal at this heptad g position was more deleterious than at position e, where R8E and K15E mutations were compatible with ₁* association and effector signaling function, even though G Lys-15 is known to interact electrostatically with ₁ Glu-18 (9). It may be that these latter residues are more important in orienting the G coil during assembly or ensuring the specificity of pairing than contributing to interhelical stability (59).

Three G mutations within the hydrophobic core were found to block * interaction in the COS cell assay: A11E, L14E, and I18E. This does not exclude the possibility that other G mutations might also prevent pairing, nor should one infer that any mutation in these codons would be equally disruptive. Similar core heptad a and d mutations in yeast G homolog STE4p, including L49E and I53E, which correspond to G₁ mutants L14E and I18E, were found to partially inhibit interaction with yeast G STE18p in two-hybrid assays (Fig. 1A).⁴ These results suggest that the coiled-coil interactions between G and G make a critical contribution to heterodimerization across a range of structurally diverse G complexes.

Mutations in heptad repeat segments that impair coiled-coil interactions of structural proteins such as keratin and spectrin have been linked to inherited human diseases (60-62). As shown in this study, assembly of the heterodimer is also vulnerable to point mutations in the coiled-coil domain. Such coiled-coil mutations may presage identification of similar loss-of-function G mutations in humans. Such G loss-of-function might impart a disease phenotype at developmental stages and/or in specific cells where redundancy among G isoforms is minimal (e.g. in retinal rod or cone cells where a single G isoform predominates; Ref. 5).

Given the homologies between multiple components of the yeast S. cerevisiae pheromone-response pathway and mammalian MAPK and JNK pathways (63), it is perhaps surprising that no loss-of-function mutations of G₁ were identified among the -competent coiled-coil mutants in the JNK screening assay. Among the G₁ mutants were several involving residues (Asp-20, Ala-21, Asp-27, and Ala-28) homologous to yeast STE4 codons where loss-of-function mutations have been identified in genetic screens (19, 20) (Fig. 1A). It is possible that the yeast genetic screening methods are more sensitive to mutations than the overexpression paradigm used for JNK assay, and that an alternative assay method might reveal loss-of-function among the pool of G₁ mutants. Inasmuch as the molecular components of neither the yeast pheromone-response nor the mammalian G protein JNK pathways are fully resolved, it may also be that mechanistic differences account for failure to uncover JNK-signaling defective G₁ mutants in this study.

Facilitation of JNK signaling by the ₁-D20K mutant implicates the coiled-coil region of the complex in the mechanism of JNK activation. In the G₁₁ crystal structure G Asp-20, in a heptad c position (Fig. 1, A and B), is seen to project from the face of the coiled coil opposite the -propeller, where it might be free to interact with other proteins (9). It is interesting to note that the residue corresponding to G₁ Asp-20 is Lys-55 in STE4, and that its alteration to an acidic Glu is sufficient to produce a signaling-defective dominant-negative (19) (Fig. 1A). Thus, there may be some evolutionary or ontogenetic advantage in restraining the JNK signaling potential of mammalian Gs at this locus. The apparent lack of effect of the ₁-D20K mutation on PLC- signaling adds to previous evidence that the structural features of the G complex important for effector signaling may differ among effectors (37). Although it remains to be seen if the D20K mutation will alter any other -responsive effector pathways, it is possible that this or a similar mutation in G might result in selective gain-of-function with an associated clinical phenotype.