Functional Interaction between Gαz and Rap1GAP Suggests a Novel Form of Cellular Cross-talk*

Gz is a member of the Gi family of trimeric G proteins whose primary role in cell physiology is still unknown. In an ongoing effort to elucidate the cellular functions of Gz, the yeast two-hybrid system was employed to identify proteins that specifically interact with a mutationally activated form of Gαz. One of the molecules uncovered in this screen was Rap1GAP, a previously identified protein that specifically stimulates GTP hydrolytic activity of the monomeric G protein Rap1 and thus is believed to function as a down-regulator of Rap1 signaling. Like Gz, the precise role of Rap1 in cell physiology is poorly understood. Biochemical analysis using purified recombinant proteins revealed that the physical interaction between Gαz and Rap1GAP blocks the ability of RGSs (regulators of G protein signaling) to stimulate GTP hydrolysis of the α subunit, and also attenuates the ability of activated Gαz to inhibit adenylyl cyclase. Structure-function analyses indicate that the first 74 amino-terminal residues of Rap1GAP, a region distinct from the catalytic core domain responsible for the GAP activity toward Rap1, is required for this interaction. Co-precipitation assays revealed that Gαz, Rap1GAP, and Rap1 can form a stable complex. These data suggest that Rap1GAP acts as a signal integrator to somehow coordinate and/or integrate Gz signaling and Rap1 signaling in cells.

Trimeric guanine nucleotide binding regulatory proteins (G proteins) 1 are key players in the transduction of a diverse array of extracellular signals to intracellular second messages (1)(2)(3). These G proteins consist of a GTP-binding ␣ subunit and two additional subunits, termed ␤ and ␥, that are tightly associated with each other and function as a single ␤␥ complex (1,2). Based on structural homology between their ␣ subunits, G proteins are divided into four groups, G s , G i , G q , and G 12 . During G protein signaling, activation by an appropriately liganded receptor stimulates GTP binding to the ␣ subunit and subsequent dissociation of ␣-GTP from the ␤␥ complex to produce the active forms of the G proteins that transmit signals to downstream targets, termed effectors (3). In addition to receptors, G proteins, and effectors, a new family of proteins has recently been discovered that is important in signaling through trimeric G proteins. These proteins, termed RGSs (regulators of G protein signaling), act as negative regulators because of their abilities to stimulate GTP hydrolysis of G protein ␣ subunits (2); in addition, some RGS proteins can apparently modulate other aspects of G protein signaling processes (4,5).
G␣ z , a member of G␣ i family, is a 41-kDa ␣ subunit that has several unique, and presumably important, properties (6 -8). First, its tissue distribution is quite restricted, being found primarily in brain, adrenal medulla, and platelets, while expression is virtually undetectable in other tissues (9,10). Another property that distinguishes G␣ z from other members of G i family is its inability to serve as a substrate for pertussis toxin-catalyzed ADP-ribosylation (8,10), making it a candidate for pertussis toxin-insensitive signaling processes. Additionally, the intrinsic rate of GTP hydrolysis by G␣ z is quite low compared with most other G protein ␣ subunits (8), suggesting that RGS or RGS-like molecules play important roles in regulation of G z signaling. Finally, G␣ z is subjected to phosphorylation by protein kinase C; this phosphorylation occurs both in vitro and in vivo and the modification interferes with the interaction of G␣ z with both the ␤␥ complex and a recently identified selective regulator, RGSZ1 (11)(12)(13)(14).
While it is not yet clear which receptors are normally coupled to G z , many receptors that couple to G i proteins can also activate G z signaling pathways if the receptors are overexpressed in cells (15,16). Similarly, although the downstream effects of G z activation are not well understood, activated G␣ z does possess an ability to inhibit some subtypes of adenylyl cyclase, a property shared with other members of G i family (15,17). Also, stable expression of mutationally activated G␣ z can transform Swiss 3T3 and NIH3T3 cells by stimulating mitogenic pathways (18). Interestingly, this stimulation is apparently unrelated to the ability of G␣ z to inhibit adenylyl cyclase, suggesting that G z signaling is coupled to additional signaling pathways (18).
As part of an ongoing effort to elucidate the signaling functions of G z , we recently undertook a yeast two-hybrid screen to identify proteins that specifically interact with a mutationally activated form of G␣ z (14). In this report, we describe one of the molecules identified via this approach as Rap1GAP, a protein that had been previously identified as a specific activator of the GTP hydrolytic activity of the monomeric G protein Rap1 (19). The ubiquitous involvement of members of the Ras superfamily in mitogenic signaling pathways makes regulators of these proteins like Rap1GAP attractive candidates to act as signal integrators and/or coordinators with G␣ z . Interestingly, the expression pattern of Rap1GAP is similar to that mentioned above for G␣ z , with predominant expression being observed in brain among normal tissues examined (19). Characterization of the interaction between G␣ z and Rap1GAP indicates that the active form of G␣ z specifically and functionally interacts with Rap1GAP and that the three proteins, i.e. G␣ z , Rap1GAP, and Rap1, can form a stable complex. These data point to an unanticipated cellular cross-talk between G z and Rap1 signaling processes, and provide new insight into potential cellular functions of G z .
Plasmid Constructs-Plasmids containing the cDNA of rat G␣ z (both wild type and the Q205L mutant), containing the Glu-Glu epitope at residues 3 to 8, were the gift of Henry Bourne (University of California at San Francisco) and have been described (21). Plasmids containing full-length Rap1GAP (designated pCAN-Rap1GAP) and fragments thereof were previously described (20). Other constructs used in yeast two-hybrid screen have also been described (14). The plasmid containing full-length Rap1GAP with an appended NH 2 -terminal His-tag (pR-SET-Rap1GAP) was constructed by subcloning a KpnI-and EcoRIdigested fragment from pCAN-Rap1GAP that contained the Rap1GAP coding sequence into the pRSET-C vector (Invitrogen).
Production and Purification of Proteins-Rap1GAP was purified as described previously (20). His-tagged Rap1GAP was produced in Escherichia coli strain BL21(DE3)/pLysS (Novagen) carrying pRSET-Rap1GAP. Briefly, cells were grown to an A 600 of 0.6 and protein production was induced by 0.5 mM isopropyl-␤-D-thiogalactopyranoside at 37°C for another 3 h. Cells were harvested by centrifugation at 5000 ϫ g and resuspended (10 ml/liter of culture) in 20 mM Tris-Cl, pH 7.7, 10 mM ␤-mercaptoethanol, and a mixture of protease inhibitors (22). The cell suspension was lysed in a French press and insoluble material was removed by centrifugation at 15,000 ϫ g. The resulting supernatant was diluted 2-fold such that the final buffer composition was 20 mM Tris-Cl, pH 7.7, 10 mM ␤-mercaptoethanol, protease inhibitors, 150 mM NaCl, and 0.1% Lubrol. This extract was applied to a 2-ml column of Ni-NTA resin (Qiagen) that had been pre-equilibrated in the same buffer. Bound proteins were eluted with a step gradient consisting of sequential 6-ml aliquots of the same buffer containing 5, 10, 25, 50, 100, and 150 mM imidazole; Rap1GAP eluted at 100 mM imidazole. G␣ s , G␣ z , and myristoylated G␣ i2 were purified from E. coli as described previously (8,23,24). Myristoylated G␣ z was purified from Sf9 cells as described previously (13).
Cell Culture and Transfection Conditions-HEK293 cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) with 10% fetal bovine serum at 37°C. Transfection was performed using LipofectAMINE TM reagent (Life Technologies, Inc.) in six-well Falcon plates. Standard protocols provided by manufacturer were used. In each transfection, 2.0 g of plasmid encoding G␣ z forms and/or 0.5 g of the plasmid encoding Rap1GAP were used whenever applicable. Sf9 cells were cultured in SF-900 II SFM medium (Life Technologies, Inc.). Preparation of membranes from Sf9 cells overexpressing adenylyl cyclase Type V were prepared as described previously (25).
Formation of Liganded Forms of G Proteins-G␣ z -GDP-(AlF 4 Ϫ ) was formed by incubation of G␣ z in 50 mM Na-HEPES, pH 7.7, 1 mM dithiothreitol, 0.1% Lubrol (Buffer A) containing 5 mM MgCl 2 , 400 M AlCl 3 , 10 mM sodium fluoride, and 5 M GDP at 30°C for 15 min. G␣ z -GTP␥S was formed by incubation of G␣ z in Buffer A containing 5 mM EDTA and 100 M GTP␥S. Both forms were isolated by gel filtration using 1-ml spin columns of G-50 Sephadex (Amersham Pharmacia Biotech) that had been prewashed with Buffer A containing 1 mM EDTA and 5 mM MgCl 2 . Rap1-GTP␥S was formed by incubation of Rap1 with Buffer A containing 1 mM EDTA, 1 mM MgCl 2 , and 100 M GTP␥S at 30°C for 10 min; the complex was isolated by the same technique used for G␣ z forms except that the G-25 Sephadex resin used in the spin column.
GAP Assays-For single turnover GTPase (commonly referred to as GAP) assays, G␣ z was first incubated with [␥-32 P]GTP at 30°C for 30 min in Buffer A containing 5 mM EDTA and 2 M [␥-32 P]GTP. The G␣ z ⅐GTP complex was then isolated by gel filtration using a 1-ml spin column of G-50 Sephadex that had been prewashed with Buffer A containing 1 mM EDTA and 5 mM MgCl 2 . GTP hydrolysis was carried out in the presence or absence of specified RGS proteins. In experiments assessing the abilities of Rap1GAP and its fragments to block RGSmediated acceleration of the GTPase reaction, the Rap1GAP proteins were incubated with the G␣ z ⅐GTP complex at 0°C for 5 min before addition of RGS proteins. GTPase activity was measured by phosphate release as described previously (26).
Adenylyl Cyclase Assay-Adenylyl cyclase activity was measured by using SpinZyme TM Acidic Alumina Devices (Pierce). The standard protocols provided by manufacturer were used. The activated form of G␣ s was formed by incubation at 30°C for 1 h with 4 mM free Mg 2ϩ and 100 M GTP␥S. Activated forms of myristoylated G␣ i2 and Sf9-produced G␣ z were produced by incubation of the proteins at 30°C for 2 h with 4 mM (for G␣ i2 ) or 5 nM (for G␣ z ) free Mg 2ϩ and 100 M GTP␥S (27,28). Free nucleotide was removed through use of a 1-ml spin column of G-50 Sephadex that had been prewashed with Buffer A containing 1 mM EDTA and 5 mM MgCl 2 . Either G␣ s -GTP␥S or forskolin was used to activate the adenylyl cyclase in the Sf9 membranes as described (17). To examine the effect of Rap1GAP on G␣ proteins, G␣ i2 -GTP␥S or G␣ z -GTP␥S and Rap1GAP were incubated for 5 min at 0°C in Buffer A containing 1 mM EDTA and 5 mM MgCl 2 before they were added into the reaction.
Co-precipitation Assays-Two types of co-precipitation assays were performed; the first series being completely in vitro using purified proteins. Purified full-length Rap1GAP containing a NH 2 -terminal Histag was incubated with G␣ z -GDP-(AlF 4 Ϫ ) or G␣ z -GTP␥S at 4°C for 1 h together with Ni-NTA resin. To determine whether Rap1GAP and Rap1 can form a stable complex, Rap1GAP and Rap1 were first incubated in Buffer A containing 5 mM MgCl 2 , 400 M AlCl 3 , 10 mM sodium fluoride, and 5 M GDP or in Buffer A containing 5 mM EDTA, 100 M GTP␥S, respectively, at 30°C for 15 min. The reactions were then diluted 5-fold with Buffer A containing 5 mM MgCl 2 to decrease the nonspecific binding to Ni-NTA resin. Following dilution, the reactions were incubated with the Ni-NTA resin for 1 h at 4°C. To assess complex formation between G␣ z , Rap1GAP, and Rap1, His-tagged Rap1, full-length Rap1GAP, and G␣ z -GTP␥S were first incubated in Buffer A containing 5 mM MgCl 2 , 400 M AlCl 3 , 10 mM sodium fluoride, and 5 M GDP at 30°C for 15 min. Reaction mixtures were then diluted 5-fold with Buffer A containing 5 mM MgCl 2 to decrease the nonspecific binding to Ni-NTA resin and the diluted reactions incubated with Ni-NTA for 1 h at 4°C. After the incubation, Ni-NTA resin was extensively washed. Resin then was boiled in SDS loading buffer to recover bound proteins and analyzed by SDS-polyacrylamide gel electrophoresis on 12% gels followed by Western blotting.
The second series of co-precipitation assays was designed to assess molecular interactions between the relevant proteins in intact cells. HEK293 cells were harvested 48 h after transfection and lysed in 20 mM Tris-Cl, pH 7.7, 1 mM EDTA, 1 mM dithiothreitol, 0.1% Nonidet P-40, and a mixture of protease inhibitors (22). Cytosolic fractions were prepared by centrifugation of cell extracts at 30,000 ϫ g for 1 h. Immobilized Rap1GAP antibody was prepared by incubation of the affinity purified antiserum with Protein A-Sepharose resin (Amersham Pharmacia Biotech) at 4°C for 12 h. Rap1GAP was immunoprecipitated from the prepared cytosolic fractions by addition of the immobilized antiserum followed by incubation at 4°C for another 2 h. Resin was then collected, washed, and analyzed as described above.

RESULTS
To identify proteins involved in G z signaling pathways, a yeast two-hybrid screen was performed in which the mutationally activated Q205L variant of G␣ z was used as "bait" to screen a human brain cDNA library. Because an active form of G␣ z was used in this screen, the expected interactors included regulators of G␣ z GTP hydrolysis and/or downstream effectors. This expectation was validated by our previously reported finding that one of the molecules identified in this screen was a novel RGS protein, RGSZ1, a selective regulator of the G␣ z GTPase reaction (14). In addition to RGSZ1, among more than 100 total positive clones identified, 24 sequences encoded a previously identified protein termed Rap1GAP, a GTPase activating protein for the monomeric GTP-binding protein Rap1.
The initial examination of the specificity of the interaction between the active form of G␣ z and Rap1GAP was performed through the use of a two-hybrid counter-screen. In this counterscreen, the Rap1GAP cDNA was co-transformed in the yeast reporter system together with the mutationally activated forms of several other G protein ␣ subunits, including two other G i family members (G␣ o and G␣ i2 ) and at least one member from each of the remaining three G protein families (Table I). No interaction could be detected in this counter-screen between Rap1GAP and any other G protein ␣ subunit with the exception of G␣ i2 , for which a very weak interaction was observed. To eliminate the possibility that the inability to detect the interaction between Rap1GAP and the G protein ␣ subunits was due to lack of expression of the ␣ subunits, an additional counterscreen between all the ␣ subunits and the aforementioned RGSZ1 was performed. Interactions between RGSZ1 and G␣ o and G␣ i2 , as well as G␣ z , were detected in the system (Table I), indicating that at least G␣ o and G␣ i2 were expressed in functional forms in yeast cells.
The observed interaction between the active form of G␣ z and Rap1GAP, and the accumulating literature indicating that both G z and Rap1 were involved in signaling pathways influencing cell proliferation and differentiation (18,29,30), prompted us to initiate an investigation of the biochemical properties and functional significance of their interaction. In the first series of experiments, the interaction between G␣ z and Rap1GAP was further evaluated both in vitro, using purified recombinant proteins, and in the context of a cell expressing both proteins. In the in vitro experiments, interaction between purified full-length Rap1GAP containing an appended NH 2terminal His-tag and recombinant G␣ z , purified from a bacterial expression system, was examined in co-precipitation assays. Interaction of activated forms of G␣ z , i.e. the GTP␥Sbound form or that activated by aluminum fluoride, with Rap1GAP was readily detected, while interaction with GDPbound form of G␣ z could not be detected (Fig. 1A). To investigate the interaction between G␣ z and Rap1GAP in a cellular environment, co-precipitation experiments were performed using extracts from HEK293 cells expressing the proteins. Cells were transfected with G z ␣ (either the Q205L constitutive-active mutant or the wild-type form) in both the absence and presence of co-transfection with full-length Rap1GAP. Cells were lysed 48 h post-transfection and cytosolic fractions isolated. A polyclonal antibody against Rap1GAP was then used to immunoprecipitate Rap1GAP along with associated proteins from this cytosolic fraction. The results of this experiment showed that the mutationally activated form of G␣ z , but not wild-type G␣ z , could be co-immunoprecipitated with Rap1GAP (Fig. 1B), even thought the wild-type protein was expressed at an equal level in the cells (not shown). These results, together with the in vitro co-precipitation results, confirmed that the activated form of G␣ z specifically interacts with Rap1GAP.
Having confirmed that activated G␣ z and Rap1GAP could indeed form a complex, experiments were designed to deter-mine whether the interaction had functional consequences in terms of measurable activities of G␣ z . Rap1GAP binding to G␣ z did not enhance the single turnover GTPase activity of the G protein ( Fig. 2C and results not shown). Because RGS proteins also bind selectively to active forms of a number of G protein ␣ subunits (2, 5), including G␣ z , the possibility that the interaction between G␣ z and Rap1GAP would interfere with the ability of RGS proteins to act on G␣ z was then examined. Two RGS proteins, RGS10 and RGSZ1, were chosen to test this hypothesis; RGS10 acts on G␣ z as well as several other members of the G i family (26), while RGSZ1 action is quite selective for G␣ z (14,31). To assess whether the interactions of Rap1GAP and the RGS proteins with G␣ z were mutually exclusive, purified recombinant proteins produced in bacteria were used in singleturnover GTPase assays. Briefly, G␣ z was loaded with [␥-32 P]GTP and the GTP-bound form of G␣ z was isolated by gel filtration. Both basal and RGS-stimulated GTP hydrolysis by G␣ z was then determined in the presence or absence of Rap1GAP. Stimulation of GTP hydrolysis of G␣ z by both RGS10 and RGSZ1 was readily observed in these assays (Fig.  2). Rap1GAP itself did not affect the basal GTP hydrolysis rate of G␣ z (data not shown), but the presence of Rap1GAP mark- The screen was performed in yeast reporter strain, Y190. Briefly, the ability of protein product of each identified clone to interact with the indicated mutationally activated G protein ␣ subunits was determined by co-transforming the identified plasmid, encoding either Rap1GAP or RGSZ1, with bait constructs containing the appropriate Gln to Leu (QL) mutants of G␣ z , G␣ i2 , G␣ o , G␣ s , G␣ q , G␣ 13 , and G␣ 15 . Results were measured by X-gal filter assay.

FIG. 1. Co-precipitation analysis of G␣ z -Rap1GAP interactions.
A, in vitro analysis using purified recombinant proteins. The indicated forms of G␣ z (all at 200 nM) were incubated either in the absence (Ϫ) or presence (ϩ) of 200 nM His-tagged Rap1GAP. Ni-NTA resin then used to precipitate Rap1GAP with associated proteins, which were then identified by immunoblot analysis as described under "Experimental Procedures." B, analysis of G␣ z -Rap1GAP interactions in a cellular context. The indicated form of G␣ z was expressed in HEK293 cells in either the absence (Ϫ) or presence (ϩ) of co-transfection with full-length Rap1GAP. Rap1GAP and associated proteins were immunoprecipitated from cell extracts using the affinity purified anti-Rap1GAP antibody and then identified by immunoblot analysis as described under "Experimental Procedures." Q205L, HEK293 cells expressed the mutationally activated form of G␣ z ; W.T., HEK293 cells expressed wild type G␣ z . For both A and B, the upper subpanel is an immunoblot in which the anti-Rap1GAP antibody was used, while lower subpanel is an immunoblot in which the anti-G␣ z antibody was used. edly attenuated the stimulation of G␣ z GTPase activity by either RGS protein. The ability of Rap1GAP to attenuate RGSmediated stimulation of G␣ z GTPase was concentration-dependent and nearly complete (Fig. 2). Additionally, the effect was quite specific for G␣ z , as Rap1GAP did not affect the ability of RGS10 to stimulate G␣ i2 GTPase activity (Fig. 2C), even though a weak interaction between Rap1GAP and G␣ i2 had been detected in the two-hybrid system (Table I).
We also sought to determine whether the interaction of Rap1GAP with G␣ z could interfere with the G protein's ability to regulate effector molecules. The only characterized activity of G z in terms of modulation of effector activity is that the active form of G␣ z can inhibit some subtypes of adenylyl cyclase (15). To examine whether Rap1GAP binding to activated G␣ z could influence this activity of the G protein, studies were performed with adenylyl cyclase type V (AC-V), which is subject to inhibition by activated G␣ z (15). AC-V was expressed in Sf9 cells and the membrane fraction containing the protein prepared. This recombinant AC-V preparation was then activated with either the GTP␥S-bound form of G␣ s or forskolin, and the influence of G␣ z in the presence or absence of Rap1GAP was determined. As previously reported (15), inhibition of AC-V activity by the GTP␥S-activated form of G␣ z could be readily observed (Fig. 3A). Addition of purified recombinant Rap1GAP attenuated the inhibition by activated G␣ z , and this attenuation increased with the increasing concentration of Rap1GAP (Fig. 3A). Similar results were obtained when forskolin-activated AC-V was used (Fig. 3B). While the effect was somewhat modest, the specificity of the interaction was confirmed by the finding that Rap1GAP had no effect on the ability of the GTP␥S-bound form of G␣ i2 to inhibit adenylyl cyclase under the same conditions (data not shown).
Having verified the binding of Rap1GAP to activated G␣ z and determined that there were indeed functional consequences to the interaction of the two molecules in terms of G␣ z activity, we next sought to determine whether Rap1 binding to Rap1GAP could influence the latter's ability to interact with G␣ z . In initial experiments, very little influence of Rap1 binding to Rap1GAP on interactions with G␣ z , or vice versa, was detected (results not shown). While the failure to observe "communication" between Rap1 and G␣ z interactions with Rap1GAP was initially disappointing, the data obtained did suggest that Rap1GAP contained distinct binding sites for Rap1 and G␣ z , otherwise at least competition between the two molecules would have been observed. Hence, we performed a preliminary structure-function study to obtain information on the region of Rap1GAP responsible for conferring the ability to interact with G␣ z . Prior to this work, the only known activity of Rap1GAP was its ability to stimulate GTP hydrolysis of Rap1; the region responsible for this activity, termed the "core domain," had been localized to residues 75 to 416 in the primary sequence (19). To determine whether this or other region(s) of Rap1GAP were required for the interaction with G␣ z , a series of NH 2terminal and COOH-terminal truncations of Rap1GAP, produced as recombinant proteins in bacteria, were examined for their abilities to interact with activated G␣ z . A Rap1GAP fragment in which the COOH-terminal 221 amino acids were deleted retained both the abilities to block RGS action on G␣ z and to function as a GAP for Rap1 (Fig. 4). However, deletion of the NH 2 -terminal 74 residues resulted in a Rap1GAP fragment that had completely lost its ability to compete for RGS action on G␣ z , even though, as previously reported (19), it retained ability to act as a GAP for Rap1. As expected from these results, deletion of both the NH 2 -terminal 74 and COOH-terminal 221 residues produced a fragment with full GAP activity toward Rap1 but no ability to block the RGS action on G␣ z . While we FIG. 2. Rap1GAP blocks GTP hydrolysis of G␣ z stimulated by RGS proteins. GTP hydrolysis by G protein ␣ subunits was determined at 0°C for 10 min as described under "Experimental Procedures." Concentrations of G␣ z -GTP and G␣ i2 -GTP was 50 nM and Rap1GAP concentration was 100 nM, or in the case of panel C as indicated in the graph. A, Rap1GAP blocks G␣ z GTP hydrolysis stimulated by RGSZ1. RGSZ1 concentration was 10 nM. B, Rap1GAP blocks G␣ z GTP hydrolysis stimulated by RGS10. RGS10 concentration was 40 nM. C, Rap1GAP blocks GTP hydrolysis of G␣ z stimulated by RGS10 (f), but not basal GTP hydrolysis (OE) or GTP hydrolysis of G␣ i2 stimulated by RGS10 (q) and the ability is dependent on Rap1GAP concentration. The RGS10 concentration was 10 nM. The star (*) indicates that Rap1GAP was heat inactivated.
have not identified a NH 2 -terminal fragment that interacts only with activated G z ␣, these data indicate that the binding domains for G z ␣ and Rap1 are dissociable in the Rap1GAP structure.
The finding that distinct domains of Rap1GAP are required for interactions with the two distinct GTP-binding proteins provides strong support for a model in which Rap1GAP functions somehow to integrate G z and Rap1 signaling pathways. Furthermore, these findings suggested that it should be possible to assemble a ternary complex that contained all three proteins, i.e. Rap1GAP, G␣ z , and Rap1. To test this hypothesis, it was necessary to develop a method to detect association between Rap1GAP and Rap1 that did not rely on a transient catalytic read-out, i.e. GTP hydrolysis by Rap1. The assay developed was based on a previous report that Rap1 and Rap1GAP form a stable complex when Rap1 is activated by aluminum fluoride (32). To confirm this finding, a co-precipitation assay was performed in which His-tagged Rap1GAP was incubated with various forms of Rap1 (Rap1-GDP, Rap1-GTP␥S, or Rap1-GDP-(AlF 4 )) prior to precipitation of Rap1GAP and associated molecules with Ni-NTA resin. Consistent with the aforementioned report, only Rap1-GDP-(AlF 4 ) co-precipitated together with Rap1GAP in this assay (Fig. 5A). The confirmation that an interaction, sufficiently stable for co-precipitation, could be induced between Rap1 and Rap1GAP allowed a variation of this assay designed to detect ternary complex formation to be performed. His-tagged Rap1, untagged Rap1GAP, and G␣ z -GTP␥S (which exhibited stronger interaction with Rap1GAP than G␣ z -GDP-(AlF 4 ), see Fig. 1A), were incubated together and Ni-NTA resin added to precipitate Rap1-Rap1GAP complex. Since His-Rap1 was the "handle" in this experiment, co-precipitation of G␣ z would indicate that both G␣ z and His-Rap1 were simultaneously bound to Rap1GAP. Indeed, only in the presence of both Rap1 and Rap1GAP could G␣ z be precipitated with the Ni-NTA resin (Fig. 5B). Increasing the amount of Rap1GAP in the assay resulted in increased precipitation of G␣ z , indicating that Rap1GAP was the limiting protein in this system and acted as a bridge to bring G␣ z and Rap1 together (Fig. 5B). These results suggest that the three proteins can form a stable complex, and that this ternary complex might perform unique function(s) in the cell. DISCUSSION Stable expression of constitutively active G␣ z can result in transformation of Swiss 3T3 and NIH3T3 cells (18); this study provided the initial link between G z signaling and mitogenic pathways. Because of the universal involvement of monomeric GTP-binding proteins in Ras superfamily in mitogenic signaling pathways (33), regulators of these proteins would be attrac- FIG. 3. Rap1GAP attenuates the ability of G␣ z to inhibit adenylyl cyclase V. Adenylyl cyclase activity was measured at 30°C for 10 min as described under "Experimental Procedures." A, adenylyl cyclase V, activated by 50 nM G␣ s -GTP␥S, was subject to inhibition by 100 nM G␣ z -GTP␥S. The inhibition was attenuated by 200 or 800 nM Rap1GAP as indicated. The star (*) indicates that Rap1GAP was heat inactivated. B, adenylyl cyclase V, activated by 50 nM forskolin, was subjected to inhibition by 50 nM G␣ z -GTP␥S. Where indicated, Rap1GAP concentration was 200 nM.

FIG. 4. The NH 2 -terminal region of Rap1GAP is required for attenuation of RGS action on G␣ z .
The ability of each Rap1GAP fragment to interact with G␣ z was determined by GAP assays, either directly for Rap1GAP action on Rap1 (19) or by competition for RGS action on G␣ z , as described under "Experimental Procedures." A, schematic representation of four different Rap1GAP fragments, their GAP activities toward Rap1, and abilities to interact with G␣ z (see panel B for details). B, abilities of Rap1GAP fragments to block RGS action on G z ␣. 40 nM RGS10 was present in the GAP assays as indicated. W.T., wild type, W.T.*, Rap1GAP was heat inactivated. tive candidates to act as signal integrators and/or coordinators with G␣ z during this transformation process. The studies in this report reveal that the active form of G␣ z interacts specifically with one such regulatory molecule, Rap1GAP. Rap1GAP is an 88-kDa peptide that specifically stimulates GTP hydrolytic activity of Rap1 (19,34); Rap1 is a member of the Ras superfamily that possesses the ability to both transform NIH3T3 cells and trigger differentiation of PC12 cells under certain conditions (29,30).
Prior to this study, the only known function of Rap1GAP was its GAP activity toward Rap1 that presumably serves to downregulate signaling through Rap1 (19). However, as with most GAPs for monomeric G proteins, only a relatively small segment of Rap1GAP is required for the GAP activity. Specifically, the catalytic domain of Rap1GAP had been identified as contained within the region from residue 75 to 416 of the primary structure (20), while no function had been identified as being associated with the NH 2 -terminal or COOH-terminal regions of the protein. The finding that the NH 2 -terminal domain of Rap1GAP is required for the protein's ability to bind activated G␣ z provides a function within its NH 2 -terminal domain that is distinct from that of its core domain. These results thus reveal Rap1GAP as a possible signal integrator between G z signaling and Rap1 signaling.
Analysis of the functional consequences of Rap1GAP binding to G␣ z revealed that the binding both blocked the ability of RGS proteins to stimulate GTP hydrolysis of G␣ z and attenuated the ability of activated G␣ z to inhibit adenylyl cyclase activity. Furthermore, the binary complex of G z ␣ and Rap1GAP retained the ability to interact with the active form of Rap1, allowing formation of a ternary complex that contained all three proteins. Structural analyses demonstrate that heterotrimeric G proteins possess distinct domains whose conformations are altered upon activation (35,36); these so-called "switch regions" have been implicated in both RGS and effector interactions of activated G␣ subunits (37,38). Presumably, Rap1GAP interacts with one or more of the switch regions of G␣ z and competes for access of RGS proteins. Similarly, interaction of Rap1GAP with one or more of these regions of G␣ z would also account for the observed attenuation of the ability of activated G␣ z to inhibit adenylyl cyclase, although this latter effect was not as pronounced as the inhibition of RGS access to the protein.
While the data indicating a functional consequence of Rap1GAP binding on G␣ z functions are quite compelling, it is less clear whether the G␣ z -Rap1GAP interaction also affects the ability of Rap1GAP to modulate Rap1 signaling properties. What is clear from both the structure-function analysis and co-precipitation experiments is that the abilities of Rap1GAP to both bind active forms of G␣ z and stimulate Rap1GTP hydrolysis are dissociable. Although our preliminary experiments indicate that the G␣ z binding to Rap1GAP does not modify the GAP activity of Rap1GAP toward Rap1 (data not shown), an alternative mode of cross-regulation may still exist in which G␣ z affects Rap1 signaling by recruiting its regulator, Rap1GAP, from cytosol to plasma membrane. Through such a "colocalization-type" mechanism, relocalization of Rap1GAP to some specific subcellular location might result in simultaneous modulation of both G z and Rap1 signaling processes.
There may be a biological precedence to the ability of a GAP for a monomeric G protein to influence signaling by a trimeric G protein. In Drosophila development, a protein termed the neurofibromatosis type 1 protein (NF1), which can function as a Ras-specific GAP, is required for the activation of adenylyl cyclase in response to the agonist PACAP-38 (39,40). The phenotype derived from loss of NF1 expression can be rescued by elevating cAMP, but not by manipulating Ras signaling (39). Therefore, both NF1 and the events leading to cAMP elevation, e.g. activation of G s or an attenuation of a G i input, seem to be somehow linked in a signaling network. Additionally, it has been reported recently that activated G␣ 12 can directly interact with, and stimulate the GAP activity of, a specific type of GAP that acts on Ras, termed RasGAP1 m (41). In yet other studies, interaction of activated G 12 with an exchange factor for the monomeric G protein Rho stimulated the exchange activity of the molecule (42,43). Together, these and current studies support the notion that complex cellular communication exists between monomeric and heterotrimeric G proteins through regulators of the monomeric G proteins. Elucidating the mechanisms and functional consequences of these communication networks should prove an exciting challenge for the future.