The α Subunits of Gz and Gi Interact with the eyes absent Transcription Cofactor Eya2, Preventing Its Interaction with the Six Class of Homeodomain-containing Proteins*

Yeast two-hybrid techniques were used to identify possible effectors for the heterotrimeric G protein Gz in human bone marrow cells. Eya2, a human homologue of the Drosophila Eya transcription co-activator, was identified. Eya2 interacts with activated Gαz and at least one other member of the Gαi family, Gαi2. Interactions were confirmed in mammalian two-hybrid and glutathione S-transferase fusion protein pull-down assays. Regions of Eya2-mediating interaction were mapped to the C-terminal Eya consensus domain. Eya2 is an intrinsically cytosolic protein that is translocated to the nucleus by members of the Six homeodomain-containing family of proteins. Activated Gαzand Gαi2 prevent Eya2 translocation and inhibit Six/Eya2-mediated activation of a reporter gene controlled through the MEF3/TATA promoter. Although G proteins are known to regulate the activity of numerous transcription factors, this regulation is normally achieved indirectly via one or more intermediates. We show here a novel functional regulation of a co-activator directly by G protein subunits.

Yeast two-hybrid techniques were used to identify possible effectors for the heterotrimeric G protein G z in human bone marrow cells. Eya2, a human homologue of the Drosophila Eya transcription co-activator, was identified. Eya2 interacts with activated G␣ z and at least one other member of the G␣ i family, G␣ i2 . Interactions were confirmed in mammalian two-hybrid and glutathione S-transferase fusion protein pull-down assays. Regions of Eya2-mediating interaction were mapped to the Cterminal Eya consensus domain. Eya2 is an intrinsically cytosolic protein that is translocated to the nucleus by members of the Six homeodomain-containing family of proteins. Activated G␣ z and G␣ i2 prevent Eya2 translocation and inhibit Six/Eya2-mediated activation of a reporter gene controlled through the MEF3/TATA promoter. Although G proteins are known to regulate the activity of numerous transcription factors, this regulation is normally achieved indirectly via one or more intermediates. We show here a novel functional regulation of a co-activator directly by G protein subunits.
The ability of a cell to respond to external cues is dependent on a large array of receptors and downstream signaling pathways. Heterotrimeric G proteins contribute to these pathways by helping to conduct the flow of information from agonistactivated heptahelical receptors to a variety of intracellular targets (1)(2)(3). G proteins have traditionally been studied within the context of second messenger regulation, where adenylyl cyclases, phospholipases, phosphodiesterases, and ion channels assume prominent roles. A substantial amount of attention has been focused more recently on forms of transduction involving protein recruitment apart from second messenger regulation. G␤␥ provides an interesting example, serving as a point of anchorage for several G protein-coupled receptor kinases (4) and, consequently, for signaling molecules recruited by the kinases themselves or arrestin (5). Additional documented interactions include those of G␤ and/or G␤␥ with Raf-1 (6), Cdc24 GEF (7), and Rho family members (8,9). G␣ subunits also participate in protein recruitment relevant to downstream events, for example interacting with regulators of G protein signaling (RGS 1 molecules), which hasten deactivation of G␣ but can coordinately activate other proteins (e.g. Rho (10,11)). G␣ z and G␣ o interact with Rap1 GTPase-activating protein (12,13), and one or more G␣ i family members interact with G protein-regulated inducers of neurite outgrowth (14) and a Ca 2ϩ -binding protein (15).
G proteins regulate the activity of numerous transcription factors typically through second messengers and quite often through other intermediates. The activation of the cAMP-responsive element binding protein CREB proceeds through phosphorylation by protein kinase A (16), which is activated by cAMP in response to activation of G s . Ternary complex factors are activated by extracellular signal-regulated kinases, whose activation through Ras and/or Raf can be initiated by G␤␥ released from G i or by G␣ q (17)(18)(19). Activation of c-Jun is accomplished by c-Jun N-terminal kinases, which respond to G␤␥, G␣ 12 , and G␣ 13 through Cdc42 and Rac and sometimes Ras (20,21). The activity of serum response factor is keyed to the depletion of G actin in response to activation of Rho (22) and can be achieved by G␤␥, G␣ q , G␣ 12 , and G␣ 13 (23). Other transcription factors regulated by G protein signaling include signal transducers and activators of transcription (STATs) (24) and nuclear factor B (25,26).
The G protein G z exists in platelets, neurons, and other highly differentiated cells exhibiting regulated exocytosis (27)(28)(29)(30). The ␣ subunit of G z hydrolyzes GTP quite slowly and is the one member of the G␣ i family that does not serve as a substrate for pertussis toxin. G z communicates with receptors that normally couple to G i (31,32) and has been demonstrated to inhibit adenylyl cyclases I and V (33), to inhibit N-type Ca 2ϩ channels (34), and to activate inwardly rectifying K ϩ channels (34,35). Interactions of G␣ z with a Rap1 GTPase-activating protein, the RGS protein RGSZ1, and inducers of neurite outgrowth have been demonstrated (12,14,36).
In an effort to identify effectors for G z that could play a role in hematopoietic development, we carried out a yeast twohybrid screen for proteins in bone marrow cells that interact with activated G␣ z . We report one such protein here, Eya2. Eya2 is a human homologue of the Drosophila eyes absent gene product and interacts not only with G␣ z but at least one other member of the G␣ i family, G␣ i2 . Eya2 in combination with Six proteins is a transcription co-activator. G␣ z and G␣ i2 interact with Eya2 in mammalian cells, and this interaction represses the transactivating activity of Eya2 in part by inhibiting Six4induced translocation of Eya2 from cytosol to the nucleus.

MATERIALS AND METHODS
cDNAs and Recombinant Plasmids-cDNAs used in this study were human G␣ z (37), G␣ z Q205L, Eya2 (38), and Six1 (39), mouse Six4 (40), and rat G␣ i2 (41), G␣ i2 Q205L, and G␣ s (Q227L). G␣ z and the rat G protein subunit cDNAs were obtained from Drs. H. Fong (University of Southern California) and N. Dhanasekaran (Temple University), respectively. Eya2 and Six1 cDNAs were isolated from a human bone marrow cDNA library (CLONTECH, Palo Alto, CA) by the polymerase chain reaction based on published sequences. Flag (DYKDDDDK)tagged Eya2 was generated by insertion of the Eya2 cDNA into pFlag-CMV2 (Eastman Kodak Co.). HA (YPYDVPDYA)-tagged proteins were generated by polymerase chain reaction using forward primers encoding the HA sequence and subsequent insertion of products into pcDNA3 (Invitrogen, Carlsbad, CA). The tags in both instances were placed at the N terminus. The glutathione S-transferase (GST)-Eya2 (269 -538) fusion protein was generated by subcloning the Eya2 (269 -538) cDNA into pGEX-5x-3 (Amersham Pharmacia Biotech) immediately downstream of the GST sequence. The luciferase reporter construct for measuring Eya2/Six-mediated transcription was generated by cloning the BamHI-EheI fragment containing the adenovirus E1b TATA box from the pNFB-Luc plasmid (Stratagene, La Jolla, CA) into pGL3 (Promega, Madison, WI) to generate pGL3-TATA; six copies of the MEF3 element (40) obtained by polymerase chain reaction were subcloned into pGL3-TATA upstream of the TATA box (SamI/SacII sites). The sequences of all constructs were verified directly. pAS2-1 and pM (CLONTECH) are vectors used to express Gal4 DNA binding domain fusion proteins in yeast and mammalian cells, respectively. pVP16 (CLONTECH) is a vector used to express VP16-activating domain fusion proteins in mammalian cells.
Yeast Two-hybrid Screen-pAG2-1-G z (Q205L) was used as a bait to screen a human bone marrow library (CLONTECH) by sequential transformation according to the manufacturer's protocol. Briefly, the yeast reporter strain Y190 was transformed with pAS2-1-G␣ z (Q205L) and plated on synthetic dropout agar plates lacking tryptophan. Control experiments confirmed that this construct had no autoactivating properties. Yeast surviving the selection protocol were then transformed with pACT2 bone marrow library plasmid DNA and plated onto synthetic dropout agar plates containing 35 mM 3-amino-1,2,4-triazole (Sigma) but lacking tryptophan, leucine, and histidine. Plates were incubated at 30°C for 7 days, and subsequent colonies were transferred to paper filters for measurement of ␤-galactosidase activity according to the manufacturer's protocol. Library plasmids were isolated from positive colonies and used to re-transform Y190 together with pAS2-1-G␣(Q205L) or, as a control, pAS2-1-p53 or empty plasmid to confirm strength of interaction and facilitate isolation of single library plasmids. In other experiments the N-and C-terminal regions of Eya2 were subcloned into pACT2 for analysis of interactions with G␣ z Q205L expressed by pAS2-1.
Mammalian Two-hybrid Assay-HeLa cells maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum were seeded in six-well plates (2 ϫ 10 5 cells/well) the day before transfection. The cells were transfected with pM and pVP16 expressing the indicated proteins or with empty vectors together with pG5CAT (CLONTECH) (0.5 g each) using SuperFect (Qiagen, Valencia, CA); pG5CAT contains five copies of the Gal4 binding site and an adenovirus E1b minimal promoter upstream of the chloramphenicol acetyltransferase (CAT) gene. 0.1 g pGL3-control, which expresses luciferase constitutively, was included as an internal control. At 48 h after transfection, cell lysates were prepared for assays of CAT activity.
In Vitro GST Fusion Protein Pull-down Assays-GST and the GST-Eya2(269 -538) fusion protein were prepared using the pGEX-5x-3 vector and Escherichia coli strain B21. [ 35 S]G␣ z Q205L, G␣ i2 Q205L, and G␣ s Q227L were generated from pcDNA3 vectors by in vitro translation using a TNT T7 coupled transcription/translation system with T7 RNA polymerase (Promega) and [ 35 S]methionine. 2-3 g of individual GST fusion proteins were attached to glutathione-agarose beads (20-l suspension) and incubated with 10 l of the 35 S-labeled subunits in 500 l of binding buffer (20 mM HEPES, pH 7.7, 100 mM NaCl, 2.5 mM MgCl 2 , 250 g/ml bovine serum albumin, 0.05% Nonidet P-40, and 10% glycerol) for 2 h at 4°C with shaking. The beads were washed 4 times with binding buffer, resuspended in 20 l of electrophoresis sample buffer, and heated (90°C) for 5 min. Proteins were visualized after SDSpolyacrylamide gel electrophoresis by autoradiography.
MEF3/TATA Reporter Gene Assay-HeLa cells were seeded into 12-well plates (1 ϫ 10 5 cells/well) the day before transfection with 0.5 g of the reporter gene construct pGL3-MEF3-Luc and the indicated expression plasmids (0.4 g each); 0.01 g of pRL-SV40 plasmid (Promega) was included as the internal control. 48 h after transfection, cell lysates were prepared, and luciferase activities were measured according to the manufacturer's protocol (Promega).
Immunofluorescence Microscopy-C2C12 cells were seeded into 12well plates (3 ϫ 10 4 cells/well) and transfected with 0.5 g of each of the indicated plasmids using Transfectam (Promega). The cells were replated 24 h thereafter onto collagen-coated glass coverslips for an additional 24 h, then fixed with 4% paraformaldehyde at room temperature for 10 min and permeabilized with 0.1% Triton X-100. Blocking was performed with 5% donkey serum and 10% fetal bovine serum in phosphate-buffered saline. The cells were then incubated for 2 h with one or more of the following antibodies in blocking buffer: mouse antiflag (M5, Sigma) at 25 g/ml for detection of FLAG-tagged Eya2, chicken anti-HA (Aves Labs, Tigard, OR) at 1:100 for detection of HA-tagged Six4, rabbit anti-G␣ i2 (Santa Cruz Biotechnology, Santa Cruz, CA) at 1:100, and rabbit anti-G␣ s (1191 (42)) at 1:100. The cells were washed with phosphate-buffered saline and incubated for 1 h with one or more of the following secondary detection reagents: donkey anti-mouse-Cy5 at 1:300, anti-chicken-Cy2 at 1:100, and anti-rabbitrhodamine at 1:75 (Jackson ImmunoResearch, Inc. West Grove, PA). Fluorescence was analyzed by confocal microscopy.

RESULTS
G␣ z and G␣ i2 Interact with Eya2-In an attempt to identify novel effectors for G z , we performed a yeast two-hybrid screen using a constitutively active variant of G␣ z , G␣ z Q205L (G␣ z QL), as bait and a human bone marrow cDNA library (3.5 million independent clones) as a source of interacting proteins; the Q205L mutation impairs the GTPase activity of G␣ z (43). Sixteen positive clones were isolated in our screen from approximately 20 million transformants. Sequencing revealed that five of the clones were the G␣ i -interacting protein GAIP (44). The other 11 clones corresponded to Eya2, a human homologue of the Drosophila eyes absent (Eya) gene. Eya is a transcription co-activator that, as a complex with the sine oculis gene product So, is essential for eye specification and the development of other tissues in Drosophila (45,46). Eya2 is one of four vertebrate forms of Eya (47,48).
The Eya2 clones contained two different lengths of inserts, one corresponding to amino acid residues 15-538 and the other to residues 141-538. Full-length Eya2 is 538 residues long and comprises N-terminal transactivation (1-269) and C-terminal "Eya" consensus (270 -538) domains (Fig. 1). To locate the G␣ z binding site on Eya2, the N-and C-terminal portions of Eya2 (residues 15-300 and 269 -538, respectively) were fused to the Gal4 activation domain and re-tested in yeast with G␣ z QL again fused to the Gal4 DNA binding domain. The C-terminal half of Eya2 alone was found to interact with G␣ z QL.
To confirm the interaction between G␣ z QL and full-length Interaction of G␣ z with Eya2 in yeast two-hybrid assays. Full-length human Eya2 or the depicted N-and C-terminal fragments were tested in yeast two-hybrid assays for interactions with G␣ z QL. Eya2 or its fragments were co-expressed with G␣ z QL as fusion proteins with Gal4 activation and DNA binding domains, respectively, in the yeast reporter strain Y190. Growth of transformants on medium lacking histidine or expression of ␤-galactosidase by the transformants was graded on a scale of no growth or no ␤-galactosidase production (Ϫ) to strong growth or strong ␤-galactosidase production (ϩϩ).
Eya2 in mammalian cells and to test the specificity of the interaction, we employed a mammalian two-hybrid system. The constitutively active forms of G␣ z (G␣ z Q205L), G␣ i2 (G␣ i2 Q205L), and G␣ s (G␣ s Q227L), as well as wild type G␣ z and G␣ i2 were cloned into the Gal4 DNA binding domain vector pM, and the full-length Eya2 was cloned into the VP16 transcription-activating domain vector pVP16. The two expression plasmids were co-transfected into HeLa cells together with a pG5CAT reporter construct. We found that G␣ z QL and G␣ i2 QL interacted strongly with Eya2, as determined by CAT expression, whereas G␣ s QL did not (Fig. 2, panel A). Wild type G␣ z and G␣ i2 subunits, which would assume an inactive conformation predominantly, interacted with Eya2 only very weakly. These data suggest that Eya2 is specific for the activated forms of the two G␣ i family members.
We generated a GST-Eya2 C-terminal domain fusion protein to ascertain in pull-down experiments whether the interaction of Eya2 with G␣ z QL and G␣ i2 QL was direct (Fig. 2, panel B). In vitro translated [ 35 S]methionine-labeled G␣ z QL, G␣ i2 QL, and G␣ s QL were incubated with the Eya2 fusion protein attached to glutathione-agarose beads. G␣ z QL and G␣ i2 QL, but not G␣ s QL, bound the fusion protein. Similar experiments were performed with lysates of human embryonic kidney 293 cells transfected with expression plasmids for wild type G␣ z and G␣ z QL. Consistent with the experiments using in vitro translated subunits, G␣ z QL but not wild type G␣ z bound Eya2 (data not shown). These results are in agreement with those of the yeast and mammalian two-hybrid experiments and support a direct interaction of G␣ z QL and G␣ i2 QL with Eya2.
Eya2 Interacts with Six Proteins-Eya2 also interacts with "Six" proteins, which are mammalian homologues of So and are reported to promote translocation of Eya2 to the nucleus (49). The Six proteins consist of a Six consensus domain near the N terminus, an adjacent homeodomain, and a variable C terminus, which in some instances (e.g. Six4) contains a transactivation domain (Fig. 3, panel A). We carried out a mammalian two-hybrid assay to confirm the interaction between Eya2 and Six proteins using Six1 as the representative (panel B); Six1 had not been evaluated previously. Six1 and the isolated consensus Six domain, both presented as Gal4 DNA binding domain fusion proteins, interacted with Eya2. We also evaluated interactions using a reporter gene (luciferase) under control of a MEF3/TATA promoter. Six proteins alone, to some extent, but the combination of Six proteins with Eya2, to a much greater relative extent, promoted gene expression (panel C); the combination of Six4 and Eya2 was particularly effective in this regard.
The Binding of G␣ Subunits and Six Proteins to Eya2 Is Mutually Exclusive-We asked whether Six1 and Six4 have any effect on the interaction of Eya2 with G␣QL subunits, and vice versa. The mammalian two-hybrid assay was performed using G␣ z QL and G␣ i2 QL subunits fused to the Gal4 DNA binding domain and Eya2 fused to the VP16 activation domain; Six1 and Six4 under the control of a cytomegalovirus promoter were also introduced (Fig. 4, panel A). As shown, both Six1 and Six4 suppressed the interaction of the G␣QL subunits with Eya2. Six1 and Six4 had no effect on the interaction of p53 with the SV40 large T antigen presented as DNA binding and activation domain fusion proteins (not shown), supporting specific-ity of the inhibition. In the converse experiment, disruption of interactions between Six and Eya2 fusion proteins using G␣ z QL or G␣ i2 QL was modest (ϳ20%; not shown), due possibly to the transport of the Eya2 fusion protein independently to the nucleus by an inserted nuclear localization signal.
We pursued the issue of competition through two additional assays, reporter gene activity and immunofluorescence microscopy. As noted above, Six4 and Eya2, when co-expressed, stimulated MEF3/TATA-controlled reporter gene activity by approximately 500-fold (Fig. 4, panel B; see also Fig. 3). We found that G␣ z QL and G␣ i2 QL suppressed the Six4/Eya2-promoted activity by 80%; G␣ s QL did not but, rather, enhanced activity to some extent. G␣ z QL and G␣ i2 QL therefore achieved a substantial inhibition of Six-promoted Eya2 regulation within the context of reporter gene expression.
Immunofluorescence microscopy demonstrated that the majority of Eya2 introduced alone into C2 myoblasts is present in the cytoplasm (Fig. 5, panel A). Eya2 introduced into human embryonic kidney 293 cells partitioned into the 100,000 ϫ g supernatant fraction (data not shown). When co-expressed with Six1 (not shown) or Six4 (panel B), Eya2 was found mostly in the nucleus. These data are consistent with the conclusion that Six proteins promote nuclear translocation of Eya proteins (49). The appearance of Eya2 in the nucleus, however, was attenuated by G␣ i2 QL (panel C). In an analysis of Six4 and G␣ i2 QL distribution specifically, we found that Six4 assumed a nuclear location regardless of Eya2 and/or G␣ i2 QL expression and that the distributions of Eya2 and G␣ i2 QL, when expressed together, were indistinguishable and similar to that of either expressed alone (data not shown). The association of G␣ subunits, including G␣ i2 (50,51) and G␣ z (52), with internal structures has been noted previously. G␣ s QL did not inhibit translocation of Eya2 (panel D). G␣ i2 QL therefore inhibits Six4promoted translocation of Eya2 to the nucleus, as it does Six4promoted transactivating activity of Eya2 in MEF3/TATA reporter gene assays. DISCUSSION Heterotrimeric G proteins are involved in the regulation of numerous transcription factors. In many instances the regulation is achieved through the manipulation of effector-controlled second messengers, mitogen-activated protein kinase cascades, and/or low molecular weight G proteins. We report here that at least two members of the G␣ i family, G␣ z and G␣ i2 , interact directly with the transcription co-activator Eya2. This interaction resembles that of G proteins with other effectors, as the subunits must assume an active conformation for the interaction to occur and the functionality of the target, Eya2, to be altered. G␣ z and G␣ i2 specifically inhibit the interaction of Eya2 with Six proteins and, thus, attenuate the import of Eya2 into the nucleus and subsequent activation of gene expression through the MEF3/TATA promoter element.
Drosophila Eya is the founding member of the Eya family. In Drosophila eye specification, Eya is required for cell proliferation in the undifferentiated epithelium of the imaginal disc, subsequent initiation and propagation of the morphogenetic furrow, and neuronal development (53). Eya normally acts as a complex with the sine oculis gene product So, where So as a homeodomain protein provides targeting through its DNA binding function, whereas Eya provides transactivation functionality (46,54). Eya2, the subject of this study, has the capacity to bind Six1 and Six4, among other mammalian homologs of So, and the resultant complexes activate the myogenin gene (49). We have corroborated the interaction between Eya2 and Six proteins by means of a mammalian two-hybrid system. Using a concatomeric MEF3/TATA reporter gene, we FIG. 4. Reciprocal inhibition by G␣QL subunits and Six1/Six4 of interactions with Eya2. Panel A shows a mammalian two-hybrid assay in which Gal4 DNA binding domain-G␣QL and VP16 activation domain-Eya2 fusion protein plasmids were introduced together with the pG5CAT reporter gene by transfection into HeLa cells with or without Six1 or Six4. Forty-eight hours after transfection, cell lysates were prepared for analysis of CAT activity, which is expressed as the mean fold stimulation Ϯ S.E. for three experiments. Panel B represents a MEF3/TATA transactivation assay in which Six4 and Eya2 were introduced by transfection of plasmids into HeLa cells together with the pGL3-MEF3-Luc reporter gene with or without G protein ␣ subunit plasmids. Forty-eight hours after transfection, cell lysates were prepared for analysis of luciferase activity, which is expressed as the mean fold stimulation Ϯ S.E for three experiments. also demonstrated that Eya2 and Six proteins act cooperatively in transactivation.
Our interest in Eya2 was prompted by identification of Eya2 cDNA sequences in a yeast two-hybrid screen with G␣ z QL as bait. Eya2 was one of two proteins identified in the screen; the other was GAIP (44). Three other proteins have been reported previously to interact directly with activated G␣ z : RGSZ1, Rap1GAP, and GRIN (12,14,36). RGSZ1 and Rap1GAP were cloned in yeast two-hybrid screens using G␣ z QL and a human brain cDNA library, whereas GRIN was cloned from a mouse embryo cDNA expression library using G z ⅐GTP␥S as a probe. Although the number of transformants screened in our assay was 5-fold the number of independent clones, some interacting proteins may have escaped detection due to the heterogeneity in cell populations contributing to the library. We find no apparent difference in the strength of interactions of Eya2 with G␣ z QL and G␣ i2 QL, which is an observation similar to that published for GAIP. RGSZ1, on the other hand, prefers G␣ z QL to G␣ i2 QL (36), whereas Rap1GAP prefers G␣ z QL to all other subunits (12).
Yeast two-hybrid screens for interacting proteins have been applied to a variety of G protein subunits. Active subunit conformations are not always required. Rap1GAP is reported to interact preferentially with unactivated G␣ o (13), whereas calnuc (nucleobindin) interacts with G␣ i3 (55), and Purkinje cell protein-2 interacts with G␣ o (56) regardless of subunit activation state. Yeast two-hybrid techniques have also been used to demonstrate interactions between G␤ subunits and Raf (6), the ankyrin repeat-containing protein (57), RhoA and Cdc42 (9), and N-type calcium channel sequences (58).
That the interaction of G␣ z or G␣ i2 with Eya2 requires an active conformation is consistent with the idea that Eya2 is an effector, but not proof. In first assessing the probability of a normal interaction between the two proteins, our concern was that Eya2 would exist primarily in the nucleus, as does Drosophila Eya. However, our work and that published by Ohto et al. (49) show that Eya2 expressed alone assumes a cytoplasmic distribution. Its co-fractionation with cytosolic protein following cell lysis suggests that it is in fact diffusable. These findings are perhaps not unexpected. While Drosophila Eya has a consensus nuclear localization signal, the vertebrate proteins do not, suggesting differences in functions related to subcellular location (38). Eya2 would have ample opportunity as a cytosolic protein, as do several G protein effectors, to interact with G␣ subunits.
Six proteins provide the means by which Eya2 can be transported into the nucleus and affixed to DNA in order to assume a transactivating role. A reasonable question is whether G␣ z QL and G␣ i2 QL can effectively compete with Six proteins in this process. The data presented here demonstrate that they can. In the MEF3/TATA transactivation assays in HeLa cells and in the immunofluorescence experiments with C2 myoblasts, the G␣QL subunits almost completely negated the actions of Six1 and Six4. The antagonism is not one-sided, however, as Six1 and Six4 can disrupt the interaction between Eya2 and G␣QL subunits in the mammalian two-hybrid assay. Although much remains to be understood about the presentation of Eya2 to appropriate gene regulatory elements, we suspect that G i2 and G z can regulate gene transcription through Eya2 in several ways. The activation of G i2 or G z in the intact cell by agonistactivated receptors involves conformational changes in G␣ i2 and G␣ z coincident with GTP binding. Subunits assuming an active conformation could bind cytosolic Eya2 at membranes or, under circumstances described for at least one other subunit (G␣ s ) (59), conceivably in the cytosol. If Eya2 is normally carried to the nucleus by Six proteins in a constitutively or regulated fashion, activated G␣ i2 or G␣ z would be anticipated to interrupt that process and thereby diminish the amount of Eya2 entering the nucleus. Additionally, Eya2 and Six proteins in the nucleus would probably not remain locked together as a transactivating complex. G␣ i2 or G␣ z might therefore decrease nuclear Eya2 based on the equilibrium established between nuclear and cytosolic forms of the protein. In either case, any gene activity under the control of Eya2 would be diminished. G␣ z hydrolyzes GTP much more slowly than most other G␣ subunits (60), a property conducive to sequestration of Eya2. G protein subunits of the G i family, moreover, are reported to translocate or diffuse into the nucleus in certain instances FIG. 5. G␣ i2 QL inhibits the nuclear translocation of Eya2 otherwise induced by Six4. FLAG-tagged Eya2 was introduced by transfection into C2 myoblasts alone (panel A), with HA-tagged Six4 (panel B), with HA-tagged Six4 and G␣ i2 QL (panel C), or with HA-tagged Six4 and G␣ s QL (panel D). The cells were fixed and permeabilized 48 h after transfection, then incubated with mouse anti-flag, chicken anti-HA, and rabbit anti-G␣ antibodies followed by Cy5-conjugated donkey anti-mouse, Cy2-conjugated donkey anti-chicken, and rhodamine-conjugated donkey anti-rabbit antibodies. Data for the mouse anti-FLAG/donkey anti-mouse antibodies alone, i.e. the distribution of Eya2 alone or as influenced by other proteins, are shown in the four panels. Cells were viewed by confocal microscopy. The data represent one of three experiments providing equivalent results. (61)(62)(63), which may also be relevant to interactions of the subunits with Eya2. In our studies, we did not distinguish between monomeric ␣⅐GDP and ␣⅐GDP complexed with G␤␥ as the inactive forms of wild type subunits; we suspect that both forms of subunit cannot interact with Eya2 and that interaction is the property of the active form only.
Six proteins need not always be considered. For cells lacking Six proteins, Eya2 would be completely cytosolic, and the interaction of activated G␣ subunits with Eya2 could be envisioned in much the same manner as that for the several already documented effectors and RGS proteins. In this context, Eya2 could exhibit activities apart from those of transactivation. Conversely, and just as significantly, Eya2 may prove relevant to the activation state of G␣ z and G␣ i2 or the ability of these subunits to interact with other effectors.
We believe that the inhibition of Eya2⅐Six1/4-activated MEF3/TATA reporter gene activity by G␣ z QL and G␣ i2 QL is most likely due to the inhibition of Eya2 translocation into the nucleus. Other routes of inhibition, for example through G␣ subunit-regulated second messenger pathways, remain possible. G␣ s QL, which by two-hybrid analysis does not interact with Eya2 and by immunofluorescence does not disrupt the interaction between Eya2 and Six4, enhances Eya2⅐Six4-promoted MEF3/TATA reporter gene activity to a small extent. A similar enhancement of activity was also found with dibutyryl cyclic AMP (not shown). If cAMP can promote the interaction or transactivating properties of Eya2 and Six4, it is not inconceivable that decreases in cAMP possibly caused by G␣ z QL and G␣ i2 QL can have the opposite effect. However, dibutyryl cyclic AMP does not attenuate the inhibition by G␣ i2 QL of Eya2/Six4promoted reporter gene activity by any but a small amount (not shown), so that the sequestration of Eya2 by the G␣ subunit remains the most plausible mode of inhibition.
Multiple members of the mammalian Eya family exist. Given the high degree of homology among consensus Eya domains, we predict that activated forms of G␣ z and G␣ i2 will interact with several other forms of mammalian Eya proteins. The extent to which these interactions influence the expression of different genes and/or the activities of G i family members is well worth exploration.