An N-terminal Region of Caenorhabditis elegans RGS Proteins EGL-10 and EAT-16 Directs Inhibition of G o Versus G q Signaling*

Transcription

1 THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 277, No. 49, Issue of December 6, pp , by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A. An N-terminal Region of Caenorhabditis elegans RGS Proteins EGL-10 and EAT-16 Directs Inhibition of G o Versus G q Signaling* Received for publication, August 9, 2002, and in revised form, September 16, 2002 Published, JBC Papers in Press, September 26, 2002, DOI /jbc.M Georgia A. Patikoglou and Michael R. Koelle From the Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut Regulator of G protein signaling (RGS) proteins contain an RGS domain that inhibits G signaling by activating G GTPase activity. Certain RGS proteins also contain a G -like (GGL) domain and a poorly characterized but conserved N-terminal region. We assessed the functions of these subregions in the Caenorhabditis elegans RGS proteins EGL-10 and EAT-16, which selectively inhibit GOA-1 (G o ) and EGL-30 (G q ), respectively. Using transgenes in C. elegans, we expressed EGL-10, EAT-16, their subregions, or EGL-10/EAT-16 chimeras. The chimeras showed that the GGL/RGS region of either protein can act on either GOA-1 or EGL-30 and that a key factor determining G target selectivity is the manner in which the N-terminal and GGL/RGS regions are linked. We also found that coexpressing N-terminal and GGL/RGS fragments of EGL-10 gave full EGL-10 activity, whereas either fragment alone gave little activity. Biochemical analysis showed that coexpressing the two fragments caused both to increase in abundance and also caused the GGL/RGS fragment to move to the membrane, where the N-terminal fragment is localized. By coimmunoprecipitation, we found that the N-terminal fragment complexes with the C-terminal fragment and its associated G subunit, GPB-2. We conclude that the N-terminal region directs inhibition of G signaling by forming a complex with the GGL/RGS region and affecting its stability, membrane localization, and G target specificity. Many hormones and neurotransmitters bind and activate heptahelical transmembrane receptors, which in turn catalyze exchange of bound GDP for GTP on G protein subunits. GTP binding induces dissociation of G from G subunits and enables these activated proteins to signal downstream effectors. Regulator of G protein signaling (RGS) 1 proteins help * This work was supported by a National Science Foundation predoctoral fellowship (to G. A. P.), by a Leukemia and Lymphoma Society Scholar award (to M. R. K.), and by grants from the National Institutes of Health and the Robert Leet and Clara Guthrie Patterson Trust. 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 U.S.C. Section 1734 solely to indicate this fact. This paper is dedicated to the memory of our late colleague Paul Sigler, who provided inspiration and insight during the early stages of this work. To whom correspondence should be addressed: Dept. of Molecular Biophysics and Biochemistry, Yale University School of Medicine, 333 Cedar St., SHM C-E30, New Haven, CT Tel.: ; Fax: ; michael.koelle@yale.edu. 1 The abbreviations used are: RGS, regulator of G protein signaling; GGL, G -like; DEP, Dishevelled/EGL-10/pleckstrin homology; RGS-N, conserved N-terminal region found in GGL-containing RGS proteins; EAT-N, -L, and -C, N-terminal, linker, and C-terminal region of EAT- 16, respectively; EGL-N, -L, and -C, N-terminal, linker, and C-terminal terminate signaling by greatly enhancing the weak intrinsic GTPase activity of G proteins, thus driving reassembly of the inactive, GDP-bound G heterotrimer (1). All RGS proteins contain a 120-amino acid region known as the RGS domain that contacts G subunits and functions as the GTPase activation domain (2). To date, more than 20 mammalian RGS proteins and about 20 mammalian G proteins have been identified (1). Although RGS proteins have been extensively studied in vitro and by overexpression in cultured cells, their in vivo G targets and physiological functions remain largely unclear. For example, the RGS proteins RGS4 and GAIP act similarly on both G i and G o in vitro (3) but show strong and opposing selectivity between these targets when assayed in cultured chick sensory neurons (4). However, when overexpressed in HEK293 cells, both RGS proteins acted similarly to block G i signaling (5). Another dilemma is illustrated by the fact that the RGS2 protein shows different preferences for G i versus G q, depending on the specific in vitro assay system used (6, 7). An additional complication arises from the fact that many studies have been carried out on small RGS domain-containing fragments of RGS proteins. RGS proteins typically contain additional conserved regions outside the RGS domain that appear to affect their functions and target specificities (1). Genetic studies have the potential to conclusively demonstrate the true physiological functions and G target specificities of RGS proteins. To date, only a few RGS proteins have been genetically characterized. A mouse knockout of RGS9 shows that it functions in the retina to limit the duration of visual signaling (8). RGS9 is the only abundant RGS protein in rod outer segments (9), which also contain only one G protein, the visual G protein G t. Thus, the selectivity of the RGS-G interaction in this case appears to be a relatively simple issue. A more complex, and probably more typical, case is revealed by genetic studies of the Caenorhabditis elegans RGS proteins EGL-10 and EAT-16. Although both are widely expressed in the nervous system, they select different G targets and thus have opposite effects on C. elegans behavior (Fig. 1C). Genetic experiments have shown that EGL-10 inhibits signaling by the C. elegans G o homolog GOA-1, which in turn inhibits C. elegans egg-laying behavior (10). EAT-16, on the other hand, inhibits signaling by the C. elegans G q homolog EGL-30, which in turn activates egg-laying behavior (11). EGL-10 and EAT-16 constitute the clearest example of RGS proteins that have been shown through rigorous genetic experiments to have distinct G target specificities. EGL-10 and EAT-16 are members of a subfamily of RGS proteins that includes RGS9 as well as the mammalian RGS6, RGS7, and RGS11 proteins. Just N-terminal to their RGS region of EGL-10, respectively; FL-N, FLAG epitope-tagged EGL-10 N-terminal fragment; HA, hemagglutinin; HA-C, HA epitope-tagged EGL-10 C-terminal fragment. This paper is available on line at

2 An RGS N-terminal Region Directs Inhibition of G Signaling domains, these proteins contain a 60-amino acid G -like (GGL) domain (Fig. 1, A and B) that mediates binding to a divergent G subunit, G 5 (12). The exact role of G 5 in signaling is unclear, but genetic and molecular studies in C. elegans show that EGL-10 and EAT-16 require association with the C. elegans G 5 ortholog, GPB-2, for their stability and function (13). All GGL-containing RGS proteins also contain a conserved N-terminal region of 220 amino acids (Fig. 1A) that has been termed the RGS-N domain (9). The significance of this region is indicated by its extraordinary conservation (e.g. 69% identity, comparing EGL-10 and RGS7; Fig. 1B). The function of the RGS-N region is not clearly understood. It contains an amino acid subregion known as the DEP domain, named for its weak sequence similarity (10 19% identity), comparing the Dishevelled, EGL-10, and pleckstrin proteins (14). Two studies analyzed the stimulation of G t GTPase activity by N-terminal truncation mutants of RGS9 deleted for the RGS-N region. One study saw little effect of the RGS-N region (15). The other (16), using a different method of kinetic analysis, found that the RGS-N region contributed to the ability of RGS9 to act preferentially on G t when G t is complexed with the subunit of its effector, phosphodiesterase. Recent studies of RGS9 have also shown that its RGS-N region binds an anchoring protein that tethers it to the membrane (17, 18). In this study, we show that the RGS-N regions of EGL-10 and EAT-16 contribute to the functions and distinct G target specificities of these RGS proteins in vivo. Using EGL-10/ EAT-16 chimeras, we found that the manner in which the RGS-N and GGL/RGS regions are linked influences whether G o or G q is selected for inhibition. Most intriguingly, we found that the RGS-N region can direct G inhibition by the GGL/RGS region even when these regions are expressed as separate polypeptides. This observation can be explained by our finding that the RGS-N region directly or indirectly binds to the GGL/RGS region and its associated G subunit GPB-2, and the complex thus formed appears to be the functional unit that acts on G targets in vivo. EXPERIMENTAL PROCEDURES Plasmids for Neural Expression of RGS Proteins A 2.2-kb PstI/ PflMI fragment of the rgs-1 promoter (19) was inserted into ppd49.26 (20) to construct pgp3, a vector that directs expression of inserted open reading frames in all neurons of C. elegans. cdna fragments coding full-length EGL-10 or EAT-16 were inserted into pgp3 to generate plasmids pgp4 and pmk340, respectively. Conserved protein subregions were identified in alignments of EGL-10 and EAT-16 with each other and their human homologs (Fig. 1B), and cdna fragments coding some of these regions were inserted into pgp3. For expression of C- terminal fragments of EGL-10 or EAT-16, an artificial AUG codon was added to initiate translation. The extents of the EGL-10 subregions, the names designating them, and the corresponding expression plasmids were as follows: residues 1 223, EGL-N, pgp39; residues , EGL-L (not expressed individually); and residues , EGL-C, pgp40. The subregions, designations, and plasmids for EAT-16 were as follows: residues 1 201, EAT-N, pgp51; residues , EAT-L (not expressed individually); and residues , EAT-C, pgp52. An alternative EGL-N fragment, consisting of residues of EGL-10, was also tested: it behaved in all respects like the smaller 223-residue EGL-N fragment (data not shown). The FL-N construct pgp91 was made by adding sequences encoding DYKDDDDKDYKDDDDK (containing two FLAG tags) immediately after the initiating AUG codon of the EGL-N construct pgp39. Similarly, the HA-C construct pgp92 was made by adding sequences encoding GYPYDVPDYAGYPYDVPDYA (containing two HA tags) after the AUG codon of the EGL-C construct pgp40. Derivatives of pgp3 were also constructed to express EGL-10/ EAT-16 chimeras composed of the N, L, and C subregions described above. The chimeras used and the corresponding expression plasmids were as follows: EGL-N/EGL-L/EAT-C, pgp55; EGL-N/EAT-L/EAT-C, pgp58; EGL-N/EAT-L/EGL-C, pgp59. Transgenic Animals Test plasmids were coinjected with a marker plasmid into the gonad of C. elegans to generate extrachromosomal FIG. 1.RGS proteins with opposing effects on C. elegans egg laying share several conserved regions. A, schematic representation of the 555-amino acid EGL-10 protein. An N-terminal conserved region of unknown function (RGS-N; gray box) contains a subregion known as the DEP domain. A poorly conserved linker region (white box) is followed by a GGL domain (hatched box) and an RGS domain (black box). In this work, we analyze N- and C-terminal fragments of EGL-10 indicated by bars under the schematic diagram. B, alignment of the EGL-10, human RGS7 (hrgs7), and EAT-16 RGS proteins. Gray, hatched, and black shaded bars (top) overlie sequences corresponding to the gray, hatched, and black shaded regions in the EGL-10 schematic in A. Amino acids identical in two or three of the sequences are shaded black. C, schematic representation of the opposing G protein signaling pathways that control egg-laying behavior in C. elegans. Signaling through cell surface receptors activates the G o and G q proteins (known in C. elegans as GOA-1 and EGL-30, respectively). GOA-1 inhibits egg laying, whereas EGL-30 activates this behavior. The RGS proteins EGL-10 and EAT-16 each exist as obligate dimers with the G 5 ortholog, GPB-2. EGL-10 inhibits GOA-1 activity, whereas EAT-16 inhibits EGL-30. Thus EGL-10 and EAT-16 have opposite effects on egg-laying behavior. transgene arrays (20). Animals from transgenic lines were recognized among the F2 progeny because they showed the phenotype induced by the marker plasmid. Rescue of egl-10 was tested by injection into animals of genotype egl-10(md176); lin-15(n765). The lin-15 rescuing

3 47006 An RGS N-terminal Region Directs Inhibition of G Signaling plasmid pl15ek (11) was coinjected at 50 ng/ l as the marker. Rescue of eat-16 was tested by injection into animals of genotype eat-16(ad702), and the marker plasmid prf4 (20), which induces a dominant Rol phenotype, was coinjected at 80 ng/ l as the marker. To test for dominant negative effects, injections were into lin-15(n765) animals carrying no RGS mutations, and pl15ek was used as the marker plasmid. Each RGS expression plasmid tested was injected at 80 ng/ l, and pgp3 (empty vector) DNA was included in certain injections so that the total expression plasmid DNA concentration was identical for every injection. Extrachromosomal transgenes expressing FLAG- and HA epitopetagged EGL-10 fragments were chromosomally integrated by irradiating transgenic animals with -rays. The resulting strains were outcrossed to egl-10(md176); lin-15(n765) animals at least two times to produce clean genetic backgrounds. The integrated FL-N and HA-C transgenes shown in this work have the allele designations vsis20 and vsis18, respectively. vsis18 male animals were mated to vsis20 hermaphrodites, and the resulting cross-progeny were allowed to selffertilize to produce animals homozygous for both vsis20 and vsis18. The genotypes of all of these strains were verified by PCR amplification of the transgenes. Behavioral Assays Unlaid eggs were counted by dissolving adult animals in bleach and counting the bleach-resistant fertilized eggs under a microscope (10). All assays were on animals selected as late L4 larvae and aged at 20 C for 30 h to produce precisely staged adults. For each extrachromosomal transgene analyzed, at least 50 animals were assayed ( 10 animals from at least five independent transgenic lines). For integrated transgenes, 30 animals were assayed. In certain cases, we also carried out a second egg-laying assay in which the developmental stages of freshly laid eggs were determined (10). In each case, this verified that the transgenes affected egg-laying behavior, not egg production, indicating that the transgenes generated normal EGL-10 and EAT-16 activities (data not shown). C. elegans Protein Extracts Worm strains carrying the egl- 10(md176) null mutation as well as the integrated transgene vsis18 and/or vsis20 were grown in liquid culture at 20 C as mixed stage populations. Worms were purified by flotation on 30% sucrose and transferred to lysis buffer (50 mm Tris, ph 7.4, 1 mm EDTA, 1 mm EGTA, 5 mm MgCl 2, 150 mm NaCl, protease inhibitors, and, in certain experiments, 1% Triton X-100). Lysis was by three passages through a French press followed by 60 s of sonication with a microtip probe (Fisher model 550 sonic dismembrator). Debris and unlysed worms were removed by centrifugation at 2,000 rpm in a clinical centrifuge. The resulting total lysates were flash-frozen in liquid nitrogen and stored at 80 C. Protein concentrations were determined by Bradford analysis. When required, total lysates were fractionated into soluble and pellet fractions by centrifugation at 100,000 g for 30 min. To assess the levels of FLAG- and HA-tagged proteins in total lysates (see Fig. 6), 50 and 160 g of total protein, respectively, were fractionated by SDS- PAGE. We further analyzed total lysates by sucrose density gradient centrifugation. Sucrose gradients were formed by successively overlaying a 49% (w/v) sucrose cushion with equal volumes of 20% sucrose and total lysate. Centrifugation was carried out in a TLS-55 swinging bucket rotor at 55,000 rpm for 2 h at 4 C. Twelve equal volume fractions were collected, with fraction 1 at the top and fraction 12 (including any pellet fraction) at the bottom. The results shown in Figs. 6 and 7 are representative of those obtained in multiple trials carried out with independently prepared lysates. Western Blotting Proteins were fractionated by SDS-PAGE and then transferred to nitrocellulose filters. The primary antibodies used to probe Western blots were as follows: mouse anti-flag (M2; Sigma); rat anti-ha high affinity (3F10; Roche Molecular Biochemicals); mouse anti- -tubulin (E7; developed by Michael Klymkowsky and obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Department of Biological Sciences); rabbit anti- GPB-2 (13); and rabbit anti-goa-1 (13). The secondary antibodies were horseradish peroxidase-coupled goat anti-mouse (Bio-Rad), goat anti-rabbit (Bio-Rad), or goat anti-rat (Pierce). Protein bands were visualized using chemiluminescence detection reagents (Pierce) and Eastman Kodak Co. BioMax MR film. The proteins studied had mobilities on SDS-PAGE analysis approximately as follows: HA-C, 28 kda; FL-N, 32 kda; tubulin, 55 kda; GPB-2 isoforms, 42 and 44 kda; and GOA-1, 40 kda. Immunoprecipitation Soluble fractions of extracts made in lysis buffer with 1% Triton X-100 were incubated with anti-flag M2 antibodies coupled to agarose beads (Sigma), tumbling for 2 h at 4 C. The beads were washed three times in lysis buffer containing 1% Triton X-100 and pelleted by centrifugation. Proteins bound to the pelleted beads were eluted in 100 mm glycine, ph 3.5, for 5 min with gentle shaking at room temperature and neutralized with a 10% elution volume of 0.5 M Tris, ph 7.4, 1.5 M NaCl. The results shown in Fig. 8 are representative of those obtained in at least three trials. RESULTS EGL-10 and EAT-16 Have Opposite Effects Even When Expressed from the Same Promoter EGL-10 and EAT-16 have precisely opposite mutant phenotypes, because EGL-10 specifically inhibits G o and EAT-16 specifically inhibits G q. In principle, these RGS proteins might achieve such specificity by being expressed in different cells that also express different G proteins. Alternatively, the RGS and G proteins might all be found in the same cells, but each RGS protein could have the ability to selectively act on only one of the G targets available to it. This latter possibility is supported by the observation that both RGS and both G proteins are expressed in all of the neurons of C. elegans, although each RGS and G protein is also additionally expressed in muscle cells that may differ for each protein (10, 11, 21 23). To distinguish between these alternative models, we constructed transgenes in which the same promoter was used to direct expression of either the EGL-10 or the EAT-16 cdna. We used the rgs-1 promoter, which is active in all neurons of C. elegans but in no other cells (19). The constructs were transformed into egl-10 or eat-16 null mutants and tested for their ability to rescue the opposite defects in egg-laying behavior seen in the two mutants. egl-10 mutants are lethargic in their egg-laying behavior and thus accumulate excess unlaid eggs compared with control animals (Fig. 2, compare A and B). Transgenic expression of EGL-10 rescued the egl-10 egg-laying defect (Fig. 2C). In contrast, expression of EAT-16 from the same heterologous promoter did not affect the egl-10 mutant (Fig. 2D). We quantitated our results by counting the number of unlaid eggs retained inside animals carrying the different transgenes (Fig. 2E). The egl-10 transgene rescued the egl-10 defect fully and even resulted in slightly hyperactive egg laying, seen as a decrease in the accumulation of unlaid eggs relative to the wild-type control (Fig. 2E, compare bars 1 and 3). Overexpressing EGL-10 has previously been shown to cause a gain-offunction effect that results in hyperactive egg laying (10). The EAT-16 protein had no detectable EGL-10 activity, since expression of EAT-16 did not reduce the accumulation of unlaid eggs (Fig. 2E, bar 4). eat-16 mutants are hyperactive in their egg-laying behavior (11). Whereas wild-type animals retain their fertilized eggs for 2hormore before laying them, eat-16 mutants lay their eggs very soon after fertilization, resulting in a decreased accumulation of unlaid eggs (Fig. 2F, compare bars 1 and 2). Transgenic expression of EAT-16 resulted in substantial rescue of the eat-16 egg-laying defect (Fig. 2F, bar 4), whereas expression of EGL-10 using the same promoter resulted in no detectable rescue of the eat-16 defect (Fig. 2F, bar 3). The rescue resulting from EAT-16 expression may have been incomplete either due to insufficient levels of expression from the heterologous promoter or because EAT-16 expression may be required outside of the nervous system to achieve full rescue. These experiments demonstrate that, even when expressed from the same heterologous promoter, EGL-10 and EAT-16 have opposite effects on egg laying. Neural expression of either RGS protein can fully (EGL-10) or substantially (EAT-16) rescue loss of endogenous expression of the same RGS protein but cannot rescue loss of the other RGS protein at all. We conclude that EGL-10 and EAT-16 have distinct effects on egg laying due to distinct properties of these RGS proteins themselves and that any small differences in their endogenous expression patterns do not account for their different functions.

4 An RGS N-terminal Region Directs Inhibition of G Signaling FIG. 2.Effects of transgenic expression of EGL-10 and EAT-16, using a heterologous neural promoter, in egl-10 and eat-16 mutant animals. A, control adult hermaphrodite that carries an empty vector transgene. B, egl-10 mutant carrying an empty vector transgene. C, egl-10 mutant carrying a transgene expressing EGL-10 in all neurons using a heterologous promoter. D, egl-10 mutant carrying a transgene expressing EAT-16 using the same promoter. The arrows in A D indicate individual unlaid fertilized eggs inside the adults. E, phenotypes shown in panels A D were quantitated by counting the number of unlaid eggs inside 50 animals for each genotype. RGS mutant backgrounds and transgenes are indicated below the graph, and bars 1 4 correspond to the genotypes shown in A D, respectively. The egl-10 mutant phenotype, seen as an accumulation of unlaid eggs, is rescued by expression of EGL-10 (bar 3) but not by expression of EAT-16 (bar 4). Error bars indicate the 95% confidence interval of the mean. F, quantitation of an experiment analogous to that shown in E but examining rescue of eat-16 rather than of egl-10. The eat-16 mutant phenotype, seen as a decrease in the number of unlaid eggs retained in adults, is substantially rescued by expression of EAT-16 (bar 4) but not by expression of EGL-10 (bar 3). Note the change of scale from E and that the control (bar 1) is the same data as shown in E (bar 1). Control data from this figure are also replotted in subsequent figures for purposes of comparison. The mutations used in this and subsequent figures were egl-10(md176) and eat-16(ad702). Each is an apparent null mutation (10, 11). The EGL-10 N- and C-terminal Fragments Need Not Be Covalently Attached for EGL-10 Function We tested the ability of subregions of EGL-10 and EAT-16 to function in vivo. The fact that the linker between the RGS-N region and the C- terminal GGL/RGS region is variable in length and sequence among different RGS proteins (Fig. 1B) suggested that a precise attachment between these two regions may not be required and encouraged us to try expressing the two regions as separate polypeptides in transgenic animals to test for their function. The same neuron-specific promoter used above for expressing full-length RGS proteins was also used to express the protein subfragments. In addition, we tested coexpression of these fragments by cotransforming expression constructs for each. The expression constructs were transformed into egl-10 or eat-16 null mutants and tested for their ability to rescue the egg-laying defects of these mutants. The constructs were also transformed into animals with no RGS mutations to test for dominant-negative effects on egg laying. Experiments in which protein fragments are expressed suffer from the problem that such fragments may not be properly folded or stable, making negative results difficult to interpret. We set a stringent criterion for dealing with this issue: we do not present or interpret results from constructs that give only negative results. However, if a construct gives strong rescuing activity in one experiment, we do interpret negative results it may give in other experiments, because the positive result demonstrates that the construct successfully expresses an active protein fragment. Fig. 3A shows the effects of EGL-10 fragment expression in an egl-10 null mutant background. We refer to the N- and C-terminal fragments of EGL-10 (indicated in Fig. 1A) as EGL-N and EGL-C, respectively. Expression of either EGL-N or EGL-C only weakly rescued the egl-10 egg-laying defect (Fig. 3A, bars 3 and 4). Surprisingly, coexpression of both fragments gave full rescue of the egl-10 mutant (Fig. 3A, bar 5), equivalent to the strong rescue previously seen by expressing full-length EGL-10 (Fig. 2E, bar 3). This result demonstrates that EGL-N and EGL-C act together to inhibit G o signaling and that EGL-N need not be covalently attached to EGL-C for full EGL-10 function. We tested the EGL-10 fragments for EAT-16 activity by expressing them in the eat-16 null mutant but found that neither fragment nor the combination of both showed much activity (Fig. 3B). We also tested the EGL-10 fragments in a background carrying no RGS mutations to test the fragments for dominant negative effects (Fig. 3C). Interference with endogenous EGL-10 function would be seen as an increased accumulation of unlaid eggs. Expression of EGL-N alone (Fig. 3C, bar 2) gave a result similar to the control (bar 1). Expression of EGL-C alone or coexpression of EGL-C and EGL-N also did not show dominant negative effects. Rather, expression of these fragments resulted in small decreases in the accumulation of unlaid eggs (Fig. 3C, bars 3 and 4), showing that these constructs gave positive EGL-10 activity, just as they did in the egl-10 mutant background (Fig. 3A, bars 4 and 5). We also tested constructs expressing N- and C-terminal fragments of EAT-16 (termed EAT-N and EAT-C, respectively) in egl-10, eat-16, and wild-type RGS backgrounds. Neither fragment nor the combination of both showed strong effects in any background tested (data not shown). According to our criterion for interpretability outlined above, we therefore do not interpret these experiments. Although we know that EAT-C is active (see below), EAT-N may simply not be folded or stable. The RGS-N Region of EGL-10 Can Direct the GGL/RGS Region of EAT-16 to Have Full EGL-10 Activity To identify the regions of EGL-10 and EAT-16 responsible for their distinct G target specificities, we generated transgenes to express EGL-10/EAT-16 chimeras, transformed them into egl-10 and eat-16 null mutants, and tested for their ability to rescue the egg-laying defects of these mutants. In these experiments, we hoped to identify discrete protein subregion(s) that determine EGL-10 activity versus EAT-16 activity. Our first strategy for generating chimeras was based on the observation that the N- and C-terminal fragments of EGL-10, when coexpressed, give full EGL-10 function. We coexpressed combinations of N- and C-terminal fragments from EGL-10 and EAT-16 to see if these combinations, which we term noncova-

5 47008 An RGS N-terminal Region Directs Inhibition of G Signaling FIG. 3.Effects of expressing EGL-10 protein fragments in egl-10, eat-16, and wild-type animals. The fragments expressed were as indicated in Fig. 1A: EGL-N, an N-terminal fragment consisting of the RGS-N domain (indicated in the figure as N); EGL-C, a C-terminal fragment containing the GGL and RGS domains (C); or coexpression of both fragments (N C). A, effects of EGL-10 protein fragment expression in the egl-10 null mutant background. Individual expression of EGL-N (bar 3) or EGL-C (bar 4) showed only partial rescue of the egl-10 egg-laying defect, whereas coexpression of both fragments showed full rescue (bar 5). B, expression of the same EGL-10 fragments as shown in A, but in an eat-16 mutant background. None of the EGL-10 fragments showed significant rescue of the eat-16 egg-laying defect. C, effects of EGL-10 protein fragment expression in a wild-type RGS background. No significant dominant-negative effects were observed. The EGL-C transgene (bar 3) and the coexpression of EGL-N and EGL-C (bar 4) both showed some positive egl-10 activity as evidenced by decreases in the number of unlaid eggs relative to the control (bar 1). lent chimeras, could give EGL-10 or EAT-16 function. We found that coexpression of the EGL-N fragment with the C-terminal GGL/RGS fragment of EAT-16 (EAT-C) gave full EGL-10 function. As shown in Fig. 4A, expression of either fragment alone in an egl-10 null background gave little rescue of the egl-10 egg-laying defect (bars 3 and 4). When coexpressed, however, the EGL-N and EAT-C fragments fully rescued the egl-10 mutant (bar 5). Indeed, the number of unlaid eggs in the animals coexpressing the fragments was below that of the wild-type control (bar 1), indicating that animals coexpressing EGL-N and EAT-C have excess EGL-10 activity and thus are slightly hyperactive for egg laying. Expression of EGL-N, EAT-C, or both gave no rescue of the eat-16 egg-laying defect (Fig. 4B, bars 3 5), showing that these fragments have no detectable eat-16 activity. We also coexpressed the EAT-N fragment with the C-terminal fragment of EGL-10 but found that this noncovalent chimera was unable to rescue either egl-10 or eat-16 mutants (data not shown). As discussed above, we have no evidence that the EAT-N fragment was successfully expressed and folded in these experiments and thus do not interpret these negative results. Our main finding from the use of noncovalent chimeras was that expression of the EGL-N and EAT-C fragments resulted in full EGL-10 activity and gave results essentially identical to those seen when EGL-N and EGL-C were coexpressed (Fig. 3). In the context of these experiments, the EGL-C and EAT-C fragments are thus equivalent and interchangeable. Whereas the GGL/RGS region of EAT-16 would normally act to inhibit G q signaling, when this region of EAT-16 is coexpressed with the RGS-N region of EGL-10, it is instead apparently directed to inhibit G o signaling. The Relationship between the RGS-N and GGL/RGS Regions Directs G Target Specificity To refine our understanding of G target specification, we generated transgenes that express full-length chimeric RGS proteins as single polypeptides. We term these covalent chimeras to distinguish them from the noncovalent chimeras described above. In designing these chimeras, we divided EGL-10 and EAT-16 into three subregions: 1) the N-terminal RGS-N domain; 2) the linker region between the RGS-N domain and the GGL domain; and 3) a C-terminal region consisting of the GGL and RGS domains and residues C-terminal to them (see Fig. 1A). We generated the complete set of six chimeras in which each of the three subregions was derived from either EGL-10 or EAT-16 in every possible combination. For example, the EGL-N/EAT-L/EAT-C chimera contains the RGS-N domain of EGL-10, followed by the linker from EAT-16 and finally the C-terminal GGL/RGS region from EAT-16. The chimeras were expressed using the same neuron-specific promoter employed in the experiments described above. As before, we present data only from chimeras that gave strong rescuing activity in either the egl-10 or eat-16 backgrounds. Purely negative results were obtained from the three chimeras containing the EAT-N region, and these results were considered uninterpretable. A key factor determining whether a chimera had EGL-10 activity or EAT-16 activity was the manner in which the RGS-N and C-terminal regions were linked. For example, the EGL-N/EGL-L/EAT-C chimera gave strong EGL-10 activity (Fig. 4A, bar 6) and little EAT-16 activity (Fig. 4B, bar 6). In this chimera, the EGL-N region apparently directs the EAT-C region to have EGL-10 activity, just as occurred when these two regions were coexpressed as separate polypeptides (Fig. 4, A and B, bars 5). However, swapping from the EGL-10 to the EAT-16 linker caused a switch from EGL-10 activity to EAT-16 activity. This is seen in the EGL-N/EAT-L/EAT-C chimera, which has strong eat-16 rescuing activity (Fig. 4B, bar 7) but little egl-10 rescuing activity (Fig. 4A, bar 7). The linker region is not the sole determinant of G target specificity. If this were true, the EGL-N/EAT-L/EGL-C chimera would be expected to show EAT-16 activity rather than EGL-10 activity. Instead, this chimera shows strong EGL-10 activity (Fig. 4A, bar 8) and little EAT-16 activity (Fig. 4B, bar 8). This chimera demonstrates that the C-terminal GGL/RGS region also contributes to G target specificity, since the EGL-10 activity of this chimera can be converted to EAT-16 activity by swapping the C-terminal region with that of EAT-16 (Fig. 4, A and B, compare bars 7 and 8). In summary, analysis of EGL-10 and EAT-16 transgenes shows that the C-terminal GGL/RGS regions of these proteins

6 An RGS N-terminal Region Directs Inhibition of G Signaling FIG. 5. Effects of chromosomally integrated transgenes expressing epitope-tagged EGL-10 fragments in egl-10 mutant animals. The fragments expressed were as follows: FL-N, a FLAG epitope-tagged N-terminal fragment consisting of the RGS-N region, and HA-C, an HA epitope-tagged C-terminal fragment containing the GGL and RGS domains. The two transgenes expressing FL-N or HA-C were chromosomally integrated to give stable expression, and the strain expressing both FL-N and HA-C was generated by genetically crossing strains carrying the individual transgenes (FL-N HA-C). The results shown are similar to those in Fig. 3A, indicating that the epitope tags used do not interfere with the function of the EGL-10 fragments and reproducing the finding that coexpression of both N- and C-terminal fragments is required for full EGL-10 rescuing activity. FIG. 4.Effects of expressing EGL-10/EAT-16 chimeric proteins in egl-10 and eat-16 mutant animals. A, effects of chimera expression in the egl-10 null mutant background. B, effects of chimera expression in the eat-16 mutant background. Egg-laying behavior was quantitated by counting unlaid eggs. The three components of the RGS proteins were as indicated in Fig. 1A: an N-terminal subregion consisting of the RGS-N domain (denoted as N); a linker subregion (L); and a C-terminal subregion containing the GGL and RGS domains (C). The label for each subregion is printed under the graph on a gray or black background to indicate that it is derived from EGL-10 or EAT-16, respectively. Bars 3 5 in A and B show the results of expressing the EGL-10 N-terminal fragment and EAT-16 C-terminal fragment as separate polypeptides in the absence of any linker region. Bars 6 8 show the results of expressing individual chimeric polypeptides in which regions from the two proteins were covalently linked. Coexpression of the EGL-10 N-terminal fragment and EAT-16 C-terminal fragment (bars 5) gave strong EGL-10 activity, whereas the EGL-N/EAT-L/ EAT-C chimeric protein (bars 7) gave substantial EAT-16 activity. Other covalent chimeras (bars 6 and 8) showed partial EGL-10 activity. are not sufficient for in vivo function but require a N-terminal conserved region to gain full function in vivo. Surprisingly, the two protein regions need not be covalently linked to function together. However, if they are covalently linked, the manner in which they are attached can determine which G protein is selected as a target. Functional Epitope-tagged EGL-10 N- and C-Terminal Fragments for Biochemical Analysis of Their Interactions Perhaps the most intriguing result from our transgenic experiments is our finding that neither the EGL-N nor EGL-C fragments are able to rescue egl-10 mutant animals but that coexpression of these fragments as separate polypeptides does give full EGL-10 function (Fig. 3A). For the remainder of this work, we present a biochemical analysis of these two protein fragments and their interactions in extracts of transgenic animals. We modified our original EGL-10 N- and C-terminal fragments to include FLAG and HA epitope tags, respectively. We refer to these modified fragments as FL-N and HA-C. We generated transgenes expressing these fragments in C. elegans and chromosomally integrated the transgenes to produce stable transgenic lines that could be grown in biochemical quantities. An additional benefit of the integrated transgenes is that we could genetically cross the two strains carrying the individual FL-N and HA-C transgenes to generate a strain carrying both transgenes. Thus, proteins expressed from the exact same FL-N or HA-C transgenes could be compared when expressed alone or in combination. Before analyzing the tagged proteins biochemically, we checked their function in vivo by testing their ability to rescue egl-10 mutant animals. We found that the tagged fragments (Fig. 5) behaved similarly to their untagged counterparts (Fig. 3A). We note one subtle but interesting difference; whereas the untagged N transgene appeared to have a small amount of egl-10 rescuing activity (Fig. 3A, bar 3), the FL-N transgene (perhaps due to a lower expression level) had no such detectable activity (Fig. 5, bar 3). Nevertheless, the FL-N transgene, when combined with a partially rescuing C-terminal transgene, could give rise to full egl-10 rescuing activity (Fig. 5, bar 5). These results suggest that the cooperation between N- and C-terminal fragments may be synergistic rather than merely additive. FL-N and HA-C Proteins Show Increased Abundance When Coexpressed We carried out Western analyses of total worm extracts from strains carrying the same FL-N and HA-C transgenes separately or together. Our results showed that the FL-N and HA-C proteins were both increased in abundance when coexpressed with each other (Fig. 6). Blots carrying different loadings of the samples to achieve equivalent signals revealed about a 2-fold enhancement in the levels of FL-N and HA-C when coexpressed (data not shown). Since the HA-C protein by itself showed partial rescue of the egl-10 mutant (Fig. 5, bar 4),

7 47010 An RGS N-terminal Region Directs Inhibition of G Signaling FIG. 6.Western blot analysis of EGL-10 protein fragment expression in transgenic animals. Total worm lysates were separated by SDS-PAGE, transferred to nitrocellulose filters, and immunoblotted with anti-ha, anti-flag, or anti-tubulin antibodies. Combining the HA-C and FL-N transgenes by genetically crossing the strains carrying them resulted in increases in the levels of both the HA-C and FL-N proteins. To control for loading, the same blots probed with anti-flag and anti-ha antibodies were also probed with an anti-tubulin antibody. the increase in HA-C protein upon coexpression with FL-N could, at least in part, explain the full rescue observed when these proteins were coexpressed (Fig. 5, bar 5). One way the FL-N and HA-C proteins might increase each other s abundance is by forming a complex and thus stabilizing each other. In order to test this hypothesis by coimmunoprecipitation, we first searched for conditions that could solubilize the FL-N and HA-C proteins for precipitation. Coexpression of FL-N with HA-C Decreases the Solubility of HA-C, and Both Proteins Can Be Solubilized with Triton X-100 We fractionated total worm extracts by 100,000 g centrifugation into soluble and pellet fractions in the presence or absence of detergent. We used Western blots to determine the solubility of the FL-N and HA-C proteins when expressed separately or together (Fig. 7A). In the absence of detergent, HA-C expressed alone was more than 50% soluble. However, when coexpressed with FL-N, almost all of the HA-C protein moved to the pellet fraction. This insoluble HA-C could be substantially solubilized by the addition of 1% Triton X-100 (compare the top two panels in Fig. 7A). In contrast to HA-C, FL-N, whether expressed alone or in the presence of HA-C, was almost entirely in the pellet fraction. The addition of 1% Triton X-100 also resulted in solubilization of a significant amount of FL-N (compare bottom two panels in Fig. 7A). To distinguish whether the insolubility of FL-N and HA-C observed in the absence of detergent was due to membrane association or due to the formation of particulate/aggregate structures, we carried out sucrose density gradient centrifugation on total lysates containing these proteins. Our results revealed that the bulk of the insoluble FL-N and HA-C floated in fractions 8 and 9 of these gradients, which comprised the 20%/49% sucrose interface, where membrane-associated proteins typically reside (top two panels in Fig. 7B). No FL-N or HA-C was detected in fraction 12, which included the pellet where any aggregates should be found. Western analysis of the same sucrose gradients showed that fractions 8 and 9, in addition to containing FL-N and HA-C, also contained endogenous membrane-associated GOA-1, the G target of EGL-10 (lower panel in Fig. 7B). FL-N and HA-C floated rather than pelleted in sucrose gradients regardless of whether these two FIG. 7. Solubility and effect of detergent on HA-C and FL-N expressed in transgenic animals. A, total worm lysates (T), soluble fractions generated as supernatants from 100,000 g centrifugation of the same lysates (S), and the insoluble pellet fractions (P) were separated by SDS-PAGE, transferred to nitrocellulose filters, and immunoblotted with anti-ha and anti-flag antibodies. This experiment was conducted either in the presence or absence of 1% Triton X-100 in the lysis buffer. The HA-C protein was largely soluble when expressed alone but moved almost entirely to the pellet when coexpressed with the FL-N protein. In contrast, the distribution of the FL-N protein remained unchanged in the presence or absence of the HA-C protein, remaining predominantly in the pellet fraction. The use of 1% Triton resulted in significant solubilization of both the HA-C and FL-N proteins. B, sucrose density gradient fractionation of total lysates prepared in the absence of detergent from worms carrying integrated transgenes expressing HA-C and FL-N. The bulk of these proteins floated in fractions 8 and 9, as did endogenous membrane-associated GOA-1. proteins were coexpressed (Fig. 7B) or expressed individually (data not shown). FL-N and HA-C therefore appear to be membrane-associated, and their insolubility is not due to aggregation. Since the RGS domain found in the EGL-10 C terminus is believed to act directly on the membrane-associated G protein GOA-1, the apparent increase in HA-C membrane localization upon coexpression with FL-N could explain, at least in part, the increase from partial to full egl-10 rescuing activity that occurs when HA-C is coexpressed with FL-N. The effect of FL-N on the solubility of HA-C suggests that FL-N may form a complex with HA-C and thereby affect not only its stability but also its membrane localization. Using the 1% Triton solubilization conditions shown in Fig. 7A, we have tested this hypothesis by coimmunoprecipitation. FL-N Forms a Complex with HA-C and Its Associated G Subunit To determine whether FL-N and HA-C form a complex, we immunoprecipitated FL-N from Triton-solubilized extracts of worms expressing both FL-N and HA-C and tested for coprecipitation of HA-C. As controls, we used extracts of worms expressing only FL-N or HA-C, which should not show any coprecipitated signal. We found that HA-C could be immunoprecipitated by the FLAG antibody only in the presence of FL-N, demonstrating that HA-C forms a complex with FL-N in worm extracts (Fig. 8). We expected that the G subunit, GPB-2, should also be in this complex, since it had previously been shown to be an obligate subunit of EGL-10, presumably via an interaction with the GGL domain (13). Indeed, we found that GPB-2 coimmunoprecipitates with FL-N (Fig. 8, lower panel). Interestingly, GPB-2 forms a complex with FL-N even in the absence of HA-C, suggesting that both the RGS-N and

8 An RGS N-terminal Region Directs Inhibition of G Signaling FIG. 8. Coimmunoprecipitation of the HA-C and FL-N proteins. Triton X-100-solubilized protein extracts from worm strains carrying integrated transgene(s) expressing HA-C and/or FL-N were subjected to immunoprecipitation (IP) with anti-flag antibodies. The pellets, along with control extracts representing 5% of the material used for the immunoprecipitations, were fractionated by SDS-PAGE, transferred to nitrocellulose filters, and immunoblotted with anti-ha, anti- FLAG, or anti-gpb-2 antibodies. Anti-FLAG antibody coprecipitates HA-C only in the presence of FL-N, indicating that the HA-C and FL-N proteins form a complex. GGL regions of EGL-10 may each have an independent ability to bind the G subunit. DISCUSSION An Experimental Approach Focused on the Physiologically Relevant Determinants of RGS-G Specificity The RGS domains of a number of RGS proteins have been found to promiscuously activate the GTPase activities of many G proteins in vitro (1). These puzzling results are at odds with the expectation that different RGS proteins might achieve distinct functions in vivo at least in part by targeting different G proteins. The G target specificity of RGS proteins might be increased in vivo by other cellular proteins or by non-rgs domain regions of RGS proteins themselves. Several studies have compared purified full-length and deleted versions of RGS proteins in in vitro GTPase activation assays to test for effects of regions outside the RGS domain on G target specificity (12, 15, 16, 24). These studies might not identify functions crucial to RGS protein activity in vivo, such as membrane localization or interaction with cellular proteins other than G. An additional problem is that, in such studies, RGS9 1 (the retina-specific isoform of RGS9) was the only RGS protein tested that had a known, physiologically relevant G target identified by genetic studies. RGS9-1 is atypical in that it acts in rod outer segments where G t is the only G protein present at a significant level. Thus, RGS9-1 does not physiologically face the challenge of selecting a G target as do other RGS proteins that are typically expressed in cells containing multiple G proteins. Another experimental approach used to analyze the basis of G target selection by RGS proteins involves expressing fulllength or deleted RGS proteins in cultured cells or Xenopus oocytes and using assays that indirectly measure signaling by G targets (reviewed in Ref. 1). In these studies, the physiologically relevant G targets of the RGS proteins used, again, are typically not known, and the signaling readouts arranged for purposes of the experiments may have no relation to the normal physiological functions of the RGS proteins studied. In contrast, our studies analyze two RGS proteins, EGL-10 and EAT-16, that genetic studies have shown to target two distinct G proteins, G o and G q, respectively. Previous studies showed that these RGS and G proteins are all expressed in the same cells, suggesting that the RGS proteins must actively select from at least two accessible G targets. We demonstrated this more clearly by using transgenes to express EGL-10 and EAT-16 from the same heterologous promoter and showing that they retained their proper G target specificities. Using this same promoter, we have also expressed subregions or chimeras of EGL-10 and EAT-16 and measured their functions in vivo using assays of egg-laying behavior. Importantly, this readout of RGS function measures the normal physiological actions of EGL-10 and EAT-16 on their genetically identified G targets. Our experimental approach also allows us to use extracts of the transgenic animals to biochemically analyze the RGS proteins expressed. Thus, our experimental system is designed to analyze RGS function and G target selectivity in a physiological setting and enables us to correlate in vivo and in vitro data. The RGS-N Region Is Essential for Activity of EGL-10 and EAT-16 and Has a Membrane Targeting Function Studies of mammalian RGS proteins have shown that their GGL/RGS regions can function in vitro as efficient G GTPase activators even when the RGS-N domain has been removed (15, 16). In contrast, our results show that the GGL/RGS regions of either EGL-10 or EAT-16 have relatively little function in vivo unless attached to or coexpressed with an RGS-N region. The differences between these results can be explained, at least in part, by the fact that the in vitro assays were carried out in the absence of membranes, whereas in vivo the RGS-N region functions to target RGS proteins to the membrane, the location of their G targets. By analyzing soluble and membrane fractions of C. elegans extracts, we found that the RGS-N and GGL/RGS regions are each independently targeted to membrane fractions, although less than 50% of the GGL/RGS region, when expressed alone, ended up in the membrane fraction. When complexed with an RGS-N region, however, the GGL/RGS region was almost entirely targeted to the membrane. Our results correlate with studies of RGS9, which also identified membrane targeting functions in both the N- and C termini of this protein. A portion of the RGS-N region contains weak similarity to a region of Dishevelled (the DEP domain ) that serves as a membrane anchor (25). In RGS9, the RGS-N domain binds the protein R9AP, which anchors it to the rod outer segment membrane (18). Lishko et al. (17) showed that this membrane association results in a 70-fold increase in the activity of RGS9 on its G target. In C. elegans, there is no clear homolog of R9AP that could serve as a membrane anchor for EGL-10 and/or EAT-16, and the nature of the membrane attachment of RGS-N domains in C. elegans remains to be elucidated. It is possible that there is a distant homolog of R9AP that does not stand out as statistically significant in homology searches of the C. elegans genome but may serve as a membrane anchor for EGL-10 and/or EAT-16. It is also possible that such a membrane anchor will be uncovered in the future by cloning additional genetically identified genes that affect C. elegans egg-laying behavior. The RGS-N Region Directs G Target Specificity How are RGS proteins directed to a specific G target? Our studies of EGL-10 and EAT-16 as well as recent work on the mammalian members of the same RGS protein subfamily (RGS6, -7, -9, and -11) are beginning to answer this question. Studies of mammalian RGS proteins have shown that the RGS domain is sufficient on its own to act as an efficient G GTPase activator (26). Thus, the RGS domain might have been expected to fully specify the G target. However, the addition of the GGL domain and the G 5 subunit appeared to restrict the specificity of the RGS domain in vitro, making it more selective for G o (12, 24, 27). RGS-G 5 complexes need not always be G o -selective in vivo, since EAT-16 is in a complex with the G 5 -like protein GPB-2 but is a G q -selective regulator (11, 13). Thus, at least in the case of EGL-10 and EAT-16, which both individually bind to GPB-2, specificity appears to be achieved by some means other than association with a G subunit.