Proteins Interacting With Caenorhabditis elegans  Gα Subunits

To identify novel components in heterotrimeric G-protein signalling, we performed an extensive screen for proteins interacting with Caenorhabditis elegans Gα subunits. The genome of C. elegans contains homologues of each of the four mammalian classes of Gα subunits (Gs, Gi/o, Gq and G12), and 17 other Gα subunits. We tested 19 of the GGα subunits and four constitutively activated Gα subunits in a largescale yeast two-hybrid experiment. This resulted in the identification of 24 clones, representing 11 different proteins that interact with four different Gα subunits. This set includes C. elegans orthologues of known interactors of Gα subunits, such as AGS3 (LGN/PINS), CalNuc and Rap1Gap, but also novel proteins, including two members of the nuclear receptor super family and a homologue of human haspin (germ cell-specific kinase). All interactions were found to be unique for a specific Gα subunit but variable for the activation status of the Gα subunit. We used expression pattern and RNA interference analysis of the G-protein interactors in an attempt to substantiate the biological relevance of the observed interactions. Furthermore, by means of a membrane recruitment assay, we found evidence that GPA-7 and the nuclear receptor NHR-22 can interact in the animal.


Introduction
Heterotrimeric G-proteins, consisting of α, β and γ -subunits, are signal transduction molecules that couple ligand-bound seven transmembrane (7-TM) receptors to a wide variety of intracellular second messenger systems (Kaziro et al., 1991;Simon et al., 1991). However, there is a growing body of evidence that 7-TM receptors can also transmit extracellular signals through mechanisms that function independently of G-protein coupling (Brzostowski and Kimmel, 2001). In addition, nonreceptor modulators and receptor-independent activators of heterotrimeric G-proteins have been identified Takesono et al., 1999).
All higher eukaryotic organisms possess Gprotein-coupled signal transduction, which has been implicated in many important biological functions, ranging from photoreception to neurotransmission and exocytosis, as well as processes such as embryogenesis, angiogenesis, tissue regeneration, and normal and aberrant cell growth. Analysis of the complete C. elegans genome (The C. elegans Sequencing Consortium, 1998) resulted in the prediction of 21 Gα, two Gβ, and two Gγ genes (E. Cuppen, unpublished results; Jansen et al., 1999). Besides one member (gsa-1, goa-1, egl-30 and gpa-12 ) of each of the mammalian Gα classes (Gs, Go/i, Gq and G12), there are 17 C. elegans specific genes (gpa-1-11, 13-17 and odr-3 ) that most closely resemble the mammalian Go/i 480 E. Cuppen et al.
class but cannot be clearly classified into any of the existing families (Jansen et al., 1999).
The Gs homologue gsa-1 is required to survive the first larval stage (Korswagen et al., 1997), and goa-1 (Go/i) and egl-30 (Gq) reciprocally function in neuromuscular processes that regulate behaviours such as locomotion, egg laying, defecation and pharyngeal pumping (Hajdu-Cronin et al., 1999;Mendel et al., 1995;Miller et al., 1999). Furthermore, goa-1 and gpa-16 are functionally redundant in asymmetric cell division (Gotta and Ahringer, 2001). At least 14 of the C. elegans-specific Gα subunits are thought to function in chemosensory signalling. They are expressed almost exclusively in the specialized structures that sense the environment, such as the amphid neurons, and loss-of-function and overexpression studies revealed chemotactic defects (Jansen et al., 1999;Roayaie et al., 1998). At this point, it is largely unknown how these G-proteins transduce their signal and how the observed specificity in signalling is established.
In this study, we tried to identify new regulatory mechanisms in heterotrimeric G-protein signalling by performing yeast two-hybrid interaction screens. Such screens may result in the identification of both upstream and downstream components in the signal transduction cascade, as well as modulators. We report here on the identification of 11 different interacting proteins, and a systematic approach by which we attempted to establish biological relevance for the observed interactions, using promoter-GFP reporter constructs and loss-of-function studies using RNAi and genetic mutants. Finally, we provide in vivo evidence to support the interaction between the C. elegansspecific Gα subunit GPA-7 and the nuclear receptor NHR-22 using a membrane-recruitment assay.

cDNA cloning
The predicted coding sequence of all C. elegans Gα subunits was amplified from total RNA (strain N2) by RT-PCR. First strand synthesis was initiated using a mixture of random hexamers and oligo-dT oligonucleotides and cDNAs were amplified by PCR, using specific oligonucleotides that include the start and stop codons and contain recognition sites for restriction enzymes for inframe cloning in the appropriate two-hybrid bait or prey vector (pGBK-T7 or pGAD-T7, Clontech). Detailed cloning information can be obtained from the authors upon request. The resulting cDNA constructs were checked by sequencing for potential differences with the spliced gene-product predicted by ACeDB and for the absence of mutations. Mutations resulting in constitutive active GTPase-deficient Gα subunits were introduced by site-directed mutagenesis using PCR and oligonucleotides harbouring the desired mutation. The amino acid changes that were induced were: GPA-5-QL, Gln203Leu; GPA-6-QL, Gln215Leu; GPA-7-QL, Gln203Leu; GPA-12-QL, Gln205Leu; GOA-1-QL, Gln205Leu; GSA-1-QL, Gln208Leu.

Two-hybrid library screening and yeast mating assay
Yeast two-hybrid interaction library screenings were done according to the manufacturer's protocol (Clontech). The C. elegans prey cDNA library that was screened (mixture of a random primed and a oligo-dT primed mixed stage C. elegans N2 cDNA library in pACT) was a kind gift of R. Barstead, OMRF Oklahoma City, OK. For mating assays, bait and prey plasmids were transformed to yeast strain AH109 (MAT a) and Y187 (MATα,), respectively, and grown overnight in selective medium lacking tryptophan or leucine, respectively. 10 µl of each culture was mixed in 80 µl YPDA in 96well microtitre plates in the appropriate matrix and grown overnight at 30 • C with shaking. 5 µl drops were seeded on plates lacking both tryptophan and leucine (LT − ) as a control for mating efficiency and on selection plates lacking tryptophan, leucine, histidine and adenine (LTHA − ) for detection of interactions.

Promoter-GFP reporter constructs
To study the in vivo expression pattern for candidate interacting genes, we generated PCR fragments harbouring 2-3 kb promoter sequence and the first 1-3 exons of the gene of interest, which are fused in-frame to the GFP coding sequence, as described previously (Hobert et al., 1999). PCRs were performed using Accutaq LA DNA polymerase mix (Sigma, St. Louis, MO), which has proofreading activity. Purified PCR fragments were directly injected at 50 µg/ml in dpy-20(e1282) animals together with the plasmid pMH86 (Han and Sternberg, 1991) at 150 µg/ml. For each promoter-GFP reporter construct, we created at least three transgenic lines. Transgenic animals carrying an extra-chromosomal array were analysed under a Zeiss Axioskop 2 fluorescence microscope for GFP expression. Cells were identified in reference to Sulston and Horvitz (1977) and White et al. (1986).

Nematode strains, culturing and manipulation
General methods used for culturing, manipulation, and genetics of C. elegans were as described (Lewis and Fleming, 1995). DNA transformation assays in C. elegans by microinjection were as described (Mello and Fire, 1995). Deletion mutants were identified as described (Jansen et al., 1997 Liu et al., 1999) a genomic region of 1.2 kb is deleted (exons 3-5), resulting in an out-of-frame mutation, deleting the ligand binding domain but leaving the DNAbinding domain intact. Male mating efficiency and behaviour of nhr-22 mutants was assayed as described (Hodgkin, 1983).

RNA interference
The cDNA inserts of the interacting candidates were subcloned from the prey vector into the XhoI site of the L4440 vector (Timmons and Fire, 1998) and transformed into the HT115(DE3) bacterial strain using standard methods. NGM-agar plates containing 1 mM IPTG were inoculated with 50 µl of an overnight bacterial culture and grown overnight. Ten L3-L4 wild-type or rrf-3 mutant animals (a strain hypersensitive to dsRNA; Simmer et al., 2002) were placed on the plate, incubated overnight and transferred to a fresh plate with nutrient that produces dsRNA of the specific cDNA insert. The F 1 progeny was monitored closely for any aberrant behavioural or developmental phenotype.

Results
cDNA cloning of all C. elegans Gα subunits C. elegans Gα subunit cDNAs were cloned based upon gene predictions and EST data present in ACeDB (release 12/98). When RT-PCRs failed, predictions were re-examined using three-frame translations of genomic sequence in ACeDB, compared with a multiple protein sequence alignment of the known C. elegans, mouse and human Gα subunits. We cloned the cDNAs for all Gα subunits, except for GPA-9, for which we were not able to amplify the 5 half of the gene, and GPA-17, which was discovered only recently. Sequence analysis shows that about 12% of the intron-exon boundaries (19 out of 154), including the manually corrected start and stop codons, were not correctly predicted by GENEFINDER and annotated in ACeDB (Version WS 12/98). Furthermore, no evidence for alternatively spliced isoforms was found.

Yeast two-hybrid interaction screen
Each of the cloned C. elegans Gα subunits and constitutively active GTPase-deficient mutants of GPA-5, GPA-7, GPA-12 and GSA-1 were used as bait in a yeast two-hybrid interaction screen (Table 1). 815 colonies grew on HTLA-plates, of which 547 turned blue on X-gal. 537 prey plasmids were successfully rescued and used to confirm the interaction specificity in a mating assay with the original and control baits. 100 interactions could not be reproduced and 137 were unspecific. The inserts of the 300 remaining clones were sequenced and compared with database sequences. Based upon the molecular identity, 103 clones were excluded from further analysis, as they encode proteins that reside in cellular compartments such as the mitochondrion, or are extracellular, these being locations in which Gα proteins are not expected to be functional. Furthermore, a set of common false-positives, the heat shock proteins, was discarded (a full list can be obtained from the authors upon request). The remaining 197 candidates were tested in a large yeast mating matrix for their interaction characteristics with all 19 cloned Gα subunits. In this experiment, additional control baits (empty vector, p53, C. elegans Gβ subunits) were included. At this point, we found that the original GPA-1 bait contained a frameshift mutation in the middle of the cDNA insert, resulting from a RT-PCR cloning error. We repeated the screen for GPA-1 with a corrected bait and used the frameshifted GPA-1 as a control for unspecific interactions. All other bait inserts were once more sequenced to exclude the presence of mutations and, in addition, all bait inserts were cloned into a bacterial expression vector. No other mutations were found and all bacterially expressed proteins were of the expected size (data not shown).
The majority of the clones (173) was found to interact with one or more of the control baits and thus could be considered unspecific interactors (a full list of the identity of these clones can be obtained from the authors upon request). The remaining 24 candidates did not interact with any of the control baits. As expected, all of these candidates did interact with the bait with which they were originally retrieved from the library. Furthermore, we found that none of the candidates interacted with any of the other C. elegans Gα subunits (data not shown), illustrating a high degree of specificity for the identified interactions.

Characteristics of the interacting candidates
The 24 remaining clones were found to encode 11 different proteins that interact with four different baits (Table 2). To confirm the observed interaction in an independent assay, we switched the bait and prey vector inserts and included GTPasedeficient mutants in a mating assay (Figure 1). Artificial interaction interfaces can be created at the junctions of the DNA binding domains and the bait, and/or the transcription activation domain and the prey. Changing the configuration can eliminate such false-positives, whereas strong interactions are expected to work in both configurations. Four prey inserts activated transcription by themselves, making them uninformative as bait.  Total  37  815  547  100  137  300  103  197  173 24 (11) * Numbers between brackets indicate the number of unique proteins. † After the two-hybrid screens a frameshift mutation was found in the middle of this bait construct. ‡ We were not able to clone the 5'end of GPA-9 cDNA. § This subunit was identified in a recently finished gap in the genome sequence during preparation of this manuscript.
For the other candidates, the interaction was confirmed in the reversed configuration assay, with the exception of T22A3.3 (a protein of unknown function). Interestingly, however, T22A3.3 did interact with K06A1.4 (NHR-22) and was also found to interact with LAG-1 [Su(H)] and MPK-1 (MAP-kinase) in other interaction screens (Walhout et al., 2000), suggesting either broad or unspecific interaction activity. We observed a strong homotypic interaction for Y50F7A.1 (Rap1GAP) and a weak homotypic interaction for F44B9.6 (LIN-36). Interaction analysis with GTPase-deficient mutants revealed that the interactions of the nuclear receptors K06A1.4 and Y45G5AM.1 with GPA-7 and GPA-13, respectively, are independent of the activation state of the Gα subunit. T22A3.3 was found to interact exclusively with the wild-type form of GPA-6 and only two (AGS3.1 and Rap1GAP) out of the seven candidates for GOA-1 interact with the GTPase-deficient mutant of GOA-1.

Expression analysis of the interacting candidates
Gene-expression patterns for the interacting candidates were obtained by fusing promoter regions to GFP-encoding sequences. Complete and partial overlap in expression pattern for the candidate genes and their respective bait can be observed for a subset of the candidates (Figure 2 and summarized in Table 3). For example, both NHR-22 and GPA-7 are expressed in different types of muscle cells and neurons, but also in the male specific neurons in the tail required for mating ( Figure 2I). GOA-1 is ubiquitously expressed in neuronal tissue and in muscles (Mendel et al., 1995;Segalat et al., 1995), and 484 E. Cuppen et al.  (2000), Gonczy et al. (2000) and Miyabayashi et al. (1999).

||
Interologs; interaction between homologous proteins was previously described in another species * * previously annotated as the homologue of human LIS-1 protein; now the gene is split in two and the interacting fragment has no homology to any known protein. † † Does also interact with wild-type GPA-7. ‡ ‡ nd, not determined, because these baits autoactivate the reporter genes. § § Bessou, et al. (1998); ¶ ¶ Thomas and Horvitz (1999);  Figure 1. Yeast mating assay. The interacting candidates identified in the screen and the corresponding Gα-subunits were cloned in both bait and prey vectors and tested in a mating matrix. Diploid strains were plated on Leu − /Trp − medium (left panel) to control for the presence of both bait and prey plasmids, and on Leu − /Trp − /His − /Ade − medium to select for two-hybrid interactions. Baits F44A6.1, K06A1.4, C05D2.6 and Y45G5AM.1 are not informative, because they autoactivate transcription of the reporter genes several of the interacting candidates are predominantly expressed in muscle cells such as AGS3.1 ( Figure 2C-E) and CalNuc, or neuronal cells like LIN-36, F59H5.1 and C01H6.9, whereas others are expressed in both neurons and muscle cells such as C05D2.6.
As for human haspin, expression of the C. elegans haspin homologue C01H6.9 is mainly observed in the nucleus ( Figure 2G). Some candidates seem to have no or little overlap with their respective bait. T22A3.3 is expressed in several neuronal cells, including a specific chemosensory neuron, AWB ( Figure 2H), but its binding partner GPA-6 is expressed in the chemosensory neurons AWA, ASI and PHB (Jansen et al., 1999).
We cannot exclude the possibility that the observed expression patterns are not complete, due to the lack of important regulatory promoter elements or to low expression levels. Furthermore, we may have missed co-expression in earlier developmental stages, as we determined the specific cells in L4 and adult stages. However, our data does show a correlation between the expression of the Gα subunits and their specific interacting partners.

Functional analysis of the Gα interactors
We performed RNAi experiments (Timmons and Fire, 1998) to functionally inactivate the genes encoding the proteins interacting with Gα. Wildtype (N2) and a mutant strain (rrf-3 ) that is hypersensitive to dsRNA (Simmer et al., 2002) were fed bacteria expressing dsRNA derived from the cDNA inserts of the prey clones. F 1 progeny was analysed for aberrant behavioural or developmental phenotypes (e.g. lethality, brood size, development, locomotion and egg-laying activity, morphology, etc.). Only RNAi of C05D2.6 in an rrf-3 mutant background resulted in a phenotype, e.g. sterile progeny (44%) and a reduced brood size (13 ±4, n = 27). In a wild-type background, no obvious phenotype for C05D2.6 RNAi is observed (brood size is 223 ±21, n = 23). The C05D2.6 (RNAi) phenotype correlates with the reduced broodsize observed in goa-1 genetic mutants (Mendel et al., 1995) and animals fed with dsRNA directed against GOA-1 (our own observations; Fraser et al., 2000). Interestingly, both GOA-1 and C05D2.6 are co-expressed in the distal tip cell of the gonad (Figure 2A; Mendel et al., 1995), and ablation of this cell is  Many neurons, muscle cells, many neurons in male tail GPA-13 § Subset of sensory neurons (ADF, ASH, AWC, PHA, PHB) GOA-1 ¶ Ubiquitously in neurons and muscles, distal tip cell of the gonad * A similar expression pattern was described in Miyabayashi et al., (1999). † Bessou et al. (1998); ‡ Thomas and Horvitz (1999); § Jansen, et al. (1999); ¶ Mendel et al. (1995). shown to result in a decreased number of germ cells (Kimble and White, 1981). Furthermore, RNAi against the candidate proteins that interact with GOA-1 in a mutant background did not modulate the goa-1 loss-and gain-of-function phenotypes ( Table 2). As knockdowns by RNAi may not be complete loss-offunction, we analysed three genetic mutants of interacting candidates: nhr-22 (K06A4.1), lin-36 (F44B9.6; Thomas and Horvitz, 1999) and nucb-1 (F44A6.1) but did not observe any phenotype for these mutants. As lin-36 has a synthetic multivulva phenotype when combined with mutations in class A synMuv genes (e.g. lin-8 ), we generated double mutants between goa-1 and lin-36 or lin-8. However, no synMuv phenotype was observed. Although there is only a single homologue for nucb-1 in C. elegans, whereas there are two in mammals, mutants behave completely like wildtype animals and no modulation of goa-1 lossor gain-of-function mutant phenotypes could be observed (Table 2), suggesting a non-essential role for this protein in C. elegans. Similarly, for a genetic mutant of nhr-22, no behavioural, developmental or male-mating defects were observed, nor modulation of gpa-7 loss-or gain-of-function mutant phenotypes (Table 2). As NHR-22 is a member of a very large, well-conserved gene family in C. elegans of more than 250 members (Enmark and Gustafsson, 2000;Miyabayashi et al., 1999;Sluder and Maina, 2001), functional redundancy, similar to that observed for the C. elegans Gα protein family (Jansen et al., 1999), cannot be excluded.

GPA-7 and NHR-22 interact in vivo
Because nhr-22 mutant animals do not show any obvious phenotypes, we examined whether there is in vivo evidence to support an interaction using a membrane recruitment assay. We co-expressed GFP coupled to wild-type or activated GPA-7 and NHR-22 coupled to a membrane localization signal in muscle cells. In the absence of NHR-22 (not shown) or the presence of wild-type NHR-22 ( Figure 3A), wild-type GPA-7 is localized in both the nucleus and cytoplasm. However, in the presence of membrane-targeted NHR-22, wild-type GPA-7 is recruited to the plasma membrane ( Figure 3B). Similar data was found for activated GPA-7 (not shown), consistent with the two-hybrid data ( Figure 1). Furthermore, the reversed experiment with GFP-tagged NHR-22 (which is, like GPA-7, localized in the cytoplasm and the nucleus) co-expressed with membranetargeted GPA-7, results in the same membrane recruitment (not shown). Taken together, these results show that GPA-7 and NHR-22 can interact both in vitro and in vivo. Such an interaction between a G-protein and a nuclear receptor has not (to our knowledge) been reported before.

Discussion
This study aimed to identify specific components in G protein signalling pathways by systematically using all C. elegans Gα subunits in yeast two-hybrid interaction screens. Although initially a large set of potential interactors was obtained, application of a series of controls resulted in a specific, but surprisingly small, set of only 11 different C. elegans proteins that interact with four Gα subunits. Among these are orthologues of proteins known to interact with Gα subunits and novel proteins with homologues in mammalian species.
The low number of interactors identified raises the questions: why did we not identify more interacting proteins?; and why did we not find interactions for the majority of the Gα subunits? First, the Gα or partner protein may not be folded properly, due to its fusion to the DNA-binding or transcription activation domain, prohibiting an interaction. Second, the proximity of the plasma membrane or the presence of additional co-factors may be necessary for proper protein folding or interaction. Indeed, we failed to reconstitute a trimeric complex of an α-, β-and γ -subunit in a yeast three-hybrid system (E. Cuppen, unpublished results; data not shown), although we were able to demonstrate two-hybrid interactions between β-, γ -and γ -likesubunits (van der Linden et al., 2001). Thirdly, Gα subunits may have to be in their active, GTPbound, state to interact with effectors. However, we only identified a single protein (K06A1.4) in the four screens with the GTPase-deficient mutants and none of the expected downstream effectors, such as adenylyl cyclases (Gilman, 1987) or phospholipases (Simon et al., 1991), were retrieved. The latter may be due to the way Gα proteins activate these downstream effectors. When these are shortlived 'kiss-and-goodbye'-type interactions, it may be impossible to detect them in an in vivo environment, limiting this screen to the identification of more structural interactions. Finally, it is possible that the interacting candidates are not present in the library that was screened, due to underrepresentation as a result of low expression levels, as is the case for some of the Gα subunits, which are expressed in only a few specific neuronal cells (Jansen et al., 1999). Illustrative of this is that most of the GPA sequences (except for GPA-4, GPA-12, GOA-1, GSA-1, EGL-30) are not present in EST and/or cDNA libraries (source: Wormbase).