The GPS Motif Is a Molecular Switch for Bimodal Activities of Adhesion Class G Protein-Coupled Receptors

Summary Adhesion class G protein-coupled receptors (aGPCR) form the second largest group of seven-transmembrane-spanning (7TM) receptors whose molecular layout and function differ from canonical 7TM receptors. Despite their essential roles in immunity, tumorigenesis, and development, the mechanisms of aGPCR activation and signal transduction have remained obscure to date. Here, we use a transgenic assay to define the protein domains required in vivo for the activity of the prototypical aGPCR LAT-1/Latrophilin in Caenorhabditis elegans. We show that the GPCR proteolytic site (GPS) motif, the molecular hallmark feature of the entire aGPCR class, is essential for LAT-1 signaling serving in two different activity modes of the receptor. Surprisingly, neither mode requires cleavage but presence of the GPS, which relays interactions with at least two different partners. Our work thus uncovers the versatile nature of aGPCR activity in molecular detail and places the GPS motif in a central position for diverse protein-protein interactions.


INTRODUCTION
Adhesion G-protein-coupled receptors (aGPCR) form the second largest of the five heptahelical (i.e., spanning seven transmembranes; 7TM) receptor families (Figures 1A and S1A) (Harmar, 2001;Fredriksson et al., 2003aFredriksson et al., , 2003b. The receptor class is essential for neuronal development, planar cell and tissue polarity as outlined by mutations throughout the spectrum of multicellular organisms. In humans, aGPCR mutations inflict bilateral frontoparietal polymicrogyria (BFPP) and Usher syndrome type IIc (Piao et al., 2004;Weston et al., 2004). In addition, aGPCR have also been implicated in the pathogenesis of attention deficit and hyperactivity disorder (ADHD) (Arcos-Burgos et al., 2010). Although it is accepted that cell-cell or cell-matrix contact mediated by the extracellular adhesion domains results in receptor signaling (Hamann et al., 1996;Stacey et al., 2003), the molecular mechanism of aGPCR activation has remained undetermined ( Figure 1A). Prevailing models of aGPCR activation and signaling have emerged from several lines of evidence. First, aGPCR can form distinct multicomponent signaling complexes with other transmembrane molecules in cis and in trans in an asymmetric layout across a single cell. This mode of signaling has been shown to occur for the cadherin-like aGPCR CELSR/Flamingo (Fmi) of Drosophila melanogaster (Usui et al., 1999;Chen et al., 2008). Second, in addition to ligand binding at the N-terminal adhesion domains interaction of a second ligand with the hormone receptor motif (HRM), a domain conserved in many but not all aGPCR, is required for signaling (Shima et al., 2004;Kimura et al., 2006). Third, the ''split personality'' receptor model proposes that N termini of aGPCR are liberated from their 7TM-module posttranslationally via cleavage at the GPS. The ectodomain fragment is then able to recombine with 7TM-modules of other aGPCR at the membrane (Silva et al., 2009).
In this study we have addressed these working models in the prototypical aGPCR lat-1, the nematode homolog of Latrophilin/CIRL/LPHN/CL, in an in vivo setting in Caenorhabditis elegans.

lat-1 Is Required for Sperm Function
We have previously selected lat-1, the C. elegans homolog of Latrophilin/CIRL/LPHN/CL ( Figure S1B), for investigations into the physiological requirements of a prototype aGPCR (Figures 1A and S1B). The null allele lat-1(ok1465) causes a distinct tissue polarity defect, which is associated with highly penetrant developmental lethality (Langenhan et al., 2009). During the screen for postembryonic phenotypes in lat-1(ok1465) mutants we discovered a second component of the Lat phenotype owing to a novel function of lat-1 in the male germline.
We employed mating experiments and feminization mutants to identify the gamete type depending on lat-1 function. In this instance, we observed that the ability to fertilize wild-type oocytes is reduced in lat-1 mutant males (N2 male x fem-1 hermaphrodite: 230 ± 8, n = 8; lat-1(ok1465) male x fem-1 hermaphrodite: 193 ± 5, n = 8; p = 0.0012; Figure 1C). This indicates that in addition to defects in embryonic tissue polarity lat-1 mutants show impaired sperm development or sperm function.
Thus, in addition to embryogenesis sperm function similarly depends on the presence of intact lat-1 signals offering a second biological context for testing latrophilin activity using modified receptor variants. We have used two lat-1 phenotypes-development and fertility-to systematically investigate the requirements of protein domains present in the LAT-1 receptor molecule via transgene complementation, and to delineate the contribution of these domains to the lat-1 wild-type gene product's activity.

The HRM Is Not Required for Receptor Activity
Previously we determined that transgenic rescue of lat-1(ok1465) absolutely requires the presence of the lectin-like RBL domain at the N terminus (DRBL and DTM2-7/DRBL ; Figures 2A-2C and  Table S1) (Vakonakis et al., 2008;Langenhan et al., 2009). The HRM has been suggested to act as a binding site for a putative secondary ligand in aGPCR signaling (Kimura et al., 2006), which would be consistent with its similarity to the ligand-binding domain of secretin-like GPCR (Figures S1B and S1C). Consequently we generated LAT-1 receptor variants with a modified HRM. Interestingly, activity ( Figure 2B and Table S1) and expression (Figures 2D,2E,S2A,S2B,and S3B) of transgenes that lack the lat-1 HRM (DHRM), or in which it is replaced by a divergent heterologous HRM from the nematode CELSR/Fmi homolog cdh-6 (HRM CDH-6 ), is indistinguishable from wild-type constructs.
Molecular models of several HRM of B1/secretin-like GPCR that have recently been solved (Grace et al., 2004;Parthier et al., 2007;Runge et al., 2008) are consistent with the rescue data on the dispensable requirement of the HRM for receptor function. The models reveal a highly conserved domain core similar to the short consensus repeat (SCR fold, also known as Sushi or CCP module) (Norman et al., 1991). Peptide-hormone ligands bind to pockets formed by an N-terminal a helix folding back onto this core. The binding is primarily mediated by hydrophobic patches in the SCR fold ( Figures 3A and 3B) (Grace et al., 2004;Parthier et al., 2007). The HRM of aGPCR lack the ligandbinding helix whereas the SCR core is conserved ( Figures 3A  and 3C). The first disulfide bond, which fixes the ligand-binding helix to the core, and the residues forming the ligand-binding pockets on the SCR are absent (Figures 3A and 3C). In contrast to secretin-like HRM, the N and C termini of the SCR fold are predicted to be located at opposing ends of the structure (Figures 3B and 3C), consistent with an extended conformation as described in non-HRM proteins (Norman et al., 1991). Molecular modeling thus strongly suggests that the HRM of aGPCR are SCR-like structural elements that lack the features required for the binding of peptide hormones by secretin receptor-like HRM (Parthier et al., 2007). We conclude that the HRM is dispensable for productive receptor activation and there is no evidence for an essential ligand-binding activity of the LAT-1 HRM.

LAT-1 Domains Relay Different Functions
We next tested the impact of the 7TM domain of LAT-1, which revealed 7TM-dependent and 7TM-independent components of the Lat phenotype. Transgenes carrying wild-type lat-1 and lat-1::gfp fusions rescue lethality and fertility phenotypes in lat-1(ok1465) animals ( Figure 2A and Table S1). In contrast, constructs expressing truncated LAT-1 variants lacking the 7TM/C terminus module (DTM2-7) or the intracellular C-terminal domain (DC-term) retain activity to complement the fertilization defect in lat-1 mutants, but did not rescue the tissue polarity phenotype ( Figure 2B and Table S1).
To rule out that 7TM-independent activity of the DTM2-7 receptor requires specific sequences in the first TM helix, we exchanged the remaining TM1 segment of the DTM2-7 receptor for a heterologous transmembrane anchor from the nematode integrin homolog pat-3. DTM2-7/TM PAT-3 exhibited no loss of activity compared to a DTM2-7 transgene ( Figure 2C and Table  S1). In contrast, functionality of the DTM2-7 transgene was still absolutely dependent on the presence of the RBL domain (DTM2-7/DRBL; Figure 2C and Table S1) (Langenhan et al., 2009).
Three possible models could explain the 7TM-independent function of LAT-1. First, the ectodomain might act as a ligand for a reciprocal receptor on an adjacent cell to provide a ''reverse signaling'' function, as has been shown for Notch-Delta signaling (Bray, 2006). Second, the split personality receptor model would suggest that the truncated constructs express an intact ectodomain, which is released by GPS cleavage and recombines with different 7TM domains to form an active receptor (Silva et al., 2009). As a third alternative, the truncated constructs might identify a distinct ''forward signaling'' state of LAT-1 receptor complex that does not require 7TM domain activation, e.g., the interaction and association with a coreceptor on the LAT-1-expressing cell (Chen et al., 2008;Strutt and Strutt, 2008).

Proteolysis of the GPS Is Dispensable for LAT-1 Activity
We investigated the 7TM-independent function of LAT-1 by mutagenesis of the GPS. The GPS is the most highly conserved feature of aGPCR ( Figures 1A and S1B), and is invariably present proximal to the 7TM domain ( Figure S1C). The polycystic kidney disease gene product Polycystin-1 (PKD1) belongs to the only other protein family containing a GPS; in this protein context the GPS is also positioned close to a transmembrane domain ( Figure S1B). The autocatalytic cleavage activity of the GPS in different molecular contexts has been investigated in detail (Lin et al., 2004;Wei et al., 2007), and crystal structures revealed that the GPS motif is embedded within a larger GAIN (GPCR-autoproteolysis inducing) domain required and sufficient for GPS proteolysis (Araç et al., 2012).
C. elegans LAT-1 and its paralog LAT-2 possess the HLY(S/T) cleavage consensus sequence and are cleaved as predicted ( Figures 4A and S4A). In a split personality receptor model, cleavage of LAT-1 molecules at the GPS is essential for activity by release of the ectodomain and re-association with other 7TM modules.
To probe whether GPS cleavage is required for LAT-1 signaling, we introduced mutations at positions À2 (GPS H528A ) or +1 (GPS T530A ) of the consensus motif, which abolish autocatalytic cleavage activity but retain cell surface expression (Lin et al., 2004;Yu et al., 2007). As expected, LAT-1 proteins containing either mutation were resistant to autocatalytic cleavage ( Figure 4B) but were expressed normally (Figures 4C,S2C,and S3B). Surprisingly, the rescue of lethality and fertility phenotypes by cleavage-deficient full-length receptors is statistically not different from cleavable wild-type controls ( Figure 4E and Table S1).
These results are consistent with the observation that the consensus sequence for autocatalytic cleavage of aGPCR is frequently lost in evolution, as has occurred in the CELSR/Fmi, EMR, and BAI aGPCR subfamilies that constitute $50% of the aGPCR family ( Figures S1A and S4B). Analysis of Ka/Ks ratios indicates persistent strong selective pressure on the GPS-coding regions even if the cleavage consensus is lost (Table S2), strongly suggesting a cleavage-independent role of the GPS.
We next introduced noncleavable GPS modifications into the truncated DTM2-7 receptor context, showing that the 7TMindependent activity of LAT-1 does also not require GPS cleavage (GPS H528A /DTM2-7, GPS T530A /DTM2-7; Figure 4F and Table  S1). This is further supported by the fact that the truncated constructs display similar activities in lat-1 single mutants and in lat-1 lat-2 double mutants (data not shown). Together, these results indicate that the 7TM-independent function cannot be explained by a transfer of the LAT-1 N-terminal fragment (NTF) to the C-terminal fragment (CTF) of LAT-2 (Volynski et al., 2004;Silva et al., 2009).

An Interaction between the GPS and 7TM Domains Is Required for LAT-1 Function
Although autocatalytic cleavage is not absolutely required and conserved, the presence of the GPS immediately next to TM domains is highly specific for aGPCR and PKD1 (Figures 1A and S1B). A secreted version of the LAT-1 ectodomain (DTM1-7; Figure 2B) and in-frame deletions of the GPS are completely devoid of rescuing activity in both truncated (DGPS/TMD2-7) (  4E, 4F, and Table S1), whereas surface expression of DGPS is intact (data not shown). These results suggest that the GPS is essential for LAT-1 signaling. These findings argue against a reverse signaling model, as the presentation of the RBL domain on a shortened membrane-tethered ectodomain should be sufficient to supply rescue activity and the membrane-proximal GPS should not be required. A clinically relevant mutation inside the GPS of the GPR56 receptor (GPS C497S ) is present in patients suffering from bilateral frontoparietal polymicrogyria (BFPP). This mutation changes an absolutely conserved cysteine residue, which is required for receptor maturation and cell surface expression (Piao et al., 2004). We confirmed this in our transgenic assay by the introduction of an equivalent mutation in lat-1 disrupting the predicted disulfide bond pattern of the GPS, which abrogated LAT-1 surface expression and resulted in an inactive receptor ( Figure 4E). Thus, the inactivity of the DGPS constructs is due to a sequence-specific requirement for the LAT-1 GPS.  Figure 2.
(E) Deletion of the GPS and a missense mutation (C497S), which abolishes its structural integrity, abrogate rescuing activity of respective full-length transgenes. In contrast, mutations that leave the GPS structure unaffected possess wild-type rescue function. A heterologous GPS from LAT-2 within the LAT-1 receptor context rescues fertility but not developmental phenotypes of lat-1 (ok1465) mutants.
(F) The remaining rescuing function of the LAT-1/ LAT-2 chimeric receptor relies on the presence of the 7TM/C terminus module but not its autocatalytic activity. For domain nomenclature see also legend in Figure 2 and Tables S1 and S4.
To test for a sequence-specific requirement of the LAT-1 GPS, we constructed chimeric receptors in which the LAT-1 GPS was exchanged for the GPS of C. elegans LAT-2 (GPS LAT-2 ; GPS LAT-2 / DTM2-7). The primary sequence of the LAT-2 GPS is similar in length and is 50% identical in amino acid sequence to the LAT-1 GPS. It retains the positions and intramolecular distances of all key structural elements ( Figure S4A). The chimeric constructs undergo autocatalytic cleavage (data not shown) and show normal expression on the cell surface ( Figures 4D, S2D, and S3B). Surprisingly, the full-length GPS LAT-2 receptor chimera rescues the fertilization defect of lat-1(ok1465) mutants but has no activity in tissue polarity signaling ( Figure 4E and Table S1), a profile indistinguishable from constructs lacking an active 7TM domain ( Figure 2B and Table S1). In contrast, the truncated chimeric construct GPS LAT-2 /DTM2-7 shows no residual activity ( Figure 4E and Table S1).
This indicates that a sequence-specific function of the GPS other than autocatalytic cleavage or surface expression is required for the 7TM-dependent and 7TM-independent activities of LAT-1.

Cross-Activation of LAT-1 Receptors
All findings regarding the different LAT-1 receptor variants are consistent with a model in which the 7TM-independent function is mediated by the localization of LAT-1 in a presignaling complex, which can be mediated by either the 7TM or the GPS. The full receptor function requires the presence of a matching pair of GPS and 7TM domains, for example by the GPS acting as a tethered endogenous ligand for the 7TM domain.
To test this model, we investigated whether a pair of mutant transgenes expressing different receptor variants each lacking the 7TM function can reconstitute full receptor activity by intermolecular complementation (Rivero-Mü ller et al., 2010). In this assay, the truncated DTM2-7 construct provides the wild-type N terminus and GPS, whereas the GPS LAT-2 construct harboring a mutant N terminus contributes a wild-type but inactive 7TM/C terminus module. Coinjection of the transgenes DTM2-7 and GPS LAT-2 , which individually only rescue the fertility defect but not lethality to wild-type level, fully reconstitutes LAT-1 activity for both requirements ( Figure 5A and Table  S1). We excluded that DNA recombination between the transgenes accounted for the intermolecular complementation as we were unable to detect repaired constructs by PCR in rescued strains. . The predicted masses of 60 kDa for a monomer and 120 kDa for a dimer were obtained from protein samples at 20 C using sedimentation equilibrium measurements. (C) A nonreducing polyacrylamide gel reveals a single band (60 kDa) in both monomer and dimer samples indicating that the dimeric form of the LAT-1 N terminus is not covalently linked. Additional bands smaller in size than 60 kDa indicate degraded protein.
We also observed that the intermolecular complementation of LAT-1 function is not dependent on GPS cleavage, as a GPS cleavage mutation on the ''donor'' (DTM2-7/GPS T530A ) or the ''recipient'' (GPS LAT-2(T874A) ) side does not prevent productive complementation ( Figure 5A and Table S1). The fact that the recipient construct can be cleavage-deficient strongly argues against the reconstitution of active monomers by transfer of N termini to 7TM domains in a split personality receptor scenario, as in our assay the inactive N terminus at the recipient receptor was inaccessible for replacement through lack of GPS cleavage.
However, both donor and recipient require the presence of the RBL domain, as the DTM2-7 construct does not complement the DRBL mutant that contains a wild-type GPS and 7TM ( Figure 5A). This suggests that ligand-induced proximity of two LAT-1 receptors mediated by the RBL domain might be required for the intermolecular complementation of GPS and 7TM domains. Our results are consistent with a model in which the ligand-induced dimerization of the NTF leads to the cross-activation of the 7TM domain by the GPS of the partner molecule. To further test this dimerization/activation model, we performed biochemical studies of the LAT-1 ectodomain expressed in HEK293 cells. Gel filtration and analytical ultracentrifugation of supernatants from cells expressing an epitope-tagged LAT-1 ectodomain show that the LAT-1 ectodomain is present in both monomeric and dimeric form ( Figure 5B). The dimeric form is very stable in low protein concentrations, but is not mediated by covalent bonds ( Figure 5C). These data are consistent with the prediction that the LAT-1 ectodomains can form stable dimers.
lat-1 and ten-1/Teneurin Are Expressed on the Same Cells and Are Not Epistatic Although the ability for intermolecular complementation sheds light on the 7TM-dependent activity of LAT-1, it does not account Figure 6. Genetic Interaction and Coexpression of lat-1 and ten-1 (A and B) Frequency of developmental (A) and fertility (B) defects in animals with different dosages of lat-1 and ten-1. Genotypes ordered in ascending severity of phenotype. Experimental error indicated as SEM. See also Table S3. (C and D) Expression of lat-1::gfp (C) and ten-1a::gfp (D) transgenes in epidermoblasts derived from the Caaa lineage during dorsal intercalation. Note that in Cpaaaa only ten-1a::gfp expression was found. Scale bars represent 10 mm. for the 7TM-independent function of the receptor. Such function might require interaction with an additional molecule(s) to yield co-reception and/or co-signaling similar to the association of the aGPCR Fmi/Stan of Drosophila melanogaster with the transmembrane proteins Frizzled (Fz) and Van Gogh/Strabismus (Vang/Stbm) (Chen et al., 2008;Strutt and Strutt, 2008).
In a recent biochemical screen Teneurin-2/Lasso/Neurestin/ DOC4/Tenascin-m was identified as a potential ligand for the Latrophilin-1 (Silva et al., 2011). The C. elegans genome contains only a single teneurin homolog, ten-1. Animals carrying the null allele ten-1(ok641) exhibit morphogenesis defects (Drabikowski et al., 2005) phenocopying defects observed in lat-1(ok1465) embryos. Consequently, we sought to confirm whether lat-1 and ten-1 function as receptor-ligand pair in vivo by investigating the genetic relationship between the null alleles lat-1(ok1465) and ten-1(ok641) on the basis of their individual and combined effects on development and fertility (Langenhan et al., 2009).
Interestingly, lat-1; ten-1 double mutants exhibited almost completely penetrant developmental arrest and a greatly reduced brood size ( Figures 6A, 6B and Table S3). We further observed that lat-1 and ten-1 alleles displayed nonallelic noncomplementation as animals carrying at least one wild-type allele for both loci already exhibited significant defects in embryogenesis and fertility ( Figures 6A, 6B and Table S4). This dosage-sensitivity indicates that both genes are acting at least partly in parallel during development and germline function and implies a synergistic rather than linear interaction between both genes. Although the devastating condition of the lat-1; ten-1 double mutants precluded the introduction of the DTM2-7 receptor as a direct test of whether the 7TM-independent activity of LAT-1 requires presence of TEN-1, our genetic data are consistent with a model where both proteins function as coreceptors for a yet unknown ligand relaying nonidentical signaling outputs into the same cell.
To corroborate this model we studied the expression patterns of lat-1 and ten-1 using transcriptional reporter transgenes. We have previously noted that an integrated lat-1::gfp transcriptional reporter transgene is expressed during epidermal dorsal intercalation of the epidermis, specifically only in left epidermoblasts (Langenhan et al., 2009). This results in a highly reproducible zebra-like pattern on the dorsal epidermis prior to syncytial fusion of these cells. In contrast, both nonsuperficial neuroblasts and right intercalating epidermoblasts are devoid of lat-1::gfp expression. Using 4D microscopy we mapped lat-1 expression to ABplaaaap daughters, ABarppaapa, and all four Caaa granddaughters ( Figures 6C and 6E). Intriguingly, an independently expressed ten-1a::gfp promoter reporter (Mö rck et al., 2010) shows expression in the same epidermal blastomeres as lat-1::gfp (Figures 6D and 6E) and additional lineages (ABaraax, MSapx, MSppx; data not shown) except one (Cpaaaa; Figures  6D and 6E). This implies that lat-1 and ten-1 act on the same cell rather than in trans to each other.
Taken together these results suggest that lat-1 and ten-1 have overlapping but not identical functions during embryonic morphogenesis and fertility, and are unlikely to act linearly as receptor and ligand in C. elegans.

DISCUSSION
The biological effects of aGPCR activation are involved in complex cellular traits such as developmental decisions. Thus, the limiting factor for the study of this receptor class has been the availability of informative assays. To overcome this difficulty, we have developed an in vivo assay based on the transgenic complementation of lat-1(ok1465), a developmentally lethal and subfertile lat-1 mutant strain. This assay evaluates the activity of engineered receptor variants (summarized in Figure 7A) interacting with wild-type up-and downstream components of the pathway irrespective of prior knowledge on their identity.
Our observations indicate that the HRM of LAT-1 is dispensable for receptor function in our assays. We find no evidence for a specific ligand-binding function of the LAT-1 HRM by domain deletion (DHRM) or domain exchange (HRM CDH-6 ) experiments. This result is in apparent contradiction to similar receptor mutants of the Drosophila aGPCR FMI, where deletion of the HRM (FMI DHRM ) resulted in loss of receptor activity (Kimura et al., 2006). However, this has been interpreted as the result of a nonspecific change of domain geometry in FMI DHRM leading to a total loss of FMI function rather than a specific loss of a putative HRM ligand binding activity. In addition, molecular models based on the 3D structures of the HRM of B1/secretin-like GPCR of the GLP receptor group show that features crucial for ligand binding are not conserved in aGPCR, although hormone binding to secretin-type HRM might differ. In contrast, we find conservation of the SCR-like fold in both 7TM families suggesting that it is a structural component of the evolutionarily ancient aGPCR family, ''shuffled'' into the proximity of the 7TM domain We further demonstrate that proteolysis at the GPS of LAT-1 is not necessary for receptor trafficking and activity in vivo. Previously, autoproteolytic function of the GPS, the common molecular denominator of all aGPCR class members (Bjarnadó ttir et al., 2007), has been a central element in models of aGPCR maturation and signaling (Okajima et al., 2010;Lin et al., 2011). Our findings are consistent with recent pharmacological data that show intact signaling capacity of the human aGPCR GPR133 carrying similar cleavage-disrupting mutations (Bohnekamp and Schö neberg, 2011). In addition, the results are also in accord with the presence of cleaved and uncleaved forms of aGPCR on the plasma membrane of native tissue (Iguchi et al., 2008), and with data suggesting that surface expression of aGPCR can occur independently of GPS cleavage Lin et al., 2004;Krasnoperov et al., 2009).
In contrast to GPS proteolysis, we observe that aGPCR signaling requires presence of a GPS as a structural element. The central finding underpinning this notion is the loss of activity in a receptor chimera carrying a heterologous GPS domain of LAT-2 in a LAT-1 context (GPS LAT-2 ) during development. A similar effect is achieved if the 7TM domain of LAT-1 is truncated but the GPS is left intact (DTM2-7). This suggests that the GPS interacts with the homologous 7TM domain of LAT-1, and possibly serves as an endogenous ligand during receptor activity. We investigated this model by intermolecular complementation assays using pairs of signaling-deficient LAT-1 variants. A combination of GPS LAT-2 and DTM2-7 receptors, each deficient in rescuing developmental defects of lat-1 mutants, reconstitutes full receptor function. This implies that homologous pairing of GPS and 7TM from two receptor molecules is required for receptor activity, which is corroborated by our biochemical evidence of LAT-1 dimers. Recent structural data of the GAIN domain, in which the GPS motif is embedded, will help to guide site-directed mutagenesis efforts to probe this putative interface in the future (Araç et al., 2012).We also provide evidence that the split personality model of aGPCR activity (Silva et al., 2009) does not account for reconstituted receptor function of hemi-receptors in the in vivo complementation assay. The biological implications of this split personality receptor model are intriguing as the putative cross-interaction of NTF and CTF derived from different aGPCR precursors could potentially re-route external signals perceived via the NTF to alternative intracellular messenger cascades activated through the CTF. The model suggests that the GPS LAT-2 -DTM2-7 reconstitution could ensue through formation of a functional receptor chimera by transfer of the NTF donated by the DTM2-7 to the CTF of the GPS LAT-2 after GPS cleavage of both parent receptors (Silva et al., 2009). Interestingly, we find that proteolysis-deficient receptor versions of either partner are sufficient to reconstitute full function in vivo.
Consequently, it appears unlikely that aGPCR signaling requires domain exchange in accord with a split personality model.
The GPS LAT-2 chimeric receptor also uncovered separate effects of lat-1 on fertility and tissue polarity/morphogenesis. lat-1 is expressed in the somatic gonad (Langenhan et al., 2009), and lat-1 mutants show reduced number of eggs laid and a high degree of sterility. Loss of fertility in lat-1 mutants but not developmental defects are efficiently rescued by a chimeric GPS LAT-2 receptor. This suggests that two different activities reside in the LAT-1 receptor molecule. One requires an interaction with the 7TM domain (LAT-1 GPS 4 LAT-1 7TM/C ) constituting a forward signal triggered by the GPS as a tethered agonist and transduced by the 7TM. The second activity is independent of the 7TM domain (LAT-1 GPS 4 X). Thus, the GPS interacts with partners, which are differentially affected in the chimeric GPS LAT-2 receptor protein. Two different models can be devised to account for the 7TM-independent function of LAT-1.
In the bidirectional signaling model a second, reverse signal is transmitted to the partner cell via engagement of the RBL domain ( Figure 7B). The forward signal is disrupted in the GPS LAT-2 chimera, whereas the reverse signal is still present. Studies in Drosophila suggest that another aGPCR design, the CELSR homolog FMI/STAN, also relays bidirectional signaling states via heterologous cis-interactions (forward signal) (Usui et al., 1999;Chen et al., 2008;Strutt and Strutt, 2008) and homophilic trans-interactions (reverse signal) (Usui et al., 1999;Kimura et al., 2006;Chen et al., 2008). Our molecular analyses do not support a noncell-autonomous activity of a diffusible LAT-1 ectodomain (Langenhan et al., 2009), a separable activity of the anchored N terminus based on adhesive properties, or the activation of counter-receptors on adjacent cells in a reverse signaling mode (this work). We do not observe partial rescue through a receptor lacking the GPS (DGPS), where the reverse signal should still be intact. But we can currently not exclude that the overall topology of the receptor is disrupted in DGPS nor can we exclude that N-terminal portions of the GAIN domain, the interspaced region between HRM and GPS (Araç et al., 2012), contributes to reverse signal.
The alternative bimodal forward signaling model encompasses a second, parallel forward signal that is still active in the GPS LAT-2 chimera ( Figure 7C). Our results are consistent with a model in which the homodimerization of the NTF could be ligand-induced through the RBL domain. This would lead to the cross-activation of the 7TM domain by the GPS of the partner molecule (Figure 7). The model is consistent with the action of the GPS as a tethered agonist of the 7TM domain (forward signal 1) and mediating protein-protein interaction with accessory proteins equivalent to RAMPs in B1/Secretin-type GPCR (forward signal 2) (Qi and Hay, 2010). An interaction candidate acting on the same cell membrane transducing the 7TM-independent activity of LAT-1, is Teneurin/ten-1, a putative LPHN1 ligand (Silva et al., 2011). Our experiments show coexpression of lat-1 and ten-1 in the developing epidermis implying that LAT-1 and TEN-1 reside in close proximity in the cell membrane, possibly as part of the same signaling complex. In addition, animals lacking lat-1 or ten-1 display dosage-dependent nonallelic noncomplementation. These results are not compatible with an epistatic genetic relationship expected from a ligand-receptor pair but rather indicate parallel functions for lat-1 and ten-1.
Finally, future studies need to focus on the possibility that features of both signaling models are present in aGPCR. Although we have found no direct evidence for a trans-acting function of LAT-1 in C. elegans, a recent report indicates that rodent Latrophilin 1/CIRL1/CL1 interacts in trans with neurexins/NRX with high affinity in cell culture . This interaction requires the presence of the olfactomedin domain, which is only present in vertebrate latrophilin homologs ( Figure S1B; Langenhan et al., 2009), and might exert a role in synapse formation and/ or function.
In summary, we have shown that the aGPCR LAT-1 is capable of relaying at least two different signals. The GPS comprises the central structural element of aGPCR function that could be instrumental for the conversion of an adhesive event into receptor output through G protein signaling, whereas the proteolytic nature of the GPS is not required for this activity. Given the general pharmacological tractability of 7TM receptors, aGPCR are an attractive yet unexploited target to modulate adhesiontriggered cell behaviors involved in neurological and immune functions, and for tumor treatment.

EXPERIMENTAL PROCEDURES
C. elegans Strains C. elegans strains were cultured and manipulated according to standard protocols (Brenner, 1974). Wild-type worms were C. elegans variety Bristol, N2. Alleles lat-1(ok1465) and ten-1(ok641) were generated by the C. elegans gene knockout consortium. Strains were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources (NCRR). A full list of transgenic worm strains is included in the Extended Experimental Procedures.

Lethality and Fertility Rescue Assay
The adult brood size and oocyte assay was conducted as previously described (Langenhan et al., 2009). All experiments were conducted at least in triplicate. A two-tailed Mann-Whitney test of data sets against a respective wild-type control construct data set were performed using Prism 5 (GraphPad Software, La Jolla, CA).

Microscopy
For analysis of transgene expression embryos were dissected from gravid hermaphrodites and mounted in M9 onto a 2% agarose pad. Images were acquired with a Deltavision Core (Applied Precision), a Leica SP5 II confocal microscope (Leica, Milton Keynes, UK) and a Zeiss confocal LSM5 setup. Four-dimensional imaging and lineage analysis of lat-1::gfp and ten-1a::gfp expressing cells were performed as previously described using SIMI Biocell software (SIMI Reality Motion Systems, Germany) (Bischoff and Schnabel, 2006;Langenhan et al., 2009). Z-stacks with spatial spacing of 0.5 mm were taken.

Protein Biochemistry
Worms were washed off culture plates with ice-cold water, pelleted by centrifugation at 13,000 rpm for 5 min and freeze-cracked in dry ice/ethanol. The lysate was boiled in 100 ml TBS (20 mM Tris, 150 mM NaCl, pH 7.5) with 1% SDS for 10 min, spun down, and supernatant and pellet separately mixed with 23 Laemmli buffer. A detailed description of methodologies used for protein analysis including western blotting, immobilized metal ion affinity chromatography, size-exclusion chromatography by gel filtration, and analytical ultracentrifugation can be found in the Extended Experimental Procedures.

SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures, four figures, and four tables and can be found with this article online at http://dx. doi.org/10.1016/j.celrep.2012.06.015.

LICENSING INFORMATION
This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 Unported License (CC-BY; http://creativecommons. org/licenses/by/3.0/legalcode).