The G protein‐coupled receptor (GPCR) Proteolysis Site (GPS) of cell‐adhesion GPCRs and polycystic kidney disease (PKD) proteins constitutes a highly conserved autoproteolysis sequence, but its catalytic mechanism remains unknown. Here, we show that unexpectedly the ∼40‐residue GPS motif represents an integral part of a much larger ∼320‐residue domain that we termed GPCR‐Autoproteolysis INducing (GAIN) domain. Crystal structures of GAIN domains from two distantly related cell‐adhesion GPCRs revealed a conserved novel fold in which the GPS motif forms five β‐strands that are tightly integrated into the overall GAIN domain. The GAIN domain is evolutionarily conserved from tetrahymena to mammals, is the only extracellular domain shared by all human cell‐adhesion GPCRs and PKD proteins, and is the locus of multiple human disease mutations. Functionally, the GAIN domain is both necessary and sufficient for autoproteolysis, suggesting an autoproteolytic mechanism whereby the overall GAIN domain fine‐tunes the chemical environment in the GPS to catalyse peptide bond hydrolysis. Thus, the GAIN domain embodies a unique, evolutionarily ancient and widespread autoproteolytic fold whose function is likely relevant for GPCR signalling and for multiple human diseases.
There is a Have you seen? (March 2012) associated with this Article.
Autoproteolysis is a widely used intramolecular post‐translational process that often converts inactive protein precursors into biologically active mature proteins. For instance, the hedgehog pathway, one of the fundamental signal transduction pathways in animal development, is activated only if the signalling domain is cleaved off from the autoproteolytic domain of hedgehog (Lee et al, 1994). Among autoproteolytic domains/motifs described in various proteins (including hedgehog, N‐terminal nucleophile (Ntn) hydrolases, nucleoporin98, protein inteins, and pyruvoyl enzymes; Perler et al, 1997), the G protein‐coupled receptor (GPCR) Proteolysis Site (GPS) is one of the few autoproteolysis‐mediating domains/motifs found in membrane proteins. The GPS motif is conserved in all 33 human cell‐adhesion GPCRs and all 5 human polycystic kidney disease (PKD) proteins (Krasnoperov et al, 1997; Hughes et al, 1999; Ichtchenko et al, 1999; Chang et al, 2003; Lin et al, 2004); representing one of the most widespread and conserved autoproteolytic domains/motifs. Recent studies indicate emerging physiological and pathological roles for these GPS‐containing membrane proteins in cancer (Kan et al, 2010), PKD (Qian et al, 2002), brain development (Piao et al, 2004), myelination of neurons (Monk et al, 2009), central nervous system angiogenesis (Kuhnert et al, 2010), attention‐deficit/hyperactivity disorder (Arcos‐Burgos et al, 2010), cell polarity in neural development (Chae et al, 1999; Usui et al, 1999; Shima et al, 2004; Langenhan et al, 2009), Usher syndrome 2 (Ebermann et al, 2009), mouse male infertility (Davies et al, 2004), and immunity (see Yona et al, 2008 for a review), but the mechanism of GPS‐mediated autoproteolysis remains unclear.
Cell‐adhesion GPCRs are characterized by long N‐terminal extracellular sequences comprising multiple domains, and constitute a poorly studied class of GPCRs that is the second largest family of GPCRs in humans (Fredriksson et al, 2003). Most cell‐adhesion GPCRs are orphan receptors with no known ligands. They are found in animals, protozoa, and alveolates, predating many other important signalling molecules. The CIRL/Latrophilins (here abbreviated as CLs, CL1–3) and brain angiogenesis inhibitors (BAIs, BAI1–3) are two distant subfamilies of cell‐adhesion GPCRs that both are primarily expressed in neurons (Krasnoperov et al, 1997; Lelianova et al, 1997; Nishimori et al, 1997; Sugita et al, 1999). CL1 is a calcium‐independent receptor for α‐latrotoxin, a black widow spider toxin that triggers massive neurotransmitter release from neurons and neuroendocrine cells (Krasnoperov et al, 1997; Lelianova et al, 1997; Sugita et al, 1998; Sudhof, 2001; Deak et al, 2009). BAIs initially received their name because an extracellular fragment of BAI1 inhibited neovascularization in experimental angiogenesis (Nishimori et al, 1997; Kaur et al, 2005). Recently, C1q‐like proteins, a family of secreted proteins, were found to bind to BAI3 and to regulate synapse formation and/or maintenance (Bolliger et al, 2011). CLs and BAIs are now known to be two of the most mutated genes in human cancers (especially in squamous lung carcinoma and lung adenocarcinoma) and may act as tumour suppressor genes (Wood et al, 2007; Jones et al, 2008; Parsons et al, 2008; Kan et al, 2010). On the other hand, PKD family proteins are not GPCRs, but they contain large extracellular sequences, multiple transmembrane regions and a GPS motif similar to cell‐adhesion GPCRs (Ponting et al, 1999). PKD1 is a 500‐kDa 11‐pass membrane protein with a large N‐terminal extracellular region that is cleaved at the GPS motif (Qian et al, 2002). It is mutated in 90% of the autosomal dominant PKD (ADPKD) patients (Rossetti et al, 2001). A mouse mutant with uncleavable PKD1 exhibits abnormal kidney development (Yu et al, 2007).
In cell‐adhesion GPCRs and PKD proteins, the GPS motif is always located at the end of their long N‐terminal extracellular regions, immediately before the first transmembrane helix of the respective protein (Krasnoperov et al, 1997; Sugita et al, 1998; Ichtchenko et al, 1999; Chang et al, 2003; Lin et al, 2004; Figure 1A). The GPS motif is a conserved sequence of ∼40 amino acids containing canonical cysteine and tryptophan residues. The GPS is referred to as a ‘domain’ in the Pfam database although no three‐dimensional structure is available, and it is unclear whether it folds autonomously. Most, if not all, cell‐adhesion GPCRs undergo autoproteolysis in the GPS between a conserved aliphatic residue (usually a leucine) and a threonine, serine, or cysteine residue (L↓T/S/C) (Krasnoperov et al, 1997, 2002b; Nechiporuk et al, 2001; Lin et al, 2004). Although previous studies suggested that the GPS motif on its own promotes autoproteolysis (Krasnoperov et al, 2002a), it is hard to imagine how such a short sequence could fold into a catalytically active autoproteolytic domain given the stability of peptide bonds. Interestingly, in cell‐adhesion GPCRs the GPS motif appears to be preceded by a conserved so‐called ‘stalk’ region, which in turn is often preceded by a Hormone Receptor (HormR) domain, but the role of these sequences remains unclear.
To shed light on the GPS‐mediated autoproteolysis of cell‐adhesion GPCRs and PKD proteins, and to explore the significance of the stalk and GPS sequences and their relation to the HormR domain, we performed structural and functional studies of fragments containing these regions. Here, we show that the GPS motif is not an autonomously folded unit but rather part of a much larger evolutionarily conserved domain that we refer to as the GPCR‐Autoproteolysis INducing (GAIN) domain. We obtained crystal structures of fragments from two distantly related cell‐adhesion GPCRs (CL1 and BAI3) that include the GAIN domain and, additionally, the HormR domain. These structures revealed a novel fold with a high degree of structural conservation of even distant GAIN domains. Functionally, the entire GAIN domain is required and sufficient for autoproteolysis. Sequence analyses showed that the GAIN domain is an ancient domain that exists in primitive ancestor organisms, and is conserved in all cell‐adhesion GPCRs and all PKD1‐related proteins. Our structures of the cleaved CL1 and uncleaved BAI3 GAIN domains, in combination with mutagenesis experiments, reveal the structural determinants for autoproteolysis and suggest a plausible mechanism of cleavage. Taken together, our data describe a newly identified, evolutionarily conserved domain that is present in 38 different human proteins and functions as an autoproteolytic fold.
Structures of CL1 and BAI3
To test the significance of the sequence arrangements of the HormR domain, the stalk region and the GPS motif, and to explore the mechanism of the GPS‐mediated autoproteolysis, we determined the structures of fragments containing these regions from two distantly related cell‐adhesion GPCRs, CL1 and BAI3 (Figure 1A). Experimental phases for the CL1 and BAI3 structures were independently obtained from sulphur and iodine single anomalous diffraction (SAD) data at 1.9 and 2.3 Å resolution, respectively (Supplementary Table S1, Materials and methods). Both structures reveal two domains (Figure 1B and C; Supplementary Figure S1A and B): at the N‐terminus is the smaller ∼70‐residue HormR domain, and at the C‐terminus is a larger ∼320‐residue domain that we refer to as the GAIN domain; both structures are similar, but with some significant differences (see below and Supplementary Figure S1C). Surprisingly, the GPS motif is an integral part of the GAIN domain (comprising the last five β‐strands). Thus, the GPS motif itself does not constitute an autonomously folded unit, but rather forms a single folded domain together with the so‐called ‘stalk’.
DALI (Holm and Sander, 1993) searches (as of January 2012) indicated that the GAIN domain is a newly identified fold, and that our CL1 and BAI3 crystal structures represent the first high‐resolution structures of this domain. The N‐terminal subdomain A of the GAIN domain is composed of six α‐helices, while the C‐terminal subdomain B consists of a twisted β‐sandwich including 13 β‐strands and 2 small α‐helices. The high similarity of the GAIN domain structures (root‐mean square deviation (r.m.s.d.)=1.38 Å; Supplementary Figure S1C) from two distantly related cell‐adhesion GPCRs (24% sequence identity) suggests structural conservation of this novel domain across the entire family of cell‐adhesion GPCRs.
The GAIN domain functions as an autoproteolytic domain
To determine whether the GAIN domain is the minimally required region for autoproteolysis, we designed deletion mutants of full‐length CL1: CL1‐GPS which lacks the HormR and GAIN domains except for the GPS motif; CL1‐HormR_GPS in which all residues of the GAIN domain except for the GPS motif are deleted; CL1‐ΔSubA in which only subdomain A of the GAIN domain is deleted; and CL1‐ΔHormR in which only the HormR domain is deleted (Figure 2A). We expressed the deletion mutants in HEK293 cells, and analysed their autoproteolysis by immunoblotting of cell lysates, and their plasma membrane transport by indirect immunofluorescence of non‐permeabilized cells. Our results show that the entire GAIN domain with both subdomains A and B is required for autoproteolysis (Figure 2B and C; Supplementary Figure S2A). Consistent with the identification of the GPS motif as an integral part of the GAIN domain, the GPS motif is not functional by itself (Figure 2B and C), presumably because it is unfolded when expressed without the other GAIN domain sequences. A protein fragment composed of only the GAIN domain of CL1 underwent autoproteolysis, demonstrating that the GAIN domain is not only necessary but also sufficient for autoproteolysis in CL1 (Figures 2D and 5D; Supplementary Figure S2B). Indeed, the electron density map of the crystal structure shows that the crystallized HormR‐GAIN fragment of CL1 is cleaved at the expected position (Figure 1D; Krasnoperov et al, 1997). In addition, the GAIN domain of GPR56 (a different cell‐adhesion GPCR, mutations of which cause bilateral frontoparietal polymicrogyria (BFPP); Piao et al, 2004) is sufficient for autoproteolysis (Figure 2E and F; Supplementary Figure S2C), suggesting that the GAIN domain mediates autoproteolysis in other cell‐adhesion GPCRs as well.
The GAIN domain is an ancient conserved fold
To study the evolution of GAIN domains, we searched for the presence of GAIN domains in the genomes of primitive evolutionarily distant organisms, and identified genes encoding such domains. Strikingly, structure‐based BLAST searches revealed that the genome of Dictyostelium discoideum, an evolutionarily ancient organism that arose before animals emerged, encodes a GAIN domain (Figure 3A), although it lacks hedgehog autoproteolytic domains as well as signalling and adhesion domains found in higher eukaryotes, such as HormR, Wnt‐signalling components, cadherins, immunoglobulins, and integrins (Eichinger et al, 2005). Moreover, a unicellular ciliate, Tetrahymena thermophila, encodes more than eight GAIN domains. Thus, the GAIN domain is among the most ancient GPCR‐related extracellular domains. Computational analysis (McGuffin et al, 2000; Pei et al, 2007) of ancient GAIN domain sequences showed that their predicted secondary structure pattern is identical to that of the CL1 and BAI3 GAIN domain (Figure 3B; Supplementary Table S2). Moreover, their GPS motifs including the putative cleavage sites are well conserved (Figure 3A), whereas those of the autoproteolytic domain of nucleoporin98 (which also exists in ciliates) are not. Furthermore, sequence alignments of cell‐adhesion GPCRs and PKD proteins revealed that all cell‐adhesion GPCRs and PKD proteins are predicted to have a GAIN domain with identical predicted secondary structure patterns to that of the CL1 and BAI3 crystal structures (Figure 3B; Supplementary Table S3). Apart from the shared GAIN domain, PKD proteins are unrelated to cell‐adhesion GPCRs but also undergo autoproteolysis (Ponting et al, 1999), indicating that the GAIN domain has been utilized by at least two different membrane protein families during evolution. These results suggest that the GAIN domain evolved prior to the evolutionary emergence of animals, that it is one of the first functional autoproteolytic domains, and thus represents an ancient feature that is not directly connected to multicellularity.
Autoproteolysis does not lead to the dissociation of the cleavage products
In general, autoproteolysis domains produce two protein fragments that then dissociate (e.g., hedgehog and protein inteins). Strikingly, the cleaved β‐strand in the CL1 structure is well ordered with full occupancy and thermal factors comparable to the surrounding protein, indicating no sign of dissociation (Figure 1D). The cleaved β‐strand is tightly bound to the rest of the GAIN domain involving an extensive network of conserved interstrand hydrogen bonds and primarily hydrophobic side‐chain interactions (Figure 1F and G). Thus, we predict that dissociation of the cleaved β‐strand would be energetically unfavourable, and would conceivably lead to unfolding of the remaining part of the GAIN domain. Consistent with this structural prediction, the N‐terminal cleavage product of full‐length CL1 and CL3 in transfected HEK293 cells remained bound to the rest of the protein on the cell surface, and was not secreted into the media (Figure 1H). Thus, the GAIN domain forms a tightly associated heterodimer upon proteolysis.
Different from CL1, the electron density of the crystal structure of the BAI3 GAIN domain showed no evidence of cleavage (Figure 1E, Supplementary Figure 7B), and BAI3 expressed in transfected HEK293 cells was uncleaved (Figure 4A). To address the question whether the BAI3 GAIN domain is functional, we raised antibodies to BAI proteins and showed that endogenous BAI proteins in brain are cleaved (Figure 4C). The conditions of BAI expression (i.e., expression levels, glycosylation machinery, and reductive environment) are very different between insect cells (used for expression of the crystallized proteins) and HEK293 cells on the one hand, and neurons on the other hand. Thus, these differences in post‐translational processing and environment could be the reason why BAI3 is uncleaved when expressed in insect or HEK293 cells, but cleaved in the brain. Nevertheless, uncleaved BAI3 was efficiently transported to the plasma membrane in HEK293 cells, indicating that cleavage at the GPS is not required for surface transport of a cell‐adhesion GPCR (Figure 4B; see also below). In contrast, autoproteolysis of the autoproteolytic protein nucleoporin98 is required for its transport to the nuclear envelope (Hodel et al, 2002).
Unique structural features of the GAIN domain enable autoproteolysis
The autocatalytic scissile bond in the GPS of the GAIN domain (star, Figure 5A and B) is positioned at a sharply kinked loop between the last two β‐strands of the GAIN domain. In the structure of the uncleaved BAI3 GAIN domain, the autocatalytic scissile bond (Figure 1E) has a distorted geometry (phi=174.0, psi=−158.2 degrees). Furthermore, numerous residues contribute to the unique structural features of the GAIN domain that localize the scissile bond at the tip of the sharply kinked loop (Figure 5A and B; Supplementary Figure S3A and B). The kink is created due to three structural features of the protein: (1) Two conserved disulphide bonds (C834‐C820 and C832‐C801 in CL1) between neighbouring β‐strands (Supplementary Figure S3A) keep this β‐turn in place. (2) The cleaved β‐strand has conserved residues involved in numerous hydrophobic interactions with other conserved residues (Figure 1F). (3) The side chain of L837 in CL1 (L856 in BAI3) at the cleavage site is trapped in a conserved hydrophobic pocket (Supplementary Figure S3B). Comparison of the cleaved and uncleaved GAIN domain structures revealed a small tilt of the cleaved β‐strand upon autoproteolysis that relaxes the sharp kink (Figure 1I).
To understand the mechanism of GAIN‐domain catalysed autoproteolysis, we performed extensive mutagenesis of the CL1 GAIN domain (Figure 5C). Among others, three mutations (H836S, L837A, and T838G) impaired autoproteolysis but did not interfere with surface transport (and consequently, protein folding) (Figure 5C and D; Supplementary Figure S4), indicating that these residues are directly involved in CL1 autoproteolysis (also providing further evidence that autoproteolysis is not required for trafficking of GAIN domain‐containing proteins). Consistent with a previously proposed mechanism for GPS autoproteolysis (Lin et al, 2004), H836 may act as a general base (although the general base is likely different in BAI3 since that residue is an Arginine; Figure 5A and B). The general base then removes the hydroxyl hydrogen from the side chain of T838 (Serine in BAI3) that in turn makes a nucleophilic attack on the carbonyl carbon of L837 (conserved in both CL1 and BAI3). Note that the T838P (Krasnoperov et al, 2002b) and T838A mutants are unfolded (Figure 5D; Supplementary Figure S4), but replacing threonine with a glycine enables proper folding, presumably by decreasing rigidity at the cleavage site.
Disease‐related mutations in PKD1 and CL GAIN domains
The GAIN domain of PKD1 is a hot spot for mutations associated with ADPKD, one of the most common monogenic diseases (1 in 1000) (Qian et al, 2002; Figure 6A). In addition, CL and BAI genes are among the most frequently mutated genes in human cancers, with mutation frequencies of 3 and 12%, respectively (Kan et al, 2010). Their GAIN domains are, again, a hot spot for cancer mutations (Figure 6C; Supplementary Figure S6A). We tested the ADPKD and cancer mutations in GAIN domains for their effect on autoproteolysis and plasma membrane localization. Almost all ADPKD mutations in the GAIN domain of PKD1 abolished autoproteolysis, either by directly interfering with catalysis or by altering their membrane localization (presumably, by impairing folding) (Figure 6A and B; Supplementary Figure S5), indicating an essential role for the GAIN domain. However, cancer mutations in CL1 and CL3 did not affect autoproteolysis or membrane localization (Figure 6C and D; Supplementary Figure S6), suggesting that the mechanism by which these mutations lead to disease is different from that of the ADPKD mutations. Since most cancer mutations are located on the surface of the CL GAIN domains, they may be involved in the interaction of the GAIN domain with a ligand or with the nearby transmembrane helices. Indeed, it is the GAIN domain of CL1 that interacts with its long‐known binding partner α‐latrotoxin, a black widow spider toxin (Sudhof, 2001; Figure 7), demonstrating the ability of the CL1 GAIN domain to function as a protein‐binding domain in addition to an autoproteolytic domain.
Cell‐adhesion GPCRs exhibit structural similarity to hormone receptors
Although the HormR domain is the second most frequently observed domain in cell‐adhesion GPCRs (found in 12 of 33 human cell‐adhesion GPCRs), no hormones have yet been identified to bind cell‐adhesion GPCRs. HormR domain‐containing cell‐adhesion GPCRs are homologous to Class B GPCRs that always include a HormR domain (Lagerstrom and Schioth, 2008). Remarkably, our HormR domain crystal structures from CL1 and BAI3 revealed an unusually high structural similarity (0.7 Å r.m.s.d.) to the genuine hormone‐binding domain of the corticotrophin releasing factor receptor (CRFR), a member of the class B GPCRs, supporting the hypothesis that cell‐adhesion GPCRs may be hormone receptors (Figure 8A).
Superposition of the CL1 structure with the structure of the CRFR HormR domain bound to corticotrophin releasing factor (CRF), a 41 amino‐acid peptide hormone, reveals that a similar hormone could not bind to CL1 HormR in the conformation observed in the crystal structure since the GAIN domain is blocking the homologous hormone binding site on the HormR domain of CL1 (Figure 8B). Clearly, hormone binding would require a conformational change in CL1 to expose the putative hormone binding site. In class B GPCRs, the HormR domain precedes the transmembrane helices and is juxtaposed to the membrane. According to the two‐domain model of class B GPCR activation, the interaction of the extracellular hormone‐binding domain with the hormone promotes the interaction of the hormone with the transmembrane helices, leading to the activation of the receptor (Hoare, 2005). However, the mechanism of cell‐adhesion GPCR activation upon hormone binding may be different, because the GAIN domain lies between the HormR domain and the transmembrane helices.
In this study, we aimed to obtain a structural understanding of the autoproteolysis function of the GPS motif that is characteristically found in cell‐adhesion type GPCRs and PKD proteins. Towards this end, we here determined the crystal structures of fragments containing the GPS motif, the stalk region and the HormR domain from two distantly related cell‐adhesion GPCRs, CL1 and BAI3. Unexpectedly, we found that a single large domain termed the GAIN domain spans the GPS motif and the stalk region (Figure 1), and that the GPS motif by itself is not a functional domain, but only a small part of the much larger GAIN domain. The GAIN domain is a novel domain, and our CL1 and BAI3 crystal structures now define this domain family.
The CL1 deletion experiments showed that the GPS motif is not functional by itself (Figure 2). Instead the entire GAIN domain is both required and sufficient for autoproteolysis. Thus, the GAIN domain functions as an autoproteolytic fold. Strikingly, structure‐based database searches revealed that primitive organisms, such as Dictyostelium discoideum that arose early in evolution before animals emerged, encode GAIN domains although they lack most other autoproteolytic domains, important adhesion and signalling domains and critical signalling pathways (Figure 3; Supplementary Table S2). These ancient GAIN domains have conserved GPS motifs and are examples of one of the most ancient functional autoproteolytic domains. Computational analysis of the GAIN domain sequences showed that all cell‐adhesion GPCRs and all PKD proteins have a GAIN domain with identical predicted secondary structure patterns to that of the CL1 and BAI3 crystal structures showing that the GAIN domain is widespread and conserved in higher eukaryotes as well as in ancient organisms (Supplementary Table S3). Our cleaved and uncleaved GAIN domain structures, in combination with extensive mutagenesis of the CL1 cleavage site, reveal the unique structural features of the GAIN domain that enable self‐cleavage and shed light on the mechanism of autoproteolysis. The discovery of the GAIN domain also revealed a possible role for the GAIN domain in a variety of diseases, including cancer, ADPKD, and BFPP. We found that nearly all ADPKD mutations interfere with autoproteolysis of the GAIN domain, whereas cancer mutations have no effect on autoproteolysis or cell‐surface localization. Thus, the mechanism by which the cancer mutations lead to disease are unrelated to autoproteolysis and the GAIN domain likely has other functions (such as interacting with ligands or transmembrane helices) in addition to the autoproteolysis function.
In summary, our experiments demonstrate that the GAIN domain is an ancient autoproteolytic domain that exhibits unique properties and is present in a large number of GPCRs and PKD proteins. Although the proteolysis of membrane proteins by sheddases and regulated intramembrane proteolysis (RIP) proteases is well studied (Blobel, 2005; Wang et al, 2006) the autoproteolysis of a membrane protein is poorly understood. GAIN domain is one of the few domains to mediate self‐cleavage of a membrane protein. Two properties of the GAIN domain‐mediated autoproteolysis make it unique and intriguing: (1) All GAIN domains that we analysed (from ancient organisms to higher eukaryotes) always immediately precede the N‐terminal transmembrane helix (Figure 9). Additionally, HormR domains, when present, always precede the N‐terminus of the GAIN domain. (2) In contrast to most other autoproteolytic domains that cause the release of the cleaved mature protein, autoproteolysis of the GAIN domain does not lead to the dissociation of the cleaved extracellular sequences containing the GAIN domain from the membrane‐embedded regions of the protein. These observations naturally lead to the hypothesis that the GAIN domain is functionally and/or structurally associated with the membrane domains and that the GAIN domain‐mediated autoproteolysis has a complex mechanism of action.
The HormR/GAIN/transmembrane helix sequence arrangement is unrelated to the autoproteolysis function of the GAIN domain although it may be significant for the regulation of receptor signalling via intramolecular interactions. Indeed, deletion of the extracellular sequences of GPR56, a cell‐adhesion GPCR with an extracellular GAIN domain that is preceded only by a short non‐GAIN domain sequence (Figure 2E), causes constitutive activation of the GPCR (Paavola et al, 2011), suggesting a possible inhibitory role for the GAIN domain. In the GAIN domain, the cleaved β‐strand is separated from the first transmembrane helix by 5–6 linker residues (Figure 9), positioning the GAIN and HormR domains into close proximity to the transmembrane helices. The small conformational change that we observe upon autoproteolysis (Figure 1I) may be critical for the relative orientation of the GAIN domain with respect to the transmembrane helices. Binding of a hormone to the HormR domain or binding of an endogenous ligand to the GAIN domain may be potential regulators of signalling via the receptor. Our work here redefines the poorly studied cell‐adhesion GPCR class to share a large unique and widespread autoproteolytic domain that may be involved in downstream signalling.
Materials and methods
Protein expression and purification
The HormR and GAIN domains of rat CL1 (residues 460–849) and human BAI3 (residues 498–868) were cloned into the Xba1–Not1 and the BamH1–Not1 sites of the pAcGP67A vector, respectively (CL1=43 kDa plus 5 carbohydrates; BAI3=41.7 kDa plus 3 carbohydrates). A 8 × His tag was added at the C‐terminus for affinity purification. Spodoptera frugiperda (Sf9) cells (Invitrogen) were transfected with pAcGP67A carrying the gene and linearized AcNPV DNA (Sapphire Baculovirus DNA; Orbigen) using Cellfectin (Invitrogen). Baculovirus was amplified in Sf9 cells in 10% (v/v) fetal bovine serum containing SF900‐II medium (Gibco).
Large‐scale protein expression of CL1 and BAI3 was performed by infection of Trichoplusia ni (High Five) cells in Insect‐Xpress medium (Biowhitaker) at a cell density of 2 × 106 cells/ml with an infection course of 72 h. The media containing the secreted, glycosylated, recombinant protein was concentrated and dialysed using a tangential flow filtration system (PALL) and subsequently purified with Ni‐NTA resin (Qiagen) at 4°C. The proteins were eluted with 200 mM imidazole (pH 7.2). CL1 and BAI3 were further purified by size‐exclusion chromatography using a Superdex 200 column (Amersham Pharmacia Biotech) in buffer containing 10 mM Hepes (pH 7.2) and 150 mM NaCl. The peak corresponding to the proteins was then concentrated to 10.7 mg/ml (for CL1) and 17 mg/ml (for BAI3) in 30 kDa Centricon (Millipore) concentrators and used for crystallization trials.
α‐Latrotoxin (residues 413–1080) with N‐terminal GST and Myc tags in a pGEX‐KG vector was expressed in Origami cells. The expression and purification were described in Li et al (2005). The latrotoxin construct used for native gel experiments corresponds to the 20 ankyrin repeats of latrotoxin with a C‐terminal His tag. The protein was expressed in High Five cells and secreted protein was purified by affinity chromatography followed by size‐exclusion chromatography.
The GAIN domain of rat CL1 (residues Thr533–Ile849) was cloned into a pCMV5 vector and fused to Ig (26 kDa) at the C‐terminus and carried an HA tag at the N‐terminus. The GAIN domain of mouse GPR56 (residues Q115–E395) was cloned into the pDisplay vector with N‐terminal HA and C‐terminal myc tags and has the PDGFR single‐spanning transmembrane domain at the C‐terminus. The full‐length human PKD construct carried N‐terminal GFP and C‐terminal Flag tags. Full‐length human CL3 was synthesized by codon optimization and carried Flag and HA tags at identical positions as in CL1.
Crystallization of CL1 and BAI3
Initial screens for crystallization of CL1 were carried out using 96‐well format kits (Hampton Index, Salt and Crystal Screen HT, Hampton Research) on a Phoenix crystallization robot (Art Robbins Instruments). Automatic plate scanning by a CrystalPro 2 imaging system (TriTek Corporation) identified initial hits in the Wizard 1–2 screen (Emerald Biosystems). After optimization, crystals of CL1 were grown using hanging drop vapour diffusion at 20°C in 24‐well format. In all, 1 μl of 10.7 mg/ml protein in 10 mM HEPES pH 7.2, 150 mM NaCl was mixed with equal volume of mother liquor consisting of 1.26 M (NH4)2SO4, 0.1 M sodium acetate pH 4.5, 0.2 M NaCl, and equilibrated against 1 ml mother liquor. Crystals grew to full size within 3 days. Crystals were cryoprotected by transferring the crystals into mother liquor with 10, 20, and 28% glycerol gradually. The best crystals diffracted to dmin=1.85 Å. Crystals of CL1 formed in space group P3121 with one molecule per asymmetric unit and 67% solvent content.
Crystallization hits for BAI3 were obtained with a 96‐well format kit (Protein Complex, Nextal) at 20°C. In all, 17 mg/ml protein in 10 mM HEPES, 150 mM NaCl pH 7.2 was mixed with equal volume of mother liquor consisting of 15% PEG6000, 0.1 M sodium citrate pH 5.5. Crystals grew to full size within 2 days and were cryoprotected with 25% glycerol. Crystals diffracted to dmin=2.3 Å. Crystals of BAI3 formed in space group C2221 with two molecules per asymmetric unit and 63% solvent content.
Structure determination and refinement of CL1
For the CL1 structure, experimental phase information was obtained from sulphur SAD data. This structure was the first structure that was determined at the Stanford Synchrotron Radiation Lightsource by sulphur SAD phasing using an optimized data collection strategy for a long wavelength at 2 Å, and one of a few de‐novo structures that has been determined by sulphur SAD at this resolution. More than 40 crystals were screened to find the best diffracting large crystal. A test radiation damage experiment was performed on a different crystal to determine the effect of radiation damage and to optimize the parameters for the experiment. Three data sets, each comprising 360° and offset by 20°, were collected on a single long crystal by translating the beam along the long edge of the crystal. The crystal was placed in a helium stream operating at 65 K to minimize background scatter at long wavelength and to further limit radiation damage. Data collection was carried out in inverse beam mode with a wedge angle of 0.5° and an exposure time of 8 s. Total data collection lasted 9 h. The data were collected at a wavelength of 2 Å, which limited the resolution to 2.3 Å (diffraction was observed to 1.8 Å resolution at a wavelength of 1 Å). The resolution limit decreased from 2.3 to 2.7 Å at the end of one 360° data collection as a result of radiation damage. All data sets were indexed, integrated, scaled, and combined in HKL2000 using the scale anomalous and high absorption options (Otwinowski, 1997). The error model was not optimized during scaling. In all, 25 sulphur sites were found (17 sites within the protein and 8 in sulphate ions from the crystallization condition) by SHELXD (Schneider and Sheldrick, 2002). SHELXD used a high‐resolution cutoff of 2.7 Å, at which the I/σ was ∼50. The sulphur sites were refined using SHARP (Vonrhein et al, 2007). Density modification was performed by DM (Cowtan and Zhang, 1999). Among all space groups tested, only space group P3121 yielded easily interpretable electron density maps. The density modified electron density maps showed good connectivity of secondary structure elements and cleavage at the GPS site (Supplementary Figure S7A). The sulphur sites corresponded to the expected side chains when compared within the density modified electron density map. Automatic model building was done with Phenix Resolve (Terwilliger, 2003) for two thirds of the structure. The rest of the model building was carried out manually using Coot (Emsley and Cowtan, 2004), and refinement performed with Phenix.refine (Adams et al, 2010) using the MLHL target function with experimental phases. The resulting model was then refined against the higher resolution native data set at 1.85 Å resolution. The subsequent refinement consisted of alternating rounds of positional minimization, individual restrained thermal factor refinement, both using the MLF maximum likelihood target function, and model building. Water molecules and N‐linked carbohydrates were added. A summary of the data collection and refinement statistics is given in Supplementary Table S1 (PDB accession code 4DLQ).
Structure determination and refinement of BAI3
BAI3 crystals were soaked into three solutions containing the crystallization condition, 0.25 M sodium iodide, and 10, 20, and 30% glycerol immediately before they were frozen in liquid nitrogen. The total soak time was about 1–2 min. Only one in ten crystals survived this treatment. The presence of increased glycerol, thus viscosity, in the soak solutions prevented the crystals from cracking. Data collection was done in inverse beam mode with a wedge angle of 25° and an exposure time of 5.5 s. The derivative crystals were not isomorphous to the native crystals, so the native data set could not be used for isomorphous phasing. The diffraction data were collected at a wavelength of 1.6 Å, resulting in diffraction data to 2.7 Å resolution. Data processing was similar to that of CL1. The BAI3 structure was independently determined with experimental phases obtained by iodine‐SAD data (a density modified map of the iodine‐SAD data around the GPS site is shown in Supplementary Figure S7B, clearly indicating lack of cleavage in contrast to the CL1 structure). SHELXD used a high‐resolution cutoff of 4 Å, at which the I/σ was ∼50, and it found 6 iodine sites. There were two molecules in the asymmetric unit and non‐crystallographic symmetry (NCS) restraints were used during initial refinement cycles using Phenix.refine but were removed at later stages. A summary of the data collection and refinement statistics is given in Supplementary Table S1 (PDB accession code 4DLO).
Cleavage and plasma membrane localization experiments
Antibodies. Mouse monoclonal anti‐HA antibody was from Covance and rabbit polyclonal anti‐Flag antibody was from Sigma.
Mutagenesis and expression constructs. CL1 deletion constructs were engineered by PCR oligonucleotide amplification. pCMV‐CL1 HA Flag was generated as follows: The Flag epitope containing constructs were obtained by subcloning HindIII–XbaI PCR fragment from pCMV‐CL1 into the same site of pCMV‐NL1‐Flag. The Flag epitope was fused at Leu25 of CL1. A single HA epitope was inserted by PCR between Tyr918 and Glu919 in the second extracellular loop of CL1. Deletion constructs were generated using pCMV‐CL1 HA Flag construct as a backbone. Deletion constructs are as follow: pCMV‐CL1‐ΔHormR HA Flag was generated by deleting Phe479–Cys532. pCMV‐CL1‐GPS HA Flag was generated by deleting Phe479–Glu792. pCMV‐CL1‐GPS_HormR HA Flag was generated by deleting Pro535–Glu792. pCMV‐CL1‐ΔSubA HA Flag was generated by deleting Pro535–Val666. pCMV‐IgCL1 LEC was generated by amplifying Met1–Tyr131 PCR fragment and subcloning into EcoRI–SalI of pCMV‐Ig. pCMV‐IgCL1 GAIN was generated by subcloning the PCR‐amplified fragment of Thr533–Ile849 into SacII–SalI of pDisplay then by excising EcoRI–SalI fragment and subcloning into the same sites of pCMV‐Ig. Therefore, the resulting pCMV‐IgCL1 GAINR construct also bears an N‐terminal HA epitope. pCMV IgN1a‐1 and pCMVN1b were described elsewhere (Ushkaryov et al, 1994; Sugita et al, 2001). pCMV‐CL1 T838A and pCMV‐CL1 T838P were generated by PCR mutagenesis. PKD mutagenesis was performed by cutting out the wild‐type human PKD with BsiWI restriction enzyme, pasting the small fragment into a carrier plasmid. After site‐directed mutagenesis was performed, the mutated fragments were subcloned back into the original plasmid. CL3 and CL1 mutagenesis was performed by the Qiagen site‐directed mutagenesis kit.
Cell culture and transfection. For immunoblotting and immunocytochemistry experiments, HEK293T cells were cultured in 6‐well plates until they reached 70–80% confluency. The cells were then transfected using 4 μl of Fugene 6® (Roche Diagnostics) and 2 μg of respective CL1 DNA constructs. For immunoblotting experiments, cells were harvested 48 h post transfection. For immunocytochemistry experiments, the cells were trypsinized 24 h post transfection and splitted into 12‐well plates containing coverslips that have been previously coated with 0.5 mg/ml poly‐l‐lysine in borate buffer. The cells were grown on the coverslips for an additional 24 h before further experiments were done. For recombinant Ig‐protein expression, HEK cells were transfected using calcium phosphate with chloroquine and 20 μg of cDNA corresponding to the various Ig proteins. Media was then harvested 4 days post transfection.
Immunocytochemistry and image acquisition. Cells transfected with CL1 constructs were washed once with PBS and fixed with 4% paraformaldehyde for 10 min on ice. Cells were washed again three times with cold PBS and incubated at room temperature for 30 min in a blocking solution containing 3% BSA in PBS with or without 0.1% Triton X‐100 meaning that cells were permeabilized or not, respectively. Mouse anti‐HA antibody was then added (1:500 ratio) and the incubation was prolonged for another 2 h. Cells were washed three times with blocking solution and incubated for 1 h at room temperature with anti‐mouse Alexa‐488 fluorescent antibody to label CL1 receptors. Cells were finally washed again three times with blocking solution and once with water before mounting on slides using media containing DaPI for nuclear staining. Slides were then analysed by confocal microscopy. Images were acquired using the confocal microscope Leica TCS2. The same confocal acquisition settings were applied to all samples of the experiment. Collected z‐section images were analysed blindly using Leica confocal software.
Immunoblotting procedures. Transfected HEK cells were harvested 2 days after transfection and solubilized for 1 h at 4°C in buffer containing 20 mM Hepes‐NaOH pH 7.4, 0.1 mM EDTA, 150 mM NaCl, 2 mM CaCl2, 2 mM MgCl2 and 1% Triton. Insoluble material was removed by centrifugation. SDS–PAGE sample buffer was added to the supernatant and the resulting sample was loaded on 6% SDS–PAGE. Gels were then transferred onto nitrocellulose membranes and processed using standard procedures. Membranes were probed with a mouse monoclonal anti‐HA antibody or a rabbit polyclonal anti‐Flag antibody followed by Horseradish peroxidase‐coupled secondary antibody, incubated with ECL reagents and revealed on X‐ray films.
Ig‐fusion protein purification and pull‐down experiments
The soluble Neurexin and CL1 Ig‐fusion proteins were affinity purified from transfected HEK293 cell supernatant using protein‐A Sepharose 4 Fast Flow (Amersham Pharmacia Biotech AB, Sweden) to bind the human‐IgG portion and washed to remove unbound proteins. The immobilized Ig‐fusion proteins were then incubated 16 h at 4°C with the eluted 3 μg GST–myc–Latrotoxin in buffer containing 20 mM Hepes buffer, pH 7.4; 0.1 mM EDTA, 150 mM NaCl, 2 mM CaCl2, 2 mM MgCl2 and 1% (v/v) Triton X‐100. The beads were then washed, incubated in boiling SDS–PAGE sample buffer and loaded on 6% SDS–PAGE. SDS–PAGE gels were either transferred onto nitrocellulose membrane and immunoblotted with anti‐myc antibody or stained using Coomassie Brilliant Blue R‐250 (Bio‐Rad).
BAI cleavage experiments
Antibodies Goat antiserum 11509 and rabbit antisera 233C and 234C were generated against a synthetic peptide (N‐CEKAGATIPLVGQDIIDLQTEV‐C; mouse BAI1). Rabbit antisera A324 and A323 were raised against recombinant GST‐fusion proteins (P1447 → V1582 of mouse BAI1 and R1151 → E1348 of human BAI3, respectively). T3743 was raised against GFP in rabbit.
Cell culture HEK293T cells were plated in 6‐well plates. At 70–80% confluency, they were transfected with 1 μg expression plasmid (FuGENE 6 from Roche), and then cultivated for 2 days. Cells were lysed in extraction buffer (20 mM Tris–HCl pH 7.5, 150 mM NaCl, 1% Triton X‐100, protease inhibitors), homogenized, and the soluble fractions were collected after centrifugation.
Immunoprecipitation Two mouse brains were homogenized in 8 ml buffer (25 mM HEPES pH 7.2, 320 mM sucrose, protease inhibitors). After centrifugation at 1500 g for 10 min, supernatant S1 was centrifuged at 150 000 g for 60 min. Pellet P2 was then homogenized in 6 ml lysis buffer (50 mM Tris–HCl pH 7.5, 100 mM NaCl, 1% Triton X‐100, protease inhibitors), and centrifuged again at 150 000 g for 60 min. In all, 3 ml of supernatant S3 was incubated with 18 μl antiserum 11509 for 60 min. Then, 180 μl protein G sepharose (GE Healthcare) was added, and reactions were incubated for 3 h. The beads were washed four times with lysis buffer, and the proteins were eluted in SDS buffer.
Immunoblotting Proteins were resolved on 6% acrylamide gels and transferred onto nitrocellulose membranes (Whatman). After a blocking step, membranes were incubated overnight with the primary antibody diluted in blocking solution (5% milk powder, 2.5% goat or rabbit serum, 0.1% Tween‐20, in TBS). After washing with TBS containing 0.1% Tween‐20, blots were incubated with a peroxidase‐conjugated secondary antibody (Cappel) for 45 min. After washing, immunoreactive bands were detected by ECL (GE Healthcare).
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank E Özkan for crystallographic advice and discussion; M Etherton, X Piao, F Qian for α‐latrotoxin, GPR56, and PKD1 cDNAs. Portions of this research were carried out at the Stanford Synchrotron Radiation Lightsource, a national user facility operated by Stanford University on behalf of the US Department of Energy, Office of Basic Energy Sciences. The SSRL Structural Molecular Biology Program is supported by the Department of Energy, Office of Biological and Environmental Research, and by the National Institutes of Health, National Center for Research Resources, Biomedical Technology Program, and the National Institute of General Medical Sciences. This study was also supported by a postdoctoral fellowship from Canadian Institutes of Health Research (CIHR, to AAB) and a grant from the NIMH (R37 MH52804‐08 to TCS). D Araç is a Howard Hughes Medical Institute Fellow of the Life Sciences Research Foundation.
Author contributions: DA designed and performed experiments and wrote the paper. AAB, DA, MFB, JN, and MS performed experiments. TCS and ATB provided guidance and wrote the paper.
This is an open‐access article distributed under the terms of the Creative Commons Attribution License, which permits distribution, and reproduction in any medium, provided the original author and source are credited. This license does not permit commercial exploitation without specific permission.
- Copyright © 2012 European Molecular Biology Organization