Autocatalytic cleavage of the EMR2 receptor occurs at a conserved G protein-coupled receptor proteolytic site motif.

Post-translational cleavage at the G protein-coupled receptor proteolytic site (GPS) has been demonstrated in many class B2 G protein-coupled receptors as well as other cell surface proteins such as polycystin-1. However, the mechanism of the GPS proteolysis has never been elucidated. Here we have characterized the cleavage of the human EMR2 receptor and identified the molecular mechanism of the proteolytic process at the GPS. Proteolysis at the highly conserved His-Leu downward arrow Ser(518) cleavage site can occur inside the endoplasmic reticulum compartment, resulting in two protein subunits that associate noncovalently as a heterodimer. Site-directed mutagenesis of the P(+1) cleavage site (Ser(518)) shows an absolute requirement of a Ser, Thr, or Cys residue for efficient proteolysis. Substitution of the P(-2) His residue to other amino acids produces slow processing precursor proteins, which spontaneously hydrolyze in a defined cell-free system. Further biochemical characterization indicates that the GPS proteolysis is mediated by an autocatalytic intramolecular reaction similar to that employed by the N-terminal nucleophile hydrolases, which are known to activate themselves by self-catalyzed cis-proteolysis. We propose here that the autoproteolytic cleavage of EMR2 represents a paradigm for the other GPS motif-containing proteins and suggest that these GPS proteins belong to a cell surface receptor subfamily of N-terminal nucleophile hydrolases.

Site-specific limited proteolysis plays an important role in a diverse range of biological processes such as the blood coagulation cascade (1), determination of cell fate (2), ligand-induced receptor activation (3), release of cell-associated growth factors (4), tissue remodeling (5), and apoptosis (6). These activities are usually carried out by specific proteolytic enzymes, some of which are themselves activated by limited proteolysis (7). A detailed understanding of the proteolytic mechanisms is not only critical for the functional studies of these biological processes but may also provide means for possible intervention and regulation.
Of the various modes of proteolytic reactions, self-cata-lyzed proteolysis, or autoproteolysis, has been recognized as an essential step in the proper folding, trafficking, and activation of several endoproteases such as furin (7)(8)(9) and other subtilisin-like proprotein convertases (10) that are involved in the activation of many secretory protein precursors. Autoproteolysis is also involved in the activation of a novel group of hydrolytic enzymes, the N-terminal nucleophile hydrolases (Ntn-hydrolases) 1 (11). The Ntn-hydrolases are activated from an inactive proenzyme by self-mediated hydrolysis of an internal peptide bond via an N 3 O or N 3 S acyl shift between a specific nucleophilic residue and its preceding amino acid (12,13). The newly generated N-terminal nucleophile then acts as a single enzymatic active site to attack its specific protein substrates. One unique feature of the Ntnhydrolases is that a single nucleophilic residue is used as the reactive nucleophile for both the autoproteolytic and enzymatic activity (14 -17). The N-terminal nucleophile is Cys in glucosamine-6-phosphate synthase (18) and asparagine synthase (19); Ser in penicillin acylase (20,21); and Thr in glycosylasparaginase (GA) (14,16,22), the proteasome (23), ␥-glutamyltranspeptidase (24), and Taspase1 (25). In addition to these proteolytic enzymes, several other proteins such as hedgehog proteins (26,27), inteins (28), and nucleoporins (29,30) also belong to the Ntn-hydrolase family.
In recent years, a proteolytic motif known as the G proteincoupled receptor (GPCR) proteolytic site (GPS) (31) has been identified in over 40 cell surface receptors (see, on the World Wide Web, smart.embl-heidelberg.de/). As suggested in its denotation, the GPS motif is primarily found in members of the class B2 GPCRs (32) or the LNB-TM7 receptors (33) that contain a large N-terminal cell adhesion-like extracellular domain coupled to a secretin receptor-like seven-pass transmembrane (TM7) domain. Examples include Flamingo (34), latrophilin (31,35), Ig-hepta (36), HE-6 (37), and the EGF-TM7 receptors (38). However, the GPS motif is not exclusively restricted to the TM7 proteins. Receptors with one-or 11-pass TM configurations such as suREJ1 (39), suREJ3 (40), and polycystin-1 (41,42) also contain the consensus GPS motif, which is characterized by a Cys-rich segment of approximate 50 amino acids located proximal to the first TM domain. Proteolytic cleavage at the GPS motif generates an extracellular (␣) and a TM (␤) subunit that associate noncovalently on the cell surface as a heterodimer. This has led to a notion that these adhesion GPCRs might couple extracellular adhesion events to intracellular signaling via the ␣and ␤-subunits, respectively (43,44). Furthermore, the highly conserved GPS motif-associated proteolysis suggests that this unique cleavage process is likely to be mediated by a common proteolytic mechanism and might be important for the function or regulation of the receptor. Indeed, the proteolysis of polycystin-1 has been found to be essential for its normal biological activity, since several autosomal dominant polycystic kidney disease-associated point mutations were shown to disrupt the GPS cleavage and the function of polycystin-1 (41).
The epidermal growth factor (EGF)-like module containing mucin-like hormone receptor 2 (EMR2) is a human myeloidrestricted EGF-TM7 receptor whose extracellular domain consists of tandem repeats of EGF-like modules followed by a Ser/Thr-rich stalk and a GPS motif (45,46). Our previous studies on the proteolysis of EMR2 have located the precise cleavage site to a conserved tripeptide (His-Leu2Ser 518 ) sequence and demonstrated that the cleavage requires not only the GPS motif but also other extracellular domains in the stalk region (47). In the present study, we elucidate the molecular basis for the cleavage of EMR2 and demonstrate that no protease is required for the proteolytic reaction. Instead, we show that EMR2 is cleaved by a self-catalyzed process characteristic of the autoproteolytic reaction commonly employed by the Ntnhydrolases (11).
Generation of the EMR2 Expression Constructs and EMR2 Fusion Proteins-The EMR2 fusion proteins employed in this report are depicted in Fig. 1A. All expression constructs were generated using standard molecular biology methods. In brief, the cDNA fragments encoding the EMR2 extracellular domain or the full-length EMR2 protein were subcloned in frame into appropriate expression vectors upstream of the protein tags via selected restriction sites. The vectors used are pcDNA3.1/myc-HIS (Invitrogen), pEGFP-N1 (Clontech), and pcDNA3.1/mFc vector as previously described (43,44). The EMR2 sitedirected mutants were made according to the protocols suggested by the manufacturer (GeneEditor Mutagenesis System; Promega). For the construction of the endoplasmic reticulum (ER)-restricted expression vector, a cDNA fragment encoding the KDEL ER retention signal was amplified by PCR using pCMV/myc/ER (Clontech) as a template. The cDNA fragment was then subcloned immediately after the EMR2-EGFP sequence. All expression constructs were subjected to DNA sequencing to confirm their identities. EMR2 fusion proteins were produced by transient transfection of cells. 48 -72 h post-transfection, the EMR2 fusion protein was collected from conditioned medium (CM) or total cell lysate (CL) of transfected cells. Briefly, CM was spun at 2,000 rpm at 4°C for 20 min followed by 100,000 rpm at 4°C for 20 min. The supernatant was collected and stored at Ϫ80°C. Total cell lysates were collected in cell lysis buffer (20 mM Tris-HCl, pH 7.4, 0.5% Nonidet P-40, 5 mM MgCl 2 , 100 mM NaCl, 1 mM sodium orthovanadate, 1 mM AEBSF, 5 mM Levamisole, 1ϫ complete TM (Roche Applied Science) protease inhibitors) at 4°C. Protein concentration was determined by a D c Protein Analysis Kit (Bio-Rad). For the purification of soluble EMR2-mFc fusion proteins, human embryonic kidney 293T cells were transfected with 40 g of DNA/175-cm 2 flask using calcium phosphate precipitation as previously described (43,44). The medium was replaced with 25 ml of serum-free Opti-MEM I 16 -18 h post transfection and incubated for a further 72 h. Conditioned medium was collected, spun, and passed through a 0.45-m filter, followed by Protein A-Sepharose 4 Fast Flow (Amersham Biosciences) column purification as previously described (43,44).
Immunoprecipitation, Western Blotting, and Other Protein Analysis-EMR2 fusion proteins were immunoprecipitated from CM or CL using appropriate Abs and/or protein A/G beads. Briefly, CM (1 ml) and CL (100 g) were either incubated with protein A/G-Sepharose directly or precleared with irrelevant Ab and protein A/G-Sepharose, followed by subsequent incubation with appropriate primary Ab and protein A/G-Sepharose, respectively. After extensive washes, the immunopurified proteins were subjected to in vitro cleavage reaction or glycosidase treatment as described. For the glycosidase treatment, the proteins were incubated with 1 unit of PNGase F (Roche Applied Science), 1 unit of endoglycosidase H (Roche Applied Science), or 0.5 milliunits of Oglycosidase (Roche Applied Science) plus 1.0 milliunit of neuraminidase in 20 mM sodium phosphate buffer, pH 7.0, at 37°C for 20 h prior to Western blot analysis. For Western blotting, proteins were denatured in reducing sample buffer, subjected to electrophoresis in 8 or 10% SDS-PAGE gels, transferred to polyvinylidene difluoride membranes (Immobilon-P; Millipore Corp.), and probed with 2A1 mAb, anti-Myc, or anti-GFP Ab. Following extensive washes, the blots were incubated with appropriate horseradish peroxidase-conjugated second Ab for ECL detection (Amersham Biosciences). The fluorescence intensity of EMR2-GFP fusion proteins was determined by a FLUOstar Galaxy fluorescence plate reader (BMG LabTechnologies Ltd., Aylesbury, UK) using an excitation wavelength at 485 nm and emission wavelength at 520 nm.
Immunofluorescence Confocal Microscopy-Transfected cells grown on glass coverslips in 24-well tissue culture plates were fixed with 4% paraformaldehyde in phosphate-buffered saline, blocked, and permeabilized in blocking buffer (phosphate-buffered saline with 0.5% bovine serum albumin, 0.1% Triton X-100, and 1% normal donkey or goat serum) for 20 min at room temperature. Cells were then incubated sequentially for 1 h at room temperature with primary antibodies (5-10 g/ml) and appropriate secondary antibody (5-10 g/ml) diluted in the same blocking buffer with extensive washing in between incubations. Cells were then mounted onto glass slides with fluorescent mounting medium (Dako, Cambridgeshire, UK). Immunofluorescence was analyzed on a Bio-Rad Radiance 2000 laser-scanning confocal microscope. The resulting images were processed in Adobe® Photoshop® 6.0.
In Vitro Cleavage of EMR2 Proteins-Immunoprecipitated EMR2 fusion proteins or those in CM and CL were incubated in cleavage buffer (50 mM Tris, pH 7.5, 20 mM NaCl, 1 mM EDTA) with or without 250 mM NH 2 OH at 37°C unless otherwise specified. At various time points, samples were withdrawn and analyzed by Western blotting. For the biochemical characterization of EMR2 autoproteolysis, samples were incubated in the cleavage buffer containing protease inhibitors or other test reagents such as EDTA, as indicated in the throughout. For the demonstration of intramolecular cleavage, EMR2-H516S-mFc fusion protein was first purified by Protein A chromatography as described above. Purified proteins were then incubated at different concentrations (0.2 and 1.0 mg/ml) in cleavage buffer alone at 37°C. At various time points, samples were withdrawn, subjected to SDS-PAGE, and stained in Simply Blue™ Safe-Stain (Invitrogen). The intensity of the uncleaved precursor protein band (ϳ110 kDa) was determined from the image captured by a Gel Doc 2000 gel documentation system (Bio-Rad).

RESULTS AND DISCUSSION
Proteolytic Cleavage at the Highly Conserved GPS Motif Occurs in the ER-We and others have previously shown that the proteolytic cleavage of the GPS motif-containing receptors (GPS-receptors) is a TM-independent process and that the GPS motif is necessary but not sufficient for proteolysis to occur (41,47,49). To further investigate this unique proteolytic process, we first compared the GPS motifs of all GPS-receptors that are known to be processed (Fig. 1B). The GPS motif is evolutionary conserved and widely present in cell surface receptors, including members of the LNB-TM7 or class B2 GPCRs as well as receptors with one-or 11-pass TM configuration such as suREJ1 (39), polycystin-1 (41,42), and suREJ3 (40). The GPS motif is always located at the membrane-proximal region, ϳ20 -30 residues from the first TM domain. The cleavage site tripeptide is highly conserved: the P Ϫ2 residue is His, the P Ϫ1 residue is Leu (or Ile in the Drosophila Flamingo protein (34)), and the P ϩ1 residue is either Ser or Thr. N-terminal to the cleavage site are two invariable Trp residues and four constrained Cys residues believed to form two intramolecular disulfide bridges. One exception to the rule is polycystin-1 that contains only two Cys residues. C-terminal to the cleavage site are 6 -8 small, hydrophobic residues that have been shown to be important both for proteolysis and noncovalent association of the cleaved subunits (47). Overall, these features indicate that there is an ordered and complex structure surrounding the GPS cleavage site and suggest that all GPS-receptors probably undergo the same proteolytic process.
To unveil the proteolytic mechanism at the GPS, we first investigated the subcellular compartment in which the proteolysis takes place. Previous pulse-chase experiments examining the cleavage of CD97, latrophilin/CL1 (calcium-independent receptor for latrotoxin (CIRL)/latrophilin), ETL, and Ig-Hepta have shown that the GPS proteolysis occurs very early (within 10 -15 min) during protein biogenesis and suggested that it might occur in the ER (36, 49 -51). To further confirm and locate the cleavage reaction in the ER, we took advantage of the specific ER retention signal, KDEL, and examined the proteolysis of KDEL-tagged EMR2-enhanced green fluorescence protein (EGFP) fusion proteins (Fig. 2). In addition to the wild type (WT) stalk, two other stalks containing a cleavage site-deficient S518A and a control S519A point mutation, respectively, were also used to demonstrate the specificity of the cleavage reaction.
The KDEL-tagged fusion proteins were confirmed to localize in the ER by the following observations. First, whereas the soluble EMR2-EGFP fusion proteins (with no KDEL signal) were detected in both CM and the total CL, the KDEL-tagged fusion proteins were found only in the CL ( Fig. 2A). Thus, the KDEL-tagged fusion proteins were expressed but retained inside the cells. Confocal immunofluorescence staining subsequently showed that the KDEL-tagged proteins co-localize with the ER-lumen resident proteins such as protein-disulfide isomerase and calreticulin ( Fig. 2B) (data not shown). On the other hand, they showed only minimum co-localization with ERGIC-53, a mannose-specific membrane lectin involved in the transport of glycoproteins from the ER to the ER-Golgi intermediate compartment (52). No co-localization of the KDELtagged proteins with mannose 6-phosphate receptor that recycles between the trans-Golgi network and endosomes was found (53) (Fig. 2B). Finally, since EMR2 is heavily glycosylated, the KDEL-tagged EMR2 proteins were subjected to glycosidase digestion. N-Linked glycosylation is initiated in the ER lumen as a high mannosyl oligosaccharide, which is then FIG. 1. A, diagrams depicting the EMR2 proteins used in this report. From the N terminus, the EGF-like modules are represented by triangles with numbers, followed by a black line representing the stalk region. The GPS cleavage site is indicated by an arrow. The EGFP is depicted by a shaded oval, whereas the mFc fragment is shown as a dimer of two circles. The TM regions are depicted as zig-zag lines. B, sequence alignments of the GPS motifs of the GPS-receptors that are known to be cleaved. The conserved residues are highlighted with a gray background. The arrow and asterisk indicate the cleavage site and the two Cys residues absent in polycystin-1, respectively. The abbreviations representing animal species are as follows. h, human; m, mouse; r, rat; d, Drosophila; su, sea urchin. modified to a complex form in the cis-Golgi compartment. Peptide:N-glycosidase F recognizes and digests N-glycosylated proteins between Asn and GlcNAc, whereas endoglycosidase H cleaves only the high mannose oligosaccharides. O-Glycosylation mainly takes place in the cis-Golgi compartment, so sensitivity to O-glycosidase digestion could also help indicate the subcellular localization of proteins. Western blot analysis showed that the KDEL-tagged proteins were sensitive to both peptide:N-glycosidase F and endoglycosidase H but were resistant to O-glycosidase digestion, indicating that these proteins have not trafficked out of the ER compartment (Fig. 2C).
When analyzed for proteolytic cleavage, the KDEL-tagged EMR2-EGFP fusion protein containing the WT stalk or a control S519A point mutant stalk was shown to be effectively cleaved to two subunits, whereas the cleavage site-deficient (S518A) stalk did not (Fig. 2C). The same results were observed in several cell lines including CHO-K1, COS-7, human embryonic kidney 293T, and NIH3T3 (data not shown). Together, these and earlier results indicate that the GPS proteolytic cleavage is likely to be carried out by a conserved proteolytic machinery in the ER.
EMR2 Proteolysis Is an Autoproteolytic Reaction-In order to identify the proposed ER-located proteolytic machinery, we first aimed to define the GPS cleavage site specificity. EMR2-mFc fusion proteins provide an efficient way for specific purification and detection of the protein and have been used previously to characterize EMR2 proteolysis (47). Thus, a series of the EMR2-mFc fusion proteins were generated, where the Ser 518 cleavage site residue was individually mutated to 19 other amino acids. As shown in Table I, proteolytic cleavage was detected in only three point mutants; the S518C and S518T mutants displayed the same efficient proteolysis (ϳ100%) as the WT protein, whereas the S518K mutant showed only a partial cleavage effect (ϳ5-10%). All 16 other Ser 518 mutants failed to undergo cleavage. A control mutant, S519A, was cleaved with the same efficiency as that of the WT protein (Table I). This indicated that the GPS proteolytic machinery only recognizes three specific cleavage site residues (Ser, Thr, and Cys).
While searching for such cleavage site-specific proteases, a novel autoproteolytic process employed by inteins, hedgehog proteins, and other Ntn-hydrolases was noted (12,54). Similar to the cleavage of EMR2 at the GPS, hedgehog and Ntn-hydrolases are cleaved at an internal peptide bond immediately preceding a Cys, Ser, or Thr residue. Furthermore, as found in the consensus GPS motif, the His residue at the P Ϫ2 position is highly conserved in most Ntn-hydrolases and has been shown to be essential for the deprotonation of the nucleophilic Cys, Ser, or Thr residue in the autoproteolytic reaction. The finding that EMR2 proteolysis follows the same requirement for the nucleophilic Cys/Ser/Thr residues at the P ϩ1 site and shares a conserved P Ϫ2 His residue as the Ntn-hydrolases suggests that the GPS cleavage might be mediated by a similar autoproteolytic mechanism. Previous studies by Guan et al. has shown that a point mutation of the P Ϫ2 His residue dramatically reduces the autoproteolytic reaction rate of GA (55). If a similar slow processing EMR2 mutant protein could be generated, it would be possible to characterize the proteolytic reaction in defined conditions and test the hypothesis of autoproteolysis. To examine the involvement of the P Ϫ2 His 516 residue in EMR2 proteolysis, five point mutants (H516A/N/Q/R/S) were generated. Similar to the GA His mutants, all five EMR2 His 516 mutants were produced as an unprocessed single chain protein, suggesting that His 516 is important in promoting EMR2 cleavage (Table I).
The slow activation of the GA His mutant is due to the inefficient N 3 O acyl shift that produces the ester intermediate. However, the processing rate of this mutant can be greatly enhanced by adding a strong nucleophile, hydroxylamine (NH 2 OH, HA) (55). HA, which is too weak to attack amides but highly reactive against (thio)esters, functions by facilitating the hydrolysis of the (thio)ester intermediate, which is the rate-limiting step in the autoproteolytic reaction (55). Interestingly, when the five immunopurified EMR2 His 516 mutants were treated with 0.25 M HA, all were cleaved, generating two protein fragments similar in size to those produced by the WT protein (Fig. 3A). HA treatment did not promote the cleavage of the S518A mutant, suggesting a fundamental difference between the His 516 and Ser 518 mutants. N-terminal sequencing of the ␤-subunit derived from the HA-assisted cleavage of the H516S mutant showed a precise match to the cleavage site of the WT protein (SSFAVLM, data not shown). This indicates that during the proteolysis of EMR2, His 516 is indeed involved in the formation of a HA-susceptible ester intermediate, resulting from the N 3 O acyl shift between Ser 518 and its preceding residue, Leu 517 .
The enhanced cleavage of the His 516 mutants by HA was only partial when treated for 6 h. This prompted us to look at the effect of HA on the EMR2 proteolytic rate. As shown in Fig. 3B, the purified H516R mutant underwent spontaneous cleavage at 37°C at a very slow rate with a half activation time () longer than 24 h. The addition of HA greatly increased the reaction rate, reducing to ϳ6 -8 h. Thus, the main function of HA is to accelerate the hydrolysis of the H516R mutant. Similar findings were also observed for other His 516 but not Ser 518 mutants (data not shown). These results are in good agreement with the autoproteolytic mechanism of GA and other Ntnhydrolases, where the rate-limiting step of the cleavage reaction is the hydrolysis of the ester intermediates. Due to the striking similarities between the proteolysis of EMR2 and the Ntn-hydrolases and because all EMR2-mFc mutants used here were immunopurified with no known protein contamination, we conclude that EMR2 proteolysis is mediated by a Ntnhydrolase type autocatalytic reaction.
To verify that the autoproteolytic reaction described above was indeed utilized by the full-length EMR2 receptor and was not due to any artifact of the recombinant mFc fusion proteins, site-directed mutants of the full-length EMR2 receptor were generated (Fig. 1A). A c-myc epitope was tagged after the TM7 domain to facilitate the purification and detection of the TM ␤-subunit. As expected, the WT and the S519A mutant TM proteins were efficiently processed, whereas the S518A mutant remained uncleaved, even in the presence of HA (Fig. 4). The H516S mutant TM protein showed an inefficient cleavage effect, which was enhanced by the addition of HA (Fig. 4). This indicates that both EMR2-mFc and TM proteins utilize the same autoproteolytic mechanism and explains the TM-independent characteristic of the GPS proteolysis. The soluble EMR2-mFc or EMR2-EGFP fusion proteins were thus used in all following experiments to further characterize and verify the autoproteolytic mechanism.
Biochemical Characterization of the EMR2 Autoproteolysis-It has been shown that the N 3 O (or N 3 S) acyl arrangement for the formation of the (thio)ester intermediate in Ntn-hydrolases is a reversible reaction and that the reaction equilibrium favors the peptide bond formation via a O 3 N (or S 3 N) reverse acyl shift (30,55). This is because the free amino group derived from the N 3 O (or N 3 S) acyl shift can not diffuse from the catalytic site and is ready to attack the ester carbonyl to restore the peptide linkage. When the EMR2 H516R mutant, either untreated or denatured first with 1% SDS, was incubated at 37°C for 2 h in the absence of HA, no or very little proteolytic processing was observed (Fig. 5A, lanes 1 and 2). However, when incubated in the presence of HA, SDS-untreated samples were readily processed, whereas SDS-denatured samples remained uncleaved (Fig. 5A, lanes 3 and 4). This indicates that no detectable ester intermediate existed in the protein samples before incubation. This was further confirmed by incubating the samples at 37°C for 2 h, followed by treatment with either HA alone or HA plus 1% SDS for a further 2 h (Fig. 5A, lanes 5 and 6). Again, proteolysis of samples was only observed when treated with HA alone but not with HA and SDS. Since the overall rate of autoproteolysis is determined by the N 3 O (or N 3 S) shift rate and the reverse O 3 N (or S 3 N) shift rate, as well as the hydrolytic rate of the (thio)ester intermediate, the most likely explanation for this finding is that although they are able to proceed through N 3 O acyl shift, the His 516 mutants produced very little ester intermediate due to the favorable O 3 N reverse shift to form a peptide bond. Thus, the end result is a very slow N 3 O acyl shift. Only when a strong nucleophile was present to facilitate the hydrolysis of the intermediate could the entire reaction be shifted toward the direction of hydrolysis. The Ser 518 mutants, on the other hand, could not initiate the first step to form ester intermediates due to the lack of the nucleophilic side chain and thus failed to be hydrolyzed by HA.
Next, the temperature effect of the HA-assisted proteolysis was examined. We found that the proteolysis can be observed even at 0°C, albeit less efficiently than at 30 or 37°C (Fig. 5B). The HA-assisted proteolysis at 0°C further ascertained that this is an autoproteolytic reaction, since the involvement of exogenous protease(s) in such condition seems highly unlikely.
Yes ϩ a Cleavage of the EMR2 mutants was determined by Western blotting. Partial, approximately 10% of the S518K mutant was cleaved.
b The expression level of the EMR2 mutants was determined by Western blotting in comparison with that of the WT protein. We next examined the effect of pH on autoproteolysis and observed that the His 516 mutants could be efficiently cleaved in alkali but not acidic conditions, even in the absence of HA (Fig.  5C). The order of the pH effect on proteolysis is pH 11 Ͼ pH 9 Ͼ pH 8 Ͼ pH 7. The cleavage efficiency at pH 11 without HA (Fig.  5C, lane 13) is nearly as good as that at pH 7 with HA (Fig. 5C,  lane 8). This suggests that the rate-limiting hydrolytic step is much more favorable in alkali conditions, where water is more nucleophilic. It will be of interest to know whether other Ntnhydrolases behave similarly in the alkali condition. Other biochemical characteristics of the autoproteolytic reactions are summarized below. Thus, the most potent HA concentration for FIG. 4. The cleavage of the full-length EMR2 protein is an autoproteolytic reaction. Cells were transfected with constructs encoding the full-length EMR2 receptor protein that was tagged with a c-myc epitope at the end of the TM7 segment. EMR2 proteins with the WT or point mutant stalks (H516S, S518A, and S519A) were immunopurified from CL (100 g) using anti-c-Myc-conjugated agarose beads, subjected to in vitro cleavage reaction as indicated, and probed with 2A1 or anti-Myc Ab. For simplicity, only the TM7 ␤-subunit (ϳ28 kDa) was shown in the anti-Myc blotting, which was overexposed to show inefficient cleavage of the H516S mutant as well as the noncleavage effect of the S518A mutant (lower panel). IP, immunoprecipitation; IB, immunoblot.
FIG. 5. Biochemical characterization of the EMR2 autoproteolysis. The purified EMR2-H516R-mFc protein was used in all of the following experiments as a representative of other EMR2 His 516 mutants. The same results were obtained from all EMR2 His 516 mutants. A, Western blot analysis of the EMR2-H516R-mFc protein with no treatment (lanes 1 and 3) or denatured first with 1% SDS (lanes 2 and  4) and then incubated in the absence (lanes 1 and 2) or the presence (lanes 3 and 4) of NH 2 OH at 37°C for 2 h. In lanes 5 and 6, samples were incubated for 2 h first and then either untreated (lane 5) or denatured with 1% SDS (lane 6) in the presence of NH 2 OH and incubated for a further 2 h. B, samples were incubated overnight in the absence or presence of NH 2 OH at different temperatures as indicated. C, samples were incubated overnight in the absence or presence of NH 2 OH at different pH values as indicated. In A, the blot was probed with 2A1 followed by horseradish peroxidase-conjugated anti-mFc to reveal the uncleaved precursor protein (ϳ110 kDa) as well as the cleaved ␣-(ϳ70 -75 kDa) and ␤-subunits (ϳ38 kDa). In B and C, the blotting was probed with horseradish peroxidase-conjugated anti-mFc to detect only the uncleaved precursor protein (ϳ110 kDa) and the cleaved ␤-subunit (ϳ38 kDa) for simplicity.

FIG. 3. EMR2 proteolysis is an autoproteolytic reaction.
A, Western blot analysis of the EMR2-mFc fusion protein with the WT or point mutant stalks. Proteins were immunopurified from CM (1 ml) using Protein A-agarose beads, subjected to in vitro cleavage reaction for 6 h as indicated. Blots were probed with 2A1, followed by horseradish peroxidase-conjugated anti-mouse Fc-specific Ab. Thus, the uncleaved precursor protein (ϳ110 kDa), the cleaved EMR2 extracellular ␣-subunit (ϳ70 -75 kDa), and the mFc ␤-subunit (ϳ38 kDa) were detected. B, purified EMR2-H516R-mFc proteins were incubated in the cleavage buffer in the presence or absence of NH 2 OH at 37°C. Samples were withdrawn at the indicated time points for analysis as described above.
Autoproteolytic Processing of EMR2 Is an Intramolecular Reaction-Autoproteolytic cleavage can be an intermolecular or intramolecular event. Studies on the autoproteolysis of Ntnhydrolases have shown that they proceed as an intramolecular reaction (22,24). To determine whether the same is true for EMR2 proteolysis, two independent experiments were carried out. First, the reaction kinetics of the spontaneous hydrolysis of the H516S mutant was examined. As shown in Fig. 6, A and B, the half-life of the EMR2 (125)-H516S-mFc precursor, ϳ24 h, is independent of the starting concentrations (0.2 or 1.0 mg/ml) of the precursor proteins. This is characteristic of a first order reaction and strongly suggests that the autoproteolytic reaction is an intramolecular event. To further confirm this, cells were co-transfected with EMR2 (1-5)-WT-EGFP and EMR2 (125)-H516Q-mFc constructs. EMR2 (1-5)-WT-EGFP, containing five EGF-like domains and a WT stalk, is active in proteolytic processing. On the contrary, the slow processing EMR2 (125)-H516Q-mFc containing alternatively spliced three EGFlike domains and a mutant stalk (H516Q), is capable of efficient cleavage only in the presence of HA. Proteolytic cleavage of these two proteins can be easily detected by the C-terminal epitope tags and the sizes of the extracellular ␣-subunit. If the cleavage reaction is an intermolecular event, the EMR2 (1-5)-WT-EGFP should be able to cleave the EMR2 (125)-H516Q-mFc protein and produce two bands corresponding to the ␣ (three EGF-like domains plus stalk) and ␤ (the mFc) subunits, respectively. Fig. 7 showed that in all co-transfection conditions tested, the EMR2 (1-5)-WT-EGFP was fully cleaved, but the proteolysis of the EMR2 (125)-H516Q-mFc protein was only achieved by the addition of HA. In the absence of HA, the co-expression of EMR2 (1-5)-WT-EGFP, even in a relatively higher concentration, did not cause cleavage of the EMR2 (125)-H516Q-mFc protein. Thus, the processing-efficient WT stalk cannot cleave the otherwise processable H516Q mutant stalk, which indicates that the proteolysis of EMR2 is indeed an intramolecular reaction.
Based upon the present and previous data and the similarity to the autoproteolytic mechanism of Ntn-hydrolases, the proposed molecular mechanism of EMR2 cleavage is depicted in Fig. 8. After translation and translocation into the ER, the newly synthesized EMR2 protein is folded properly into a specific "cleavage" conformation, where the reactive hydroxyl group of Ser 518 is close enough to be deprotonated by His 516 . This is followed immediately by a nucleophilic attack in cis on the ␣-carbonyl carbon of Leu 517 to form a transitional tetrahedral intermediate. An ester intermediate is then formed via an N 3 O acyl shift. An attack by water finally hydrolyzes the ester bond and produces two polypeptide fragments (Fig. 8). Although only the cleavage site tripeptide was shown in this model for simplicity, it is evident from previous studies that other residues are certainly involved in forming the specific "cleavage" conformation ( Fig. 8). From this model, it is not known how the two subunits are held noncovalently together, another characteristic shared by EMR2 and many Ntn-hydrolases. Future studies on the structure of the EMR2 molecule, especially the stalk region, should reveal the detail of this intriguing proteolytic mechanism.
The Potential Functional Significance of the Autoproteolytic Cleavage at the GPS-Results from data base searches indicate that the GPS-receptors are ubiquitously expressed in animal species (see, on the World Wide Web, smart.embl-heidelberg.de/). Intriguingly, however, the GPS motif has not yet been found in proteins of lower eukaryotes such as yeast and of prokaryotes such as bacteria, suggesting a potential functional role for these receptors in multicellular organisms. The GPS proteolytic cleavage seems to be an inherent part of these receptors, since the GPS motif is almost always associated with the class B2 TM7 domain. This suggests that the GPS motif and the class B2 GPCR have co-evolved, probably by an early exon-shuffling event. Therefore, the GPS cleavage might be important for the function/regulation of these receptors during evolution.
Earlier studies of the processing of latrophilin/CL-1 by Krasnoperov et al. (49) have suggested that proper cleavage of CL-1 at the GPS might be a prerequisite for efficient receptor trafficking to the cell surface. To determine whether the same is true for EMR2, the expression levels of all EMR2 point mutants were compared. Interestingly, no consistent relationship between proteolytic cleavage and protein expression could be found (Table I). For example, both S518T and S518C mutants were efficiently cleaved, but the S518C mutant was expressed at a much lower level than the S518T mutant and the wild type protein. Similarly, some but not all uncleaved Ser 518 mutants were expressed at levels comparable with that of the wild type protein. More importantly, the same finding was observed in both mFc fusion proteins and the TM7 proteins (data not shown). Since all expression constructs are the same except for the point mutations, we conclude that the proteolytic cleavage of EMR2 protein per se does not play a role in the regulation of receptor expression. The differences in protein expression most likely are due to the conformational stability of the individual proteins. This conclusion is consistent with the GPS cleavage of polycystin-1, in which point mutations in the receptor for egg jelly (REJ) domain next to the GPS motif affect receptor cleavage but not its cell surface expression (41).
To date, the best example linking the GPS cleavage and the FIG. 8. The proposed autoproteolytic mechanism for the cleavage of EMR2 at GPS. The newly synthesized EMR2 receptor is translocated into the ER lumen, where a correctly folded conformation is achieved, allowing the autoproteolytic reaction to proceed to produce a mature heterodimeric receptor. The EGF-like modules are represented by small triangles, the stalk region by a pair of half-ovals, the GPS motif by a triangle with two disulfide bonds, and the TM regions by zig-zag lines. The hypothesized "cleavage" conformation introduces a tight bend at the GPS cleavage site, where the autoproteolysis occurs. The autoproteolytic reaction is shown inside a box between the precursor and the mature protein. For simplicity, only the cleavage tripeptide is shown. The arrow represents electron transfer and nucleophilic attack. His 516 is the proton donor/acceptor for the generation of the tetrahedral intermediate. The formation of ester intermediate is derived from the cleavage of the C-N bond through the protonation of the amino group of Ser 518 (N 3 O acyl shift). Hydrolysis of the ester bond by water produces a carbonyl group on Leu 517 and a hydroxyl group on Ser 518 . receptor function is from the study of polycystin-1 (41), the product of the PKD1 gene whose mutation is responsible for the major form of the autosomal dominant kidney disease in humans. Qian et al. (41) have clearly demonstrated that polycystin-1 is cleaved at the GPS and that the REJ domain N-terminal to the GPS motif is required for proteolysis. Most importantly, they found that the GPS cleavage is essential for the biological functions of polycystin-1, since it is disrupted by some mutations associated with the autosomal dominant polycystic kidney disease. Polycystin-1 has been found to mediate Ca 2ϩ influx by interacting with polycystin-2 on the cell surface (56). It is possible that the cleavage of polycystin-1 at the GPS leads to a conformational change of the TM and the cytoplasmic domains, which in turn influences the ability of polycystin-1 to co-assemble with polycystin-2 to form a calcium-permeable nonselective ion channel. On this note, it is interesting that both latrophilin/CL-1 and suREJ1 have also been shown to support Ca 2ϩ influx, although it is not known whether the GPS cleavage is required for this function (57,58). We have recently shown that both EMR2 and CD97 can act as an adhesion molecule, capable of binding to the cognate ligands on the cell surface (44,48). However, attempts to reveal intracellular signaling, including Ca 2ϩ influx, have been mostly unsuccessful. 2 Therefore, it remains unknown at present whether the cleavage of EMR2 and CD97 at the GPS can trigger Ca 2ϩ influx or any other signaling events.
By virtue of the similarities between the GPS-receptors and the Ntn-hydrolases, it is possible that these receptors might also possess similar enzymatic functions. The majority of the Ntn enzymes are amidases with unique and specific protein substrates. For example, GA specifically hydrolyzes the amide bond between Asn and N-acetylglucosamine (59), ␥-glutamyltranspeptidase is involved in glutathione metabolism (24), and Taspase1 is responsible for the cleavage and activation of the mixed lineage leukemia (MLL) gene product that is frequently disrupted in human infant leukemia (25). The identification of the Ntn enzyme type autoproteolytic mechanism described here certainly merits further investigation to examine whether the GPS-receptors can function as hydrolytic enzymes.
In addition to the potential functions of the receptors, the detailed mechanical framework of the autoproteolytic reaction also deserves further attention. Although the autoproteolysis of most Ntn-hydrolases can be recapitulated in heterologous expression systems such as Escherichia coli, attempts to demonstrate the same EMR2 autoproteolytic reaction in E. coli as well as in in vitro transcription and translation systems have been unfruitful. 2 This not only reflects the fact that the GPS proteolysis is identified only in receptors of higher eukaryotes but also is suggestive of the involvement of a multifactorial mechanism. As the extracellular stalk of EMR2 contains multiple potential glycosylation sites and disulfide bridges, it is possible that additional protein modification steps are required for the GPS autoproteolysis. Recently, it has been shown that the autoproteolysis of GA is preceded by the dimerization of the precursor protein in the ER (17,60,61). It will be of great interest to determine whether the same is true for EMR2 and other GPS-receptors.
In summary, we have characterized the proteolytic cleavage of EMR2 receptor at the GPS motif and have presented strong evidence demonstrating that it is an intramolecular autoproteolytic reaction similar to the cis-autoproteolysis of Ntn-hydrolases. We suggest that all GPS-receptors will undergo the same autoproteolytic reaction and that they form a novel cell surface subfamily of the Ntn-hydrolase clan.