Protein-tyrosine Phosphatase- (cid:1) Is a Novel Member of the Functional Family of (cid:2) -Latrotoxin Receptors*

Receptor-like protein-tyrosine phosphatase sigma (PTP (cid:1) ) is essential for neuronal development and function. Here we report that PTP (cid:1) is a target of (cid:2) -latro-toxin, a strong stimulator of neuronal exocytosis. (cid:2) -La-trotoxin binds to the cell adhesion-like extracellular region of PTP (cid:1) . This binding results in the stimulation of exocytosis. The toxin-binding site is located in the C-terminal part of the PTP (cid:1) ectodomain and includes two fibronectin type III repeats. The intracellular catalytic domains of PTP (cid:1) are not required for the (cid:2) -latro-toxin binding and secretory response triggered by the toxin in chromaffin cells. These features of PTP (cid:1) resem-ble two other previously described (cid:2) -latrotoxin receptors, neurexin and CIRL. Thus, (cid:2) -latrotoxin represents an unusual example of the neurotoxin that has three independent, equally potent, and yet structurally dis-tinct

␣-Latrotoxin stimulates spontaneous neurotransmitter release by massive synaptic vesicle exocytosis (1). The first steps in the sequence of ␣-latrotoxin action include extracellular binding of the toxin to cell surface membrane proteins called ␣-latrotoxin receptors (2,3). Once bound, toxin molecules interact with the lipid bilayer that results in their insertion into the membrane and formation of cation-permeable pores (4,5). The cation fluxes are responsible for at least a part of the ␣-latrotoxin effects (6). In addition, the inserted fragment of the toxin molecule may interact with some as yet unidentified intracellular effectors coupled to the exocytosis machinery (7).
The first studies of ␣-latrotoxin receptors demonstrated that they interact with the toxin with high affinity, that their density is low, and that in the neuro-muscular junctions they co-localize with the active zones of the nerve terminals (8,9). Therefore, the ␣-latrotoxin receptors have been long viewed as specific markers for the presynaptic release sites. Two neuronal ␣-latrotoxin-binding proteins were described previously, neurexin (10,11) and CIRL 1 (calcium-independent receptor of ␣-latrotoxin), also called latrophilin, lectomedin, and CL (12,13,14). Although they do not share any sequence homology, either of them can serve as a functionally active receptor of ␣-latrotoxin (12,15). Their signal-transducing membranes and intracellular domains are not required to support the stimulatory effect of ␣-latrotoxin in secretory cells (14,16). It was therefore proposed that these receptors serve as targeting sites to bring the toxin to the necessary location on the cell surface.
Neurexin is a cell surface protein with a single transmembrane domain and a large extracellular region containing six laminin G domains interspersed with three EGF domains (17). It binds ␣-latrotoxin only in the presence of micromolar concentrations of calcium. Neurexin has thousands of isoforms due to the existence of three highly homologous genes, alternative splicing at multiple sites, and the use of two alternative promoters. The physiological importance of neurexins remains unknown. Interestingly, neuroligin, an endogenous ligand of neurexin (18), can function as a post-synaptic trigger for the de novo formation of presynaptic structure in CNS neurons (19).
CIRL has seven transmembrane domains, suggesting that it functions as a G protein-coupled receptor. Like neurexin, CIRL has a large extracellular region with several cell adhesion-like domains. Three highly similar genes encoding CIRL isoforms are present in the mammalian genome (19 -21). The CIRLs are members of the emerging family of heptahelical orphan receptors that are chimeras of cell adhesion molecules and G proteincoupled receptors (reviewed in Refs. 22 and 23). Although information about function is quite limited and available only for a few members of this family, their structural features suggest the potential to couple cell-to-cell and cell-to-matrix interaction to intracellular G protein signaling.
Here we report that a known receptor-like protein-tyrosine phosphatase sigma (PTP) can function as the third receptor of ␣-latrotoxin in the brain. This membrane phosphatase was discovered independently in several laboratories and is known as PTP (24), PTP NE-3 (25), PTP-P1 (26), and LAR-PTP2 (27). Genetic studies in mouse and Drosophila identified PTP as a protein essential for neuronal development and axonal pathfinding (28 -30). We have shown that, similarly to two other ␣-latrotoxin receptors, PTP interacts with ␣-latrotoxin via its extracellular cell adhesion-like region, and the catalytic protein phosphatase domains are not required for the toxin-stimulated exocytosis. Thus, ␣-latrotoxin targets three structurally unrelated cell adhesion receptors that can activate multiple signaling pathways.

EXPERIMENTAL PROCEDURES
␣-Latrotoxin was freshly purified from Black Widow spider venom as described (31). The plasmid containing PTP cDNA and polyclonal antibodies 322 against the N-terminal 80 kDa subunit and 320 against the C-terminal 76 kDa subunit of PTP were kindly provided by Drs. J. Schlessinger (NYU School of Medicine, New York) and A. Ullrich (Max Planck Institute for Biochemistry, Martinsried). Antibodies against the C-terminal peptide of neurexin (32) and the anti-p120 and -p85 subunits of CIRL-1 (12) have been described previously.
Human growth hormone (hGH) was expressed in pXGH5 plasmid under control of the mouse metallothionein I promoter (33). Incubation with a heavy metal was not necessary to obtain adequate hGH expression. The plasmid encoding CIRL (pCDR7) is described (12). Reagents were received from the following sources: [  SDS-PAGE and Western blotting with ECL detection were performed according to Bio-Rad and Amersham Biosciences protocols, respectively. Transfection of COS cells was performed with ExGen-500 reagent according to the manufacturer's protocol (Fermentas). Bovine adrenal chromaffin cells were prepared and maintained in culture as described previously (34), except that the medium used was Dulbecco's modified Eagle's medium/Ham's F-12. Cells for transfection were grown in 12-well plates (Costar Corp., Cambridge, MA) and were transfected by calcium phosphate precipitate 14 -18 h after plating (35). Experimental (PTP, 1-875, or CIRL; 3 g/well) and control (pCMVneo; 3 g/well) plasmids were each mixed with pXGH5 (2 g/well) prior to generating the precipitates. Previous work in the laboratory has demonstrated that this ensures that Ͼ95% of transfected cells express both hGH and the protein of interest (36,37). hGH is stored in chromaffin granules and is released concomitantly with endogenous catecholamine by various secretagogues. Thus, hGH serves as a selective marker for secretion from transfected cells.

Purification of ␣-Latrotoxin Receptors
Twenty grams of rat brains were homogenized in 300 ml of an ice-cold buffer containing 50 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, and 0.5 mM phenylmethylsulfonyl fluoride, pH 8.0. After centrifugation for 30 min at 50,000 ϫ g the obtained pellet was extracted with 2% Triton X-100 in a buffer with 50 mM Tris-HCl, 2 mM EDTA, and 0.5 mM phenylmethylsulfonyl fluoride, pH 8.0, for 40 min in the final volume of 200 ml. The extract was centrifuged for 1 h at 100,000 ϫ g, and the supernatant was supplemented with 100 mM NaCl and divided into two equal portions. Each portion was loaded onto 2 ml of either ␣-latrotoxin or BSA-Sepharose overnight. Matrices were washed with 200 ml of 50 mM Tris-HCl, 130 mM NaCl, 2 mM EDTA, and 0.1% Triton X-100 buffer, pH 8.0, for 8 h, and the retained proteins were eluted with the same buffer containing 1 M NaCl.

Protein Identification by Mass Spectrometry
Silver-stained protein bands on an SDS-PAGE were excised under a tissue culture hood to minimize contamination and de-stained using a freshly prepared 1:1 solution of 30 mM potassium ferricyanide and 100 mM sodium thiosulfate. The de-stained gels were cut into 1-mm 3 pieces and washed three times alternatively with distilled water and acetonitrile. The proteins were in-gel reductively alkylated and digested (38,39) with an excess of sequencing grade unmodified trypsin (Roche Molecular Biochemicals). The resulting peptide mixture was extracted and dried under vacuum, re-suspended in 0.1% trifluoroacetic acid and desalted using a ZipTip TM (Millipore) C-18 reverse phase micro column. The peptides were eluted with 2-3 l of 70% acetonitrile, and the final solution was adjusted to 1% acetic acid. The peptide mixture was loaded into a medium-tip length nanospray needle (Protana, Denmark) and infused at a flow rate of ϳ40 nl/min into a quadrupole time-of-flight mass spectrometer (Micromass, Beverly, MA). Doubly or triply charged tryptic peptide precursor ions were selected in the quadrupole and fragmented by collision with Argon, and the masses of fragment ions were measured in the time-of-flight analyzer. Mass spectra were acquired and analyzed using MassLynx (Micromass) software. Charge state deconvoluted mass spectra were either directly submitted to the search engine Mascot (Matrix Science, UK) or manually interpreted with the help of PepSeq (Micromass) to search the NCBI non-redundant protein data base using BLAST.

Immunoprecipitation
One-milliliter aliquots of either solubilized COS cells transfected with PTP or solubilized rat brain membranes were incubated with 15 l of protein A-agarose and 5 l of either rabbit anti-CIRL (anti-p85) or rabbit anti-␣-latrotoxin or anti-PTP 320 antibodies. Each set was immunoprecipitated either with or without 10 nM ␣-latrotoxin. After overnight incubation, the immunobeads were washed three times with cold 20 mM Tris-HCl, 0.15 M NaCl, 2 mM EDTA, and 0.05% Triton X-100, pH 8.0. Remaining proteins were eluted with sample buffer for analysis by Western blotting with chicken anti-p120 antibody.

␣-Latrotoxin Binding Analyses
␣-Latrotoxin Binding with Immunoprecipitated Rat Brain Membranes-1-ml aliquots of solubilized rat brain membranes were immunoprecipitated with anti-PTP antibody 320 as described above. Binding assays were performed in a volume of 0.15 ml in the presence of 0.5 nM 125 I-␣-latrotoxin. No brain extract and anti-PTP antibody 320 or brain extracts with the preimmune antibody were used as negative controls. For an additional control to measure nonspecific binding, 50-fold excess of unlabeled toxin was added to the second set of immunoprecipitates.
Transfected COS Cells-COS cells transfected with PTP and rat brain membranes were solubilized with 2% Triton X-100 and clarified by centrifugation. Obtained supernatants were immunoprecipitated with 5 l of antibody 320 and 15 l of protein A-agarose overnight in the presence of 0.5 nM 125 I-␣-latrotoxin. The preimmune serum or the solubilization buffer was used as negative control. To control for nonspecific binding, a second set of immunoprecipitates was incubated with 25 nM unlabeled ␣-latrotoxin.
Generation of Soluble C-terminal Deletion Mutants of PTP-Three pairs of primers: 323 and 306, 323 and 307, 323 and 308 (see sequences below) and the plasmid pPTP_1-880 as a template were used in PCR to generate soluble C-terminal deletion mutants of PTP. After amplification the products of PCR were digested with EcoRI/XhoI and ligated with a 4982-bp fragment isolated after digestion of pPTP_1-880 with EcoRI/XhoI. The resulting plasmids encoded 1-850 aa, 1-592 aa, and 1-311 aa and were named pPTP_1-850, pPTP_1-592, and pPTP_1-311, respectively.

Secretion Assays
Physiological salt solution contained 145 mM NaCl, 5.6 mM KCl, 5.6 mM glucose, 0.5 mM ascorbate, 15 mM HEPES pH 7.4, 2.2 mM CaCl 2 , and 0.5 mM MgCl 2 , unless otherwise indicated. hGH was measured with a luminescent assay kit from Corning Nichols Institute Diagnostics (San Juan Capistrano, CA) (37). Endogenous catecholamines were measured by spectrofluorometric assay (40). Stimulated release is calculated as the amount of hGH released into the incubation medium divided by the total hGH (i.e. hGH released ϩ hGH remaining in the cells). Data are expressed as means Ϯ S.E. of the mean unless otherwise indicated. Significance was determined by Student's t test. Error bars smaller than symbols were omitted from figures.

RESULTS
To understand better the mechanism of ␣-latrotoxin action, we performed a search for ␣-latrotoxin-interacting proteins by affinity chromatography of solubilized brain membranes on immobilized ␣-latrotoxin. Because the concentration of ␣-latrotoxin receptors in crude brain membranes is low (about 200 fmol/mg protein), one-step affinity chromatography results only in partial purification of the receptors. To distinguish between specifically retained proteins that interact with ␣-latrotoxin and the nonspecific background, we used chromatography on BSA-agarose as a control. Detergent extracts of brain membranes were loaded onto these two columns in identical conditions, and the eluates with a high-salt buffer were compared by SDS gel electrophoresis. As a result of this analysis, several bands were identified that were present only in the eluate from the ␣-latrotoxin column but not from the BSA column (Fig. 1A, left panel). In addition to the bands containing the previously described neurexin (160 and 200 kDa) and CIRL FIG. 1. Presence of PTP in ␣-latrotoxin-agarose eluates. A, solubilized rat brain membranes were loaded onto either BSA-or ␣-latrotoxin-agarose as described under "Experimental Procedures." The aliquots of 1 M NaCl eluate were analyzed by silver staining (left panel) and Western blotting (right panels) with the indicated anti-PTP antibodies. The arrows on the silver-stained gel indicate the extracellular p80 subunit of PTP (filled), the intracellular p70 subunit of PTP (gray), and the extracellular p120 subunit of CIRL (open). The heavily stained high-molecular weight protein bands in the ␣-latrotoxin column eluate represent multiple isoforms of neurexin (160 -220 kDa) together with the p85 heptahelical subunit of CIRL, which typically forms oligomers and aggregates on SDS gels. AB#322, antibody 322; AB#320, antibody 320. B, the domain structure of PTP.
(120 kDa and high-molecular weight aggregates), two protein bands of ϳ80 and 70 kDa were observed. These protein bands were excised from the gel, and their partial amino acid sequences were determined by tandem mass spectrometry. The peptide sequences obtained (Table I) unambiguously identified these protein bands as p80 and p70 subunits of PTP that derive from the intracellular proteolytic cleavage of the proreceptor (41). p80 is an extracellularly oriented hydrophilic protein with three immunoglobulin and four fibronectin (the first three of a high homology to the generic consensus and the fourth of a low homology) domains, whereas p70 consists of a single transmembrane domain and two cytoplasmic catalytic domains (Fig. 1B).
To further verify the identity of the discovered protein and specificity of its interaction with ␣-latrotoxin, the eluates were analyzed by Western blotting with two anti-PTP antibodies that recognize either p80 or p70 PTP subunits. Both antibodies produced strong staining of the respective protein bands in the ␣-latrotoxin column eluate preparation but not in the control one (Fig. 1A, right panel).
Using anti-PTP antibody, we also performed the experiment that is reciprocal to the affinity chromatography on immobilized ␣-latrotoxin. PTP was immunoprecipitated with antibody, and its affinity to ␣-latrotoxin was probed by addition of radioactively labeled toxin to the immunoprecipitation mixtures. To assure the specificity of the phosphatase-toxin interaction, three negative controls were performed. First, excess unlabeled ␣-latrotoxin was added to displace the labeled toxin. Second, the brain extract was not added to the mixture to exclude a possibility of direct sorption of ␣-latrotoxin on either protein A or anti-PTP antibody. Finally, preimmune serum was used in a similar immunoprecipitation reaction. The precipitates were washed and counted for radioactivity. The resulting data confirmed the specific interaction of PTP with ␣-latrotoxin (Fig. 2).
Two known neuronal receptors of ␣-latrotoxin, neurexin and CIRL, differ in their Ca 2ϩ requirement for ␣-latrotoxin binding. Neurexin interacts with the toxin only in the presence of micromolar concentrations of Ca 2ϩ , whereas ␣-latrotoxin binding activity of CIRL does not depend on the Ca 2ϩ presence at all. We therefore tested whether there was any difference in PTP sorption onto the ␣-latrotoxin column when chromatography was performed with either high-Ca 2ϩ or low-Ca 2ϩ buffer systems. No significant difference in the amount of purified PTP under these two conditions was observed on the Western blot (Fig. 3). Thus, the interaction of PTP with ␣-latrotoxin does not depend on Ca 2ϩ . These results also suggest that the interaction does not depend on neurexin presence, since neurexin is not retained on ␣-latrotoxin-agarose in low-Ca 2ϩ buffers (10, 31).
Because the experiments showing PTP-␣-latrotoxin interaction were performed using crude membrane extracts, they may be interpreted as either direct binding of the toxin to phosphatase or indirect complexing that would involve another ␣-latrotoxin-binding protein, e.g. CIRL. To investigate the possibility of PTP interaction with CIRL, we immunoprecipitated crude brain membrane extracts either in the presence or absence of ␣-latrotoxin with anti-PTP antibody and with anti-p85 CIRL and anti-␣-latrotoxin antibodies as controls. The immunoprecipitates were processed for Western blotting staining with anti-p120 CIRL antibodies. CIRL was efficiently precipitated with the anti-CIRL antibody independently of ␣-latrotoxin presence. However, its precipitation with either anti-␣latrotoxin or anti-PTP antibody was strictly dependent on ␣-latrotoxin (Fig. 4A).
We also performed an experiment in which PTP was immunoprecipitated and beads were thoroughly washed and analyzed for ␣-latrotoxin binding. In a parallel experiment, the interaction of CIRL with ␣-latrotoxin was assayed. The results of these experiments suggest that PTP can be precipitated separately from CIRL and still interact with ␣-latrotoxin (Fig. 4B).
To verify further the identity of PTP as ␣-latrotoxin receptor, we analyzed the ␣-latrotoxin binding activity of overex-TABLE I Partial amino acid sequences of the proteins specifically retained on ␣-latrotoxin-agarose Detergent extracts of crude rat brain membranes were chromatographed on ␣-latrotoxin-agarose and analyzed by SDS-PAGE, and their partial amino acid sequences were determined as described under "Materials and Methods." Leucine and isoleucine, which are indistinguishable by low energy tandem mass spectrometry used to sequence the peptides, were assigned according to the similarity of the peptides to known rat PTP sequence. pressed PTP in non-neuronal cells. Either COS cells or 293 cells were transfected with a plasmid encoding full-length PTP, and its expression was confirmed by Western blotting with anti-PTP antibodies ( Fig. 5A and data not shown). The two-subunit structure of PTP results from the intracellular cleavage of the single chain precursor in its ectodomain close to the transmembrane region. The results of the Western blot analysis suggested that PTP in COS cells was processed essentially in a similar manner. However, the presence of noncleaved forms on the top of the gel indicate lower efficiency of the cleavage in this artificial expression system as compared with the processing under physiological condition in brain cells. The expressed PTP and even its non-cleaved precursor bound to the ␣-latrotoxin column in a specific manner as indicated by soluble ␣-latrotoxin competition and lack of binding to BSA-agarose.
The direct analysis of 125 I-␣-latrotoxin binding by the transfected cells revealed a significantly lower, as compared with CIRL-transfected cells, level of specific binding that may be explained by lower surface expression of PTP in non-neuronal cells (data not shown). The low signal/background ratio did not allow us to estimate the affinity of the interaction precisely. As an alternative approach to assess the specificity of the interaction between PTP and ␣-latrotoxin, the cells overexpressing PTP were solubilized and analyzed for the binding of 0.2 nM radioactively labeled ␣-latrotoxin by immunoprecipitation with anti-PTP antibody. To control for the specificity of the interaction, excess non-labeled toxin (5 nM) was added as competitor. In addition, similar precipitation reactions were set up with the preimmune serum. The results of these experiments (Fig. 5B) indicate that PTP is indeed an ␣-latrotoxin-binding protein. The fact that PTP exogenously expressed in nonneuronal cells binds ␣-latrotoxin indicated that this interaction does not involve any other neuronal ␣-latrotoxin receptor, CIRL or neurexin, which are not present in COS cells (data not shown).
In two other ␣-latrotoxin-binding proteins, neurexin and CIRL, the toxin-binding sites are located within the extracellular domains (14,15). It has been shown that the extracellular region of PTP can be cleaved off and released into the medium (41). This cleavage is secondary in relation to the first intracellular processing that results in the two-subunit complex of PTP. We tested whether the soluble ectodomain of PTP expressed in COS cells can be precipitated with ␣-latrotoxinagarose. The cells were transfected with a plasmid encoding FIG. 4. The interaction of PTP with ␣-latrotoxin does not require CIRL. A, detergent extracts of crude brain membranes were immunoprecipitated with either anti-␣-latrotoxin (Anti-LTx), two anti-CIRL (anti-p85pept and anti-p85prot), or anti-PTP antibody 320 or without any added antibody. The precipitates were analyzed by Western blotting with anti-CIRL (anti-p120) antibody. All reactions were performed in the presence or absence of 5 nM ␣-latrotoxin. The competition with antigen peptides was used as additional control for the specificity of immunoprecipitation with anti-PTP antibodies (ϩPept). B, detergent extracts of crude brain membranes were immunoprecipitated with either anti-CIRL (anti-p85pept) or anti-PTP 320 antibody. After overnight incubation the immunomatrices were extensively washed and analyzed for the binding of 150,000 cpm of 125 I-␣-latrotoxin (gray bars). In controls, excess non-labeled toxin was included (black bars), or membrane extracts were omitted (white bars).

FIG. 5. ␣-Latrotoxin binding activity of recombinant PTPs. A,
COS cells were transfected with the PTP expression plasmid. Triton X-100 extracts of the harvested cells were chromatographed on either ␣-latrotoxinor BSA-agarose. The column eluates were analyzed by Western blotting with the anti-PTP antibody 322. As an additional control for the specificity of the interaction, 40 nM soluble toxin was added to the extract prior to the loading onto ␣-latrotoxin-agarose. As a reference, native brain PTP purified on ␣-latrotoxin-agarose was electrophoresed on the same gel (right lane). B, the detergent extracts of the transfected COS cells were immunoprecipitated with the antibody 320 (1) either in the presence of 150,000 cpm of 125 I-␣-latrotoxin only (gray bars) or with a 50-fold excess of unlabeled toxin (black bars). In the controls, immunoprecipitation reactions were set up either without the brain extract (2) or with the preimmune antibody instead of the immune antibody (3). full-length PTP, and the conditioned media were incubated with either ␣-latrotoxin or BSA-agarose. The absorbed proteins were eluted and analyzed by Western blotting with the anti-PTP antibody that recognizes the extracellular domain. The staining confirmed the expression of PTP precursor and its processed forms in the cells. It also demonstrated the specific interaction of the soluble ectodomain of PTP with immobilized ␣-latrotoxin (data not shown).
Because the extracellular region of PTP is large and consists of multiple structural domains we performed an experiment to identify the ␣-latrotoxin-binding site more precisely. We prepared expression constructs that represent a series of deletion mutants encoding soluble fragments of the PTP ectodomain (Fig. 6A) and used them to transfect COS cells. The expressed proteins were tested for ability to bind to the ␣-latrotoxin matrix (Fig. 6B). For most constructs, we used the conditioned medium in the binding assay labeled with black horizontal bars at the top of the blot. However, some mutant proteins were not secreted efficiently. In those cases, the cells were first solubilized with Triton X-100, and the extracts were precipitated with ␣-latrotoxin agarose (white bars at the top of the blot). The expression of the recombinant proteins was confirmed by direct Western blotting of cell extracts (Fig. 6C). The results of the binding analysis indicated that the ␣-latrotoxinbinding site of PTP is located between amino acid residues 407 and 685 and may include two fibronectin repeats but none of the immunoglobulin repeats.
The in vitro experiments with solubilized brain membranes suggested that PTP is an ␣-latrotoxin-binding protein. The question arises whether the interaction of ␣-latrotoxin with PTP may occur in vivo and whether the formation of the toxin-receptor complexes would result in the stimulation of exocytosis. As shown earlier, brain membranes contain two other families of functional ␣-latrotoxin receptors, the neurexins and CIRLs. Similarly to CIRL, PTP interacts with ␣-latrotoxin independently of Ca 2ϩ presence in the medium. Thus, the experiments with neurons or neuronal membranes cannot distinguish easily between the effects related to the interaction of ␣-latrotoxin with PTP and CIRL.
To prove that the complexes of ␣-latrotoxin with PTP may exist in parallel with other receptor-toxin complexes at similar low-toxin concentrations, we designed the experiment when radioactively labeled ␣-latrotoxin pre-bound to native brain membranes at a nanomolar concentration was chemically cross-linked to the membrane receptors. The preparation was further dissolved under denaturing conditions, and the toxinreceptor complexes were separately assayed by immunoprecipitation with antibodies against neurexin, CIRL, PTP, and another receptor tyrosine phosphatase (RPTP␣) as control. The precipitated 125 I-␣-latrotoxin was counted for radioactivity (Fig. 7A), and the precipitates were further electrophoresed in SDS and autoradiographed to detect the cross-linked bands (Fig. 7B). In several independent experiments, we reproducibly observed cross-linked toxin-receptor complexes with neurexin I␣, CIRL-1, and PTP but not with RPTP␣. The amount of precipitated ␣-latrotoxin-PTP complexes was lower than for neurexin and CIRL, which was consistent with our independent estimates of relative concentration (Fig. 1). Although the cross-linking produced high-molecular weight fuzzy bands, the band pattern was clearly different among PTP, neurexin, and CIRL. When antibody against the extracellular subunit of PTP was used, a band of ϳ200 kDa was reproducibly observed that corresponded to the one-to-one complex of ␣-latrotoxin monomer and the p80 subunit of PTP. Virtually no radioactivity could be immunoprecipitated with any antibody when the cross-linking reagent was not added to the membrane prep-aration (Fig. 7A), indicating that only covalent receptor-toxin complexes could be precipitated in our experiments. In other important controls, when antigen peptides were used as competition in immunoprecipitation reactions or preimmune sera, no significant amount of radioactivity could be detected in the precipitates ( Fig. 7 and data not shown).
To test whether ␣-latrotoxin binding to PTP results in the functional response typical for this toxin, we analyzed secretion from chromaffin cells transfected with the PTP expression plasmid. Chromaffin cells respond to ␣-latrotoxin stimulation by robust exocytosis of catecholamines (42,37). To distinguish the secretion from only a minor portion of actually transfected cells, we used an earlier developed method based on the cotransfection of the human growth hormone cDNA followed by the quantification of the human growth hormone secretion (43). FIG. 7. Chemical cross-linking of ␣-latrotoxin complexes with PTP, neurexin I␣, and CIRL-1 in brain membranes. Brain membranes with pre-bound 125 I-␣-latrotoxin were treated with chemical cross-linking reagent BS 3 as described under "Experimental Procedures." The preparation was further dissolved in an SDS-containing buffer and subjected to immunoprecipitation with a panel of antibodies against neurexin I␣, PTP subunits, CIRL-1 p120 subunit, and receptor tyrosine phosphatase PTP␣ as negative control. The precipitates originated from either cross-linked membranes or the similarly processed preparation without BS 3 addition were counted for radioactivity (A), electrophoresed, and autoradiographed for 24 h at Ϫ70°C. B, the competition with antigen peptides was used as additional control for the specificity of immunoprecipitation with anti-PTP antibodies (ϩPept).
As a negative control, co-transfection of the empty vector plasmid and human growth hormone construct was performed. In a parallel experiment, chromaffin cells were similarly transfected with CIRL, a functional ␣-latrotoxin receptor that is known to significantly increase the sensitivity of the cells to the toxin.
As shown in Fig. 8A, the dose dependence of ␣-latrotoxinstimulated secretion of human growth hormone from PTPtransfected chromaffin cells was shifted significantly to the lower concentrations of the toxin as compared with mock-transfected cells. To control for the cell viability and secretion efficiency, the release of catecholamine was measured in parallel (Fig. 8B). Since about 95% of all cells in the preparations were not transfected, no significant difference was observed in the catecholamine secretion from CIRL or PTP-transfected cells in comparison with mock-transfected cells. These data indicate that PTP is a functional ␣-latrotoxin receptor.
A similar shift in the dose dependence for ␣-latrotoxin was achieved when PTP-transfected cells were preincubated with ␣-latrotoxin in the absence of calcium, and the toxin was removed before addition of calcium-containing buffer (data not shown). This result indicates that the calcium-dependent secretion from the PTP-transfected cells is due to the calciumindependent interaction of the toxin with receptor.
To address the question of whether PTP signaling is required for the stimulatory effect of ␣-latrotoxin, we transfected chromaffin cells with a PTP C-terminal deletion mutant that did not have its intracellular catalytic phosphatase domains but contained the transmembrane domain. This truncated receptor could support ␣-latrotoxin-stimulated release in chromaffin cells as well as CIRL and better than wild-type PTP (Fig. 8A). This may be explained either by higher cell surface expression of the non-signaling mutant or by toxic effects on the cells of overexpressed wild-type PTP that has full signaling potency.

DISCUSSION
Stimulation of neurotransmitter release with ␣-latrotoxin requires binding to neuronal cell surface proteins. The data presented in this paper identify receptor protein-tyrosine phosphatase PTP as an ␣-latrotoxin-binding protein and functionally competent ␣-latrotoxin receptor. The finding that PTP is a functional ␣-latrotoxin receptor is unexpected because two other ␣-latrotoxin receptors, neurexin and CIRL, have already been described. Although each of these three receptors can support ␣-latrotoxin action efficiently and independently, no obvious homology can be found between their sequences. Neurexin is a one-transmembrane cell adhesion protein, whereas CIRL is a hybrid of a cell adhesion molecule with a G proteincoupled receptor (Fig. 9).
The strongest evidence that PTP is yet another ␣-latrotoxin receptor is that its overexpression in chromaffin cells results in a significant increase in sensitivity to stimulation with low concentrations of ␣-latrotoxin. The biochemical data with native and recombinant PTP confirm the specificity of the direct FIG. 9. Domain structure of the ␣-latrotoxin receptors.
FIG. 8. Expression of PTP and its C-terminally truncated mutant increases the sensitivity of intact chromaffin cells to stimulation by ␣-latrotoxin. Chromaffin cells were transfected with plasmids for hGH and either CIRL, PTP, PTP_1-875, or neo (a control) as described. Four days later, cells were incubated with the indicated concentrations of ␣-latrotoxin in physiological salt solution for 6 min. The amounts of hGH (A) and catecholamine (B) released into the medium and the amounts remaining in the cells were determined as described (12). Stimulated secretion was calculated by subtracting the secretion in the absence of ␣-latrotoxin, which was similar for all groups, from total secretion in the presence of toxin. Secretion in the absence of ␣-latrotoxin was 1.57 Ϯ 0.15% (hGH) or 0.23 Ϯ 0.03% (catecholamine). n ϭ 4 wells/group. interaction between ␣-latrotoxin and PTP and identify the ␣-latrotoxin-binding site in the extracellular region of this receptor. In particular, we have shown by chemical cross-linking that the ␣-latrotoxin/PTP complexes can form in brain membranes at a low physiologically relevant concentration of ␣-latrotoxin. Northern blotting data obtained by other groups indicate that the highest concentration of PTP is found in brain tissues (24,25,44), which is also compatible with its proposed role as a functional ␣-latrotoxin receptor.
The proposed existence of the third functional ␣-latrotoxin receptor is in good agreement with the phenotype of recently described double knock-out mice with inactive neurexin I␣ and CIRL-1 genes. These mice showed a physiological response to ␣-latrotoxin and its binding to brain membranes at ϳ25% level as compared with the wild-type animals (45). In addition to the low-affinity receptors CIRL-2 and neurexin isoforms, PTP may constitute a major contribution to the residual ␣-latrotoxin sensitivity of the neurexin I␣-and CIRL-1-deficient mice.
What are the roles of PTP and the other two receptors in the ␣-latrotoxin mechanism? Our data suggest that the toxin interacts with the extracellular domain of PTP. In the chromaffin cell secretion assay, a PTP mutant with a deleted intracellular region was shown to support ␣-latrotoxin stimulation even better than wild-type PTP. We may thus conclude that the direct interaction of the toxin with the phosphatase catalytic domains is unlikely and that the PTP-mediated signaling is not essential for the ␣-latrotoxin stimulatory effect in chromaffin cells. Essentially similar data have been reported previously for CIRL and neurexin exogenously expressed in chromaffin and PC12 cells (14,15). Thus, at least in secretory cells, ␣-latrotoxin receptors serve as targeting sites that attract the toxin and facilitate its partial insertion into the cell membrane that ultimately results in the stimulation of exocytosis via channel formation and/or another yet unknown mechanism. Since small synaptic vesicle exocytosis has features different from granular exocytosis, further physiological tests in transfected neurons would be required to establish unequivocally the role of neuronal receptors in the ␣-latrotoxin action.
It seems unlikely that evolution has developed a toxin that efficiently targets three seemingly unrelated proteins. We may therefore speculate that the ␣-latrotoxin receptors share not only affinity to the toxin but also a physiological role that is related to the presynaptic function. A common structural feature of all three ␣-latrotoxin receptors is that they contain large extracellular cell adhesion-like domains (Fig. 9). It is likely that in neurons PTP, CIRL, and neurexin function together as regulators of cell-to-matrix and cell-to cell interactions in synaptic and possibly other junctions. A possible mechanistic explanation of the link between the ␣-latrotoxin receptor would be the existence of an endogenous ligand that is shared by all receptors and resembles the toxin.