Neurexin Iα Is a Major α-Latrotoxin Receptor That Cooperates in α-Latrotoxin Action*

α-Latrotoxin is a potent neurotoxin from black widow spider venom that binds to presynaptic receptors and causes massive neurotransmitter release. A surprising finding was the biochemical description of two distinct cell surface proteins that bind α-latrotoxin with nanomolar affinities; Neurexin Iα binds α-latrotoxin in a Ca2+-dependent manner, and CIRL/latrophilin binds in a Ca2+-independent manner. We have now generated and analyzed mice that lack neurexin Iα to test its importance in α-latrotoxin action. α-Latrotoxin binding to brain membranes from mutant mice was decreased by almost 50% compared with wild type membranes; the decrease was almost entirely due to a loss of Ca2+-dependent α-latrotoxin binding sites. In cultured hippocampal neurons, α-latrotoxin was still capable of activating neurotransmission in the absence of neurexin Iα. Direct measurements of [3H]glutamate release from synaptosomes, however, showed a major decrease in the amount of release triggered by α-latrotoxin in the presence of Ca2+. Thus neurexin Iα is not essential for α-latrotoxin action but contributes to α-latrotoxin action when Ca2+ is present. Viewed as a whole, our results show that mice contain two distinct types of α-latrotoxin receptors with similar affinities and abundance but different properties and functions. The action of α-latrotoxin may therefore be mediated by independent parallel pathways, of which the CIRL/latrophilin pathway is sufficient for neurotransmitter release, whereas the neurexin Iα pathway contributes to the Ca2+-dependent action of α-latrotoxin.

␣-Latrotoxin is a potent neurotoxin from black widow spider venom that binds to presynaptic receptors and causes massive neurotransmitter release. A surprising finding was the biochemical description of two distinct cell surface proteins that bind ␣-latrotoxin with nanomolar affinities; Neurexin I␣ binds ␣-latrotoxin in a Ca 2؉dependent manner, and CIRL/latrophilin binds in a Ca 2؉ -independent manner. We have now generated and analyzed mice that lack neurexin I␣ to test its importance in ␣-latrotoxin action. ␣-Latrotoxin binding to brain membranes from mutant mice was decreased by almost 50% compared with wild type membranes; the decrease was almost entirely due to a loss of Ca 2؉ -dependent ␣-latrotoxin binding sites. In cultured hippocampal neurons, ␣-latrotoxin was still capable of activating neurotransmission in the absence of neurexin I␣. Direct measurements of [ 3 H]glutamate release from synaptosomes, however, showed a major decrease in the amount of release triggered by ␣-latrotoxin in the presence of Ca 2؉ . Thus neurexin I␣ is not essential for ␣-latrotoxin action but contributes to ␣-latrotoxin action when Ca 2؉ is present. Viewed as a whole, our results show that mice contain two distinct types of ␣-latrotoxin receptors with similar affinities and abundance but different properties and functions. The action of ␣-latrotoxin may therefore be mediated by independent parallel pathways, of which the CIRL/latrophilin pathway is sufficient for neurotransmitter release, whereas the neurexin I␣ pathway contributes to the Ca 2؉ -dependent action of ␣-latrotoxin.
␣-Latrotoxin is a potent neurotoxin from black widow spider venom that triggers neurotransmitter release (reviewed in Ref. 1). ␣-Latrotoxin is thought to elicit neurotransmitter release by binding to presynaptic receptors. Only nanomolar concentrations of ␣-latrotoxin are required to trigger exocytosis, suggesting the presence of high affinity receptors. Binding activates an unknown signal transduction cascade that leads to exocytosis of virtually all synaptic vesicles. ␣-Latrotoxin also induces neurotransmitter release in the absence of Ca 2ϩ in a manner that circumvents the normal Ca 2ϩ -dependent stimulation pathway for synaptic vesicle exocytosis (2). The mechanism of action of ␣-latrotoxin and the nature of the intracellular signaling cascades involved are unknown.
Two membrane proteins that bind ␣-latrotoxin with high affinity have been isolated from brain. The first such protein identified was neurexin I␣, which was isolated by affinity chromatography of brain proteins on immobilized ␣-latrotoxin (3,4). cDNA cloning revealed that neurexin I␣ is a member of a family of three genes (4 -6). Each of the neurexin genes contains two promoters that direct synthesis of ␣and ␤-neurexins. Neurexins are highly polymorphic and neuron-specific in expression (7). Their structure and interactions with putative endogenous ligands (neuroligins and neurexophilin) suggest a receptor function (8 -10). Experiments with recombinant neurexin I␣ confirmed that it binds to ␣-latrotoxin with high affinity but showed that this interaction is completely dependent on Ca 2ϩ (11). Despite the wealth of biochemical data on neurexins and their interactions with endogenous ligands and ␣-latrotoxin, the in vivo roles of neurexins are unknown.
The observations that ␣-latrotoxin binds to neurexin I␣ only in the presence of Ca 2ϩ but triggers neurotransmitter release also in the absence of Ca 2ϩ suggested that there must be a second ␣-latrotoxin receptor in addition to neurexin I␣. Based on this finding, a second high affinity receptor for ␣-latrotoxin called CIRL or latrophilin was identified (12,13). Surprisingly, CIRL has no structural similarity to neurexins. The presence of two distinct high affinity binding proteins for ␣-latrotoxin in brain raises the possibility that ␣-latrotoxin may physiologically act via two independent receptor pathways or that the in vitro binding of ␣-latrotoxin by one of the two putative receptors may represent an artifact. In the current study, we have addressed this question using knockout mice. We have generated a mouse line in which expression of neurexin I␣ was abolished and investigated the importance of neurexin I␣ for brain function and ␣-latrotoxin action. Our data confirm that neurexin I␣ constitutes a major ␣-latrotoxin receptor. We show that neurexin I␣ is not required for ␣-latrotoxin action in the absence of Ca 2ϩ but is essential for full ␣-latrotoxin action in the presence of Ca 2ϩ .

EXPERIMENTAL PROCEDURES
Genomic Cloning of Neurexin Genes-A mouse genomic library was screened for the 5Ј ends of the neurexin genes by high stringency hybridization as described (14). Clones were isolated, mapped, and sequenced using general molecular biology techniques (14,15). Sequences were analyzed on a personal computer using DNA-STAR software.
Generation and Maintenance of Knockout Mice-A knockout vector was constructed from the genomic clone for neurexin I␣ as diagrammed in Fig. 1. Embryonic stem cells were electroporated with the vector and selected with G418 (Life Technologies, Inc.) and FIAU essentially as described (16). Resistant embryonic stem cell clones were analyzed by polymerase chain reaction for homologous recombination (primers used: 676 [GAGCGCGCGCGGCGGAGTTGTTGAC] and 918 [AGC-CAATACTTCTGGGAAGACAGACT]) and confirmed by Southern blotting of genomic DNA digested with SpeI with the probe indicated in Fig.  1. Positive clones were injected into blastocysts, resulting in the generation of a single mouse line that was bred to homozygosity and genotyped by Southern blotting. To analyze the effect of the mutation in the neurexin I␣ gene on mouse survival, mice heterozygous for the neurexin I␣ mutation were mated with each other, and the number of adult surviving offspring was determined by genotyping.
Antibodies and Immunoblot Analysis-Antibodies against the cytoplasmic tails of neurexins were raised in rabbits using recombinant bacterially expressed proteins in which the N terminus of the cytoplasmic tail sequence was fused to a hexahistidine sequence for purification. Antibodies were affinity-purified on immobilized glutathione S-transferase fusion proteins of the same sequences, and their specificity was confirmed using recombinant neurexins. All other antibodies were described previously (17).
␣-Latrotoxin Binding Measurements-␣-Latrotoxin binding measurements were performed by a rapid centrifugation assay essentially as described (18). Mouse brains were homogenized in 0.15 M NaCl, 1 mM EDTA, 20 mM Tris-HCl, pH 7.8, and crude membranes were prepared by centrifugation. The specific binding of 0.5 nM 125 I-␣-latrotoxin to crude brain membranes from mice was analyzed in triplicates in a 0.15-ml volume with 0.2 mg of protein in the presence of 2 mM Ca 2ϩ or 2 mM EDTA. A 50-fold excess of unlabeled ␣-latrotoxin was added to control for nonspecific binding. Ca 2ϩ -independent binding was defined as ␣-latrotoxin binding observed in the presence of EDTA, and Ca 2ϩdependent binding was defined as the difference between total ␣-latrotoxin binding measured with Ca 2ϩ and Ca 2ϩ -independent binding. Binding assays for Scatchard analyses were performed similarly except that ␣-latrotoxin concentrations from 0.17 to 17.0 nM were used.
Embryonic Cultures and Electrophysiology-Cultures from embryonic hippocampus from wild type or mutant mice were prepared as described (19) and analyzed by electrophysiological recordings (20). Spontaneous miniature excitatory postsynaptic currents were monitored as a function of the application of 1 nM ␣-latrotoxin in the presence or the absence of Ca 2ϩ . To analyze the currents when there was a high degree of overlap of miniature excitatory currents (especially after ␣-latrotoxin application), currents were integrated over 200-ms intervals. The charge transfer obtained in this manner is proportional to the frequency of miniature excitatory postsynaptic currents, the quantal release rate (21).

Measurements of [ 3 H]Glutamate
Release from Synaptosomes-Synaptosomes were prepared from the neocortex of adult mice by a Percoll gradient centrifugation method modified from Ref. 2. The crude mitochondrial fraction (P 2 ) was resuspended in 8.5% (v/v) Percoll suspended in 0.25 M sucrose, 5 mM HEPES-NaOH, pH 7.4, and layered on top of an 12%/20% Percoll step gradient in the same buffer. After centrifugation at 16,000 ϫ g for 20 min, synaptosomes were recovered from the 12%/20% Percoll interface. Percoll was removed by the addition of 30 volumes of 0.32 M sucrose and centrifugation. Pelleted synaptosomes were resuspended in 2 ml of ice-cold gassed (95% O 2 /5% CO 2 ) Krebs-Henseleit-HEPES buffer (KHH buffer), pH 7.4 (composition in mM: NaCl 118, KCl 3.5, CaCl 2 1.25, MgSO 4 1.2, KH 2 PO 4 1.2, NaHCO 3 25, HEPES-NaOH 5 at pH 7.4, glucose 11.5, 0.1 g/liter bovine serum albumin (Sigma catalog number A6793)). Synaptosomes were loaded with 3 H-labeled glutamate by a 5-min incubation in KHH buffer containing 0.17 M [ 3 H]glutamate (24 mCi/mmol; NEN Life Science Products) in a 95% O 2 /5% CO 2 atmosphere at 34°C. Afterward synaptosomes (100 l) were captured on a glass fiber filter (GF/B) in a superfusion chamber, overlaid with 50 l of a 50% slurry of Sephadex G-25 in KHH buffer, and superfused (flow rate, 0.8 ml/min) at 34°C using two stimulation protocols: 1) Synaptosomes were superfused with KHH buffer for at least 12 min for equilibration, and then the perfusate was collected for 2 min to establish the base-line release rate of [ 3 H]glutamate, and 0.5 nM ␣-latrotoxin was added to the superfusion buffer for 1 min followed by continued superfusion with KHH buffer without ␣-latrotoxin. The perfusate was collected throughout the procedure. 2) Synaptosomes were superfused with Ca 2ϩ -free KHH buffer (KHH buffer in which 0.1 mM EGTA was present instead of 1.25 mM CaCl 2 ) gassed with 95% O 2 /5% CO 2 for at least 12 min for equilibration. Then the perfusate was collected for 2 min to establish the base-line release rate of [ 3 H]glutamate, after which 0.5 M sucrose was added to the superfusion buffer for 30 s to stimulate release of glutamate. Thereafter the synaptosomes were superfused for an additional 2.5 min with Ca 2ϩ -free KHH buffer to re-equilibrate them, and finally a 1-min pulse stimulation of ␣-latrotoxin in the same buffer was applied followed by continued superfusion without ␣-latrotoxin. The perfusate was continuously collected. [ 3 H]Glutamate levels in the perfusate were determined by scintillation counting. The fractional release rate was calculated by dividing the amount of radioactivity released at any given time point by the total amount of radioactivity remaining with the synaptosomes at that point. Two mice from each genotype were analyzed in independent experiments.

RESULTS
Structure of the 5Ј End of the Neurexin I␣ Gene-Upon screening a mouse genomic library, we isolated a clone containing a single large exon from the 5Ј end of the neurexin I␣ mRNA (Fig. 1). Sequence analysis revealed that the exon encoded the N terminus of neurexin I␣, including the signal peptide, the first LNS domain (LNS domains are repeat sequences found in laminins, neurexins, and sex hormone-binding globulins; Ref. 26), and the first epidermal growth factorlike sequence. The exon ended at the 5Ј boundary of the first site of alternative splicing in neurexin I␣ corresponding to residue 253 (Fig. 1). In addition to the 5Ј end of the coding region, the exon also contained the entire 5Ј-untranslated region that we previously sequenced in rat and bovine cDNA clones (0.89 kilobase pairs; Refs. 5, 7), suggesting that the exon present in this genomic clone constitutes the first exon of the gene and that the 5Ј-flanking sequences present in the genomic clone represent the neurexin I␣ promoter.
Thus the 5Ј end of the neurexin I␣ gene is composed of a large exon (Ͼ1.5 kilobase pairs) that includes the entire 5Ј-untranslated region and extends to the first site of alternative splicing. Interestingly, the first site of alternative splicing in neurexin I␣ is among the most polymorphic and exhibits at least seven different variants (7). The presence of a large exon preceding this site may facilitate regulation of alternative splicing. Preliminary studies on the structures of the neurexin II␣ and III␣ genes suggest that their 5Ј ends also contain a single large first exon that extends until the position of the first site of alternative splicing (not shown). This result agrees well with the fact that these neurexins are also extensively alternatively spliced at this position.
Generation of Knockout Mice for Neurexin I␣-Using the neurexin I␣ genomic clone, we constructed a targeting vector in which the entire first exon and several kilobases of 5Ј-flanking sequence (presumably containing the neurexin I␣ promoter) were replaced by a neomycin gene cassette as a positive selectable marker (Fig. 1). The short arm of the targeting vector was constructed from the sequence of the first intron and followed by two copies of a Herpes simplex virus thymidine kinase gene cassette for negative selection. The long arm of the vector was obtained from further 5Ј-flanking sequences (Fig. 1). As a result, the entire first exon and part of the promoter of the neurexin I␣ gene were deleted in the targeting vector.
We transfected embryonic stem cells with the targeting vector. Clones emerging after positive and negative selection (with neomycin and FIAU, respectively) were analyzed by polymerase chain reaction using primers flanking the short arm (A and B in Fig. 1). Several clones with putative homologous recombination were obtained and injected into blastocysts. In this manner we generated a single mouse line that transmitted the mutation through the germline. Southern blotting demonstrated that in the mutants, the wild type and mutant neurexin I␣ genes were allelic and were transmitted in a Mendelian fashion (Fig. 2).
Mice Lacking Neurexin I␣ Are Viable and Fertile-Mice carrying the deletion of the first exon of the neurexin I␣ gene were bred to homozygosity and analyzed. Homozygous mutant mice were indistinguishable in appearance from wild type mice.
They were fertile and survived for more than a year. The only abnormality that we observed was that female knockout mice were less able than control mice to attend to litters, either their own mutant litter or wild type substituted control litter. As a consequence, when mouse pups were cared for by neurexin I␣-deficient females, more pups died independent of the genotype. These data indicate that the neurexin I␣ mutation does not cause a major impairment in mouse survival or brain functions but may have subtle behavioral effects. Exact definition of these potential behavioral changes will require extensive behavioral analyses.
The lack of a strong phenotype in the neurexin I␣-deficient mice raised the possibility that we mutated an inactive pseudogene instead of an active gene. To address this possibility, we raised affinity-purified polyclonal antibodies to the cytoplasmic tails of neurexins I and III and analyzed membrane proteins from wild type and knockout mice by immunoblotting. Because these antibodies were raised against the cytoplasmic tails of neurexins, they recognize both ␣and ␤-neurexins, and because the cytoplasmic tails of neurexins are so similar, the antibodies are weakly cross-reactive with other neurexins. Immunoblotting revealed that in wild type mice, a cluster of bands of 160 -200 kDa and a less diverse set of bands of 90 -100 kDa were reactive with neurexin I and III antibodies (Fig. 3). The larger proteins correspond in size to ␣-neurexins, and the smaller proteins correspond to ␤-neurexins. The multitude of bands that we observed for the ␣-neurexins agrees well with their extensive alternative splicing (7) and provides evidence that this alternative splicing indeed results in different protein products. Analysis of the knockout mice by immunoblotting showed that the signal for the neurexin I␣ bands was greatly diminished (Fig. 3). The neurexin III␣, I␤, and III␤ signals, however, were unchanged. Some residual 160 -200-kDa reactivity was observed with the neurexin I antibodies in the neurexin I␣ knockouts, probably because of the cross-reactivity of the neurexin I antibody with neurexins II and III (5). Together these data confirm that we introduced a mutation into an active neurexin I␣ gene that interferes with the expression of neurexin I␣.
␣-Latrotoxin Binding Is Impaired in the Neurexin I␣ Knockouts-We prepared brain membranes from wild type mice, neurexin I␣ knockout mice, and two different control knockout lines and studied binding of 125 I-labeled ␣-latrotoxin to these membranes in the presence and the absence of Ca 2ϩ (Fig. 4). In the knockouts, a major decrease of Ca 2ϩ -dependent binding of ␣-latrotoxin was observed. In contrast, no major changes in Ca 2ϩ -independent binding were detected. To gain insight into the relative affinities of Ca 2ϩ -dependent and Ca 2ϩ -independent ␣-latrotoxin binding sites, we performed Scatchard plot analyses of ␣-latrotoxin binding to brain membranes from wild type and knockout mice. These analyses revealed that Ca 2ϩdependent and Ca 2ϩ -independent binding exhibit similar affinities (Fig. 5). Ca 2ϩ -dependent binding accounted for a major

FIG. 2. Southern blot analysis of neurexin I␣ knockout mice.
Genomic tail DNA from mice offspring from matings between heterozygous mutant neurexin I␣ mice was analyzed by digestion with SpeI and hybridization with the probe shown in Fig. 1. WT, wild type; KO, knockout.
part of the total binding capacity but was largely absent in the knockout mice, suggesting that neurexin I␣ constitutes the major Ca 2ϩ -dependent high affinity binding site for ␣-latrotoxin in mice. In addition, a smaller decrease in the Ca 2ϩindependent binding of ␣-latrotoxin was occasionally observed in the knockout mice (Fig. 5). The Ca 2ϩ -dependent binding site has the same affinity and a similar abundance as the high affinity Ca 2ϩ -independent binding site that is probably provided by the CIRL/latrophilin protein (12,13).

␣-Latrotoxin Still Activates Neurotransmission in Neurons Lacking
Neurexin I␣-To test if ␣-latrotoxin still stimulated neurotransmitter release in neurexin I␣-deficient neurons, we cultured hippocampal neurons from knockout and wild type mouse embryos and studied miniature excitatory postsynaptic currents as a function of ␣-latrotoxin. 1 nM ␣-latrotoxin, a concentration considerably higher than the K D values of the Ca 2ϩ -dependent and Ca 2ϩ -independent binding sites (Fig. 5), was applied in the presence or the absence of Ca 2ϩ . An increase in neurotransmitter release was observed in both types of mice, with or without Ca 2ϩ (data not shown). These data demonstrate that neurexin I␣-deficient mice still respond to relatively high concentrations of ␣-latrotoxin, suggesting that neurexin I␣ is not required for ␣-latrotoxin stimulated neurotransmitter release. However, it is difficult to quantitate the number of synapses whose release events give rise to the signal in electrophysiological experiments under the particular conditions used here. As a consequence, the electrophysiological results give no insight into the relative magnitude of the effect of ␣-latrotoxin in the two types of mice. containing or Ca 2ϩ -deficient buffer, and then stimulated release with ␣-latrotoxin at low concentration (0.5 nM) or with sucrose (0.5 M). When we superfused wild type synaptosomes with ␣-latrotoxin in the presence of Ca 2ϩ , we observed a large prolonged increase in [ 3 H]glutamate release (Fig. 6A). [ 3 H]Glutamate release was activated by ␣-latrotoxin within 30 s and lasted for at least 10 min, the length of the experiment. In Brain membranes from wild type mice (ϩ/ϩ) and neurexin I␣ knockout (k/o) mice (Ϫ/Ϫ) were analyzed by immunoblotting with antibodies to the cytoplasmic tail of neurexin I (A) or of neurexin III (B), to dynamin (C), and to rab3A (D). The antibodies to neurexins I and III used in A and B were raised against their highly homologous cytoplasmic tails and still exhibit partial cross-reactivity ater affinity purification. The antibodies recognize in brain membranes proteins of 160 -200 kDa in size that are likely to correspond to ␣-neurexins (Nrx I␣, neurexins I␣; Nrx III␣, neurexin III␣), and proteins of 90 -100 kDa that are probably made up of ␤-neurexins (Nrx I␤, neurexin I␤; Nrx III␤, neurexin III␤). Especially the ␣-neurexins exhibit size heterogeneity in agreement with their extensive alternative splicing (7). Note that there is no major change in the immunoreactivities of the proteins shown in the knockout except for a loss of neurexin I␣.

Quantitation of [ 3 H]Glutamate Release from Synaptosomes
FIG. 4. ␣-Latrotoxin binding to brain membranes from wild type mice, neurexin I␣ knockout mice, and control knockout mice. Binding of 0.5 nM 125 I-labeled ␣-latrotoxin to crude membranes from individual mouse brains was measured in the absence and the presence of 25 nM unlabeled ␣-latrotoxin to control for nonspecific binding. Brains were from wild type control mice, neurexin I␣ knockout mice, and two lines of control knockout mice. Data show the means Ϯ S.D. (n ϭ 3) of specific binding measured in the presence of Ca 2ϩ or EDTA. Ca 2ϩ -independent binding equals the binding of ␣-latrotoxin in the presence of EDTA, and Ca 2ϩ -dependent binding was calculated by subtracting Ca 2ϩ -independent binding from total ␣-latrotoxin binding obtained in the presence of Ca 2ϩ . neurexin I␣-deficient synaptosomes, however, the extent of [ 3 H]glutamate release stimulated by ␣-latrotoxin was significantly decreased (Fig. 6A). These data could either mean that neurexin I␣ contributes to ␣-latrotoxin action in the presence of Ca 2ϩ and is required for full action of ␣-latrotoxin or that there is a general decrease in the amount of [ 3 H]glutamate that can be released from neurexin I␣-deficient synaptosomes.
To differentiate between these two hypotheses, we quantitated the amount of [ 3 H]glutamate release triggered by hypertonic sucrose and ␣-latrotoxin in the absence of Ca 2ϩ . Hypertonic sucrose was used because it had been shown in electrophysiological studies to trigger exocytosis of docked vesicles and can therefore be used as an indirect measure of the releasable pool of glutamate (23). A 30-s pulse of 0.5 M sucrose stimulated similar amounts of [ 3 H]glutamate release from wild type and neurexin I␣-deficient synaptosomes, suggesting that there is no principal impairment of neurotransmitter release in the mutants and no major difference in pool size (Fig. 6B). Furthermore, when we applied ␣-latrotoxin in the absence of Ca 2ϩ , no difference in [ 3 H]glutamate release between wild type and neurexin I␣-deficient synaptosomes was observed. Thus the neurexin I␣-deficient synaptosomes exhibit a selective impairment of ␣-latrotoxin-stimulated glutamate release in the presence of Ca 2ϩ . DISCUSSION Neurexins are a family of highly polymorphic cell surface proteins with a receptor like structure. There are three ␣and three ␤-neurexins, each of which is alternatively spliced. Neur- FIG. 5. Scatchard plot analysis of ␣-latrotoxin binding to brain membranes from wild type (A) and knockout mice (B) in the presence and absence of Ca 2؉ . Crude brain membranes from wild type and knockout mice were used to measure binding of 125 I-labeled ␣-latrotoxin at different concentrations (0.17-17.0 nM). The specific binding was calculated for each toxin concentration assayed and analyzed in a Scatchard plot. Calculated affinities (K D ) for the data shown are 0.3 and 0.5 nM for wild type and knockout mice in the presence of Ca 2ϩ and 0.45 nM for both genotype in the absence of Ca 2ϩ . Although the ␣-latrotoxin binding affinities do not change in knockouts, the majority of Ca 2ϩ -dependent binding is lost, and a small, currently unexplained decrease in Ca 2ϩ -independent binding is also occasionally observed. exins interact with at least two endogenous ligands: ␣-neurexins bind to neurexophilin (10), and ␤-neurexins bind to neuroligins (8,9). In addition, neurexin I␣ but not other neurexins bind ␣-latrotoxin with high affinity (3,4,11). The structures and known ligands of neurexins indicate the possibility of dual functions for neurexins as cell adhesion molecules and signal transduction receptors (24,25). However, the exact in vivo roles of neurexins are unclear.
As a first step to probe the functions of neurexins and their role in ␣-latrotoxin action, we have now produced mice that lack neurexin I␣. Our data show that these mice are remarkably normal, suggesting that neurexin I␣ is not an essential gene and not required for basic nervous system functions. This finding suggests that neurexin I␣ is functionally redundant or that it performs a more subtle function that is not immediately apparent in the analysis performed here. The presence of multiple neurexins with overlapping expression patterns (7) would agree well with functional redundancy. This indicates that neurexins may substitute for each other functionally. On the other hand we found that the maternal behavior of the neurexin I␣ knockout mice appears to be abnormal, suggesting that subtle defects may exist in the single neurexin I␣ knockout. Future experiments using knockouts of multiple neurexins and a detailed behavioral analysis will be required to resolve these questions.
␣-Latrotoxin is an excitatory neurotoxin that is a component of black widow spider venom and produces massive neurotransmitter release after binding to presynaptic nerve terminals (reviewed in Ref. 1). Although ␣-latrotoxin binding to neurexin I␣ requires Ca 2ϩ (11), ␣-latrotoxin induces neurotransmitter release even in the absence of Ca 2ϩ , indicating that a second receptor for ␣-latrotoxin may mediate its toxicity in the nerve terminal. Such a receptor was recently identified in CIRL/ latrophilin, a membrane protein that resembles G-proteinlinked receptors and binds ␣-latrotoxin with high affinity in the absence of Ca 2ϩ (12,13).
It is puzzling that neurons should express two distinct ␣-latrotoxin receptors with different binding properties (Ca 2ϩ -dependent versus Ca 2ϩ -independent) but similar affinities. Although the two types of ␣-latrotoxin receptors have been characterized thoroughly biochemically, it is unknown if they function as ␣-latrotoxin receptors in vivo. To address this question, we analyzed ␣-latrotoxin action in the neurexin I␣-deficient mice that we had generated. Our data demonstrate that in the neurexin I␣-deficient mice, most of the Ca 2ϩ -dependent ␣-latrotoxin binding activity is lost, whereas Ca 2ϩ -independent binding is unchanged. The ␣-latrotoxin affinities of neurexin I␣ and CIRL are very similar, as is the relative abundance of the Ca 2ϩ -dependent binding site due to neurexin I␣ and the Ca 2ϩindependent binding site presumably due to CIRL. Thus neurexin I␣ accounts for the bulk of the Ca 2ϩ -dependent high affinity binding of ␣-latrotoxin. We found, however, that ␣-latrotoxin is still capable of triggering neurotransmitter release in cultured hippocampal neurons or synaptosomes from neurexin I␣-deficient mice. Therefore neurexin I␣ is not absolutely required for the excitotoxic action of ␣-latrotoxin. Nevertheless, we observed that in the presence of Ca 2ϩ , glutamate release triggered by ␣-latrotoxin from synaptosomes is decreased in the knockout mice compared with wild type mice. If release is stimulated by ␣-latrotoxin or by sucrose in the absence of Ca 2ϩ , no change is observed, indicating that the general release apparatus is intact. This finding also agree well with the electrophysiological data and the mild phenotype of the knockout mice. Viewed together, our data demonstrate that neurexin I␣ is not essential for the ability of ␣-latrotoxin to trigger neurotransmitter release but contributes to ␣-latrotoxin action in the presence of Ca 2ϩ .
What could the action of neurexin I␣ be compared with that of CIRL? CIRL is the presumptive Ca 2ϩ -independent ␣-latrotoxin receptor that probably mediates the ␣-latrotoxin-dependent activation of neurotransmitter release observed in the neurexin I␣ knockout mice. Our finding that in the presence of Ca 2ϩ , neurexin I␣ is required for a full response to ␣-latrotoxin at low concentrations suggests two possibilities. Either neurexin I␣ and CIRL represent independent pathways for ␣-latrotoxin action that work in parallel, or neurexin I␣ assists CIRL in the presence of Ca 2ϩ in triggering neurotransmitter release. In either case, neurexin I␣ represents a target for ␣-latrotoxin. Futhermore, it is possible that ␣-latrotoxin may have as yet unidentified effects other than triggering neurotransmitter release, which could be mediated by neurexin I␣. Taken together, with the identification of two structurally different, presumably cooperative receptors the action of ␣-latrotoxin is much more complex than previously envisioned.