α-Latrotoxin Induces Exocytosis by Inhibition of Voltage-dependent K+ Channels and by Stimulation of L-type Ca2+ Channels via Latrophilin in β-Cells*

The spider venom α-latrotoxin (α-LTX) induces massive exocytosis after binding to surface receptors, and its mechanism is not fully understood. We have investigated its action using toxin-sensitive MIN6 β-cells, which express endogenously the α-LTX receptor latrophilin (LPH), and toxin-insensitive HIT-T15 β-cells, which lack endogenous LPH. α-LTX evoked insulin exocytosis in HIT-T15 cells only upon expression of full-length LPH but not of LPH truncated after the first transmembrane domain (LPH-TD1). In HIT-T15 cells expressing full-length LPH and in native MIN6 cells, α-LTX first induced membrane depolarization by inhibition of repolarizing K+ channels followed by the appearance of Ca2+ transients. In a second phase, the toxin induced a large inward current and a prominent increase in intracellular calcium ([Ca2+]i) reflecting pore formation. Upon expression of LPH-TD1 in HIT-T15 cells just this second phase was observed. Moreover, the mutated toxin LTXN4C, which is devoid of pore formation, only evoked oscillations of membrane potential by reversible inhibition of iberiotoxin-sensitive K+ channels via phospholipase C, activated L-type Ca2+ channels independently from its effect on membrane potential, and induced an inositol 1,4,5-trisphosphate receptor-dependent release of intracellular calcium in MIN6 cells. The combined effects evoked transient increases in [Ca2+]i in these cells, which were sensitive to inhibitors of phospholipase C, protein kinase C, or L-type Ca2+ channels. The latter agents also reduced toxin-induced insulin exocytosis. In conclusion, α-LTX induces signaling distinct from pore formation via full-length LPH and phospholipase C to regulate physiologically important K+ and Ca2+ channels as novel targets of its secretory activity.

property has been extensively exploited to investigate the molecular mechanisms underlying exocytosis (1,2). Toxin action requires first the binding to a surface receptor, and three distinct receptors for ␣-LTX have been identified: the latrophilins (LPHs), which contain a large extracellular adhesion molecule domain and a C-terminal portion bearing the signature of G-protein-coupled receptors (3,4), neurexin Ia and ␤ (2), and the receptor-like protein-tyrosine phosphatase (5). It is generally accepted that the toxin inserts subsequently as a tetramer into membranes to form a stable, cation-permeable pore (6), and the ensuing Ca 2ϩ influx plays a major role in the activation of exocytosis. Indeed, expression of a C-terminally truncated form of LPH lacking all except the first transmembrane domain is sufficient to establish toxin-induced pores leading to calcium influx in epithelial HEK293 cells and sensitization of exocytosis in chromaffin cells (7)(8)(9).
Although these findings indicate that receptor-mediated signal transduction is not required for the action of ␣-LTX, other observations suggest that pore-mediated Ca 2ϩ influx is not sufficient to explain the action of the toxin. ␣-LTX sensitizes exocytosis to Ca 2ϩ in chromaffin cells and in synaptosomes (10,11). Moreover, a point mutated toxin increases exocytosis in the absence of ion fluxes through a toxin-induced pore (12). The intracellular actions provoked by ␣-LTX to induce exocytosis are not fully resolved, apart from pore-mediated Ca 2ϩ influx. Depending on the system, they may implicate phospholipase C with subsequent activation of protein kinase C and of release of Ca 2ϩ from the intracellular stores (10,11,(13)(14)(15).
We have previously demonstrated that ␣-LTX receptors are also expressed on primary ␤-cells and the toxin induces exocytosis of the peptide hormone insulin (16). Clonal ␤-cells cell lines provide a useful model for toxin-induced exocytosis of large dense core vesicles, because they differentially express LPH: whereas high levels of LPH are found in the toxin-sensitive MIN6 cells, HIT-T15 cells express only very low amounts and are toxin-insensitive (16). This situation is clearly distinct from PC12, chromaffin, or HEK293 cells, which are toxin-sensitive (7)(8)(9). ␤-Cells therefore provide a very suitable model and, moreover, ion channels, signal transduction events and the molecular mechanism underlying insulin exocytosis are relatively well characterized (17)(18)(19).
Using this model in combination with truncated or full-length receptors, we addressed the issue of receptor-mediated signaling and pore formation-dependent events in the action of latrotoxin on intracellular calcium, membrane conductances, and exocytosis. To obtain a better understanding of the underlying events, we also compared the effect of native ␣-LTX to the action of recombinant mutated latrotoxin, LTX N4C , which is devoid of pore formation (12). Our data demonstrate for the first time a regulation of K ϩ channels and of L-type Ca 2ϩ chan-nels by latrotoxin that is mediated by its receptor latrophilin via phospholipase C and contributes to the secretory effect.

EXPERIMENTAL PROCEDURES
Materials-Monoclonal anti-Myc and anti-HA antibodies were purified from culture medium of myeloma cells generously provided by Dr. K. Matter (Université de Genève), polyclonal anti-Myc antibodies were obtained from Research Diagnostics (Flanders, NJ). Anti-Ca v 1.2 and anti-Ca v 1.3 antibodies were obtained from Alomone (Jerusalem, Israel). ␣-LTX was prepared and iodinated in Dr. Ushkaryov's laboratory (12). Commercial preparations used (Calbiochem) gave qualitatively similar results. Recombinant LTX N4C was produced and purified as described previously (12). Protein kinase and phospholipase C inhibitors, A23187, apamin, and iberiotoxin were from Calbiochem, calciseptine from Latoxan (Apt, France). The m9 subclone of MIN6 cells (20) was kindly provided by Dr. S. Seino (Chiba, Japan) and was used throughout this study. PC12 cells were generously provided by Dr. B. Rudkin (Ecole Nationale Supérieure, Lyon, France) and used between passages 2 and 9.
Molecular Cloning-The N-terminal HA-affinity epitope was obtained by inserting a sequence coding for HRQDLPGNDNSTAGNS between amino acids 24 and 25. To this end the adjacent sequences were amplified from full-length LPH using the primers pairs N1 (5Ј-GCGG-TACCTGGATAGCGGTTTGACTC-3Ј; 5-GCGGTACCTGGATAG-CGGTTTGACTC-3Ј) and N2 (5Ј-CGGGATCCAGGGCATAACGT-AGATACGG-3Ј; 5Ј-GGAATTCCCTGAGCCGGGCTGGAC-3Ј). The amplicons were subcloned into pKSϩ at the restriction sites ClaI-/KpnI and BamHI/EcoRI, respectively. Two synthetic oligonucleotides (5Ј-AATTACCTGCTGTACTGTTATCATTTCCCGGGAGATCT-TGT-3Ј; 5Ј-CGACAAGATCTCCCGGGAAATGATAACAGTACA-GCAGGT-3Ј) were annealed and inserted into the EcoRI/ClaI sites. The construct was subsequently excised by restriction with HindIII and BamHI and subcloned into the HindIII/BamHI sites of LPH constructs. LPH-TD1 Myc (LPH 1-890 ) was generated by PCR from LPH-HA using the primer 5Ј-GTGGTATATGATGGTGCC-3Ј and 5Ј-ATCGAATT-CGTGGATGGTGTTGCGGTCGG-3Ј. The amplicon was digested with XhoI/EcoRI and subcloned into the corresponding sites of pcDNA3ϩ encoding an in-frame 3Ј-Myc tag and complemented by insertion of the 5Ј-portion of LPH into the HindIII/XhoI sites. To introduce a C-terminal Myc tag, a C-terminal fragment of LPH was amplified to replace the stop codon by an EcoRI site using the primers 5Ј-CGAATCCGGAGGATGTGG-3Ј and 5Ј-ATCGAATTCGAGAC-TAGTGACCAACTGC-3Ј. The amplicon was digested with KpnI/ EcoRI, subcloned into the KpnI/EcoRI sites of pcDNA3ϩ encoding an in-frame 3Ј-Myc tag and complemented by insertion of the 5Ј-portion of LPH into the HindIII/KpnI sites. All sequences were verified by sequencing of both strands. The generation of LPH-TD1-7 and of LPH-TD1-5 has been described before (21). The fusion protein syt2-C 2 AB-eGFP was constructed by PCR amplification of syt2-C 2 AB (amino acids 101-422) and insertion in-frame into peGFP.
Cell Culture, Transient Transfection, Secretion, and Immunoblotting-Cell culture, transient transfection, and secretion assays for endogenous insulin or human C-peptide were performed as described previously (22,23) using enzyme-linked immunosorbent assay. Human growth hormone was used as reporter gene in the case of MIN6m9 cells. To this end the open reading frame of human growth hormone was excised by EcoRI from the plasmid pMMTV-GH (Nichols Institute, San Clemente, CA) and inserted into the corresponding sites of pcDNA3.1ϩ. In control transfections plasmids encoding for LPH or its truncated forms were replaced by peGFP. For immunoblotting, cells were detached (23), resuspended in cold phosphate-buffered saline with 1% Triton X-100, incubated for 30 min on ice, and disrupted by brief sonication. Proteins were solubilized in sample buffer for 30 min at 37°C and separated on 8% SDS-PAGE. Protein blotting, antibody incubation, and detection were performed as described previously (23) except that liquid transfer was used (8 h, 150 mA).
Immunohistochemistry and Imaging-To observe surface labeling, cells were incubated with primary antibody at 4°C on ice, washed and fixed for 5 min with 5% paraformaldehyde on ice before adding the second antibody. To observe intracellular labeling under the same conditions, primary antibodies were added after fixation. All procedures were carried out on ice. Imaging was performed using an inverted microscope (Nikon TD300 equipped with a Z-drive) coupled to a monochromator (Till Photonics) and appropriate emission filters. Images were recorded by a charge-coupled device camera (Micromax 1300Y HS, Roperts Scientific) using Metamorph software (Universal Imaging) and deblurred by deconvolution (Autodeblur, Universal Imaging). The same set-up was used for imaging of living cells. In this case cells on coverslips were kept in 1 ml Krebs-Ringer buffer (KRB) (23) supplemented with 0.05% BSA (KRB-BSA) at 37°C on a heated stage during acquisition. Toxin in KRB-BSA was pressure ejected (5 p.s.i., 10 s) from a micropipette held at ϳ20 mm from the cell.
Carbon Fiber Amperometry-Pheochromocytoma PC12 cells (50,000 cells per 35-mm dish) were transfected for 8 h using Lipofectamine and assayed 48 h later. Prior to recording, cells were loaded for 1 h (1 mM dopamine, 1 mM ascorbic acid, pH 7.4, in culture medium) and washed twice. During amperometry cells were kept at ambient temperature (23°C) in modified Ringer buffer (145 mM NaCl, 5 mM KCl, 1.3 mM MgCl 2 , 3 mM CaCl 2 , 20 mM Hepes, pH 7.4, 10 mM glucose) on the stage of the inverted microscope described above and cotransfected cells identified by fluorescence of eGFP. Cells were stimulated by a 20-s ejection at 5 p.s.i. from a micropipette mounted on an Eppendorf microinjector (Femtojet) held at ϳ2 cell diameters (40 m) from the cell to be recorded. Only single, round cells were used, and carbon fibers were positioned using a piezoelectric driver (PCS-1000, Burleigh Instruments). The micropipette and the carbon fiber (ProCFE, 5 m, Axon Instruments) held at 700 mV were kept at approximately the same distance throughout experiments. Fibers were used only if the root mean square was Ͻ0.5 nA at 700 mV. Currents were recorded using an EPC9 at 2.5 kHz, low pass filtered at 1 kHz, and spikes were analyzed with a program generously provided by Dr. F. Borges (Universidad de La Laguna, Tenerife, Spain) (24).
Measurements of [Ca 2ϩ ] i -HIT-T15 cells grown on coverslips were cotransfected with LPH/CIRL constructs and a plasmid expressing DsRed fluorescent protein to identify transfected cells. 72 h after transfection, cells were loaded with 13 mM of the fluorescent probe indo pentaacetoxymethyl ester (indo-1/AM, Sigma) in KRB (3 mM glucose) for 45 min at 37°C, washed and maintained at room temperature in KRB (3 mM glucose and 0.05% BSA) before fluorescence measurements. MIN6 cells were handled identically except that transfection was omitted. [Ca 2ϩ ] i measurements were carried out as already described (25). Test substances were applied in KRB (3 mM glucose and 0.05% BSA) through a "pouring" pipette with a tip opening of 10 -20 mm and positioned at a distance of 50 -100 mm from the cell. The absence of mechanical artifacts due to drug application was confirmed by applying external medium (KRB) to the cells.
Electrophysiological Recordings-The whole cell recording mode of the patch clamp technique was used as already described (25). Transiently transfected cells were identified as given above. KRB was used as standard extracellular solution contained, and the osmolality was adjusted to 300 -310 mosM/kg with sucrose. The recording pipette was filled with an artificial intracellular saline containing (in mM): 150 potassium chloride, 2 MgCl 2 , 1.1 EGTA, and 5 HEPES (pH 7.3 Ϯ 0.01 with KOH; osmolality: 290 mosM/kg). Drugs were applied as described for calcium measurements, and all experiments were performed at room temperature (20 -22°C). To isolate voltage-dependent Ca 2ϩ currents, K ϩ currents were eliminated by replacing K ϩ gluconate used to formulate the intracellular electrolyte with isomolar N-methylglucamine gluconate. The solution was then buffered to pH 7.3 with HEPES-gluconate. Because Ca 2ϩ currents were not always stable and often disappeared during whole cell recording, current-voltage (intensityvoltage) relations for calcium currents were constructed using cells that showed little or no rundown within 10 -15 min after impalement. Results are expressed as mean Ϯ S.E. Each experiment was repeated several times.
Toxin Binding-Cells were transiently transfected in 24 wells and detached 48 h later by the use of 10 mM EDTA in KRB at 37°C. Cells were centrifuged (5 min, 2,000 ϫ g, 4°C) and resuspended in KRB containing 1.3 mM CaCl 2 but no EGTA. Aliquots were kept for cell counting and protein determination. 200 ml of cell suspension was transferred to 1.5-ml Eppendorf tubes and binding initiated by addition of radioiodinated toxin (26) for 15 min at 37°C. The binding was stopped by the addition of ice-cold buffer and immediate centrifugation at 10,000 ϫ g for 5 min at 4°C. The pellets were washed once with ice-cold KRB, centrifuged at 10,000 ϫ g for 2 min at 4°C, and counted for radioactivity in a ␥-counter. Nonspecific binding was determined in the presence of 50 nM native toxin.

Surface Expression of Full-length and C-terminally Truncated LPH
Constructs-C-terminal truncations of LPH, which lack the domains homologous to G-protein-coupled receptors, have been reported to suffice for actions of ␣-LTX in several cell systems, although these constructs are not always efficiently expressed (7,8,21). In our attempt to test their function in toxin-mediated exocytosis of insulin-containing large dense core vesicles we transiently expressed full-length and truncated LPH in toxin-insensitive hamster HIT-T15, or in mouse MIN6 insulinoma cells, which express endogenous LPH and are toxin-sensitive (16). Three truncated constructs were made: LPH-TD1, terminated after the first transmembrane domain; LPH-TD1-5 (amino acids 1-1020), limited to five transmembrane domains; and LPH-TD1-7 (amino acids 1-1099), truncated after the last transmembrane domain.
We examined first their surface expression after transient transfection using constructs, which carry an HA tag at the extracellular N terminus and an Myc tag at their intracellular C termini (Fig. 1A). To identify surface expression, antibody binding to HA tags was conducted on ice prior to fixation. The results observed confirmed surface expression of LPH and LPH-TD1 in HIT-T15 cells as well as LPH-TD1 in MIN6 cells (Fig. 1B). Surface expression was also observed for LPH-TD1-7, whereas LPH-TD1-5 was visualized only in a small number of cells (data not shown). As shown in Fig. 1C, expression levels of total LPH and LPH-TD1 were comparable. The antibody used is directed against the N-terminal hemagglutinin epitope. Latrophilin is processed in the endoplasmic reticulum into two non-covalently bound subunits, N-terminal p120 and C-terminal p85 (25). Only the 120-kDa extracellular N-terminal form was detected on Western blots, which indicated that transiently expressed full-length LPH was completely cleaved in HIT-T15 cells.
To further quantify the expression, we analyzed binding of 125 I-␣-LTX to transiently expressed constructs in intact HIT-T15 cells (Fig.   1D). Similar affinities (K d values of ϳ0.25 nM) were observed for LPH, LPH-TD1, LPH-TD1-7, and LPH-TD1-5, whereas only negligible binding was detected in control cells. Moreover, the amount of binding sites expressed was comparable for all constructs except for LPH-TD1-5. These observations indicated that LPH, LPH-TD1, and LPH-TD1-7 were efficiently expressed and bound the toxin with comparable characteristics in insulin-secreting cells. To assess whether these proteins are capable to mediate ␣-LTX-induced effects, we determined functional responses of transiently expressed LPH or its truncated forms first on exocytosis of large dense core vesicles.
Full-length LPH Mediates Exocytosis Induced by ␣-LTX or LTX N4C -Next we determined the effects of LPH and LPH-TD1 in ␣-LTX-induced exocytosis and transiently cotransfected these receptors in HIT-T15 cells with a plasmid coding for human prepro-insulin as reporter gene ( Fig. 2A). Membrane depolarization by 48 mM KCl induced a 3-to 4-fold increase, which was not altered by expression of the different constructs. As expected, ␣-LTX up to 2 nM did not induce release of human C-peptide in control transfected HIT-T15 cells, whereas cells expressing full-length LPH increased human C-peptide release more potently than KCl to ϳ10-fold of basal secretion. The LPH construct LPH-TD1-7 behaved similarly though a slightly greater efficacy was constantly observed. Cells expressing LPH-TD1 demonstrated only a marginal increase in toxin-evoked hormone secretion ( Fig. 2A). This observation stems from a large series of experiments done at different passage numbers. Similar results were obtained using LPH-TD1 devoid of HA or Myc tags (data not shown).
LPH-TD1 has been shown to suffice for ␣-LTX binding and induction of exocytosis in neuroendocrine PC12 cells (7,8). One major difference between PC12 and HIT-T15 cells resides in the expression levels of endogenous LPH. For this reason we also tested the effect of LPH-TD1 expression in MIN6 cells, known to express endogenous LPH receptor (16) (Fig. 2B). In control cells, ␣-LTX increased the release of reporter gene product. Co-expression of full-length LPH sensitized secretion to ␣-LTX, and this effect was also apparent in the case of expression of LPH-TD1.
We also tested whether a mutant toxin, LTX N4C , can induce secretion, because this protein does not assemble in pore-forming tetramers (21). As shown in Fig. 2 (C and D), LTX N4C induced insulin secretion in the subnanomolar range. Again transient expression of LPH was required in HIT-T15 cells to observe an effect (Fig. 2C), whereas responses similar to those induced by 48 mM KCl were observed in native MIN6 cells (Fig. 2D).
Our LPH-TD1 construct differed by one amino acid from those truncated forms, which have been shown to enhance sensitivity to ␣-LTX in pheochromocytoma cell line PC12 cells (7,8). To control further for expression and function of our construct we also measured exocytosis in PC12 cells. Exocytosis assessed by amperometry in non-transfected cells demonstrated a vigorous secretory response after application of 1 nM ␣-LTX (Fig. 3, A and C) but only a spurious response in the case of 0.1 nM toxin (concentrations in the pipette). We therefore used this concentration to test whether LPH or our LPH-TD1 construct may sensitize these cells to the effects of the toxin. Transient expression of either construct sensitized cells to the toxin to similar extent (Fig. 3, A and C). In addition, mutant toxin LTX N4C induced exocytosis albeit to a lesser extent than wild-type toxin (Fig. 3, B and C). This demonstrates that the construct used by us behaved similar to the ones previously reported in PC12 cells (7,8) and that LPH-TD1 is indeed sufficient to sensitize exocytosis in cells expressing already endogenous LPH.
The requirement of full-length LPH for ␣-LTX-induced secretion in HIT-T15 cells as well as the effects of LTX N4C in clonal ␤-cells and PC12 cells clearly indicated the presence of signaling pathways distinct from pore-induced influx of calcium and other ions. We therefore investigated these putative pathways by examining first the effects of the toxins on [Ca 2ϩ ] i .   4A), and two different patterns were observed. First, a transient effect, termed phase I, appeared (Fig. 4A). This phase is characterized by either the appearance of calcium spikes or an increase in calcium spikes amplitude and frequency in already active cells. It lasted from 40 to 550 s with a mean of 211 Ϯ 162 s (n ϭ 26). Second, a long-lasting increase was observed with a clear plateau-phase, termed phase II (Fig. 4A). This response was still present at the end of the recordings and lasted always more than 550 s (n ϭ 17).

Full-length LPH, but Not C-terminally Truncated LPH-TD1, Mediates ␣-LTX-induced Oscillations in [Ca
In cells responding by both phases (IϩII) the extent of the [Ca 2ϩ ] i rise above baseline (370 Ϯ 12 nM, n ϭ 13) was comparable to that induced by KCl-evoked depolarization (381.5 Ϯ 26.5 nM, n ϭ 7). In cells responding by type I or type II alone the amplitude of the response was lower (phase I alone: 130 Ϯ 11.5 nM, n ϭ 13; type II alone: 85 Ϯ 8.5 nM, n ϭ 4). 30 out of 43 HIT-T15 cells expressing LPH responded, and the absence of response may in part be due to the method employed to detect cotransfected cells (see "Experimental Procedures"). 13 cells showed only phase I, 4 only phase II, and 13 phase I and II.
In the case of transient expression of the LPH-TD1, phase I responses were never observed and 4 of 11 cells responded with phase II (Fig. 4B). The amplitude of the responses was comparable to that observed in HIT-T15 cells expressing LPH (phase II alone): 97.5 Ϯ 8.5 nM (n ϭ 4). Exposure to ␣-LTX increased [Ca 2ϩ ] i also in MIN6 cells, which express endogenous LPH (Fig. 4C). All cells tested (18) responded to the toxin, and a large majority (12) demonstrated only the transient phase I response, whereas concomitant phase I and II were present in 6 cells (phase I: 191.5 Ϯ 10 nM, n ϭ 12; phase II: 336.5 Ϯ 24 nM, n ϭ 6).
LTX N4C Increases [Ca 2ϩ ] i in MIN6 Cells by Calcium Mobilization and by Influx Through L-type Voltage-dependent Ca 2ϩ Channels-Our preceding observations suggested that the effect of ␣-LTX includes two distinct phases and that phase I differs from pore formation. We therefore examined the effects of LTX N4C in MIN6 cells, because this recombinant toxin should not induce pore formation. As shown in Fig. 5A, the mutated toxin provoked a sharp, transient rise in [Ca 2ϩ ] i in all cells tested (332 Ϯ 8 nM above baseline; n ϭ 46). In stark contrast to the native toxin, LTX N4C never induced a plateau phase (compare Figs. 5A and 4C). We sometimes observed a slowly decreasing phase as shown in Fig. 5A (trace a; 14 out of 46 cells). In the absence of extracellular calcium [Ca 2ϩ ] i still increased by 91 Ϯ 13 nM in 15 out of 28 cells (see traces a and b). This suggests that LTX N4C is capable of inducing the mobilization of Ca 2ϩ from intracellular stores.
To determine the implication of Ca 2ϩ influx and mobilization we employed the L-type channel blocker calciseptine (27)  A rise in [Ca 2ϩ ] i induces exocytosis, and binding of C2-domain-containing proteins, such as the cytosolic domain of synaptotagmin, to the plasma membrane is thought to be important for this process (33,34). We employed this mechanism to compare further the effect of ␣-LTX and LTX N4C in living MIN6 cells. To this end MIN6 cells were transiently transfected with a plasmid expressing the two cytosolic C2-domains of synaptotagmin 2 linked to eGFP (syt2-C 2 AB-eGFP). As shown in Fig. 5B (upper panel), exposure of these cells to 1 nM ␣-LTX resulted in complete translocation to the plasma membrane, whereas in the case of LTX N4C the extent of translocation was less prominent (n ϭ 4 for each condition). Analysis of the time course revealed that ␣-LTX provoked first a partial translocation followed by complete translocation of syt2-C 2 AB-eGFP. In contrast, LTX N4C induced only partial translocation with clear oscillatory patterns (Fig. 5B, lower panel).
Because inhibitors of L-type calcium channels and PKC abolished the effect of LTX N4C on [Ca 2ϩ ] i , we asked whether these agents might alter toxin-induced insulin secretion. As expected, the Ca 2ϩ channel blockers nifedipine and calciseptine reduced secretion subsequent to membrane depolarization by KCl (Fig. 5C, upper panel). The blockers also reduced secretion evoked by LTX N4C indicating the role of L-type channels in its action. Nifedipine did not alter hormone release induced by the calcium-ionophore A23187 underscoring the specificity of the inhibitors at the concentrations used. Similarly the PKC inhibitors Gö 6983, BIS I, and staurosporine reduced LTX N4C -evoked insulin secre-

␣-Latrotoxin Regulates Ion Channels and Exocytosis via LPH
tion, although relatively high concentrations had to be used (Fig. 5C,  lower panel). It has to be noted that these reagents increased KClevoked release in MIN6 cells already at concentrations of 1-5 mM (data not shown).

␣-LTX Induced Membrane Depolarization Distinct from Pore Formation and Inhibited K ϩ Currents in HIT-T15 Cells Expressing LPH-Because the toxin provoked oscillatory increases in [Ca 2ϩ
] i and translocation of syt2-C 2 AB-eGFP, we investigated more closely its effects on membrane potential and the induction of ionic currents by patch-clamp recordings (Fig. 6). Immediately after establishment of whole cell recording, the mean-resting membrane potential of HIT/full-length LPH, HIT/LPH TD1, and MIN6 cells was Ϫ52 Ϯ 2.8 mV (n ϭ 11), Ϫ54 Ϯ 1.9 mV (n ϭ 11), and Ϫ56 Ϯ 1.5 mV (n ϭ 12), respectively.
As could be expected for excitable cells, HIT-T15 and MIN6 cells exhibit spontaneous action potentials of various amplitude and kinetics at rest (see Figs. 6A and 7A). Application of 48 mM K ϩ solution elicited a rapid and reversible membrane depolarization in HIT-T15 cells (Fig.  6A). In native HIT-T15, application of 2 nM ␣-LTX did not elicit any appreciable response (Fig. 6A). In contrast, in HIT-T15 cells expressing LPH, application of 2 nM ␣-LTX induced a slower but stronger membrane depolarization (45-60 mV) than KCl. Toxin effects on membrane potential established more slowly than the observed effects on [Ca 2ϩ ] i , probably due to differences in recording conditions imposed here by the patch-clamp approach, including dilution of cytosolic components in the patch pipette. However, similar to responses in [Ca 2ϩ ] i , two phases were clearly present (as indicated by vertical dashed lines). Action potentials were transiently elicited during the onset of the depolarization indicated as phase I in Fig. 6A. In 3 out of 4 cells, this membrane depolarization was followed by a transient membrane hyperpolarization of various amplitudes (from Ϫ5 to Ϫ30 mV). Then, membrane potential became depolarized and remained so for the rest of the recordings (phase II). In HIT-T15 cells expressing LPH-TD1, ␣-LTX (2 nM) elicited membrane depolarization, although of lower amplitude (10 -20 mV), and membrane potential slowly recovered. In contrast, action potentials indicative of phase I were never recorded (Fig. 6A). Toxin-induced membrane depolarization was insensitive to diazoxide, excluding an involvement of the K ATP channel (data not shown).
We investigated subsequently the time course in more depth in HIT-T15 cells expressing full-length LPH (Fig. 6B). To monitor membrane conductance and K ϩ current, respectively, hyperpolarizing and depolarizing voltage steps of constant amplitude were repeatedly applied (every 10 s, Fig. 6B, trace a). During an initial period ␣-LTX did not induce any significant change on steady-state current (Fig. 6B, trace a, phase I) but slowly reduced the K ϩ current evoked by electrical depolarization from Ϫ40 to ϩ40 mV (Fig. 6B, trace b). The maximal effect was obtained 80 s after the beginning of toxin ejection. During this time no significant effect on basal membrane conductance could be detected (Fig. 6B, trace c). Second, 60 -120 s after application of the toxin a huge inward current appeared (from 600 pA to Ͼ1 nA, Fig. 6B, trace a; phase II as indicated by vertical dashed line) and was associated with an increase in membrane conductance (from 1.1 Ϯ 0.1 nS to 5.35 Ϯ 0.2 nS, n ϭ 7; Fig. 6B, trace c). This current slowly developed in stages and was not reversible within 15 min. It could be responsible for the phase II of the calcium response described above. Furthermore, the driving force (V h Ϫ E rev ) underlying the long-lasting inward current (I) response to the toxin was estimated to obtain an indication on the nature of the ions implicated in this phase II event. Using Ohm's law as applied to membranes (I ϭ G(V h Ϫ E rev ), where E rev is reversal potential), the conductance (G) can be derived by measuring throughout the time course changes in current responses to brief hyperpolarizing commands (20 mV from a holding potential, V h , of Ϫ40 mV). This in turn permitted repeated estimates of E rev . For example, the 24 estimates of the driving force calculated for the cell illustrated in Fig. 6B (trace d) yielded an average reverse potential E rev of ϩ35 Ϯ 2 mV. This value is consistent with the contribution of Ca 2ϩ ions to this current.
Because K ϩ channels participate in the control of the membrane potential (31), we investigated the effect of ␣-LTX on these channels more closely (Fig. 6C). Cells were voltage-clamped at Ϫ40 mV, close to the mean resting membrane potential. Contamination of K ϩ current recordings with Na ϩ was avoided by the use of the Na ϩ channel blocker tetrodotoxin (2 mM) in the external solution. Under these conditions, voltage steps of increasing intensity elicited an outward current of increasing amplitude. This outward current was greatly reduced by external application of 48 mM KCl (data not shown) confirming that it  MARCH 3, 2006 • VOLUME 281 • NUMBER 9 represents a pathway for K ϩ efflux. In HIT-T15 cells expressing LPH, a prominent inhibition of outward currents was observed (Fig. 6C, upper panels) in response to ␣-LTX (ϳ70%), which was observed over the whole voltage range (Ϫ40 to ϩ60 mV). In contrast, in cells expressing LPH-TD1 (Fig. 6C, lower panels), ␣-LTX was ineffective on K ϩ currents elicited by depolarization from Ϫ40 to Ϫ10 mV, and it only slightly reduced K ϩ current (ϳ30%) for higher potentials. These data are in agreement with the current-clamp and [Ca 2ϩ ] i experiments showing that ␣-LTX produces phase I events only in HIT-T15 cells expressing full-length LPH.

␣-Latrotoxin Regulates Ion Channels and Exocytosis via LPH
LTX N4C Induced Oscillation in Membrane Potential and Inhibited BK-type K ϩ Channels in MIN6 Cells via Phospholipase C-We next examined the behavior ␣-LTX and of LTX N4C in MIN6 cells. The K ϩ channel blocker TEA (5 mM) is capable to elicit oscillations of the membrane potential as shown in Fig. 7A. Application of wild-type ␣-LTX induced initially similar oscillations, and only later constant membrane depolarization was observed. The two events were clearly distinct and therefore again indicative for the presence of two phases. In stark contrast, the mutated toxin LTX N4C provoked only oscillations of the membrane potential of considerable amplitude (Fig. 7A).
Using the same protocol as described for Fig. 6B, we investigated the time course and compared it to other inhibitors of K ϩ channels. A run-down of the K ϩ outward current was evident under our experimental conditions when comparing the earliest and latest time points in Fig.  7B (trace a). LTX N4C produces a pronounced and transient inhibition of the K ϩ outward current from 455 Ϯ 24 to 98 Ϯ 8 pA (n ϭ 11). This was similar to the first phase observed in LPH-expressing HIT-15 cells upon exposure to ␣-LTX (see Fig. 6B). However, LTX N4C did not produce a second phase as demonstrated by the return to the base line (Fig. 7B,  trace a). A similar effect was produced by iberiotoxin, a peptide inhibitor of BK-type K ϩ channels, whereas apamin, an inhibitor of SK-type K ϩ channels, was ineffective in MIN6 cells (Fig. 7B, trace b). Indeed, iberiotoxin reduced outward currents to 94 Ϯ 10 pA (n ϭ 10). Moreover, inhibition of BK-type K ϩ channels by iberiotoxin completely abolished the effect of LTX N4C (Fig. 7Bc; n ϭ 5) indicating that the effects of LTX N4C and of iberiotoxin are mediated by the same type of channels. We next examined whether phospholipase C is implicated in the action of LTX N4C on K ϩ channels (Fig. 7B, trace c). Indeed, U73122 blocked the effect of the recombinant toxin on K ϩ currents, whereas the stereoisomer U73343 was ineffective. Therefore, regulation of K ϩ currents and of [Ca 2ϩ ] i (see above) implied phospholipase C.
The inhibition of the voltage-dependent K ϩ outward current was further characterized in MIN6 cells (Fig. 7D) as before in HIT-T15 cells (see above). LTX N4C provoked a considerable inhibition of the currents and a right-shift of the I-V curve. We next asked whether inhibition of the K ϩ outward currents are per se sufficient to induce insulin secretion in MIN6 cells and may contribute to the secretory activity of the toxin. Indeed, TEA and iberiotoxin alone were capable to induce a significant increase in insulin secretion (Fig. 7E). TEA and iberiotoxin also augmented insulin secretion evoked by 35 mM KCl by 70.3 Ϯ 8.8 and 95 Ϯ 36%, respectively (n ϭ 4).
LTX N4C Transiently Increases L-type Voltage-dependent Ca 2ϩ Currents Independently from Its Action on Membrane Potential-The LTXinduced events on membrane potential are seemingly sufficient to induce opening of voltage-dependent Ca 2ϩ channels and to provoke influx of the cation. To our surprise, LTX N4C had still an effect on calcium currents when MIN6 cells were clamped to 0 mV thus obliterating any contribution from LTX-induced changes in membrane potential (Fig. 8A). LTX N4C enhanced inward currents both at the beginning and throughout the applied pulse. This led to a downward shift of the I-V curve indicating that the effect was not voltage-dependent (Fig. 8B). Note that this increase is considerable and amounts to some 50%. Timecourse experiments using voltage jumps, shown in Fig. 8C, demonstrate the transient nature of the effect induced by LTX N4C as its application is accompanied by a short increase in Ca 2ϩ currents. Insulin-secreting cells express a variety of voltage-dependent Ca 2ϩ channels. In addition to the well established role of L-type channels, R-type channels may also control insulin exocytosis to a certain extent (35). To further determine the nature of the Ca 2ϩ -current involved, we employed the L-type-specific blocker calciseptine, which abolished oscillations in Ca 2ϩ and in insulin secretion evoked by LTX N4C (Fig. 5, A and C). Calciseptine completely blocked the voltage-evoked inward Ca 2ϩ current and abrogated the effect of LTX N4C (Fig. 8D). This blockade of calcium currents and LTX N4C effects was observed throughout the time course (Fig. 8E).
The molecular identity of L-type Ca 2ϩ channels involved in insulin secretion is a matter of debate and may be formed by Ca v 1.2 (␣ 1C ) or Ca v 1.3 (␣ 1D ) (36). As the expression pattern has not been established in the MIN6 subclone used here (m9), we have performed immunoblots using isoform-specific antibodies. As shown in Fig. 8F, both proteins are expressed and may be regulated by LTX N4C .

DISCUSSION
Pancreatic ␤-cells secrete the peptide hormone insulin by exocytosis and require a machinery similar to that of neuroendocrine cells (17).
Our analysis of ␣-LTX action in these cells has now revealed previously unidentified and physiologically important targets, i.e. K ϩ channels and L-type Ca 2ϩ channels. Their regulation required the G-protein-coupling moiety of latrophilin and stimulates exocytosis.
Various ionic events have been observed previously in native endocrine and neuroendocrine cells after application of ␣-LTX, including "spikes" and "plateaus" in primary chromaffin and ␤-cells, but they were mainly interpreted in terms of pore formation (9,37,38). We have now been able to clearly define two distinct phases by the use of different cell lines, receptor constructs, and wild-type or mutated toxin. In phase I oscillatory membrane depolarizations appear as well as calcium spikes. In contrast, during phase II a large and mixed transmembrane conductance appeared representing known pore formation (39) and provoked a sustained increase in [Ca 2ϩ ] i as had been reported for secretory and non-secretory cells (7,9,40). Several observations demonstrated that phase I was independent from pore formation. First, this phase required the full-length receptor and was linked to the inhibition of voltage-dependent K ϩ fluxes. The inhibition of these K ϩ fluxes cannot be explained by pore formation, because it occurred prior to them (see Fig.  6B). Moreover, toxin-induced pores are known to be poorly selective and should induce increases but not reduction in ion fluxes (41). Most notably, phase I did not exhibit any significant effect on steady-state current or basal membrane conductance, which would have been indicative of pore formation. Finally, LTX N4C only induced phase I effects; they were of transient nature and sensitive to an inhibitor of phospholipase C.
Further support for a two-step event is given by observations on ␣-LTX-induced exocytosis of glutamate from synaptosomes that can also be divided into two phases (11). An initial, receptor-and phosphorylation-dependent secretion of neurotransmitters is followed by   MARCH 3, 2006 • VOLUME 281 • NUMBER 9 release depending on pore formation. Therefore, not only mutant, but also wild-type toxin activates specific signaling pathways distinct from pore formation in neuronal and endocrine cells.

␣-Latrotoxin Regulates Ion Channels and Exocytosis via LPH
The specific phase I events required full-length latrophilin, including its G-protein coupling domains in HIT-T15 cells that are naturally toxin-insensitive. LPH-TD1 did not suffice to support efficiently toxininduced stimulation of insulin exocytosis despite increases in [Ca 2ϩ ] i . We do not think that reduced increases in [Ca 2ϩ ] i in LPH-TD1-expressing cells (as compared with LPH-expressing ones, see Fig. 4) explain the absence of measurable exocytosis. Although changes in [Ca 2ϩ ] i were observed only in one-third of the cells expressing LPH-TD1, their amplitude corresponded approximately to those induced by KCl over the stimulation period. Ca 2ϩ channels and exocytotic sites are coupled in ␤-cells (42), and it is feasible that pores induced during phase II in HIT-T15 cells expressing LPH-TD1 may not be coordinated with release sites.
Previous reports suggested that expression of the extracellular domain of LPH (LPH-TD1), if anchored to the plasma membrane, is sufficient for toxin-induced increases in [Ca 2ϩ ] i or enhancement of exocytosis (4,7,8,43). In contrast to HIT-T15 cells, the cellular systems used in these reports (PC12, HEK293, or chromaffin cells) express endogenous LPH and are sensitive to the toxin (7,10,44). Similar to those observations we have observed enhancement of ␣-LTX effects upon overexpression of our LPH-TD1 construct in cells that harbor endogenous LPH, such as PC12 and MIN6 cells. However, this was clearly not the case in HIT-T15 cells that do not contain functional endogenous toxin receptors. This suggests complementation between endogenous full-length and transiently expressed truncated receptor as the base for LPH-TD1 effects. Indeed, G-protein-coupled receptors form oligomers, and truncated receptors can be salvaged by interaction with full-length receptors (45). Cooperativity is also indicated by coimmunoprecipitation of different toxin receptors provided that ␣-LTX is present (5). Moreover, toxin-induced direct interactions have been reported between the extracellular N-terminal portion of LPH, as generated here by LPH-TD1, and its C-terminal membrane-spanning portions (15). Our data also suggest that neurexins may not compensate for truncations in LPH as HIT-T15 cells, but not MIN6 cells, express considerable levels of endogenous neurexin I␣ and I␤ (16).
Our data demonstrate for the first time that two types of ion channels are regulated as downstream effectors by binding of toxin to latrophilin: inward-rectifying K ϩ channels and voltage-dependant Ca 2ϩ channels (VDCCs). As inhibition of phospholipase C abolished membrane depolarization and increases in [Ca 2ϩ ] i , the following sequence of events can be established. Activation of phospholipase C leads to closure of inward-rectifying K ϩ channels, and the ensuing membrane depolarization will induce opening of VDCCs followed by Ca 2ϩ influx. In addition, toxin binding to LPH further stimulated VDCCs independently from its action on membrane depolarization. Clearly, combining both effects will considerable enhance the efficiency of the toxin in terms of exocytosis.
Pharmacological approaches using specific agents identify the VDCC as L-type, which constitutes the major but not only VDCC in ␤-cells in terms of exocytosis (35,36). It is currently still a matter of debate whether Ca v 1.2 or Ca v 1.3 constitutes the major molecular form implicated in insulin secretion (46). Similar to primary cells, HIT-T15 and MIN6 cells express Ca v 1.2 and Ca v 1.3 channels (see Ref. 36 and our study). Thus, both may be targets of LPH-mediated signaling, although Ca v 1.3 seems to be expressed in these cells to a greater extent.
The observed increase in [Ca 2ϩ ] i required the activation of protein kinase C according to pharmacological criteria. L-type VDDCs are sub-ject to phosphorylation by PKC subsequent to the activation of receptors coupled to the G-protein G q , and the functional outcome depends on the splice variant present (47,48). Such a mechanism could eventually underlie the stimulation of L-type VDDCs by LTX N4C observed here, although other pathways may apply (36). In ␤-cells an intriguing parallel to LPH-mediated signaling is given by the recent description on fatty-acid-induced insulin secretion via the G-protein-coupled receptor GPR40 (29). Similar to LPH, GPR40 regulates phospholipase C and L-type VDCCs.
Different K ϩ currents with specific characteristics and roles at distinct stages modulate secretion in ␤-cells, mainly the ATP-sensitive K ATP , the voltage-dependent K V and the calcium-and voltage sensitive K Ca (49). Glucose metabolism alters the ATP/ADP ratio, which leads to the closure of K ATP channels, membrane depolarization, Ca 2ϩ influx via VDCCs, and insulin exocytosis (18). Repolarization of ␤-cells implies mainly voltage-dependent (K v ) and perhaps Ca 2ϩ -activated K Ca channels (50,51). Our observations of voltage-dependent effects and their resistance to diazoxide exclude the K ATP channel as target. Genetic or pharmacological inhibition of K v leads to increased stimulated insulin secretion in primary, HIT-T15, and MIN6 cells (52,53). In addition, HIT-T15 and primary cells express functional voltage-dependent Ca 2ϩactivated K Ca channels (54). As demonstrated here for MIN6 cells by the use of iberiotoxin and apamin, these cells express the BK-, but not the SK-type of these K Ca channels. The observed inhibition of latrotoxin effects on K ϩ currents by iberiotoxin suggests that latrotoxin specifically inhibits BK-type channels. Moreover, inhibition of repolarization by BK channels using iberiotoxin was sufficient to induce insulin exocytosis. Regulation of BK channels by ␣-LTX therefore contributes to the stimulation of insulin exocytosis and may well explain the enhancement of depolarization-induced secretion by the native toxin in chromaffin and primary ␤-cells (37,38,55).
Using pharmacological approaches and taking advantage of the mutant toxin LTX N4C we have in part delineated the downstream signaling of latrophilin in clonal ␤-cells. The increase in [Ca 2ϩ ] i depends on activation of phospholipase C, and similar pathways have been described previously in synaptosomes, in neurons, in neuroblastoma cells transiently expressing latrophilin, as well as in the case of the recently identified latrophilin-like orthologue in Caenorhabditis elegans (12,13,15,56). In concordance with recordings from pyramidal neurons, the observed release from intracellular stores in MIN6 cells implied the inositol 1,4,5-trisphosphate receptor (13). Latrotoxin-induced inhibition of K ϩ channels or Ca 2ϩ influx through VDCCs has not been reported in these neuronal preparations. L-type VDDCs are of physiological importance in neurons, however, they are not directly linked to neuroexocytosis in contrast to exocytosis in endocrine cells (57). As far as BK channels are concerned, their potential role in neuroexocytosis is not fully established (58,59).
Inhibition of calcium influx only partially inhibited toxin-induced exocytosis of insulin. This suggests that LPH-mediated stimulation implies downstream targets in addition to ion channels. Indeed, subnanomolar concentrations of ␣-LTX stimulate exocytosis even in the absence of changes in [Ca 2ϩ ] i (60) and on top of Ca 2ϩ in permeabilized cells (10). Protein kinases A and C are known to augment exocytosis by increasing the size of the releasable pool of vesicles in insulin-secreting cells (61). A most recent detailed analysis demonstrated that both ␣-LTX and LTX N4C increase exocytosis in ␤-cells by sensitizing the release process to calcium via PKC (62).
Collectively, our data support the view that ␣-LTX activates exocytosis via latrophilin employing specific signaling mechanisms in addition to formation of a cation-permeable pore. Most importantly, the receptor-mediated mechanisms are mediated by phospholipase C and impinge on targets relevant for the physiological regulation of insulin exocytosis, namely repolarizing K ϩ currents and L-type calcium channels.