Latrophilin, Neurexin, and Their Signaling-deficient Mutants Facilitate α-Latrotoxin Insertion into Membranes but Are Not Involved in Pore Formation*

Pure α-latrotoxin is very inefficient at forming channels/pores in artificial lipid bilayers or in the plasma membrane of non-secretory cells. However, the toxin induces pores efficiently in COS-7 cells transfected with the heptahelical receptor latrophilin or the monotopic receptor neurexin. Signaling-deficient (truncated) mutants of latrophilin and latrophilin-neurexin hybrids also facilitate pore induction, which correlates with toxin binding irrespective of receptor structure. This rules out the involvement of signaling in pore formation. With any receptor, the α-latrotoxin pores are permeable to Ca2+ and small molecules including fluorescein isothiocyanate and norepinephrine. Bound α-latrotoxin remains on the cell surface without penetrating completely into the cytosol. Higher temperatures facilitate insertion of the toxin into the plasma membrane, where it co-localizes with latrophilin (under all conditions) and with neurexin (in the presence of Ca2+). Interestingly, on subsequent removal of Ca2+, α-latrotoxin dissociates from neurexin but remains in the membrane and continues to form pores. These receptor-independent pores are inhibited by anti-α-latrotoxin antibodies. Our results indicate that (i) α-latrotoxin is a pore-forming toxin, (ii) receptors that bind α-latrotoxin facilitate its insertion into the membrane, (iii) the receptors are not physically involved in the pore structure, (iv) α-latrotoxin pores may be independent of the receptors, and (v) pore formation does not require α-latrotoxin interaction with other neuronal proteins.

One of the major effects of LTX binding to its receptors is the induction of cation-permeable channels, or pores, in the cell membrane (2,6,15). The pores play a profound role in the action of the toxin; (i) influx of Ca 2ϩ or other cations through such pores can cause neurotransmitter release (12, 16 -18), and (ii) cytoplasmic neurotransmitters, including ATP, can leak out directly through these pores (1,12,19,20). However, the mechanism by which LTX induces the pores and the role of the receptors in this action remain largely unknown. The receptors could (i) mediate intracellular signaling leading to the opening of endogenous channels, (ii) participate in the pore structure or (iii) simply tether LTX to the cell surface, thus allowing its insertion into the membrane to form pores. In addition, other (non-receptor) proteins of the exocytotic apparatus could participate in pore induction upon direct stimulation by the toxin. On the other hand, LTX preparations can permeabilize artificial lipid bilayers (21)(22)(23) and form pore-like tetrameric structures in liposomes (24), suggesting that, under some conditions, the toxin is able to make pores even in the absence of any receptors. It is unclear, however, whether LTX acts in the same way upon binding to LPH or NRX and whether these receptors are physically involved in the structure of the LTX pore.
The objective of this work was to analyze the roles that LPH and NRX play in pore formation by LTX. We found that, in contrast to the crude venom, highly purified LTX in physiological concentrations is very inefficient in forming pores in lipid bilayers or receptor-less cells. When expressed in non-secretory cells, the receptors, including signaling-deficient mutants, tether LTX to the plasma membrane with subsequent formation of large pores permeable to ions and small molecules.
These pores remain open in the membrane even after LTX dissociation from the receptor(s) and can be blocked by anti-LTX antibody (Ab). Our findings suggest that LTX is the sole component of the pores but only inserts efficiently into the plasma membrane if recruited by any of its receptors. dectus lugubris as described (25). Briefly, lyophilized venom was dissolved in buffer B (50 mM Tris-HCl, pH 8.3, 0.15 M NaCl) containing 20 mM EDTA, applied on to a high performance gel filtration column KW-803 (Waters) and eluted with buffer A. Fractions containing LTX were subjected to ion exchange chromatography on a Protein-Pak Q AP-1 column (Waters) eluted with a gradient of NaCl. The LTX fraction (corresponding to conventional toxin preparations (34)) was further purified by preparative native electrophoresis using a 3.5-cm-long 6% polyacrylamide gel cast in a model 491 Prep Cell (Bio-Rad). The electrophoresis buffer contained 25 mM Tris-HCl, 192 mM glycine, pH 8. 3. Electrophoresis was carried out overnight at 200 V, with continuous elution of proteins (monitored at 280 nm). LTX was pooled and concentrated to 2-7 M using a Centriprep 30 unit (Amicon). Proteins obtained after each stage of purification were carefully tested by SDS-polyacrylamide gel electrophoresis and immunoblotting and assayed for toxicity, receptor binding, and transmitter release (25).
Channel Formation in Lipid Bilayers-Planar bilayers were formed from pure lipids suspended at a total concentration of 30 mg/ml in n-decane (Aldrich). Membranes containing POPE, equimolar POPE/ POPS, or an equimolar mixture of POPE, POPS, POPC, and cholesterol were cast across a 0.3-mm-diameter hole in a polystyrene partition separating two solution-filled chambers as described elsewhere (48,49). Unless otherwise stated, the bilayers were exposed to symmetric 100 mM KCl buffered with 10 mM Tris-HCl, pH 7.4. The membranes were voltage-clamped using a Biologic RK-300 patch clamp amplifier (Intracel, UK), and transmembrane currents were low pass-filtered at 1-10 kHz (Ϫ3-db point, 8-pole Bessel-type response) and recorded.
Experiments were only carried out with membranes that thinned spontaneously to give a capacitance of at least 250 picofarads and a conductance of Ͻ10 picosiemens; junction potentials were routinely nulled to within Ϯ 1 mV. The cis side of the bilayer (the side to which LTX was added) was voltage-clamped relative to the trans side, which was grounded, and we followed the standard electrophysiological current convention (i.e. positive currents flowing cis to trans are shown as positive, up-going currents). The contents of the chambers were changed by perfusion (Ն10 volumes) as required. Data were post-filtered and digitally sampled at 5 times the filter corner frequency for analysis using pClamp 6 (Axon) and other programs. Relative ion permeabilities were calculated from reversal potentials measured under bi-ionic conditions using appropriate forms of the Goldman-Hodgkin-Katz and Fatt-Ginsborg equations.
Recombinant Receptors-Rat latrophilin cDNA in pBlueScript (13) was cleaved with HindIII and EcoRI restrictases and subcloned into the same sites of the pcDNA3.1 vector (Invitrogen). To produce LPH-FS lacking the 5Ј-untranslated region, the fragment between the HindIII and BamHI sites of the initial construct was replaced with a shorter fragment, which was polymerase chain reaction (PCR)-amplified between primers CGTGCCCGCCCCAAGCTTTCGCCA (N77) and GGGTCCGCAGGTCATATTTG. Latrophilin deletion mutants LPH-7TM and LPH-5TM were prepared by PCR between N77 and oligonucleotides GCACTTGCTGTACTTCTAGAGCACCTTTTT or ATGAGCT-TCGGATCATCTAGAGCAGGGTC, respectively, using latrophilin cDNA as a template. These fragments were then digested with HindIII and XbaI and ligated into the respective sites of pcDNA3.1. For LPH-1TM, LPH-FS was cleaved with HpaI and NotI, and the deleted piece was replaced with a fragment obtained by PCR between primers GGAGAATGCCACAGTGAAGCTGGCAGGTGAG and AGGAAGCG-GCCGCTGCACAGTTACTTGTGGATG (cut with the same enzymes). To make the LPH-NRX hybrid, an intermediate construct was prepared by replacing the XhoI/XbaI segment of the LPH-FS with a PCR-amplified fragment between primers AAGGGAACTCGAGGAATTGCCTCG and GGCCTCTAGAGAAGCAGAAGGTGG. Another fragment, obtained by PCR (between CTTCTCTATGCTAGCTACAAGTACAG and TAAAAACAGTCTAGACTTCCTGATTGCA) on the NRX template and cut with NheI and XbaI, was then introduced into the XbaI site of the intermediate construct. The full-size neurexin construct was prepared by directly subcloning a KpnI-BglII fragment of the bovine cDNA for neurexin I␣ into pcDNA3.1. All mutations introduced were verified by sequencing. DNA for transfection of mammalian cells was prepared using Qiagen kits. Transfection of COS-7 cells maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) was carried out using the SuperFect transfection reagent (Qiagen) according to the manufacturer's protocol.
Antibody Preparation-Rabbit polyclonal Ab recognizing the N-terminal domain of LPH was obtained by injecting rabbits with bovine brain LPH purified by LTX affinity chromatography and SDS-electrophoresis (9). The C terminus-specific Ab was obtained by immunizing rabbits with a fusion protein consisting of glutathione S-transferase and the LPH C terminus (cDNA fragment 3976 -4856). LPH fusion proteins were constructed to affinity purify these Abs; the N-terminal domain (nucleotides 630 -2957) was subcloned into the pGEX-4X vector in-frame with glutathione S-transferase, whereas the C-terminal domain (nucleotides 3976 -4856) was cloned into the pQE40 vector (Qiagen) in-frame with dihidrofolate reductase. The fusion proteins were expressed in Escherichia coli, purified by glutathione or Ni 2ϩ affinity chromatography, and spot-blotted onto pieces of Immobilon membrane (Millipore). These were used to purify the LPH-specific Ab from respective sera as described (50). Anti-NRX Ab was elicited in rabbits injected with keyhole limpet hemocyanin conjugated with the NRX C-terminal peptide (CSANKNKKNKDKEYYV) and affinity-purified using this peptide immobilized on thiol-Sepharose (Amersham Pharmacia Biotech).
Immunostaining Procedures-COS-7 cells were grown on multi-well chamber slides (Nunc) and transfected as described above. One day after transfection, the cells were washed with PBS, fixed with 4% paraformaldehyde in PBS for 30 min at room temperature, and washed again. Nonspecific binding sites were blocked by a 20-min incubation with 3% (w/v) goat serum (Life Technologies, Inc.). In some experiments, the cells were permeabilized with 0.05% Triton X-100. The specimens were then incubated for 1 h with the respective primary Ab, washed, incubated with a fluorescein-conjugated anti-rabbit IgG (Sigma), washed, mounted using Vectashield mounting medium (Vector Laboratories), and analyzed by confocal laser-scanning microscopy. In some experiments, a fluorescent derivative of LTX was used, which was obtained using the FluoroLink mAb labeling kit (Amersham Pharmacia Biotech) according to the manufacturer's protocol. In immunocytochemical experiments, the distribution pattern of the fluorescently labeled LTX was identical to that of immunostained toxin.
For immunoblotting, COS-7 cells were detached from plates in icecold PBS and pelleted by centrifugation. The pellets were then dissolved in electrophoresis sample buffer containing 2% SDS, 100 mM dithiothreitol, and 8 M urea (10 5 cells/50 l of buffer) for 30 min at 37°C and separated in 6% SDS-polyacrylamide gels not containing urea. After electrophoresis, the proteins were transferred onto Immobilon membrane (Millipore) and visualized using primary Ab, goat anti-rabbit peroxidase-conjugated IgG, and chemiluminescent substrate (Pierce) followed by exposure onto x-ray film.
Confocal Microscopy-Specimens were imaged with a LSM 510 system (Zeiss) mounted on an upright microscope (Axioplan-2, Zeiss) using dual excitation at 488 and 633 nm. Light emitted by doubly stained preparations was analyzed using a combination of an FITC-type narrow band-pass filter block (505-530 nm) and a long-pass filter block (Ն650 nm). Images were collected using oil immersion objective (Plan-Neofluar, 40ϫ/1.3, Zeiss) for fixed samples or a water immersion objective (Achroplan, 40ϫ W/0.8, Zeiss) for submerged unfixed samples. To quantify the degree of co-localization, scatter diagrams for doubly stained samples were produced using the LSM 510 software (Zeiss). In such diagrams, the brightness of each pixel in the red channel (Cy5-LTX) is interpreted as X coordinate, and that in the green channel (immunostained NRX) is interpreted as Y coordinate; coordinates of pixels that have similar intensity in both channels (co-localization) appear near the diagonal running from bottom left to top right of the graph.
Binding Experiments-LTX was iodinated as described earlier (34). When binding experiments were performed on culture plates, the cells were incubated with 1 nM iodinated LTX in buffer A (145 mM NaCl, 5 mM KCl, 1.2 mM NaH 2 PO 4 , 2 mM MgCl 2 , 10 mM glucose, 20 mM HEPES, pH 7.4) in the presence or absence of 2 mM CaCl 2 , then washed with this buffer. Bound toxin was eluted from plates with 1% Triton X-100. Alternatively, cells were scraped off the plates into PBS, counted, then incubated with increasing concentrations of [ 125 I]LTX and filtered through GF/F glass fiber filters (Whatman). Nonspecific binding was determined in the presence of a 100-fold excess of unlabeled toxin. The radioactivity of the Triton eluate from plates and of the filters was measured in a ␥-spectrometer.
LTX-induced Influx into COS-7 Cells-For influx experiments, cells were plated at 50% confluence on 6-well tissue culture plates, transfected with control or receptor plasmids, and 24 h later, used in measurements of LTX-stimulated influx of 45  For 45 Ca 2ϩ influx, the cells were washed twice with buffer A and incubated at room temperature for 5 min in the same buffer supplemented with 2 mM CaCl 2 and 2-4 Ci/ml 45 Ca 2ϩ . One to five min after the addition of 1-2 nM LTX, the cells were quickly washed twice with buffer A. Alternatively, the toxin pores were pre-formed in cells before the addition of 45 Ca 2ϩ . For this purpose, the cells were incubated with 1 nM LTX, washed with 0.2 mM EGTA, and kept in buffer A without Ca 2ϩ for a further 10 -30 min at room temperature (anti-LTX serum or purified IgG fraction was included in some experiments at this stage); influx was initiated by the addition of Ca 2ϩ / 45 Ca 2ϩ . In all cases, after the final wash, the cells were lysed in 0.5 ml of 10 mM EDTA and 1% Triton X-100. The lysate was mixed with 10 ml of UltimaGold scintillation mixture (Packard), and its radioactivity was determined in a ␤-spectrometer.
[ 3 H]NE influx was measured in receptor-or vector-transfected cells either by simultaneously adding LTX and 0.5-2.5 Ci of [ 3 H]NE or by pre-forming LTX pores as above, with subsequent addition of the label. Influx was determined after 5 min by washing and lysing the cells.
To measure FITC influx, the cells were briefly washed with buffer A (plus/minus 2 mM Ca 2ϩ ) before the addition of 5 M FITC with or without 4 nM LTX. After a 10-min incubation, the cells were extensively washed with buffer A and immediately examined by confocal microscopy using a water immersion objective.

LTX Pores in Lipid Bilayers and Cells-Recently
, we introduced a new improved LTX purification procedure (25) that uses an additional step (preparative native electrophoresis) to remove several proteins usually contaminating conventional LTX preparations (26). The resulting highly purified LTX is fully active in biochemical and physiological tests (25). Here, we compared the ability of this pure LTX and the crude spider venom to form channels in artificial planar lipid bilayers (21) containing different lipids and in the presence or absence of Ca 2ϩ . The addition of small amounts of the crude venom (containing as little as 10 pM pure LTX) always gave rise to channel activity within 2 min (Fig. 1A). Larger doses of the venom (equivalent to 10 nM LTX) typically gave rise to macroscopic conductances of Ͼ500 nanosiemens. The channels were permeable to both Ca 2ϩ and K ϩ and had conductances ranging from 20 to 200 picosiemens, suggesting that the venom contains several different channel-forming components. Relative Ca 2ϩ /K ϩ selectivities (measured in bi-ionic conditions in palmitoyloleoylphosphatidylethanolamine (POPE) bilayers) ranged from 0.45 Ϯ 0.2 (mean Ϯ S.D., n ϭ 6) for 60-picosiemens channels to 1.6 Ϯ 0.7 (mean Ϯ S.D., n ϭ 6) for 150 -200picosiemens channels. Partially purified toxin (corresponding to conventional LTX preparations) was much less active than the crude venom, typically producing only 1-2 channels despite prolonged exposure (Fig. 1B).
In contrast, 10 nM pure LTX (i.e. 1000-fold more concentrated than in the venom) did not give rise to any channel activity after periods of 20 -60 min when added to more than 20 bilayers containing POPE and palmitoyloleoylphosphatidylserine (POPS), POPE alone, or a POPE/POPS/palmitoyloleoylphosphatidylcholine (POPC)/cholesterol mixture, irrespective of the presence of 2 mM Ca 2ϩ . The result was the same if membranes were broken and reformed in the continued presence of LTX. To illustrate these findings, Fig. 1C shows an experiment using POPE/POPS/POPC/cholesterol bilayer. 10 nM LTX had no effect over a prolonged period (note a different time scale), but there was near-immediate channel activity on stirring in the venom equivalent to 10 pM LTX (confirming that the bilayer remained intact and fusion-competent). These findings suggest that many pore-forming constituents of the venom are gradually removed from LTX by the purification procedure, resulting in a pure toxin that is unable to form pores in pure lipid membranes at physiological concentrations. In one experiment, we added a bolus of 2 M LTX to produce a final concentration of more than 100 nM (Fig. 1D). This did result in immediate channel activity, suggesting that pure LTX can induce membrane pores provided it is present at extremely high concentrations. Unfortunately, the need for these very large amounts of pure toxin precluded further, more detailed experiments on planar bilayers.
On the other hand, nanomolar LTX is known to induce pores readily in the membranes of neurons and endocrine cells or even non-secretory cells expressing the toxin receptors (e.g. see . To investigate this phenomenon, we applied LTX to COS-7 cells in the presence of extracellular 45 Ca 2ϩ . Consistent with its poor ability to permeabilize pure lipid bilayers, 2 nM LTX was unable to induce influx of Ca 2ϩ into vector-transfected cells (Fig. 1E). Only when the toxin concentration was elevated to more than 100 nM, did some Ca 2ϩ influx occur in control cells (Fig. 1E). However, 2 nM LTX from the same toxin batch induced a large accumulation of Ca 2ϩ (Fig. 1E) by cells that had been transfected with an LPH expression construct (13); this action required millimolar extracellular [Mg 2ϩ ] (12,25). These results indicated that the receptor was required for the induction of LTX pores by a mechanism to be determined.

Generation and Features of LTX Receptor Constructs-
To study the possibility that LPH-mediated signal transduction is involved in the LTX pore formation, we produced a series of LPH deletion mutants ( Fig. 2A). Heptahelical receptors require their cytoplasmic loops and the cytoplasmic tail in order to interact functionally with G proteins (30 -33), so these domains were progressively removed in the receptor mutants. As a result, functional coupling to G proteins was abolished 2 in the mutants LPH-N5TM, LPH-N3TM and LPH-N1TM, which contain five, three, and one transmembrane domain (TM), respectively. To independently test whether the receptor itself makes part of the pore, a structurally different LTX receptor, NRX I␣, was also expressed ( Fig. 2A). In addition, a LPH-NRX hybrid was obtained by fusing the cytoplasmic domain of NRX to the C terminus of LPH-N1TM.
When expressed in COS-7 cells, the recombinant LPH mutants had molecular masses consistent with the addition of approximately 15 kDa due to glycosylation (Fig. 2B). All these constructs could be purified from solubilized cells by LTX affinity chromatography (not shown), indicating that expression in non-neuronal cells and the mutagenesis did not abolish the affinities of the receptors for LTX.
Immunostaining of non-permeabilized cells transfected with the LPH mutants revealed that most of these proteins were transported to the plasma membrane (Fig. 3A). However, cells that abundantly expressed LPH-N1TM (which could be isolated from solubilized cells on an LTX column) did not display this mutant on their surface and did not bind LTX (see Fig. 4A), probably because the mutant protein lacked plasma membrane delivery signals. All the other LPH mutants were surfaceexpressed and bound LTX independently of Ca 2ϩ , although with different affinities. Thus, LPH-N7TM that lacks only the cytoplasmic tail was indistinguishable from the full-size LPH (LPH-FS, K d ϳ 4 nM), whereas other constructs containing fewer transmembrane domains (including the LPH-NRX hybrid, not shown) had lower affinities for the toxin and reduced numbers of receptors on the cell surface (Fig. 3B). COS-7 cells transfected with NRX also bound the toxin but in a strictly Ca 2ϩ -dependent manner (K d ϳ 7.5 nM) (Fig. 3B); in addition, fewer receptor sites were present on the surface of NRX-expressing cells than in the case of LPH-FS.
LTX-induced Pores Are Permeable to Ca 2ϩ and Small Molecules-As shown in Fig. 1E, LTX stimulates 45 Ca 2ϩ influx into LPH-expressing COS cells. Similarly, NRX-transfected cells also accumulated 45 Ca 2ϩ in the presence of LTX (Fig. 4A). Most importantly, even LPH mutants unable to activate G proteins (e.g. LPH-N3TM and LPH-NRX) stimulated pore formation. After a 5-min exposure to LTX (Fig. 4A), the total amount of 45 Ca 2ϩ uptake by the receptor-transfected cells correlated with the binding capability of the expressed receptor, i.e. the more toxin was bound to the cells, the more tracer was accumulated irrespective of receptor structure. This was also reflected in a slower initial rate of Ca 2ϩ uptake by the NRX cells compared with the LPH cells (Fig. 4B). This was not due to slower LTX association with NRX; in fact, toxin binding to NRX reached a maximum even faster but remained lower than in the case of LPH (Fig. 4C). Our experiments also directly showed that to induce the pore, LTX must physically bind to the cell surface receptor rather than just contact the cell. Indeed, when NRX cells were treated with 10 nM LTX in the absence of Ca 2ϩ , they did not accumulate 45 Ca 2ϩ added after toxin removal (Fig. 4A).
Previously, we demonstrated that the toxin pores in neuronal plasma membrane are permeable not only to cations but also to small molecules such as norepinephrine (NE), D-aspartate, and fluorescein isothiocyanate (FITC) (12). This approach was also used here to study the channels induced by the toxin in non-neuronal cells. When COS-7 cells, transfected with LPH, NRX, or vector DNA, were incubated with FITC without added toxin, they did not accumulate the dye (not shown). However, when FITC was added in the presence of LTX, receptor-expressing, but not control, cells became fluorescent (Fig.  5A). Upon washout, the dye quickly dissipated from the cytosol, suggesting that the toxin pore allowed FITC diffusion into and out of the cytoplasm. Importantly, the nuclei of these cells also accumulated and retained the dye (Fig. 5A). Diffusion of FITC from the cytoplasm into the nucleus (probably via the nuclear pores) strongly suggests that the dye was not simply present in vesicles taken up by endocytosis.
Neurotransmitters are known to leak out of the neuronal cytoplasm via the LTX pore (12,19); therefore, we tested the permeability of the toxin pore in non-secretory cells to [ 3 H]NE. COS-7 cells expressing the receptors took up this neurotransmitter but only on the addition of LTX (Fig. 5B); again, uptake correlated with the toxin binding (not shown). Overall, the results of the influx experiments suggest that (i) when exogenous receptors are expressed on the surface of cells not normally responsive to LTX, the toxin permeabilizes the membrane of such cells to cations and small molecules, (ii) the toxin pores are similar to those made by LTX in neurons containing native receptors, and (iii) receptor-mediated intracellular signaling is unlikely to be involved in the induction of the pore.
LTX Inserts into the Cell Membrane-To further characterize the interaction of LTX with the surface of receptor-expressing COS-7 cells, we used confocal immunofluorescent microscopy. Upon binding of LTX covalently labeled with a fluorophore (Cy5), the cells were fixed and then permeabilized to immunostain the C-terminal domains of LPH or NRX with appropriate fluorescently labeled Abs. We found that the toxin bound only to the cells expressing recombinant receptors and, as demonstrated in Fig. 6, A and B, remained on the cell surface where it co-localized with the receptors in a patchy pattern (arrows). In the LPH cells, toxin fluorescence coincided with the staining for LPH at all times (Fig. 6A). However, in the case of NRX, some of the toxin fluorescence appeared separate from the receptor staining (Fig. 6B, arrowhead), suggesting that, upon binding, LTX can remain on the cell surface even in the absence of direct contact with the receptor protein.
The nature of this toxin interaction with the cell surface was further studied in NRX-expressing cells because LTX-NRX interaction is strictly Ca 2ϩ -dependent (11,12), and the removal of this cation provides a convenient way to dissociate the toxin from the receptor. Furthermore, even solubilized NRX does not interact with the toxin in the absence of Ca 2ϩ (12), suggesting that not only surface-bound but also membrane-inserted LTX should lose its contact with NRX upon the addition of EGTA.

FIG. 1. LTX only forms pores in membranes containing LTX receptors.
A, examples of channels formed by crude venom. An aliquot of black widow spider venom containing 10 pM LTX was added to an equimolar POPE/POPS bilayer cast in 100 mM KCl voltage-clamped at Ϫ100 mV (cis minus trans). K ϩ currents flowing trans to cis appear as downwards deflexions. The closed (zero current) level is indicated by the upper horizontal line. Data were filtered at 100 Hz. B, example of channel formation by conventional (incompletely purified) LTX. Shown is a 15-s recording from a POPE bilayer in symmetric 100 mM KCl voltage-clamped at Ϫ100 mV and exposed to the ion exchange fraction (25, 34) (containing the equivalent of 1 nM pure LTX) added 2 min earlier to the cis chamber. Data were filtered at 50 Hz. C, effect of adding Ca 2ϩ and venom to a bilayer exposed to pure LTX. Continuous recording from a bilayer containing equimolar POPE/POPS/POPC/cholesterol in symmetric 100 mM KCl voltage-clamped at ϩ40 mV and exposed to 10 nM pure LTX (added by stirring in for 1 min before starting the recording). 2 mM CaCl 2 and venom (containing 10 pM LTX) were added as indicated. Note the rapid increase in bilayer currents (upward deflections indicate net positive currents flowing cis to trans) on adding the venom. D, channel recordings from a very high concentration of pure LTX. 15 s recording from a bilayer containing equimolar POPE/POPS/POPC/cholesterol in symmetric 250 mM KCl voltageclamped at Ϫ40 mV and exposed to Ͼ100 nM pure LTX, added as a bolus of 2 M LTX stock solution to the cis chamber. Channel openings give rise to downwards deflections, and the upper horizontal line indicates the main closed level. At least two channels appear to be present. E, COS-7 cells (10 6 cells/plate) were transfected with an empty expression vector or an LPH construct and incubated with 45 Ca 2ϩ in the presence or absence of 2 or 100 nM LTX. LTX-stimulated influx of Ca 2ϩ into the cells was measured as described under "Experimental Procedures." Values are the mean Ϯ S.E. At physiological concentrations, LTX stimulates Ca 2ϩ accumulation only in cells expressing LPH but not in control cells; however, very high toxin concentrations can permeabilize the receptor-deficient cells as well.
Here, we first allowed the fluorescent LTX derivative to bind to the NRX cells in a Ca 2ϩ -containing buffer and then washed the cells with EGTA. After incubation in the absence of Ca 2ϩ , the cells were fixed and immunostained for NRX. As predicted, the co-localization of the toxin and the receptor largely disappeared, although LTX remained on the cell surface (Fig. 6C,  arrowheads). Statistical analysis (Fig. 6D) demonstrates that, in the presence of Ca 2ϩ (left panel), there was no free toxin on the cells (absence of pixels along the red axis) as it all colocalized with NRX (spots concentrating along the diagonal). However, upon Ca 2ϩ removal (right panel), almost all cellbound toxin appeared separate from NRX (arrowheads). This could be due to lateral diffusion of membrane-inserted LTX.
To confirm that LTX did indeed insert into the membrane, we applied toxin to broken membranes from NRX-expressing cells at low or high temperatures in the presence of Ca 2ϩ . When the LTX-NRX interaction was subsequently disrupted by the addition of EGTA, the bound toxin dissociated from the cells much faster if the binding had been induced at 4°C (Fig. 7A); conversely, when the binding had been done at 28°C, a large proportion of LTX remained bound ( Fig. 7A; see also Fig. 6D). Notably, only the initial rate of dissociation was greatly different, whereas the later, very slow dissociation phase followed the same pattern in both cases. This suggests that the residual LTX binding at both temperatures had the same nature and was much stronger. Since endocytotic uptake by broken cell membranes was not possible, the very strong interaction of LTX with the cell surface was most likely due to membrane insertion of the toxin, which was facilitated by elevated temperatures. Likewise, temperature strongly affected the interaction of the toxin with rat brain nerve terminal membranes, where LTX displays a lower dissociation constant at 37°C (K d ϳ 1.5 nM) than at 0°C (K d Ͼ 6 nM) (Fig. 7B, left). At intermediate temperatures, both dissociation constants are regularly observed (Fig. 7B, right) (e.g. Refs. 34 and 35). Importantly, LTX shows similar two dissociation constants even when only one receptor (LPH) is expressed in the nerve terminals of NRX knockout mice (35), indicating that the complex binding pattern is not associated with the presence of two distinct receptor proteins. Taken together, these findings are consistent with the idea that LTX binding to receptors is followed by its insertion into the lipid bilayer, which is facilitated by permissive temperatures, although some incorporation can occur even at low temperature.
LTX Continues to Form Pores after Dissociating from Receptor-Our interesting observation that LTX stays inserted into the membrane but can dissociate physically from NRX suggested a method for testing whether this receptor was forming a part of the toxin-induced pore (Fig. 8A). We therefore studied the permeability of the cell membrane after LTX had been allowed to dissociate from NRX. As shown in Fig. 8B, Ca 2ϩ (added to NRX-expressing cells after pre-treatment with LTX and a 10 -30-min incubation in the absence of this cation) very quickly entered the cells through the LTX pore. However, the re-introduced Ca 2ϩ could potentially stimulate some re-association of the membrane-inserted LTX with NRX. To avoid this uncertainty, we looked at the influx of NE through the preformed toxin pores detached from NRX; this allowed us to monitor the pore in the continued absence of Ca 2ϩ . Under these conditions, NE still accumulated in NRX cells (Fig. 8C). Moreover, the pre-formed LTX pores were also permeable to FITC (Fig. 8D). The influx of both compounds was even stronger in the absence of Ca 2ϩ than in its presence. Thus, the persistence of the pore was most likely due to the ability of the toxin to form pores in the membrane even in the absence of its direct interaction with NRX.
The LTX Pore Can Be Blocked by Anti-LTX Antibodies-Our experiments (Figs. 4, 5, and 8) strongly implied that intracellular signaling was not involved in the pore induction but could not rule out the possibility that LTX, by directly interacting FIG. 2. LTX receptors and their mutants used in this work. A, the structures, putative membrane topographies, and nomenclature of the recombinant receptors. B, SDS-gel analysis of the receptors and mutants expressed in COS-7 cells. Cells were transfected with respective constructs and solubilized as described under "Experimental Procedures." Proteins were separated by electrophoresis (10 5 cells/lane) in 6% gels, transferred onto Immobilon membrane, and stained with the primary Ab (as indicated), peroxidase-conjugated secondary Ab, followed by chemiluminescent visualization. The upper bands represent dimers of some mutant proteins. Note that the C-terminal Ab recognizes LPH-FS whose proteolytic cleavage (51) is undetectable in this experiment.
with receptors or some endogenous cellular proteins, induced pore formation by these proteins themselves. To provide an unambiguous demonstration that LTX itself is involved in the pore structure, we applied anti-LTX Ab to NRX cells containing pre-formed LTX pores dissociated from the receptor. The immune serum inhibited Ca 2ϩ influx through such pores, whereas the pre-immune serum was ineffective (Fig. 8E). This result strongly suggests that LTX alone forms the pore. DISCUSSION Channel formation by LTX has been puzzling: it was unclear why this toxin forms pores in lipid bilayers but not in cells that do not express LTX receptors. Unfortunately, previous work did not compare the same toxin preparations on both lipid bilayers and receptor-deficient cells. Such comparison made in our work proves unequivocally that LTX alone is a very inefficient pore former; experiments with lipid bilayers and control COS-7 cells (Fig. 1) demonstrate that pure LTX at concentrations up to 10 nM is unable to form pores in membranes (artificial or biological) in the absence of LTX receptors. Moreover, the strong pore-forming activity of the venom gradually disappears with LTX purification. Therefore, previous observations of pores in artificial lipid membranes made at low nanomolar LTX concentrations (21-23) must be attributed to residual To demonstrate the synthesis of mutant proteins, transfected cells were permeabilized and stained with the LPH N-terminal Ab or NRX C-terminal Ab (left). Surface delivery of the expressed proteins was assessed using the same Abs on non-permeabilized cells (right). Note the lack of surface exposure of LPH-1TM despite its abundant production by the cells (NRX cannot be detected in non-permeabilized cells because the Ab only recognizes its cytoplasmic domain). B, Scatchard plot analysis of LTX binding to some receptors. Cells were dislodged from plates 24 h after transfection (2 ϫ 10 5 cells/point), and their binding of [ 125 I]LTX was measured as explained under "Experimental Procedures." Note that LPH-7TM has the same binding capacity and affinity as LPH-FS, whereas LTX binding to NRX, LPH-5TM, and LPH-N3TM is 2-, 4.5-, and 3-fold weaker, respectively. LPH-1TM lacking all but the first transmembrane domain does not bind LTX due to the lack of surface delivery. contamination by other venom components. Indeed, LTX used in bilayer experiments was routinely purified by one-step ionexchange chromatography (e.g. Refs. 22 and 23), although even the original paper describing toxin purification pointed out the incomplete homogeneity of LTX even after a two-step procedure (26). The large variety of channels obtained with the crude venom (Fig. 1A) and with different toxin preparations (e.g. Refs. 17,23,28,36,and 37) also suggests that other venom components either form channels themselves or help LTX to insert into the membrane. In fact, at concentrations above 100 nM, the pure toxin can insert into liposomes (24), planar bilayers (Fig. 1D), and biological membranes (Fig. 1E). However, receptors present on the cell surface facilitate LTX pore formation by more than 100-fold (Fig. 1E).
What is the role of the receptors in membrane pore formation by LTX? When the toxin pores are studied in native secretory cells (neuroblastoma (17), chromaffin cells (6), gonadotropes (38)) or in secretory cells transfected with exogenous receptors (39,40), one cannot totally exclude the involvement of intracellular signaling in pore induction because such cells possess endogenous NRX and LPH as well as receptor-linked pathways and non-receptor exocytotic proteins. Similarly, injection of non-secretory cells (Xenopus oocytes) with heterogeneous neuronal mRNA cannot demonstrate which receptor(s) or other neuronal proteins are involved in LTX pore formation (28). A more analytical approach was used in recent work (37,41) based on electrophysiological characterization of LTX pores induced in non-secretory (HEK293 and BHK, respectively) cells transfected only with LPH, NRX, or mutant receptors. The results demonstrated that LTX makes pores on binding to any of its receptors.
In our present work too, introduction of LTX-binding proteins, including various receptor mutants, into non-secretory cells rendered them susceptible to the pore-forming activity of the toxin (Figs. 1, 4, 5, and 8). COS-7 cells are not capable of regulated exocytosis and lack most of the components of the exocytotic machinery present in endocrine or neuronal cells. FIG. 6. Co-localization of LTX with receptors in cell membrane. A and B, LPH-and NRX-expressing cells were incubated with Cy5-labeled LTX in buffer A containing 2 mM CaCl 2 for 10 min, washed, fixed, and stained with respective anti-receptor Ab, and visualized by confocal fluorescent microscopy. Scale bars, 20 m. Note the complete co-localization of LTX with the LPH C-terminal staining (arrows) and partial co-localization of LTX and NRX (free toxin is shown by an arrowhead). C, LTX remains membrane-bound after dissociating from receptor in NRX-expressing cells. The NRX cells were first incubated for 10 min with LTX in the presence of 2 mM Ca 2ϩ , washed with 0.2 mM EGTA, incubated in an EGTA-containing buffer for a further 10 -30 min, then fixed and immunostained with anti-NRX Ab as above. D, scatter diagrams (see "Experimental Procedures") showing analysis of co-localization of Cy5-labeled LTX with immunostained NRX after incubation of cells in the continued presence of Ca 2ϩ (left, n ϭ 6) or after subsequent removal of cation (right, n ϭ 9). Spots along the diagonal line correspond to LTX co-localizing with NRX; spots along the x axis (arrowheads) correspond to free toxin.
Thus, specific neuronal proteins cannot be involved in the formation of the toxin pore in these cells. In addition, pore formation depends on the amount of bound LTX rather than on the type of receptor expressed (Fig. 4). Moreover, even receptor mutants that cannot mediate signaling (LPH-5TM, LPH-3TM, and LPH-NRX) allow the toxin to form pores, indicating that signal transduction is not involved in the induction of the pore.
Do LPH or NRX contribute to the structure of the pore or just recruit the toxin to the membrane? LPH and NRX have very different membrane domain structures; therefore, if the receptors indeed formed the pore together with LTX, such pores would have had different characteristics. However, the pores induced by LTX in HEK293 or BHK cells transfected with LPH, NRX, or their mutants have very similar electrophysiological parameters (37,41). In our experiments too, the features of the toxin pores do not depend on the receptor structure (Figs. 4,5,and 8). Furthermore, at least in the case of NRX, the toxin can dissociate from the receptor and remain in the membrane (Figs. 6 and 7), continuing to form pores with the same macro characteristics (Fig. 8). Although such dissociation apparently cannot be achieved for LPH, this receptor is also unlikely to be physically involved in the pore because its mutants with different numbers of transmembrane domains support the induction of similar LTX pores ( Fig. 4 and data not shown).
Thus, our results suggest that the toxin itself forms pores, but how does this hydrophilic toxin insert into the membrane? Using cryo-electron microscopy, we recently visualized the toxin pores in the membrane of liposomes and revealed the likely mechanism of pore formation (24). LTX in solution exists as a homodimer (24,25), but when its local concentration increases (e.g. upon receptor binding) and, especially in the presence of divalent cations (25), the dimers assemble into tetramers; this rearrangement renders one side of the tetramer hydrophobic. The tetramers can then incorporate into a lipid bilayer. In the center, the tetramer has a channel of 10 -25 Å in diameter that spans the membrane. Strikingly, a very similar diameter (ϳ19 Å) was estimated for LTX water pores in lipid bilayers (23). Our tetramer/pore model (24,25) is consistent with the permeability of the toxin pore to different cations, amino acids, NE, and FITC (largest dimension, 12 Å) but not to small dextrans or cytoplasmic proteins (12,14). Based on this model, the receptors may act both by increasing the nearmembrane toxin concentration (thus, aiding LTX tetramerization) and by orientating the tetramers with their hydrophobic side toward the membrane. Receptors serving as specific anchors for inefficient pore-forming toxins can also explain the phylum specificity of the members of latrotoxin family, which have very similar mechanisms of action but only interact with receptors in their target animals (26,42).
Can pore formation by LTX explain all of its dramatic effects on a secretory cell? We showed that most of LTX-induced secretion correlates directly with formation of tetramers and pores (25). Massive influx of Ca 2ϩ and other cations through these pores can stimulate exocytosis directly (6,17,40,43) or indirectly by mobilizing intracellular Ca 2ϩ (12,17,38,44); in addition, leakage of cytosolic constituents (12,19,20) can further destabilize and stimulate the cell. But do the receptors mediate at least some of LTX action? To answer this question properly, one should use a LTX derivative that is unable to form pores; otherwise, the robust effects of the membrane pore always mask the subtle receptor-mediated mechanisms. Indeed, even signaling-deficient receptor mutants, by binding LTX, support toxin pore formation and Ca 2ϩ influx. Ironically, an increased sensitivity of endocrine cells to LTX (in the presence of Ca 2ϩ ) upon super-transfection with receptor mutants was used to suggest that the receptors do not transmit exocytotic signals (40,45,46). However, neither these results nor the data presented in this paper exclude the possibility that receptor-mediated signaling may be involved in some aspects of secretion stimulated by LTX, especially at low extracellular Ca 2ϩ , as has been suggested by many authors (12,14,39,44,47). Thus, although the pore formation is very important for the action of the toxin, the natural role of LTX receptors remains uncertain. Future work will be needed to address the functions of both LTX receptors.