Regulated secretion: SNARE density, vesicle fusion and calcium dependence

SNAREs such as VAMP, SNAP-25 and syntaxin are essential for intracellular trafficking, but what are their exact molecular roles and how are their interactions with other proteins manifest? Capitalizing on the differential sensitivity of SNAREs to exogenous proteases, we quantified the selective removal of identified SNAREs from native secretory vesicles without loss of fusion competence. Using previously established fusion assays and a high sensitivity immunoblotting protocol, we analyzed the relationship between these SNARE proteins and Ca2+-triggered membrane fusion. Neither the extent of fusion nor the number of intermembrane fusion complexes per vesicle were correlated with the measured density of identified egg cortical vesicle (CV) SNAREs. Without syntaxin, CVs remained fusion competent. Surprisingly, for one (but not another) protease the Ca2+ dependence of fusion was correlated with CV SNARE density, suggesting a native protein complex that associates with SNAREs, the architecture of which ensures high Ca2+ sensitivity. As SNAREs may function during CV docking in vivo, and as further proteolysis after SNARE removal eventually ablates fusion, we hypothesize that the triggered steps of regulated fusion (Ca2+ sensitivity and the catalysis and execution of fusion) require additional proteins that function downstream of SNAREs.

Ca 2+ -dependent assembly of SNARE complexes (Xu et al., 1999;Scales et al., 2001). However, the exact function of SNAREs is difficult to establish because the regulated exocytotic pathway is interconnected and cyclic: interfering with any step eventually blocks secretion. By studying the massively synchronous cortical vesicle (CV) exocytosis of fertilization, which occurs but once, the cycle is bypassed and the final Ca 2+ -dependent fusion steps are isolated. The exocytosis of CV in vitro occurs with physiological [Ca 2+ ]free (Baker and Whitaker, 1978).
To differentiate between the density of CV SNAREs, changes in the Ca 2+ sensitivity of CV fusion, and the rate and extent of CV fusion, advantage was taken of the size, purity homogeneity and high preparative yields that make CV useful for coupled functional-biochemical analyses Tahara et al., 1998;Zimmerberg et al., 2000). Endogenously docked CVs from eggs of Strongylocentrotus purpuratus are fully Ca 2+ sensitive and release-ready, and they fuse with the plasma membrane (PM) within milliseconds of exposure to optimal [Ca 2+ ]free (exocytosis in vitro) (Shafi et al., 1994). Ca 2+ -triggered homotypic CV fusion retains the essential features of regulated exocytosis Zimmerberg et al., 2000); only by removing CVs from the PM is there full access to all vesicle surface proteins. Our approach uses proteases with differing spectra of endogenous substrates on intact CVs, to quantitatively change the density 2087 SNAREs such as VAMP, SNAP-25 and syntaxin are essential for intracellular trafficking, but what are their exact molecular roles and how are their interactions with other proteins manifest? Capitalizing on the differential sensitivity of SNAREs to exogenous proteases, we quantified the selective removal of identified SNAREs from native secretory vesicles without loss of fusion competence. Using previously established fusion assays and a high sensitivity immunoblotting protocol, we analyzed the relationship between these SNARE proteins and Ca 2+triggered membrane fusion. Neither the extent of fusion nor the number of intermembrane fusion complexes per vesicle were correlated with the measured density of identified egg cortical vesicle (CV) SNAREs. Without syntaxin, CVs remained fusion competent. Surprisingly, for one (but not another) protease the Ca 2+ dependence of fusion was correlated with CV SNARE density, suggesting a native protein complex that associates with SNAREs, the architecture of which ensures high Ca 2+ sensitivity. As SNAREs may function during CV docking in vivo, and as further proteolysis after SNARE removal eventually ablates fusion, we hypothesize that the triggered steps of regulated fusion (Ca 2+ sensitivity and the catalysis and execution of fusion) require additional proteins that function downstream of SNAREs.
of different sets of proteins; the [Ca 2+ ]free-dependent rate and extent of CV fusion is then tested, and the density of SNARE proteins measured. This approach (1) makes no assumptions as to which proteins are essential, (2) allows for the involvement of low-abundance proteins, (3) uses native vesicles of endogenous size and composition, and (4) is not subject to issues of compensatory proteins as are genetic knockout studies. This approach does require quantitative assays of absolute protein amount that are antibody dependent (Coorssen et al., 2002).
On average, isolated CV have ~5500, ~700 and ~330 copies of VAMP, SNAP-25 and syntaxin, respectively (Coorssen et al., 2002). We combined (1) a sensitive, quantitative immunoblotting assay and polyclonal antibodies (to minimize underestimation due to epitope loss) to measure SNARE proteins (Coorssen et al., 2002); (2) fusion assays to assess the effects of SNARE removal; and (3) a method for determining the average number of active fusion complexes per vesicle, n -, that combines the relationship between the extent of fusion and n -, with the exponential decrease in nafter protease treatment (Vogel et al., 1996;Coorssen et al., 1998). This analysis was important because there is redundancy in the number of fusion complexes on CV; for any one vesicle, fusion will not be inhibited until all functional fusion complexes on that vesicle are inactivated, although the rate of fusion will be progressively inhibited by reductions in the density of critical components (Vogel et al., 1996;Blank et al., 2001).
The determination of proteolytic fragmentation patterns for syntaxin and SNAP-25 (Fig. 1) provided a guide for cutting the blotting membrane into three pieces that could be probed independently for SNAREs. In most experiments, using the molecular weight markers as a guide, the membrane was cut between syntaxin and SNAP-25 and just below the smallest SNAP-25 fragment. Myoglobin, recognized by the SNAP-25 antibody, was added only when blotting the entire membrane for syntaxin and SNAP-25 and was not included when processing blotting membranes for the presence of VAMP.
For kinetic assays, CVs were centrifuged onto coverslips and transferred to a perfusion chamber, permitting constant monitoring both by time-resolved light scattering and by direct visual inspection (Blank et al., 1998a;Blank et al., 1998b;Coorssen et al., 1998). A flow rate of ~1 ml/s was used to wash CV lawns and exchange buffers. All experiments used B-IM containing [Ca 2+ ]free of ~100 nM, 150 µM or 1 mM. Light-scattering signals were normalized so that [Ca 2+ ]free of ~100 nM and 1 mM gave 0% and 100% fusion, respectively. In all cases, aliquots of the same CV suspensions were used to determine and verify Ca 2+ activity.
The analysis of fusion is based on the number of active fusion complexes/CV, a Poisson distributed random variable in which the average number of committed fusion complexes, n -, varies between 0 and ~9 depending on [Ca 2+ ]free (Vogel et al., 1996;Coorssen et al., 1998;Blank et al., 2001). Although 1 mM [Ca 2+ ]free is routinely used to saturate n -Max and establish the sigmoid plateau, ~100% fusion occurs at much lower [Ca 2+ ]free; cumulative log-normal sigmoid curves can be characterized by the distribution midpoint and width and/or the midpoints and corners. Thus, for comparing activity curves, both the width and the [Ca 2+ ]free associated with 5% and 50% fusion are used. The relationships between extent of fusion (%F), n -, and trypsin treatment are %F=100×(1-exp(-n -)) and n -=n -Max× exp(-[trypsin]/τ), respectively, where τ is the concentration decay constant (Vogel et al., 1996). Changes in %F as a function of the trypsin concentration were fit using the equation %F=100× (1-exp(-n -Max×exp(-[trypsin]/τ))). For kinetic studies, low solution flow was used to minimize disruption of CV on coverslips. Under Journal of Cell Science 116 (10) Fig. 1. Full-length western blot probed for syntaxin and SNAP-25. 4 hour control, clostripain-and papain-treated samples (5, 15 and 20 µl load) were evaluated. The normalized fluorescence scans represent the load-adjusted intensity as a function of the normalized western blot coordinate (see Coorssen et al., 2002). these conditions, the transition from active to committed state is not resolved; the kinetic data were fit using the original two-state model and a temporal offset (Vogel et al., 1996;Blank et al., 2001). The initial rate of fusion can be approximated by n -×p (Vogel et al., 1996;Blank et al., 2001). With p=κ×n - (Blank et al., 2001), where κ represents a characteristic efficiency constant, the initial rate can be expressed as κ×n -2 . This function suggests that κ is proportional to a rate constant, whereas nis analogous to a concentration term. With κ proportional to a fusion rate constant, the change in energy resulting from papain treatment can be calculated using the relationship κcontrol/κpapain=exp(-(Econtrol-Epapain)/kT).
The exponential correlation between SNARE density and the midpoint of the calcium activity curve, Midpoint=A× exp(-Density/B)+C, was evaluated with weighting using 1/(s.e.) 2 of the midpoint estimates. The error in SNARE density was ignored in the fitting after determining that the error range in the parameter estimates (95% confidence) was greater than changes in the estimates arising from fitting the midpoint vs. the density values shifted by ± the error in the density. Overall, the error in the density divided by the density was 19%, 16% and 5% for VAMP, SNAP-25 and syntaxin, respectively. We could determine syntaxin density to better than 1 molecule/CV † .

Results
The Ca 2+ activity and kinetics of CV fusion are described by an analysis developed for CV-PM fusion (Vogel et al., 1996;Coorssen et al., 1998;Blank et al., 2001). CV-CV fusion is characterized by sigmoidal Ca 2+ activity curves; midpoints and widths of the Ca 2+ threshold distributions were determined ( Fig. 2A; Table 1) Blank et al., 1998a). The Ca 2+ activity curves for CV-CV fusion are minimally shifted from those for CV-PM fusion , which predict [Ca 2+ ]free that are comparable to those measured during fertilization (Vogel et al., 1996); 5%/50% fusion occurs at a [Ca 2+ ]free of 3/6 µM. One hour in buffer has little effect on CV fusion, with 5%/50% fusion occurring at [Ca 2+ ]free of 3/7 µM ( Fig. 2A; Table 1); after 4 hours in buffer, there is a significant shift in the midpoint of the Ca 2+ activity curve and 5%/50% fusion occurs at a [Ca 2+ ]free of 36/76 µM. Thus, results of protease treatments were compared to parallel, untreated controls.
Trypsin had a systematic and progressive affect on Ca 2+ activity (sensitivity and extent) ( Fig. 2A; Table 1). 700 units/ml trypsin (1 hour) shifted Ca 2+ activity rightward, but with no loss in maximal extent of fusion elicited by a saturating [Ca 2+ ]free; the 5%/50% fusion level occurs at 3/17 µM. 3500 units/ml trypsin (1 hour) caused a further rightward shift in Ca 2+ activity and ~30% loss of fusion. Prolonged incubations or higher trypsin doses further digested certain CV proteins, correlating with losses in Ca 2+ activity and progressively greater inhibition of fusion; 700 units/ml (3 hours) resulted in ~70% inhibition (data not shown). In comparison, 35,000 units/ml trypsin (1 hour) rendered CV fusion incompetent (99±1% inhibition at saturating [Ca 2+ ]free). Like N-ethylmaleimide (NEM), trypsin inhibited fusion in a manner consistent with a single, crucial site of action and the number of these sites, n -Max=8.4±1.0, agrees with previous determinations (Vogel et al., 1996;Coorssen et al., 1998). By contrast, papain (2000 units/ml, 1 and 4 hours) shifted and broadened Ca 2+ activity without altering the maximal extent of fusion ( Fig. 2A,B; Table  1); 5%/50% fusion occurs at 3/45 and 68/197 µM [Ca 2+ ]free, respectively, following 1 and 4 hour treatments. Clostripain (100 units/ml, 1 and 4 hours) had minimal effect on Ca 2+ activity ( Fig. 2A,B; Table 1, 5%/50% fusion occurs at 2/11 and 40/69 µM [Ca 2+ ]free, respectively, following 1 and 4 hour treatments), but distinct effects on the CV protein profile (Fig.  2C). When added hourly, fresh papain (but not clostripain) gradually inhibited fusion (~20-30% fusion at saturating [Ca 2+ ]free after 3 hours treatment, data not shown). In all cases, comparable results were obtained after simply settling proteasetreated CV into contact, confirming that the measured fusion is dependent on proteins rather than on centrifugation . Similar results were also obtained for exocytosis in vitro after 1 hour treatment of CV-PM preparations ( Fig. 2D; Table 2). Trypsin does not appear to affect docking per se, as CV undocking was not observed. The effects of papain (3000 units/ml) were less dramatic, suggesting limited access to the CV-PM contact site. Thus, each of these proteases had a different effect on CV fusion, enabling us to distinguish between changes in the Ca 2+ sensitivity or extent of fusion.
Assuming that the relative error for each of these terms is 5% (0.05), an extremely conservative value for all terms except the extraction efficiency, then the total error can be approximated by σCopy #/CV=Copy #/CV×0.1.
Using a weighted average of the results obtained at the two different protein loads (30 and 15 µl of sample), the average copy numbers are 0.107±0.008 and 0.404±0.032. If the calculated error is representative of the underlying error then the Copy #/CV is significantly less than one copy at the 99% confidence level (3×σCopy #/CV). If the calculated error represents an estimate of the unknown underlying distribution then the appropriate t value with one degree of freedom (two samples were averaged, n-1 degrees of freedom) must be used. In this case, the t value for one degree of freedom at the 95% confidence interval is 12.71 and the error would be 0.107±0.102 and 0.404±0.407; these values also indicate significantly less than one copy at the 95% confidence level. Therefore, within >95% probability, these two experiments show that our determination of less than one copy of syntaxin per CV is statistically significant. SNARE proteins, we used the fact that known SNAREs have substantial numbers of potential trypsin, papain and clostripain cleavage sites throughout their cytoplasmic domains (Table 3). The initial number of accessible proteolytic cleavage sites is expected to be lower due to structural constraints (Hubbard et al., 1998;Hubbard, 1998). Using the program Nickpred , examination of SNAREs in the Brookhaven Protein Data Base predicted multiple limited proteolytic sites. Subsequent loss of structure is expected to expose additional sites to further proteolysis. Indeed, ladders of multiple proteolytic fragments migrating below the main VAMP and SNAP-25 protein bands were detected with lower protease concentrations or shorter treatment time (1 hour). Following 4 hour protease treatments, syntaxin appears to be either intact or fully proteolyzed, as no fragments were detected (Fig. 1). This observation is consistent with the complete loss of syntaxin complex immunoreactivity observed following trypsin proteolysis (Lawrence and Dolly, 2002). SNAP-25 Journal of Cell Science 116 (10)  immunoreactivity, which may correspond to fragments, was detected in both control and treated samples (Fig. 1); VAMP immunoreactivity often appeared as a doublet (data not shown). The amounts of these SNARE fragments decreased with increased protease concentration or treatment time and were included in the quantitative determinations of SNARE density. Thus, unlike Clostridial toxins, these proteases could destroy all known SNAREs (including toxin-insensitive homologs). Trypsin (700 units/ml, 1 hour) digested substantial amounts of VAMP, SNAP-25 and syntaxin but had no effect on the extent of fusion (Fig. 3). Increased trypsin (3500 units/ml) removed >90% of VAMP and syntaxin; ~60% SNAP-25 remained (Fig. 4). Higher trypsin doses (Fig. 4) or longer incubation times (700 units/ml, 3 hours) did not reduce SNAREs to levels lower than detected after 1 hour with 3500 units/ml, but did block fusion. Papain (2000 units/ml) and clostripain (100 units/ml) removed most SNAREs after 1 hour; there was little added affect of 4 hour incubations. Papain reduced VAMP, SNAP-25 and syntaxin by 94%, 91% and >99%, respectively, and clostripain by 92%, 87% and 94%; CV remained fully Ca 2+ sensitive and 100% fusion competent (Fig.  3). A post-protease wash with chaotropic buffer resulted in some further loss of SNARE fragments nonspecifically bound to CV, but there was no change in fusion at saturating [Ca 2+ ]free, indicating that free SNARE fragments were not promoting fusion; although ineffective, these fragments were still included in our density assessments to ensure exhaustive SNARE quantification. Intact SNAREs are not removed by the chaotropic buffer wash (data not shown). Post-protease fusion was not due to nonphysiological Ca 2+ -induced membrane destabilization, or fusion would remain high after extensive trypsinization; this was not the case (Figs 2, 3 and Fig 4A; 35,000 units/ml trypsin). Thus, despite SNARE removal, fusion activity remained. Furthermore, when added hourly, fresh papain, but not clostripain, gradually inhibited fusion despite comparable reductions in SNARE densities. Fusion appears to require proteins other than the SNAREs.
Reducing syntaxin to <1 copy/CV ('biochemical knockout'; 0.1 or 0.4 copy/CV at the 95% confidence level for each of two samples with a range from 0-0.8 † ) by the most potently SNARE-destructive treatment (papain; 2000 units/ml replaced every hour for 3 hours) (Fig. 5) decreased but did not abolish the extent of fusion elicited by saturating [Ca 2+ ]free (two trials, final extent 20-30%). Eight more trials with 2000 units/ml papain gave 100±1% fusion with only 1-<3 syntaxin copies/CV in 4 trials and >3 copies/CV in 4 trials; 50% fusion occurred at 45 and ~200 µM [Ca 2+ ]free after 1 and 4 hour treatments, respectively. Overall, the reduced syntaxin density as a function of papain treatment ranged from <1-80 copies/CV depending upon the papain concentration (300-5000 units/ml) and treatment time (1-4 hours). For comparison, the reduced syntaxin density as a function of clostripain treatment ranged  from 5 to 50 copies/CV. If an individual SNARE complex catalyzes fusion, and proteolysis is random, then Poisson analysis (Vogel et al., 1996;Coorssen et al., 1998;Blank et al., 2001) can be used to predict the extent of fusion at saturating [Ca 2+ ]free. The extents of fusion predicted by the measured syntaxin densities following papain treatment are not consistent with the observed fusion (χ 2 , P<0.05); neither syntaxin nor syntaxin-containing complexes have the properties of a Poisson-distributed fusogen. Three SNARE complexes per vesicle fusion event have been suggested (Hua and Scheller, 2001), and it is thus significant that the maximum extent of fusion (100%) elicited by saturating [Ca 2+ ]free is independent of syntaxin density from 1 to 330 copies/CV (papain treated relative to control).
Shifts in Ca 2+ activity preserving fusion competence are observed in both control (4 hours) and papain-treated samples, suggesting disruption of a Ca 2+ regulatory system mediating fusion. As clostripain-treated samples did not have large shifts in Ca 2+ activity relative to control and the Ca 2+ activity midpoint was not correlated with SNARE density (slope not significantly different from zero), it is somewhat surprising that, for papain-treated samples, the midpoints of the Ca 2+ activity curve were correlated with SNARE density (Fig. 6). For all three SNAREs, the correlation between [Ca 2+ ]free (EC50) and SNARE density was described by the exponential relationship, midpoint=A×exp(-density/B)+C, where A+C is the midpoint (EC50, µM) of the Ca 2+ activity curve in the absence of SNAREs and B is the density decay constant. The extrapolated midpoint at high SNARE density (parameter C) varied from 13 to 20 µM, depending on the SNARE, and is consistent with control midpoints (6-20 µM for 0, 1 and 2 hour incubations). When low-density values are normalized to the Journal of Cell Science 116 (10) Fig. 5. Proteolytic ablation of syntaxin: average density <1 copy/CV. Using a calibration curve derived from serial dilution of purified glutathione S-transferase (GST)-syntaxin (Coorssen et al., 2001) 0.73 fmoles, corresponding to 113 molecules/CV, were detected in 0 hour controls. No syntaxin was detected in papain-treated (2000 units/ml, replaced every hour for 3 hours) samples. The limit of detection for this calibration was <0.039 fmoles or 0.3 Treated Equivalent (T.E.) molecules/CV for the same protein load (CV number) used to visualize the papain-treated sample. The maximal syntaxin density in this sample is 0.3 molecules/CV. The plot of normalized fluorescence vs. Western Blot Coordinate shows how the signal associated with 0.039 fmoles is easily quantified. The curves are drawn with an offset for visual comparison.  6. Ca 2+ activity and SNARE density. The midpoint of the Ca 2+ activity curve following papain treatment at either 1 hour (blue) or 4 hours (red) is correlated with SNARE density (Copy #/CV). The data were fit using the relationship A×exp(-density/B)+C, where A+C (µM) is the midpoint of the Ca 2+ activity curve in the absence of SNAREs (248±10, 433±66 and 305±22) and B is the density decay constant (212±63, 42±10 and 11±3) for VAMP, SNAP-25 and syntaxin, respectively. When SNARE density was normalized by the density decay constant, the pooled transformed data were described by the relationship A×exp(-normalized density/B)+7.1; the midpoint of the Ca 2+ activity curve in the absence of SNAREs (A+7.1) is 260±22 µM. density at 1 hour and the high-density midpoint constrained to the value of the 1 hour control (7.1 µM), all three SNAREs were described by a common exponential decay with a lowdensity midpoint of ~250 µM [Ca 2+ ]free. Thus, papain treatment reveals an intrinsic Ca 2+ -dependent process with an activity midpoint that can be modulated by proteolysis betweeñ 6 and ~250 µM [Ca 2+ ]free. If this reflects the inherent Ca 2+ sensitivity of the native regulated fusion complex, then the EC50 of this endogenous fusion complex is ~250 µM [Ca 2+ ]free.
Decreases in the density of components crucial to triggered fusion are predicted to inhibit the rate of fusion (Vogel et al., 1996;Blank et al., 2001). As clostripain (100 units/ml, 4 hours) failed to affect Ca 2+ activity (Fig. 2B) and had no effect on fusion kinetics (Fig. 7A), but did digest SNARES, it is unlikely that the identified SNARES in this system are crucial to fusion. Papain (2000 units/ml, 4 hours) reduced the kinetics and extent triggered by 150 µM Ca 2+ but not the final extent of fusion triggered by saturating Ca 2+ , indicating that both the probability per unit time (p) and ndecreased (Fig. 7A). However, a limited papain treatment (2000 units/ml, 3 hours) altered the kinetics triggered by 150 µM Ca 2+ with minimal changes in the extent (Fig. 7B). Analysis (see Materials and Methods) ruled out a direct reduction in n -, indicating that papain reduced P, the probability that any given vesicle will fuse (Vogel et al., 1996;Blank et al., 2001). After 3 hours, p/n -(a measure of efficiency) in papain-treated CV is only ~17% of control, but is unaffected by clostripain (4 hours) (Fig. 7A). The changes in fusion rate after 3 hours papain treatment are consistent with an increased energy barrier for fusion. Treating p/nas a rate constant (see Materials and Methods), the energy contributed by papain-sensitive proteins is estimated to be ~2 kT (0.6 kcal/mol) per fusion complex. This is about half that required to overcome the hydration layer at membrane surfaces (Leiken et al., 1993) or to mediate molecular rearrangements thought necessary for bilayer merger (Kuzmin et al., 2001). Thus, papain (but not clostripain or low doses of trypsin) removes or modifies crucial modulatory proteins that have pronounced effects on the efficiency of triggered fusion.

Discussion
The ability to determine with precision the number of copies of syntaxin/CV allowed a quantitative evaluation of SNARE hypotheses, as the SNARE core complex requires SNARE motif helices from syntaxin that cannot be substituted by other SNAREs (Scales et al., 2000a). Ca 2+ -and proteindependent fusion of native secretory vesicles proceeded after substantial removal of identified SNAREs, even after syntaxin was stoichiometrically eliminated (biochemical 'knockout'). It should be noted that the effective CV SNARE densities remaining after proteolysis may be even lower than the values we report, as we included all detectable SNARE fragments in the overall quantification of our high sensitivity immunoblots. If ineffective SNARE fragments are present (as indicated by the use of chaotropic washes), they have, for the sake of thoroughness, nevertheless been included in the quantitative evaluations of density. Thus, although the identified SNAREs do not effect fusion, the Ca 2+ sensitivity data suggest that they may interact with modulators of Ca 2+ efficiency that directly influence the capacity for fusion in this Ca 2+ -triggered system.

Lack of a direct role for SNAREs in membrane fusion: alternate interpretations
Our conclusion, that SNAREs are neither driving CV membranes together nor inextricably linked to Ca 2+ -regulated fusion via an absolute requirement for the presence of their cytosolic domains, is dependent on a correct interpretation of SNARE proteolysis, antibody interactions and the fusion process. Two classes of alternative hypothesis that preserve a direct role for SNARE proteins can be considered. The first class is that the 'wrong' SNAREs were evaluated in this study: the necessary SNAREs were still present on protease-treated CV at a sufficient density to support fusion. This class can be subdivided into SNARE homologs and SNARE fragments that do not cross-react with the antibodies used. The second class is that multiple fusion pathways exist, and include SNAREdependent and -independent mechanisms; the latter can be subdivided into protein-mediated and protein-independent mechanisms.

Do we have the right SNAREs?
The interpretation of genetic studies of SNARE activity can be influenced by the presence of gene homologs (e.g. Vilinsky et al., 2002). By contrast, the biochemical knockout approach relies on the structural conservation of putative homologs. The conservation of the different potential protease cleavage sites in all known SNARE homologs is very high (Table 3), particularly for the trypsin, papain and clostripain cleavage sites closest to the PM (i.e. cleavage at this site would result in the complete release of the cytoplasmic domain) and in the potential coiled-coil domains postulated to be essential to SNARE-mediated fusion. Thus, although we do not know all the epitopes recognized by the antibodies employed, complete destruction of other SNARE homologs should also occur under the conditions of our experiments. If this interpretation is incorrect, then a putative protease resistant 'SNARE' would either have a sequence that is significantly different from all known SNAREs, such that not even one functional proteolytic site exists (a feature that would question the nature of SNARE homology), or exist in a stable conformation such that its cleavage sites were never accessible to protease action during the extended incubation times used in our experiments.
Antibodies raised specifically against nonconserved regions of the protein can certainly identify different homologs of SNAREs. In this study, polyclonal antibodies raised against the whole cytosolic region (e.g. including the extended highly conserved domains) were used. Because SNAREs maintain highly conserved regions it is more likely that our antibodies would cross-react with any protein containing these conserved regions than would monoclonal antibodies. We have used antibodies against SNAREs from other species in the past Coorssen et al., 2002;Tahara et al., 1998), and have always identified only these SNAREs on the CV membrane. To accept the alternate hypothesis, that SNAREs drive the fusion step, occult SNARE homologs to (or proteolytic fragments of) syntaxin, VAMP and SNAP-25 need to exist in protease-resistant conformations that are not recognized by our polyclonal antibodies (i.e. lack all the epitopes recognized by the antibodies used here) at all protein concentrations and dilutions assessed with an ultrasensitive immunodetection protocol (Coorssen et al., 2002), and the putative proteolytic fragments would have to remain stably bound in the presence of chaotropic buffers. Although this is an alternative explanation, we think it simpler and thus more likely that CV-CV fusion can occur in either the absence of syntaxin or with 91-99% overall loss of SNARE proteins. This interpretation is also most consistent with the substantial differences in SNARE densities measured on native secretory vesicles compared with those required for fusion in reconstituted preparations (Weber et al., 1998;Coorssen et al., 2002). In the absence of SNARE and other modulatory influences, the fusion process has an endogenous Ca 2+ -activity whose midpoint is ~250 µM [Ca 2+ ]free.
Are there multiple protein-dependent or proteinindependent fusion mechanisms? Our analysis of the Ca 2+ activity and kinetics of fusion is based on the hypothesis that an increase in [Ca 2+ ]free increases the number of participating fusion complexes and that these complexes are randomly distributed among CVs (Vogel et al., 1996;Blank et al., 2001). This analysis has consistently explained the experimental data observed in this system (Vogel et al., 1996;Blank et al., 1998a;Blank et al., 1998b;Coorssen et al., 1998;Tahara et al., 1998;Blank et al., 2001;Ikebuchi et al., 2001). However, if changes in the Ca 2+ activity and the kinetics of fusion do indicate a shift from one fusion pathway to another, then perhaps SNARE proteins normally mediate fusion, but in their absence alternative mechanisms become evident; redundant pathways may exist. This is an intriguing and provocative alternative hypothesis. However, no changes occurred in kinetics following clostripain treatment, despite substantial decreases in SNARE density, whereas comparable decreases in SNARE density following papain treatment did alter the kinetics of fusion. This is perhaps the most direct evidence against an essential role for the SNAREs in fusion, as any change in the density of an essential component of the fusion complex would be predicted to inhibit the rate of fusion (Vogel et al., 1996;Blank et al., 2001). The differential effects of protease treatments are difficult to reconcile with a simple switch from a SNARE-mediated to a SNARE-independent, protein-mediated fusion pathway without postulating the existence of multiple classes of SNAREs such that clostripain preserves the SNARE-mediated pathway by acting only on nonfunctional SNAREs while papain and trypsin target only the functional SNAREs. Rather, the demonstrated relationship between SNARE density and Ca 2+ activity, seen in the papain data (Fig. 6), clearly indicates the existence of a single fusion pathway (with an inherent Ca 2+ activity). The Ca 2+ sensitivity of this fundamental fusion pathway is modulated in the 'physiological' range of [Ca 2+ ]free either by SNAREs coupled with other modulatory components such as 'Ca 2+ sensors,' or by other components alone that track the papain-dependent changes in SNARE density via similar proteolytic sensitivity. Redundancy in modulatory factors may be a conserved hallmark of triggered exocytosis; the data indicate that trypsin and papain more broadly affect these modulatory factors than does clostripain.
A switch from a SNARE-mediated to a protein-independent mechanism (e.g. Ca 2+ -mediated lipid fusion) can not explain the data because extensive trypsinization ablates the fusion response, even to several millimolar [Ca 2+ ]free; if any of the fusion observed was nonphysiological or purely lipidmediated, then it should also have been observed under these conditions. As the fusion response is also lost after treatments with other broad-spectrum proteases (e.g. chymotrypsin; data not shown), it is unlikely that purely lipid-mediated fusion occurs under our experimental conditions. In fact, even when using pure lipid vesicles (LV) as the target membrane (CV-LV fusion; Ca 2+ -dependent fusion at <60 µM [Ca 2+ ]free), proteinaceous machinery is still essential to the triggered fusion steps, as determined by thiol sensitivity (Vogel et al., 1992). The shift in the midpoint of the Ca 2+ activity curve for CV-CV fusion never enters the range of [Ca 2+ ]free required to induce purely lipid fusion; fusion of lipid membranes containing mixtures of neutral and negatively charged lipids requires tens of millimolar [Ca 2+ ]free (Cohen et al., 1980;Cohen et al., 1984;Duzgunes et al., 1981;Coorssen and Rand, 1995;Zimmerberg and Chernomordik, 1999). The Ca 2+ activity curves for all conditions, except for papain where the distribution of Ca 2+ sensitivity is believed to be significantly altered, indicate that >95% fusion occurs between ~10 and 160 µM [Ca 2+ ]free, a range that is consistent with the physiology of Ca 2+ -triggered exocytosis. Because it takes time for the proteases to work, we must also consider a shift in mechanism due to incubation time. However, the invariance of p/nwith incubation time (0-4 hours), indicating no change in the estimated energy of fusion, makes a time-dependent shift in mechanism unlikely.

SNAREs and the process of exocytosis
The simplest hypothesis is that there is only one endogenous mechanism of protein-mediated fusion in this system, and our dissection of this mechanism results in altered responses. According to this hypothesis, although SNARES have key roles in the pathway of exocytosis, they are not members of any minimal protein set required for Ca 2+ -triggered fusion in this regulated exocytotic system -certainly not in any capacity that requires their intact cytosolic domains; their rapid intermembrane binding does not appear to be essential to the minimal mechanism of native membrane merger . Of course, SNAREs are enormously important to the process of exocytosis; there is good evidence that SNAREs affect an early stage of target membrane recognition, vesicle docking and/or priming (Pelham, 2001;McNew et al., 2000;Scales et al., 2000a;Scales et al., 2000b). Indeed, our new estimates indicate that SNAREs, together with other modulatory proteins, can contribute energy to the fusion mechanism, although not enough to directly trigger fusion, except perhaps at densities in excess of those in native vesicle membranes (Weber et al., 1998;Coorssen et al., 2002). SNAREs probably play a role in trafficking and the localization of CV to PM docking sites, which occurs during the oocyte-to-egg transition (Berg and Wessel, 1997), because fusion can be disrupted by clostridial toxins injected into eggs whose CVs have been de-docked (Bi et al., 1995); these PM docking sites may be associated with fusion complexes (Ikebuchi et al., 2001).
In other systems, genetic knockouts of VAMP and SNAP-25 affect the regulation of exocytosis but not the ultimate capacity to release vesicular contents (Nonet et al., 1998;Yoshihara et al., 1999;Schoch et al., 2001;Rao et al., 2001;Washbourne et al., 2002), leading to two hypotheses -that SNAREs catalyze fusion reactions by stabilizing transition states (Schoch et al., 2001) and that multistep, Ca 2+ -dependent assembly of SNAREs forces membranes together to promote fusion (Xu et al., 1999;Scales et al., 2001). Considering the findings presented here, neither of these hypotheses is probable because the rate of CV fusion (rather than of multiple late steps in the exocytotic pathway) in clostripain-treated samples was independent of SNARE density, and the removal of SNAREs and other modulatory proteins suggests an inherent Ca 2+ activity in the triggered steps of fusion.
In regulated fusion, SNAREs may function as structural proteins that prepare or optimize local release sites , perhaps by coordinating the right vesicles (Rothman, 1994;McNew et al., 2000) to the right Ca 2+ channels (Rettig et al., 1997;Jarvis et al., 2000), for which they then act as regulatory components. This represents another interpretation of SNARE knockout data (Schoch et al., 2001;Washbourne et al., 2002). If VAMP loss alters presynaptic architecture such that vesicles dock less effectively (outside domains of increased intracellular Ca 2+ ), leading to diminished functional docking by ~10%, then spontaneous release and measures of docked, primed vesicles (i.e. sucrose pools) would be diminished by ~10% (Parsons et al., 1995) and the fast evoked response to Ca 2+ would be lost (Xu et al., 1998).
A calcium regulatory complex that is papain sensitive Although we do not yet know whether the fusion complex itself is inherently Ca 2+ sensitive, or whether this property is resident to an associated sensor, this sensitivity can be correlated to SNARE density following proteolysis by papain. This suggests a native papain-sensitive complex that includes SNAREs and associated proteins. However, this complex is not essential per se in the triggered fusion steps of exocytosis because in our work higher [Ca 2+ ]free overcame the block to fusion (seen at 'physiological' [Ca 2+ ]free) caused by proteolysis. Such recovery has also been documented in earlier experiments with Clostridial toxins (Dreyer and Schmitt, 1983;Ahnert-Hilger and Weller, 1993;Bittner and Holz, 1993;Lawrence et al., 1994;Glenn and Burgoyne, 1996;Lawrence et al., 1996;Capogna et al., 1997;Land et al., 1997;Fassio et al., 1999), suggesting that the concept of a regulatory complex may extend beyond the present system. In general, the testing and rescue of exocytosis at only one [Ca 2+ ]free after complex disruption tells us a great deal about the modulatory role of the complex. Incomplete Ca 2+ activity curves provide only a partial analysis of the pathway in question, leading to models that explain only a very circumscribed stage of exocytosis, upstream of the fusion steps. Complexes of SNAREs and accessory proteins may promote fusion efficiency in vivo. Our finding that proteolysis by papain reduced the probability of fusion supports this interpretation. Although similar influences of SNAREs on the probability of fusion have been noted in other systems (Finley et al., 2002;Stewart et al., 2002), the mechanism of this action was not discernible; here, we hypothesize that a complex of SNAREs and associated proteins may modulate the inherent Ca 2+ sensitivity of the native fusion complex, and that these modulatory proteins are more sensitive to papain than to clostripain.
In summary, the isolated CV preparation is unique in its stage specificity, allowing the dissection of post-SNARE, Ca 2+dependent steps of exocytosis. Proteolysis has revealed two different functional activities involving the Ca 2+ sensor(s) and the fusogen, in accordance with other studies indicating the involvement of additional proteins acting after SNAREs Tahara et al., 1998;Ungermann et al., 1998;Peters et al., 2001). In the current study we achieved a biochemical 'knockout' of syntaxin in stage-specific native secretory vesicles, without the loss of triggered fusion. These studies have also provided intriguing estimates of the inherent Ca 2+ sensitivity of the endogenous fusogen-fusion complex separate from a regulatory complex that appears to include SNAREs.