Photolysis of Caged Ca²⁺, Whole-Cell Capacitance, and Ca²⁺ Measurements.

Experimental procedures related to flash photolysis were performed as described (11, 13). Nitrophenyl-EGTA was supplied by G. Ellis-Davies (MCP Hahnemann University, Philadelphia).

For calcium ramp experiments (Fig. ​(Fig.4),4), calcium was measured by using a mixture of 400 μM Fura-4F (Molecular Probes; K_d for calcium, 1 μM), and 400 μM furaptra (Molecular Probes; K_d, 40 μM). The signal (ratio of fluorescence with excitation at 350 and 380 nm) was calibrated in vivo by making whole-cell patches with eight different pipette solutions with known calcium concentrations (12). The ramp was elicited by the fluorescence excitation light alternating between 350 and 380 nm at 40 Hz such that photolysis of nitrophenyl-EGTA could be combined with simultaneous measurement of [Ca²⁺]_i. For off-line analysis, the simultaneously obtained capacitance trace was averaged to give one capacitance value per calcium measurement (Fig. ​(Fig.44b, second trace). The rate between two consecutive measurements was calculated as

1

where C_m,j is the averaged capacitance around the jth calcium measurement, and Δt is the time between measurements (Fig. ​(Fig.44b, third trace). To calculate the rate constant, an estimate for the vesicle pool size is required. This estimate is difficult to obtain, because the contributions of individual vesicle pools are not always obvious in ramp experiments. We adopted the technique of numerically differentiating the (raw) capacitance trace and fitting a Gaussian distribution to the (smoothed) derivative to find the time of fastest capacitance increase (the top point of the Gaussian). We assumed that the pool would comprise 2 times the capacitance change up to this point. This method measures the rate constants of the vesicle pool that fuses first, i.e., the pool with the highest rate constant at low [Ca²⁺]_i. For control cells this will be the readily releasable pool; however, some contamination from the slowly releasable pool cannot be excluded. We assume that pool estimates in practice were correct within 20–30%.

Open in a separate window

Figure 4

Double mutation in the D58/E170/Q177 site changes the calcium dependence of exocytosis. (a) Example of a control cell stimulated by continuous weak UV illumination. Capacitance (line, Bottom) and calcium (small boxes, Top) measurements are shown. (b) Rate constant calculations. The capacitance trace was averaged (second trace) around each calcium measurement (first trace), and the local rate was calculated (third trace). Based on an estimate of the size of the readily releasable pool (see a), the remaining pool size was calculated (fourth trace), and finally the rate constant (fifth trace) was calculated as the rate divided by remaining pool size. (c) Plot of rate constant versus [Ca²⁺]_i on a double logarithmic scale. (d) Double logarithmic plot of rate constant versus calcium for 15 control cells (black squares) and 13 cells overexpressing the SNAP-25 E170A/Q177A mutant (gray circles). For comparison with ramp data, the rates for the corresponding flash data (from fig. 1i) are indicated with open symbols with crosses. For the control cells the rate for the readily releasable pool was taken, and for the mutant cells the rate for the slowly releasable pool was chosen. The flash rates confirm that the apparent saturation of ramp rates at around 1 s is artificial, probably caused by some contamination of the pool size estimate with slower components at later times. Fitted to control data are the three binding-sites secretory model in f, whereas fit to mutant data are the two binding-sites model. (e) Double logarithmic plot of rate constant versus calcium for 13 control cells (black squares) and 13 cells overexpressing the synaptobrevin S75A/E78A/T79A mutant (gray circles). The line is the same fit of the three-step model as the one in a. (f) Sequential calcium-binding models for exocytosis. (Upper) Standard model for chromaffin cells, with three equivalent calcium-binding sites. (Lower) Modified model, with only two equivalent calcium-binding sites. The models were solved numerically by using a fourth order Runge-Kutta method with automatic step-size control and a typical calcium ramp as forcing function. A simplex method was used to fit the models to the estimated data in d. Fitted parameters in the three binding-sites model (fit to controls in d; only α was varied, β and γ were taken from ref. 12): α = 5.7 μM⁻¹s⁻¹; β = 56 s⁻¹; and γ = 1450 s⁻¹. Fitted parameters in the two binding-sites model (fit to SNAP-25 E170/177A data): α = 4.2 μM⁻¹s⁻¹; β = 69 s⁻¹; and γ = 4.7 s⁻¹.

By using this estimate, we calculated the remaining pool size (Fig. ​(Fig.44b, fourth trace) corresponding to each rate measurement (pool_j) and calculated the rate constant, rc_j (Fig. ​(Fig.44b, fifth trace), as rc_j = rate_j/pool_j. Finally, the rate constant was plotted as a function of the interpolated [Ca²⁺]_i between the two measurements, giving rise to the rate estimate (Fig. ​(Fig.44c). The underlying assumption of this plot is that the calcium ramp was slow enough that the calcium sensor can be considered to be in equilibrium with calcium at all times (“quasi steady state”). Simulations of the standard secretory model (Fig. ​(Fig.4f4f Upper) showed that this was the case for the calcium ramps used in this paper.

To bin results from several cells (Fig. ​(Fig.44 d and e), paired results of calcium and rate constants from each cell (see Fig. ​Fig.44c) were transformed by (x,y) → [log x, log(y + 1)], where x is [Ca²⁺]_i, and y is the rate constant [transformation of the rate by log(y + 1) was necessary, because measured rates could be negative]. After binning for calcium on this logarithmic scale, the mean ± SEM was calculated on transformed data, and the mean and limits were back-transformed to give the plots in Fig. ​Fig.44 d and e.

The Calcium Dependence of Secretion at Low Calcium Concentrations.

If the calcium sensor for exocytosis is either linked to the SNARE complex or integral to it, mutation in the SNARE complex may alter the calcium dependence of secretion. To address this question, we devised a method to derive the dependence of the secretory rate constant on [Ca²⁺]_i at the release site. The ideal calcium concentration for these studies is 0.5–5 μM, because in this concentration range the calcium sensor is expected to be far from saturation. At the same time, the limitation of being able to give only two or three stimulations to a single cell (due to run-down) combined with significant cell-to-cell variability made it advantageous to derive the entire calcium dependence from a single stimulation.

The solution was to use a calcium ramp technique where calcium was liberated slowly from the calcium cage over the course of 2–5 s by weak UV illumination. The UV illumination was alternated between 350 and 380 nm, allowing the simultaneous fluorimetric measurement of [Ca²⁺]_i. In the control cell shown in Fig. ​Fig.44a, a calcium ramp triggered a faster phase of capacitance increase followed by a slower, more sustained phase. Data analysis was carried out as shown in Fig. ​Fig.44b. Based on a calculation of the local rate of capacitance change around each calcium measurement and an estimate of the size of the releasable pool (Fig. ​(Fig.44a), the rate constant was calculated around each calcium measurement. The rate constant then was replotted as a function of the [Ca²⁺]_i, yielding the desired relationship (Fig. ​(Fig.44c).

Fig. ​Fig.44 d and e show plots of rate constants versus [Ca²⁺]_i obtained from several cells. From Fig. ​Fig.44d it is evident that for low calcium concentrations (0.5–2 μM), control cells displayed a steeper relationship of the rate constant on [Ca²⁺]_i than cells overexpressing the SNAP-25 E170A/Q177A mutants. The slope was 2.7 for the control and 1.5 for the mutant cells. This result already indicates a difference in the number of calcium ions involved in exocytosis. To achieve a more precise description, we fitted the relationships in Fig. ​Fig.44d with sequential calcium-binding models for exocytosis (Fig. ​(Fig.44f). The simplest possible such model assumes that calcium binds sequentially and noncooperatively to three similar binding sites on the calcium sensor for exocytosis to occur (12). A satisfactory fit of this model to mean control data (upper blue line in Fig. ​Fig.44d) was obtained with only minor changes compared with the previously published values (ref. 12; α = 5.7 μM⁻¹s⁻¹ instead of 4.4 μM⁻¹s⁻¹). In contrast, to fit the shallower slope of the SNAP-25 E170A/Q177A relationship, the number of calcium-binding sites in the model was reduced by one (Fig. ​(Fig.4f4f Lower). With this change, then, the model showed a satisfactory fit to the mutant data (Fig. ​(Fig.44d, lower blue line) by assuming a much lower maximal rate (γ = 4.7 s⁻¹ instead of 1,450 s⁻¹ for the controls) but in the absence of large changes in on- or off-rates for calcium (see Fig. ​Fig.44 legend). This is the expected result if a binding site for calcium had been eliminated by the mutation and simultaneously it had been made possible for vesicles to fuse by using the two remaining binding sites.

The corresponding calcium ramp measurements on the synaptobrevin S75A/E78A/T79A mutant showed that the slope in the double logarithmic plot was almost unchanged (Fig. ​(Fig.44e), although the rates for the mutants were on average a little lower than in control cells. Therefore, we conclude that although mutation of the S75/E78/T79 site leads to a depression in secretion (Fig. ​(Fig.3),3), the fastest part of the remaining secretion (which is the part that in effect is analyzed in calcium ramp experiments, see Materials and Methods) exhibits an almost unchanged dependence on [Ca²⁺]_i. The integrity of the S75/E78/T79 site therefore is important for maintaining the size of primed vesicle pools before stimulation, but apparently it does not modulate Ca²⁺-dependent triggering.

In Fig. ​Fig.44d we compared the rate data obtained by the ramp method with the rates obtained at higher [Ca²⁺]_i values by flash experiments (symbols with crosses) using the rapidly releasable pool rate for the control situation and the slowly releasable pool rate for the E170A/Q177A mutant (Fig. ​(Fig.11i). These rates fall on the extension of the model fits to higher calcium values. This result validates the ramp method and also indicates that in effect the measured rate for the E170A/Q177A mutant is that of the slowly releasable pool, because the rapidly releasable pool in this case is too small to be resolved in ramp experiments. It is an open question whether the slowly releasable pool in E170A/Q177A-expressing cells has the same molecular basis as the slowly releasable pool in control cells, but if this was the case, it would mean that the slow release from the slowly releasable pool can be viewed as being that of a rudimentary fusion machine, characterized by a shallower calcium dependence.

Because mutation in the D58/E170/Q177 binding pocket abolished one calcium-binding site in the calcium sensor for exocytosis, apparently this site interacts with or constitutes part of the calcium sensor for exocytosis. There are two possible ways in which this could come about. First, the D58/E170/Q177 site could interact directly with calcium. Because the site does not provide the full coordination sphere required for calcium binding, simultaneous binding of another protein such as synaptotagmin I would be necessary to generate a full calcium-binding site. The other possibility is that the D58/E170/Q177 site could bind to another protein, which then in turn could bind calcium. The “rudimentary fusion machine” discussed above might represent a SNARE complex without such a protein bound. If calcium was to bind directly and exert an electrostatic effect, charge reversals in the pocket might simulate the calcium-bound state and create a functional SNAP-25 mutant. However, charge-reversal mutations (D58K, E170K, and D58K/E170K were studied) eliminated the burst component of secretion (Fig. 5, which is published as supporting information on the PNAS web site, www.pnas.org), and spontaneous fusion was not observed. This is the expected result if the site binds a calcium sensor distinct from the core complex. However, we cannot rule out that SNAP-25 creates a calcium-binding site in a complex with another, unknown partner and that the detailed structural integrity of this binding site is critical for exocytosis to occur.

In conclusion, charged putative binding sites on the outside of the SNARE complex are essential for normal fast calcium-triggered secretion as measured with high time-resolution techniques. Mutation of these sites hardly affects complex stability but has a profound effect on secretion, probably because of loss of an interaction with auxiliary factors/proteins. The finding of a mutant with a reduced calcium dependence of the exocytotic rate means that the calcium sensor of exocytosis is linked intimately to SNARE complex function. The SNARE complex and the calcium sensor therefore constitute parts of an integrated fusion machine.