Single secretory granules of live cells recruit syntaxin-1 and synaptosomal associated protein 25 (SNAP-25) in large copy numbers

Abstract

Before secretory vesicles undergo exocytosis, they must recruit the proteins syntaxin-1 and synaptosomal associated protein 25 (SNAP-25) in the plasma membrane. GFP-labeled versions of both proteins cluster at sites where secretory granules have docked. Single-particle tracking shows that minority populations of both molecules are strongly hindered in their mobility, consistent with their confinement in nanodomains. We measured the fluorescence of granule-associated clusters, the fluorescence of single molecules, and the numbers of unlabeled syntaxin-1 and SNAP-25 molecules per cell. There was a more than 10-fold excess of SNAP-25 over syntaxin-1. Fifty to seventy copies each of syntaxin-1 and SNAP-25 molecules were associated with a single docked granule, many more than have been reported to be required for fusion.

The three proteins syntaxin-1 (Syx), synaptosomal associated protein 25 (SNAP-25), and vesicle-associated membrane protein 2 (VAMP-2)/synaptobrevin are collectively called “soluble N-ethylmaleimide-sensitive factor attachment protein receptors” (SNAREs) and catalyze membrane fusion during exocytosis in neurons and neuroendocrine cells (1). Syx and SNAP-25 inhabit the plasma membrane and are collectively called target SNAREs (tSNAREs). One expects them to collect at the exocytic site before a vesicle or granule can fuse there. When we transfected PC12 cells with GFP-labeled Syx-1A (Syx-GFP) and selectively illuminated the cell surface of PC12 cells, we indeed saw that Syx-GFP clustered in the plasma membrane where single secretory granules had docked. Such “on-granule” clusters facilitated exocytosis of the associated granule and disassembled once exocytosis was complete (2). Unlike “off-granule” clusters (2, 3), they failed to form when Syx lacked its N-terminal domain.

It is estimated that three (4), eight (5), or more than three SNARE complexes (6) participate in the exocytosis of a secretory granule, 10 to 15 participate in the exocytosis of a synaptic vesicle (7), and a single complex participates in the fusion of liposomes (8). Here, we determined the number of t-SNAREs recruited by a single granule in live cells. We measured the fluorescence of on-granule Syx-GFP and GFP-SNAP-25 clusters and compared it with the fluorescence of single Syx-GFP and GFP-SNAP-25 molecules. By determining the amounts of Syx-1 and SNAP-25 per cell, we obtained the expression ratio of labeled and unlabeled t-SNAREs. We estimate that each docked granule recruits 50–70 copies each of Syx and SNAP-25.

Results

Clusters Saturate as the Syx-GFP Expression Level Rises.

In cells cotransfected with the granule marker neuropeptide-Y (NPY)-mCherry and the t-SNARE Syx-GFP, NPY-mCherry showed punctate red fluorescence when viewed with total internal reflection fluorescence (TIRF). Each spot represents a single granule (e.g., Fig. 1A, Left). At low expression levels of Syx-GFP, green fluorescence is punctate as well, with a subset of spots representing “on-granule” clusters precisely aligned with granules. With increasing Syx-GFP expression, both the cell surface and on-granule clusters initially brightened in proportion (2), but when the CMV promoter forced strong expression, Syx clusters no longer were clearly visible in single images (Fig. 1A, Right). They remained present, nonetheless, because when square regions were centered on granules and many such regions were copied into the Syx images and excised, the resulting average Syx image (Fig. 1B, Right) clearly showed a bright spot at the center. The spot had the same size as that obtained at 35-fold lower Syx-GFP expression levels. The difficulty of seeing Syx spots at high expression levels suggests that granule-associated Syx sites had become saturated while the plasma membrane continued to fill with granule-unrelated Syx.

Fig. 1.

Clusters vary with expression levels. (A) Red (Left) and green fluorescence (Right) of a cell transfected with NPY-mCherry and with Syx-GFP using the CMV promoter. Strong expression of Syx makes S = 15,700. (B) (Left) Averages of ΔF and S were determined for each cell. The averages were sorted by S and binned at 15 cells per point. Next, ΔF was determined in regions placed randomly onto the same cells, plotted against S, and fitted by a straight line (slope ΔF/ S = 0.011). Finally, the difference was formed. ●, Measurements with a crippled CMV promoter (dCMV, n = 224 cells); ○, measurements with an intact CMV promoter (n = 122 cells). Curve is the best fit of ΔF = B_max x S/(k + S) to the combined data, with B_max = 297 CU and k = 2,225 CU. Coordinates on the top and right are based on the conversion factors in rows 5 and 6 of Table 1. (Right) Syx clusters at expression levels corresponding to S = 200–1000 CU (mean, 485 CU, 63 cells, dCMV) and S = 12,400–23,000 CU (mean 17,543 CU, 16 cells, CMV). Gaussian fits gave essentially identical widths. For clarity, most of the plasma membrane fluorescence was subtracted. (Scale bar: 0.5 μM.) (C) Percentage of granules associated with Syx-GFP. Pairs of images centered on single granules (2) were viewed with auto-scaling and scored visually by two independent viewers as to whether they contained a spot centered to within 89 nm. The fraction scoring positive was determined for each cell. Each point averages multiple cells with similar S values. ●, Measurements with the dCMV promoter (n = 224 cells); ○, measurements with an intact CMV promoter (n = 122 cells; ▲, association of Syx-GFP with random locations. (D) Schematic with two granules (shaded), one associated with a Syx cluster in the plasma membrane (blue). Vertical lines represent GFP-labeled (red) and endogenous Syx molecules (black). (E) As in B but for cells expressing GFP-SNAP-25 (dCMV promoter). (Left) ΔF as a function of S as in B (Left), and in bins of 20. (Right) Average of granules as in B (Left) and at S = 200–380 CU (mean 290 CU). ΔF represents 0.8 fluorescent molecules.

As described in ref. 2, the fluorescence per pixel was measured in specific regions and given as camera units (CU). Small circles (radius 0.312 μm) were centered on granules and copied into the Syx image, and the fluorescence there (C) was compared with that in a surrounding annulus (S). S is attributed to Syx-GFP fluorescence that is spatially unrelated to granules, and the difference, ΔF = (C − S), is attributed to Syx-GFP molecules specifically bound by granules. When results from large numbers of cells were plotted against S, ΔF was seen to saturate (Fig. 1B, Left). As expression levels rose, Syx-GFP evidently occupied an increasing fraction of available sites on granules and ultimately displaced the endogenous Syx. The results were well described by a model where Syx binds to a fixed number of sites, B_max, with a dissociation constant, k.

Some granules failed to attract Syx-GFP in the plasma membrane (2). To test whether they might succeed at higher expression levels, granule locations were scored as to whether they coincided with a cluster. On average, 56% scored positive, regardless of expression level and of the promoter used (Fig. 1C). Because Fig. 1 is based on images collected over 1-s intervals, we suggest that, in any given second, two kinds of granules are visible in TIRF (Fig. 1D). 56% are molecularly docked at the plasma membrane, where they induce Syx clusters that are in equilibrium with both freely diffusing Syx molecules and off-granule Syx clusters. The others do not interact with Syx, perhaps because they are too far from the cell surface.

SNAP-25.

GFP-SNAP-25 is functional, because it rescues exocytosis both in SNAP-25 knockout mice (9) and after cleavage of endogenous SNAP-25 by Botulinum toxin E (10, 11). Nonetheless, the fluorescence of GFP-SNAP-25 was diffuse or mottled (Fig. S2). The following experiments showed that the diffuse fluorescence did not arise from soluble SNAP-25 that had failed to become membrane associated. When cells coexpressing GFP-SNAP-25 and free mCherry were permeabilized with digitonin, all but 5% of the mCherry was lost, but 88% of the GFP-SNAP-25 stayed behind (Fig. S1). Nonetheless, 87% of SNAP-25 was too mobile in the plane of the plasma membrane (see Fig. 5) to be located on submembrane organelles. Evidently the fluorescence visible in TIRF is almost entirely plasma membrane-associated.

Fig. 5.

Movement of single Syx-GFP molecules. (A–C) Image in a 50-Hz movie. The area outlined in red in A is magnified in B and bandpass-filtered in C. (D) The area outlined in red in C is shown at 20-ms intervals. A Syx-GFP molecule first moves downward and then moves toward the right; red points indicate its center of mass. The dot vanished temporarily in the penultimate frame as the GFP entered a dark state (“blinked”). Blinking terminated a track. (E) Histograms of distances traveled in 200 ms were fitted (red line) by Eq. [1] (22):

Here r is the distance traveled during the time t (200 ms in this case), D₁ and D₂ are the diffusion coefficients, and A₁ and A₂ are proportional to the fractions of fast and slow molecules, respectively. With a single diffusion coefficient and A₂ forced to zero, the fit is poor for both Syx-GFP (black line) and for SNAP-25. These calculations are based on 8,811 trajectories in 17 cells (Syx-GFP) and 5,799 trajectories in 10 cells (GFP-SNAP-25). (F) Overlay of 50-Hz movies (0.14 mJ per frame at 488 nm) of granules (red) and Syx-GFP molecules (green). Granules were stationary during the observation period, and their images are averages. Green movies were low-pass filtered. (Upper) Syx-GFP was captured at the granule site (*) and bleached there about 0.44 s later. (**) was recorded 20 ms after the preceding image to show abrupt bleaching. (Lower) Syx-GFP was present during the first 0.24 s and then released. Times are relative to the beginning of the sequences (cell SB2278).

Although GFP-SNAP-25 clusters were hard to detect in individual images, they were clearly visible in averages (Fig. 1E, Right). After correcting for the possible green fluorescence of NPY-mCherry (2), ΔF/S was 0.018 ± 0.006 when 200 < S < 1,000 CU (n = 34 cells), much less than for Syx-GFP (ΔF/S = 0.137 ± 0.011, n = 89 cells) (2). GFP-SNAP-25 showed no saturation over the range of S explored (Fig. 1E, Left).

Clusters and Single Molecules.

We can determine how many molecules inhabit a cluster if the brightness of single Syx-GFP and GFP-SNAP-25 molecules is known. Single molecules can be distinguished from clusters because the former bleach in single steps (12–14). At the transfection levels and excitation intensities used so far, bleaching caused Syx-GFP spots to dim with a time course that was noisy but not clearly stepwise (Fig. 2 A–C). At 10-fold higher excitation intensity cells with much lower Syx-GFP expression could be imaged. Bleaching was much faster, the fluorescence was invariably punctate, and about 70% of the spots bleached in single steps (Figs. 2 D and E).

Fig. 2.

Photobleaching of Syx-GFP clusters and single molecules. (A). Averages of five consecutive 20-ms exposures (0.1 mJ at 488 nm). Time starts on illumination. (Cell MK2316) (B) C vs. time for the spot at the center. (C) Average of nine traces as in B. Exponential time constant is 1.62 s. (D) A stationary Syx-GFP molecule in a cell with extremely low Syx-GFP expression. Consecutive 20-ms exposures at 10-fold stronger illumination (1 mJ per exposure at 488 nm). (Cell SB2823) (E) C vs. time for the spot at the center. (F) Average of 50 traces as in E (four cells). Time constant is 0.17 s.

Brightness of a Single Syx-GFP Molecule.

Single Syx-GFP molecules were dim. To collect the maximum number of photons from each, we imaged fixed cells in which molecules did not move appreciably and could be observed until they bleached. Fig. 3A is an average of the first 10 frames of a movie and shows numerous spots of varying brightness. Spots were located automatically, their locations were superimposed onto the movie, and the fluorescence at each location was plotted against time (Fig. 3B). When the light was turned on, bright fluorescence fluctuated strongly and then vanished in steps as single molecules bleached.

Fig. 3.

Single molecules in paraformaldehyde-fixed cells. (A) Average of the first 10 frames in a 20-Hz movie (1 mJ per exposure). (Cell MK3012) (B) Fluorescence at six locations in one cell. Excitation (488 nm) started at the vertical red line; vertical black lines mark the moment of bleaching, and dashed horizontal lines indicate the background. Bleaching is detected as a downward step followed by an abrupt lessening of noise. (Cell SB2450) (C) Mean time to bleaching (latency) as function of step size. Spots are located in 10-frame averages as in A. Step size (ΔC) is the fluorescence before bleaching minus fluorescence in the first 1.5 s thereafter, in traces as in B. Data are sorted by step amplitude and binned into packets of 10 points. The plot overestimates the latencies of the brightest molecules because we excluded spots bleaching in <200 ms. The plot underestimates the latencies of the dimmest molecules because our observation time was <7.5 s. Red line is taken from simulation in D. (D) Histogram of step amplitudes. Red line indicates a Monte-Carlo simulation assuming a normal distribution for both the brightness of molecules (coefficient of variation 0.3) and their distances from the coverslip (SI Text and Fig. S2). (Inset) Region near the origin on an expanded abscissa. Red line indicates fraction of detected synthetic spots plotted as a function of their brightness.

Both the amplitudes and times of steps varied strongly and in inverse proportion (Fig. 3C). Because dim molecules bleached most slowly, we wondered whether even dimmer molecules might be detected during longer observations. Indeed, when spots were located in 100-frame averages and the detection threshold was lowered, 40% additional spots were found, and each generated an additional trace similar to those in Fig. 3B. The combined step-size histogram (Fig. 3D) showed a peak at about 60 CU; the mean step size was ΔC = 87 ± 5 CU (n = 384 steps in eight cells). Results with GFP-SNAP-25 were similar (mean, 82 ± 3 CU, 344 steps in four cells). In live cells expressing Syx-GFP (Fig. 2 D and E), the mean step size was somewhat larger (122 ± 11 CU per molecule, 65 steps in six cells), but because dim molecules tended to diffuse away before they bleached, our analysis in live cells is biased against them, and we consider the result from fixed cells more reliable.

We tested whether the comparative rarity of step sizes <60 CU in Fig. 3D resulted from a failure to detect dim spots in images such as Fig. 3A. Images of Gaussian point-spread functions were placed into 100-frame averages of cells. The images represented synthetic “single molecules” whose brightness could be varied systematically. When their brightness was 10 CU, we detected 90% of them (red symbols, Fig. 3D Inset); hence detection failures cannot explain the peak in the histogram.

Like others working with TIRF on live cells (15, 16), we obtained a histogram that is skewed and broader than obtained in epifluorescence recordings in cell-free systems (14). Because our results are based on single steps, the skewing cannot be explained by multiple molecules per spot. Instead, Fig. 3C shows that the dimmest molecules bleach most slowly, suggesting that these molecules were not intrinsically dim but instead excited less intensely. The excitation may vary, e.g., if the field of view is not illuminated evenly or if the distance between the plasma membrane and the glass coverslip varies. Indeed, when the distance was allowed to vary in a simulation, a normal distribution of molecule brightness became skewed, and brightness varied inversely with the time to bleaching (Fig. 3D, red line; SI Text, and Fig. S2).

Molecule brightness also depends on highly local factors. In a striking example (Fig. 3B, Bottom Trace), a spot bleached in three steps, presumably representing three fluorescent molecules in a cluster. The first and last steps differed eightfold in amplitude, even though Gaussian fits to difference images reported the two molecules to be within 15 nm of each other in the plane of the coverslip. The illumination intensity is unlikely to vary strongly over such a short distance, and we do not understand all factors influencing molecule brightness. On average, however, such factors may be assumed to operate regardless of whether we image single molecules or many.

Numbers of Fluorescent Molecules.

The mean step size of ΔC = 87 CU for Syx-GFP translated into 10.4 CU at the lower illumination energy used in Fig. 1 (Methods). Hence one Syx-GFP molecule generated an average of 10.4 CU in each of the 39 pixels of the circle from which we collect its light (area, 0.31 μm²). Because molecules and clusters were analyzed with the same circles, the number of fluorescent molecules in a cluster equals ΔF/10.4. Some Syx-GFP molecules failed to fluoresce in our cells, either because they bleached while being imaged or because they had not matured (Methods, SI Text, and Fig. S3). As a result of both effects, the average Syx-GFP peptide chain provided a fluorescence intensity of 5.6 CU in a 0.31-μm² circle (Table 1, row 5).

Table 1.

Single molecules and abundance of t-SNAREs

GFP-tagged	Syx-GFP	GFP-SNAP-25
ΔC @ 1.0 mJ (CU/molecule)	87 ± 5	82 ± 3
ΔC @ 0.1 mJ (CU/molecule)	10.4 ± 0.6	9.8 ± 0.4
Fraction fluorescent	0.66 ± 0.09	1.02 ± 0.12
Bleaching factor	0.82 ± 0.01	0.82
ΔF (CU/peptide chain)	5.6	8.0
F (CU) @ 1 peptide chain/μm2	1.7	2.5

Endogenous	Syx	SNAP-25
Molecules/cell (x103)	428 ± 37	7,440 ± 850
Molecules/μm2	540	7,500

Rows in the table body are numbered from top to bottom. (The row “Endogenous” is not numbered.) Row 1: Step amplitudes ΔC in 1.0-mJ exposures as in Fig. 3. Row 2: Row 1 divided by 3.36 (Methods) to convert to 0.1-mJ exposures. Rows 1 and 2 are expressed in CU per fluorescent molecule. Rows 3 and 4: Factors correcting for bleaching and immature GFP (Methods). Row 5: Average fluorescence generated by a single GFP-tagged t-SNARE, equal to the product of Rows 2, 3, and 4. Row 6: Row 5 multiplied by the area of the circle in which we measure ΔC and ΔF (0.31 μm²). Rows 5 and 6 are expressed in CU per peptide chain. Row 7 is based on Fig. 4. Row 8 is derived from Row 7 by assuming that all Syx and 80% of all SNAP-25 molecules reside in the plasma membrane and that the plasma membrane area is 796 μm² per cell. For details see main text and Methods.

Fluorescence also could be calibrated in terms of molecules per square micrometer. If one fluorescent molecule generates 10.4 CU over a 0.31-μm² area, then 3.24 molecules generate the same fluorescence over 1 μm², an area 3.24 times larger. Hence a fluorescence intensity of 10.4/3.24 = 3.2 CU represents a density of one fluorescent molecule/μm², and a fluorescence intensity of 3.2 × 0.66 × 0.82 = 1.7 CU represents one peptide chain/μm² (Table 1, row 6). Values for GFP-SNAP-25 were found similarly. Rows 5 and 6 in Table 1 formed the basis for the top and right-hand coordinates in Fig. 1 B and E.

Number of Molecules per Cell and per Granule.

To determine the number of endogenous SNAREs, we compared samples of purified protein at known concentrations with cell lysates from known numbers of cells in quantitative Western blots (Fig. 4 A and B). Values for Syx ranged from 369,000–511,000 molecules/cell (means from each of four different antibodies). The overall mean (428,000 molecules/cell, n = 4 antibodies) is in fair agreement with an earlier value (17). Following others (17–19), we attribute all Syx to the plasma membrane. With a cell-surface area of 800 μm² by capacitance measurement (7.96 ± 0.76 pF, n = 10 cells), there are ∼540 molecules/μm². Means of values for SNAP-25 ranged from 6.06 × 10⁶ to 8.26 × 10⁶ molecules/cell with three antibodies (overall mean, 7.22 × 10⁶ molecules/cell, n = 3 antibodies). Given the cell-surface area, and that 80% of the SNAP-25 in PC12 cells resides in the plasma membrane (18), there were ∼7,500 SNAP-25 molecules/μm². Competition from this large excess of endogenous SNAP-25 explains why GFP-SNAP-25 clusters appeared dim relative to background and why GFP-SNAP-25 failed to saturate on-granule sites at the expression levels explored.

Fig. 4.

An excess of SNAP-25 over Syx. (A) Twofold serial dilutions of cell lysate (Left) and of recombinant Syx-1A (Right), both probed with antibody SC-12736. Numbers of cells (Left) and amounts of protein (Right) are given. We calculated 321,000 molecules per cell from this example. (B) Twofold serial dilutions of cell lysate (Left) and of recombinant SNAP-25 (Right) probed for SNAP-25 with antibody SySy 111002. We calculated 5,190,000 molecules per cell in this example. (C) Data in Fig. 1B (Left) are replotted, including all results with S > 50/μm². E = 540/μm² from Table 1. B = ΔF/R is the total amount of Syx bound per granule. Curve is the best fit to B = B_max R/(R + k) with B_max = 50 per granule and k = 467/μm². Results are weighted by the reciprocals of their variance. ●, Measurements with the dCMV promoter; ○, measurements with the intact CMV promoter.

From the surface densities, E, of endogenous Syx and SNAP-25, one may calculate the fraction of surface-associated SNARE that is GFP-tagged in each cell. If G is the surface concentration of GFP-labeled SNARE, the fraction is R = G/(G+E). The total number of Syx or SNAP-25 molecules bound by the average granule, B, then can be calculated as B = ΔF/R.

For SNAP-25, G was estimated as in Table 1 (row 6) either as S or as F_cell, where F_cell was the fluorescence averaged over the “footprint” of the cell. In cells with 0.4% < R < 5.4%, granules are expected to be far from being saturated with GFP-SNAP-25, and indeed no saturation was observed (Fig. 1E). Therefore we may ignore the occupation of sites by G, and B applies to untransfected cells. B was 37 ± 11 or 42 ± 13 molecules per granule (n = 47 cells), respectively, depending on whether R was obtained from S or from F_cell. The estimates scale with E and would be halved if there were half as many endogenous molecules per cell.

For Syx, saturation makes our estimate less sensitive to the value of E. Fig. 4C shows a plot of B vs. (S + E). The line is the best fit of a model wherein Syx binds to a fixed number of sites. It extrapolates to B = 27 molecules per granule at the endogenous surface density of E = 540/μm². In other PC12 cells, larger endogenous surface densities (17, 19) may result in larger values. However, the saturating value in Fig. 1B (∼53 Syx-GFP molecules per granule) gives an upper limit that is independent of E.

Mobility of t-SNAREs.

In single-particle tracking studies, plasma membrane-anchored proteins show alternating periods of rapid and slow diffusion as they are captured in nanodomains and escape from them (15, 20). Given that some Syx and SNAP-25 molecules are similarly bound in clusters, we tested whether single-particle tracking (SI Text) would distinguish populations with different mobilities in movies of cells expressing fluorescent SNAREs (Fig. 5 A–D). Spots assumed to represent single molecules were located and tracked (21), the tracks were cut into 200-ms nonoverlapping segments, and the distances traveled in the 200-ms intervals were plotted as a histogram (Fig. 5E). Brownian motion (22) fails to account for a large population of fast molecules if only a single diffusion coefficient is allowed but fits well if two-thirds of the molecules move rapidly and the remainder moves 10-fold more slowly (Fig. 5E). The finding suggests that one-third of Syx molecules are sequestered in nanodomains.

Fast Syx molecules (diffusion coefficient D = 0.106 μm²/s) were nearly as fast (23) as the protein linker for T-cell activation (LAT) (24). Like Syx, LAT has essentially no extracellular domain that could slow diffusion by binding to extracellular partners. The slow Syx-GFP molecules were barely faster (D = 0.010 μm²/s) than Syx-GFP in fixed cells (Table S1), perhaps because our 200-ms sampling interval was too short to track slow movement accurately. The value D = 0.014 μm^2/s in a 1- to 2-min measurement of fluorescence recovery after photobleaching (FRAP measurement) (17) may be more reliable for slow molecules, but that method did not detect the faster molecules observed here.

At least some Syx molecules are slow because they are temporarily immobilized beneath granules. In Fig. 5F a Syx-GFP molecule was seen to approach a granule, become captured at the granule site, and bleach about 440 ms later as the green fluorescence was lost and the dot in the center turned from orange to red. The abrupt loss of green fluorescence suggests the bleaching of a single molecule. At another granule, a Syx-GFP molecule was released after more than 0.24 s of residence (see also Movies S1 and S2). Evidently, on-granule Syx clusters are in equilibrium with the remainder of the plasma membrane. In many other cases GFP molecules entered and left a granule site without clearly stopping. Probably the capture radius of a Syx cluster is less than the optical resolution of our microscope.

Fast and slow diffusion coefficients also were found for GFP-SNAP-25. Fast SNAP-25 molecules had a value (0.24 μm²/s) similar to that of other molecules anchored in the membrane solely by acyl chains (25). One in eight SNAP-25 molecules seemed hindered in its diffusion, possibly because it was bound to diffusionally hindered Syx. The other SNAP-25 molecules diffused too rapidly to be bound to Syx.

Discussion

We have shown that both Syx and SNAP-25 molecules form clusters at sites aligned with secretory granules. Minority populations of both proteins were strongly hindered in their diffusion within the plasma membrane, consistent with their being caught in clusters. In the case of Syx, single molecules could be observed as they were trapped at granule sites or released from them. Finally, the local concentration of Syx molecules at granule sites underwent rapid fluctuations that ceased in fixed cells (2). In combination, these findings suggest that Syx, and probably SNAP-25, reversibly congregate in the plasma membrane subjacent to individual granules.

Most Syx clusters at granule sites were present only intermittently and formed and dispersed in a coordinated fashion (2). Here we show that, in any given second, 50–60% of granules accumulate Syx clusters, regardless of Syx expression level. Taken together, the findings suggest that granules transit between being capable or incapable of recruiting a Syx cluster. Granules with Syx clusters are molecularly docked, and apparently such docking is reversible. We were unable to test whether SNAP-25 clusters form and disassemble synchronously with Syx clusters. GFP-SNAP-25 clusters were strikingly hard to see, probably because GFP-SNAP-25 had to compete with a more than 10-fold excess of endogenous molecules. The large excess will tend to favor 1:1 Syx/SNAP-25 heterodimers over the possibly unproductive Syx₂/SNAP-25 heterotrimers that form in vitro (26).

By single-molecule recordings, we have determined the surface densities of fluorescent Syx-GFP and GFP-SNAP-25, as well as the number of molecules per cluster. Knowing the amounts of endogenous protein, we could calculate that the average granule of an untransfected cell bound 27 Syx molecules and 37 SNAP-25 molecules. Given experimental error, both molecules may be present in equal numbers, as expected if Syx and SNAP-25 combine in a heterodimer (26). Because the on-granule Syx-GFP signal results from only 56% of the granules seen in TIRF, one Syx-associated granule binds 48 Syx molecules. Applying the same percentage to SNAP-25 suggests that 66 molecules are bound per SNAP-25-associated granule. We suggest that fusion of secretory granules occurs near a metastable cluster containing some 50–70 Syx/SNAP-25 heterodimers. Perhaps coincidentally, a single synaptic vesicle contains 70 VAMP-2 molecules (27). Hence fusion sites harbor many more SNAREs than are directly needed for fusion of a secretory granule (4–8). It remains to be seen whether the supernumerary t-SNAREs are present just for safety (8, 27) or have another role.

In studying the interaction of plasma membrane proteins with single granules, we have used the granules for the spatial alignment of multiple images and have detected the presence of the protein near the granule in averages of aligned images. The method (which we call “location-guided averaging”) will be applicable to other proteins and other organelles. In Fig. 1D a single SNAP-25 molecule gave a clear image, and there are no reasons why, by averaging more images, one would fail to detect amounts representing an average of less than one molecule.

Methods

Cells, plasmids, light microscopy, and image analysis are described in ref. 2. Rat GFP-SNAP-25 was made from EYFP-SNAP-25b (a gift of J. W. Taraska, National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, MD) by replacing EYFP with GFP, using Age1/Not1 sites. The fluorescence F_cell of the cell footprint was measured by drawing an outline that enclosed all granules and copying it into the channel showing labeled SNAREs. Experiments in Figs. 1, 2, and 5 were done on live cells. Recording temperature was 28 °C. Results are given ± SEM. Significance was tested with Student's t test.

Cell Harvest and Western Blots.

Cells were plated at a density of 4 × 10⁶ in 6-cm dishes and, if necessary, were transfected with lipofectamine. After 48 h in the incubator, cells were harvested in Eppendorf tubes, pelleted at low speed, and resuspended in lysate buffer [100 mM KCl, 20 mM Hepes (pH 7.5), 1 mM EDTA-Na₂, 1 mM DTT, 1% Triton-X100, protease inhibitors tablet, Roche Diagnostics)]. They then were incubated on ice for 1–12 h and vortexed twice. Supernatants were obtained by microcentrifugation at 16,000 × g for 10 min at 4 °C. NuPAGE 4xLDS sample buffer (NP0007; Invitrogen) was added, and the supernatants were incubated for 10 min at 95 °C, run on SDS/PAGE, and transferred onto PVDF membranes (Invitrogen). Membranes were incubated with primary antibody and then with an infrared dye-conjugated secondary antibody (1:10,000 donkey anti-rabbit IgG), IRdye 800 (611–732-127; Rockland). Signals were detected and quantified with the Odyssey Infrared Imaging system (Li-COR Biosciences).

Single Molecules in Fixed Cells.

Coverslips were sonicated in 2% SDS for 10 min, then rinsed, soaked in bleach overnight, rinsed again, and stored in deionized water. Cells were cotransfected with NPY-mCherry and either rat syntaxin-1A-GFP or GFP-SNAP-25, fixed 20–30 h later in paraformaldehyde (2–3 mM) for 10 or 30 min at 25 °C, and then transferred to imaging buffer. Cells were located by the red fluorescence of their granules at 568 nm excitation, and a 10-s movie (20 Hz) at 488 nm and 568 nm excitation was collected. We selected the movies showing the very dimmest green fluorescence; that fluorescence was invariably punctate. A 240 × 240 pixel region was excised, and the Flatten Background routine (structure size, 10 pixels) was run. Frames 2–11 (or 2–101 in a second analysis) were averaged to increase the signal/noise ratio. Fluorescent spots probably representing single molecules were found with the Find Spot routine in Metamorph (thresholds were 25 CU for 10-frame averages and 5 CU for 100-frame averages). Onto each spot, Find Spot placed a circular region of the size also used for the analysis of clusters (seven-pixel diameter and 0.312-μm radius). The regions were accepted if the spots in them were round and appeared no larger than a 200-nm bead. If the centers of any two regions were closer than 0.62 μm, then both were deleted.

The remaining regions were superimposed on the movie, the average fluorescence intensity in each was read into an Excel worksheet using the Graph Intensity routine in Metamorph, and off-cell background was subtracted. The results were stored as “traces” and plotted against time; each spot generated one trace. Traces were deemed to show the bleaching of a single Syx-GFP molecule if an abrupt drop in fluorescence to background levels was accompanied by a visually clear decrease in noise. The amplitude of the fluorescence step, ΔC, was taken to represent the brightness of the molecule and was measured as the difference between the time averages of fluorescence before bleaching and during the first 30 frames thereafter. The time to bleaching (latency) was determined as well. When two downward steps occurred at the same location, only the second was analyzed. About 20–50% of spots in each cell were rejected because their fluorescence did not decline in steps or did not decline to background levels. We also rejected traces that bleached in <250 ms after the excitation light was turned on.

The excitation light (20 mW at 488 nm) used for single molecules was stronger than that use in other recordings (5 mW or less). To allow for this difference, cells were imaged in 50-ms exposures with powers alternating between 20 mW and 4.4 mW. At the lower power, cells were 3.7 ± 0.1-fold dimmer. By linear extrapolation, they would be 3.36 times dimmer at 5 mW than at 20 mW; this value was used to calculate row 2 in Table 1.

To test for autofluorescence, 100-frame averages were formed from cells expressing red granule marker but not Syx-GFP. Find Spots found 0.024 ± 0.007 spots/μm² in the footprint of 10 cells, 35 times fewer than in the nine cells chosen for analysis in Fig. 3 (0.78 ± 0.15 spots/μm²). Hence the spots did not result from autofluorescence.

Nonfluorescent GFP.

GFP molecules may be nonfluorescent because they bleached during a measurement. Cells coexpressing Syx-GFP and NPY-mCherry were imaged as averages of 1-s movies (50 Hz, 0.1 mJ per exposure at 488 nm). The Syx-GFP fluorescence of their “footprint” on the glass slide was plotted as a function of time. During the first second, the average fluorescence was 0. 82 ± 0. 01 times as large as the initial fluorescence (n = 8 cells).

Some GFP peptide chains may have failed to mature into a fluorescent form in our cells. Lysates from cells expressing GFP, Syx-GFP, or GFP-SNAP-25 were analyzed in a fluorimeter and assayed by quantitative Western blot, and results were compared (SI Text and Fig. S3). All GFP-SNAP-25 but only 66% of the Syx-GFP in our cells was fluorescent. With 0.1-mJ exposures at 488 nm, therefore, the average Syx-GFP peptide chain provided a fluorescence intensity of 0.82 × 0.66 × 10.4 CU, or 5.61 CU in the 0.31-μm² radius circle used to measure C and ΔF. For GFP-SNAP-25 the value was 0.82 × 9.8 CU = 8.0 CU per peptide chain (rows 5 and 6 in Table 1).

Calibration in Terms of Fluorescent Molecules.

Relative to the light collected by the camera from a cluster or molecule, what fraction is collected by our 0.312-μm radius circle? Syx-GFP images were aligned to granules as in figure 1C of ref. 2 and were collected in an average image. Then single molecules were identified by stepwise bleaching, spatially aligned, and collected in a second average image. Both images were well fitted (28) by 2D Gaussians, with SDs of 0.125 μm for clusters (730 images from 62 cells) and 0.114 μm for molecules (266 images from five cells). Within our circle, the two Gaussians contained 95.7% of the light from clusters and 97.7% from single molecules. The small difference was ignored (SI Text). Because the same size circle was used to analyze single molecules and clusters, a cluster x times as bright as a single fluorescent molecule contains x fluorescent molecules. Values for single fluorescent molecules are given in Table 1 (row 2).

For Syx-GFP, an average fluorescence intensity of 3.2 CU represents a density of one fluorescent molecule/μm², as follows. The total light collected from a single fluorescent Syx-GFP was 10.4 CU in each of the 39 pixels of our circle. Essentially the same light (maximally 1/0.977 times as much) would be collected in any larger area. Specifically, a 1-μm² area contains 126 pixels, and 39 of them each receive 10.4 CU, whereas the other 87 receive nothing. When distributed over 1 μm², therefore, the light from one fluorescent GFP molecule provides an average fluorescence per pixel of 10.4 × 39/126 = 3.2 CU. Correcting for immature or bleached t-SNAREs yielded the value given in row 6 of Table 1, with a corresponding value for GFP-SNAP-25.

t-SNARE Content of Cells.

Cultured PC12 cells were harvested and counted by hemocytometry. Lysis buffer and denaturation medium were added in volumes chosen such that 1 μL of denatured supernatant contained the extract of 25,000 cells. A sample of recombinant rat Syx-1A (kindly provided by T. Lang, LIMES Institute, University of Bonn, Germany) was assayed by amino acid analysis. As a standard, we used 329 fmole of syntaxin in 5 μL of 1xLDS buffer. Standard and cell supernatants were subjected to serial twofold dilutions. The resulting aliquots were probed by quantitative Western blot with four Syx–specific antibodies: Santa Cruz Biotechnology sc12736 (1:200), Sigma 1172 (1:4,000) and s0466 (1:2,000), and Synaptic Systems 110111 (1:2,000). Cell harvests from three cultures were processed independently with each of the four antibodies. Densities were plotted as functions of Syx amounts or cell numbers and fitted with straight lines through the origin. The slopes provided the density per amount of protein and the density per cell, and the ratio of slopes provided the amount of protein per cell. Recombinant SNAP-25b (MBS2030125; Biosource) was used similarly as a standard for SNAP-25. On a Coomassie blue stain, 95% of the density ran as a single peak at the molecular mass of SNAP-25. Aliquots containing 40 pmoles and cell supernatants were subjected to serial dilution and compared. SNAP-25 antibodies were from Santa Cruz Biotechnology (sc-7538; 1:200), Sigma (S-9684; 1:4,000) and Synaptic Systems (SY111002; 1:1,000).