Insulin granule exocytosis was evoked by a depolarizing solution containing high K⁺ (70 mM KCl and 20 mM glucose), and was continuously recorded in real-time for 10 min at a frame rate of 291 Hz using the TIRF-SIM microscope. The fusion events were characterized by a transient fluorescence increase at the center of the VAMP2-pHluorin molecule intensity, followed by a decay to baseline (Figures 1A–D).

Regarding the pHluorin-labeled events, we observed four categories of fusion modes according to their fluorescence intensity profile and diffusion pattern: ring fusion (full fusion with a “ring” structure, Figure 1A), dot fusion (full fusion with puncta, Figure 1B), kiss-and-run (K&R, Figure 1C) and kiss-and-stay (K&S, Figure 1D) according to their fluorescence intensity profile and diffusion pattern. In full fusion events (ring and dot fusion), corresponding to spreading or discharge, a robust fluorescence increase occurred both at the center and in the annular area of VAMP2-pHluorin molecules (Figures 1A,B,E), indicating the full collapse of a vesicle’s one-time diffusion to the plasma membrane and a dilated fusion pore. In contrast, confined events (kiss-and-run and kiss-and-stay) showed a brightening of the central intensity but no or a limited fluorescence increase in the annular area (Figures 1C–E), representing transient opening and reclosure of a restricted fusion pore that limits vesicle collapse.

Markedly, 80% of the total vesicle exocytotic events were full fusion (Figure 1F), including ring fusion (25%, FWHM ∼202 ± 27 nm in diameter, Figures 1F,G) and dot fusion (55%, FWHM ∼92 ± 19 nm in diameter, Figures 1F,G), when depolarizing stimulation using high K⁺ solution were applied to INS-1 cells. The percentages of kiss-and-run and kiss-and-stay events were 15 and 5%, respectively (Figure 1F). Collectively, the data indicate that after stimulation by membrane depolarization, full fusion and Kiss-and-run fusion are the dominant release modes in INS-1 cell exocytosis.

Next, we characterized the kinetics of single fusion events. A previous study showed that for the ring and dot fusions, with an improved frame rate of 291 Hz [via a “rolling” SIM reconstruction (Huang et al., 2018), four kinetically distinct steps could be dissected (Huang et al., 2018)]: (1) an initial rapid increase in the fluorescence intensity (fast rise, duration t₁), which appeared to represent the initial pore opening when vesicular H⁺ efflux occurred; (2) a slower increase (slow rise, duration t₂) due to the movement of the fusing vesicle toward the plasma membrane; and (3) a stage in which the fluorescence remained elevated (plateau phase, duration t₃) and during which the enlarged fusion pore formed; and (4) the final return of fluorescence to the baseline value (decay phase, the time constant τ), which refers to the final dilation of the vesicle (Figure 1A, higher magnification inset, Figure 5). Meanwhile, for kiss-and-run and kiss-and-stay events, the fusion of single vesicles exhibited only the fast rise phase t₁ and the decay phase (time constant τ); the slow-rise and plateau phases were absent (Figures 1C,D). The rising phase probably represents fusion pore opening, and the decay phase may depict the vesicle resealed and reacidification of the vesicle lumen at the kiss-and-run events (Figure 5).

Further, we found that the average t₁, t₃, and τ values for ring fusion were 1.5–3.5-fold longer than those for dot fusion events. Given the rate of change in fluorescence was the same in the two populations according to a previous study (Huang et al., 2018), more time is needed for the dequenching or diffusion of VAMP2 in vesicles undergoing ring events. Simultaneously, the mean peak intensity of ring events was 1.5-fold higher than that of dot events, suggesting that ring events involve vesicles with larger membrane surfaces (Figure 1H). Among the confined events, the rise times are similar between kiss-and-run (∼60 ms) and kiss-and-stay (∼64 ms) events (Figure 1I), but were markedly slower than full fusion events, in agreement with smaller fusion pores. In contrast, the vesicle collapse times of full fusion were quite similar (Figure 1J). In addition, ring events remained in the enlarged fusion pore intermediate phase (t₃) for more extended periods (Figure 1K). Finally, the decay time (τ) of kiss-and-stay events was 3.5-fold longer than that of kiss-and-run events, suggesting very slow diffusion of vesicular proteins on the plasma membrane (Figure 1L).

To better explore the mechanism by which chronic hyperglycemia impairs GSIS, we established a glucose toxicity model in INS-1 cells. Briefly, glucotoxicity conditions were created by exposing INS-1 cells to RPMI 1640 full culture medium containing 20 or 30 mM glucose for 120 h. The cells were then used to detect cell function under GSIS conditions. Using ELISA experiments to detect insulin secretion, we showed that (1) for INS-1 cells cultured in the presence of basal (5 mM) glucose, stimulation with 16.7 mM glucose produced a threefold enhancement of insulin secretion; and (2) with increasing concentrations of glucose in the medium, the secretory response to 16.7 mM glucose was reduced by 40–68% in cells cultured under high glucose (p < 0.01) (Supplementary Figure 1).

Next, to determine whether high glucose treatment inhibits vesicle synthesis, we counted the number of vesicles labeled with VAMP2-pHluorin after NH₄Cl incubation (50 mM, 2 min) under control (11 mM glucose) and hyperglycemic conditions (high glucose: 20 and 30 mM) (Figure 2A). Although diameters of granules in cells cultured in 20 and 30 mM glucose remained unchanged (Figure 2H and Supplementary Figure 3), hyperglycemia significantly reduced the number of secretory vesicles (Figure 2C). This result agreed with the reduced insulin granules immunofluorescently labeled in INS-1 cells cultured under elevated glucose concentrations (Supplementary Figure 2). Meanwhile, the hyperglycemic conditions severely reduced the fusion frequency (Figures 2B,D) and the probability of vesicle release (Figure 2E), confirming the reduced exocytosis in cells after long-term incubation in high glucose concentrations (Supplementary Figure 1).

Thus, we examined the fusion mode. Surprisingly, long-term exposure to high glucose concentrations reduced the full fusion percentage (Figure 2F, ring fusion: ∼25–5%; dot fusion: ∼55–30%) and triggered more kiss-and-run events (∼15–60%), while the percentage of kiss-and-stay events remained unchanged (∼5%) (Figure 2G). Regarding whether high glucose treatment altered the kinetics of the confined events, the results showed a slower rise time for kiss-and-run events, which reflected a longer brightening time for the pHluorin molecules and a smaller fusion pore (Figure 2I). In contrast, kiss-and-stay events showed more tolerance to glucotoxicity since the rise and decay times showed no significant differences compared to the events under control glucose concentrations (Figure 2J).

Next, we investigated spatiotemporal kinetics of full fusion events revealed by the Hessian SIM, and compared the fusion pore sizes at t₂ (vesicle collapse) and t₃ (enlarged pore formation) under normal (11 mM glucose) and long-term hyperglycemia (20 mM and 30 mM glucose, 120 h) conditions. The results showed that glucotoxicity reduced the fusion pore size (Figures 3A,B), from 198 ± 15 nm (Control, n = 30) to 153 ± 11 nm (20 mM, n = 25) and 120 ± 9 nm (30 mM, n = 35) at the t₂ stage, and from 230 ± 12 nm (Control, n = 30) to 168 ± 12 nm (20 mM, n = 30) and 132 ± 10 nm (30 mM, n = 20) at the t₃ stage. Correspondingly, long-term exposure to glucose also delayed ring fusions. A longer period of time was needed to observe the first appearance of pore structures (Figure 3C), along with delayed initial fusion pore opening (Figure 3D, t₁), vesicle collapse, and extended large pore formation (Figure 3D, t₂, t₃). The final dilation and diffusion of the vesicular membrane, however, was not affected. Similarly, t₁, t₂, and t₃ were all slowed in dot fusion events from cells been treated with high glucose for 120 h (Figure 3E).

To better explore the underlying mechanisms, we further determined whether glucotoxicity altered expressions of key regulators in vesicle fusion, SNARE proteins. Surprisingly, we found that core SNARE complex expression was significantly reduced at higher glucose concentrations (Figure 4A and Supplementary Figure 4). Correspondingly, knocking down syntaxin-1A and SNAP-25 in INS-1 cells decreased the fusion frequency (Figure 4B), triggered more kiss-and-run events (Figure 4C), and restricted the fusion pore size (Figure 4D), which were similar to the effects of hyperglycemia.

Next, we determined how the fusion pore dynamics changed when interfering with syntaxin-1A and SNAP-25. For ring fusion, decreased syntaxin-1A expression slowed early pore opening and inhibited vesicle collapse to the plasma membrane (Figure 4E), while decreased SNAP-25 expression slowed early pore opening, inhibited vesicle collapse, and hindered the enlarged pore formation process (Figure 4F). Similar results occurred in dot fusion (Supplementary Figures 5A,B). In addition, decreased SNARE expression also inhibited the initial pore opening of kiss-and-run events (Figure 4G) but had no significant effect on kiss-and-stay events (Supplementary Figure 5C), phenocopied the effects of long-term exposure to high glucose in INS-1 cells.

Hessian SIM microscopy enabled the direct visualization of the insulin fusion process and led us to the following conclusions:

To study the fusion pore dynamics, INS-1 cells were transfected with VAMP2-pHluorin using Lipofectamine 2000 reagent (Thermo Fisher Scientific, 11668019) and plated onto polylysine-coated coverslips. The experiments were conducted after 16–24 h of culturing in an incubator at 37°C. For the knockdown and rescue experiments, cells were transfected with BoNT/C α51, BoNT/E, syntaxin1A-EGFP, and SNAP-25-EGFP.

The rat insulinoma β-cell line INS-1 was cultured as previously described (Zhou and Misler, 1996). The cells were maintained at 37°C in a humidified incubator supplemented with 5% CO₂ and were subcultured twice per week. The cells were maintained in RPMI 1640 medium containing 11 mM D-glucose supplemented with 10% FBS (fetal bovine serum), 100 mg/ml penicillin-streptomycin, 10 mM HEPES, 1 mM sodium pyruvate, 2 mM L-glutamine, and 50 μM beta-mercaptoethanol (Invitrogen, Saint Aubin, France).

All experiments were performed at 37°C. INS-1 cells were plated on glass coverslips and incubated with bath solution containing the following (in mM): 136 NaCl, 4.2 KCl, 2.4 CaCl₂, 1.2 KH₂PO₄, 1.2 MgSO₄, 5 glucose, 10 HEPES, and 1 L-glutamine (pH 7.4). Individual coverslips were then placed in a metal chamber mounted on a heated stage. The INS-1 cells were stimulated with a solution containing the following (in mM): 70 NaCl, 70 KCl, 2.4 CaCl₂, 1.2 KH₂PO₄, 1.2 MgSO₄, 20 glucose, 10 HEPES, and 1 L-glutamine (pH 7.4) to trigger insulin granule exocytosis under the TIRF-SIM microscope.

Glucotoxicity conditions were obtained by exposing INS-1 cells to 20 mM and 30 mM glucose (high glucose) for 120 h. The control cells were INS-1 cells exposed to standard culture medium containing 11 mM glucose (Control) for 120 h. Each respective culture medium was changed after 48 h and replaced with the same culture medium until 120 h.

The following primary antibodies were used for immunofluorescence and western blot analysis: polyclonal guinea pig anti-insulin antibody (1:200, Dako, Carpentaria, CA, United States); mouse monoclonal anti-syntaxin-1A (1:1000, SySy); mouse monoclonal anti-SNAP-25 (1:1000, SySy); rabbit monoclonal anti-VAMP2 (1:1000, Abcam); rabbit monoclonal anti-tubulin (1:1000, Abcam). The secondary antibody was DyLight 488 goat anti-guinea pig IgG (1:500, Thermo); IRDye 800CW goat anti-rabbit IgG (1:1000, LI-COR Biosciences); and IRDye 680CW goat anti-mouse IgG (1:1000, LI-COR Biosciences).

INS-1 cells cultured in control or high glucose conditions were washed with 1 ml of KRBB buffer (125 mM NaCl, 5.9 mM KCl, 2.56 mM CaCl₂, 1.2 mM MgCl₂, 1 mM L-glutamine, 25 mM HEPES, and 1 g/L BSA) containing 5 mM glucose, preincubated for 1 h at 37°C and then transferred to KRBB containing 16.7 mM glucose for 20 min at 37°C. The incubation solution was then collected, cell lysates were prepared by incubating for 30 min in RIPA buffer at 4°C, and the insulin levels were quantified using a rat/mouse insulin ELISA kit according to the manufacturer’s instructions (EZRMI-13K, Millipore). Insulin secretion was normalized to the total insulin content determined from the cell lysates, and the values shown in Supplementary Figure 1 were detected by normalizing 4 × 10⁵ cells.

Total RNA was extracted from INS-1 cells using the RNeasy Mini Kit (74104, QIAGEN). First-strand complementary DNA was synthesized from total RNA using TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix (AT311-03, TransGenBiotech). Real-time PCR was performed on an Eppendorf RealPlex2 system using TransStart Top Green qPCR SuperMix (AQ131-03, TransGenBiotech). The qPCR primers used to detect VAMP2, syntaxin-1A, and SNAP-25 expression were VAMP2-F (5′-GGGAGTCTGGACTTTTGGGG-3′), VAMP2-R (5′-GAAACGGGGTAAGGGAAG-3′), syntaxin-1A-F (5′-CATGGACTCCAGCATCTCGAA-3′), syntaxin-1A-R (5′-TCCATGAACATGTCGTGCAGC-3′), SNAP-25-F (5′-TCGGGAACCTCCGTCAC-3′), and SNAP-25-R (5′-AATTCTGGTTTTGTTGGAATCAG-3′). RNA transcript levels were quantified using the 2^–ΔΔCt method.

For immunofluorescence labeling, cell samples were fixed with 4% paraformaldehyde in PBS for 15 min, followed by permeabilization in PBS containing 0.5% Triton X-100 (MERCK, Billerica, MA, United States) for 10 min and blocking in PBS containing 5% bovine serum for 60 min. The samples were incubated for 60 min in PBS containing primary antibodies and 2.5% bovine serum and then exposed to fluorescent dye-conjugated secondary antibodies for 60 min at 37°C. The cell samples were scanned with an Olympus IX81 (Olympus, Tokyo, Japan).

Insulin granules of INS-1 cells were labeled with anti-insulin antibodies. Besides, we performed the Z-series analysis of the whole-cell and calculated the averaged insulin vesicle number by normalizing per cell size. The insulin granule density in Supplementary Figure 2 was defined as the number of insulin puncta per cell volume (number/μm³). All the Immunofluorescence images were shown as maximum intensity projection throughout the cell and were determined by ImageJ (National Institutes of Health¹).

The cell samples were washed with PBS and homogenized on ice with lysate buffer [150 mM NaCl, 1% NP-40, 0.5% deoxycholic acid, 0.1% SDS, 50 mM Tris, pH 8.0, 1 mM PMSF, and 2% proteinase inhibitor (539134, Calbiochem)]. The homogenates were centrifuged at 16,000 g for 15 min at 4°C, and the centrifuged supernatant was adjusted to same total protein concentration following protein quantification by a standard BCA method. Proteins were electrophoresed and transferred to nitrocellulose filter membranes. Each membrane was blocked by incubation for 1 h with PBS containing 0.1% Tween-20 (v/v) and 5% non-fat dried milk (w/v). After washing with 0.1% Tween-20 containing PBS (PBST), the blots were incubated with primary antibodies at 4°C overnight in PBST containing 2% bovine serum albumin (BSA). Secondary antibodies were then applied at room temperature for 1 h. Blots were scanned with an Odyssey infrared imaging system (LI-COR Biosciences) and quantified with ImageJ. At least three independent western blots were conducted, and one typical blot is presented.

All confocal microscope images in Supplementary Figure 2 were generated using an inverted fluorescence microscope (Olympus, IX81) and a Laser Scanning Confocal system (Yokogawa, CSU-X1) with a 100×, 1.30 numerical aperture (NA) oil objective lens. The confocal settings used for image capture were held constant when samples were being compared. Images were quantified and analyzed using ImageJ software (National Institutes of Health).

TIRF-SIM imaging. The imaging system has been schematically illustrated (Huang et al., 2018). Briefly, it was based on a commercially available inverted fluorescence microscope (IX83, Olympus) equipped with a TIRF objective (Apo N 100×/1.7 HI Oil, Olympus), and images were captured by a sCMOS camera (Flash 4.0 V2, Hamamatsu, Japan). For TIRF-SIM, nine raw images are required to reconstruct one super-resolved image. Reaching a frame rate of 873 Hz, the temporal resolution for the non-overlapping reconstruction of SR images was 97 Hz. Because the three images in one orientation were independent of the images in the other two orientations, we proposed the reconstruction of time-lapse SIM images using overlapping raw image sequences (“rolling”) (Huang et al., 2018). By implementing this algorithm, we further increased the temporal resolvability to 291 Hz. All the images were obtained at a frame rate of 291 Hz except for Figure 2A and Supplementary Figure 2.

A 488-nm laser was used to excite GFP/pHluorin. To visualize VAMP2-pHluorin-labeled granule numbers (Figure 2A), the TIRF illumination was adjusted to 2D-SIM. Each pixel from the camera was 32.5 nm × 32.5 nm. The release probability in Figure 2E was defined to describe the chances of vesicles to fuse to the plasma membrane, which can be calculated as (number of fusion events) / (granule number per cell).

Image processing was primarily performed using ImageJ software. The reconstruction of the images acquired by TIRF-SIM or 2D-SIM was accomplished using customized MATLAB software (2016a). The data analysis and curve fitting were performed using Igor Pro software (6.11, WaveMetrics, Lake Oswego, OR, United States) and GraphPad 7.0 software. To analyze different stages of vesicle fusion, we determined the time interval between the initiation of the fast rise in fluorescence intensity (the intersection of the stable linear fit at the bottom and the fitted fast rise) to the initiation of slow rise (the intersection of the fitted fast rise slope and the fitted slow rise) as the t₁; the time interval between the initiation of the slow rise in fluorescence intensity to the time point of the initiation of stable fluorescence (the intersection of the fitted slow rise and the stable linear fit at the top) as the t₂; the time interval between the initiation of stable fluorescence to the time point of the initiation of fluorescence decay (the intersection of the stable linear fit at the top and the fitted exponential decay) as the t₃.

The results are presented as the mean ± SEM. Statistical significance was evaluated using Student’s t-test for single Gaussian-distributed datasets or the Mann–Whitney rank sum test for non-single Gaussian-distributed datasets. The asterisks ^∗, ^∗∗, and ^∗∗∗ denote statistical significance at p < 0.05, p < 0.01, and p < 0.001, respectively.