Membrane lipids couple synaptotagmin to SNARE-mediated granule fusion in insulin-secreting cells

Abstract

Insulin secretion depends on the Ca²⁺-regulated fusion of granules with the plasma membrane. A recent model of Ca²⁺-triggered exocytosis in secretory cells proposes that lipids in the plasma membrane couple the calcium sensor Syt1 to the membrane fusion machinery (Kiessling et al., 2018). Specifically, Ca²⁺-mediated binding of Syt1’s C2 domains to the cell membrane shifts the membrane-anchored SNARE syntaxin-1a to a more fusogenic conformation, straightening its juxtamembrane linker. To test this model in live cells and extend it to insulin secretion, we enriched INS1 cells with a panel of lipids with different acyl chain compositions. Fluorescence lifetime measurements demonstrate that cells with more disordered membranes show an increase in fusion efficiency, and vice versa. Experiments with granules purified from INS1 cells and recombinant SNARE proteins reconstituted in supported membranes confirmed that lipid acyl chain composition determines SNARE conformation and that lipid disordering correlates with increased fusion. Addition of Syt1’s C2AB domains significantly decreased lipid order in target membranes and increased SNARE-mediated fusion probability. Strikingly, Syt’s action on both fusion and lipid order could be partially bypassed by artificially increasing unsaturated phosphatidylserines in the target membrane. Thus, plasma membrane lipids actively participate in coupling Ca²⁺/synaptotagmin-sensing to the SNARE fusion machinery in cells.

• The calcium-binding protein synaptotagmin is known to couple calcium signaling to exocytosis in secretory cells, including in insulin secretion from pancreatic β cells. However, the molecular mechanism of this coupling is still debated.

• This work shows that the calcium-dependent translocation of synaptotagmin C2 domains to membrane lipids alters the properties of the membrane, which in turn alters the conformation of SNAREs in insulin-secreting cells and in a hybrid reconstituted system.

• The results imply that dietary, genetic, or pharmacologic factors that change lipid behavior and composition in β cells contribute to the etiology of diabetes.

INTRODUCTION

Professional secretory cells mediate signaling to other cells through regulated exocytosis of neurotransmitters and hormones. For example, pancreatic beta cells regulate circulating glucose levels in the blood by releasing insulin, whereas neurons release neurotransmitters across the synapse to control action potentials in the recipient cell. In regulated exocytosis, cells encapsulate their molecular cargo destined for secretion within vesicles originating at the trans Golgi network, and release the content by fusing the vesicle membrane with the plasma membrane (PM) upon a stimulating signal. These signals transiently raise intracellular calcium, which triggers a molecular machinery to facilitate membrane fusion. The components and features of these machineries are common among many types of secretory cells (Jahn and Fasshauer, 2012; Sudhof, 2013). The energy for vesicle fusion is provided by the SNARE proteins (Weber et al., 1998), whereby the four SNARE motifs in the t-SNAREs (e.g., syntaxin-1a and SNAP-25) and the v-SNARE (e.g., synaptobrevin-2, VAMP-2) zipper to form a coiled-coil four helical bundle (Sutton et al., 1998). In models for SNARE-mediated fusion, it is assumed that the SNAREs initially interact between the membranes in a flexible prefusion trans complex that straightens upon fusion into a rigid postfusion cis complex with continuous helices between the SNARE motifs and transmembrane domains (Kiessling and Tamm, 2003; Stein et al., 2009; Jahn and Fasshauer, 2012).

SNARE proteins alone can mediate membrane fusion without calcium in reconstituted fusion systems (Weber et al., 1998); however, secretion in vivo occurs in a highly regulated, Ca²⁺-dependent manner and requires other regulatory proteins and lipids. To mimic the physiological fusion process experimentally, a minimal set of regulatory components are required to assemble a prefusion complex that inhibits Ca²⁺-independent fusion and enhances Ca²⁺ sensitivity (Malsam et al., 2012; Ma et al., 2013; Lai et al., 2014; Kreutzberger et al., 2017). This minimal set of components is comprised of the calcium-sensitive membrane protein synaptotagmin-1 (Syt1) (Brose et al., 1992; Tucker et al., 2004; Wang et al., 2011; Kiessling et al., 2013), Munc18/SM proteins (Dulubova et al., 1999; Ma et al., 2013; Kreutzberger et al., 2017; Jiao et al., 2018), MUN domain−containing proteins like Munc13 or CAPS (Ann et al., 1997; Varoqueaux et al., 2002; Ma et al., 2013; Kreutzberger et al., 2019), the soluble protein complexin (Tang et al., 2006; Giraudo et al., 2008; Kreutzberger et al., 2017), as well as the anionic lipids phosphatidylserine (PS) and phosphatidylinositol-4,5-bisphosphate (PI(4,5)P₂) in the PM.

While it is known that Syt family proteins are the key molecular players in responding to the influx of calcium upon depolarization, the exact molecular interactions underlying Syt-triggered fusion remain elusive (Brunger et al., 2018; MacDougall et al., 2018; Brose et al., 2019; Holz and Zimmerberg, 2019). Syt1 binds Ca²⁺ and phospholipids via its C2A and C2B domains (Davletov and Sudhof, 1993). There are two proposed mechanisms for the interactions between SNAREs and Syt1’s C2 domains: 1) Syt1 acts directly with SNARE proteins to stimulate fusion, as suggested by a crystal structure of C2 domains with the postfusion SNARE complex (Ernst and Brunger, 2003; Zhou et al., 2015), and 2) C2−lipid interactions are the primary fusion triggering interactions under physiological condition (Park et al., 2015; Perez-Lara et al., 2016). Consistent with the importance of C2−lipid interactions, we recently proposed a mechanism where lipid order links the Ca²⁺ stimulated membrane binding of Syt1 to SNARE protein−mediated fusion (Lai et al., 2011; Kiessling et al., 2018). In this model, Ca²⁺-mediated membrane binding of Syt1’s C2 domains induces disorder into the membrane, which in turn changes SNARE proteins from a bent, trans- to a straight cis- conformation, providing force that pulls the vesicle in close apposition to the PM and induces fusion between the two membranes.

To further test this model in live cells and extend it to insulin secretion, we investigated the fusion of insulin granules from a rat pancreatic beta cell line (INS1 cells), both in live cells and in a well-defined hybrid reconstituted fusion system with purified insulin granules and the relevant lipid and protein components reconstituted in supported membranes (Kreutzberger et al., 2019; Kreutzberger et al., 2020). The measurements of exocytosis in INS1 cells and single vesicle fusion in the reconstituted system strongly support the notion that disordering of the lipid components in the PM, measured by the fluorescence lifetime of the order-sensitive FlipperTR probe (Colom et al., 2018), changes the orientation of the SNAREs from a nonfusogenic to a fusogenic conformation. Importantly, membrane engagement of the C2 domains of Syt1 causes the same disordering of the lipids in the target membrane and the same fusion promoting conformational changes in the embedded SNAREs as an artificial supplementation of the target membrane with unsaturated lipids, providing strong support for a model in which Ca²⁺/synaptotagmin controls the fusion of secretory vesicles through changes in the lipid phase of the PM.

RESULTS

Changing lipid order in INS1 cell plasma membranes with lipid-loaded methyl-β-cyclodextrin

To test how membrane order of the PM may affect fusion in living cells, we first adapted a method to efficiently enrich the PM with lipids of various acyl chain compositions, utilizing methyl-β-cyclodextrin (MβCD) (Figure 1A) (Kainu et al., 2010). Liposomes composed of chol:PC:PS (2:1:1) were solubilized by MβCD and used to enrich the PM with lipids composed of either dipalmitoyl (referred to as “DP,” chol:DPPC:DPPS, 2:1:1) or dioleoyl (referred to as “DO,” chol:DOPC:DOPS, 2:1:1) acyl chains. Palmitoyl-oleoyl (referred to as “PO,” chol:POPC:POPS, 2:1:1) was used as a control that approximates the acyl chain composition of the native PM (Hicks et al., 2006; Lorent et al., 2020).

FIGURE 1:

Enrichment of INS1 cells with lipids. (A) INS1 cells were treated with MβCD loaded with lipids (PC, PS, and cholesterol) composed of either DP, PO, or DO acyl chains and then imaged after treatment. (B and C) Fluorescence lifetime imaging of INS1 cells stained with the FlipperTR probe. Images show the lifetime component tau₂ that was obtained from fitting a 2-exponential decay to the raw data (B), with quantification of the lifetime (tau₂) based on photons collected from the PM (C). (D and E) Imaging of cholesterol in fixed INS1 cells using Filipin III stain after treatment with exogenous lipids delivered by MβCD (D), with quantification of average Filipin III intensity in the PM of fixed INS1 cells (E). *** indicates p < 0.001, **** p < 0.0001. Each data point represents a single cell. Scale bars indicate 5 μm.

Changes in the membrane’s physical properties upon lipid exchange with MβCD were quantified using a recently developed push-pull probe, FlipperTR (Colom et al., 2018). The fluorescence lifetime of FlipperTR responds to changes in the lipid acyl chain packing which is directly related to lipid acyl chain order. A decrease in lifetime indicates a less well packed and more disordered acyl chain environment. Cells were subjected to a panel of lipid treatments and the lifetime of FlipperTR was determined by a double-exponential fit to the time course of the total photon count (Figure 1B). We found that the fluorescence lifetime of FlipperTR decreases in cells enriched with PO and DO lipids (Figure 1B,C), while cells treated with DP lipids did not show a significant effect on the fluorescence lifetime of FlipperTR. Empty MβCD showed only a mild decrease of FlipperTR lifetime, while treatment with MβCD loaded with cholesterol did not have a significant effect (Supplemental Figure S1A).

Because MβCD is commonly used to deplete cholesterol, we reasoned that the employed MβCD-mediated exchange of phospholipids might also change the cholesterol concentration in the PM. To quantify PM cholesterol content, we stained fixed cells with Filipin and measured the fluorescence intensity at the PM as a readout of its cholesterol content. Filipin is a fluorescent polyene macrolide antibiotic that binds to sterols and is commonly used to track cholesterol distribution in cells (Sezgin et al., 2016; Chaudhuri and Anand, 2017). Sensitivity of Filipin staining was confirmed with INS1 cells treated with empty MβCD and MβCD loaded with cholesterol (Supplemental Figure S1B). Indeed, Filipin staining (Figure 1, D and E) shows that the DP and PO lipid enrichment procedure also significantly depleted the membrane of cholesterol, although less than empty MβCD (Supplemental Figure S1B), while the DO treatment did not affect the cholesterol content (Figure 1D). Thus, for DP and PO lipids, both phospholipid enrichment and cholesterol depletion contribute to changes in the observed fluorescence lifetime of FlipperTR, while for DO lipids the observed effect can be primarily attributed to the enrichment of DO lipids (Figure 1B). None of the treatments resulted in PMs with an increase in membrane order relative to untreated cells. However, relative to the control treatment with PO lipids, DP lipid treatment results in increased order and DO lipid treatment results in decreased order.

Enrichment of the plasma membrane with DP acyl chains inhibits stimulated insulin vesicle fusion, while enrichment with DO acyl chains has an opposite effect

To measure the effects of lipid treatment on secretion of proinsulin-C-peptide-GFP from INS1-derived GRINCH cells, total internal reflection fluorescence (TIRF) microscopy was used to record fusion of insulin granules (vesicles) after membrane depolarization with 90 mM potassium which causes an intracellular increase in calcium (Bendahmane et al., 2018; Kreutzberger et al., 2020) (Figure 2, A and E). Vesicles that underwent fusion after stimulation were counted. The various lipid treatments did not affect the number of docked granules per observed area of plasma membrane (Supplemental Figure S2A). Relative to controls (see “Untreated” and “PO” lanes), cells enriched with DP lipids showed a significant decrease in fusion probability, while cells enriched with DO lipids showed a significant increase in fusion probability relative to both untreated cells and control cells treated with PO lipids (Figure 2B). Vesicle fusion probability in PO-treated control cells was similar to the fusion probability in untreated cells (Figure 2B), suggesting that the measured variation in cholesterol (Figure 1E) did not significantly impact this aspect of exocytosis. We also compared the number of fusion events occurring over time. The insulin granule fusion kinetics are essentially the same after DP and DO lipid treatment and may have marginally slowed in PO-enriched cells (Supplemental Figure S2B).

FIGURE 2:

Lipid acyl chain order regulates insulin granule fusion in INS1 cells. Quantification of GRINCH cells after treatment with DP, PO, or DO lipids. (A and B) TIRF imaging of C-peptide-GFP INS1 cells after stimulation with 90 mM KCl to stimulate fusion of insulin vesicles, with examples of fusion circled (A), with quantification of fusion probability (number of granules which fused divided by total number of granules) (B). (C and D) C-peptide-GFP INS1 cells expressing the R-GECO calcium indicator fluorescent protein before, 5s after, and 50s after stimulation with 90 mM KCl (C), with quantification of the peak value of signal from the fluorescent R-GECO calcium indicator of live GRINCH cells stimulated with 90 mM KCl (D) shown in median box plots. The peak ΔF/F values (D) were not significantly different between treatments. (E) Examples of characteristic C-peptide-GFP intensity versus time graphs used to identify granules as non-release or release events. (F) Example of the calcium indicator R-GECO normalized fluorescence over time, where KCl stimulation occurs at 10 s. **** indicates p<0.0001. Scale bars indicate 5 μm.

Enrichment with lipids does not alter calcium influx upon depolarization

High potassium depolarizes the membrane potential and increases intracellular Ca²⁺ via pathways regulated by membrane lipids (Roberts-Crowley et al., 2009). Thus, it is possible that the effects of different lipid treatments on membrane fusion may be secondary to alterations in Ca²⁺ influx or handling. To address this possibility, intracellular Ca²⁺ levels were monitored in TIRF using the genetically encoded Ca²⁺ indicator, R-GECO (Zhao et al., 2011). Representative fluorescence intensity versus time records in cells stimulated by 90 mM KCl and exposed to different lipid treatments are shown in Figure 2C and F. The max ΔF/F, calculated from the time courses of cytosolic Ca²⁺ signals, was not significantly different between treatments (Figure 2D). The total R-GECO response over time after KCl stimulation essentially did not change either in response to the different lipid treatments (Supplemental Figure S2C). This suggests that alterations in Ca²⁺ influx are unlikely to be responsible for the observed effects on fusion probability.

Reconstituted fusion of insulin vesicles with biomimetic membranes depends on the acyl chain composition of target membranes

Biochemical in vitro model membrane assays allow the isolation of the membrane fusion reaction with defined lipid and protein compositions, as well as the application of biophysical methods that are not applicable to live cells. To directly demonstrate the effects of the acyl chain order in the target membrane on the fusion probability of insulin granules, we used a reconstituted single vesicle fusion assay with supported membranes and GRINCH cell−derived insulin granules. In this assay, recombinant Syx1a and dSNAP25 (Kreutzberger et al., 2016) are reconstituted into a well-defined lipid environment (Domanska et al., 2009) in the supported PM surrogate. Fusion of insulin granules purified from GRINCH cells is detected by following the fluorescence of C-peptide-GFP with a TIRF microscope (Figure 3A) (Kreutzberger et al., 2019). Using purified dense core vesicles (DCVs) from the secretory PC12 cell line derived from the rat adrenal gland, we showed previously that a minimal system consisting of SNARE proteins in the target membrane and soluble C2AB domains of Syt1 reproduces Ca²⁺ dependence of fusion (Kreutzberger et al., 2017; Kiessling et al., 2018). Addition of Munc18 and complexin-1 (Cpx1) proteins efficiently suppresses fusion before arrival of Ca²⁺.

FIGURE 3:

Phosphatidylserine acyl chain order controls fusion of purified insulin granules with reconstituted supported target membranes and SNARE conformation in these membranes. (A−C) Fusion of purified insulin vesicles with supported membranes harboring syntaxin-1a and dSNAP25 measured with TIRF microscopy. (A) Cartoon of the experimental setup. (B) Example of a single granule docking and fusion event. (C) Fusion probability of granules with target membranes with different lipid compositions. Only the PS component of the PM lipid mix was changed as indicated. Red bars indicate 100 μM EDTA buffer, white bars indicate 100 μM CaCl₂ buffer. (D−F) Measurement of Syx*192 (Alexa546) distance from the membrane surface by FLIC microscopy. (D) Cartoon of experimental setup. (E) Example of fit to an experimental FLIC data set. (F) Distances of Syx residue 192 from the surface of target membranes containing Syx/SNAP25/Syb(1-96) complexes with different lipid compositions. Only the PS component of the PM lipid mix was changed as indicated. Red bars indicate 100 μM EDTA buffer, white bars indicate 100 μM CaCl₂ buffer and 0.4 μM Syt C2AB domains. (G−I) Lipid order measured in supported membranes measured by FLIM using the FlipperTR probe. (G) Cartoon of experimental setup and lipid order and tension sensitive FlipperTR dye. (H) example images and fits of FLIM decay curve before and after adding C2AB and Ca²⁺, where scalebars indicate 25 μm. (I) Fluorescence lifetimes of FlipperTR probe in supported membranes with different lipid compositions. Only the PS component of the PM lipid mix was changed as indicated. Red bars indicate 50 μM EDTA buffer, white bars indicate 100 μM CaCl₂ buffer and 0.5 μM Syt C2AB. * indicates p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Each data point represents a single supported membrane.

To mimic the PM, supported membranes were prepared from porcine brain−derived phospholipids with a headgroup composition consisting of PC, PE, PS, and PI(4,5)P₂ (34:30:15:1) and 20 mol% cholesterol, referred to as “brain PM” or bPM. The acyl chain composition was altered by replacing the 15% brain-derived PS component in the vesicle facing leaflet of the supported membrane with equivalent amounts of either DPPS, DOPS, or POPS. Purified insulin granules were added either in presence of 100 μM EDTA or 100 μM Ca²⁺ and movies of GFP fluorescence were recorded. Docking and fusion of individual vesicles were distinguished by analyzing the characteristic fluorescent signal from regions containing single vesicles (Figure 3B) (Kreutzberger et al., 2019). With brain PM as the target bilayer, 28% of vesicles fused in the absence of Ca²⁺, while 40% fused in 100 μM Ca²⁺ (Figure 3C). The fusion efficiencies increased to 34% and 49%, respectively, when brain PS was replaced with POPS. PO lipids approximate the major components of the PM’s acyl chain composition of primarily monounsaturated and saturated acyl chains. Thus, in both of these “control” conditions, we observed Ca²⁺-stimulated fusion. In contrast, when brain PS was replaced with either saturated DPPS or doubly unsaturated DOPS, Ca²⁺ sensitivity was lost. With DPPS, the fusion efficiencies with and without Ca²⁺ were 24% and 25%, respectively, while in the DOPS case, the fusion efficiencies with and without Ca²⁺ were 32% and 36%, respectively. These results demonstrate that Ca²⁺-dependent regulation of insulin vesicle fusion requires a target membrane composition where acyl chains can be modulated: the fusion stimulating effect of Ca²⁺ via the endogenous synaptotagmins embedded in the insulin granule membrane can only be observed when PS contains mixed (saturated and unsaturated) acyl chains.

Conformation of the SNARE complex depends on the acyl chain composition of supported membranes

We next aimed to understand how the effect of lipid order on insulin granule fusion may relate to SNARE protein conformation and its alteration by synaptotagmin. It was previously shown that fusion of DCVs with SNARE containing supported membranes of different lipids with and without Ca²⁺ correlates with the activity of Syx1a. The likelihood or fraction of Syx1a being in a highly fusogenic state was quantified by measuring the average distance of the N-terminus of the SNARE motif within the ternary SNARE complex (Syx1a residue 192) from the membrane surface (Kiessling et al., 2018). These distances were measured previously by site-directed fluorescence interference contrast (sdFLIC) microscopy (Lambacher and Fromherz, 2002; Kiessling and Tamm, 2003; Liang et al., 2013) as illustrated in Figure 3, D and E (Kiessling et al., 2018), under identical lipid conditions as the fusion experiments in Figure 3C. Measurements were taken in the absence and presence of Ca²⁺ and Syt1’s C2AB domains. When brain PS and POPS were included in the supported membrane, addition of Ca²⁺ and C2AB triggers a conformational change of the SNARE complex to a more upright orientation (Figure 3F). The conformational change was impaired when brain PS was replaced with either DPPS or DOPS (Figure 3F). This behavior exactly parallels the Ca²⁺-mediated fusion behavior observed under the same lipid conditions described in Figure 3C, showing that SNARE conformation responds to changes in acyl chains in the membrane or addition of C2AB similarly to fusion probability.

Binding of Syt1 C2AB domains decreases lipid order in ordered supported lipid bilayers, but not in those containing the disordered DOPS species

Because our data showed that synaptotagmin/calcium’s effect on conformation and fusion depends on the bilayer’s acyl chain composition, we then reasoned that synaptotagmin might induce acyl chain disorder as part of its calcium triggering activity in this system (Lai et al., 2011). We therefore sought to measure lipid order and its potential distortion by C2AB domains with the FlipperTR dye (Figure 3G).

Supported membranes (without proteins) were prepared on cover slips with the same lipid compositions as in the fusion and FLIC assays. Fluorescence lifetimes of FlipperTR in supported membranes were significantly lower for all lipid compositions than the values that were measured in cell membranes (Figure 3H,I). This might be due to the close proximity of the fluorophore to the buffer/glass interface when it is in supported bilayers (Hellen and Axelrod, 1987; Kiessling and Tamm, 2003) and/or composition differences between cell membranes and supported bilayers, including the presence of membrane proteins and glycocalyx at relatively high densities in cell membranes and a PM cholesterol content higher than the 20% used in supported bilayers here (van Meer et al., 2008). However, FlipperTR reported the expected effects of lipid composition on acyl chain order: membranes prepared with DPPS had the longest fluorescence lifetime, DOPS the shortest, and POPS and brain PS were intermediate (Figure 3I).

To compare effects on lipid order with its effect on SNARE conformation and calcium-triggered fusion, purified C2AB domains and Ca²⁺ were added to each supported lipid bilayer composition (Figure 3, H and I). Ca²⁺ alone did not have a significant effect on the lifetime of FlipperTR in lipid bilayers (Supplemental Figure S3A). The amount of fluorescently labeled C2AB domains that bound to lipid bilayers of different lipid compositions did not significantly depend on the lipid composition, except that C2AB had a marginally higher affinity to bilayers containing DOPS than DPPS (Supplemental Figure S3B). C2AB/Ca²⁺ decreased FlipperTR lifetime in brain PM bilayers, suggesting decreased lipid order (Figure 3I). Substituting brain PS with DPPS or POPS in bPM bilayers also decreases the lifetime upon addition of Ca²⁺ and C2AB, although in the case of DPPS, the lifetime was not decreased below the values of the other lipid compositions. Replacing brain PS with DOPS results in the lowest measured fluorescence lifetimes and no further decrease upon addition of Ca²⁺ and C2AB domains. The results of Figure 3I therefore show that changes in the lipid order of supported membranes correlate with the ability of the SNAREs to straighten their conformation (Figure 3F) and support fusion (Figure 3C).

DISCUSSION

In this work, we present evidence that the order of membrane lipids in the PM regulates exocytosis and insulin secretion from rat pancreatic beta cells (INS1 cells). INS1 cells serve as a clinically relevant cellular model of regulated exocytosis and membrane fusion, in which insulin granules are arrested in a primed, calcium-sensitive state at the PM (Sudhof, 2013; Thurmond and Gaisano, 2020). Using this system, we studied the role of bulk membrane lipids to test the model that the actions of Syt on the fusion activity of SNARE proteins are mediated through the membrane, specifically through modulation of lipid acyl chain order. We show that triggering of exocytosis by intracellular Ca²⁺ includes a disordering of lipids in the PM, which is initiated by the interaction of the Ca²⁺ sensor Syt with acidic lipids in the cell membrane.

The proposed steps of this model are depicted in Figure 4: the SNAREs syntaxin and synaptobrevin that are anchored in the two respective membranes are flexibly connected to their transmembrane domains after early N-terminal SNARE assembly steps are completed and before fusion is initiated, that is, when the system is “primed” for fusion. This conformation, a trans SNARE complex, is stabilized by accessory proteins such as Munc18, Munc13, and complexin. When intracellular calcium rises, Ca²⁺ binds to the C2 domains of Syt, leading to insertion of its calcium-binding loops into both membranes (C2A into the granule and C2B into the PM, respectively (Nyenhuis et al., 2019)), which bends the membranes and disorders the lipids in them. The disordering of the lipids in turn changes the linker region of the SNAREs, which becomes stiffer and connects the SNARE motifs and transmembrane domains of the two proteins in a continuous helix. This disordered-to-helical transition exerts force on the transmembrane domains pushing the two membranes into close apposition, ultimately forcing them to fuse.

FIGURE 4:

A model for synaptotagmin-triggered fusion of insulin granules. Proposed steps involved in calcium-mediated fusion, with the interplay of syntaxin’s linker region and synaptotagmin’s C2B domain with the bilayer highlighted in the inset. Prior to calcium entering the cell, the N-terminal end of the SNARE complex resides at an intermediate distance from the membrane surface due to a flexible linker. When calcium rapidly enters the cell, synaptotagmin’s C2 domains introduce acyl chain disorder in the membrane, which straightens the SNARE linker region, thereby promoting fusion by transducing force.

Previous studies have tested the effect of free fatty acid treatments, such as palmitic acid, as a model metabolite in cells that mimics the insulin secretion characteristics of type-2 diabetes (Cen et al., 2016; Kreutzberger et al., 2020). In contrast, here we were interested in the contribution of the bulk phospholipid acyl chain composition in the PM to the SNARE-mediated membrane fusion reaction. Unsaturated lipids (containing dioleoyl acyl chains) increased the fusion probability of insulin granules with cellular PMs and supported model membranes. This result is in agreement with the previously reported effects of acyl chain composition in purified PC12 DCV fusion (Kiessling et al., 2018).

In this study, we quantified the altered membrane order by measuring the fluorescence lifetime of the push-pull dye FlipperTR in the PM of INS1 cells and in reconstituted membranes, which is sensitive to membrane tension and acyl-chain composition in cell and model membranes. Changes in cell membrane order after lipid enrichment correlate well with observed fusion efficiencies: DO-lipid enrichment produced the lowest membrane order and highest fusion efficiency, while DP-lipid enrichment produced the highest membrane order and lowest fusion efficiency. We did not observe an increase of fluorescence lifetime upon DP lipid treatment, whereas both PO and DO lipid treatments resulted in decreased fluorescence lifetimes. Some of these effects may be attributable to a decreased cholesterol content caused by the MβCD treatment, which may counter some of the effects of the saturated acyl-chains on the FlipperTR lifetime but not on fusion probability. We additionally cannot rule out that changes in the extracellular leaflet by MβCD’s interaction with cholesterol and sphingomyelin (Zidovetzki and Levitan, 2007; Suresh and London, 2022) might influence some of our observations of fusion. However, our measurements of cholesterol and the preservation of calcium influx as well as the inclusion of PM-mimicking PO lipids as a control suggest major off-target effects from lipid delivery are unlikely. Despite these caveats, the experiments suggest that acyl chain order in the PM acts as an important modulator of insulin release.

To specifically isolate the effect of lipid order, we relied on a biochemical reconstituted fusion assay where the lipid composition can be precisely controlled. In these experiments, only the PS component was replaced with synthetic DPPS, POPS, or DOPS in the t-SNARE containing target membranes (i.e. only 7.5 mol% of the total lipid content). Strikingly, the fusion probabilities mirrored those measured in live cells and similarly depended strongly on acyl chain order. While the inhibition of fusion with DPPS is not as strong as when all lipids have DP acyl chains (Kiessling et al., 2018), calcium-dependent fusion probability is strongly impaired with DPPS. Conversely, replacing bPS with DOPS in the target membrane increased fusion without Ca²⁺ close to fusion observed with Ca²⁺, that is, eliminated the exquisite control by Ca²⁺ observable with brain lipids in the reconstituted system and with the unperturbed natural membrane in INS1 cells. It is interesting that the loss of dynamic range of Ca²⁺ control is not observed upon replacement with POPS, which has only one cis double bond in the sn-2 chain and matches most closely the natural distribution of unsaturation found in the lipids of cell membranes.

This dynamic range becomes more evident when the fusion probability of insulin granules is plotted as a function of the fraction of SNAREs that are standing upright as determined from the FLIC measurements (Supplemental Figure S3C) (Kiessling et al., 2018). From the more upright orientation of the syntaxin anchored SNARE complex, resembling the helically extended cis-SNARE complex (Stein et al., 2009), in more disordered membranes, we conclude that the increased fusion probability is a result of the linker between the SNARE motif and transmembrane domain being shifted to a more rigid fusogenic straight conformation compared with a more flexible bent conformation in more ordered membranes. This transition from the bent trans-SNARE complex to an upright cis-SNARE complex may therefore be interpreted as an “activation” of the SNARE complex.

We also found that C2AB of Syt disorders the lipids in supported bilayers, which is in agreement with a previous study using Fourier transform infrared spectroscopy to measure lipid order (Lai et al., 2011). This lipid disordering corresponds to the increased fraction of Syx in the upright “activated” state (Supplemental Figure S3D), where the distance of 12 nm from the crystal structure of the helically extended cis SNARE complex may be considered as 100% of all SNAREs in the active state and smaller distances as smaller fractions of SNAREs transitioning to the fully active state. Similarly, the plot of fusion probability as a function of FlipperTR lifetimes (Supplemental Figure S3E) also suggests that decreased membrane order increases fusion. This explains the range of fusion efficiency with similar fractions of SNAREs in the active state in the absence of Ca²⁺ (Supplemental Figure S3C). However, FlipperTR has also been used frequently to report on changes in membrane tension (Colom et al., 2018). We confirmed that in supported lipid bilayers, which lack tension from any attached proteins, the probe responds as expected simply from lipid acyl chain packing (Figure 3I). This sensitivity of FlipperTR to lipid order is consistent with previous reports (Colom et al., 2018; Ward et al., 2023). While we interpret the fluorescence lifetime changes to quantify order changes in the lipid bilayer, other physical changes of the bilayer measured by the FlipperTR probe such as water penetration or lipid dynamics could conceivably also cause the conformational changes in the linker region of SNAREs and activate them to become more fusogenic. For example, it is possible that the juxtamembrane region of Syx increases its interaction with the hydrophobic core of the lipid bilayer, thereby inducing the helical extension from the SNARE motif observed in the crystal structure of the cis-SNARE complex (Stein et al., 2009). Finally, we note that FlipperTR lifetimes are shorter in protein-free model membranes than in cell membranes which has been observed by others before (Colom et al., 2018). This is not surpising because cell membranes are much more densily packed with proteins, which surely affects lipid packing and mobility. Therefore, we only compare relative lifetime differences under changed lipid conditions in each system.

Both cell and biochemical experiments show that the lipid environment of the PM are an intrinsic part of the molecular release machinery during regulated exocytosis. The two experimental systems also reveal important differences, apart from differences in quantification of exogeneous lipid incorporation. In cells the different lipid treatments result in a steady increase of release probability from more ordered (DP) to more disordered (DO) lipids, while in the reconstituted single vesicle fusion assay, DP and DO lipids result in a reduced dynamic range of Ca²⁺ stimulation compared with brain and PO lipids. The major difference between the two assays is that in the in vitro assay we purposely isolated the Ca²⁺ stimulated fusion reaction without the influence of preceding docking, SNARE assembly, and priming steps that involve regulatory proteins such as Munc18, complexin, and Munc13 (Jahn and Fasshauer, 2012; Sudhof, 2013; Kiessling et al., 2018), where the calcium-dependent increase on purified insulin granule fusion with SNARE-containing bilayers is almost entirely due to the effect of endogenous synaptotagmins embedded in the granule membrane (Kreutzberger et al., 2019). The isolated Syt/SNARE machinery contrasts with cells, where the granule is arrested and primed at the PM prior to its calcium-dependent release. The reduced effect of Ca²⁺ in the minimal Syt/SNARE in vitro experiments therefore likely reflects impaired action of Syt, whereas in cells the fusion competency following depolarization might also originate from effects on upstream fusion inhibition, priming, and earlier steps in the secretory pathway such as sorting. Based on the results presented here and known interactions of other proteins (e.g., complexin [Zdanowicz et al., 2017] or Munc13 [Ma et al., 2011]) with the lipid bilayer, we conclude that the lipid environment including its acyl chain composition are integral parts of the machinery that controls vesicle docking and fusion. Beyond the C2AB−lipid−SNARE interactions focused on here, more efficient inhibition and priming before Ca²⁺ influx through a local reorganization of the lipid bilayer by a higher availability of unsaturated lipid species could contribute to the higher observed insulin release probabilities.

Our results are broadly consistent with known influences of lipid order on viral membrane fusion (White et al., 2023). However, in this current case, we find that SNARE conformation is controlled by lipid disorder, which subsequently provides the force for fusion. In viral fusion, most emphasis has been on how lipid order may change the energy landscape for fusion, for example by changing line tension between ordered and disordered lipid domains providing the energy for HIV-mediated fusion (Yang et al., 2016) or changing membrane bending rigidity in the formation of fusion intermediates (Chernomordik and Kozlov, 2008; Ward et al., 2023).

It has been questioned in the literature whether altered membrane lipid composition is a cause or effect in diabetes (Pilon, 2016), where insulin secretion from pancreatic beta cells is impaired in type 2 diabetes and at the onset of type 1 diabetes (Steele et al., 2004; Guillausseau et al., 2008). For example, erythrocytes from diabetic patients have been reported to have a less fluid membrane, but the causal meaning of this observation is unclear (Kamada et al., 1992; Pilon, 2016). Our data indicate that lipid composition in the insulin releasing cells themselves is important and needs to be tightly controlled to properly regulate insulin release from these cells.

In summary, our data show that lipids are intimately involved in the core fusion machinery of secretory cells. The combination of cellular and biochemical reconstitution studies allowed us to achieve a more complete picture of what function acyl chain composition of the membrane plays in a pancreatic β-cell−derived cell line and very likely also other secretory cells including neurons. To be functional and achieve the full control of regulation by Ca²⁺, the lipid acyl chain composition must be highly regulated. The right balance of fusion-promoting and fusion-inhibiting acyl chains must be achieved in the vicinity of the fusion site so that SNAREs can switch from a conformation with a flexible linker between the SNARE motif and transmembrane domain to an activated conformation with a rigid linker between these two protein domains. The latter SNARE conformation is promoted by decreasing the lipid acyl chain order in the membrane, which can be achieved by the Ca²⁺-dependent binding of synaptotagmin’s C2AB domains to lipids in the vicinity of the SNARE linker. Therefore, Ca²⁺ activation of the exocytic fusion machinery occurs through a lipid mediated activation of SNAREs by the Ca²⁺ sensor Syt. It would be informative and potentially lead to better treatments of diabetes to further study how membrane lipid order—and its potential alterations by dietary, genetic, and pharmacological factors—relates to diabetic etiology in primary beta cells in their native islet environment.

MATERIALS AND METHODS

Materials

For the cellular experiments, 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC), 1,2-dipalmitoyl-sn-glycero-3-phospho-l-serine (DPPS), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC),1,2-dioleoyl-sn-glycero-3-phospho-L-serine (DOPS) were obtained from Avanti Polar Lipids; methyl-beta-cyclodextrin, cholesterol, glucose, CaCl₂, HEPES, beta-mercaptoethanol from Sigma. KCl from EMD chemicals; FBS, PBS (no Ca, no Mg), Roswell Park Memorial Institute (RPMI) media, sodium pyruvate, penicillin/streptomycin, puromycin, and G418 from Gibco.

For the reconstitution experiments, the following additional materials were purchased and used without further purification: porcine brain L-α-phosphatidylcholine (bPC), porcine brain L-α-phosphatidylethanolamine (bPE), porcine brain L-α-phosphatidyl­serine (bPS), porcine brain phosphatidylinositol-4,5-bisphosphate (PI(4,5)P₂), 1,2-dipalmitoyl-sn-glycero-3-phosphol-serine (DPPS), 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-l-serine (POPS); DMPE-PEG2000-triethoxysilan (DPS) was custom-synthesized by Shearwater Polymers (Huntsville, AL); cholesterol, sodium cholate, EDTA, calcium, Opti-Prep Density Gradient Medium, sucrose, MOPS, glutamic acid potassium salt monohydrate, potassium acetate, magnesium (Mg²⁺) salt, and glycerol were from Sigma; CHAPS and DPC were from Anatrace; HEPES was from Research Products International; chloroform, ethanol, Contrad detergent, all inorganic acids and bases, and hydrogen peroxide were from Thermo Fisher Scientific. Water was purified first with deionizing and organic-free 3 filters (Virginia Water Systems) and then with a NANOpure system from Barnstead to achieve a resistivity of 18.2 MΩ cm⁻¹.

Treatment of cells with lipid-loaded MβCD

INS1 cells (for FLIM or cholesterol staining) and INS1 cells stably expressing human proinsulin-C-peptide-GFP (also called GRINCH cells; for TIRF micro­scopy) were cultured at 37°C, 5% CO₂ in RPMI1640 culture medium (10% FBS, 10 mM HEPES, 1 mM sodium pyruvate, 50 μM beta-mercaptoethanol, 1x penicillin/streptomycin; with G418 with proinsulin-C-peptide-GFP cells), as described previously (Haataja et al., 2013; Hussain et al., 2018). Medium was changed every 2 d and cells were passed at 70−80% confluency. For imaging, cells were passed onto 35-mm dishes and imaged within 1-2 d of passage.

To treat cells with MβCD, a previously described method was adapted (Kainu et al., 2010). To prepare liposomes, the desired lipids in chloroform were first mixed in a 12 × 75 mm culture tube. Chloroform was evaporated using a stream of nitrogen gas and subsequent vacuum desiccation for at least an hour, yielding a film of lipids. Lipid films were rehydrated using PBS (no Ca, no Mg) to reach a concentration of 12 mM lipids. After freeze−thaw cycles, liposomes were extruded through a polycarbonate membrane with a 50-nm pore (Avanti). On the day of imaging, a stock of 32 mM MβCD was prepared in phenol red-free RPMI growth media. The liposomes and MβCD were mixed and warmed for 20−30 min at 37°C. The cells were washed with warmed phenol-red free culture medium 2−3 times before adding lipid-loaded MβCD. Lipid-loaded MβCD was added to the cells to reach a final concentration of 8 mM MβCD, 3 mM lipids; in the case of the untreated condition, PBS and RPMI (no MβCD, no lipids) were used in place of lipid-MβCD. Cells were kept warm at 37°C, 5% CO₂ for 2 h while treated with lipid-MβCD. After the 2-h treatment period, cells were washed three times with warmed PBS to remove any remaining exogenous lipid-MβCD complexes. PBS was replaced with phenol red−free RPMI culture medium before proceeding to imaging.

Cholesterol stain

To quantify cholesterol content of the PM, a commercially available, cell-based Filipin III kit was used (Cayman Chemicals). After treatment with MβCD described above within 35-mm imaging dishes, cells were immediately fixed and the commercial staining protocol was used as described, except scaled to 35-mm dishes rather than 96 well plates. The cells were stained for 30 min before washing and imaging with confocal microscopy on an Zeiss LSM 880 microscope with a 63x objective.

Fluorescence lifetime microscopy of cells

INS1 cells were imaged in phenol red−free RPMI culture medium after treatment with lipid-MβCD (described above). Cells were stained with 1 μM FlipperTR (Colom et al., 2018) and immediately imaged. A 63x objective was used, 480 nm excitation, 533-741 nm emission on a Leica FLIM SP8 microscope. Frame repetitions (typically 4) were used to assure sufficient photon counts for fitting of the data. For each cell, the PM was selected with an ROI and fit to a two-exponential decay and the longer lifetime tau₂ was used to report lipid packing (Colom et al., 2018). A single datapoint within Figure 1C represents the fit result for the tau₂ of one cell’s PM.

TIRF imaging of cellular C-peptide-GFP secretion

Following treatment with lipid-MβCD, GRINCH cells were prepared for imaging with TIRF microscopy. Cells were first chased with RPMI culture medium for 30 min in a 37°C, 5% CO₂ incubator. Cells were then incubated with a glucose-starved RPMI medium (1x penicillin/streptomycin, 50 μM BME) for 20−40 min. To track secretion in real-time, previously described methods to measure secretion with TIRF microscopy were followed using an Olympus cellTIRF-4Line microscope as previously described (Bendahmane et al., 2020; Kreutzberger et al., 2020). Briefly, culture media was replaced with physiological salt solution (15 mM HEPES, 145 mM NaCl, 5.6 mM KCl, 0.5 mM MgCl₂, 2 mM CaCl₂, 3 mM glucose, pH 7.4). To stimulate secretion, a needle was placed next to the field of view where a cell can be seen. When recording began, physiological salt solution was perfused for 10 s, followed by an immediate perfusion of 90 mM KCl stimulating salt solution (15 mM HEPES, 50 mM NaCl, 90 mM KCl, 0.5 mM MgCl₂, 5 mM CaCl₂, 3 mM glucose, pH 7.4) for typically 60−70 s.

Fusion probability was analyzed as previously described (Kreutzberger et al., 2020). Granules containing fluorescent C-peptide-GFP that were present in the evanescent field at the start of the recording were considered to be “docked”, whereas those that entered the field after the start are newcomers. The number of fusing granules (including both docked granules and newcomers) after KCl depolarization was divided by the total number of granules to obtain a measure of “fusion probability.” Fusion events were identified by an increase in GFP fluorescence followed by a lateral spread of the fluorescence (Rao et al., 2017; Kreutzberger et al., 2020).

For the analysis of the fusion kinetics in Supplemental Figure S2B, Igor Pro (WaveMetrics, Inc.) was used to fit a first order kinetics curve to the normalized cumulative distribution of fusion events following KCl perfusion. The manual activation of the KCl perfusion results in a small uncertainty of the exact moment of stimulation of a few seconds.

Calcium imaging

A commercially available kit was used to transduce GRINCH cells with the R-GECO fluorescent calcium indicator protein (Montana Molecular). The protocol for stimulating cells with 90 mM KCl was followed as described for the secretion experiments above and recorded with TIRF microscopy. The internal area of each cell after stimulation was traced. Background was subtracted from the fluorescence over time graph by fitting each curve to an exponential decay function, excluding the peak; after fitting the curve to an exponential decay, ΔF/F was calculated at each timepoint in the curve.

Protein purification for reconstitution experiments

Syntaxin-1a (1-288 full length construct) and SNAP-25 from Rattus norvegicus were expressed in Escherichia coli strain BL21(DE3) cells (New England Biolabs, Ipswich, MA) under the control of the T7 promoter in the pET28a expression vector and purified as described previously using Ni-NTA affinity chromatography and removal of the N-terminal His tags by thrombin cleavage (Liang et al., 2013). SNAP-25 was quadruply dodecylated through disulfide bonding with dodecyl methanethiosulfonate (Toronto Research Company, Toronto, Ontario) to its four native cysteines to obtain dSNAP25 (Kreutzberger et al., 2016).

Expression and purification of Syt1 C2AB (residues 136–421) from Rattus norvegicus was described previously (Kiessling et al., 2018; Nyenhuis et al., 2021). Briefly, C2AB was expressed and purified using a pGEX-KG vector with an N-terminal GST-tag. C2AB was expressed in BL21(DE3) cells (Invitrogen, Grand Island, NY), grown in LB media, with induction using 0.1 mM IPTG at an OD₆₀₀ of 0.8–1.0. The protein was purified using a GST affinity column and subsequent thrombin cleavage followed by ion exchange chromatography. Purity was verified by SDS Page and UV-Vis to ensure that protein was free of nucleic acid contaminant.

For labeled C2AB used in the TIRF binding assay (Supplemental Figure S2B), the native cysteine, residue 277, was mutated to an alanine and the site of interest, T329C, was mutated to a cysteine using site directed mutagenesis. C2AB T329C was reacted with two-fold molar excess of Alexa-546. Labeled proteins were separated from free dye using a PD10 column (Millipore Sigma, Burlington, MA) and subsequent dialysis in physiological buffer (20 mM HEPES, 150 mM KCL, pH 7.4). Protein was concentrated using an Amicon Ultra-15 10 K concentrator (Millipore Sigma, Burlington, MA).

Insulin granule purification from INS1 cells

Insulin granules were purified from INS1 cells stably expressing human proinsulin-C-peptide-GFP (Haataja et al., 2013). The purification protocol was exactly as described previously (Kreutzberger et al., 2017; Kreutzberger et al., 2019). Fraction 9 (14.5%/30% iodixanol interface) of the step gradient centrifugation was harvested for our experiments because it is highly enriched in insulin granules that are either mostly or fully mature. Less dense organelles that minimally sediment through the 14.5% iodixanol layer and are prominent in Fraction 3 include condensing vacuoles and earliest immature granules and vesicles derived from the ER, Golgi, and trans Golgi network where proinsulin-GFP is much more enriched relative to cleaved C-peptide-GFP, as well as synaptic-like microvesicles and endosomes.

Reconstitution of SNAREs into proteoliposomes

For the single-insulin vesicle fusion assay, SNARE proteins were reconstituted in sodium cholate, as described previously (Domanska et al., 2009). For bPM-supported membranes, bPC:bPE:bPS:bPI(4,5)P₂:cholesterol (34:30:15:1:20) lipids were used for the proteoliposomes (with bPS substituted with different acyl chains as indicated in text). The desired lipids were mixed, chloroform was evaporated using a stream of nitrogen, and was placed under vacuum desiccation for at least an hour. The dried lipid films were dissolved in 25 mM sodium cholate in buffer (20 mM HEPES, 150 mM KCl, pH 7.4) followed by the addition of an appropriate volume of the desired SNARE protein(s) in their respective detergents to reach a final protein/lipid ratio of 1:3000 syntaxin, 1:1500 dSNAP25 (Kreutzberger et al., 2016). After 30 min of equilibration at room temperature, the mixture was diluted to reach a sodium cholate concentration of 16 mM, close to the critical micellar concentration, by adding more buffer to the desired final volume. The sample was then dialyzed against 1 liter of buffer with one buffer change after ∼2 h and subsequent dialysis overnight.

Preparation of planar-supported bilayers containing acceptor t-SNARE complexes

Planar-supported bilayers with reconstituted PM SNAREs were prepared by the Langmuir–Blodgett vesicle fusion technique as described in previous studies (Kalb et al., 1992; Wagner and Tamm, 2000, 2001). Quartz slides (Quartz Scientific, Fairport Harbor, OH) were cleaned by dipping in 3:1 sulfuric acid:hydrogen peroxide for 15 min using a Teflon holder. Slides were then rinsed thoroughly in water. The first leaflet of the bilayer was prepared by Langumir–Blodgett transfer directly onto the quartz slide using a Nima 611 Langmuir–Blodgett trough (Nima) by applying a lipid mixture of 77:20:3 PC:cholesterol:DPS from a chloroform solution using PC acyl chains to match the conditions indicated in the text (e.g. brain PC for bPM mix in the outer leaflet). After allowing the solvent to evaporate for 10 min, the monolayer was compressed at a rate of 10 cm²/min to reach a surface pressure of 31 mN/m. After equilibration for 5−10 min, a clean quartz slide was rapidly (68 mm/min) dipped into the trough and slowly (5 mm/min) withdrawn, while a computer maintained a constant surface pressure and monitored the transfer of lipids with headgroups onto the hydrophilic substrate. Proteoliposomes were incubated with the Langmuir–Blodgett monolayer to form the outer leaflet of the planar supported membrane. A concentration of 0.1 mM total lipid was used. After incubation of the proteoliposomes for 1 h, excess proteoliposomes were removed by perfusion with 5−10 ml of buffer (120 mM potassium glutamate, 20 mM potassium acetate, 20 mM HEPES, pH 7.4, for fusion experiments). During the experiments, buffers were exchanged with buffers that contained either 100 μM CaCl₂ or 100 μM EDTA in addition to the above compositions.

TIRF microscopy to measure reconstituted single vesicle fusion

Experiments examining single-vesicle docking and fusion events were performed on a Zeiss Axiovert 35 fluorescence microscope (Carl Zeiss), with a 63 × water immersion objective (Zeiss, numerical aperture, 0.95) and prism-based TIRF illumination. The light source was an OBIS 532 LS laser or an OBIS 488 LS laser from Coherent Inc. Fluorescence was observed through a 610 nm band pass filter (D610/60, Chroma) by an electron multiplying charge coupled device (CCD) (DU-860E, Andor Technology). The electron multiplying CCD (EMCCD) was cooled to −70°C, and the gain was set at 200. The prism-quartz interface was lubricated with glycerol to allow easy translocation of the sample cell on the microscope stage. The beam was totally internally reflected at an angle of 72° from the surface normal, resulting in an evanescent wave that decays exponentially with a characteristic penetration depth of ∼100 nm. An elliptical area of 250 μm × 65 μm was illuminated. The laser intensity, shutter, and camera were controlled by a homemade program written in LabVIEW (National Instruments).

Single vesicle fusion assay

The assay to measure the docking and fusion of single vesicles to planar-supported membranes has been previously described (Kreutzberger et al., 2019; Kreutzberger et al., 2017; Domanska et al., 2009). Planar supported membranes containing acceptor t-SNARE complexes were washed with buffer (120 mM potassium glutamate, 20 mM potassium sulfate, 20 mM HEPES, pH 7.4) containing 100 μM EDTA or 100 μM CaCl₂ as indicated. Purified secretory vesicles (insulin granules) were diluted into 500 μl of buffer and injected into the microscope flow chamber containing the planar-supported membrane. The fluorescence of the C-peptide-GFP containing vesicles was recorded with a 488 nm laser. After injection, the microscope was focused within 30s before images were taken at 200 ms exposure times for insulin granules. All experiments were performed at room temperature.

Single-vesicle fusion data were analyzed using a homemade program written in LabVIEW (National Instruments). Stacks of images were filtered by a moving average filter. The maximum intensity for each pixel over the whole stack was projected on a single image. Vesicles were located in this image by a single-particle detection algorithm described previously (Kiessling et al., 2006). The central pixel and mean fluorescence intensities of a 5-pixel by 5-pixel area around each identified center of mass were plotted as a function of time for all particles in the image series. The exact time points of docking and fusion were determined from the central pixel similar as in previous work (Domanska et al., 2009). Numbers of docking and fusion events were counted for each membrane. The fusion probability was determined by the number of vesicles that underwent fusion compared with the total number of observed vesicles.

Fluorescence lifetime microscopy of supported membranes

Direct fusion (Kalb et al., 1992) was used instead of the combined Langmuir−Blodgett/vesicle fusion method to prepare supported bilayers. Attempts at polymer-supported, asymmetric bilayers with FlipperTR led to heterogeneous structures that were not suitable for FLIM, while direct fusion lead to a more homogeneous plane.

To form the supported bilayers, bPM-1%FlipperTR liposomes (bPC:bPE:bPS:bPI(4,5)P₂:cholesterol:FlipperTR, 33:30:15:1:20:1) were prepared by drying down FlipperTR in DMSO, redissolving FlipperTR in chloroform, adding lipids in a chloroform solution, and drying down to produce a lipid film. The same sodium cholate detergent concentrations and dialysis were used for the liposomes as described above for SNARE proteoliposomes, except in this case SNAREs were not included in the bPM-FlipperTR liposomes. On the day of imaging, coverslips were cleaned with piranha solution and washed with water. 0.1 mM liposomes were added to the coverslip for 1 h to produce a symmetric supported lipid bilayer via the direct vesicle fusion method (Kalb et al., 1992). Excess liposomes were removed by washing with a 50 μM EDTA reconstitution buffer and imaged with FLIM. Afterward, the bilayer was washed with 100 μM Ca²⁺, and then 100 μM Ca²⁺ with purified 0.5 μM Syt1 C2AB. C2AB was incubated for 15−20 min before imaging.

To image supported bilayers with FLIM, a similar protocol as for FLIM imaging of cells was used using the Leica SP8 FLIM microscope. Samples were excited at 488 nm and emission was collected in the 533-741 nm range. To acquire sufficient photon counts, typically 4 frame repetitions were applied. The entire field of view was used to fit a 2-exponential decay and tau₂ was recorded. For each bilayer preparation, typically at least 3 different areas of the bilayer were imaged and tau₂ was averaged for each data point.

Binding of Synaptotagmin-1 C2 Domains to Supported Membranes

C2-domain binding experiments by TIRF microscopy were conducted under the same membrane and buffer conditions as the FLIC and single vesicle fusion experiments in the presence of 100 μM Ca²⁺. One milliliter of a 0.4 μM Alexa Fluor 546-labeled Syt1 C2 domain solution was injected into the flow-through chamber containing a supported membrane and the fluorescence was collected through a 40 × water immersion objective (Zeiss, numerical aperture, 0.75) under TIRF illumination by a OBIS 532 LS laser on the same set-up as described above for the single vesicle fusion experiments. The binding process was monitored for 20 min by taking images every 30 s and storing the mean intensity from all 128 × 128 pixels. First-order kinetics were fitted to the fluorescence intensities and the resulting saturation value was used to quantify binding under the chosen condition.

Statistical analysis

GraphPad PRISM was used for statistical analysis. Where statistical significance is mentioned in the text, an unpaired Student t test was used, unless otherwise indicated. Paired t tests were used for the FLIC (Figure 3F) and FLIM (Figure 3I) results before and after treatments with calcium and C2AB. In all bar graphs, mean ± SEM is indicated, unless otherwise indicated.

Abbreviations used:

bPM

plasma membrane-mimicking supported bilayer, formed with porcine brain lipids

bPS

porcine brain−derived phosphatidylserine

C2AB

C2A and C2B domains of synaptotagmin-1

Chol

cholesterol

DO

phospholipid with dioleoyl acyl chains

DOPS

dioleolyl phosphatidylserine

DP

phospholipid with dipalmitoyl acyl chains

DPPS

dipalmitoyl phosphatidylserine

dSNAP25

dodecylated synaptosome associated protein 25 (SNAP25)

EDTA

ethylenediaminetetraacetic acid

FLIM

fluorescence lifetime imaging microscopy

GRINCH cells

INS1 cells stably expressing insulin C-peptide-GFP

INS1 cells

rat insulinoma cells

MβCD

methyl-beta-cyclodextrin

PC

phosphatidylcholine

PE

phosphatidylethanolamine

PM

plasma membrane

PO

phospholipid with palmitoyloleoyl acyl chains

POPS

palmitoyloleoyl phosphatidylserine

PS

phosphatidylserine