Membrane tension increases fusion efficiency of model membranes in the presence of SNAREs

The large gap in time scales between membrane fusion occurring in biological systems during neurotransmitter release and fusion observed between model membranes has provoked speculations over a large number of possible factors that might explain this discrepancy. One possible reason is an elevated lateral membrane tension present in the presynaptic membrane. We investigated the tension-dependency of fusion using model membranes equipped with a minimal fusion machinery consisting of syntaxin 1, synaptobrevin and SNAP 25. Two different strategies were realized; one based on supported bilayers and the other one employing sessile giant liposomes. In the first approach, isolated patches of planar bilayers derived from giant unilamellar vesicles containing syntaxin 1 and preassembled SNAP 25 (ΔN-complex) were deposited on a dilatable PDMS sheet. In a second approach, lateral membrane tension was controlled through the adhesion of intact giant unilamellar vesicles on a functionalized surface. In both approaches fusion efficiency increases considerably with lateral tension and we identified a threshold tension of 3.4 mN m−1, at which the number of fusion events is increased substantially.

| Cross-sectional view of the substrate. A) Scheme of the PDMS stretching device. B) A cross-sectional view shows the channel system and spanned PDMS layer on top. In the centre, a big channel is visible that is connected to the surrounding air. The free standing PDMS sheet is around 180 µm thick. The channels to the left and to the right are referred to as the side chambers. Here, vacuum can be applied to stretch the PDMS sheet. The outer channels are the connection to the syringe pump system (see A)).

A B
1 µm Section 2: Dilatation of supported lipid bilayers on hydrophilic PDMS surface Supported lipid bilayers (SLBs) on hydrophilic surfaces are produced by spreading of vesicles. SUVs and LUVs normally form bilayers that fully cover the surface of the substrate if the appropriate vesicle concentration in the solution is used.
As depicted in Figure S1, the membrane-covered PDMS sheet was stretched to reach a defined area dilatation of the surface. Fluorescent beads incorporated into the thin PDMS sheet were used to document the local area dilatation. Figure S2 shows the dilatation of a membrane fully covering the PDMS surface. Green fluorescent beads were incorporated into the thin PDMS sheet. Applying lower pressure to the adjacent side channels of the PDMS layer leads to dilatation of the SLB firmly attached to the PDMS sheet. Prior to stretching, the surface is completely covered with a membrane (red fluorescence) without displaying any wrinkles or cracks. Figure S2 B shows the surface after applying vacuum to the side channels. The stretching of the PDMS surface results in a surface dilatation of 4.1%, which is slightly above the maximal area increase for this lipid composition of 3.6%. Therefore, it was possible to observe small cracks in the membrane. The image in figure S2 C shows a ruptured membrane on the stretched PDMS-support. Vertical cracks are aligned in parallel. At this stage, the surface dilatation was around 7.6%, which is even beyond the predicted area-increase at lysis tension. Releasing the surface dilatation by increasing the air pressure in the side channels, yields back a relaxed membrane. In the left part of the picture in figure S2 D small membrane tubes emerge indicating that the SLB area is now compressed. Tube formation can be explained by a slightly compressed surface. The membrane tube formation also confirms membrane fluidity because the lipids in the membrane are mobile and can form tubes through compression.
This documentation of dilating SLBs on the fabricated PDMS surface of the membrane stretcher device shows that it is possible to dilate membranes between 0 -7.6% of their initial area, which can be measured by fluorescence microscopy. Cracks and defects in the SLB occurred at around 4% area increase and could be clearly identified in the fluorescent images.

Figure S2 | Dilatation of PDMS surface fully covered with a lipid bilayer.
A) The SLB (red) covers the entire PDMS surface. Green fluorescent beads were incorporated into the thin PDMS sheet below the surface to measure the dilatation. B) The uniaxial dilatation between the beads at the left and right was around 4% and the SLB on the surface started to rupture. C) Vertical cracks in the SLB occur because of the high area dilatation of around 7.6%. D) Relaxing of the PDMS surface led to compression of the SLB and formation of membrane tubes visible at the left side of the image. E) Diffusion constants of a lipid bilayer deposited on stretchable PDMS as a function of strain e.

A B
C D E Passivation of PDMS surface LUV fusion to SLBs and spreading of the LUVs on the PDMS substrate was tested with a sample without any further passivation of the substrate surface around the membrane patches and in the absence of SNARE proteins. Figure S3 shows two images representing the two fluorescently labeled lipids, A594 in the SLBs (A) and A488 (green) in the LUVs (B). The right image reveals that LUVs are everywhere but on the membrane patches documenting that it is necessary to passivate the surface to prevent unwanted LUV adsorption and subsequent fusion at the edges of the SLBs as well as spreading of the LUVs onto the hydrophilic PDMS surface. As a consequence, the protein BSA was used to protect the hydrophilic PDMS surface.

Figure S3 | Control sample without passivation. A) Spread
GUVs on the PDMS surface that from SLBs (red). B) Addition of LUVs (green) in the absence of SNAREs. LUV adsorb on the uncovered surface and even might fuse with SLBs. Cracks in the PDMS surface also promote LUV fusion to the SLBs.

A B
6 Control experiments with passivated surfaces in the absence of SNAREs Figure S4 shows the three different control experiments ruling out non-specific fusion. In figure S4 A an experiment is shown where the membrane patches on the stretched PDMS surface lack SNARE proteins but the added LUVs are equipped with the ΔN49-complex, composed of syntaxin and SNAP25. Even after stretching of the PDMS surface (2 mL vacuum) no membrane fusion or docking of the LUVs to the membrane patches could be observed after more than 30 minutes of incubation.
In figure S4 B, a control sample without any area change of the membrane patches is shown. The membranes were equipped with SNARE proteins on both membrane sites, GUV (ΔN49) and LUV (Syb). This control sample shows no increase in fluorescence intensity originating from LUV fusion after 30 minutes of incubation time documenting that fusion requires tension. In rare cases, larger LUVs fuse with the membrane patch.
In figure S4 C two membrane patches (blue) are shown after incubation of LUVs (red) in the absence of SNARE proteins in the sample. Additionally, the PDMS surface was dilated (2 mL vacuum) but no LUV fusion on membrane patches could be detected. The corresponding fluorescence intensity area or line scans for the SLB dye A390 and LUV dye A594 are shown on the right-hand side.
The data shows that passivation of the hydrophilic PDMS surface around the membrane patches with BSA was successful to prevent undefined fusion events and that SNAREs in combination with pre-stressed membranes are necessary to trigger fusion. shown in figure S6-S8. A mean relative LUV dye intensity increase of A594 was measured on each membrane patch by thresholding of the channel for A390 and selecting the corresponding red ROI for the membrane patch.
This ROI was used in the red channel (A594) to provide the mean intensity I LUV, A594 on the membrane patch. The fluorescence intensity I max of the dye A594 on an adhered LUV on the PDMS surface was taken as a reference.
The fusion efficiency F eff was calculated from the channel of the dye A594 for every membrane in all images pictures that were taken after the LUV incubation. In figure S6 (left image) the membrane patches at the border emit more fluorescence since these patches were not within the previously scanned area and thus were not subject to photobleaching as much as those in the central region. The image on the right-hand side shows the fluorescence intensity of the LUV dye A594. Many LUVs adhere non-specifically to the PDMS surface. Also in this image, some membrane patches appear brighter in the red channel of dye A594 indicative of fusion with the membrane patches.

Figure S7 | Experiment 3:
A very large membrane patch showing an area change of (1.4±0.1)% after stretching the underlying PDMS substrate provoked a large amount of docked and fused LUVs (red dye A594). Docked vesicles appear as bright red spots, while fused LUVs generate a homogeneous fluorescence in the membrane patch. Figure S7 shows a single large membrane patch with a size of 6246 µm 2 after stretching of the PDMS sheet and after addition of LUVs. For this membrane patch an area change of (1.4±0.1)% was found leading to a fusion efficiency of (5.8±0.1)%.

Figure S8 | Experiment 4:
A) Two unstressed membrane patches. B) Stretching of the PDMS substrate led to a strong increase in area of ROI 1 but only a small area-change of ROI 2 was observed. Upon stretching small holes appear in the supported bilayers. C) Addition of LUVs (A594) to the pre-stressed membrane patches. Fusion of LUVs with the membrane patches was predominately found with ROI 1. D) Fluorescence image of the red channel (A594) after incubation with LUVs. The LUVs showed a higher affinity to the stretched ROI 1. E) The measured membrane area changes and relative fluorescence intensities I LUV = F eff are shown. Scale bar: 20 µm.
In figure S8 two membrane patches were shown that exhibit a different membrane tension after PDMS dilatation.
Holes in the membrane were subtracted prior to area determination. The membrane patch labelled as ROI 1 exhibits an area increase of (1.6±0.3)% after stretching of the PDMS substrate and shows a much higher fusion efficiency F eff = (4.7±0.3)% compared to the membrane patch labelled ROI 2 (ΔA/A 0 = (0.4±0.3)%, F eff = (1.0±0.1)%). Docking, hemi-fusion and fusion of LUVs occurs more frequently on ROI 1.

FRAP measurements of PDMS-based SLBs
Experimentally, a high-energy LASER pulse was used to bleach a region of interest (ROI) on the supported lipid bilayer with the actual bleaching area radius r. The diffusion coefficient D can be obtained from equation (I), 1 where r is the bleaching radius of the LASER and t 1/2 the half-time recovery for the intensity: In figures S9-S10 FRAP measurements performed on a SLBs were shown. It was ensured that indeed LUV fusion occurred by bleaching the lipid dye A594 to monitor the mobility of the newly inserted dye. A fast recovery occurred on the SLBs with only a small immobile fraction (15%) and a diffusion coefficient of (0.9±0.1) µm 2 s -1 . Figure

A B C D
Bleaching of the edges of the SLBs and the LUVs that adhered to the PDMS surface ( Figure S11) proved that LUVs adhere to the PDMS but do not spread to form a lipid bilayer because no recovery of the bleached area could be detected.

Section 4: Lipid mixing of LUVs with adhered GUVs Experiment 1: Low membrane tension
In experiment 1 the GUVs were slightly adhered to the functionalized glass surface to generate a low pre-stress.
The adjusted low membrane tension on the GUVs serves as a reference for highly tensed membranes to compare the docking and fusion efficiency of LUVs. Typical membrane tensions in biological cells are in the range of 0.01 mN/m to 0.3 mN/m. 2 Therefore, the membrane tension of a total of 13 GUVs was adjusted to a similar range between 0.17 mN/m and 1.2 mN/m. In the figures S13-S16, all measured GUVs are shown in cross-sectional images. In order to categorize docking and fusion of LUVs on each GUV, the image on the left side is composed as a two-channel image for both dyes to visualize docking of LUVs to the freestanding GUV-membrane. On the right side of figures S13-S16 the cross-sectional views of the channel for the LUV dye are shown. Adhesion of LUVs to the sample surface was found to occur in all measurements and the fluorescence intensity of the adhered LUVs served as a reference (100 %) for the docked LUVs on the GUVs.   In figure S13, four slightly adhered GUVs (1-4) with membrane tensions between 0.3 mN/m to 0.56 mN/m are shown that exhibit nearly no docked LUVs. GUVs 5-7 are shown in figure S14. Docking of LUVs to the GUVmembrane occurred mainly on GUV 5, which shows a larger adhesion radius compared to the other GUVs. The calculated membrane tension of 1.2 mN/m for GUV 5 is significantly higher compared to that of the other GUVs.
The adhesion site of GUV 5 shows no fluorescence intensity originating from the LUV dye. Therefore, lipid mixing with the membrane of GUV 5 can be excluded. In figure S14 A and B, the surrounding solution of the adhered GUVs contained a higher concentration of LUVs compared to image C. The GUVs 8-12 in figure S15 were incubated with LUVs for 50 minutes to see whether LUVs dock to the membranes over a prolonged time more frequently. Even after such a long time an increased docking and fusion of the LUVs to GUVs with membrane tensions lower than 0.51 mN/m could not be observed.
In figure S16, GUV 13 is shown with many docked LUVs at the freestanding membrane. Compared to GUV 7 in figure S14 C with a membrane tension of 0.78 mN/m the docking on GUV 13 is significantly increased, which could be related to a higher LUV concentration in the surrounding solution. Already at this moderate tension, fusion of LUVs with GUV 13 can be monitored. The lipid mixing between the vesicles after fusion leads to the diffusion of the fluorescently labeled lipids into the adhesion area of the GUV. After the addition of the LUVs, GUV 13 exhibited a larger adhesion radius.
In summary, the docking probability of LUVs is increased at membrane tensions above 0.8 mN/m ( Fig. S16 and S14 A). Generally, membrane tensions below 1.2 mN/m on adhered GUVs do not lead to a measureable amount of fusion.

Experiment 2: Elevated membrane tension
Comparable membrane tensions to those occurring in cells do not lead to a significant increase of vesicle fusion to pre-stressed GUVs as it is described in experiment 1. Therefore, in experiment 2 GUVs with membrane tensions from 1.1 mN/m to 8.6 mN/m served as target membranes for the investigation of the docking and fusion efficiency of LUVs. We used an elevated Mg 2+ concentration to foster adhesion as the driving force for stretching.
In figures S17-S19 the GUVs 14-23 with an elevated membrane tension are shown after the incubation of LUVs.
GUV 14 shown in figure S17 has the largest membrane tension of 8.6 mN/m of all measured GUVs, close to lysis tension. The two-channel image on the left side of the figures shows both fluorescently labeled lipids and on the right-side the channel for the LUV dye is shown (GUV labeled with ATTO A390, LUVs with ATTO A594 or A488, respectively). All highly-tensed GUV-membranes reveal a large number of docked LUVs. At the adhesion site of the tensed GUVs, the fluorescence intensity of the LUV dye was increased indicative of fusion with the GUVmembrane. To prove that the fluorescently labeled dye originating from LUVs can diffuse freely in the GUVmembrane FRAP experiments were carried out at the adhesion site as described in section 5.
In figures S19 C-E two touching GUVs (  immobile fraction of about 17% was found. In figures S21 F-J the edge of the adhesion area was bleached and a recovery was detected, whereby the fluorescence intensity recovered to around 60% of its initial value.   The FRAP-measurement shown in figure S25 shows that the GUV content can be bleached by the LASER without recovery confirming that the GUVs are not leaky. To further exclude diffusion of the dye ATTO ® 488 from the surrounding buffer solution into the GUV a control sample without SNAREs in neither the GUVs nor the LUVs is shown in figure S26. Even strongly adhered GUVs with a high membrane tension do not contain any fluorescence intensity so that non-specific fusion and diffusion of the water-soluble LUV dye through the GUV-membrane can be excluded.

Figure S25 | Content mixing -FRAP measurement.
The images were taken before (A) and after (B) bleaching of the GUV content at 6 s. Image (C) was the last one more than 300 s after the GUV content was bleached.

Control measurements in the absence of SNAREs
Content mixing in the absence of SNAREs due to leaky GUVs was exclude by preparing samples with the water soluble fluorescent dye ATTO 488 ® carboxy (green). Fusion of LUVs, containing the water-soluble fluorescent dye inside, with the adhered GUVs was not observed ( Figure S26). Also, LUVs containing synaptobrevin did not fuse with an adhered GUV (Fig. S26 A). For the two GUVs 1 and 2 from figure S26 the corresponding membrane tension is listed in the table (C). Content mixing in the absence of the DN49-complex at the adhered GUV (Fig. S26 A) or in the absence of both synaptobrevin orthe DN49 complex was not observed even for highly tensed membranes ( Fig. S26 B+C).

Figure S26 | Content mixing without SNAREs.
A) Content mixing assay with synaptobrevin at the incubated LUVs but without the reconstituted DN49-complex in the adhered GUVs. The green fluorescent dye inside the LUVs did not diffuse into the GUV confirming that the vesicles are not leaky. B) GUVs in the absence of SNAREs shows no content mixing after addition of LUVs filled with a green fluorescent dye. The cross-sectional view shows that the GUVs strongly adhere on the surface generating a highly-tensed membrane but still do not show fusion. C) The table compiles the measured radii, area changes and the corresponding membrane tension of the GUVs 1 and 2 from B. LUVs were produced by the detergent dilution method as described in the materials and methods section of the paper. In figure S27, the typical diameter size distribution is shown measured by dynamic light scattering in buffer solution.