Synaptotagmin 7 outperforms synaptotagmin 1 to open and stabilize nascent fusion pores via robust membrane penetration

Ca 2+ -triggered exocytosis is a crucial aspect of neuronal and neuroendocrine cell function, yet many of the underlying molecular mechanisms that regulate these processes are unknown. Here, we contrast the biophysical properties of two prominent neuronal Ca 2+ sensors, synaptotagmin (syt) 1 and syt7. In both proteins, four Ca 2+ -binding loops partially penetrate bilayers that harbor anionic phospholipids, and mutagenesis studies suggest that these interactions are important for function. However, these mutations also alter the interaction of syts with the SNARE proteins that directly catalyze membrane fusion. To directly assess the role of syt membrane penetration, we took a different approach and found that Ca 2+ -syt1-membrane interactions are strongly influenced by membrane order; tight lateral packing of phosphatidylserine abrogates syt1 binding to lipid bilayers due to impaired membrane penetration. Function could be restored by making the membrane penetration loops more hydrophobic, or by inclusion of cholesterol. In sharp contrast, syt7 unexpectedly exhibited robust membrane binding and penetration activity, regardless of the lipid acyl chain structure. Thus, syt7 is a ‘super-penetrator’. We exploited these observations to specifically isolate and examine the role of membrane penetration in syt function. By altering bilayer composition, rather than protein structure, we disentangled the roles of syt-membrane versus syt-SNARE interactions. Using nanodisc-black lipid membrane electrophysiology, we demonstrate that membrane penetration underlies the ability of syts to directly regulate reconstituted, exocytic fusion pores in response to Ca 2+ . Quantification of a liposome titration in the presence of NBD-labelled syt1 C2AB. The protein concentration was fixed a 250 nM and the fluorescence at 525 nm was monitored as the concentration of lipid increased from 0-500 µM. Error bars represent standard error of the mean from triplicate experiments performed on separate days with fresh materials.


Introduction
EPSC amplitude and Ca 2+ -sensitivity, compared to WT syt1 (43,44). However, the 4W and 4A mutations have other effects on syt1 biochemistry, including altered interactions with SNARE proteins (39). Additional approaches are needed to unambiguously define the precise role of syt1-membrane interactions in fusion.
Here, we examine how membrane binding and penetration by the C2-domains of syt1 and syt7 contribute to Ca 2+ -triggered membrane fusion. Specifically, we use phospholipid bilayer order as a tool to explore the link between Ca 2+ -syt membrane penetration and the regulation of reconstituted fusion pores, which represents the first crucial intermediate in the fusion reaction. We demonstrate that syt function can be controlled by manipulating PS acyl chain structure. In particular, syt1 cannot bind or penetrate into bilayers containing phosphatidylserine that has saturated acyl chains; without efficient membrane penetration, syt1 fails to respond to Ca 2+ to regulate fusion pores. Interestingly, we observed that syt7 displays far more robust membrane penetration activity as compared to syt1, and efficiently penetrated all bilayers that were tested, regardless of the membrane order. Furthermore, we report the first use and characterization of reconstituted full-length syt7 and determine, via comparisons with syt1, that the ability of C2domains to penetrate membranes underlies how of these proteins to regulate fusion pores.

Syt7, but not syt1, penetrates membranes that harbor saturated PS
To compare how syt1 and syt7 interact with phospholipid bilayers, we first performed a well-described syt-loop penetration assay (34, 45). For this, we labeled the cytoplasmic domains (denoted C2AB) of each isoform with a solvatochromic fluorescent dye, NBD, on a cysteine residue placed in the distal tip of a membrane penetration loop (loop-3) in C2B. In the absence of Ca 2+ (EGTA), the C2AB domains do not interact appreciably with 100 nm PC/PS (80: 20) liposomes, resulting in a low fluorescence signal (Figure 1A-C). Upon binding Ca 2+ , the C2AB domains rapidly associate with the membrane and partially insert into the hydrophobic core of the bilayer (34, 35), causing a significant increase in blue-shifted NBD fluorescence (Figure 1A-C). To specifically examine the contribution of hydrophobic interactions to the binding reaction, we compared the ability of syt1 and syt7 to penetrate into bilayers containing 20% PS with either saturated (16:0/16:0) or unsaturated acyl chains (18:1/18:1), also notated as dipalmitatephosphatidylserine (DPPS) and dioleate-phosphatidylserine (DOPS), respectively; the remaining 80% of the lipids in both conditions were unsaturated dioleate-phosphatidylcholine (DOPC) (Figure S1). Knowing that syt1 and syt7 are PS binding proteins, we postulated that tight lateral packing of saturated PS acyl chains in the liposomes might render the bilayers refractory to penetration, thus restricting hydrophobic interactions. As expected, syt1 and syt7 were both able to penetrate bilayers containing unsaturated DOPS (Figure 1B & 1C). However, if the PS acyl chains are fully saturated, syt1 C2AB penetration was significantly reduced (Figure 1B), despite having 80% of the bilayer composed of unsaturated DOPC. Surprisingly, we found syt7 exhibited comparable penetration into saturated and unsaturated PS bilayers (Figure 1C), indicating a more robust mode of binding.
To establish the generality of the syt1 penetration defect into saturated PS-bilayers, we examined the membrane penetration performance of other C2-domain-containing proteins under the same conditions as were used with syt1 and syt7. We engineered a single cysteine substitution into the third loop of the C2-domains from protein kinase C (PKC) and cytosolic phospholipase A2 (cPLA2), as well as the third loop in the C2B domain of Doc2β C2AB. We found that all three proteins also displayed significantly impaired penetration into membranes with saturated acyl chains in response to Ca 2+ (Figure S2), thus the robust penetration of syt7 into DPPS bilayers appears unique.

Molecular dynamics simulations show saturated PS clusters resist syt1 C2B-domain penetration
Next, to further examine how syt1 and syt7 interact with bilayers containing saturated and unsaturated PS, we conducted all-atom molecular dynamics (MD) simulations of their C2B domains with phospholipid bilayers, again using the same lipid compositions used above. We first performed lipid-only MD simulations to monitor the stability of DOPS and DPPS clusters in the DOPC/DxPS (80: 20) bilayers. We postulated that DPPS may form a tight cluster, thus restricting C2-domain penetration. For this initial test, a small cluster of 32 PS lipids, either DOPS or DPPS, was placed in the center of a DOPC membrane ( Figure S3A) and equilibrated for 450 ns using MD simulations. Figure 2A shows that DPPS lipids (cyan) are much more compact than DOPS lipids (blue), suggesting that the saturated PS lipids are likely to form bigger, more stable clusters in a saturated-unsaturated lipid mixture (Supplemental movies 1 & 2). We then quantified how clustered the two PS species were at the end of the simulations by measuring the midpoint distances between the PS molecules; the radial distribution function (RDF), shows that the DPPS molecules remain in close proximity with a peak centered around 0.75 nm. In contrast, the DOPS molecules are more dispersed, as indicated by a shorter peak height with a broad distribution ( Figure 2B). These initial MD simulations were also validated via atomic force microscopy (AFM) imaging of supported lipid bilayers containing DOPS or DPPS; we found that Ca 2+ reversibly caused DPPS to segregate into 15 ± 7 nm clusters in the bilayer ( Figure S4A). In contrast, no lipid clustering was observed in the DOPS condition ( Figure S4B).
We then carried out a second set of MD simulations to examine how PS acyl chain structure affects C2-domain-membrane interactions. The isolated C2B domain of syt1 (orange) or syt7 (yellow) was placed on top of the membrane such that the membrane penetration loops were positioned close to the PS cluster (Figures S3B & S3C). The system was solvated and charge neutralized by adding Ca 2+ and Clions. Upon equilibration of the C2B domains, we found that Ca 2+ -bound loops 1 and 3 of syt1 and syt7 spontaneously inserted into, and remained inside, the DOPC/DOPS membrane throughout the 1000 ns long MD simulations (Figures 2C).
In both cases, the insertion depth of loop 3 was greater than the depth of loop 1, while loop 2 remained away from the lipid head groups (Figures 2C, lower panel). Insertion of loops 1 and 3 allows the hydrophobic residues, V304 and I367 of syt1, and I298 and L361 of syt7, to access the hydrophobic bilayer core, strengthening the binding interaction. This result matches well with previous experimental results from our laboratory (45). In contrast to the syt1-DOPS simulation, tight packing of the PS lipids in the DOPC/DPPS membrane does not allow strong binding or insertion of the syt1 C2B domain (Figure 2D, left panel). However, in line with our experimental results (Figure 1C), we found that the syt7 C2B loops penetrated well into the DPPS cluster ( Figure 2D, right panel).

Cholesterol rescues the syt1 membrane penetration defect
The experiments and MD simulations above reveal that acyl chain packing strongly affects the ability of syt1, but not syt7, to penetrate membranes in response to Ca 2+ . We therefore tested whether 'loosening' this lateral packing can rescue syt1 function using bilayers that harbored DPPS. Physiological lipid bilayers contain approximately 30 -40% cholesterol (6,46). Cholesterol displays unique membrane modulating properties by reducing the rotational mobility (stiffening) of unsaturated acyl chains, while also adding space between (loosening) tightly packed saturated acyl chains. We therefore hypothesized that introducing cholesterol into the DPPS bilayer would loosen the PS clusters and allow the syt1 C2B domain to penetrate the otherwise refractory membrane. After repeating the DPPS MD simulations in the presence of 30% cholesterol, we indeed found that loops 1 and 3 of syt1 C2B could again penetrate the DPPS membrane (Figure 3A & 3B). Correspondingly, in AFM experiments, Ca 2+ failed to induce detectable clusters in DPPS supported lipid bilayers that contained 30% cholesterol ( Figure S4C).
Next, we revisited the syt1 NBD-loop penetration experiments described in Figure 1 to further examine the effects of cholesterol, experimentally. First, we used soluble cholesterolloaded methyl-β-cyclodextrin (sChol) to introduce cholesterol into DOPC/DPPS bilayers while monitoring changes in NBD fluorescence in real-time ( Figure 3C). As described above ( Figure  1B), Ca 2+ -bound syt1 C2AB can only minimally penetrate DPPS-containing bilayers, as indicated by a relatively low fluorescence signal under this condition ( Figure 3D, grey trace). However, after of the addition of sChol, to incorporate cholesterol into the liposomes, syt1 C2AB regained the ability to penetrate the bilayer ( Figure 3D, black trace), showing maximal penetration after approximately 10 minutes. This penetration rescue was also directly visualized in by microscopy of DOPC/DPPS giant unilamellar vesicles (GUVs) with and without 30% cholesterol ( Figure 3E). No NBD fluorescence was observed on the GUV surface in the absence of Ca 2+ ; upon addition of Ca 2+ , syt1 C2AB became associated with the cholesterol containing GUVs, as indicated by the appearance of the NBD signal on the outer membrane surface. Notably, the DOPC/DPPS GUVs without cholesterol fail to show any NBD signal on the GUV surface in response to Ca 2+ ( Figure 3E).
We then generated DOPC/DPPS liposomes with increasing amounts of cholesterol, up to 30%, while monitoring membrane penetration by fluorometery. As with the real-time addition of cholesterol, discrete increases in the cholesterol concentration allowed syt1 C2AB to penetrate membranes, with an EC50 of 17.6%, reaching maximal membrane penetration at 30% cholesterol ( Figure 3F). Without cholesterol, approximately ten-fold more lipid to was needed to drive maximal penetration of syt1 C2AB into liposomes bearing DPPS (protein:lipid, 1:4000), compared to DOPS (protein:lipid, 1:400) ( Figure 3G). However, if cholesterol is present, syt1 penetration is comparable for all PS species, across a wide range of protein/lipid ratios ( Figure  3G). This effect of cholesterol on syt1 binding to DPPS rationalizes how Kiessling et al. (2018) observed syt1 triggering membrane fusion using bilayers containing saturated PS, as these membranes also contained 20% cholesterol (47). Together, these results support the conclusion that syt1 fails to penetrate DPPS-containing membranes, due to the rigidity of the bilayer.
However, the attributes of syt7 that enabled robust penetration into DPPS bilayers remained unclear; this is addressed further below.

C2-domain binding to phospholipid bilayers is coupled with membrane penetration
To determine whether the compromised membrane penetration of syt1 was also accompanied by a failure to bind electrostatically to the liposome surface, we performed a protein-liposome co-sedimentation assay. The C2AB domains were mixed with 100 nm liposomes that were composed of 20% saturated (DPPS) or unsaturated PS (DOPS), in 0.2 mM EGTA or 0.5 mM free Ca 2+ , followed by ultracentrifugation to pellet the liposomes ( Figure 4A); bound proteins co-sediment with the liposomes. C2AB domains from both syt1 and syt7 efficiently bound DOPS-containing liposomes in the presence of Ca 2+ , as indicated by a depletion of protein from the supernatant ( Figure 4B). However, analogous to the penetration data, only 20-25% of syt1 C2AB bound DPPS-bearing liposomes in Ca 2+ (Figure 4C). This validates that membrane insertion is a prerequisite for syt1-membrane binding, presumably by providing hydrophobic interactions; electrostatic interactions are apparently insufficient to enable syt1 binding in isolation. This conclusion is consistent with earlier mutagenesis studies (39) In contrast to syt1, and in agreement with the penetration results, we found that syt7 maintained robust binding to liposomes bearing either unsaturated or saturated PS ( Figure 4B & 4C), further supporting the idea that membrane binding is coupled with penetration. Notably, we consistently observed that approximately 25-30% of the syt7 bound to the liposomes in EGTA; this might reflect the propensity of DPPS to form clusters, thereby increasing the local concentration of this lipid. Indeed, a previous report showed that syt7 exhibits robust Ca 2+independent binding to liposomes, as a function of the mole fraction of PS (48).

Hydrophobic residues are critical for C2-domain membrane binding
The membrane binding and penetration assays described above revealed that the C2domains of syt1 and syt7 engage distinctly with lipid bilayers. We therefore compared the amino acid sequences of their membrane penetration loops; overall, the sequences are highly conserved ( Figure 4D). The most notable differences are that the syt1 loops contain three more hydrophobic residues, and three fewer positive charged residues (excluding the histidine residues in the syt1 loops at physiological pH) than the syt7 loops. This could be expected to equip syt1 with comparatively enhanced penetration activity, and syt7 with prominent electrostatic binding component; instead, we found that syt7 displayed enhanced membrane binding (Figure 4B & 4C) and penetration ( Figure 1C). To delve into this further, we generated a series of mutants to examine how the hydrophobic and cationic character of the loops influenced binding. These included a syt1 C2AB construct that contained five tryptophan substitutions in the penetration loops (syt1 5W) to increase hydrophobicity, or a poly-cationic mutant (L171K, H198K, K236R, N333R, Y364K, K369R) that substituted all the cationic residues of the syt7 loops on to syt1 (syt1 s7+). Note that while K236R and K369R in the syt1 s7+ construct do not affect the charge at these two positions, arginine residues have been reported to exhibit enhanced interfacial binding to phospholipids, compared to lysines (49). We also examined a syt7 C2AB mutant with reduced hydrophobicity (F167, F229, I298, L361A) in the penetration loops (syt7 4A).
We first tested how syt1 5W, syt1 s7+ and syt7 4A performed against the refractory bilayers containing saturated DPPS. Co-sedimentation assays revealed that syt1 5W was indeed able to enhance membrane binding in response to Ca 2+ ; the tryptophan substitution enabled syt1 5W to bind equally well to DOPS and DPPS (Figure 4E & 4F). Moreover, NBDlabelled syt1 5W was able to penetrate efficiently into DPPS bilayers ( Figure S5). In contrast, the syt1 s7+ construct did not improve binding to DPPS in response to Ca 2+ (Figure 4E & 4F). In fact, syt1 s7+ also displayed a partial reduction in binding to DOPS in Ca 2+ ; in short, this mutant exhibited the opposite of the predicted effect. When examining syt7 4A, we found this construct had a reduced ability to bind DPPS in Ca 2+ , compared to WT syt7 ( Figure 4C, 4E & 4F), again highlighting the importance of hydrophobic residues in the loops.
To further explore how electrostatic interactions contribute to syt-membrane interactions, we conducted protein-liposome co-sedimentation experiments as a function of increasing ionic strength. We observed that syt1-membrane binding was highly sensitive to the salt concentration, whereas WT syt7 was entirely resistant (Figure 4G & 4H), as reported previously (34, 45,50). We then repeated these experiments using syt1 5W, syt1 s7+ and syt7 4A. We found that increasing the hydrophobicity of syt1 (syt1 5W) had a greater impact on overcoming salt sensitivity, compared with positive charged residue substitutions ( Figure 4G & 4H); bestowing the electrostatics of syt7 loops onto syt1 (syt1 s7+) did not improve salt resistance. This syt1 s7+ result is surprising considering this construct contains prominent hydrophobic and electrostatic character yet is still outperformed by WT syt7. Moreover, syt1 5W was also outperformed by WT syt7 in this assay. In line with the DPPS co-sedimentation experiments ( Figure 4E & 4F), we also observed that hydrophobicity is critical for salt-insensitivity of syt7, as the 4A mutation enabled modest increases in salt (200 mM) to disrupt binding ( Figure 4G & 4H). To tease apart contributions of each syt7 C2-domain, we generated and tested syt7 C2AB constructs with reduced loop hydrophobicity in either the C2A (syt7 C2A2AB) domain or the C2B (syt7 C2AB2A) domain. These results demonstrated that the C2A domain of syt7 is responsible for the salt insensitivity, whereas the syt7 C2AB2A mutation had no effect ( Figure 4G & 4H). We also found, via stopped-flow rapid mixing experiments, that the membrane dissociation kinetics of each of these constructs shared the same trends as the co-sedimentation studies ( Figure 4I). Specifically, the 5W mutation slowed syt1 dissociation from liposomes and both the syt7 4A and syt7 C2A2AB mutations dramatically increased the dissociation rate ( Figure 4I), further supporting the importance of loop hydrophobicity. Curiously, we found that the syt7 C2AB2A mutation reproducibly slowed the dissociation kinetics ( Figure 4I). At present, a mechanism for this perplexing C2AB2A mutation effect is unknown.
Our data support previous conclusions (39,43) that C2-domain loop hydrophobicity is critical for membrane binding, especially when contrasting our syt7 4A and syt1 s7+ results ( Figure 4E & 4F). However, considering syt7 has higher membrane binding affinity ( Figure S6), with fewer hydrophobic residues in the loops, compared to syt1 (Figure 4D), these results highlight the somewhat counterintuitive complexity of C2-domain-membrane interactions, which cannot easily be predicted from the primary sequence. These results indicate that the hydrophobicity of the C2-domain loops may be more important than the cationic properties in . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; governing membrane binding. However, additional factors, other than the sum of hydrophobic and electrostatic residues in the loops, appear to influence C2AB-membrane interactions.

Membrane penetration is required for synaptotagmin 1 to dilate and stabilize fusion pores
Having established that PS acyl chain order can govern how C2-domains interact with lipid bilayers, we proceeded to investigate functional consequences of restricting syt membrane penetration. We reasoned that PS acyl chain structure could be exploited to specifically interrogate the relationship between membrane penetration and syt-mediated fusion pore expansion (51). For this, we used the recently developed nanodisc-black lipid membrane (ND-BLM) planar lipid bilayer electrophysiology approach (51,52). Previously, we reported that 13 nm nanodiscs (ND) with three copies of the SV SNARE, synaptobrevin 2, (ND3) make small unstable fusion pores (40,52) in BLMs containing the t-SNAREs syntaxin-1a and SNAP-25. However, incorporation of syt1 into these NDs caused the fusion pores to expand into larger, more stable structures in response to Ca 2+ (51). This experimental system is, therefore, ideal to examine the direct role of syt1 in fusion pore formation and expansion/dilation. By substituting PS species within the BLM, we can examine how syt membrane penetration influences pore kinetic properties and dilation.
Classically, BLM experiments are conducted by first dissolving a lipid mixture in a solvent, such as n-decane, followed by using a fine tipped brush or glass capillary to paint the lipids across a small aperture to form a planar lipid bilayer (51). We found the critical lipid in this study, DPPS, to be insoluble in n-decane, making the standard BLM protocol unusable. We therefore developed a new strategy to form a planar lipid bilayer across the aperture of a standard BLM cup that could accommodate all lipids used in this study (described in detail in the Methods section). This approach involves resuspending the lipids in pure water and then pipetting a droplet of lipid across the aperture of the BLM cup ( Figure 5A). The cup is then placed into a vacuum desiccator to evaporate the water and form a lipid film. Buffer is then applied, and the lipids self-assemble to form a bilayer across the aperture. In contrast to other BLM protocols, this new strategy also has the advantage of allowing the incorporation of t-SNAREs directly into the BLM lipid mixture, rather than requiring a second step to donate the t-SNAREs into the BLM after formation, as was done previously (51). This drying method was successfully able to form t-SNARE-incorporated bilayers, enabling DPPS ND-BLM experiments.
In line with Das et al. (2020) we validated that in the absence of Ca 2+ (0.5 mM BAPTA), ND3-syt1 formed small and unstable fusion pores in all conditions (Figure 5B & 5C). The addition of Ca 2+ caused a significant increase in the current passing through ND3-syt1 fusion pores that form in the DOPS-containing BLM (Figures 5B & 5C), indicative of an increase in pore size. Ca 2+ significantly increased the fraction of time that the ND3-syt1 pores remained in the open state in the DOPS-containing BLM ( Figure 5D). The peak of the open dwell time distribution was also shifted to longer open times by >100-fold ( Figure 5E). Interestingly, we found that Ca 2+ failed to trigger syt1-mediated fusion pore expansion in the DPPS-containing BLM (Figure 5B-5E was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint state were also unaffected by Ca 2+ . Together, these data suggest that syt1 must penetrate the target membrane in order to drive stable fusion pore opening and dilation.

Synaptotagmin 7 robustly penetrates ordered PS bilayers to dilate and stabilize fusion pores
To further correlate membrane penetration with fusion pore opening and dilation we aimed to purify and reconstitute full-length syt1 5W (FL-syt1 5W), and full-length syt7 (FL-syt7) into NDs to compare with the WT syt1 ND-BLM recordings, as the C2AB domains of syt1 5W and syt7 were able to efficiently bind (Figure 4C & 4F) and penetrate (Figure 1, 2 & S5) into DPPS-containing bilayers. However, to date, the use of recombinant FL-syt7 has not been reported, presumably due to difficulty in purification. We were able to express and purify functional FL-syt7, using high volumes of starting material and a stringent protocol, which enabled us to examine the impact of FL-syt7 on fusion pores, as compared to FL-syt1 and FL-syt1 5W. We reasoned that if syt1 5W and syt7 C2AB domains were able to penetrate saturated PS bilayers, and if penetration is indeed critical for regulating pores, then both FL-syt1 5W and FL-syt7 should be capable of opening and dilating fusion pores that form in a DPPS-bearing BLM, in response to Ca 2+ . Indeed, increased hydrophobicity was previously shown to enhance fusion pore expansion in HeLa cells expressing flipped t-SNAREs (53) and in PC12 cells (54). In line with enhanced penetration performance, compared to WT syt1, we found that FL-syt1 5W and FL-syt7 reconstituted into ND3 both yielded significant increases in fusion pore current in the presence of Ca 2+ (Figure 5B & 5C). The addition of Ca 2+ also significantly increased the fraction of time that ND3-syt1 5W and ND3-syt7 pores were in the open state ( Figure 5D) and the open dwell time distribution shifted to approximately 10-fold longer open times ( Figure 5E).
When examining the kinetics of fusion pore transitions in the presence of Ca 2+ , we found that the closing rates for ND3-syt1 fusion pores formed in the DPPS-BLM were significantly faster than for the DOPS condition ( Figure 5F, upper panel); FL-syt1 failed to stabilize the fusion pore open state when membrane penetration was impaired. Moreover, the closing rates of the ND3-syt1 fusion pores in the DPPS-containing BLM were also significantly faster than the penetration-competent ND3-syt1 5W and ND3-syt7 samples when Ca 2+ was present (Figure 1, 5F & S5), further supporting the conclusion that stabilization of the open state is likely a direct consequence of C2-domain penetration into the target BLM. Interestingly, while Ca 2+ enabled ND3-syt1 to significantly increase the fusion pore opening rate in the DOPS BLM, we found no difference in the opening rates between ND3-syt1, ND3-syt1 5W and ND3-syt7 in the DPPS condition ( Figure 5F, lower panel & Figure S7). The discrepancy in opening rates between the penetration competent samples, ND3-syt1 in DOPS, as well as ND3-syt1 5W and ND3-syt7 in DPPS, may suggest that C2-domain membrane penetration is preceded by SNARE zippering. Alternatively, the presence of DPPS at the fusion pore site may alter the conformation of the t-SNAREs (47) or increase the local rigidity of the phospholipid bilayer (55), to affect pore opening. Indeed, it was previously demonstrated that saturated acyl chains reduce SNAREmediated fusion (56). It is also noteworthy that in the ND-BLM BAPTA conditions, PC and PS (80% and 20%, respectively) are expected to be uniformly distributed throughout the BLM ( Figure S4A). Therefore, SNARE-alone fusion pores could form in a PS-independent manner (i.e., by forming a malleable, DOPC rich fusion pore) and are thus less affected by the acyl . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint chain structure of the PS constituent. However, in the DPPS BLM, Ca 2+ is expected to cause the syts to act at the site of, or adjacent to, a rigid PS cluster (Figure 2 & S4A), which may affect the kinetics of pore opening. Regardless, upon pore opening, the penetration competent syts act to restrict closure and drive fusion pore dilation.

Discussion
During membrane fusion, discrete phospholipid bilayers are tethered and drawn together by proteins to overcome significant energy barriers and drive membrane merger. Recently, we described that an accessory protein, complexin, promotes SNARE-mediated membrane fusion by stabilizing these curved intermediate structures via insertion of a C-terminal amphipathic helix at the fusion pore site (40). Syt1 has also been suggested to promote fusion via membrane insertion (34, [37][38][39]. However, with the multitude of functions assigned to syt1, precisely how this molecule synchronizes Ca 2+ influx into nerve terminals with the release of neurotransmitter remains unclear. In particular, the specific contribution of syt1 membrane penetration in triggering the fusion reaction has not been rigorously established. For example, syt1 is thought to inhibit SV exocytosis in the absence of Ca 2+ (a fusion clamp) by binding SNAREs and preventing complete zippering (9). In principle, Ca 2+ binding could drive synchronized SV exocytosis, simply by re-directing syt1 away from the trans-SNARE complex to release the fusion clamp and trigger fusion, making membrane penetration inconsequential. Previous studies have examined the role of syt1 membrane penetration by mutagenesis to increase or decrease the hydrophobicity of the C2-domain penetration loops (39,53,57). These studies support the importance of membrane penetration, but -as described above -these mutants may have unintended effects on protein function that confound conclusions.
To specifically assess the role of syt1 membrane penetration in the fusion reaction, we took an alternative approach by primarily working with WT proteins and then manipulating the composition of the target membrane to control syt1-membrane interactions. We found that syt1 cannot bind or penetrate into phospholipid bilayers containing saturated PS (Figure 1-4). We reasoned that this phenomenon could be exploited to isolate the role of membrane penetration from SNARE binding to gain detailed insights into how syt1 triggers fusion. We also found that other C2-domain containing proteins (i.e. cPLA2, PKC and Doc2β) failed to penetrate into bilayers containing saturated acyl chains (Figure S2), which could be explored in future studies. Surprisingly, however, we discovered that syt7 exhibited robust membrane penetration into bilayers containing either saturated or unsaturated PS (Figure 1 & 2). This distinction enabled direct comparisons between syt1 and syt7, to tease apart the specific role of membrane penetration in syt function.
The distinct membrane penetration performance of syt1 and syt7, both experimentally (Figure 1) and via MD simulations (Figure 2), then led us to investigate a mechanism that enables syt7 to act as a 'super-penetrator'. We therefore examined the relative contributions of hydrophobic and electrostatic interactions in mediating syt-membrane binding. We demonstrate that Ca 2+ -dependent binding of C2-domains to phospholipid bilayers requires membrane penetration (Figure 4); in isolation, electrostatic interactions are insufficient. Indeed, although Ca 2+ -bound syt1 binds to di-acylated anionic lipids with high affinity, syt1 does not associate with mono-acylated PS (lyso-PS) (58), which originally suggested that syt1 membrane binding is . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint a combination of electrostatic and hydrophobic interactions. Hydrophobic residues in the membrane penetration loops of both syt1 and syt7 are indeed critical for maintaining binding to phospholipid bilayers (Figure 4). Specifically, increasing syt1 hydrophobicity overcomes the defect in binding to saturated PS and slows the membrane dissociation kinetics (Figure 4E, 4F  & 4I), while reduced hydrophobicity renders syt7 susceptible to elevated salt (Figure 4G & 4H) and speeds up the dissociation kinetics ( Figure 4I). Curiously, the primary amino acid sequence of the syt1 and syt7 penetration loops (Figure 4D) failed to predict the binding properties of these C2-domains, suggesting that membrane binding is more complex than simply a sum of their charged and hydrophobic residues.
After identifying that syt1 membrane binding can be controlled by manipulating the PS acyl chain structure, and that syt7 and a gain-of-function syt1 (syt1 5W) can overcome the saturated PS binding defect observed with WT syt1, we proceeded to test these three variants functionally using the ND-BLM approach. To our knowledge, this is the first reported purification and successful reconstitution of FL-syt7. We aimed to determine whether membrane penetration is required for syt1/7 to drive fusion pore opening and dilation. A new BLM method was developed in order to enable the formation of planar lipid bilayers containing all the lipids used in this study ( Figure 5A). As predicted, we observed that stable fusion pore opening and dilation were directly correlated with the ability of syts to penetrate membranes. Specifically, when a C2-domain is incapable of penetrating into a bilayer, fusion pore stabilization and dilation were blocked. Conversely, conditions that permit the syts to penetrate membranes in response to Ca 2+ (syt1 into DOPS, or syt1 5W and syt7 into DPPS) result in larger, more stable fusion pores (Figure 5). This validates that syt membrane penetration directly mediates, at least in part, the regulation of fusion pores.
Kinetic analysis revealed that membrane penetration competent syts significantly reduced the fusion pore closing rates, thereby keeping pores in the open state for longer periods ( Figure 5F). Notably, the ND-BLM system traps fusion pores in a reversible intermediate state, due to the rigid scaffold that surrounds the NDs. However, in vivo, no such restricting scaffold is present; the penetration action of the syts in neurons may instead favor a one-way reaction by stabilizing the nascent fusion pore, reducing the propensity to reverse towards closure, and supporting pore dilation. In this view, Ca 2+ binding by syts directs the C2domains to penetrate into the target membrane, thus lowering the energy barrier for full fusion of SV with the plasma membrane.
It is interesting to note that, while syt7 was determined herein to exhibit more robust binding and penetration into lipid bilayers, as compared to syt1, recent findings suggest that syt7 does not act directly as a Ca 2+ sensor for SV exocytosis (21,23). Instead syt7 functions as a dynamic, Ca 2+ -regulated SV docking protein on the axonal plasma membrane that feeds vesicles to Doc2α during asynchronous neurotransmitter release (27). In this context, the extent to which syt7 membrane penetration contributes to SV exocytosis is unknown. Moreover, with highly curved ~42 nm SVs as the target membrane in this case, syt7 binding and penetration, to mediate SV docking, rather than fusion, may be the critical function. However, syt7 has also been reported to be a direct exocytic Ca 2+ sensor residing on lysosomes, contributing to lysosome-plasma membrane fusion (28) and on dense-core granules in chromaffin cells, to release neuropeptides (30). Yet another study suggested that syt7 on dense core vesicles plays a role in docking or priming, rather than fusion per se (32). Clearly, additional work is needed to . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; clarify the function of syt7, but an appealing idea is that syt7 functions differently when targeted to the presynaptic membrane versus when it is targeted to dense core vesicles or lysosomes. In this study, we reconstituted FL-syt7 into NDs to facilitate a direct comparison with syt1 using the ND-BLM approach to assess whether it can directly modulate fusion pores, and if so, how membrane penetration influences this regulation. We established that syt7 exhibits unusually robust membrane binding properties and that membrane penetration is indeed a critical step that enables both syt1 and syt7 to directly regulate fusion pores in response to binding Ca 2+ . We note that interrogation of syt1 and syt7 function in the ND-BLM experiment mimics, in a sense, dense-core granule exocytosis, due to both isoforms residing on the ND (i.e., the vesicle face). Ongoing studies will incorporate FL-syt7 into the BLM and syt1 in the ND to model SV biology, in vitro, with increased complexity. Moreover, with recombinant FL-syt7 readily in hand, future studies will use FL-syt7 to reconstitute the SV docking step, with the goal of developing an in vitro model to build a more complete understanding of the SV cycle. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Main figures
The copyright holder for this preprint (which this version posted July 11, 2022. ; same conditions as panel B. Error bars represent standard error of the mean. **** represents p<0.0001 and ns represents a non-significant difference between conditions. . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; to assay for protein depletion. B) Representative Coomassie stained SDS-PAGE gels of syt1 and syt7 co-sedimentation samples, in the presence and absence of Ca 2+ , using liposomes composed of DOPC/DOPS or DOPC/DPPS. Throughout the figure, "Input" refers to the proteinonly control sample. C) Quantification of the replicated syt1 and syt7 co-sedimentation assays comparing binding to DOPS and DPPS containing liposomes. D) Amino acid sequences of the penetration loops of syt1 and syt7 C2A and C2B domains. Unique residues at comparable positions between the two sequences are indicated by an asterisk (*). Unique residues with a differences in charge are indicated by asterisk and exclamation point (*!). E) Representative Coomassie stained SDS-PAGE gels of mutant syt1 and syt7 co-sedimentation samples, in the presence and absence of Ca 2+ , and liposomes composed of DOPC/DOPS or DOPC/DPPS. F) Quantification of the mutant syt1 and syt7 co-sedimentation assays comparing binding to DOPS-and DPPS-bearing liposomes. G) Representative Coomassie stained SDS-PAGE gels of WT and mutant syt1 and syt7 co-sedimentation samples containing between 100 and 500 mM salt. H) Quantification of the replicated WT and mutant syt1 and syt7 co-sedimentation samples containing between 100 and 500 mM salt. I) Disassembly of Ca 2+ -dependent WT and mutant syt1 and syt7 complexes with liposomes, measured by stopped flow rapid mixing with the Ca 2+ chelator, EGTA. Error bars represent standard error of the mean. Within the bar graphs, ****, ***, * and ns represents p<0.0001, p<0.001, p<0.5 and a non-significant difference, respectively. . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint Figure S1 -Phospholipid structures Chemical structures of the phospholipids used in this study . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Supplemental Figures
The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; Figure S6 -Syt7 displays higher affinity for PS-bearing liposomes than syt1 Representative Coomassie stained SDS-PAGE gel of a co-sedimentation assay (described in Figure 3A) with syt1 and syt7 C2AB domains in the presence of Ca 2+ and liposomes with increasing mole fractions of PS. Depletion of protein from the samples indicates binding to liposomes. Input refers to the protein-only control sample.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint

NBD labeling
The cPLA2 and PKC C2 domains were engineered to harbor a single cysteine residue in loop 3 of their single C2 domains, while syt1, syt1 5W, syt7 and Doc2β C2AB domains had a single cysteine in loop 3 of their C2B domains. The mutations for the single cysteine variants are listed above in the protein purification section. To conjugate NBD to the cysteine thiol group, 50 μM of each protein was incubated with at least 10fold excess iodoacetamide-NBD overnight at 4 ℃, followed by removal of free dye using a PD-10 desalting column. Protein labeling efficiency (>80 %) was calculated using the Beers-Lambert law (A = εcl) with the NBD extinction coefficient of 25000 M -1 cm -1 at 480 nm to determine NBD content, and this value was compared to the protein concentration.

Preparation of unilamellar vesicles
Large unilamellar vesicles (LUVs) were prepared by pipetting chloroform suspended lipids into a glass tube, followed by drying the sample under a stream of nitrogen gas.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint After formation of a thin lipid film, the sample was then placed in a lyophilizer for at least 60 minutes to remove residual organic solvent. The lipids were hydrated with 25 mM HEPES, 100 KCl pH 7.4 and placed at 50 ⁰C to aid in resuspension, followed by vortexing to generate multilamellar vesicles. The sample was then repeatedly passed through a 100 nm filter using a mini extruder system (Avanti Polar Lipids) to result in uniform LUVs.
Giant unilamellar vesicles (GUVs) were prepared by electroformation from a lipid film on indium tin oxide (ITO) coated glass slides, as previously described (59) After 2 hours, the frequency was changed to 0.5 Hz for 5 -10 minutes to stimulate the release of vesicles from the glass surface. After electroformation, the ITO slides were separated and the GUV solution was recovered from within the o-rings. The collected GUVs were then washed with 20 mM HEPES, 100 mM KCl (pH 7.4) and gently filtered to purify GUVs greater than 3 μm in diameter. Finally, the GUVs were transferred to BSA-treated glass bottom 96 well plate and imaged with a Zeiss 880 airyscan microscope using a 63x objective.

Co-sedimentation assay
One hundred nm LUVs (0.5 mM) composed of DOPC/DOPS or DOPC/DPPS (80: 20) were mixed with 2.5 µM of the indicated C2AB proteins in 0.2 mM EGTA or 0.5 mM free Ca 2+ . The LUVs, or protein-LUV complexes, were then pelleted by ultracentrifugation at 65,000 RPM in a TLA-100 rotor (Beckman). The unbound fraction C2AB in the supernatants was quantified by densitometry analysis of Coomassie blue stained SDS-. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; PAGE gels. For experiments with increasing salt, the samples received progressive increases in NaCl.

Atomic force microscopy
Supported lipid bilayers composed of DOPC/DOPS, DOPC/DPPS (80:20) or DOPC/DPPS/Cholesterol (56:14:30) were prepared by first generating 1 mM LUVs, as described above, followed by incubating a 10-fold diluted sample with freshly cleaved mica discs in a liquid cell. The AFM imaging of the supported lipid bilayers was performed using an Agilent 5500 Scanning Probe Microscope, as previously described (60).

Molecular dynamics simulations
The syt1 C2B protein crystal structure was taken from the PDB database (PDB ID 1K5W).
The lipid models were generated using CHARMM-GUI software (61). The CHARMM36 forcefield (62) was used to model all the components of the system. To check the stability of PS clusters, we first built a cylindrical cluster containing 32 DPPS or DOPS lipids. This PS cluster was then inserted into the center of a DOPC bilayer ( Figure S3A). The lipid bilayer was then solvated using TIP3P water box of (~9.4 × 9.4 × 8.0 nm 3 ) and 32 Ca 2+ and 32 Clions were added to the system. The solvated system was first energyminimized using conjugate gradient method to remove any bad contacts between the solvent and solute atoms. This step was followed by a short NVT simulation in which phosphorous atoms of the lipid heads were restrained. The NVT equilibrated system was then subject to restraint-free NPT equilibration at the atmospheric pressure and room temperature. The temperature and pressure of the system were controlled using Nose-Hoover thermostat with a time constant of 1 ps and Parrinello-Rahman barostat with a time constant of 5 ps, respectively. The covalent bonds involving light hydrogen atoms were constrained using LINCS algorithm to enable the simulation time step of 2 fs.
Periodic boundary conditions were enforced in all three directions and the long-range electrostatic interactions were calculated using particle-mesh-Ewald (PME) method. The short-ranged van der Waals forces was smoothly decayed to zero in the range of 1.0 to . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint 1.2 nm using a switching function. All simulations were carried out using the Gromacs suite of programs (63).

Stopped-flow rapid mixing
Rapid mixing was performed using a SX-18.MV stopped-flow spectrometer (Applied Photophysics). Samples constituting protein (4 µM), liposomes (1 mM lipids composed of 70% DOPC, 25% DOPS, and 5% dansyl-PE), and CaCl2 (250 µM) were allowed to rapidly mix with an equal volume of EGTA (2 mM) at room temperature (23° C). Before mixing, the samples were allowed to equilibrate for 5 min in their respective syringes of the spectrometer. Endogenous protein tryptophan residues were excited at 295 nm and the emission of the dansyl, due to FRET, was monitored via using a 470 nm long-pass filter (KV470, Schott). The reaction volume was set to 120 µl in the stopped-flow instrument. Using Applied Photophysics Pro-Data SX software, the average traces were fitted with either single or double exponential functions and the kinetic values were selected based on the minimum chi-square value of the fitted curve. A few initial data points (within 1 ms) were omitted from the fit to account for the dead-time of the instrument. The experiments were done in biological as well as technical triplicates.

Nanodisc-black lipid membrane electrophysiology
Reconstitution of three copies of syb2 with three copies of syt1, syt1 5W or syt7 into nanodiscs and t-SNARE SUVs were performed as previously described (40). In contrast to previous nanodisc-black lipid membrane (ND-BLM) experiments (40), planar lipid bilayers were formed by drying lipids that were resuspended in water, rather than decane. See Figure 5A for an illustration of this procedure. This method was developed as DPPS is not soluble in the typical solvent used to form the bilayers, decane, used in classical approaches. Throughout this manuscript, the same planar lipid bilayer formation strategy was used, regardless of lipid mixture to ensure that samples can be directly compared. Bilayers were formed by resuspending 30 mM of either DOPC/DOPS or DOPC/DPPS in pure water. Five µl of the lipids were mixed with 1 µl t-SNARE reconstituted SUVs (~ 400 µM lipids, 400 nM protein) and pipetted as a droplet over the aperture in the bilayer cup. The bilayer cup, with the lipid droplet is then place . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 11, 2022. ; https://doi.org/10.1101/2022.07.07.499215 doi: bioRxiv preprint horizontally in a vacuum desiccator for 60 minutes to evaporate the water and form a lipid film across the aperture. As soon as the lipid is dried, the bilayer cup is placed in the bilayer chamber and 25 mM HEPES, 10 mM KCl pH 7.4 buffer is added to the trans chamber. To break through the dried lipid, and 'unplug' the cup aperture, press firmly on top of the cup to force a small volume of buffer through the aperture, followed by filling the cis chamber with 25 mM HEPES, 100 mM KCl. Next, use a fine tipped paint brush, dipped in decane, to re-seal the cup aperture and form the planar lipid bilayer. Note that while DPPS is not soluble in pure decane, when the lipids are in the presence of water, this mixture of polar and non-polar solutions facilitates the formation of a bilayer harboring DPPS. After the bilayer is formed, a cis chamber buffer exchange is performed to remove free-floating lipid and t-SNAREs. Subsequent ND-BLM experiments are conducted and analyzed as previously described (40).