
==== Front
bioRxiv
BIORXIV
bioRxiv
2692-8205
Cold Spring Harbor Laboratory

39257774
10.1101/2024.08.30.610324
preprint
1
Article
Sec18 side-loading is essential for universal SNARE recycling across cellular contexts
http://orcid.org/0000-0003-0201-2796
Khan Yousuf A. 12349
Ian White K. 12345
Pfuetzner Richard A. 12345
Singal Bharti 6
Esquivies Luis 12345
Mckenzie Garvey 7
Liu Fang 7
DeLong Katherine 1
Choi Uchoer B. 12345
Montabana Elizabeth 6
Mclaughlin Theresa 7
Wickner William T. 8
http://orcid.org/0000-0001-5121-2036
Brunger Axel T. 123459
1 Department of Molecular and Cellular Physiology, Stanford University, Stanford, CA, USA
2 Department of Neurology and Neurological Sciences, Stanford University, Stanford, CA, USA
3 Department of Structural Biology, Stanford University, Stanford, CA, USA
4 Department of Photon Science, Stanford University, Stanford, CA, USA
5 Howard Hughes Medical Institute, Stanford University, Stanford, CA, USA
6 Stanford Cryo-EM microscopy center, Stanford University, Palo Alto, CA, USA
7 Stanford University Mass Spectrometry, Stanford University, Palo Alto, CA, USA
8 Department of Biochemistry and Cell Biology, Geisel School of Medicine at Dartmouth, Dartmouth College, Hanover, NH 03755.
Author contributions

Y.A.K and A.T.B conceived and designed experiments. K.I.W, B.S., and E.M. assisted Y.A.K. with cryo-EM data collection. Y.A.K. performed and analyzed all experimental data. K.I.W. assisted Y.A.K. with cryo-EM data processing and model building. R.A.F. and L.E. assisted with protein purification. G.M., F.L., and T.M. assisted with high-resolution MS proteomics data collection and analysis. K.D. assisted with bulk disassembly assays. U.B.C. collected and assisted with smFRET data. W.T.W. performed and analyzed vacuolar fusion assays. Y.A.K., A.T.B., and K.I.W. wrote the manuscript.

9 Correspondence: yousuf@stanford.edu (YAK), brunger@stanford.edu (ATB)
01 9 2024
2024.08.30.610324https://creativecommons.org/licenses/by-nc-nd/4.0/ This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which allows reusers to copy and distribute the material in any medium or format in unadapted form only, for noncommercial purposes only, and only so long as attribution is given to the creator.
nihpp-2024.08.30.610324.pdf
Summary

SNARE proteins drive membrane fusion as their core domains zipper into a parallel four-helix bundle1,2. After fusion, these bundles are disassembled by the AAA+ protein Sec18/NSF and its adaptor Sec17/ α-SNAP3,4 to make them available for subsequent rounds of membrane fusion. SNARE domains are often flanked by C-terminal transmembrane or N-terminal domains5. Previous structures of the NSF–α-SNAP–SNARE complex revealed SNARE domain threaded through the D1 ATPase ring6, posing a topological constraint as SNARE transmembrane domains would prevent complete substrate threading as suggested for other AAA+ systems7. Here, in vivo mass-spectrometry reveals N-terminal SNARE domain interactions with Sec18, exacerbating this topological issue. Cryo-EM structures of a yeast SNARE complex, Sec18, and Sec17 in a non-hydrolyzing condition shows SNARE Sso1 threaded through the D1 and D2 ATPase rings of Sec18, with its folded, N-terminal Habc domain interacting with the D2 ring. This domain does not unfold during Sec18/NSF activity. Cryo-EM structures under hydrolyzing conditions revealed substrate-released and substrate-free states of Sec18 with a coordinated opening in the side of the ATPase rings. Thus, Sec18/NSF operates by substrate side-loading and unloading topologically constrained SNARE substrates.

ATPases associated with diverse cellular activities (AAA+)
N-ethylmaleimide-sensitive factor (NSF)
Sec18
Soluble N-ethylmaleimide sensitive factor Attachment protein Receptor (SNARE)
Exocytosis
Quality control
cryo-electron microscopy
single-molecule fluorescence resonance energy transfer
==== Body
pmcCellular compartmentalization, growth, hormone secretion, transport, neurotransmission, and many other pathways depend on precise, rapid, and regulated membrane fusion8,9. Membrane fusion in eukaryotic cells is mediated by a highly conserved superfamily of SNAREs (Soluble N-ethylmaleimide sensitive factor Attachment protein Receptors)5. All SNAREs share a characteristic 60–70 amino acid SNARE domain often flanked by a C-terminal transmembrane domain, membrane anchors, and a folded N-terminal variable domain specific to the SNARE’s function and intracellular pathway10. SNAREs on opposing membranes interact primarily through their SNARE domains to form a parallel trans-SNARE complex, juxtaposing two different membranes11,12. Membrane fusion commences when these SNARE domains zipper together in a directed fashion1,2. After fusion, the SNAREs form a highly stable parallel helical bundle, the so-called cis-SNARE complex1. This highly stable four-helix cis-SNARE complex is disassembled by Sec18 (N-ethylmaleimide-sensitive factor, NSF, in higher eukaryotes)4,13,14 to provide the energy for subsequent membrane fusion events15.

Sec18/NSF is a universally conserved AAA+ (ATPases associated with diverse cellular activities) protein translocase16–18. Sec18/NSF consists of an N-domain, an active D1 AAA+ domain, and a catalytically inactive D2 AAA+ oligomerization domain19. It was initially discovered as a critical complementation group required by the yeast secretory pathway, and it is now known for its role in recycling SNAREs20 and SNARE assembly quality control for proper SNARE complex assembly21–24. Sec18/NSF, with the adaptor protein Sec17 (α-SNAP in higher eukaryotes), disassembles cis-SNARE complexes. The disassembly requires multiple ATP hydrolysis events in the presence of Mg2+ ; as few as 6 ATP hydrolysis events are sufficient25,26.

Cryo-EM structures of the mammalian 20S complex (NSF, α-SNAP, and neuronal SNAREs ternary complex) in a non-hydrolyzing condition (i.e., in the absence of divalent cations) revealed a supramolecular architecture in which the NSF N-domains bind α-SNAP molecules, three or four of which in turn bind the four-helix SNARE complex consisting of the SNARE domains of syntaxin-1A, SNAP-25, and synaptobrevin3,13. The N-terminal residues of the SNARE domain of SNAP-25 was bound to the D1 ring pore without apparent ATP hydrolysis, where it interacts with several conserved tyrosine amino acids in a spiral staircase-like pattern13,16,27. In these EM maps, no ordered density was observed in the D2 ring pore, consistent with the notion that it is catalytically inactive and primarily responsible for NSF oligomerization rather than substrate engagement28. The interaction between the SNARE substrate and the D1 pore is like that observed for other AAA+ translocases and suggests a conserved mechanism for substrate threading through the D1 pore. However, the membrane anchors and domains of the SNAREs would seemingly prevent complete threading of the type suggested for other AAA+ systems7,29,30, posing a topological challenge. Furthermore, SNAREs contain globular N-terminal domain(s) of variable length and structure; for example, the N-terminal domain of syntaxin consists of a three-helix bundle (Habc domain) involved in regulating its function31, and the N-terminal domains of Use1 and Sec20 in part form a stable 255 kDa tethering complex32. Complete threading would imply that such N-terminal domains are somehow unfolded. These topological constraints surrounding SNARE loading, processing, and release through Sec18/NSF are further compounded by the observation that Sec18/NSF disassembles all SNAREs in all cellular contexts5,10,33, which all contain a variety of different N- and C-terminal domains and membrane arrangements/linkages.

In vivo cross-linking mass-spectrometry with Sec18

These questions led us to investigate the space of Sec18/NSF—SNARE interactions through in vivo protein crosslinking mass-spectrometry (XL-MS) in yeast. Due to its power as a model system and relatively simple SNARE proteome, S. cerevisiae is an excellent model for investigating Sec18 interactions with different substrates in live cells. As such, we developed a protocol for in vivo crosslinking mass-spectrometry (XL-MS) in yeast to identify binding partners of Sec18 (Extended Data Figure 1A, Methods).

In total, 35 identified proteins are shared between disuccinimidyl glutarate crosslinker (DSG) treated and untreated conditions (i.e., they are non-specific proteins) (Figure 1A). The nine proteins observed only in the untreated condition had few unique spectra mapping to their identified proteins, suggesting that they were not detected in the treated condition due to their low abundance and difficulty of consistent detection. The 205 proteins unique to the 5 mM DSG treated condition were processed using spatial analysis of functional enrichment (SAFE) analysis to visualize the various cellular processes to which the identified proteins contribute34. At increasing levels of significance thresholds (p = 0.01, 0.001, and 0.0001), the identified proteins only enriched the vesicle trafficking processes. Gene ontology (GO) analysis also revealed lesser-known processes, such as vacuolar acidification and ergosterol biosynthesis (Supplementary Figure 1), consistent with Sec18’s functions in these contexts35–37. The enrichment of these processes validated our crosslinking protocol for specifically targeting Sec18 and its binding partners inside the cell.

Within our enriched protein dataset, we found many yeast SNAREs involved in different pathways and compartments of the cell. We then mapped crosslinked residues of these SNAREs to those of Sec18 (Figure 1B). Using a series of empirical constraints (Methods), we classified a crosslink as high confidence if it met all constraints, medium confidence if it met some but not all, and low confidence if it did not. Since crosslinking was performed in live cells, we expect these crosslinks to represent interactions between SNAREs and Sec18 during substrate loading, disassembly, and substrate release.

Previous cryo-EM structures of the complex of NSF, α-SNAP, and neuronal SNAREs in a non-hydrolyzing condition revealed that the four-helix SNARE bundle interacts with between two and four α-SNAP molecules, which in turn are bound by the N-domains of NSF; the N-terminal end of one of the SNAREs is bound to the pore of the D1 ring3,13,27. Consistent with these structures, we found crosslinks between SNARE domains and the N-domain or D1 domain (examples in Figure 1C, E, F). Additionally, we observed several high-confidence crosslinks of yeast SNARE proteins to the D2 domain, an unexpected result given an absence of interactions between the SNAREs and the D2 ring in the cryo-EM structures and that the D2 domain is catalytically inactive19 and has had no previously reported role in SNARE recycling. Considering this unexpected result, we thus focused on SNAREs with at least one high-confidence crosslink to the D2 domain of Sec18 (Figures 1C–F).

Specifically, seven of ten high-confidence D2 crosslinks connect to regions N-terminal to a SNARE domain (Figures 1C–F). For example, for Sso1, the high-confidence crosslink to D2 involves its Habc domain (Figure 1D, Extended Data Figure 1B). The Habc domain is a stable three-helix bundle31; thus, assuming complete threading through the D1 and D2 pores, the Habc domain (~30 Å diameter) would have to be unfolded transiently. The remaining three high-confidence D2 crosslinks connect to C-terminal regions within SNARE domains. At first glance, this suggests complete substrate threading through both the D1 and D2 rings, akin to other AAA+ protein translocases38. However, considering the topology imposed by membrane domains, anchors, and folded N-terminal domains, how is a SNARE substrate loaded and released? To answer these vexing topological questions, we next determined cryo-EM structures of this orthologous yeast complex together with Sec18 and Sec17.

The Sec18—Sec17—Sso1—Snc1—Sec9 (y20S) complex

To corroborate the surprising LC-MS/MS results, we determined cryo-EM structures of a yeast SNARE complex together with Sec18 and Sec17. We chose the Sso1—Sec9 —Snc1/Snc2 complex (referred to as ySNARE complex) since each component of this complex crosslinked with high confidence to the D2 domain of Sec18 (Snc1 is highly homologous to Snc2). We prepared the Sec18—Sec17—Sso1—Snc1—Sec9 (y20S) complex (Extended Data Figure 2, Methods), and after single particle cryo-EM data collection and processing, we obtained 381,591 high-quality particles that yielded eight 3D classes into which models were built (Figure 2A–E, Supplementary Figure 2A–B, Supplementary Table 1).

In the structures of all eight classes, Sso1 is threaded through the D1 pore of Sec18 (Figure 2A). A characteristic phenylalanine side chain density in Sso1 is present at the same position relative to Sec18 in all classes, allowing for reliable indexing of Sso1. The nucleotide states and the arrangements of D1 around the substrate were also the same among all eight classes, with the ADP-engaged E protomer forming the top of a spiral staircase of tyrosine residues. Protomers D, B, and C were all ATP bound and formed the middle of the staircase. The bottom protomer, A, forms the base of the staircase. Protomer F, with no apparent density for nucleotide in D1, was not associated with Sso1 (Figure 2B). This arrangement of the protomers is driven, in part, by the angle between the more mobile D1 domain and rigid D2 domain. From protomers E to A, the angle between the D1 and D2 decreases monotonically (Figure 2C). The protomer at the top of the staircase maintains the largest angle between D1 and D2, while the one at the bottom has the smallest. The eight classes from the non-hydrolyzing y20S dataset also show differences in the ySNARE-Sec17-Sec18-N-domain subcomplex arrangement (“spire”) and Sec17 stoichiometry above the D1 ring of the y20S complex before disassembly (Figure 2D–E, Extended Data Figure 3B–C; Supplementary Discussion).

The high-confidence crosslinks for Sec18 intra- and inter-protomer interactions are consistent with the cryo-EM structure, specifically with the arrangements of the N-domain and D2-domain (Figure 3A, Extended Data Figure 3A). Moreover, inter-protomer crosslinks were found between pore loop regions, and between the inter-protomer loop between residues 476 and 490 and N-domain linker regions of D1. These high-confidence crosslinks were within the theoretical maximum crosslinking distance (Nζ–Nζ: ~40 Å)39.

Remarkably, the structures reveal that Sso1 is also threaded through the D2 pore and that the Habc domain of Sso1 interacts with the outside of the D2 ring, with class 1 having the most well-defined, discrete density (Figure 3A–E). In class 1, the density for Sso1 within Sec18 is continuous from the D1 pore entrance to the D2 pore exit, where the Habc domain begins (Figure 3A, D). The three-helix bundle of the Habc domain is then positioned outside the D2 pore, where it packs between two small D2 subdomains (protomers A and B) in an interaction driven by charge complementarity. Indeed, the electrostatic potential surface of the Habc domain forms a roughly opposite that of the bottom of the D2 surface against which it packs (Figure 3B–C). Given that this charge distribution is replicated between each pair of D2 subdomains around the D2 surface, it is unsurprising that the Habc domain adopts multiple corresponding rotational states over the eight classes, presumably due to the flexible linker that follows it (Figure 3E). This structural observation is consistent with XL-MS data showing a crosslink between the Habc and the D2 domain of Sec18 (Figure 3A, Extended Data Figure 1B). Although the putative distance between the crosslinked residues (K570 in Sec18 and K111 of Sso1) observed in the model for class 1 (42 Å) is at the limit for a crosslinking distance, the conformational flexibility of the Habc domain to rotate and exist in multiple conformations about the D2 ring (Figure 3E) likely produces conformations well within the crosslinking distance threshold.

Sec18 does not unfold the Sso1/syntaxin Habc domain

Given that this complex was assembled in non-hydrolyzing conditions, we next asked how Sso1 is loaded into Sec18 given the diameter of the Habc domain of Sso1 is ~30 Å whereas the D1 pore of Sec18 has a diameter of ~11 Å. In other AAA+ translocases, such as ClpX, ATP hydrolysis drives complete substrate threading through the AAA+ pore7. However, in the case of Sec18 in a non-hydrolyzing condition, such complete threading would require the Habc domain to unfold without any energetic input from Sec18.

We employed two orthogonal approaches to test if the Sso1 Habc domain unfolds during SNARE disassembly. In our first approach, we turned to a single-molecule Fluorescence Resonance Energy Transfer (smFRET) system developed previously for NSF, α-SNAP, and neuronal SNAREs14 in which individual SNARE domains were synthetically linked and stochastically labeled with FRET pairs. This system allows one to observe multiple rounds of SNARE disassembly as the disassembled SNARE domains readily reassemble into a cis-SNARE complex due to the covalent linkages. We chose this system since it is well-established. Moreover, it is relevant for the Sec18 system studied here since the primary sequence is conserved in the core regions of Sec18/NSF responsible for ATP hydrolysis and substrate processing (Extended Data Figure 4A). In addition, the D1 and D2 rings are structurally conserved when comparing the D1 and D2 rings of Sec18/NSF with engaged substrates, with an average Cα RMSD of 1.64 Å (Extended Data Figure 4B–C). The only protomer that exhibits much difference is the F protomer, which is likely due to its mobility and not due to inherent differences between the structures. Furthermore, we tested the ability of Sec18 and Sec17 to disassemble fluorophore-labeled neuronal SNARE and, vice versa, NSF and α-SNAP to disassemble fluorophore-labeled exocytic ySNARE complex (Extended Data Figure 5A–E, Supplementary Discussion). We found that both could process the species-ortholog SNARE complex, albeit at different rates, further corroborating the interchangeability of NSF and Sec18.

We thus used the smFRET assay to study the effect of NSF on the dynamics of the three-helix bundle Habc domain of syntaxin during disassembly. We either stochastically labeled the SNARE domains or stochastically labeled the Habc domain at two distinct residue positions to monitor either SNARE disassembly or the folded state of the Habc domain (Figure 4A). As in our previous work14, we observed repeated rounds of disassembly and re-assembly for the SNARE domains as indicated by the changes in single FRET intensity over tens of seconds (Figure 4B–C). In stark contrast, we did not observe a change in single-molecule FRET intensity for the labeled Habc domain in non-disassembly and disassembly conditions, producing a FRET intensity distribution consisting only of a high-FRET state (the small peak at 0 is due to traces where acceptor photobleaching occurred since the particular traces never transitioned back to high FRET and no change in the peak intensity between non-disassembly and disassembly conditions was observed) (Figure 4D–F). This result suggests that the Habc domain does not unfold during NSF activity.

To corroborate this finding with an orthogonal approach for the yeast system, we double-crosslinked the Habc domain of Sso1 by incorporating an unnatural amino acid, 4-azido-L-phenylalanine40 (Extended Data Figure 5F–G). This double crosslink was introduced to prevent the Habc domain from unfolding during any Sec18-driven threading. After crosslinking, we performed a disassembly assay on fluorescently labeled un-crosslinked and crosslinked complexes and monitored progress by native gel electrophoresis (Figure 4G). The top band, representing the fully assembled exocytic SNARE complex, is present in both conditions but disappears when adding Mg2+ to initiate hydrolysis and the disassembly reaction. There was no qualitative difference in the disassembly between uncrosslinked and crosslinked conditions after normalizing for labeling efficiency (Figure 4H). Thus, the folded state of the Habc domain does not change during yeast SNARE disassembly, and crosslinking the Habc domain does not affect the disassembly kinetics. Taken together, these results for both the yeast and neuronal systems further argue against a threading model of Sso1 engagement since the Habc domain would be unable to pass through the D1 and D2 domains.

Substrate-released Sec18 structures reveal coordinated ring opening

Together, these observations beg the question of how Sso1 enters the Sec18 ATPase rings. To address this topological challenge, we next determined structures of the y20S complex after initiating Sec18 hydrolysis by adding Mg2+ (referred to as “hydrolyzing condition”). Informed by a fluorescent protein disassembly assay (Extended Data Figure 5H), we initiated the disassembly reaction and waited 7 seconds before sample vitrification.

In the resulting cryo-EM dataset, we found a small number of particles (6,356 or 2.26% of final particles, class 1) that consist of entire y20S assemblies (Figure 5A), largely similar to those observed in the non-hydrolyzing condition (Figure 2). While the resolution of this reconstruction is low (10.88 Å), the D1 ring is flattened relative to reconstructions from the non-hydrolyzing condition (Figure 5B). Considering that our sample consists of purified y20S complexes before initiating disassembly by adding Mg2+, this D1 ring flattening likely occurs due to Mg2+ binding or ATP-hydrolysis at one or more subunits.

The remaining particles (274,883 or 97.74 % of final particles) were substrate-free (classes 2–4 in Figure 5C–E). The D1 ring is also flat in all these substrate-free classes, with all D1 protomers bound to ADP. This uniform binding to ADP, as opposed to the spiral staircase of apo, ADP, and ATP-bound protomers, likely explains the flattened nature of the protomers. In the D2 ring all protomers are still ATP bound, consistent with its role in oligomerization. Class 4 consists of a configuration where both the D1 and D2 rings are heptameric (Figure 5C), and class 3 is similar to this heptameric class, except the 7th protomers is only partially occupied (Figure 5D).

Class 2 has a well-resolved hexameric configuration with a coordinated split in class 2 in the D1 and D2 rings (Figure 5E). This split spans ~20 Å between protomers and is large enough to accommodate a polypeptide chain entering or leaving the rings. This conformational change is likely induced by the nucleotide state, and specifically coupled to the presence of ADP throughout the ring following hydrolysis and SNARE substrate processing; this opening allows the processed SNARE substrate to be released from the side. This side-release mechanism thus solves one part of the topological challenge state above; threading occurs, but the substrate exits from the side, avoiding the membrane domains and linkages in SNARE proteins.

Substrate-free Sec18 structures reveal coordinated ring opening

Next, we asked if this same side-opening of Sec18 occurs in the hydrolyzing condition without SNARE substrate. We prepared purified wild-type Sec18 in the presence of ATP and Mg2+ in a three-stage procedure that yielded a pure sample used for cryo-EM studies (Extended Data Figure 6A–C). Two bands were visible in native gels, indicating two oligomeric states (Extended Data Figure 6D). Processing and classification led to three classes, two of which appeared to be heptameric and identical (280,935 particles and 73,253 particles, respectively; Figure 6A) and the third class, which contained indeterminate density in the D1 ring (70,513 particles). Density for the N-domains was not well resolved in any of the classes, presumably due to the conformational flexibility of the N-domains without substrate or adaptor present.

3D variability analysis (3DVA)41 of the third class suggests that it consists of a mixture of heptameric and hexameric Sec18 (Extended Figure 6E), consistent with the two bands seen on a native gel of the sample. The hexameric class reconstructed from 3DVA particle slices from the first mode (Figure 6B) is like class 2 observed for the y20S complex under hydrolyzing conditions (Figure 5C), wherein hexameric Sec18 has a large, coordinated opening in its D1 and D2 domain rings without substrate. This similarity in Sec18 conformations provides further evidence that the substrate is both loaded and released through the side of the rings, explaining why the Habc domain is not unfolded during Sec18/NSF activity (Figure 3) and that crosslinking the Habc domain does substantially affect the disassembly kinetics (Figure 4H). This solves the second part of the topological question stated above.

The heptameric class (Figure 6A) is nearly identical to class 4 observed in the y20S complex in a hydrolyzing condition (Figure 4A). Each protomer in the heptamer is similar to the others, with an average RMSD of <1Å between them, with the angle between the D1 and D2 rings varying only with a small range of 61.9°–67.1°, in contrast to hexamer, with a range of 60.8°–92.3° (Figure 2C, Extended Data Figure 6F). The D1 pore diameter is 30 Å, larger than substrate-bound hexamer, with a diameter of ~11 Å. The substrate-free structures (hexamer and heptamer) are nearly identical (RMSD = 0.48 Å) to the substrate-released structures.

To test if these D1 ADP-bound, ring-flattened states of Sec18 observed by single particle cryo-EM convert to a disassembly-competent state, we tested the same protein preparation used for cryo-EM with a vacuolar/lysosome proteoliposome fusion assay42,43 (Extended Figure 6H–J, Supplementary Discussion); this assay confirmed that the Sec18 preparation used for cryo-EM was active and processed vacuolar SNAREs. Moreover, NSF also adopts this flattened ring state and a heptameric or hexameric arrangement when exchanged into a buffer containing ATP and Mg2+ without substrate (Extended Data Figure 7, Supplementary Discussion), consistent with our observations that Sec18 and NSF share the same functional mechanism.

Implications for ATPase activity

The observed structures of Sec18 in substrate-free, substrate-loaded, and substrate-released states suggest large conformational change upon ATP hydrolysis, consistent with previous observations3,27,44. In the substrate-free and substrate-released states, examination of the D1 nucleotide binding pockets reveals critical catalytic residues far from the nucleotide (Figure 6C). However, in the substrate-engaged state, these conserved residues are tightly bound to nucleotide in a state preceding the binding of divalent cation required to drive hydrolysis. Employing the same statistical workflow used to cluster the different y20S classes (Methods), we identified the most significant residues that vary between the no substrate and substrate-bound state by generating an ensemble of models for each class via ensemble refinement (residues colored red in Figure 6D, variance contribution per component in Extended Figure 5I). For both the arginine finger loop and Walker B motif, we observed substantial changes to side chain and backbone conformations (Figure 6E–F), wherein coordinating residues are positioned closer to the nucleotide-binding pocket in the presence of substrate, presumably establishing a hydrolysis-ready state. In the case of the arginine fingers, the side chain conformational change is likely coupled to nucleotide identity, as hydrolysis disrupts phosphate coordination with ATP, weakens the interprotomer interface, and leads to rigid body motion of the D1 large subdomain as the arginine fingers disengage. The case of the shift in the Walker B element, on the other hand, appears to be the flexing of the loop without substantial changes to the sidechains of D349 and E350, induced by the motions of pore loop 2 (GVG motif) that bind to the substrate in the pore.

Discussion

Based on our LC-MS/MS, structural, and functional results, we propose the following model for SNARE recycling (Figure 6G), consisting of four phases. Sec18/NSF transiently forms split-open conformation of both D1 and D2 rings in the resting phase with no substrate. Once the cis-SNARE complex has been coated with adaptor Sec17/α-SNAP molecules, the Sec18 N-domains can bind the subcomplex and position the unstructured linker connecting the N-terminal Sso1 Habc domain to the SNARE domain proximal to the D1/D2 split, thus promoting the loading of substrate into the catalytic core of the enzyme. More generally, this side-loading mechanism would allow Sec18/NSF to process complexes incorporating all SNAREs in the cell regardless of the presence of folded N-terminal domains, provided that, in the latter case, a long enough linker connects them to the following SNARE domain. Upon the formation of the 20S particle, ATP-driven hydrolysis likely fuels the disassembly of the SNARE complex by some degree of threading, accompanied by D1 ring flattening. In the case of Sso1, since full threading cannot occur due to the transmembrane domain, partial melting of the N-terminal end of the SNARE domain could be sufficient to destabilize the SNARE complex. Finally, the pore-bound SNARE is released through the side-split of the D1 and D2 rings, regardless of membrane anchors or transmembrane domains in SNARE proteins. This model explains how Sec18/NSF can disassemble SNAREs with constrained topologies.

As observed in our cross-linking LC-MS/MS experiments (Figure 1), the Sec18 D2 ring can interact with not just with the Habc domain of Sso1 but also with other N-terminal accessory domains (GOS1, TLG1, TLG2, etc). In turn, N-terminal domains often interact with other factors. For example, the trimeric Dsl1 complex binds to the SNAREs Use1, Ufe1, and Sec20 through interactions with N-terminal SNARE domains. This 255 kDa complex tethers endoplasmic reticulum membranes together before forming the trans-SNARE complex and likely remains complexed before, during, and after fusion32. Another example is HOPS, a hexameric complex that tethers lysosomal/vacuolar membranes and orchestrates their fusion45. HOPS has two Rab-binding subunits for tethering and a SNARE binding subunit that catalyzes initial SNARE complex assembly46. Sec18/NSF and Sec17/α-SNAP cooperate with HOPS to ensure rapid and efficient SNARE-mediated fusion47,48. For the case of Dsl1, the N-terminal domains of SNAREs remain in a supramolecular arrangement that must be bypassed. For vacuolar/lysosomal fusion, SNAREs must be preprocessed and handed off to the HOPS complex. These complexes pose topological constraints on both ends of the SNAREs that can be resolved by side-loading and side-release.

Finally, we note that distantly related AAA+ clamp loaders and their clamp load their nucleic acid substrate from the side as well49. DNA is loaded into a split opened clamp and AAA+ clamp loader. Once the clamp loader recognizes the substrate, the ATPase is triggered, and the clamp loader closes the complex around the DNA. Once the clamp is fully secured, the clamp-loader is then released. This shared mechanism suggests the possibility of side-loading and release across the entire AAA+ protein family.

In summary, Sec18/NSF can load and release SNARE substrates via a coordinated opening in the D1 and D2 rings. This mechanism does not preclude the more conventional notion of complete substrate threading through the pores of Sec18/NSF if the SNARE substrate is not topologically restricted by membrane domains or anchors or by N-terminal domains or complexes. However, considering these topological considerations are ubiquitous for SNAREs, the side-loading/release mechanism is likely conserved for the entire Sec18/NSF class of AAA+ proteins.

Methods

HA-Sec18 S. cerevisiae strain preparation and treatment

Briefly, we created an S. cerevisiae strain containing HA-tagged Sec18 and treated it with zymolyase and lyticase to dissolve its cell wall but maintain its integrity as a living cell (spheroplast). We then incubated the spheroplasts with disuccinimidyl glutarate (DSG), a membrane permeable protein crosslinking agent, for 30 minutes before cellular lysis and anti-HA co-immunoprecipitation (coIP) to isolate Sec18 and any bound proteins to it. This sample was then subjected to SDS-PAGE gel electrophoresis, and the bands corresponding to crosslinked Sec18 were excised and sent for liquid chromatography-mass spectrometry (LC-MS/MS) analysis. More specifically, tagged Sec18 (pYK175, 3x-Flag N-terminal, HA-tag C-terminal) plasmid was transformed into S. cerevisiae S288C using the Frozen-EZ Yeast Transformation Kit (Zymo Research T2001). Expression of tagged Sec18 was accomplished by growing yeast in 1% raffinose and 1% galactose media. Cells were spun down and treated with 100 units of zymolyase (Zymo Research #1004) at 37°C for 1 hour. After incubation, spheroplasts were pelleted with a 1000 xg spin. DSG diluted in 1x PBS was used to resuspend the pellet, and the sample was incubated for 30 minutes. The excess DSG was then quenched with 1M Tris and incubated for another 15 minutes. Cells were washed with cold PBS and were resuspended in lysis buffer (5mM EDTA, 1mM ATP, 1mM TCEP, 150 mM NaCl, 10% glycerol, 0.5% NP40, and 50 mM HEPES pH 7.4) and glass beads (BioSpec 11079105). This bead-lysis-cell solution was vortexed for 5 minutes at 4°C three times with ice incubations between each vortexing. The solution was then spun down at 10,000 xg for 5 minutes at 4°C. The supernatant was taken and then spun down again at 10,000 xg for 15 minutes at 4°C. The supernatant was then taken and subjected to magnetic-bead anti-HA coIP (Thermo Scientific 88838). The eluant of the beads was run on an SDS-PAGE gel, and the bands corresponding to Sec18 were excised.

XL-MS sample preparation

Protein samples were embedded in Coomassie-stained gel bands, fixed in 1% acetic acid. The gels were washed several times in 50 mM ammonium bicarbonate, followed by reduction with 10 mM of dithriothreitol (DTT) and incubation for twenty minutes at room temperature. They were then alkylated using 30 mM acrylamide for 30 minutes. This was proceeded by overnight digestion at 37 degrees Celsius with 500ng of mass spectrometry grade trypsin/LysC mix (Promega). Post-digestion, samples were quenched with formic acid (adjusted to a pH ~3) and desalted using MonoSpin C18 Solid-Phase Extraction (SPE) columns (GL Sciences). Finally, the samples were dried via SpeedVac (Thermo Scientific, San Jose, CA) and exchanged into LC-MS reconstitution buffer (2% acetonitrile with 0.1% formic acid in water) for instrumental analysis.

LC-MS/MS analysis

Mass spectrometry experiments were performed using an Orbitrap Eclipse Tribrid mass spectrometer (Thermo Scientific, San Jose, CA) attached to an Acquity M-Class UPLC system (Waters Corporation, Milford, MA). The UPLC system was set to a flow rate of 300 nL/min, where mobile phase A was 0.2% formic acid in water and mobile phase B was 0.2% formic acid in acetonitrile. The analytical column was prepared in-house with an I.D. of 100 microns pulled to a nanospray emitter using a P2000 laser puller (Sutter Instrument, Novato, CA). The column was packed with Dr. Maisch 1.9 micron C18 stationary phase to a length of approximately 25 cm. Peptides were directly injected into the column with a gradient of 3–45% mobile phase B, followed by a high-B wash over a total of 80 minutes. The mass spectrometer was operated in a data-dependent mode using CID fragmentation for MS/MS spectra generation.

The RAW data were analyzed using Byonic v5.2.5 (Protein Metrics, Cupertino, CA) to identify peptides and infer proteins. Initial Byonic analyses used a concatenated FASTA file containing the Uniprot Saccharomyces cerevisiae proteins and other likely contaminants and impurities. Once sample complexity was determined, a second round of Byonic analyses was completed using a targeted FASTA file, which included only the sequences present in the samples and allowances for crosslinked peptides with the appropriate linker. Proteolysis with Trypsin/LysC was assumed to be fully specific with up to two missed cleavage sites. The precursor ion tolerance was set to 12 ppm. The fragment ion tolerance was set to 0.4 Da. Cysteine modified with propionamide was set as a fixed modification in the search. Variable modifications included oxidation on methionine and acetylation on protein N-terminus. Proteins were held to a false discovery rate of 1% using the standard reverse-decoy technique (Elias & Gygi Nat. Meth. 2007).

Potential crosslinked peptides were then validated and graded as described previously using Byologic v5.2.31 (Protein Metrics) (Kobilka & Benovic, Cell, 2017). Briefly, a series of empirical constraints were used for inspection, including minimum peptide and crosslink partner length, chromatographic profile, MS1 coelution and MS/MS spectrum fragmentation quality. Following this, identified crosslinked peptides were further categorized into three groups: high confidence, medium confidence, and low confidence, where “high confidence” succeeded on all these rules and “low confidence” failed at least in three of these rules.

Preparation of y20S in the non-hydrolyzing condition

The y20S complex consists of the Sso1–Snc1–Sec9 SNARE (ySNARE) complex, the Sec18 oligomeric complex, and Sec17 adaptor proteins, and it was assembled in a multi-stage workflow. Starting with the yeast exocytic complex, individual SNARE components Sec9 (401–651), Sso1 (1–265), and Snc1 (1–93) were purified individually. All three proteins were cloned into the pET28b plasmid vector with a 6x-His tag and a TEV cleavage sequence N-terminal to the SNARE protein sequence.

For Sec9, we used a lysis buffer of pH 7.5, 50 mM NaPi, 300 mM NaCl, 10 mM Imidazole, 0.5 mM TCEP, 1% Triton X-100 and SIGMAFAST Protease Inhibitor Cocktail Tablet (Sigma-Aldrich S8830) which was used to resuspend bacterial cell pellets from an 8L culture of auto-induced One Shot BL21 (DE3) E. coli (Thermofisher C600003). This resuspended lysate was subjected to sonication for 20 minutes (3 seconds on, 9 seconds off, 60% amplitude), followed by a 30-minute 4000 xg spin and a 1-hour 40,000 xg spin. The supernatant was bound (1 hour, 4°C), run through a Ni-NTA column, and washed with pH 7.5, 50 mM NaPi, 30 mM Imidazole, 300 mM NaCl, and 0.5 mM TCEP buffer. The sample is then eluted with a pH 8 50 mM NaPi, 400 mM Imidazole, 300 mM NaCl, and 0.5 mM TCEP buffer. This sample is then digested overnight with TEV protease in a dialysis buffer (pH 8 20 mM Tris, 250 mM NaCl, 0.5 mM TCEP, and 1 mM EDTA). The cleaved sample was then subjected to a MonoS 5/50 GL with buffer A consisting of pH 8 20 mM Tris, 50 mM NaCl, 0.5 mM TCEP, 1mM EDTA and with buffer B consisting of pH 8 20 mM Tris, 500 mM NaCl, 0.5 mM TCEP, and 1mM EDTA. Taking the majority peak corresponding to Sec9, we then subjected this peak to size exclusion chromatography (SEC) with the HiLoad 16/60 Superdex 200. The peak corresponding to Sec9 was concentrated and flash-frozen. We followed the same protocol for individually purifying Snc1 and Sso1.

Once all three individual SNAREs were purified, a 1:1:1 molar ratio of these proteins was added to a 6M GdHCl solution. This solution was then slowly dialyzed overnight at 4°C into a solution of 250 mM NaCl, 50 mM HEPES pH 7.6. This sample was then diluted to a NaCl concentration of 75 mM and then subjected to MonoQ with the low salt buffer at 75 mM and the high salt buffer at 500 mM with both buffers in 50 mM HEPES pH 7.6. The peak corresponding to a 1:1:1 ratio of all three bands, indicative of the yeast exocytic complex, was then concentrated for y20S formation without freezing.

The complete Sec18 protein coding sequence was cloned into the pMZ0002/pYK103 backbone with a 6x-His tag and TEV cleavage sequence N-terminal to Sec18. The 8L culture of One Shot BL21 (DE3) E. coli (Thermofisher C600003) was spun down and resuspended in lysis buffer of pH 7.5 100 mM HEPES, 500 mM KCl, 5 mM ATP, 5mM MgCl2, protease inhibitor tablets, and benzonase. This lysate was sonicated for 20 minutes (3 seconds on, 9 seconds off, 60% amplitude) and clarified with a 45,000 xg spin for 30 minutes. The lysate was supplemented with 20 mM Imidazole bound to Ni-NTA beads (4°C) and washed with pH 7.5 20 mM HEPES, 480 mM KCl, 0.5 mM ATP, 0.5 mM TCEP, 20 mM Imidazole, 1 mM MgCl2, and 10% glycerol. The sample was then eluted with pH 7.5 20 mM HEPES, 480 KCl, 0.5 mM ATP, 0.5 mM TCEP, 500 mM Imidazole, 1 mM MgCl2, and 10% glycerol. The protein-containing peak was then equilibrated in SEC buffer pH 6.8 20 mM PIPES, 125 mM KCl, 0.2M sorbitol, 5 mM MgCl2, 2 mM ATP, 2 mM DTT, and 10% glycerol and injected a HiLoad 16/60 Superdex 200. The fractions were pooled, concentrated, and snap-frozen for later usage. The band right below Sec18 is an oligomeric E. coli contaminant, which is removed in the final steps of y20S complex purification (Figure S2A, C). Moreover, this contaminant is absent in the protocol for preparing high-purity Sec18 (see below and Figure S6B).

The full Sec17 coding sequence was cloned with a 10x-His tag and a TEV cleavage sequence fused N-terminally. The 8L culture of One Shot BL21 (DE3) E. coli was spun down and resuspended in a lysis buffer of pH 8 50 mM NaPi, 300 mM NaCl, 20 mM Imidazole, and 0.5 mM TCEP. This mixture was then sonicated (3 seconds on, 9 seconds off, 60% amplitude) and clarified at 45,000 xg for 30 minutes. This supernatant was then incubated with Ni-NTA beads for an hour and then washed with lysis buffer. The protein was eluted with lysis buffer supplemented with 500 mM Imidazole and 5 mM EDTA. TEV protease was added to pooled fractions and slowly dialyzed overnight at 4°C against a pH 8 50 mM Tris, 100 mM NaCl, 1 mM EDTA, 0.5 mM TCEP buffer. This sample was then diluted to a NaCl concentration of 50 mM and run on a MonoQ 10/100 from a 50 mM to 500 mM salt gradient. The major peak corresponding to Sec17 was then pooled and injected into a HiLoad 16/60 Superdex 200 SEC. The major peak from this run was taken, pooled, and snap-frozen for later usage.

To assemble the entire y20S complex, Sec17 was added to the freshly prepared yeast exocytic SNARE complex, followed by EDTA quenched Sec18 with a final ratio of Sec18:SNARE:Sec17 at 1:1.67:10, where a total of 5580 picomoles of Sec18 was used, in the following buffer: pH 8 50mM Tris, 150 mM NaCl, 1 mM TCEP, 1 mM ATP, and 1 mM EDTA. Following this assembly, this mixture was injected onto a Superose 6 10/300 increase column, and the peak corresponding to the y20S complex, as assessed by elution volume and corresponding SDS-PAGE gel, was concentrated to ~40 mg/mL.

Single particle Cryo-EM grid preparation of y20S in the non-hydrolyzing condition

Quantifoil R1.2/1.3 200 mesh gold grids were treated with chloroform and dried overnight. Grids were glow discharged, and 5 microliters of the sample (at a final concentration of 20 mg/mL with 0.05% v/v Nonidet P-40) was blotted onto the grids and further blotted and vitrified in liquid ethane using an FEI Vitrobot (ThermoFisher Scientific). Grids were blotted for 4 seconds with a 5-second wait time (reduced to 3 seconds in hydrolyzing condition).

Sample preparation and single particle Cryo-EM of y20S in the hydrolyzing condition

Samples were prepared identically to y20S in the non-hydrolyzing condition, except that excess MgCl2 was blotted onto grids immediately before vitrification, and the total blot and wait time was reduced to 7 seconds before freezing.

Sample preparation and single particle Cryo-EM of substrate-free Sec18 in the hydrolyzing condition

To prepare high-purity, substrate-free Sec18, an alternative protocol was required since low-purity Sec18, which can form y20S, was not amenable to high-quality cryo-EM studies. We used a protocol similar NSF18, but at the final reassembly step, we used MgCl2 instead of EDTA. The cryo-EM studies were performed similarly to the studies of Y20S in the non-hydrolyzing condition.

Sample preparation and single particle Cryo-EM of NSF in the hydrolyzing condition

A sample of high-purity NSF was prepared, as described in the previous section. This sample was exchanged into a buffer supplemented with 1mM magnesium chloride instead of EDTA and incubated before a final SEC and freezing.

Single particle cryo-EM data collection, processing, and model building

Single particle cryo-EM data collection and processing workflow are described thoroughly in the Extended Table 1 and Supplementary Figure 2. An FEI Titan Krios (ThermoFisher Scientific) cryo-electron microscope with either a K3 Gatan or a Falcon F4i camera was used for data collection. Micrographs were analyzed using a combination of initial Relion 3.151 processing followed by additional rounds of 3D classification or heterogenous refinement, and final refinements in CryoSPARC v452. A general workflow description is provided in Supplementary Figure 2, and specific dataset information is provided in Supplementary Figures 2–6 as well.

y20S models were constructed by first docking the crystal structure of the exocytic SNARE complex (PDB ID: 3B5N) and AlphaFold253 models for Sec17 into the unsharpened density first obtained in Relion 3.151. Next, sharpened cryoSPARC-v452 maps were then used to model the Sec18 complex and the region of Sso1 that extended from the spire into the D1 pore. For class 1, 3DFlex analysis in cryoSPARC-v4 was performed. A mixture of the Relion 3.1 and cryoSPARC-v4 unsharpened and sharpened maps, and the 3Dflex map were used to model the Habc domain outside of the D2 ring. The resulting atomic models were iteratively refined with a combination of COOT54, Phenix real-space refinement55, and ISOLDE56.

Models of substrate-free Sec18 and NSF were built de novo in COOT using cryoSPARC sharpened maps, and iteratively refined with ISOLDE and Phenix.

smFRET disassembly/re-assembly assay

smFRET assays with linked neuronal SNAREs, NSF, α-SNAP, and reagents were prepared as previously reported14. Briefly, linked SNARE complex (L-SNARE) composed of SNAP25A, synaptobrevin-2, and syntaxin-1A were biotinylated and flowed onto the chamber that was passivated by 10 mg/mL egg phosphatidylcholine 50 nm liposomes coated with 1 mg/mL of biotinylated BSA to mimic the lipid environment inside the cell. 0.1 mg/mL streptavidin was added to surface-tether the biotinylated and labeled L-SNARE complexes. L-20S was then assembled by first adding αSNAP and then later NSF (if in disassembly conditions). L-SNARE was diluted such that about 500 molecules of L-SNARE per 45 × 90 µm2 field of view was visible. The raw data was imported into tMaven44 for analysis. The standard, recommended tMaven workflow was employed. Traces were not subject to any preprocessing workflow. Automatic photobleach detection was used to determine the locations of photobleaching, and a vbGMM mixture model was used to detect the different states.

Crosslinking assays

Cysteines were introduced into Sec9 and Sso1 for labeling with Oregon Green Maleimide 488 (O6034). UAG stop codons were introduced at V57 and F83 in Sso1 for crosslinking. Cys-Sec9, Cys-Sso1 (with or without stop codon depending on if used for crosslinked), and 6x-His-TEV-Snc1 were co-expressed (if crosslinked, also transformed with pEVOL-pAzF AddGene #31186, 2mM AzF and 0.2% v/v arabinose) and induced with IPTG. Cell pellets were lysed with pH 7 50 mM Tris, 200 mM NaCl, 1% v/v Triton X-100, 20 mM Imidazole, 0.5 mM TCEP and protease inhibitor tablets. Lysate was sonicated for 20 minutes (3 seconds on, 9 seconds off, 60% amplitude) and then clarified at 40,000 xg for 30 minutes. Following a 1-hour incubation at 4°C, the beads were washed with two different buffers. Buffer 1 is simply a lysis buffer without Triton X-100, and Buffer 2 is a high salt wash of pH 7 50 mM Tris, 1M NaCl, 50 mM Imidazole, 0.5 mM TCEP. The complex was eluted with pH 7 50 mM Tris, 200 mM NaCl, 350 mM Imidazole, and 0.5 mM TCEP. The sample was then digested overnight with TEV protease and dialyzed into pH 7 20 mM Tris, 200 mM NaCl, 0.5 mM TCEP. This sample was then injected onto a HiLoad 16/60 Superdex 200. The fractions containing all three in equimolar ratios were taken and concentrated. Samples were labeled with Oregon Green Maleimide 488 in molar excess overnight before cleanup with a Zeba spin column (ThermoFisher 89889). If the sample was meant to be crosslinked, it was subjected to UV irradiation before Oregon Green Maleimide labeling.

For native gel disassembly assays, reactions were assembled in test tubes, initiated with MgCl2, and quenched with EDTA. Samples were then loaded into Any kD Mini-Protean gels (Bio-Rad 4569033) and used Tris-glycine running buffer (Bio-Rad ) supplemented with 1 mM ATP to prevent Sec18 from falling apart in the gel. Gels were first visualized in an iBright 1500 (ThermoFisher Scientific) to visualize ySNARE bands only before coomassie staining. Gel densitometry was assessed in ImageJ.

FRET and fluorescent dequenching disassembly assays

Neuronal disassembly assays were carried out as previously reported10,45, leveraging fluorescence dequenching of the neuronal SNARE complex labeled with Oregon Green Maleimide 488. For yeast SNARE disassembly assays, a strategy was devised to tag Snc1 with mTurq and Sso1 with mVenus. When assembled, excitation by a 434 nm laser leads to mVenus emission. 434 nm laser excitation leads to mTurq emission. Disassembly is followed by monitoring excitation/emission at 434/474 nm such that as the SNARE complex is disassembled, mTurq emission increases. We used the same instrument (FlexStation II 384, Molecular Devices) and a 384-well format to perform both the dequenching and FRET assays that were initiated by titrating MgCl2 in excess of the reactions.

Structural analysis

Statistical structural analysis was performed to identify the most significant backbone motions between the many different structural states observed. Using Python 3, principle component analysis (PCA) on backbone ɸ/ψ angles was performed, and the single mode that contributed most to the variance (>70%) was analyzed. y20S models were then clustered in this reduced 5-dimensional space. When comparing the substrate-free and substrate-engaged states, ensembles consistent with a given map were generated to avoid model-building bias using Phenix’s simple molecular dynamics implementation for 2000 steps followed by a round of phenix.refine55. These ensembles were then subjected to backbone ɸ/ψ PCA, and the mode that contributed to the largest variance between the two clusters for the two classes was used to determine which residues contributed most to the difference in structures.

Vacuolar fusion assay

Assays were performed as previously reported42,43 with the same substrate-free Sec18 protein sample/batch that was purified and used for CryoEM. Briefly, proteoliposomes with vacuolar SNAREs were prepared with either Phycoerythirin-biotin (PhycoE) or Cy5 and incubated in a fluorescence plate reader for 40 minutes. FRET between PhycoE and Cy5 was measured in intervals of 5–15s. Reaction mixtures contained HOPS complex, Sec17, and variable amounts of Sec18.

Extended Data

Extended Data Figure 1: In vivo mass-spec workflow and additional analysis

a, Workflow of how samples were processed before mass-spec. b, Representative mass-spectrometry raw data of high confidence (Methods), suggesting that the Habc domain is crosslinked to the D2 domain. In the sequence diagram, the cross-linked lysine residues are shown in orange.

Extended Data Figure 2: Purification and assembly of y20S complex

a, Sec18 SDS-PAGE gel after the final SEC run. Pooled fractions were frozen and stored. b, Sec17 SDS-PAGE gel after final SEC run. Pooled fractions were frozen and stored. c, Schematic of y20S assembly. Snc1, Sso1, and Sec9 were purified individually before being co-folded for complex formation (Methods). The resulting SNARE complex was added to purified Sec17, EDTA was added, and then purified Sec18 was added. The resulting y20S complex was purified by SEC, and fractions were pooled. The final gel shows reference inputs for Sec18, Sec17, and yeast SNARE complex, the final concentrated y20S complex, and the fractions from SEC that were pooled (yellow lines).

Extended Data Figure 3: Intra and inter-protomer crosslinks mapped on the y20S complex structure

a, Two Sec18 protomers of the y20S complex are shown. The black dashed lines represent crosslinks designated as excellent (passing all manual verifications) and superimposed over the diagram. The two insets show inter-protomer crosslinks. b, Electrostatic surface potentials of the interaction between SNAREs and Sec17. Negatively charged regions are colored red, and positive ones are colored blue. c, Electrostatic surface potentials of the interaction between Sec17 and the N-domains of Sec18. Alternating views of these interactions are shown to demonstrate the complementarity of the electrostatic interactions.

Extended Data Figure 4: Sequence and structural conservation between NSF and Sec18

a, Primary sequence alignment between NSF c. griseus and Sec18 s. cerevisiae. b, cartoon representation of an atomic model of Sec18 D1/D2 focused model colored by Cα root-mean-square-difference (RMSD) to the structure of the binary SNARE complex with NSF and α-SNAP27. c, Focused view of substrate in D1. The substrate bound to the D1 pore is represented as spheres.

Extended Data Figure 5: Cross-complementation disassembly assays

a, Scheme of fluorescent-protein labeling. Snc1 was labeled with mTurq, and SSO1 was labeled with mVenus (Methods, referred to as fpySNARE). b, SDS Page gel of the final SEC purification of y20S complex. c, Representative traces from the disassembly assay (Methods). The mTurq emission is monitored, and an increase in emission as the complex suggests that SNARE disassembly occurs (loss of fluorescence resonance energy transfer to mVenus upon disassembly). d-e, Disassembly plots of SNAREs and NSF/Sec18 combinations. f, Residues that were replaced with p-azido-l-phenylalanine are shown as yellow spheres in a black box g, MS raw data of double crosslinked yeast exocytic SNARE complex. h, Michaelis-Menten diagram of the y20S disassembly reaction. The red line shows the fitted line used to calculate M-M metrics, whereas the black points and error bars show the actual data. i, Variance contributed by each principal component.

Extended Data Figure 6: Functional studies of Sec18.

a, Schematic of purification of Sec18. b, SDS-PAGE gel of final SEC fractions and the fractions that were pooled for CryoEM studies of the y20S complex and the vacuolar fusion assay. c, Corresponding SEC trace showing the Sec18 oligomer eluting at expected volume. d, Native gel of final sample showing two oligomeric species. e, Representative slices of class 3 of the y20S complex under the hydrolyzing condition, subjected to 3DVA revealing split hexamer class. f, Alignment of all seven protomers, where each protomer is colored uniquely. g, Cartoon depicting the large pore in the heptameric state of the y20S complex under the hydrolyzing condition. h, Cartoon of the vacuolar fusion assay 42,43 (Methods) used to determine the function of the same Sec18 sample used for CryoEM studies of the y20S complex. Fusion only occurs when Sec18 prepares SNAREs for fusion, leading to fluorescence. i, Representative traces of fusion activity. j, Dot-plots of all activity assays.

Extended Data Figure 7: CryoEM structures of substrate-free NSF in the hydrolyzing condition

a-d, CryoEM maps (sharpened cryoSPARC) in three orientations (top to bottom rows) of the four classes of NSF in the hydrolyzing condition. The NSF heptamer (class 1) is designated as class 1. Classes 2–4 showed hexameric NSF in similar orientations with varying resolution for the substrate, so the best one, class 3 (initial) (c), was chosen for 3DVA analysis in CryoSPARC. e, The main principal component of the 3DVA analysis was used to select a subset of particles clearly showing an NSF N-terminal domain that is engaged with the D2 ring and then re-refined against this subset of particles to yield the final map for class 3. f-g, insets showing that an N-domain acts as a bound substrate in class 3 (final).

Extended Data Table 1 | Cryo-EM data collection, refinement, and validation statistics

	Y20S, non-hydrolyzing condition	Y20S, hydrolyzing condition	Sec18, hydrolyzing condition	NSF, hydrolyzing condition	
	#1 Class 1 (EMDB-) (PDB )	#2 Class 2 (EMDB-) (PDB )	#3 Class 3 (EMDB-) (PDB )	#4 Class 4 (EMDB-) (PDB )	#5 Class 5 (EMDB-) (PDB )	#6 Class 6 (EMDB-) (PDB )	#7 Class 7 (EMDB-) (PDB )	#8 Class 8 (EMDB-) (PDB )	#9 All Merge D1/D2 (EMDB-) (PDB )	#10 Class 1 (no model deposited)	#11 Class 2 (EMDB-) (PDB )	#12 Class 3 (EMDB-) (PDB )	#13 Class 4 (EMDB-) (PDB )	#14 Class 1 (EMDB-) (PDB )	#15 Class 2 (EMDB-) (PDB )	#16 Class 1 (EMDB-) (PDB )	#17 Class 2 (EMDB-) (PDB )	
Data collection and processing																		
Magnification	130,000×	130,000×	130,000×	130,000×	130,000×	130,000×	130,000×	130,000×	130,000x	130,000x	130,000x	130,000x	130,000x	130,000x	130,000x	130,000x	130,000x	
Camera	Gatan K3	Gatan K3	Gatan K3	Gatan K3	Gatan K3	Gatan K3	Gatan K3	Gatan K3	Gatan K3	Falcon 4i	Falcon 4i	Falcon 4i	Falcon 4i	Gatan K3	Gatan K3	Gatan K3	Gatan K3	
Voltage (kV)	300	300	300	300	300	300	300	300	300	300	300	300	300	300	300	300	300	
Electron exposure (e–/Å 2 )	26.8	26.8	26.8	26.8	26.8	26.8	26.8	26.8	26.8	51.48	51.48	51.48	51.48	25.5	25.5	39.04	39.04	
Defocus range (μm)	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	−1.0 to −2.0	
Pixel size (Å, super-res)	0.548	0.548	0.548	0.548	0.548	0.548	0.548	0.548	0.548	0.49	0.49	0.49	0.49	0.548	0.548	0.548	0.548	
Exposure time (seconds)	4	4	4	4	4	4	4	4	4	6.93	6.93	6.93	6.93	4	4	4	4	
Number of frames per exposure	80	80	80	80	80	80	80	80	80	2133 (EER fractionation of 42)	2133 (EER fractionation of 42)	2133 (EER fractionation of 42)	2133 (EER fractionation of 42)	80	80	80	80	
Number of movies	12,439	12,439	12,439	12,439	12,439	12,439	12,439	12,439	12,439	14455	14455	14455	14455	7821	7821	13375	13375	
Symmetry imposed	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	C1	
Final particle images (no.)	41,159	35,723	79,060	26,603	33,084	64,857	69,608	31,497	381,591	6,356	86,542	91,962	96,291	280,935	16,074	72,069	11,844	
Map resolution (Å)	3.89	3.91	3.73	4.61	4.29	3.73	3.75	4.13	3.4	10.88	3.38	3.36	3.18	2.99	6.09	3.60	4.70	
FSC threshold	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	0.143	
Model information																		
Initial model used (PDB code)	3B5N, Sec17/18 from AlphaFold2	3B5N, Sec17/18 from AlphaFold2	3B5N, Sec17/18 from AlphaFold2	3B5N, Sec17/18 from AlphaFold2	3B5N, Sec17/18 from AlphaFold2	3B5N, Sec17/18 from AlphaFold2	3B5N, Sec17/18 from AlphaFold2	3B5N, Sec17/18 from AlphaFold2	Sec18 from AlphaFold2	Y20S EDTA Class 1	De novo	De novo	De novo	De novo	De novo	3J94	3J94	
Bond RMSD (Å)	0.12	0.012	0.007	0.007	0.010	0.005	0.005	0.010	0.006	No model deposited	0.004	0.013	0.007	0.007	0.006	0.006	0.007	
Angle RMSD (°)	1.461	1.473	1.186	1.168	1.206	1.023	1.027	1.243	0.920	No model deposited	1.203	1.444	1.186	1.160	1.090	0.715	1.313	
Molprobity score	1.68	1.79	1.58	1.79	1.72	1.73	1.62	1.76	1.55	No model deposited	1.23	1.78	1.78	1.51	1.80	2.00	1.72	
Clashscore, all atoms	5.77	8.55	5.80	8.26	8.37	9.22	6.68	7.71	4.01	No model deposited	2.67	8.27	4.49	4.52	9.44	5.37	9.81	
Ramachandran. Favored/Allowed/Disfacored (%)	94.77
4.85
0.38	95.35
4.49
0.16	96.15
3.75
0.10	95.12
4.54
0.34	96.11
3.69
0.20	96.41
3.45
0.14	96.24
3.67
0.08	95.81
3.88
0.31	94.80
5.01
0.19	No model deposited	96.95
2.92
0.14	95.16
4.65
0.18	95.13
4.58
0.29	95.95
4.05
0.00	95.73
4.17
0.00	95.64
4.36
0.00	97.17
2.26
0.57	
Rotamer outliers (%)	0.61	0.44	0.43	0.69	0.88	0.55	0.52	1.14	0.55	No model deposited	0.08	0.00	1.96	0.93	0.97	3.45	1.19	
Cβ outliers (%)	0.02	0.04	0.00	0.04	0.02	0.02	0.00	0.02	0.00	No model deposited	0.04	0.10	0.00	0.00	0.00	0.00	0.00	
CaBLAM outliers (%)	2.36	2.08	2.06	2.22	2.09	1.99	2.15	2.14	3.06	No model deposited	0.93	1.36	1.20	0.99	0.99	1.44	1.46	

Supplementary Material

Supplement 1

Acknowledgments

We thank William I. Weis and Joseph D. Puglisi for stimulating discussions. We thank the National Institutes of Mental Health (NIMH) for support (RO1MH63105 to A.T.B., 1F31MH134477 to Y.A.K.). Y.A.K was additionally supported in part by an NSF GRFP and the Knight-Hennessy Scholarship. K.I.W. was supported by a postdoctoral fellowship from the Helen Hay Whitney Foundation, supported by the Howard Hughes Medical Institute. W.T.W. is supported by a grant from the NIH (2R35GM118037). We thank the Vincent Coates Foundation Mass Spectrometry Laboratory, Stanford University Mass Spectrometry (RRID:SCR_017801) for utilizing the Thermo Orbitrap Eclipse nanoLC/MS system (RRID:SCR_022212) that was purchased with funding from the National Institutes of Health Shared Instrumentation Grant 1S10OD030473, the Stanford Cancer Institute Proteomics/Mass Spectrometry Shared Resource (NIH P30 CA124435).

Data availability statement

The cryo-EM maps will be deposited at the Electron Microscopy Data Bank (EMDB) and the coordinates will be deposited in the Protein Data Bank. Motion-corrected micrographs will be deposited in the Electron Microscopy Public Image Archive (EMPIAR). The data will be made publicly available as of the date of publication. Original code reported in this paper will be deposited on a publicly available GitHub page. The mass spectroscopy data will be deposited in a publicly available repository.

Figure 1: In vivo crosslinking mass-spectrometry of Sec18

a, Top left: Venn diagram of proteins found in the condition without crosslinking agent (red) added and with crosslinking agent added (green). Bottom right: SAFE-analysis34 of yeast proteins that specifically crosslink to Sec18. Blobs represent enrichments in the pathway. Blobs of different colors represent enrichments that are specific to different p-value thresholds (brown < 0.01, blue < 0.001, red < 0.0001). b, Heatmap of yeast SNARE proteins crosslinking to specific Sec18 positions. The Sec18 primary sequence along the x-axis is colored coded by domain, and yeast SNARE proteins with at least one medium confidence crosslink are shown along the y-axis. Red squares represent positions where one or more manually verified, high-confidence crosslinks were found (Methods). Orange squares represent where one or more manually verified medium-confidence crosslinks were found. White squares represent positions in which manually verified crosslinks were not found or were of low confidence. c-f, Crosslinking schematic of Sec18 and SNAREs that contain at least one high-confidence crosslink to the Sec18 D2 domain. Red dashed lines represent high-confidence crosslinks, and orange dashed lines represent medium-confidence crosslinks.

Figure 2: The y20S supramolecular complex

a, Domain diagrams of the components used to prepare the y20S complex; coloring follows this throughout (top left). The structure of class 1 of y20S, with ATP and ADP nucleotides (red and blue, respectively; bottom right). Cryo-EM density for class 1 and the Sso1 atomic model around residue F173 (inset, bottom left). b, Sso1 substrate is engaged by Y315 in all D1 protomers with bound nucleotide. c, Conformations of the protomers of Sec18 when engaged to substrate. D, Top-down views of class 1 and class 8 cryo-EM maps. e, Cartoon top-down views of eight y20S classes, which differ primarily by spire configuration, i.e., the pattern of N-domain and ɑ-SNAP engagement.

Figure 3: Sso1 is threaded through the D1 and D2 domains

a, y20S class 1 atomic model emphasizing the position of Sso1. Stars represent the crosslink identified by crosslinking mass spectrometry (Figure 1B). b, Bottom view of y20S class 1 D1 ring; the Habc domain is packed against the D2 ring surface, between protrusions corresponding to the C-terminal helical regions of two D2 small subdomains. All surfaces are colored grey, except for protomers A and B, which are colored by their electrostatic potential c, Rotated view of y20S class 1 relative to b showing the Habc tucked between protomers A and B with now the Habc domain colored by its electrostatic potential d, Composite cryoEM map (blue, Relion-unsharpened map: Habc domain and D1 density and CryoSPARC 3DFlex map: D2 density) of y20S class 1. Nucleotides are shown as red (ATP) or blue (ADP) spheres. e, Cryo-EM maps of all eight y20S classes reveal the Habc domain associated with the Sec18 D2 ring in all classes (red). In 6/8 classes, the Habc density is radially averaged into multiple discrete positions around the D2 ring, leading to more diffuse density.

Figure 4: Sec18/NSF does not unfold the Habc domain

a, Diagram of the smFRET disassembly/re-assembly assay with linked neuronal SNAREs. The linked SNAREs are either stochastically labeled within the SNARE domains or the syntaxin Habc domain. b, Representative traces of labeled neuronal SNARE complex in the non-hydrolyzing condition. Violet represents the acceptor, and orange represents the donor-dye fluorescence-intensity time traces. n = 53. c, Representative time trace of labeled SNARE complex in disassembly conditions, the black line represents fitting by a vbGMM50. n = 114. d, Representative time-trace of labeled Habc domain in the non-hydrolyzing condition. Green represents the donor, and red is the acceptor dye fluorescence intensity time-traces. n = 27. e, Representative time-trace of labeled Habc domain in disassembly conditions. The black line represents fitting a vbGMM. n = 22. f, Histograms of E-FRET for all four conditions. The blue dotted line represents a Gaussian fit of the data. The y-axis is the normalized frequency of occurrence of the E-FRET states in the traces. Top left: SNARE E-FRET for the non-hydrolyzing condition, top right: SNARE E-FRET for the hydrolyzing condition. Habc E-FRET for the non-hydrolyzing condition (bottom left). Habc E-FRET for the hydrolyzing condition (bottom right). g, Native gel of labeled uncrosslinked or crosslinked yeast SNARE complex that has been labeled with Oregon Green Maleimide 488. h, Quantification of gel densitometry of n = 3 independent disassembly experiments. Error bars represent the standard deviation of the mean. Densitometries were all normalized to 0 min values.

Figure 5: Post-disassembly of y20S reveals substrate-released states

a, Cryo-EM map of y20S in the hydrolyzing condition, colored by domain. b, The distance between the D1 and D2 domains decreases for some protomers under hydrolyzing conditions, leading to D1 ring flattening. The orange model corresponds to protomer E from the merged D1/D2 focused class in the non-hydrolyzing condition, which was rigid-body fit into the density of the hydrolyzing condition to generate the blue model. c, Cryo-EM map and model of class 4 of Sec18, showing a heptamer. d, Cryo-EM map and atomic model of class 3 of Sec18, suggesting a transition state. e, Cryo-EM map and atomic model of class 2 of Sec18, showing a split hexamer. In panels c-e, the top is the EM-map, while the middle and bottom show atomic models in top and side views, respectively.

Figure 6: Structures of substrate-free Sec18 in the hydrolyzing condition

a, Atomic model of the substrate-free heptameric state of Sec18 (class 1). b, Atomic model of the split substrate-free hexameric state of Sec18 (derived from 3DVA of class 3). c, The D1 nucleotide binding pocket is remodeled as a function of SNARE substrate binding and nucleotide state. The arginine fingers R406 and R409 are contributed by the neighboring protomer. The substrate-free and substrate-released states are nearly identical in conformation. Water molecules are shown as red spheres. d, Atomic model of Sec18 with protomer A colored blue if not a significant residue from conformational analysis and colored red if significant (p-value < 0.05). e, Atomic models showing the difference of the arginine finger loop in the no substrate (transparent) and substrate condition. f, Atomic models of the loop proximal to the Walker B motif in the no substrate (transparent) and substrate condition. For d-f, residues are colored red if they are in the top 5% of residues that significantly vary between conditions. g. We propose a general model of SNARE recycling, regardless of the cellular context. First, a cis-SNARE complex is coated with at least 1 Sec17/α-SNAP adaptor molecule, allowing Sec18/NSF to recognize it. The SNARE substrate is then loaded into a split hexamer with a coordinated opening in both D1 and D2 rings through the side, bypassing whatever N-terminal domains may be present. The substrate is then threaded coaxially through the Sec18/NSF pore. Upon completion of processing, the SNARE substrate is released through the side, bypassing the topological constraints of the membrane. Finally, Sec18 returns to its ‘resting’ state until more substrate is encountered. Yellow arrows indicate the direction of loading and release.
==== Refs
References

1. Sutton R. B. , Fasshauer D. , Jahn R. & Brunger A. T. Crystal structure of a SNARE complex involved in synaptic exocytosis at 2.4 Å resolution. Nature 395 , 347–353 (1998).9759724
2. Weber T. SNAREpins: Minimal Machinery for Membrane Fusion. Cell 92 , 759–772 (1998).9529252
3. Zhao M. Mechanistic insights into the recycling machine of the SNARE complex. Nature 518 , 61–67 (2015).25581794
4. Söllner T. , Bennett M. K. , Whiteheart S. W. , Scheller R. H. & Rothman J. E. A protein assembly-disassembly pathway in vitro that may correspond to sequential steps of synaptic vesicle docking, activation, and fusion. Cell 75 , 409–418 (1993).8221884
5. Jahn R. , Cafiso D. C. & Tamm L. K. Mechanisms of SNARE proteins in membrane fusion. Nat. Rev. Mol. Cell Biol. 1–18 (2023) doi:10.1038/s41580-023-00668-x.36319695
6. White K. I. , Zhao M. , Choi U. B. , Pfuetzner R. A. & Brunger A. T. Structural principles of SNARE complex recognition by the AAA+ protein NSF. eLife 7 , e38888 (2018).30198481
7. Maillard R. A. ClpX(P) Generates Mechanical Force to Unfold and Translocate Its Protein Substrates. Cell 145 , 459–469 (2011).21529717
8. Wickner W. & Schekman R. Membrane fusion. Nat. Struct. Mol. Biol. 15 , 658–664 (2008).18618939
9. Südhof T. C. Neurotransmitter Release: The Last Millisecond in the Life of a Synaptic Vesicle. Neuron 80 , 675–690 (2013).24183019
10. Ungar D. & Hughson F. M. SNARE Protein Structure and Function. Annu. Rev. Cell Dev. Biol. 19 , 493–517 (2003).14570579
11. Rizo J. Molecular Mechanisms Underlying Neurotransmitter Release. Annu. Rev. Biophys. 51 , 377–408 (2022).35167762
12. Liang B. , Kiessling V. & Tamm L. K. Prefusion structure of syntaxin-1A suggests pathway for folding into neuronal trans-SNARE complex fusion intermediate. Proc. Natl. Acad. Sci. 110 , 19384–19389 (2013).24218570
13. White K. I. , Zhao M. , Choi U. B. , Pfuetzner R. A. & Brunger A. T. Structural principles of SNARE complex recognition by the AAA+ protein NSF. eLife 7 , e38888 (2018).30198481
14. Choi U. B. NSF-mediated disassembly of on- and off-pathway SNARE complexes and inhibition by complexin. eLife 7 , e36497 (2018).29985126
15. Yoon T.-Y. & Munson M. SNARE complex assembly and disassembly. Curr. Biol. 28 , R397–R401 (2018).29689222
16. Khan Y. A. , White K. I. & Brunger A. T. The AAA+ superfamily: a review of the structural and mechanistic principles of these molecular machines. Crit. Rev. Biochem. Mol. Biol. 57 , 156–187 (2022).34632886
17. Hanson P. I. & Whiteheart S. W. AAA+ proteins: have engine, will work. Nat. Rev. Mol. Cell Biol. 6 , 519–529 (2005).16072036
18. Puchades C. , Sandate C. R. & Lander G. C. The molecular principles governing the activity and functional diversity of AAA+ proteins. Nat. Rev. Mol. Cell Biol. 21 , 43–58 (2020).31754261
19. Nagiec E. E. , Bernstein A. & Whiteheart S. W. Each Domain of the N-Ethylmaleimide-sensitive Fusion Protein Contributes to Its Transport Activity (*). J. Biol. Chem. 270 , 29182–29188 (1995).7493945
20. Novick P. , Field C. & Schekman R. Identification of 23 complementation groups required for post-translational events in the yeast secretory pathway. Cell 21 , 205–215 (1980).6996832
21. Ma C. , Su L. , Seven A. B. , Xu Y. & Rizo J. Reconstitution of the vital functions of Munc18 and Munc13 in neurotransmitter release. Science 339 , 421–425 (2013).23258414
22. Lai Y. Molecular Mechanisms of Synaptic Vesicle Priming by Munc13 and Munc18. Neuron 95 , 591–607.e10 (2017).28772123
23. Gao Y. Sensory deficit screen identifies nsf mutation that differentially affects SNARE recycling and quality control. Cell Rep. 42 , 112345 (2023).37027300
24. Leitz J. Beyond the MUN domain, Munc13 controls priming and depriming of synaptic vesicles. Cell Rep. 43 , 114026 (2024).38809756
25. Ryu J.-K. Spring-loaded unraveling of a single SNARE complex by NSF in one round of ATP turnover. Science 347 , 1485–1489 (2015).25814585
26. Ryu J.-K. , Jahn R. & Yoon T.-Y. Review: Progresses in understanding N-ethylmaleimide sensitive factor (NSF) mediated disassembly of SNARE complexes. Biopolymers 105 , 518–531 (2016).27062050
27. White K. I. NSF converts syntaxin clusters to a priming-ready state. Unpublished.
28. Whiteheart S. W. N-ethylmaleimide-sensitive fusion protein: a trimeric ATPase whose hydrolysis of ATP is required for membrane fusion. J. Cell Biol. 126 , 945–954 (1994).8051214
29. de la Peña A. H. , Goodall E. A. , Gates S. N. , Lander G. C. & Martin A. Substrate-engaged 26S proteasome structures reveal mechanisms for ATP-hydrolysis–driven translocation. Science 362 , eaav0725 (2018).30309908
30. Rizo A. N. Structural basis for substrate gripping and translocation by the ClpB AAA+ disaggregase. Nat. Commun. 10 , 2393 (2019).31160557
31. Fernandez I. Three-dimensional structure of an evolutionarily conserved N-terminal domain of syntaxin 1A. Cell 94 , 841–849 (1998).9753330
32. DAmico K. A. Structure of a membrane tethering complex incorporating multiple SNAREs. Nat. Struct. Mol. Biol. 1–9 (2024) doi:10.1038/s41594-023-01164-8.38253665
33. Wang T. , Li L. & Hong W. SNARE proteins in membrane trafficking. Traffic 18 , 767–775 (2017).28857378
34. Baryshnikova A. Systematic Functional Annotation and Visualization of Biological Networks. Cell Syst. 2 , 412–421 (2016).27237738
35. Thomas P. D. PANTHER: Making genome-scale phylogenetics accessible to all. Protein Sci. 31 , 8–22 (2022).34717010
36. Kato M. & Wickner W. Ergosterol is required for the Sec18/ATP-dependent priming step of homotypic vacuole fusion. EMBO J. 20 , 4035–4040 (2001).11483507
37. Ungermann C. , Wickner W. & Xu Z. Vacuole acidification is required for trans-SNARE pairing, LMA1 release, and homotypic fusion. Proc. Natl. Acad. Sci. 96 , 11194–11199 (1999).10500153
38. Baker T. A. & Sauer R. T. ClpXP, an ATP-powered unfolding and protein-degradation machine. Biochim. Biophys. Acta BBA - Mol. Cell Res. 1823 , 15–28 (2012).
39. Merkley E. D. Distance restraints from crosslinking mass spectrometry: Mining a molecular dynamics simulation database to evaluate lysine–lysine distances. Protein Sci. Publ. Protein Soc. 23 , 747–759 (2014).
40. Chin J. W. Addition of p-Azido-l-phenylalanine to the Genetic Code of Escherichia coli. J. Am. Chem. Soc. 124 , 9026–9027 (2002).12148987
41. Punjani A. & Fleet D. J. 3D variability analysis: Resolving continuous flexibility and discrete heterogeneity from single particle cryo-EM. J. Struct. Biol. 213 , 107702 (2021).33582281
42. Zucchi P. C. & Zick M. Membrane fusion catalyzed by a Rab, SNAREs, and SNARE chaperones is accompanied by enhanced permeability to small molecules and by lysis. Mol. Biol. Cell 22 , 4635–4646 (2011).21976702
43. Zick M. , Stroupe C. , Orr A. , Douville D. & Wickner W. T. Membranes linked by trans-SNARE complexes require lipids prone to non-bilayer structure for progression to fusion. eLife 3 , e01879.
44. Shein M. Characterizing ATP processing by the AAA+ protein p97 at the atomic level. Nat. Chem. 16 , 363–372 (2024).38326645
45. Shvarev D. Structure of the HOPS tethering complex, a lysosomal membrane fusion machinery. eLife 11 , e80901 (2022).36098503
46. Song H. , Orr A. S. , Lee M. , Harner M. E. & Wickner W. T. HOPS recognizes each SNARE, assembling ternary trans-complexes for rapid fusion upon engagement with the 4th SNARE. eLife 9 , e53559 (2020).31961324
47. Song H. , Torng T. L. , Orr A. S. , Brunger A. T. & Wickner W. T. Sec17/Sec18 can support membrane fusion without help from completion of SNARE zippering. eLife https://elifesciences.org/articles/67578 (2021) doi:10.7554/eLife.67578.
48. Orr A. & Wickner W. Sec18 supports membrane fusion by promoting Sec17 membrane association. Mol. Biol. Cell 33 , ar127 (2022).36103252
49. Kelch B. A. , Makino D. L. , O’Donnell M. & Kuriyan J. Clamp loader ATPases and the evolution of DNA replication machinery. BMC Biol. 10 , 34 (2012).22520345
50. Verma A. R. , Ray K. K. , Bodick M. , Kinz-Thompson C. D. & Gonzalez R. L. Increasing the accuracy of single-molecule data analysis using tMAVEN. Biophys. J. (2024) doi:10.1016/j.bpj.2024.01.022.
51. Scheres S. H. W. A Bayesian view on cryo-EM structure determination. J. Mol. Biol. 415 , 406–418 (2012).22100448
52. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination - PubMed. https://pubmed.ncbi.nlm.nih.gov/28165473/.
53. Jumper J. Highly accurate protein structure prediction with AlphaFold. Nature 596 , 583–589 (2021).34265844
54. Emsley P. & Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60 , 2126–2132 (2004).15572765
55. Adams P. D. The Phenix Software for Automated Determination of Macromolecular Structures. Methods San Diego Calif 55 , 94–106 (2011).
56. Croll T. I. ISOLDE: a physically realistic environment for model building into low-resolution electron-density maps. Acta Crystallogr. Sect. Struct. Biol. 74 , 519–530 (2018).
