Synaptic vesicles (SVs) are storage organelles of neurotransmitters located in the presynaptic bouton1,2. When SVs fuse with the presynaptic membrane, neurotransmitters are released to propagate chemical signaling between neurons1,2,3. To sustain rounds of synaptic transmission, SVs have to be recycled, requiring tight control of their biogenesis4,5. A characteristic property of SVs is their small uniform size, which is important for brain function, because it enables the storage of well-defined amounts of neurotransmitters and thus reliable synaptic transmission. The membrane of SVs is composed of a small number of lipids and organelle-specific, abundant transmembrane (TM) proteins6. Cell biological and biochemical studies have shown that both the SV TM proteins and lipids influence the morphology of SVs1,7,8,9,10,11. However, little is known how SV membrane proteins and lipids cooperate to sculpt SVs.

A major family of SV membrane proteins comprises the tetraspan vesicle membrane proteins synaptophysin and synaptogyrin6,11. Tetraspan vesicle membrane proteins are integral membrane proteins composed of four TM helices11. Multiple studies have linked synaptogyrin and synaptophysin to high membrane curvature and the formation of SVs11,12,13,14,15,16,17. Ectopic expression of synaptophysin in non-neuronal cells leads to the formation of small cytoplasmic vesicles12, while removal of synaptogyrin from SVs increases SV size13. Supporting this connection, the name synaptogyrin (from Greek γυρος, meaning circle) references the uniform spherical shape of SVs14,15,16. The synaptogyrin family consists of three isoforms, with synaptogyrin 1 and synaptogyrin 3 being neuronal6,11. Single and double knockout mice lacking synaptogyrin 1 and/or synaptophysin 1 did not have detectable morphological or biochemical changes in their SVs, but had reduced short- and long-term synaptic plasticity16. Mice in which synaptogyrin 1 and 3, as well as synaptophysin 1 and 2, were knocked out had elevated SV release probability18.

Synaptogyrins have been linked to several human diseases. Synaptogyrin 1 levels are decreased in the outer layer of the dentate gyrus in people with Alzheimer’s disease19. Synaptogyrin 3 plays an important role in presynaptic dysfunction induced by the protein Tau20, the protein that aggregates into the neurofibrillary tangles inside neurons, a pathological hallmark in people with Alzheimer’s disease21. Lowering synaptogyrin 3 expression rescued Tau-induced memory deficits in mice22. Mutations in synaptogyrin 1 have been identified in people with schizophrenia23,24. In addition, synaptogyrins have been associated with stroke and viral diseases25,26. Despite the important role of synaptogyrins in SV function and human disease, the structure of synaptogyrins, or any other tetraspan vesicle membrane protein, has remained enigmatic.

To gain insights into the molecular mechanisms that determine how SVs achieve their uniform morphology, we determined the high-resolution structure of synaptogyrin 1, studied its interaction with the SV-abundant lipid phosphatidylserine (PS), and investigated the ability of synaptogyrin to remodel the SV membrane. Our results establish a molecular mechanism for the sculpting of the SV membrane by synaptogyrin.


Human synaptogyrin is a stable monomer

To gain insight into the molecular basis of SV sculpting, we combined high-resolution structural analysis of synaptogyrin with membrane remodeling assays, nuclear magnetic resonance (NMR)-based lipid-binding experiments, and site-directed mutagenesis. We successfully overexpressed isoform b of human synaptogyrin 1 (hereafter named synaptogyrin; Fig. 1a) in Escherichia coli and directly extracted it from the membrane without refolding27. To enable high-resolution structural analysis by NMR spectroscopy, synaptogyrin was labeled with 15N and 13C and perdeuterated, and reconstituted into different detergents, as well as into isotropic bicelles formed by 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 1,2-diheptanoyl-sn-glycero-3-phospocholine (DHPC)28. Bicelles are an excellent medium for studying membrane proteins29,30,31.

Fig. 1: Structure of human synaptogyrin.
figure 1

a, Schematic representation of a SV (diameter: ~30–50 nm) with major TM proteins (NT transporter represents neurotransmitter transporters)6. Synaptogyrin is highlighted; other tetraspan vesicle membrane proteins are displayed in black. b, Superposition of 1H-15N TROSY-HSQC spectra of synaptogyrin in isotropic bicelles (red; DMPC/DHPC, q = 0.3) and n-undecyl β-maltoside (UDM) micelles (black). c, Residue-specific secondary structure score and random coil index (RCI S2) derived from the NMR chemical shifts of synaptogyrin. Four TM helices are identified: TM1 (P17–E46), TM2 (N66–F92), TM3 (K99–Q131) and TM4 (T144–S176). d, Ten lowest-energy structures of synaptogyrin in cartoon representation. The N- and C-terminal tails are located in the cytoplasm. e, Paramagnetic relaxation enhancement by 16-DSA and gadodiamide on synaptogyrin structure in UDM micelles. Residues with 16-DSA relaxation enhancement rates of >77 mM−1 s−1 (blue) are located inside the hydrophobic environment. Loop and tail residues have gadodiamide relaxation enhancement rates of >6.5 mM−1 s−1 (red). f, Interactions between conserved tryptophan residues inside the TM structure of synaptogyrin.

Source data

We observed high-resolution liquid-state NMR spectra with well-defined and dispersed synaptogyrin cross-peaks (Fig. 1b and Extended Data Fig. 1). Circular dichroism revealed the presence of α-helical structure (Extended Data Fig. 2), in agreement with four predicted TM helices11. We then estimated the average transverse NMR relaxation rates of the amide resonances of synaptogyrin, which are in agreement with a synaptogyrin monomer (Extended Data Fig. 2c). Synaptogyrin directly extracted from SVs by the detergent 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS) is also monomeric14. The data demonstrate that synaptogyrin is a stable monomer in the investigated environments.

High-resolution structure of synaptogyrin

To resolve the three-dimensional (3D) structure of synaptogyrin to high resolution, we determined the protein’s backbone resonance assignment. NMR assignment of multispan α-helical TM proteins is complicated by strong signal overlap, which arises from the large line widths and limited chemical shift dispersion in α-helical TM regions32. We overcame this challenge by combining multi-dimensional NMR experiments optimized for high-molecular-weight membrane proteins with amino acid type-specific labeling (Supplementary Table 1 and Extended Data Figs. 3 and 4a)32,33. Through this combined approach, we achieved backbone assignment for 98% of the residues. On the basis of the assigned NMR chemical shifts, we identified four TM helices formed by residues P17–E46 (TM1), N66–F92 (TM2), K99–Q131 (TM3) and T144–S176 (TM4) (Fig. 1c). The three loops between the TM helices, as well as the amino- and carboxy-terminal tails, are disordered (Fig. 1c). Comparison of the 1H and 15N chemical shifts of synaptogyrin in isotropic bicelles and detergent micelles showed that chemical shift changes occur predominantly in the disordered regions (Extended Data Fig. 2a).

We then recorded side chain-specific NMR experiments and assigned 83% of the side chain protons. Distance restraints were subsequently derived from 3D NOESY spectra (Supplementary Table 2 and Extended Data Fig. 4b,c) and paramagnetic relaxation enhancement data of a single-cysteine mutant protein tagged with the nitroxide spin label MTSL (S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)methyl methanethiosulfonate). In addition, we measured 1H-15N residual dipolar couplings (RDCs) of synaptogyrin weakly aligned in a polyacrylamide gel. The combined NMR data were subjected to structure calculation in Rosetta34. The ten resulting lowest-energy structures have a r.m.s. deviation (r.m.s.d.) of 0.5 Å (Fig. 1d, Table 1 and Extended Data Fig. 5). Paramagnetic relaxation confirmed the hydrophobic burial of the TM region and the solvent accessibility of the loops and tails (Fig. 1e and Extended Data Fig. 6).

Table 1 NMR constraints and structural statistics for the ensemble of ten best energy structures of synaptogyrin

The three-dimensional structure reveals a tight bundle of four α-helices (Fig. 1d). Previous studies have shown that the N- and C-terminal tails are located on the cytoplasmic side, and the TM1–TM2 and TM3–TM4 loops point to the interior of SVs11. The four TM helices are curved and twist around a central axis (Fig. 1d and Extended Data Fig. 7). Helix TM1 kinks at proline P24. The short loop between TM2 and TM3 is well defined and is the least flexible loop, according to the chemical-shift-derived flexibility parameter (Fig. 1c).

The sequence compositions of the four TM helices of different synaptogyrin isoforms, as well as of the tetraspan vesicle membrane protein synaptophysin, are well conserved (Fig. 2a,b). Synaptogyrin 1a–c and synaptogyrin 3 contain four cysteine residues, with C59 being located in the TM1–TM2 loop, C68 and C82 in TM2, and C124 in TM3. In synaptogyrin 2, C82 in TM2 is replaced by serine (Fig. 2a). The 3D structure of synaptogyrin 1b shows that the cysteines are spatially separated and do not form disulfide bonds. The 3D structure of synaptogyrin 1b further reveals contacts between the aromatic rings of W115 and W119 from TM3 together with W161 from TM4 (Fig. 1f). These three tryptophan residues are fully conserved in different synaptogyrin isoforms (Fig. 2a).

Fig. 2: Sequence conservation of synaptogyrins.
figure 2

a, Sequence alignment of the different isoforms of synaptogyrin together with the amino acid sequence of synaptophysin (SYP 1). The lysine-arginine patches on the cytoplasmic (cyt) and the intravesicular (iv) membrane side are highlighted by blue boxes. b, Mapping of conserved residues (red) onto the 3D structure of synaptogyrin.

Synaptogyrin sculpts small vesicles

Next, we prepared liposomes that mimic the lipid composition of SVs (SVs have a diameter of ~30–50 nm)6,35,36. Liposomes of different sizes were visible in electron micrographs (Fig. 3a and Extended Data Figs. 8 and 9). Using dynamic light scattering (DLS), we estimated that the average diameter of the liposomes was ~200 nm (Fig. 3b and Extended Data Figs. 8 and 9). We then reconstituted synaptogyrin at increasing concentrations into the liposomes. For protein:lipid molar ratios of 1:2,000, 1:1,000 and 1:500, the average liposome diameter only slightly decreased (Fig. 3b). At the protein:lipid ratio of 1:500, however, the surface of the liposomes appeared more rugged when compared with that at lower synaptogyrin concentrations (Fig. 3a, middle row). When we further increased the concentration of synaptogyrin to protein:lipid molar ratios of 1:50 and 1:20, we observed only small vesicles, with average DLS-derived diameters of ~54 nm and ~42 nm, respectively (Fig. 3b, top row, and Extended Data Fig. 8). Synaptogyrin thus dramatically remodels SV-like membranes, generating small vesicles with a diameter comparable to SVs.

Fig. 3: Synaptogyrin sculpts SV-sized vesicles.
figure 3

a, Negative-stain electron micrographs (scale bar, 200 nm) of liposomes and vesicles comprising phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylserine (PS), phosphatidylinositol (PI) and cholesterol (PC:PE:PS:PI:cholesterol, 40:32:12:5:10)35 with increasing concentrations of synaptogyrin. Insets display individual liposomes or vesicles. b, Average DLS-derived diameter of liposomes and vesicles without or with increasing concentrations of synaptogyrin. Lipid compositions (wt/wt): top panel, PC:PE:PS:PI:cholesterol, 40:32:12:5:10; second panel, PC:PE:PS:PI, 43:34:15:7; third panel, PC:PE:PI:cholesterol, 47:36:6:10; bottom panel, PC:PE:PI:cholesterol, 40:32:17:10. Error bars represent s.d., based on three repeated measurements.

Source data

The SV membrane is rich in cholesterol6,36, and cholesterol binds to the tetraspan SV protein synaptophysin8. To determine whether cholesterol is required for synaptogyrin’s membrane-sculpting activity, we repeated the experiments using liposomes without cholesterol. Without synaptogyrin and cholesterol, the DLS-derived liposome diameter was ~130–140 nm (Fig. 3b), smaller than in presence of cholesterol, which is known to increase membrane stiffness and thus makes membranes harder to bend37. When we reconstituted synaptogyrin at increasing concentrations, the DLS-derived diameter decreased, and ~40- to 50-nm-large vesicles formed (Fig. 3b). Cholesterol is thus not required for synaptogyrin-induced sculpting of the SV-like membrane.

The most abundant negatively charged lipid in the SV membrane is PS6,35, a phospholipid with unique physical and biochemical properties38. To investigate the role of PS, we replaced it with the anionic lipid phosphatidylinositol (PI) and tested two different PI concentrations. Electron microscopy and DLS showed that, even with 17% PI and at the highest concentration of synaptogyrin, no membrane remodeling occurred (Fig. 3b, lower panel). Remodeling of the SV-like membrane is thus specific for both synaptogyrin and PS.

Membrane remodeling depends on synaptogyrin-phosphatidylserine interaction

To gain molecular insights into synaptogyrin-based SV membrane sculpting, we reconstituted 15N-2H-labeled synaptogyrin into bicelles formed by DMPC and DHPC, as well as into DMPC/DMPS/DHPC-bicelles with 1,2-dimyristoyl-sn-glycero-3-phospho-L-serine (DMPS) in 40-fold molar excess over synaptogyrin. The 1H-15N correlation spectra of synaptogyrin in the two bicelle environments closely superimposed with selected cross-peaks displaying DMPS-induced changes (Extended Data Fig. 10 and Figs. 4a and 5a, top panel). DMPS changes both the chemical shifts and the intensities of several residues (Fig. 4a,c and Extended Data Fig. 10). In addition, the cross-peaks of some residues, such as G41, G71, G75, A78, L80 and K136, were split into 2–3 new signals in the presence of DMPS (Fig. 4a,c). K136 together with K134 forms a positively charged patch on the intravesicular side (Fig. 5b). K102 and R170, which are located on the cytoplasmic side of synaptogyrin (Fig. 5b), were also perturbed in the presence of DMPS (Fig. 4a).

Fig. 4: Membrane remodeling depends on synaptogyrin-phosphatidylserine interaction.
figure 4

a, Selected lysine and arginine cross-peaks in 1H-15N TROSY-HSQC spectra of synaptogyrin in isotropic bicelles (DMPC/DHPC, q = 0.3) without (red) or with DMPS (20%; blue). b, Change in average DLS-derived diameters (Φ) of liposomes and vesicles in the presence of increasing concentrations of wild-type synaptogyrin or different synaptogyrin mutants (protein:lipid molar ratios of 1:50 (dark gray), 1:100 (gray) and 1:500 (light gray)) when compared with liposomes without protein (ΦRef.). Lipid composition (wt/wt) when not noted otherwise: PC:PE:PS:PI:cholesterol, 40:32:12:5:10. Synaptogyrin displays a lysine-arginine patch at both the intravesicular (iv) and the cytoplasmic (cyt) membrane side (displayed as blue boxes) containing K102 and K103, and K134 and K136, respectively. The data for wild-type synaptogyrin in liposomes lacking PS are shown for comparison. The protein containing the substitutions C68A, C82S and C124A serves as control. Error bars represent s.d., based on three repeated measurements. c, DMPS binding induces different TM structural states in synaptogyrin, as evidenced by multiple cross-peaks for individual residues. Cross-peaks observed in isotropic bicelles without and with 20% DMPS are displayed in red and blue, respectively. Tentative cross-peak assignments in the presence of DMPS are indicated by dashed arrows. Residues that display multiple cross-peaks in the presence of DMPS cluster together in the synaptogyrin structure (marked in red).

Source data

Fig. 5: Phosphatidylserine shapes synaptogyrin TM structure.
figure 5

a, Residue-specific changes in the 1H-15N chemical shifts (CSP, chemical shift perturbation) of wild-type (top) and mutant synaptogyrin (K102Q, middle; K134Q K136Q, bottom) in isotropic bicelles (DMPC/DHPC, q = 0.3) without and with 20% DMPS. The dashed red line marks the average of DMPS-induced CSPs observed in wild-type synaptogyrin. Substitution sites are marked by green arrows. The location of the cytoplasmic (cyt) and the intravesicular (iv) lysine-arginine patch is marked in orange. Red boxes mark the location of residues that display both chemical shift changes and peak splitting. L86 and A87 in TM2 (pink box) also display DMPS-induced CSPs. The location of the four TM helices is shown above. b, CSPs induced by DMPS in wild-type synaptogyrin (a, top panel) mapped onto the 3D structure. Larger CSPs are represented by increased tube thickness and a brighter red color. Lysine and arginine residues located on the cytoplasmic (cyt) and the intravesicular (iv) membrane side are displayed with side chains (blue). c, Superpositions of cross-peaks from the 1H-15N TROSY-HSQC spectra of wild-type synaptogyrin without (red) or with DMPS (blue), as well as the two synaptogyrin mutants in the presence of DMPS (K102Q synaptogyrin, orange; K134Q K136Q synaptogyrin, cyan). Tentative cross-peak assignments in the presence of DMPS are indicated by the dashed arrows. d, Scheme for the remodeling of the SV membrane by synaptogyrin. Simultaneous binding of phosphatidylserine (PS; red) to the lysine-arginine cluster (blue) on the cytoplasmic and the intravesicular side of synaptogyrin (left) induces conformational changes in the TM structure of synaptogyrin (gray and orange, middle) that trigger the bending of the SV membrane (right).

Source data

Residue-specific analysis of the chemical shift changes confirmed that both lysine- and arginine-rich regions were perturbed in the presence of DMPS (Fig. 5a, top panel). The DMPS-induced chemical shift changes of residues 138–140 might be caused by binding of DMPS to the nearby K134 or K136 (Fig. 5b). The NMR signal perturbations support binding of DMPS, likely through its negatively charged head group, to the two positively charged lysine-arginine clusters located on the cytoplasmic and the intravesicular side of synaptogyrin.

We next probed the importance of PS binding to synaptogyrin by combining site-specific protein mutation with membrane-remodeling assays (Fig. 4b). Either one or two positively charged residues in the two lysine-arginine clusters were mutated to glutamine. None of the charge-attenuated synaptogyrin mutants was able to remodel liposomes into SV-sized vesicles (Fig. 4b, Extended Data Fig. 8 and Supplementary Fig. 1). By contrast, a control mutant protein, in which the three cysteines, C68, C82 and C124, were mutated, retained its membrane-sculpting activity (Fig. 4b). The experiments demonstrate that the two lysine-arginine clusters of synaptogyrin are required for the protein’s ability to remodel the SV-like membrane.

Besides the PS-associated changes of the NMR signals in the two lysine-arginine clusters, we found pronounced signal perturbations for residues 68–80 in TM2, as well as for residues 14–23 and 35–48 in TM1 (Fig. 5a, top panel, b). In addition, we observed two to four cross-peaks for individual residues in the regions 35–41 and 71–80 in the presence of PS (Fig. 4c). For example, a single cross-peak is present for L80 without DMPS that splits into four cross-peaks with PS. None of the 35–48 or 68–80 residues is positively charged, making a direct interaction of these residues with PS unlikely. Instead, the detection of multiple NMR signals for individual residues provides evidence for the formation of distinct TM structural states, which are in slow exchange on the NMR time scale. In the high-resolution structure of synaptogyrin, residues 35–48 and 68–80 cluster together at the cross-over of TM1 and TM2 (Figs. 4c and 5b). Binding of PS to the two lysine-arginine patches thus induces a conformational rearrangement of the TM1–TM2 interface of synaptogyrin.

Phosphatidylserine binding reshapes synaptogyrin structure

To gain further insight into the connection between PS binding and conformational changes in synaptogyrin, we prepared the 15N-2H-labeled mutant proteins synaptogyrin-K102Q and synaptogyrin-K134Q K136Q. We reconstituted the mutant proteins into DMPC/DHPC-bicelles without or with DMPS and recorded two-dimensional 1H-15N correlation spectra. For both mutant proteins, a 40-fold excess of DMPS did not induce chemical shift changes in the respective mutated lysine-arginine cluster (Fig. 5a, middle and lower panels). In addition, residues 14–16 at the N terminus of TM1, which are close to the cytoplasmic K102, K103 and R170 patch, were not perturbed by DMPS in the synaptogyrin-K102Q (Fig. 5a, middle panel, b). For all three proteins, DMPS induced changes for residues 20–23, suggesting a direct interaction of DMPS with R22 (Fig. 5a and Extended Data Fig. 10b). Notably, a substitution in either the cytoplasmic or the intravesicular lysine-arginine cluster did not abolish the DMPS-induced perturbations in the cluster on the other side of the membrane (Fig. 5a). For example, DMPS caused chemical shift changes for residues 101–103 in synaptogyrin-K134Q K136Q (Fig. 5a, lower panel). DMPS thus binds independently to the two lysine-arginine clusters located on the cytoplasmic and the intravesicular side of synaptogyrin.

Alterations in the cytoplasmic or the intravesicular lysine-arginine cluster abolished the ability of synaptogyrin to remodel liposomes into small vesicles (Fig. 4b). Therefore, DMPS-induced changes in the NMR spectra that are inhibited by both mutations are particularly interesting. This is the case for the DMPS-induced chemical shift perturbation and peak splitting of residues 68–80 in TM2 and 35–48 in TM1. For both synaptogyrin-K102Q and synaptogyrin-K134Q K136Q, the presence of DMPS did not cause perturbations of residues 68–80, and smaller changes for residues 35–48 (Fig. 5a). DMPS binding to the non-mutated, solvent-accessible lysine-arginine cluster thus did not generate multiple structural states for the residues in the TM regions. Instead, the peak positions of these residues in the mutant proteins in the presence of DMPS were similar to those of the wild-type protein without DMPS (Fig. 5c and Supplementary Fig. 2). In addition, p.K102Q as well as p.K134Q and p.K136Q abolished or attenuated the DMPS-associated changes of residues 85YLA87 in TM2 and V30 in TM1 (Fig. 5a,b). When we inhibited binding of PS to either the cytoplasmic or the intravesicular lysine-arginine cluster, we blocked the phospholipid-induced changes at the TM2–TM1 interface (Fig. 5b), and the ability of synaptogyrin to remodel membranes into small vesicles (Fig. 4b).


Protein-lipid interactions govern the functional properties of cellular membranes and have been suggested to shape SVs1,7,8,9,10,11,39,40. An important morphological feature of SVs is their uniform size, which ensures that well-defined amounts of neurotransmitters are stored in each SV1. Although several complementary mechanisms may act in concert to enable a uniform morphology of SVs, our study establishes a connection between the 3D structure of the tetraspan vesicle membrane protein synaptogyrin, its interaction with the abundant SV-lipid PS and the formation of small vesicles. Binding of PS to the two lysine-arginine clusters located at the cytoplasmic and intravesicular side of synaptogyrin induces a rearrangement of the TM2–TM1 interface (Fig. 4c). Binding of PS to either the cytoplasmic or the intravesicular side is not sufficient to trigger the structural changes in synaptogyrin (Fig. 5a). Only when PS binds to both clusters, the TM structure of synaptogyrin is changed, and small vesicles are formed (Figs. 3 and 4b,c). This suggests that the binding processes of PS to the cytoplasmic and the intravesicular basic patch work together. The additive/cooperative action might be enabled through the direct interaction of the cytoplasmic lysine-arginine cluster with the N-terminal end of TM1 (Fig. 5b, cyt view), and of the intravesicular basic patch with the N-terminal end of TM2 (Fig. 5b, iv view).

While the PS-associated generation of multiple cross-peaks for residues at the TM1–TM2 interface indicates that PS binding changes the TM structure and likely results in two to three different synaptogyrin TM1–TM2 conformations, which are in slow exchange on the NMR chemical shift time scale (Fig. 4c), the high-resolution structures of these synaptogyrin states are currently unknown. Determination of these structures is complicated not only by the underlying dynamics, but also by the high molecular weight of the synaptogyrin-bicelle assembly that is a challenge for NMR studies. In addition, the strong aliphatic resonances of PS in NOE spectra complicates the extraction of NOE contacts for synaptogyrin in complex with PS. In the absence of a high-resolution structure of synaptogyrin with PS molecules simultaneously bound to both the cytoplasmic and the intravesicular basic patch, we can only hypothesize that PS binding changes the relative orientation of the TM1 and TM2 helices and thus induces a more curved synaptogyrin structure, which promotes the bending of the membrane (Fig. 5d).

Synaptogyrin localizes predominantly to SVs at nerve terminals, as well as to microvesicles in endocrine cell types outside the nervous system17. Cell studies have further shown that synaptogyrin regulates Ca2+-dependent exocytosis41. In mice with a synaptogyrin-synaptophysin double knockout, short-term and long-term synaptic plasticity were reduced16. In vivo evidence for a direct role of synaptogyrin in the formation of small vesicles was obtained when a synaptogyrin null mutant was studied in Drosophila13. The Drosophila genome encodes a single synaptogyrin isoform and lacks a synaptophysin homolog, which minimizes compensatory effects associated with synaptogyrin knockouts that might occur in other models systems. The flies that lacked synaptogyrin had SVs with increased diameters13. In addition, the flies lacking synaptogyrin showed changes in transmission, that is the study connected synaptogyrin with the diameter of SVs and neurotransmitter-associated function13. Further support for the importance of synaptogyrin and synaptophysin for sculpting of small vesicles was provided by ectopic expression of synaptophysin in non-neuronal cells, which induced the formation of small cytoplasmic vesicles12. The synaptogyrin-phosphatidylserine-dependent remodeling of liposomes into small vesicles reported in the current work (Fig. 3) demonstrates, through a bottom-up reconstitution approach, that synaptogyrin alone, that is in the absence of other SV proteins, is able to remodel SV-like membranes and induce the formation of SVs. Changes in membrane morphology already occur at synaptogyrin:lipid ratios of 1:500 (Fig. 3a). In addition, when the membrane contains only PC, PS and cholesterol, the vesicle diameter is decreased from 195 to 124 nm at the synaptogyrin:lipid ratio of 1:500 (Extended Data Fig. 9a). Formation of SV-sized vesicles with a diameter of 40–50 nm required synaptogyrin:lipid ratios of 1:50 (Fig. 3a). The data suggest that the high density of proteins in SVs6, for example through steric exclusion and the formation of synaptogyrin clusters, might contribute to the formation of 40- to 50-nm-sized SVs.

The high sequence conservation between different synaptogyrins, including the two lysine-arginine clusters (Fig. 6), suggests that other synaptogyrins might also be able to remodel the SV membrane in a PS-dependent manner. The related tetraspan vesicle membrane protein synaptophysin, however, lacks some of the lysine and arginine residues (Fig. 6). Future studies will thus have to show whether synaptophysin is able to remodel the membrane of SVs and elucidate the influence of different SV lipids, including cholesterol8, that is whether the membrane remodeling activity of synaptogyrins, synaptoporins and synaptophysin might be interchangeable. Such an interchangeability between different tetraspan vesicle membrane proteins in shaping the SV membrane could be responsible for the mild functional deficits observed in synaptogyrin-synaptophysin knockout studies16,18,41, and could at the same time ensure that the formation of uniformly sized SVs in the mammalian brain is highly robust. Because the SV membrane contains a high concentration of TM proteins6, the membrane-remodeling activity of synaptogyrin, and potentially other tetraspan vesicle membrane proteins, might be amplified by homo- or heterophilic protein interactions or through indirect protein interactions via the bilayer6,42.

Fig. 6: The TM structure of tetraspan vesicle membrane proteins is conserved.
figure 6

Comparison of the structures of different isoforms of synaptogyrin, as well as the structure of synaptophysin (SYP), as predicted by AlphaFold2. When available, the AlphaFold2-predicted structures were downloaded from the AlphaFold Protein Structure Database (,46; otherwise, they were predicted using the AlphaFold2 notebook ( Views from the cytoplasmic (cyt) and the intravesicular (iv) membrane side are displayed to the right. Side chains are shown and labeled for selected lysine and arginine residues of synaptogyrin 1b. N- and C-terminal tails were removed for better visualization.

We determined the high-resolution structure of isoform b of human synaptogyrin 1, a protein from the family of tetraspan vesicle membrane proteins. We further show that binding of PS changes the TM structure of synaptogyrin and is critical for the ability of synaptogyrin to remodel liposomes into small vesicles. Because synaptogyrin has been linked to several diseases of the brain19,20,21,22,23,24,43,44, the findings from our study are likely relevant for both normal and perturbed brain activity.


Molecular cloning, protein expression, and purification

The construct of synaptogyrin 1b (Uniprot: O43759-2) was cloned into the pET16b vector with an N-terminal histidine tag. The synaptogyrin mutants were engineered using site-directed mutagenesis (Thermo Fisher Scientific) and confirmed by DNA sequencing (Supplementary Table 3). Expression was performed in the Escherichia coli strain Lemo21 (DE3) with 250 µM l-rhamnose, following previously established procedures27. Cells were grown in Luria Bertini or minimal medium containing 15NH4Cl and/or [13C6]glucose as the sole nitrogen or carbon source in H2O or D2O, respectively. Single or double amino acid selective labeling was obtained using [15N]Ala, [15N]Leu, [15N]Cys, [15N]Lys, [15N]Phe and [15N]Tyr as precursors, and the other amino acids were unlabeled in the minimal medium to suppress scrambling. Protein methyl-labeled on Ileδ1, LeuproS and ValproS was expressed using 13CH3-methyl specifically labeled precursors (NMR-Bio) in fully deuterated medium. The precursor of 2-(D3) methyl, 2,4-(13C2) acetolactate was added 1 h prior to the deuterated M9 medium. The 2-ketobutyric acid 4-(13C), 3,3 (D2) sodium salt was added 20 min before induction by addition of IPTG.

For synaptogyrin production, the temperature was reduced to 20 °C at an optical density at 600 nm (OD600) of 0.8–1.0 after seeding the preculture. Cells were induced with 0.2 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) after 1 h and were further incubated at 20 °C for 16 h before collection. The cell pellet was resuspended in 50 ml lysis buffer (20 mM NaPi, pH 7.5, 300 mM NaCl, 1 mM DTT, 1 mg ml–1 lysozyme, 0.01 mg ml–1 DNaseI, 1 mM phenylmethylsulfonyl fluoride, and one Complete Protease Inhibitor Cocktail-EDTA free (Roche Applied Science)) and homogenized by stirring for 50 min at 4 °C. The suspension was lysed with an Emulsiflex (Avastin) and centrifuged (10,000g, 10 min, 4 °C) for removal of cell debris. The membrane fraction was collected from the supernatant by ultracentrifugation (Beckman Ti45 rotor) at 100,000g at 4 °C for 50 min. The membranes were resuspended and solubilized by solubilization buffer (20 mM NaPi, pH 7.5, 300 mM NaCl, 1 mM DTT, and one Complete Protease Inhibitor Cocktail-EDTA free, 1 % dodecyl-ß-d-maltoside (DDM)) at 4 °C for 1 h. The solubilized membrane fraction was loaded onto a histidine-trap open column, and the imidazole concentration was increased stepwise to 400 mM. When required, the DDM detergent was exchanged for different detergents, such as 0.005% 2,2-didecylpropane-1,3-bis-β-d-maltopyranoside (NG310) or 0.1% n-undecyl β-maltoside (UDM) in this step. The protein was eluted with 400 mM imidazole, followed by removal of the histidine-tag with TEV protease during dialysis (20 mM sodium phosphate, pH 7.5, 300 mM NaCl, 1 mM DTT, 5% glycerol, 0.03% DDM (0.1% UDM or 0.005% NG310)). After cleavage, the TEV and the histidine tag were removed by loading onto a second histidine-trap column. The pure protein was further purified by gel filtration (HiLoad Superdex 200 column) in 20 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM DTT, 5% glycerol, 0.03% DDM (or 0.1% UDM or 0.005% NG310). The purified protein was verified by SDS polyacrylamide gel electrophoresis. The protein sample buffer was dialyzed against NMR buffer (20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.1 % UDM) for 18 h at 4 °C.

Circular dichroism

Far-ultraviolet CD measurements were performed at 40 °C using 10 μM synaptogyrin in 20 mM sodium phosphate, pH 7.5, 1 mM DTT, 0.1% UDM on a Chirascan spectrometer (Applied Photophysics). The path length of the cuvette was 0.2 mm. CD spectra were recorded from 190 to 280 nm with an integration time of 0.5 s. The measurement was repeated three times. The final spectrum was obtained by baseline subtraction using the buffer-only measurement.

Liposome morphology

DLS measurements were performed at 25 °C using a DynaPro NanoStar instrument with a detector positioned at a 90° angle concerning the incident light. The monochromatic laser with a wavelength of 662 nm (auto-attenuation turned on) irradiated the liposome sample in disposable COC cuvettes. Liposome samples (0.4 mM lipid) containing increasing synaptogyrin concentrations were incubated for 18 h at 25 °C in 20 mM HEPES, pH 7.5, and 1 mM TCEP prior to the measurement. DLS data were acquired with 20 cycles and 5 s of acquisition time for three repeated measurements and were analyzed with DYNAMICS v7.10.0.23. Error bars represent the s.d. of three measurements. Microsoft Excel v16.43 was used to fit data, as well as for statistical analysis.

In parallel, samples from the same preparation were adsorbed onto 400-mesh carbon-coated copper grids, and the buffer was removed using filter paper. Samples were then stained by adding 1% uranyl acetate solution, which was dried with a filter paper. The grids were imaged using a FEI Tecnai Spirit electron microscope with a TVIPS F416 4K camera.

NMR spectroscopy

NMR spectra were recorded at 40 °C on Bruker 600, 700, 800, 900, and 950 MHz Bruker NMR spectrometers using cryogenic probes. 1H spin-spin echo experiments with echo times from 0.1 to 8.1 ms were recorded on 0.1–0.3 mM synaptogyrin in 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.1% UDM (or 0.005% NG310 or 0.03% DDM) at 40 °C. Relaxation times (T2) were calculated using peak intensities of NMR signals between 9.5 ppm and 8.5 ppm.

The backbone resonances were assigned by TROSY-based 3D triple resonance through bond scalar correlation HNCO, HN(CA)CO, HNCA, HN(CO)CA, and HN(CA)CB experiments47,48,49 using 2H-13C-15N-labeled synaptogyrin (0.8 mM synaptogyrin in 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.1 % UDM). Backbone resonance assignment was further supported by a 3D 15N-edited NOESY-HSQC (NOE mixing time of 150 ms) experiment recorded on 0.8 mM 2H-15N-labeled synaptogyrin in 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.1% UDM, as well as 1H-15N TROSY-HSQC experiments recorded on [15N]Ala/[15N]Leu-, [15N]Leu/[15N]Tyr]-, [15N]Cys/[15N]Phe-, [15N]Thr-, and [15N]Lys-labeled synaptogyrin (0.8–1.0 mM). Side chain assignments were determined using a 3D (H)CCH-TOCSY (mixing time 10 ms), two 3D 13C-edited NOESY-HSQC (80 ms and 300 ms mixing time; 1.0 mM 13C-15N-labeled synaptogyrin in 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.03 % d25-DDM, 100% D2O), and two 3D 15N-edited NOESY-HSQC (80 ms and 300 ms mixing time; 0.9 mM 15N-labeled synaptogyrin in 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.03 % d25-DDM, 10 % D2O) experiments. For [U-2H-15N; Iδ1 /LVproS 13CH3]-labeled synaptogyrin, 2D 1H-13C HSQC and 13C-edited NOESY-HSQC (200 ms mixing time) experiments were recorded and used for the assignment of the methyl groups of the Ile, Leu and Val residues. Long-range NOE restraints between HN backbone resonances and side chain methyl protons were extracted from 3D 15N-resolved NOESY-HSQC experiments. recorded for 0.4 mM [U-2H-15N; Iδ1 /LVproS 13CH3]-labeled synaptogyrin in 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.03% d25-DDM, 10% D2O.

Residual 1DNH dipolar couplings were obtained by taking the difference in the J splitting values measured in oriented (7% acrylamide gel alignment medium) and isotropic sample conditions using interleaved 2D 1H-15N TROSY-HSQC/2D 1H-15N HSQC spectra recorded on 0.8 mM 2H-15N-labeled synaptogyrin in 20 mM sodium phosphate, pH 8.0, 300 mM NaCl, 1 mM TCEP, 0.1% UDM. NMR spectra were processed with Topspin 3.6.1 (Bruker) and analyzed using the software CcpNmr (Analysis 2.4.2)50. Residue-specific secondary structure scores and random coil indices (RCI S2) were determined from the assigned NMR chemical shifts of synaptogyrin using TALOS+ (ref. 51). The RDC values were analyzed using the PALES program52.

For the titration experiments of wild-type and mutant synaptogyrin with DMPS, 0.2 mM 2H-15N-labeled synaptogyrin was solubilized in isotropic bicelles without or with 20 % DMPS ((DMPC:DMPS = 8:2)/DHPC, q = 0.3) corresponding to a protein:DMPS ratio of 1:40. The buffer was 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP. 1H and 15N normalized weighted average chemical shift perturbations were calculated using ΔδNH = [((ΔH)2 + (ΔN)2 / 5) / 2]1/2, where \(\Delta H\) and \(\Delta N\) correspond to the 1H and 15N chemical shift differences between two states (for example, without and with DMPS).

To measure T2, we used a 90°–180° pulse sequence. Fitting the intensity decay curve to I0 (–t / T2) with I0 being the starting intensity gives the characteristic time constant T2. The correlation time τc was subsequently calculated according to:

$$\frac{1}{T2}=\frac{K}{2}\left[3{\tau }_{c}+\frac{5{\tau }_{c}}{1+\,{\omega }_{0}^{2}{\tau }_{c}}+\frac{{\tau }_{c}}{1+4{\omega }_{0}^{2}{\tau }_{c}}\right]$$
$$K=\frac{3{\mu }_{0}^{2}}{160{\pi }^{2}}\frac{{\hslash }^{2}{\gamma }^{4}}{{r}^{6}}.$$

The relaxation times are approximately associated with the correlation time by the equations according to Bloembergen–Purcell–Pound (BPP) theory, where ω0 is the rotational frequency of the signal, µ0 is the magnetic permeability of free space (4π × 10−7 H m−1), ħ is the reduced Planck constant 1.054571726 × 10−34 J s–1, γ is the gyromagnetic ratio of the nucleus (for 1H, it is 2.67513 × 108 rad s−1 T−1) and r is the distance between magnetically active spin-½ nuclei. The molecular weight of synaptogyrin in different micelle conditions was then calculated according to:

$${\mathrm{Molecular}}\; {\mathrm{weight}}=\frac{{\tau }_{c}{\mathrm{RT}}}{\eta (v+h)}$$

η is viscosity, v is the partial specific volume, h is the degree of hydration, R is the ideal gas constant and T is temperature.

Paramagnetic relaxation enhancement

For the spin labeling on Cys124 of synaptogyrin with MTSL (S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)methyl methanesulfonothioate), the three other native cysteines in synaptogyrin were mutated to either serine or alanine (C59A C68A C82S). To remove residual reducing agent, synaptogyrin was passed through a desalting column (HiPrep 26/10 Desalting; GE Healthcare). The protein was collected into a solution containing tenfold excess of MTSL and incubated at 4 °C for 18 h. Using the desalting column, we removed the excess MTSL, then exchanged the protein into NMR buffer containing 20 mM NaPi, pH 8.0, 150 mM NaCl, and 0.1% UDM. To determine the paramagnetic relaxation enhanced-broadening of the NMR signals, a 1H-15N TROSY-HSQC experiment was recorded for 1.0 mM 15N-labeled mutant synaptogyrin (C59A C68A C82S) in 20 mM sodium phosphate, pH 8.0, 150 mM NaCl, 1 mM TCEP, 0.1 % UDM.

To probe the hydrophobic burial and solvent exposure of synaptogyrin residues, we used the detergent-soluble 16-doxylstearic acid (16-DSA) and the water soluble [5,8-bis-(carboxymethyl)-11-[2-(methylamino)-2-oxoethyl]-3-oxo-2,5,8,11-tetraazatridecan-13-oato(3-)]-gadolinium (gadodiamide), respectively. The paramagnetic reagents were stepwise added to 0.8 mM U-[2H-15N]-synaptogyrin solubilized in UDM micelles. 1H-15N TROSY-HSQC spectra were then recorded at 40 °C using Bruker 900 and 800 MHz spectrometers.

Structure calculation

The recorded backbone chemical shifts (HN, N, C, Cα and Cβ) of residues 16–176 of synaptogyrin were used in RASREC CS-Rosetta to select 200 fragments of 3 and 9 residues in length. The N- and C-terminal tails of synaptogyrin were not included into the structure determination process, because they were predicted to be disordered by TALOS+ (ref. 51) on the basis of the experimental NMR chemical shifts. In addition, the loop regions formed by residues 47–65, 91–98 and 132–143, which had a TALOS+-predicted order parameter of less than 0.7, were removed from the Rosetta score computation, but were part of the structure calculation. NOE restraints from 3D 15N-edited NOESY-HSQC experiments were manually assigned. In the later stages of the structure calculation, NOE restraints determined automatically by CYANA were also included. The NOE-based distances (d0) were restrained with a ROSETTA flat-bottom potential with fixed lower bound (lb) = 1.5, upper bound (ub) = d0 + 0.15 Å, and inverse curvature (c) = 0.3 Å. In addition, 68 long-range PRE-based distances were restrained between the Cβ atom of the MTSL-labeled residue (Cys124) and the corresponding backbone amide protons with error bounds of ±4 Å in order to account for the size and flexibility of the Cys-MTSL53. NOE, PRE and RDC data maintained weights of 5 and 0.1 for scoring in the low-resolution sampling stage and the all-atom sampling stage, respectively. A weight equal to 5 was used for the NOE, PRE and RDC data for the centroid and all-atom structures to select a pool of iterated structures. The standard Rosetta all-atom energy function was used with the experimental weight set mpframework_smooth54. The energy function was adjusted for TM proteins to model the hydrophobic TM region55. TM regions were determined from the amino acid sequence using the OCTOPUS method56. Five thousand structures were calculated, and the ten structures with the lowest Rosetta all-atom energy score were selected for further analysis. Structure figures were prepared using the PyMOL Molecular Graphics System (Version

Liposome assays

Liposomes were formed from L-α-phosphatidylcholine (PC), L-α-phosphatidylethanolamine (PE), L-α-phosphatidylserine (PS), L-α-phosphatidylinositol (PI), and cholesterol (Avanti Polar Lipids) in selected wt/wt ratios. After addition of each phospholipid to the glass tube, the phospholipid mixture was evaporated off chloroform-methanol (1:1) under a low stream of nitrogen. The organic solvent was lyophilized to obtain a thoroughly dried lipid film. The dry lipid film was hydrated with 1 mL of liposome buffer (20 mM NaPi HEPES, pH 7.5, 1 mM TCEP). To produce unilamellar liposomes, the rehydrated lipid was sonicated three times for 2 min using an ultrasonic water bath sonicator. Next, the lipid film was extruded through a filter with a 200-nm pore size and thus liposomes of more uniform size were obtained. In order to reconstitute DDM-solubilized synaptogyrin into liposomes, DDM was then removed using Bio-Beads SM-2 Resin (Bio-rad). After DDM removal, the synaptogyrin sample was directly added to the liposomes. The sample was incubated for 18 h at 25 °C in 20 mM HEPES, pH 7.5, 1 mM TCEP. For NMR experiments, the sample was further concentrated using 30-kDa cut-off vivaspin (Satorious). NMR spectroscopy was used to confirm removal of DDM.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.