CryoCLEM-targeted cryoET of mouse forebrain synapses

(A) Schematic of the ‘ultra-fresh’ preparation of synapses for cryoCLEM. Psd95GFP/GFPknockin mice were culled, forebrain was dissected, homogenized in ice cold artificial cerebrospinal fluid, and plunge-frozen on cryoEM grids. 2 minutes was the time taken to cull, dissect and cryopreserve samples on EM grids.

(B) CryoCLEM of cryoEM grid square containing an ultra-fresh synapse preparation. Top left, cryoFM image of a holey carbon grid square. Top right, cryoEM image of the same grid square shown in top left. Bottom left, merged image of cryoFM and cryoEM images indicating location of PSD95-GFP puncta. Scale bar, 5 µm. Black box indicates region enlarged in Bottom right, showing PSD95-GFP associated with PoSM. White rectangle, indicates region where images for the tomogram shown in C were acquired. Scale bar, 500 nm.

(C) Tomographic slice of PSD95-contaning glutamatergic synapse. The PoSM (green) was identified by PSD-GFP cryoCLEM and the PreSM (cyan) was identified by its tethering to the PoSM and the prevalence of synaptic vesicles. Salient organelle and macromolecular constituents are indicated: Purple arrow, mitochondrion. Yellow arrow, microtubule. Cyan arrow, synaptic vesicle and cyan arrowhead indicating intermediate of vesicle fission/fusion. Gold arrow, F-actin filament. Pink arrow, putative endosomal compartment. Orange arrowheads, transsynaptic macromolecular complex bridging PreSM and PoSM. Magenta arrowheads, postsynaptic membrane proteins with extracellular domains extending 13-14 nm into the synaptic cleft. Blue arrow, lateral matrix of macromolecules in synaptic cleft. Scale bar, 20 nm. The tomographic map shown here is representative of 93 tomograms obtained across 5 mouse forebrain preparations.

(D) 3D segmentation of membranes and macromolecules in a representative tomographic volume of a PSD95-containing glutamatergic synapse. Coloured as in C.

(E) Prevalence of branched filamentous actin networks in presynaptic membrane compartments (PreSM, cyan), PSD95-containing postsynaptic membrane compartments (PoSM, green), and neighbouring non-synaptic membrane compartments (Vicinal, grey).

Data points are per mouse, for 5 adult Psd95GFP/GFP mouse forebrain samples. Error bars, SEM.

(F) Same as E but for microtubules.

(G) Molecular crowding of the PreSM and PoSM cytoplasm in ultra-fresh synapse preparation. Top, schematic showing the measurement of molecular density estimated with multiple line profiles plots of voxel intensity within each synapse tomogram. Bottom, voxel intensity (a.u., arbitrary units) profile plots of presynaptic (cyan) and postsynaptic (green) cytoplasm. Profiles were aligned to the lipid membrane peak of the PreSM and PoSM (grey arrows). The average intensity profile of 27 synapses from 3 mice is shown in black with 1 standard deviation in grey. Red and purple bar, proximal and distal regions of PoSM cytoplasm, respectively, as analysed in H.

(H) Comparison of average molecular density profile in regions 5-30 nm (proximal) and 50-200 nm (distal) from the PoSM of each synapse. Regions proximal to the PoSM correspond to locations in which a PSD is a conserved feature in conventional EM. Blue, red and green datapoints, synapses with cytoplasm proximal the PoSM that contained significantly higher, lower, and not significantly different density from distal regions, respectively (two-tailed t test with Bonferroni correction, P< 0.05, n=682 and 5673 voxels).

The in-tissue molecular architecture of glutamatergic synapses in the adult mammalian brain

(A) Schematic showing cryoCLEM and cryoET workflow using thin vitreous sections from forebrains of adult Psd95GFP/GFP knockin mice to determine the in-tissue architecture of glutamatergic synapses. Mice were culled and dissected. 100 μm acute slices were collected, from which 2 mm diameter biopsies of cortex were high-pressure frozen. 70-150 nm thin cryo-sections were cut from vitrified tissue and attached to cryoEM grid for cryoCLEM and cryoET.

(B) CryoCLEM of cryoEM grid square containing 150 nm thin cryo-section. Top left, cryoFM image of a holey carbon grid square. Top right, cryoEM image of the same grid square shown in top left. Bottom left, merged image of cryoFM and cryoEM images indicating location of PSD95-GFP puncta. Scale bar, 5 µm. Black box indicates region enlarged in Bottom right, showing PSD95-GFP associated with PoSM. White box indicates region where images for the tomogram shown in ‘c’ were acquired. Scale bar, 500 nm.

(C) Tomographic slice of PSD95-GFP containing synapse within thin vitreous cryo-section of adult mouse cortex. Cyan, PreSM. Green, PoSM. Scale bar, 50 nm.

(D) 3D segmentation of membranes and macromolecules in a representative tomographic volume of a PSD95-containing glutamatergic synapse within thin vitreous cryo-section of adult mouse cortex. The PoSM (green) was identified by PSD-GFP cryoCLEM and the PreSM (cyan) was identified by its tethering to the PoSM and the prevalence of synaptic vesicles. Salient organelle and macromolecular constituents are indicated: Magenta, membrane proteins within synaptic cleft. Purple, mitochondrion. Yellow, microtubule. Pink, putative endosomal compartment. Brown, myelin. Grey, vicinal membrane-bound subcellular compartments.

(E) Molecular profiles same as Figure 1G but for 21 in-tissue synapses of vitreous cryo-sections from acute brain slices of seven Psd95GFP/GFP mice.

(F) Comparison of average molecular density profile in regions 5-30 nm (proximal) and 50-200 nm (distal) from the PoSM of each synapse. Blue, red and green datapoints, synapses with cytoplasm proximal the PoSM that contained significantly higher, lower, and not significantly different density from distal regions, respectively (two-tailed t test with Bonferroni correction, P< 0.05, n=682 and 5673 voxels).

(G) Conventional EM of synapses in chemically fixed, resin-embedded, heavy metal-stained acute brain slice biopsies from Psd95GFP/GFP knockin (left) and wildtype (right), respectively. Red arrowhead, postsynaptic density. Scale bar, 50 nm.

(H and I) Same as e and f but for acute brain slice biopsies imaged by conventional EM (chemically fixed, resin imbedded, and heavy metal-stained) from 3 Psd95GFP/GFP mice. PSD, postsynaptic density evident in non-native, conventional EM samples.

Architecture of lipid membrane bilayers in glutamatergic synapses

(A-C) Organelles in PSD95-containing synapse shown top, as a tomographic slice and bottom, 3D segmentation. Top inset, PSD95-GFP cryoCLEM image of synapse. Scale bar, 500 nm. Cyan, PreSM. Green, PoSM. Red, green and blue arrows indicate x-, y-, and z-axis of tomogram, respectively. Scale bar, 100 nm.

(A) Flat/tubular membrane compartment and large spheroidal membrane compartment pseudo-coloured orange and red, respectively.

(B) Multivesicular bodies pseudo-coloured orange.

(C) Polyhedral membrane vesicle pseudo-coloured orange

(D) 3D segmentation (left) and schematic (right) of PSD95-GFP-containing synapses showing various topologies of connectivity. Top, unimodal single input and single output. Middle, bifurcated synapse with single input and output. Bottom, divergent synapse with single input and two outputs. Cyan, PreSM input. Green, PoSM output.

(E) Plot of fluorescence intensity (from cryoFM of PSD95-GFP) versus PoSM volume suggesting that there is no correlation between apparent amount of PSD95-GFP and PoSM volume. Linear regression (green line) with Pearson’s r=0.14, P=0.41.

(F) Plot showing the copy number of F-actin cytoskeletal elements in PoSM compartment versus PoSM volume of each synapse. The copy number of actin filaments was estimated from 3D segmented models of 7 nm filaments within the tomographic map. F-actin branching from another filament were counted as a separate filament. Linear regression (green line) with Pearson’s r=0.83, P<0.0001.

(G-K) Snapshots of membrane remodeling within glutamatergic synapses.

(G) Prevalence of membrane fission/fusion intermediates within PreSM, PoSM, and non-synaptic membranes (vicinal) from tomograms of 5 adult Psd95GFP/GFP mice. Error bars, SEM.

(H) Same as E but for clathrin-coated membrane.

(I) Top, tomographic slice of a clathrin-coated (red arrowhead) synaptic vesicle and bottom, a 3D tomographic density map for the region shown on top. Clathrin cage and membrane is shown with red and yellow voxels respectively. Scale bar, 20 nm.

(J) Top, tomographic slice of a clathrin-coat (red arrowhead) encapsulating part of an internal membrane within the PreSM compartment of a PSD95-GFP containing synapse and shown bottom, a masked 3D tomographic map. Clathrin cage and membrane is shown with red and yellow voxels, respectively. Scale bar, 20 nm.

(K) Top, tomographic slice, and bottom, masked 3D tomographic map showing clathrin-coated endocytic pit (red arrowhead) within the cleft of a PSD95-GFP containing PoSM. Clathrin cage and membrane is shown with red and green voxels, respectively. Scale bar, 20 nm.

Structural variables of synaptic strength

(A) Distribution of cleft height distances of all synapses is shown as a kernel density estimation (KDE). Inset, schematic depicting nearest neighbour method for defining the synaptic cleft between PreSM (cyan) and PoSM (green) compartments.

(B) Top left, ‘side’ view of atomic structure of AMPA subtype of ionotropic glutamate receptor (PDB ID: 3kg2). Dashed line, boundary of transmembrane domain. Top right, ‘top’ views of distal amino-terminal domain layer, proximal ligand-binding domain layer, and transmembrane domain layer. Bottom left, tomographic slice of PSD95-GFP containing membrane oriented approximately parallel to the electron beam. Putative ionotropic glutamate receptor pseudo-coloured in magenta. Bottom right, tomographic of PSD95-GFP containing membrane oriented approximately orthogonal to the electron beam. Multiple putative ionotropic glutamate receptor pseudo-coloured in magenta. Scale bar, 20 nm.

(C) In situ structure of ionotropic glutamate receptors within PSD95-GFP containing PoSM determined by subtomogram averaging of 2,368 sub-volumes. Atomic model of AMPA receptor (GluA2, PDB: 3kg2) docked in the determined map. Subunits of the model are coloured in red, yellow, cyan and purple.

(D) Cluster analysis of identified ionotropic glutamate receptors depicted as a 3D model of the postsynaptic compartment with ionotropic glutamate receptors in the orientation and position determined by sub-tomogram averaging. Each cluster is coloured differently: magenta, orange, and yellow. Presynaptic and postsynaptic cleft membrane are shown in cyan and green, respectively. Postsynaptic membrane outside of the cleft is shown in grey.

(E) Receptor number per cluster for three different cluster locations: in the cleft (magenta), partly inside and outside the cleft (boundary, yellow), and completely outside the cleft (perisynaptic, grey) from tomograms of 3 adult Psd95GFP/GFPmice.

(F) Prevalence of receptor clusters for three different cluster locations: in the cleft (magenta), partly inside and outside the cleft (boundary, yellow), and completely outside the cleft (perisynaptic, grey) from tomograms of 3 adult Psd95GFP/GFP mice.

CryoFM, docked vesicles and actin in PSD95-GFP-containing synapses

(A) Cryogenic fluorescence microscopy of forebrain (cortex and hippocampus) ultra-fresh synapse preparation from Psd95GFP/GFP (top panels) and WT (bottom panels). Left, detection of GFP with excitation and emission filters of 480 nm and 527 nm, respectively. Middle, detection of red fluorescence with excitation and emission filters of 538 nm and >590 nm, respectively. Right, bright field image. Submicron-sized puncta in the GFP channel only detected exclusively in Psd95GFP/GFPsamples.

(B) Tomographic slice showing representative docked synaptic vesicles. Docked synaptic vesicles were identified by the macromolecular complexes (open cyan arrow heads) that tethered synaptic vesicles 2-8 nm from the presynaptic membrane. 96% PreSM compartments contained docked synaptic vesicles (see Figure 1–table 1). Scale bars, 20 nm.

(C) Masked tomographic map of a branched filamentous actin (F-actin) network in PSD95-containing PoSM compartment. Red spherical markers, spaced 7.1 nm apart on the outside edges of a filament.

(D) Molecular architecture of F-actin in the presynaptic compartment.

Left, tomographic slice through tomogram of synapse located by PSD95-GFP cryoCLEM. F-actin parallel bundle indicated by yellow arrowhead within PreSM (cyan) compartment that was adhered to PoSM (green). See also Figure 1–video 2. Top inset, PSD95-GFP cryoCLEM image of synapse. White box, indicates White box, indicates region where images for the tomogram were collected. Scale bar, 500 nm.

Right, 3D segmented model of F-actin (yellow tubes) in PreSM (cyan) opposing PoSM (green). Synaptic vesicles (orange). Short F-actin filaments connect a fraction of synaptic vesicles to the plasma membrane. Scale bar, 20 nm.

Molecular density analysis within synapse tomographic volumes

(A) Schematic showing an error that arises when projecting a tomographic volume of synapse with a pre-(blue) and post-synaptic (green) compartments that are not perfectly aligned with the z-axis of the tomogram. Molecular crowding has been previously analysed using a projection (hatched red rectangle) of multiple tomographic slices (red lines). Projection of such an unaligned volume generates an erroneous molecular density profile in which the plasma membrane and extracellular proteins contribute to measurements of the postsynaptic cytoplasm. Black arrow, direction of the missing wedge.

(B) Example of synapse tomographic volume with postsynaptic membrane not perfectly oriented with z-axis. Left and right, tomographic slices sub-volume (red rectangle) of top, middle and bottom of tomographic sub-volume, respectively. Blue and green arrowheads, position of PoSM and PreSM in each tomographic slice of the sub-volume, respectively. Note, the position of the PoSM changes in different slices of the sub-volume.

(C) Projection of unaligned tomographic slices with black bracket showing error caused by regions of molecular density from the membrane and synaptic cleft smeared into the postsynaptic cytoplasm

(D) Schematic showing how accurate measurements of molecular density of the postsynaptic cytoplasm within a tomographic sub-volume were obtained (Figure 1G-H and 2E-F). To measure molecular density profiles of each tomographic slice were aligned to the PoSM before averaging. Tomographic slices in which the PoSM was tilted in the x-y plane (see top tomographic slice in panel B) were also excluded from the profile.

Tomographic slices and molecular density profiles of 27 ultra-fresh PSD95-GFP synapses. All tomograms within the ultra-fresh synapse dataset are shown in which the synaptic cleft is approximately parallel to the beam (PreSM and PoSM perpendicular to the x-y plane to enable accurate profile measurements of molecular density), and PoSM and PreSM compartments were closed. Left, tomographic slice at central regions of synaptic cleft. Red rectangle, region from which profile measuring voxel density were plotted. Scale bar, 20 nm. Right, average molecular density profile (36 nm wide and 450 nm long) was measured at central region of the cleft, close to docked synaptic vesicles and clusters of putative ionotropic glutamate receptors, because this is where one would expect to identify a postsynaptic density. Profiles were aligned to the lipid membrane peak of the PreSM and PoSM (grey arrows). Voxel intensity (a.u., arbitrary units) profile plot of presynaptic (cyan) and postsynaptic (green) cytoplasm. Profiles show that the appearance of high molecular density corresponding to a PSD is not conserved in these samples compared to synapse imaged by conventional EM (see Figure 2G-H and Figure 2–figure supplement 2).

Molecular density profiling and tomographic slices of 21 in-tissue cryosection tomograms containing PSD95-GFP synapses

(a) CryoCLEM and cryoET of two synapses in vitreous cryo-sections collected from adult Psd95GFP/GFP knockin mouse cortex. Left, PSD95-GFP cryoCLEM image of in-tissue synapse. Scale bar, 500 nm. White box, field of view in right, tomographic slice of PSD95-GFP containing synapse within thin vitreous cryo-section of adult mouse cortex. Cyan, PreSM. Green, PoSM. Scale bar, 50 nm. See also Figure 2–video 2.

(b) Tomographic slices and molecular density profiles of 21 in-tissue PSD95-GFP synapses. All tomograms within the in-tissue synapse dataset are shown in which the synaptic cleft is approximately parallel to the beam (PreSM and PoSM perpendicular to the x-y plane to enable accurate profile measurements of molecular). Left, tomographic slice at central regions of synaptic cleft. Red rectangle, region from which profile measuring voxel density were plotted. Scale bar, 20 nm. Right, Average molecular density profile (36 nm wide and 450 nm long) was measured at central region of the cleft, close to docked synaptic vesicles and clusters of putative ionotropic glutamate receptors, because this is where one would expect to identify a postsynaptic density. Profiles corresponding to a 37 nm sub-volume were aligned to the lipid membrane peak of the PreSM and PoSM (grey arrows). Voxel intensity (a.u., arbitrary units) profile plot of presynaptic (cyan) and postsynaptic (green) cytoplasm. Profiles show that the appearance of high molecular density corresponding to a PSD is absent in these samples compared to synapse imaged by conventional EM (see Figure 2G-H and Figure 2–figure supplement 2).

Conventional EM of adult brain glutamatergic synapses

(A) Conventional EM of synapses in, resin-embedded, heavy metal-stained Psd95GFP/GFP knockin (left) and wildtype (right) mouse. Red arrowhead, postsynaptic density. Scale bar, 50 nm. Top, Electron micrograph showing multiple post-synaptic densities (PSDs, red arrowheads) from perfusion fixed cortex. Scale bar, 500 nm. Black box, PSD shown in middle close up. Bottom, Conventional EM of ultra-fresh forebrain synapses chemically fixed for conventional EM. Scale bar, 50 nm.

(B) Molecular crowding of the PreSM and PoSM cytoplasm of perfusion fixed Psd95GFP/GFPknockin mouse cortex. Pixel intensity (a.u., arbitrary units) profile plots of presynaptic (cyan) and postsynaptic (green) cytoplasm. Profiles were aligned to the lipid membrane peak of the PreSM and PoSM (grey arrows). The average intensity profile of 9 synapse tomograms is shown in black. PSD, postsynaptic density.

Structural categories of membrane-bound organelles in PSD95-containing synapses in the adult brain

(A) Prevalence of membrane bound organelles within the PreSM (cyan) and PoSM (green) compartments identified in tomograms of forebrain PSD95-GFP containing synapses from 5 adult Psd95GFP/GFP mice. Error bars, SEM.

(B) Prevalence of structural classes of membrane-bound organelles, including polyhedral vesicles, flat/tubular membrane compartments, large spheroidal membrane compartments (>60 nm diameter), multivesicular bodies, and polyhedral vesicles within the PreSM (cyan) and PoSM (green) compartments of PSD95-containing forebrain synapses from 5 adult Psd95GFP/GFP mice. Error bars, SEM.

(C-F) Classification of membrane-bound organelles in PSD95-GFP-containing synapses. Gallery of tomographic slices showing intracellular organelle subtype (pseudo-coloured magenta) of four different PSD95-GFP containing synapses of (C) Flat tubular twisted membranes.

(D) Clustered membrane proteins in vesicles.

(E) Multivesicular bodies.

(F) Polyhedral membranes.

Scale bar, 20 nm

(G-I) Synaptic mitochondria.

(G) Bar chart showing the prevalence of mitochondria in the pre-(PreSM) and post-synaptic membrane (PoSM) compartments of glutamatergic synapses.

(H) Tomographic slice of synapse located by PSD95 cryoCLEM showing PreSM containing mitochondria identifiable by surrounding outer membrane (solid white arrowhead) and inner cristae membrane (open white arrowhead). Electron-dense deposits in the mitochondrial matrix are indicated with open orange arrowheads. Scale bar, 50 nm. Top inset, PSD95-GFP cryoCLEM image of synapse. Scale bar, 500 nm.

(I) Same as h, except ultra-fresh synapses (see Methods) were prepared with ACSF omitting calcium chloride. Electron dense deposits within mitochondrial matrix were detected in all (3 out of 3) PreSM compartments. Scale bar, 50 nm.

(J) Tomographic slices of synapses with intermediates of vesicle fusion/fission on the presynaptic side of the cleft of a PSD95-GFP containing synapse. Left, small intermediate. Right, Large intermediate. Scale bar, 20 nm.

Quantification and variability of cleft height in glutamatergic synapses

(A) Tomographic slice of PSD95-GFP containing synapse. 3D model of computationally determined coordinates of synaptic cleft are shown with presynaptic and postsynaptic cleft membrane coordinates depicted in cyan and green, respectively. Scale bar, 100 nm.

(B) Measuring the nearest neighbour distance between presynaptic and postsynaptic coordinates of the cleft were used to quantify the cleft heights for each synapse. The distribution of cleft height distances is shown using kernel density estimation (KDE) plotted with a different colour for each synapse.

(C and D) Synapses with a bimodal distribution of cleft hights. Left, tomographic slice of PSD95-GFP containing synapses showing subsynaptic regions with varying cleft height between PreSM (cyan) and PoSM (green). Red arrowheads, 5-15 nm transsynaptic adhesion complexes. Orange arrowheads, 22-28 nm transsynaptic adhesion complexes. Yellow arrowheads, 32-36 nm transsynaptic adhesion complexes. Scale bar, 20 nm. Top inset, PSD95-GFP cryoCLEM image of synapse. Scale bar, 500 nm. Right, the corresponding distribution of cleft height distance for each synapse is plotted using kernel density estimation (KDE).

Subtomogram averaging and anchoring of ionotropic glutamate receptors

(A) Estimation of the resolution of subtomogram averaged ionotropic glutamate receptors shown by plotting with a linear scale the resolution (x-axis) versus global Fourier shell correlation (y-axis).

(B) Cytoplasmic anchoring of perisynaptic ionotropic glutamate receptor clusters (pseudo-coloured in magenta). Cluster was defined using DBSCAN. Tomographic slices through tomogram of PSD95-GFP containing postsynaptic compartment. Slices separated by 16, 10, 10 and 10 nm, showing top-down views of i) extracellular membrane protein domains, ii) proximal and iii) distal cytoplasmic macromolecular complexes, iv) associated branched f-actin network (orange open arrowhead). Scale bar, 20 nm.