BGPT: Paper Review: The molecular infrastructure of glutamatergic synapses in the mammalian forebrain.

Explore by Goal

Quick Explanation Copied

Rapid scientific take

Peukes et al. use an ultra-fresh cryoCLEM→cryoET pipeline with Psd95 (Dlg4)-GFP labeling in adult mouse forebrain to test a long-standing idea: that glutamatergic synapses have a conserved, PSD-like high-density postsynaptic layer near the membrane.

Their core finding is that a conventional-EM-style “PSD density layer” is not conserved under near-physiological cryoET conditions; instead, molecular crowding near the postsynaptic membrane is variable, with widespread presence of cytoskeletal networks (notably F-actin) and organelles, and with ionotropic receptor nanodomains/clusters in cleft and perisynaptic space.

Long Explanation

Paper Review (visual-first): Peukes et al., eLife 2025 — “The molecular infrastructure of glutamatergic synapses in the mammalian forebrain”

Link: eLife article • DOI: 10.7554/eLife.100335

What they set out to test: whether a conserved, conventional-EM-style postsynaptic density (PSD) density layer is a true native feature of adult mammalian glutamatergic synapses, or instead an artifact of chemical fixation/heavy-metal staining/resin embedding.

Figure 1 — PSD-like “proximal density increase” is heterogeneous

The paper quantifies molecular crowding profiles along the postsynaptic compartment. They report that only about half the synapses show a significant relative increase in proximal (5–30 nm) vs distal (50–200 nm) cytoplasmic density near the PoSM (with the remainder showing no increase or decreased relative density).

Evidence basis (from paper text): the authors state that “half the population of synapses” shows significant relative increase in density proximal to the PoSM (5–30 nm vs 50–200 nm), while other subpopulations do not show higher density or show lower relative density.

Figure 2 — Cytoskeletal and organelle compartment prevalence differs by synaptic subcompartment

Reported fractions of compartments containing F-actin and microtubules in presynaptic vs postsynaptic membrane-bound compartments (PreSM vs PoSM), plus mitochondria prevalence (noting mitochondria mostly in PreSM).

Evidence basis (from paper text):

F-actin identifiable in all PoSM compartments and in 49 ± 7% of PreSM compartments (they note actin could be monomeric/short undetectable in tomograms where not seen).
Microtubules identified in 15 ± 2% of PreSM compartments and absent from PoSM.
Mitochondria present in PreSM more abundantly (44 ± 5% of PreSM), with near-rare PoSM proximity.

Figure 3 — Ionotropic glutamate receptor clusters (AMPA-like) are common but frequently not confined to cleft

They identify putative ionotropic receptors in most PoSM compartments, quantify clusters using subtomogram positions, and report that many clusters are entirely or partly outside the cleft.

Evidence basis (from paper text):

Putative ionotropic glutamate receptors identified in 95 ± 2% of PoSM compartments.
Clusters identified in 78% of synapses; average 2 clusters per synapse.
Clusters can contain up to ~60 receptors; average about 10 receptors per cluster.
60% of synapses maintain multiple large clusters partially or completely outside the cleft.

Figure 4 — Synaptic cleft height is ~33 nm on average but varies locally

They computationally define cleft between nearest PreSM and PoSM surfaces and report mean cleft height of ~33 nm in ultra-fresh data, with a synapse-to-synapse range reported as 27–37 nm.

Evidence basis (from paper text): mean ~33 nm, range 27–37 nm for ultra-fresh synapses; they also report broader distributions in some in-tissue and ultra-fresh cases.

Mechanistic interpretation (what is known vs inferred vs uncertain)

1) Known from the data (within the paper’s imaging/analysis)

No conserved, PSD-defining dense protein layer in near-physiological cryoET: molecular crowding proximal to PoSM is not consistently higher and is heterogeneous.
Cellular architecture is preserved variably (F-actin networks widespread; microtubules in PreSM; organelle diversity including membrane-bound compartments, mitochondria; they provide mitochondrial aggregate characterization and argue against excitation during preparation via experiments without exogenous calcium).
Receptors cluster: ionotropic receptor subtomogram averaging supports AMPA receptor-like density and they quantify nanodomains/clusters that frequently extend outside the cleft.
Preparation dependence is directly assessed by conventional EM of comparable samples: they reproduce a PSD-like dense region in conventional EM, and argue it arises during sample preparation.

2) Inferred (plausible, but not uniquely determined)

PSD-like density in conventional EM likely reflects fixation/staining/resin-embedding-induced redistribution or extraction rather than a native conserved layer—supported by their paired workflow comparison, but still dependent on what conventional EM does to specific molecular species and on how ‘density’ maps across modalities.
Synaptoplasm organization shifts the emphasis from a single membrane-adjacent scaffold layer to a distributed 3D organization of cytoskeleton, receptor clusters, and transsynaptic assemblies.

3) Uncertain / key unknowns

Resolution-limited specificity: subtomogram averaging yields ~25 Å resolution for receptor densities, which supports identification at a coarse structural level but cannot assign isoforms/complex compositions at fine detail (e.g., exact receptor subtype mix, auxiliary subunits) from density alone.
Functional state: cryoET captures snapshots; without direct time-resolved or activity-state labeling, ‘synaptic strength’ implications remain partly inferential.
Generality: experiments focus on adult mouse forebrain and rely on Psd95-GFP labeling; while they test preparation dependence, transfer of conclusions to other brain regions, developmental ages, or synapse types still requires further datasets.

Critical appraisal (skeptical, evidence-weighted)

Strengths

Methodological rigor around ‘PSD artifact’: the paper directly addresses the classic fixation/staining concern by comparing cryoET molecular density profiles to conventional EM prepared from the same labeling context, and reports reproduction of PSD-like density in conventional EM.
Correlative targeting: cryoFM-guided cryoCLEM maps Psd95-GFP puncta to cryoEM coordinates, enabling in situ tomogram selection and reducing ambiguity about glutamatergic maturity.
Quantification discipline: they explicitly avoid projection artifacts in molecular density profiling by aligning individual slices to membrane peaks before averaging.
Structural “infrastructure” cataloging: organelle and cytoskeletal features are annotated, including cytoskeletal prevalence and evidence against sample-prep induction for mitochondrial dense aggregates.

Potential blind spots / limitations

Single-scaffold labeling bias: Psd95-GFP guides synapse selection and PoSM definition; while this is a rational marker for mature glutamatergic synapses, it may bias sampling toward Psd95-rich subtypes or states.
Missing wedge / tomographic geometry: cryoET missing wedge can affect segmentation and density estimates; while they exclude tomograms when PoSM cytoplasm is obscured or oriented on the missing wedge, quantification can still be sensitive to reconstruction choices.
Density is a proxy: voxel intensity → “molecular crowding” assumes comparable contrast and interpretability across compartments and preparation states; even if relative density changes are robust, the mapping to specific protein complexes is not direct.
Cluster identification thresholds: receptor clustering uses computational proximity/meshing choices; while described, cluster membership can change with algorithmic parameters.
Sample size and generalization: adult male mice from cortex/hippocampus are studied; extrapolation to all mammalian forebrain circuits and across species requires more cryoET panels.

Broader context: why this matters

The classic PSD is often defined from conventional EM as a dense postsynaptic protein layer opposite the active zone. The paper’s cryoET result—absence of a conserved PSD-like density layer—directly challenges the universality of that textbook structural rule and reframes the synapse as a 3D, heterogeneous “synaptoplasm” whose measurable density gradients depend on near-physiological preservation.

Targeted queries you may want next (BGPT)

Author review links

Buttons open BGPT bespoke author-review pages for each named author.

Feedback:

Updated: April 28, 2026