💡 Suggest / Edit
← Wiki Index / Genetics & Evolution / Postsynaptic Density
🧬 Genetics & Evolution

Postsynaptic Density

The postsynaptic density (PSD) is a ~1-gigadalton, 30-to-50-nm thick biomolecular condensate and layered protein scaffold (PSD-95/DLG4, GKAP, Shank, and Homer) beneath excitatory glutamatergic synapses that anchors NMDA/AMPA receptors and originated in nerveless basal metazoans.

12 min read • Updated November 2026 • Knowledge Wiki • AI-Assisted & Human-Supervised

The postsynaptic density (PSD) is a disk-shaped, electron-dense protein megacomplex positioned immediately beneath the postsynaptic plasma membrane of excitatory glutamatergic dendritic spines in the central nervous system, aligned directly opposite the presynaptic vesicle release active zone across the ~20-nm synaptic cleft. First identified in 1956 by George Palade and Sanford Palay as an osmiophilic thickening in transmission electron micrographs and biochemically isolated by Philip Siekevitz and Carl Cotman in the 1970s, a single mammalian PSD measures 200 to 500 nm in diameter, 30 to 50 nm in thickness, and has an ensemble molecular mass of ~1.0 to 1.2 gigadaltons (GDa)—packing roughly 1,000 to 2,000 individual polypeptide chains spanning ~400 distinct core proteins into a semi-crystalline, phase-separated organelle.

Far from a static structural plaque, the PSD is a dynamic, four-layered laminar scaffold built around four master multivalent hub families—PSD-95 (DLG4), GKAP (SAPAP1–4), Shank (SHANK1–3), and Homer (HOMER1–3)—that undergoes liquid-liquid phase separation (LLPS) and expands in volume within seconds of calcium influx through NMDA receptors and CaMKII activation during long-term potentiation (LTP), the cellular basis of learning and memory. Strikingly, comparative evolutionary genomics has revealed that the entire four-tier DLG–GKAP–Shank–Homer scaffold evolved before the origin of neurons and synapses, existing in full modular form in nerveless sponges (Porifera) and placozoans (Trichoplax adhaerens), while conspicuously lacking key components in the independently evolved nervous system of comb jellies (Ctenophora).

Four-Layer Laminar Architecture of the Excitatory Postsynaptic Density

Super-resolution STORM/PALM microscopy, immunogold electron tomography (Valtschanoff & Weinberg, 2001; Chen et al., 2008), and X-ray crystallography (PDB: 1BE9, Doyle et al., 1996) demonstrate that the PSD is stratified vertically from the postsynaptic lipid bilayer down into the dendritic spine cytoplasm across four distinct molecular layers:

LAMINAR STRATIFICATION OF THE EXCITATORY POSTSYNAPTIC DENSITY (0 to 50 nm Depth)

Synaptic Cleft (~20 nm; Trans-Synaptic Neuroligin-1 : Neurexin-1β + N-Cadherin Nanocolumns)

Postsynaptic Membrane: [NMDAR (GluN1/GluN2A/2B -ESDV-COOH)] [AMPAR (GluA1/2) + Stargazin (TARP γ2)]

Layer 1 (0–12 nm): [PSD-95 / DLG4] --> PDZ1-2 bind GluN2B & TARP -x-T/S-x-V-COOH (PDB: 1BE9 GLGF loop) --> PDZ3 binds Neuroligin-1; SH3-GK hook binds GKAP

Layer 2 (12–24 nm): [GKAP / SAPAP1-4] -> N-terminal 14-aa repeats bind PSD-95 GK; C-term -QTRL binds Shank PDZ

Layer 3 (24–40 nm): [Shank1-3 (PROSAP)] -> ANK + SH3 + PDZ + Proline-rich + C-terminal SAM helical polymer

Layer 4 (30–50+ nm): [Homer1b/c-3 Tetramers] (EVH1 binds Shank + mGluR1/5 + IP3R) + [Cortactin / F-Actin]

### Layer 1 (0–12 nm Below Membrane): Ionotropic Receptors and PSD-95 (DLG4) MAGUKs
The membrane-proximal zone is dominated by ~300 copies of PSD-95 (DLG4) (alongside its vertebrate 2R ohnologs SAP97/DLG1, PSD-93/DLG2, and SAP102/DLG3), which belong to the Membrane-Associated Guanylate Kinase (MAGUK) superfamily. Anchored to the inner leaflet of the postsynaptic membrane by dual N-terminal palmitoylation at Cys3 and Cys5, PSD-95 contains five folded modular domains arranged in a conserved PDZ1–PDZ2–PDZ3–SH3–GK supramodule:
- PDZ1, PDZ2, and PDZ3 Domains (PDB: 1BE9): Each ~90-residue PDZ domain (named for PSD-95, Drosophila Discs Large, and tight-junction Zonula Occludens-1) folds into a six-stranded β-barrel capped by two α-helices (αA and αB). In the landmark crystal structure of PSD-95 PDZ3 bound to its cognate C-terminal peptide (PDB: 1BE9, Doyle et al., 1996), the target peptide's extreme C-terminal consensus motif -X-Thr/Ser-X-Val-COOH (Type I PDZ ligand, e.g., -ESDV-COOH on NMDA receptor subunits GluN2A/GluN2B and -TTPV-COOH on the AMPA receptor auxiliary subunit Stargazin/TARP γ2) inserts as an antiparallel β-strand into the groove between strand βB and helix αB. The free C-terminal carboxylate (-COO−) is coordinated by a conserved four-residue GLGF carboxylate-binding loop (Gly322–Leu323–Gly324–Phe325 in PDZ3), while the side-chain hydroxyl of Thr/Ser at position P−2 forms a direct hydrogen bond with His372 on helix αB.
- The Intramolecular SH3–GK Supramodule: Rather than binding proline-rich peptides or catalyzing GMP phosphorylation, the C-terminal SH3 and enzymatically inactive Guanylate Kinase-like (GK) domains of PSD-95 fold together via an antiparallel β-strand hook into an integrated phosphopeptide-binding module that specifically captures phosphorylated GKAP.
- Trans-Synaptic Nanocolumns: PDZ3 of PSD-95 binds the C-terminus of Neuroligin-1, which reaches across the 20-nm synaptic cleft to bind presynaptic Neurexin-1β (alongside LRRTM2 and N-cadherin), aligning postsynaptic AMPA/NMDA receptor clusters within ~80-nm "trans-synaptic nanocolumns" directly beneath presynaptic active-zone RIM1/2 vesicle fusion hotspots.
### Layer 2 (12–24 nm): GKAP / SAPAP Vertical Linkers
Bridging the upper PSD-95 layer to the deeper Shank matrix are ~150 copies of GKAP (Guanylate Kinase-Associated Protein, encoded by DLGAP1–4 / SAPAP1–4). The N-terminal half of GKAP contains two to five conserved 14-amino-acid repeats that bind the GK domain of PSD-95 with nanomolar affinity once phosphorylated by CaMKII or PKA, while the extreme C-terminus of GKAP terminates in a canonical Type I PDZ-binding motif (-QTRL-COOH) that slots directly into the PDZ domain of Shank.
### Layer 3 (24–40 nm): The Shank1–3 (SHANK1–3) Master Polymer Lattice
Occupying the core structural belly of the PSD are ~150 to 200 copies of Shank1, Shank2, and Shank3 (also called ProSAP). Full-length Shank (~200 kDa) is a five-domain multivalent scaffold comprising N-terminal Ankyrin repeats (ANK), an SH3 domain, a central PDZ domain (which binds GKAP), a long proline-rich region (which binds Homer and cortactin), and a C-terminal Sterile Alpha Motif (SAM) domain. In the presence of micromolar Zn2+ ions released from glutamatergic synaptic vesicles and buffered inside the spine, Shank3 SAM domains polymerize head-to-tail into helical fibers that cross-link side-by-side into extensive 2D sheets, creating the rigid deep lattice of the PSD.
### Layer 4 (30–50+ nm): Homer Tetramers, mGluRs, and the Spine F-Actin Cytoskeleton
At the cytoplasmic boundary of the PSD, long-form Homer proteins (HOMER1b/c, HOMER2, HOMER3) assemble into dumbbell-shaped, antiparallel coiled-coil tetramers presenting two N-terminal EVH1 (Enabled/VASP Homology 1) domains at each end. By simultaneously binding -PPxxF- proline motifs on Shank, Group I metabotropic glutamate receptors (mGluR1/mGluR5 at the perisynaptic rim), IP3 receptors on the spine apparatus endoplasmic reticulum, and Dynamin-3, Homer tetramers weld the Shank sheet into a 3D meshwork coupled via cortactin, drebrin, and α-actinin to the branched F-actin cytoskeleton of the dendritic spine head.
| PSD Scaffold Layer & Protein | Human Genes & Copy Number per PSD | Key Structural Domains & PDB Reference | Primary Binding Partners | Neurodevelopmental / Psychiatric Disease Associations |
| :--- | :--- | :--- | :--- | :--- |
| Layer 1: PSD-95 family (MAGUKs) | DLG4 (PSD-95), DLG1–3 (~300 copies/PSD) | PDZ1–PDZ2–PDZ3 (PDB: 1BE9 GLGF loop) + SH3–GK supramodule | NMDAR (GluN2A/B -ESDV), AMPAR/TARPs (Stargazin), Neuroligin-1, Kv1.4, nNOS, GKAP | DLG4 truncating variants cause DLG4-related synaptopathy (intellectual disability, epilepsy, autism spectrum disorder) |
| Layer 2: GKAP / SAPAP | DLGAP1–4 (~150 copies/PSD) | N-terminal phosphorylated 14-aa repeats + C-terminal -QTRL-COOH motif | Bridges PSD-95 GK domain (upper layer) to Shank PDZ domain (deep layer) | DLGAP1/3 variants and Sapap3 knockout cause severe obsessive-compulsive disorder (OCD) and trichotillomania phenotypes |
| Layer 3: Shank / ProSAP | SHANK1, SHANK2, SHANK3 (Chr22q13.3; ~200 copies/PSD) | ANK – SH3 – PDZ – Proline-rich – Zn2+-binding SAM polymer | GKAP, Homer (-PPxxF-), Cortactin, ABI1, WAVE complex | SHANK3 haploinsufficiency/deletion causes Phelan-McDermid Syndrome (22q13.3 deletion) and ~1–2% of Autism Spectrum Disorder (ASD) |
| Layer 4: Homer (Long vs. Short) | HOMER1 (1b/c vs. IEG Homer1a), HOMER2, HOMER3 (~120 copies) | N-terminal EVH1 domain + C-terminal tetrameric coiled-coil (absent in Homer1a) | Shank, mGluR1/5, IP3R (ER calcium release), Ryanodine receptors, TRPC channels | Activity-induced Homer1a acts as a dominant-negative monomer that uncouples Shank-Homer hubs during homeostatic synaptic scaling |
| Dynamic Kinase Hub: CaMKII | CAMK2A (α), CAMK2B (β) (~1,000+ holoenzymes in spine; 80–150 in PSD) | 12-Subunit Dodecameric Holoenzyme (hub ring + calmodulin-binding kinase domains; Thr286) | Translocates upon Ca2+/CaM influx to bind GluN2B (residues 1290–1309), phosphorylating GluA1 Ser831 and Stargazin | Autonomous Thr286 autophosphorylation acts as a molecular memory switch required for LTP and spatial learning |

Liquid-Liquid Phase Separation (LLPS) and CaMKII Dynamics in LTP

Reconstitution of the PSD as a Biomolecular Condensate (ePSD)

In 2016–2018, Mingjie Zhang and colleagues (Zeng et al., 2016, 2018) demonstrated that when purified recombinant PSD-95, GKAP, Shank3, and Homer3 (together with the NMDA receptor GluN2B cytoplasmic tail and Stargazin) are mixed in physiological buffer at their native micromolar synaptic stoichiometries, they spontaneously undergo liquid-liquid phase separation (LLPS) to form micron-sized, membrane-less biomolecular condensate droplets (ePSD). This phase transition is driven by high-valency specific domain interactions (3 PDZ + 1 GK sites per PSD-95 × multivalent GKAP × polymeric Shank SAM × tetrameric Homer coiled-coils). Remarkably, reconstituted ePSD condensates selectively concentrate glutamatergic components (GluN2B, Stargazin, SynGAP) by >100-fold while actively excluding inhibitory GABAergic synaptic scaffold protein Gephyrin, explaining how excitatory and inhibitory synapses remain biochemically segregated on the same dendrite without a bounding membrane.

SynGAP Dispersion, CaMKII Translocation, and Homer1a Homeostatic Remodeling

In the basal (unpotentiated) state, the PSD contains high concentrations of SynGAP (SYNGAP1), a Ras/Rap GTPase-activating protein present at nearly 1:1 stoichiometry with PSD-95. SynGAP forms a trimeric coiled-coil that binds PSD-95 PDZ slots in a phase-separated condensate, competitively keeping AMPA receptor/TARP complexes out of the core PSD and suppressing ERK signaling.

When high-frequency presynaptic glutamate release coincides with postsynaptic depolarization, Mg2+ is expelled from NMDA receptors, triggering a surge of Ca2+ (~10–20 μM) into the dendritic spine head:

  1. CaMKII Activation and Translocation: Ca2+/calmodulin binds the 12-subunit dodecameric holoenzyme of CaMKIIα/β, triggering inter-subunit autophosphorylation at Thr286 (which locks CaMKII into an autonomous active state even after Ca2+ falls) and driving rapid translocation of CaMKII from the spine shaft directly into Layer 1 of the PSD, where it binds the cytoplasmic tail of GluN2B.
  2. SynGAP Dispersion and AMPAR Slot Capture: Active CaMKII phosphorylates SynGAP, reducing its affinity for PSD-95 and causing SynGAP to disperse out of the PSD condensate into the dendritic shaft. Simultaneously, CaMKII phosphorylates the C-terminal tail of Stargazin (TARP γ2), neutralizing its positively charged arginines so it releases from the negatively charged lipid bilayer and binds the newly vacated PDZ1–2 slots on PSD-95, trapping laterally diffusing AMPA receptors (GluA1/GluA2) in the synapse to produce long-term potentiation (LTP).
  3. Homer1a Immediate-Early Uncoupling: Conversely, during prolonged network hyperexcitability, neuronal activity induces transcription of the immediate-early splice variant Homer1a, which possesses the N-terminal EVH1 domain but lacks the C-terminal coiled-coil tetramerization tail. Monomeric Homer1a competitively displaces tetrameric Homer1b/c from Shank, dissolving the ePSD condensate back into the soluble phase to drive homeostatic synaptic downscaling.

Deep Evolutionary Origin of the Postsynaptic Complex Before Neurons

One of the most profound insights of comparative neurogenomics came from sequencing the nerveless marine sponge Amphimedon queenslandica (Sakarya et al., 2007; Srivastava et al., 2010), the placozoan Trichoplax adhaerens (Placozoa), and the unicellular/colonial choanoflagellates (Monosiga brevicollis and Salpingoeca rosetta):

  • Unicellular Holozoa (Choanoflagellates & Capsaspora): Already encode recognizable ancestors of DLG/PSD-95 (with intact PDZ1-3–SH3–GK domain architecture), Shank, and Homer (which localizes to the flagellar collar/nucleus), as well as PMCA calcium pumps and vesicular SNAREs.
  • Porifera (Sponges) and Placozoa: Possess a virtually complete, colinearly interacting "proto-postsynaptic" scaffold comprising DLG/PSD-95 (whose PDZ3 crystal structure in Amphimedon exhibits the exact human GLGF peptide-binding groove), GKAP, Shank, Homer, Cortactin, and CaMKII. In sponge larvae, this proto-PSD complex is co-expressed in ciliated flask cells that sense environmental cues without forming chemical synapses.
  • Ctenophora vs. ParaHoxozoa: Whereas Cnidaria and Bilateria co-opted this complete DLG–GKAP–Shank–Homer scaffold together with ionotropic glutamate receptors to construct excitatory chemical synapses, Ctenophora (comb jellies) lack GKAP and canonical Neuroligin orthologs and rely on an independently assembled presynaptic-triad and syncytial neural architecture—reinforcing the macrosyntenic evidence for independent neural centralization in Ctenophora versus ParaHoxozoa.

Frequently Asked Questions

How does the PSD stay assembled for weeks or months without a surrounding lipid membrane, while individual PSD-95 and AMPA receptor molecules exchange in minutes?

The PSD is a multivalent biomolecular condensate formed by liquid-liquid phase separation (LLPS). Just as water molecules inside a raindrop constantly exchange with water vapor while the droplet itself persists, individual PSD-95, GKAP, Shank, and Homer proteins have modest pairwise binding affinities (Kd ≈ 0.1–10 μM) and exchange on timescales of seconds to minutes; however, because each scaffold molecule possesses 3 to 12 binding sites (high valency), the cross-linked 3D network remains thermodynamically stable as a condensed phase anchored to transmembrane NMDA/AMPA receptors and Zn2+-stabilized Shank SAM polymers.

Why do mutations in SHANK3 and SYNGAP1 cause severe autism spectrum disorder (ASD) and intellectual disability?

Both SHANK3 (mutated in Phelan-McDermid syndrome and ~1–2% of all ASD cases) and SYNGAP1 (accounting for ~1% of all non-syndromic intellectual disability) encode dosage-sensitive master hubs of the excitatory PSD. Shank3 haploinsufficiency collapses the deep Layer 3/4 Shank–Homer–Cortactin lattice, causing dendritic spine shrinkage and loss of NMDA/AMPA receptor currents in striatal and cortical circuits. Conversely, SynGAP1 haploinsufficiency removes the basal phase-separated brake on PSD-95 PDZ slots and Ras/ERK signaling, causing premature, unsaturated insertion of AMPA receptors during early postnatal critical periods and Occluding future LTP.

If sponges and Trichoplax have no neurons or synapses, what does the PSD-95–Shank–Homer complex do in their cells?

In nerveless sponges (Amphimedon queenslandica) and placozoans (Trichoplax adhaerens), the DLG–GKAP–Shank–Homer complex functions as an apical sub-membrane signaling plaque in polarized epithelial, sensory flask, and peptidergic gland cells. There, it couples extracellular receptors and Ca2+ channels to the cortical actin cytoskeleton to regulate cell polarity, ciliary beating, and paracrine exocytosis—an ancestral sensory-secretory module that was later concentrated opposite presynaptic vesicle release sites to create the first eumetazoan chemical synapses.

References

  1. Doyle, D. A., Lee, A., Lewis, J., Kim, E., Sheng, M., & MacKinnon, R. (1996). "Crystal structures of a complexed and peptide-free membrane protein-binding domain: molecular basis of peptide recognition by PDZ." Cell, 85(7), 1067–1076. https://doi.org/10.1016/S0092-8674(00)81307-081307-0)
  2. Sheng, M., & Hoogenraad, C. C. (2007). "The postsynaptic architecture of excitatory synapses: a more quantitative view." Annual Review of Biochemistry, 76, 823–847. https://doi.org/10.1146/annurev.biochem.76.060805.160029
  3. Sakarya, O., Armstrong, K. A., Adamska, M., Adamski, M., Wang, I. F., Tidor, B., ... & Kosik, K. S. (2007). "A post-synaptic scaffold at the origin of animals." PLOS ONE, 2(6), e506. https://doi.org/10.1371/journal.pone.0000506
  4. Zeng, M., Chen, X., Guan, D., Xu, J., Wu, H., Tong, P., & Zhang, M. (2018). "Reconstituted postsynaptic density as a molecular platform for understanding synapse formation and plasticity." Cell, 174(5), 1172–1187. https://doi.org/10.1016/j.cell.2018.06.047
🧬 Support Open Science
Explore interactive science games on LowKey Games or browse scientific apparel & accessories in the CRV Science Shop.
💡 Spot an inaccuracy or have primary literature to expand this article?
Submit a citation or revision note using your shared CRV Science / LowKey Games profile ID.
Suggest Revision