Mesoscale model of postsynaptic density proteins.

(A) Schematics and molecular representation of the four PSD components in this simulation study. AMPAR and NMDAR diffuse on membrane, N-terminal part of PSD-95 is embedded on membrane due to the palmitoylation, and CaMKIIα is present beneath the membrane. (B) Specific domain-domain interactions. Every PDZ domains in PSD-95 can interact with TARP and with GluN2Bc. Only active CaMKII can interact with GluN2Bc by its kinase domains. (C) The 3D system simulation box. Core spherical particles of AMPAR and NMDAR are substituted by fluorescent proteins DsRed2 and eqFP670, respectively. (D) The 2D system setup and simulation box. Particles representing AMPAR, NMDAR, and N-terminal of PSD-95 diffuse on the planar membrane (orange), whereas other particles diffuse in the cytosol beneath the membrane (sky-blue).

Trajectories and morphologies in simulations of four-component PSD systems.

(A,B) Representative trajectories in the 3D (A) and 2D (B) systems. Bar chart to the right of the trajectory shows the average molecular compositions of AMPARs (A, red), NMDARs (N, orange), PSD-95s (P, blue) and CaMKIIs (C, green) in the largest cluster in the final 1000 snapshots. Dashed horizontal lines represent the whole number of proteins in the system. (C, E) Overview (left) and cross-sectional view (right) final snapshots of simulations in the 3D system with all CaMKIIs in the activated state (C) and in the inactive state (E). (D, F) Density profiles in the slice of the droplet along z axis (left) and the molecular distribution as a function of the distance from the center of mass of the cluster (right) with all CaMKIIs in the active state (D) and in the inactive state (F). (G, I) Top view (left) and side view (right) snapshots of simulations in the 2D system with CaMKII in the active (G) and inactive (I) states. (H, J) Density profiles in the slice of the cluster on membrane along X axis (left) and molecular distribution on membrane along the distance from the center of mass of the cluster (right) with all CaMKIIs in the active state (H) and in the inactive state (J). Shaded areas indicate standard errors for three independent simulations. The colors of the molecules, trajectories, bar graphs and distributions are corresponding.

Molecular interaction and CaMKII architecture in the four-component PSD assembly.

(A) Number of specific interaction (top) and non-specific interaction (bottom) with active CaMKII in 3D (blue), inactive CaMKII in 3D (sky-blue) active CaMKII in 2D (orange) and inactive CaMKII in 2D (pale orange). Each index indicates the interaction type where A, P, N, and C represent AMPAR, PSD-95, NMDAR, and CaMKII, respectively. (B) Distribution of the number of NMDARs bound to a CaMKII molecules in 3D system (top) and 2D system (bottom). (C) Protein distribution normal to the membrane plane (the z-coordinate) in the 2D system with the active (left) and inactive CaMKIIs (right). The orange and light blue background represent the membrane and cytoplasmic regions, respectively. (D) Two-dimensional kernel density estimation plot of the position and orientation of CaMKII molecules. The horizontal axis represents the distance from the membrane, while the vertical axis shows the angle between the membrane plane and each two hexagonal planes of CaMKII kinases. The height of each histogram and the color of KDE plot are shown in logarithmic scale. (E) Representative snapshots of the 2D systems with active CaMKII. The hexagonal plane of CaMKII is parallel to the membrane plane (left) or perpendicular to the membrane plane (middle). A snapshot of a whole cluster looking from the cytoplasm (right). Orange plane corresponds to the membrane center (z = 0).

Multiphase morphology for CaMKII with reduced volume in the 3D system.

(A) Schematics of CaMKIIs used in each simulation. Original sized CaMKII (WT) (left), CaMKII (r=2/3) with 2/3 of its radius (middle), and CaMKII (r=1/2) with 1/2 of its radius (right). (B) Snapshots of cross-sectional view with original CaMKII (WT) (left), CaMKII (r=2/3) (middle), and CaMKII (r=1/2) (right). (C) Density profile of AMPAR (red) and NMDAR (orange) in the slice of the cluster along Z axis for CaMKII (WT) (left), CaMKII (r=2/3) (middle), and CaMKII (r=1/2) (right). (D) Molecular distribution of AMPAR (red) and NMDAR (orange) from the center of mass of the cluster (right) of the simulation with CaMKII (WT) (top), that with 2/3 in its radius (middle), and with 1/2 in its radius (bottom). As the molecular radius decreases, peak of AMPARs shifted outside while that of NMDARs moved close to the cluster center. (E) Molecular composition of each phases in the simulation with CaMKII (WT) (top), CaMKII (r=2/3) (middle), and CaMKII (r=1/2) (bottom). In each case, the three bars indicate the numbers of molecules in the AMPAR-containing phase (top bar), the NMDAR-containing phase (middle bar) and the dilute phase (bottom bar).

Multiphase morphology for CaMKII with reduced valency in the 2D system.

(A) Schematics of CaMKII used in each simulation. Original 12 valence CaMKII (WT) (left), 6 valence CaMKII (V=6) (middle), and 3 valence CaMKII (V=3) (right). (B) Snapshots of cross-sectional view with CaMKII (WT) (left), that with CaMKII (V=6) (middle), and with CaMKII (V=3) (right). (C) Density profile of AMPAR (red) and NMDAR (orange) in the slice of the cluster along X axis of the simulation with CaMKII (WT) (left), CaMKII (V=6) (middle), and CaMKII (V=3) (right). As the valency decreases, the architecture of the core-shell cluster shifts between AMPARs and NMDARs. (D) Molecular distribution of AMPAR (red) and NMDAR (orange) from the center of mass of the cluster (right) of the simulation with CaMKII (WT) (top), that with CaMKII (V=6) (top middle), that with CaMKII (V=3) (bottom middle), and that with inactive CaMKII (bottom). As the molecular valence decreases, peak of AMPARs shifts inside while that of NMDARs moves far from the cluster center. (E) Molecular composition of each cluster in the simulation with CaMKII (WT) (top), that with CaMKII (V=6) (top middle), that with CaMKII (V=3) (bottom middle), and that with inactive CaMKII (bottom). The three bars indicate the number of molecules in each of the phases containing AMPARs (top bar), NMDARs (middle bar) and others (bottom bar), respectively. Composition of phase with NMDAR does not change drastically between different CaMKII sizes.

Interfacial tensions explain the multiphase morphology in 3D system.

(A) Representative snapshots of 3D slab simulation used in the measurement of interfacial tension between A (AMPAR-containing, red) and N (NMDAR-containing, green) phases, with CaMKII (WT) (top), CaMKII (r=2/3) (middle), and CaMKII (r=1/2) (bottom). (B) Interfacial tensions between the A and dilute phases (γA, red line), between the N and dilute phases (γN, orange line) and between the A and N phases (γAN, cyan line) with error bars at each condition. (C) Phase diagram on the γA⁄γAN and γN⁄γAN plane. Dashed lines represent theoretical borders of multiphase morphology.

Conditions of concentrations in 3D system

Concentrations of each molecules in dilute phase

Conditions of molecular number in 2D system

Conditions of molecular number in 2D system

Three pairs of molecules with specific interactions between those domains.

PDZ domains in PSD-95 can interact with PDZ binding motif (PBM) in TARP (top left) and with PBM in C-terminus of GluN2B subunit (top right). Only active CaMKII can interact with GluN2Bc by its kinase domains (bottom). Interaction strengths between all pair of molecules are set to reproduce the Kd values obtained from previous studies.

Number of specific and nonspecific interactions in each system and each CaMKII state.

(A) Number of specific interactions in 3D system with active CaMKII (top left) and that in 3D system with inactive CaMKII (top right), that in 2D system with active CaMKII (bottom left) and that in 2D system with inactive CaMKII (bottom right). (B) Number of non-specific interactions in 3D system with active CaMKII (top left) and that in 3D system with inactive CaMKII (top right), that in 2D system with active CaMKII (bottom left) and that in 2D system with inactive CaMKII (bottom right). Results in both conditions are obtained by averaging the last 100 trajectories of 3 independent trajectories.

Multiphase morphology for CaMKII with reduced valency in the 3D system.

(A) Schematics of CaMKIIs used in each simulation. Original 12 valence CaMKII (WT) (left), 6 valence CaMKII (V=6) (middle), and 3 valence CaMKII (V=3) (right). (B) Snapshots of cross-sectional view with original CaMKII (WT) (left), CaMKII (V=6) (middle), and CaMKII (V=3) (right). Inactive kinase domains are expressed by grey particles in the snapshots. (C) Density profile of AMPAR (red) and NMDAR (orange) in the slice of the cluster along z axis for CaMKII (WT) (left), CaMKII (V=6) (middle), and CaMKII (V=3) (right). (D) Molecular distribution of AMPAR (red) and NMDAR (orange) from the center of mass of the cluster (right) of the simulation with CaMKII (WT) (top), that with CaMKII (V=6) (middle), and with CaMKII(V=3) (bottom). Despite the molecular valency reduction, both peaks of AMPARs and NMDARs remain unchanged, while the distribution of NMDARs moved more broadly to the cluster center as the valency decrease. (E) Molecular composition of each phases in the simulation with CaMKII (WT) (top), CaMKII (V=6) (middle), and CaMKII (V=3) (bottom). In each case, the three bars indicate the numbers of molecules in the AMPAR-containing phase (top bar), the NMDAR-containing phase (middle bar) and the dilute phase (bottom bar).

Multiphase morphology for CaMKII with reduced volume in the 2D system.

(A) Schematics of CaMKIIs used in each simulation. Original sized CaMKII (WT) (left), CaMKII (r=2/3) with 2/3 of its radius (middle), and CaMKII (r=1/2) with 1/2 of its radius (right). (B) Snapshots of cross-sectional view with CaMKII (WT) (left), that with CaMKII (r=2/3) (middle), and with CaMKII (r=1/2) (right). (C) Density profile of AMPAR (red) and NMDAR (orange) in the slice of the cluster along X axis of the simulation with CaMKII (WT) (left), CaMKII (r=2/3) (middle), and CaMKII (r=1/2) (right). As the valency decreases, the inward propensity of NMDAR-containing cluster decreased and the AMPARs and NMDARs are separated into their own clusters. (D) Molecular distribution of AMPAR (red) and NMDAR (orange) from the center of mass of the cluster of the simulation with CaMKII (WT) (top), that with CaMKII (r=2/3) (middle), and that with CaMKII (r=1/2) (bottom). (E) Molecular composition of each cluster in the simulation with CaMKII (WT) (top), that with CaMKII (r=2/3) (middle), and that with CaMKII (r=1/2) (bottom). The three bars indicate the number of molecules in each of the phases containing AMPARs (top bar), NMDARs (middle bar) and others (bottom bar), respectively. Composition of phase with NMDAR does not change drastically between different CaMKII sizes.

Alleviating membrane constraints controls the partitioning state of receptors.

(A) Final snapshots with weaker membrane constraints (k = 0.01 kJ/mol) in 2D system. Top view (top left) and side view (bottom left) are the results with active CaMKII, and top view (top right) and side view (bottom right) are the results with inactive CaMKII. (B) Density profile of AMPAR (red) and NMDAR (orange) in the slice of the cluster along X axis of the simulation with active CaMKII (left), and inactive CaMKII (right). (C) Molecular distribution of AMPAR (red) and NMDAR (orange) from the center of mass of the cluster of the simulation with active CaMKII (top) and inactive CaMKII (bottom). (D) Final snapshots with weaker membrane constraints (k = 0.01 kJ/mol) and wider membrane region (thickness = 50 nm) in 2D system. Top view (top left) and side view (bottom left) are the results with active CaMKII, and top view (top right) and side view (bottom right) are the results with inactive CaMKII. (E) Density profile of AMPAR (red) and NMDAR (orange) in the slice of the cluster along X axis of the simulation with active CaMKII (left), and inactive CaMKII (right). (F) Molecular distribution of AMPAR (red) and NMDAR (orange) from the center of mass of the cluster of the simulation with active CaMKII (top) and inactive CaMKII (bottom).

Switching dissociation constants of AMPAR-PSD-95 and NMDAR-CaMKII has little effect on the multiphase morphology both in 3D and 2D system.

(A) Overview (left) and cross-sectional view (right) final snapshots of switching the Kd of the AMPAR-PSD-95 and NMDAR-CaMKII specific interactions in 3D system. (B) Density profiles in the slice of the droplet along z axis (left) and the molecular distribution as a function of the distance from the center of mass of the cluster (right). (C) Top view (left) and side view (right) snapshots of simulations in the 2D system in the condition of switching the Kd. (D) Density profiles in the slice of the cluster on membrane along x axis (left) and molecular distribution on membrane along the distance from the center of mass of the cluster (right).

Fixed molecular composition does not change the relative relationship between the interfacial tensions between the phases in 3D system.

(A) Interfacial tensions of NMDARs (orange solid line) and AMPAR-NMDAR (cyan solid line) with error bars at each condition. In the fixed molecular composition, simulation system of N phase contains 180 copies of NMDARs, 180 copies of PDS-95, 60 copies of CaMKIIs. Interfacial tension between A phase and N phase is recalculated using the result of the fixed molecular composition of N phase. Dashed lines are the same plots as shown in the Fig.6B, merely for reference. (B) Plot with fixed molecular composition (violet solid line, denoted as “fixed”). Violet dashed line (variables) is the same plot as shown in the Fig.6C, merely for reference.

Validation results of our mesoscale model.

(A) Estimation results of the concentration of PSD-95 in the dilute phase for each upper cutoff distance limit. Among the cutoff value from 1.0 nm to 1.8 nm, we choose 1.5 nm as a cutoff limit since it was the shortest distance which almost reproduces the critical concentration of PSD-95. (B) Snapshots of the simulation result in three cutoff conditions: rc = 1.0 nm (top), rc = 1.4 nm (middle), and rc = 1.8 nm (bottom). The volume of the condensed phase decreases as the cutoff distance decreases. (C) Estimated dissociation constants (Kd) between AMPAR-PSD-95 (top), GluN2Bc-PSD-95 (middle), and GluN2Bc-CaMKII kinase (bottom), calculated by using single molecule simulations based on our previous methods (14). The numbers in the boxes represent the logarithm of the dissociation constant. For the molecular dissociation reaction in each molecular pair, the X-axis value of koff, which takes the value of dissociation constant shown in yellow color in the diagram, was adopted as the reaction rate constant and used in the subsequent simulations.