Structural plasticity with preserved topology in the postsynaptic protein network

Abstract

The size, shape, and molecular arrangement of the postsynaptic density (PSD) determine the function of excitatory synapses in the brain. Here, we directly measured the internal dynamics of scaffold proteins within single living PSDs, focusing on the principal scaffold protein PSD-95. We found that individual PSDs undergo rapid, continuous changes in morphology driven by the actin cytoskeleton and regulated by synaptic activity. This structural plasticity is accompanied by rapid fluctuations in internal scaffold density over submicron distances. Using targeted photobleaching and photoactivation of PSD subregions, we show that PSD-95 is nearly immobile within the PSD, and PSD subdomains can be maintained over long periods. We propose a flexible matrix model of the PSD based on stable molecular positioning of PSD-95 scaffolds.

Fig. 1.

Individual PSDs undergo continuous morphological plasticity. (A) Individual spines from hippocampal neurons 4 weeks in culture expressing PSD-95-GFP. Scale bar, 1 μm. (B) Distribution of PSD sizes in spines of hippocampal neurons 4 weeks in culture, transfected with PSD-95-GFP (n = 392 PSDs in 7 neurons). Gray area indicates the average length and breadth of subresolution particles (100 nm beads) acquired and analyzed under the same conditions (average 312 nm). (C) Time-lapse imaging of a single PSD in the spine of a hippocampal neuron. Images acquired at several z planes were maximally projected and interpolated for display. (Scale bar: 1 μm.) See Movie S1. (D) Simultaneous and coordinate reshaping of PSD-95-mCherry with stargazin-GFP. (Scale bar: 1 μm.) (See Movie S2.)

Fig. 2.

PSD structural plasticity is rapid, graded, continuous, and regulated by synaptic activity. (A) Rapid time course of PSD change. Pairwise subtraction of images was used to measure intensity differences at individual PSDs (n = 220 PSDs in 5 cells) over varying time delays between image stacks. Solid line shows the best-fitting sigmoidal logistic equation (half-time of 16 sec and a power of 1.2). Images were not thresholded or binarized for this analysis. (B) Projected images from a time lapse acquired at 1 stack/min were interpolated and segmented, and the morphology (Upper) and intensity (Lower) of the resulting 2D objects were measured. See Movie S4. (C) Asynchronous plasticity of neighboring PSDs. Elliptical form (EF) of several PSDs over time. (Scale bar: 5 μm.) See Movie S5. (D) Thin colored lines plot the EF of 20 PSDs from 1 neuron, and the thick line shows the mean of all PSDs measured. (E) Modes of PSD morphological dynamics defined by the rate of change in EF. Gray region indicates ± 1 standard deviation of dEF/dt for PSDs in 10 control neurons. (F) Continuous entry to dynamic Mode 1 by changing sets of 24 PSDs of a single neuron. Although individual PSDs switch between Mode 1 and Mode 0 as in E, the fraction in Mode 1 remains nearly constant. (G and H) Degree of PSD shape change during alterations in network activity. CV of EF measured in 10 min running bins. Neurons were incubated with 1 μM TTX for 4 days before experiments and during baseline (black bar). (G) Red bar indicates switch to activity-promoting solution containing 50 μM bicuculline (Bic) and 25 μM 4-aminopyridine (4-AP) (n = 5). (H) Red and blue bars indicate change to control solution containing Bic and 4-AP along with 50 μM d-APV and 25 μM DNQX (n = 7). (I) Same experiments as in G and H. CV of EF measured in the 5 bins before solution change and at the peak of the response to increased activity. Individual neurons are plotted as connected points and group mean ± SEM is displayed on the Right. (*, P < 0.02, Mann–Whitney U test) (J) Activity increases dynamic Mode 1 occupancy. Modes were calculated as in E and Mode 1 PSDs were plotted as a ratio of the occupancy in cells that received Bic/4AP to the occupancy in cells exposed to Bic/4-AP+DNQX/APV.

Fig. 3.

Actin drives spatial fluctuations of PSD-95 molecular density in individual PSDs. (A) Fractional protein density maps of single PSDs from several cells. The color scale shows the proportion of total molecules per 1,000 nm². (Spatial scale bar: 1 μm.) (B) Time-lapse density maps of a PSD over time. (Spatial scale bar: 1 μm.) Color scale as in A. (C) Density maps 1 min after application of 2 μM latrunculin A. (Spatial scale bar: 1 μm.) Color scale as in A. See Movie S6. (D–G) Quantification of EF and local protein density of individual PSDs in 3 cells. (D) Elliptical form; (E) CV of EF measured in 10 min running bins; (F) proportion of PSDs occupying dynamic Mode 1; (G) CV of maximum protein density within each PSD measured in 10-min running bins. Arrowheads indicate time of latrunculin A application.

Fig. 4.

Limited mixing of molecules within the PSD. (A) Two alternative models of PSD structure: a “corral” that contains movable molecules within a defined border (Left), or a “matrix” of molecules held in fixed positions within the structure (Right). (B) Linescan time-lapse imaging of PSD-95-GFP. A single line through an individual PSD was scanned repeatedly at 200 Hz. The resulting fluorescence along the line is shown as an xt plot with time advancing downward. An excitation intensity was chosen which resulted in little bleaching during this protocol (“Unbleached”). To photobleach, the intensity of the laser was raised on a portion of the line for 80 ms, appearing in the plot as a thin white line. The bleach was targeted to encompass the entire PSD (“Fully bleached”) or only 1 side of the PSD (“Partially bleached”). (Scale bars: vertical, 3.5 sec; horizontal, 1 μm). (C) XY scans through the PSD shown in B, taken at the Z plane of the linescan before and after the linescan FRAP protocol. Brackets indicate the bleached PSD subregion, which remained dimmer after the recovery period. (D) Intensity profile of the PSD in C along the scanned line before bleaching (thin line), immediately after bleaching (thick line), and 14 sec after bleaching (dashed red line). (E) Average recovery for PSD-95-GFP in fully bleached PSDs (black, n = 10), in partially bleached PSDs (red, n = 12), or in glia (blue, n = 4). PSD-95-GFP in glia recovered with a t_1/2 of 1.7 sec. The blue line is the best fit to the glia recovery data of a double exponential function with time constants 0.4 sec and 6.7 sec predicting a recovery saturating at 85%. (F) Fluorescence overlay photobleaching. Neurons were cotransfected with 2 cDNAs encoding PSD-95 tagged with cerulean or with citrine, and citrine was selectively bleached (yellow box) with 514 nm illumination. Brackets indicate the region of persistent reduction of fluorescent PSD-95-citrine in a subdomain of the PSD. Dashed line in Right middle frame shows the line along which intensity profiles were measured in G. (G) Intensity profile of the PSD in F before bleaching (thin line), 30 sec after bleaching (thick line), and 4 min after bleaching (dashed red line). (H) A PSD microdomain optically labeled through photoactivation of PSD-95-PAGFP within a small region. After imaging over the next minute, the entire PSD was photoactivated. (Scale bar: 1 μm.) (I) GFP intensity measured within subregions of the single PSD shown in H. Measured regions are shown in the Inset. Intensity was normalized to the intensity of each subregion after activation of the whole PSD. The stable intensities in each subregion of the PSD demonstrate the near immobility of synaptic PSD-95. See Fig. S6 for further analysis.

Fig. 5.

Flexible matrix model of PSD architecture. (Left) Scaffold proteins (yellow) establish matrix coordinates within the larger PSD complex (gray). (Right) Components of the matrix can be added (green) or removed (red), and this molecular exchange takes place at defined coordinates, preserving overall structure. Morphological plasticity of the complex is independent of component protein exchange and may control local protein density while protein molecular number remains constant.