Destabilization of the postsynaptic density by PSD-95 serine 73 phosphorylation inhibits spine growth and synaptic plasticity

Abstract

Long-term potentiation (LTP) is accompanied by dendritic spine growth and changes in the composition of the postsynaptic density (PSD). We find that activity-dependent growth of apical spines of CA1 pyramidal neurons is accompanied by destabilization of the PSD that results in transient loss and rapid replacement of PSD-95 and SHANK2. Signaling through PSD-95 is required for activity-dependent spine growth and trafficking of SHANK2. N-terminal PDZ and C-terminal guanylate kinase domains of PSD-95 are required for both processes, indicating that PSD-95 coordinates multiple signals to regulate morphological plasticity. Activity-dependent trafficking of PSD-95 is triggered by phosphorylation at serine 73, a conserved calcium/calmodulin-dependent protein kinase II (CaMKII) consensus phosphorylation site, which negatively regulates spine growth and potentiation of synaptic currents. We propose that PSD-95 and CaMKII act at multiple steps during plasticity induction to initially trigger and later terminate spine growth by trafficking growth-promoting PSD proteins out of the active spine.

Figure 1. PSD-95 Regulates Activity-Dependent Spine Growth

(A) (Top) Example of a region of apical dendrite of a CA1 pyramidal neuron expressing dsRed that was imaged repetitively over 31 min. The acquisition times in minutes relative to the start of imaging are given. White arrowheads in this and all figures indicate the spines that were stimulated by 2PLU of MNI-glutamate and the time of plasticity-inducing stimulus (PS) onset. PS triggers an enlargement of the targeted spine. (Bottom) As above for neurons in the presence of 10 μM CPP, which blocks NMDARs and prevents activity-dependent spine growth. (B) Time course of head cross-sectional area of stimulated spines (red, n = 22/4 spines/cells), unstimulated neighboring spines (black, n = 16/4 spines/cells), or spines stimulated in the presence of CPP (blue, n = 12/3 spines/cells). In all summary graphs, the black bar indicates the timing of the PS. * and # indicate statistical difference of p < 0.05 for the area of stimulated compared to unstimulated spines either 1 min (*) or averaged 20–30 min (#) after the stimulus. (C) (Left) Example of LTP of uEPSCs evoked by PS delivered to a visualized spine of a voltage-clamped neuron filled with Alexa Fluor 594. The red points correspond to uEPSCs during the PS. (Right) Single optical slice showing the stimulated spine (top) and the uEPSC (bottom) before (1) and after (2) the PS. (D) Time course of the changes in uEPSCs before and after the PS (n = 5/5 spines/cell). (E) As in panel (A) for hippocampal neurons expressing dsRed and shPSD-95. (F) Time course of spine head area of neurons expressing dsRed and shPSD-95 (red, n = 41/10 spines/cells). For comparison, data from panel (B) for dsRed-only expressing neurons are replotted (gray). In all plots of spine area, * and # indicate p < 0.05 compared to the data plotted in gray, respectively, at 1 min (*) or averaged 20–30 min (#) after PS. Error bars depict the SEM.

Figure 2. PSD-95 Transiently Leaves the Spine Head during Activity-Dependent Growth

(A) Images of spines from neurons expressing dsRed (red) and PSD-95-PAGFP (green). PSD-95-PAGFP in spines in the indicated areas (white boxes) was photoactivated at minute 0. Fluorescence intensity was monitored at the indicated times in minutes. At the end of the imaging period, the spines were exposed to a second photoactivating pulse (PA′). (B) As in panel (A), with delivery of PS to the spine (arrowhead) at minute 10. (C) Time course of PSD-95-PAGFP fluorescence in unstimulated spines following photoactivation (n = 45/9 spines/cells). (D) Time course of PSD-95-PAGFP fluorescence in spines after photoactivation and stimulation with PS at minute 10 (red, n = 23/7 spines/cells). For comparison, the data from unstimulated spines are replotted (gray). Open symbols indicate statistical difference of p < 0.05 compared to the same time points for unstimulated spines. (E) Images of a spine expressing dsRed (red) and WT PSD-95-GFP (green) that was stimulated by PS at 10 min. (F) Time course of WT PSD-95-GFP fluorescence for spines stimulated with PS at 10 min (black, n = 11/3 spines/cells). The open symbol indicates statistical difference (p < 0.05) between the fluorescence at 11 min compared to 10 min. Error bars depict the SEM.

Figure 3. CaMKII-Dependent Phosphorylation of PSD-95 Negatively Regulates Activity-Dependent Spine Growth

(A) Images of stimulated dendritic spines from neurons pretreated with 10 μM of KN-93 and expressing dsRed alone (top) or dsRed and WT PSD-95 (bottom). (B) Time course of head area (red) of stimulated spines from neurons expressing dsRed alone (left, n = 10/3 spines/cells) or dsRed and WT PSD-95 (right, n = 17/5 spines/cells) in the presence of KN-93. The gray area shows the data for spines in the absence of KN-93. (C) Images of stimulated spines from neurons expressing dsRed and either S73A (top) or S73D (bottom) PSD-95. (D)Time courses of head area (red) of stimulated spines from neurons expressing S73A (left, n = 20/7 spines/cells) or S73D (right, n = 21/7 spines/cells) PSD-95. For comparison, data from neurons expressing WT PSD-95 are replotted (gray). Error bars depict the SEM.

Figure 4. Regulation of PSD-95 Serine 73 Is Not Necessary for the Effects of PSD-95 on Synaptically Evoked AMPAR Currents

(A) Amplitudes of AMPAR (left) and NMDAR (right) EPSCs of neurons transduced with a lentivirus encoding shPSD95 and WT PSD-95-GFP plotted against those recorded simultaneously in uninfected neighboring neurons. Each symbol represents the results of a single experiment, with the exception of red symbols, which show the mean ± SEM across experiments. The insets show example average traces from infected (gray) and uninfected (black) neurons from a single experiment. (B and C) As in (A) for neurons transduced with shPSD95 S73A PSD-95 or shPSD95 S73D PSD-95, respectively. (D) Summary of AMPAR and NMDAR EPSC amplitudes in neurons expressing shPSD95 PSD-95-GFP, shPSD95 S73A PSD-95-GFP, or shPSD95 S73D PSD-95-GFP expressed as a ratio to those in neighboring uninfected neurons (n.s. indicates p > 0.05). Error bars depict the SEM.

Figure 5. LTP Requires Regulation of PSD-95 at Serine 73

(A) Representative images before (left) and 1 min after (right) delivery of PS. The white arrowhead indicates the spine that was stimulated. Below the right panel are shown uEPSCs recorded ~20 min after PS from control spines (1 and 3) and from the stimulated spine (2). The line and shaded regions show the mean and mean ± SEM uEPSC for each spine. (B) (Left) uEPSC_control measured in neurons expressing GFP (n = 65/15 spines/cells). (Right) Normalized uEPSCs measured from control (black) and stimulated (red, n = 15/15 spines/cells) spines from GFP-expressing neurons. The dashed line depicts the normalized amplitude of one of the control uEPSC. (C–E) As in panel (B) for neurons expressing GFP and either WT PSD-95 ([C], control n = 69/11, stimulated n = 11/11 spines/cells), S73A PSD-95 ([D], control n = 58/11, stimulated n = 11/11 spines/cells), or S73D PSD-95 ([E], control n = 58/12, stimulated n = 12/12 spines/cells). (F) Ratio of potentiation (R_pot) of the uEPSC at the stimulated spine compared to unstimulated neighboring spines in neurons expressing GFP alone or GFP and either WT PSD-95, S73A PSD-95, or S73D PSD-95. All differences across conditions with the exception of WT PSD-95 versus S73A PSD-95 are significant. Error bars depict the SEM.

Figure 6. Regulation at S73 Controls Basal and Activity-Dependent Trafficking of PSD-95

(A) Images of spines from neurons expressing dsRed and S73A PSD-95-PAGFP. PAGFP in a spine was photoactivated (white box, 0 min), and subsequent changes in fluorescence intensity were monitored. The same spine was subjected to a second photoactivation pulse (PA′) at the end of the imaging period. (B) As in panel (A) except that the spine received a PS between minutes 10 and 11 (white arrow). (C) Average time course of S73A PSD-95-PAGFP fluorescence in spines after photoactivation (red, n = 35/14 spines/cells). For comparison, the time course of fluorescence in spines of WT PSD-95 transfected neurons is replotted in gray. (D) As in panel (C) for spines stimulated with PS at minute 10 (red, n = 14/5 spines/cells). The gray region shows the behavior of S73A PSD-95-PAGFP in unstimulated spines replotted from panel (C). (E and F) As in panels (A) and (B) for neurons expressing S73D PSD-95-PAGFP in basal conditions or receiving PS at minute 10, respectively. (G) As in panel (C) for neurons expressing S73D PSD-95-PAGFP (red, n = 35/9 spines/cells). Open markers indicate p < 0.05 compared to the time course of WT PSD-95-PAGFP fluorescence (gray). (H) As in panel (D) for neurons expressing S73D PSD-95-PAGFP (red, n = 23/8 spines/cells). The gray region depicts the behavior of S73D PSD-95-PAGFP in unstimulated spines from panel (G). Error bars depict the SEM.

Figure 7. N- and C-Terminal Interactions of PSD-95 Are Necessary for Activity-Dependent Spine Growth

(A) Representative images of spines from neurons expressing dsRed and ΔPDZ1, 2 PSD-95-GFP. The indicated spines were stimulated with PS between minutes 10 and 11. (B–D) As in panel (A) for neurons expressing dsRed and ΔGK PSD-95-GFP (B), S73A-ΔGK PSD-95-GFP (C), or S73D-ΔGK PSD-95-GFP (D). (E) Average time course of head area of stimulated spines from neurons expressing dsRed and ΔPDZ1, 2 PSD-95-GFP (red, n = 38/11 spines/cells). Data from stimulated spines overexpressing WT PSD-95 are shown in gray. (F–H) As in panel (E) for neurons expressing dsRed and either ΔGK PSD-95-GFP ([F], n = 10/3 spines/cells), S73A-ΔGK PSD-95-GFP ([G], n = 17/5 spines/cells), or S73D-ΔGK PSD-95-GFP ([H], n = 13/4 spines/cells). Error bars depict the SEM.

Figure 8. Activity-Dependent Spine Growth Triggers Rapid and Transient Translocation of SHANK2 out of the Spine Head

(A) Images from neurons expressing dsRed and PAGFP-SHANK2. Selected spines were photoactivated (white box, 0 min), and fluorescence was monitored over time as in Figure 4. (B) Time course of PAGFP-SHANK2 fluorescence in spine heads after photoactivation (red, n = 19/3 spines/cells). The data corresponding to WT PSD-95-PAGFP are replotted (gray) for comparison, and statistically significant (p < 0.05) differences are indicated by open symbols. (C) Images of spines from neurons expressing dsRed and PAGFP-SHANK2. A single spines was photoactivated (white box, 0 min) and stimulated with PS (arrowhead, 10 min). (D) Time course of stimulated spine areas from neurons expressing dsRed and PAGFP-SHANK2 (red, n = 13/4 spines/cells). The gray area corresponds to activity-dependent spine growth measured in dsRed-expressing cells. (E) Time course of PAGFP-SHANK2 fluorescence in spine heads after photoactivation and PS (red, n = 8/3 spines/cells). The gray area shows the PAGFP-SHANK2 fluorescence of unstimulated spines. Open red circles indicate statistically significant differences (p < 0.05) between stimulated and unstimulated spines. (F) Time-lapse images of spines expressing dsRed and GFP-SHANK2. A single spine received PS (arrowhead, 10 min). (G) Time course of GFP-SHANK2 fluorescence in spines after PS (black, n = 9/3 spines/cells). The black open circle indicates a statistically significant difference (p < 0.05) between the data at the 10 and 11 min time points. Error bars depict the SEM.

Figure 9. The GK Domain and S73 of PSD-95 Regulate Activity-Dependent Trafficking of SHANK2

(A) Images from neurons expressing dsRed, PAGFP-SHANK2, and either WT PSD-95 (top) or ΔGK PSD-95 (bottom). Spines were photoactivated (white box, 0 min) and stimulated with PS (arrowhead, 10 min). (B) Time course of head area of stimulated spines from neurons expressing dsRed, WT PSD-95, and PAGFP-SHANK2 (red, n = 15/5 spines/cells). The gray area corresponds to data for spines expressing dsRed and PAGFP-SHANK2. (C) Time course of PAGFP fluorescence from spines expressing dsRed, WT PSD-95, and PAGFP-SHANK2 that received PS (red, n = 15/5 spines/cells). The gray area corresponds to data for spines expressing dsRed and PAGFP-SHANK2 and stimulated by PS. (D) As in panel (B) for spines of neurons expressing dsRed, ΔGK PSD-95, and PAGFP-SHANK2 (red, n = 20/5 spines/cells). (E) As in panel (C) for spines of neurons expressing dsRed, ΔGK PSD-95, and PAGFP-SHANK2 and stimulated with PS (red, n = 20/5 spines/cells). Open circles indicate p < 0.05 compared to data for spines expressing dsRed and PAGFP-SHANK2 (gray). (F) Summary graph of relative areas at minute 11 (black) or averaged between minutes 21 and 31 (red) of spines from neurons of the indicated genotypes. * indicates p < 0.05 for comparisons within each genotype (black versus red bars). # indicates p < 0.05 for the comparison across genotypes to the data from WT PSD-95 and PAGFP-SHANK2 expressing neurons. (G) Summary graph of PAGFP fluorescence at minutes 10 (black) and 11 (red) for spines of neurons of the indicated genotype and that received PS between these time points. * and # as in (F). Error bars depict the SEM.

Figure 10. Model of Regulation of Activity-Dependent Structural Plasticity by CaMKII-Dependent Phosphorylation of PSD-95 at S73

(A) In the basal state, PSD-95 molecules are stably incorporated in the PSD such that the rate of exchange of proteins across the spine neck is ~0.001/s. In contrast, SHANK molecules are exchanged at a higher rate of ~0.01/s. The basal rate of exchange of SHANK is independent of interactions with PSD-95 via GKAP (yellow circle) since overexpression of ΔGK PSD-95 or expression of destabilized PSD-95 mutants (S73D) does not alter the exchange rate of SHANK. Orange arrows represent protein movement, whereas black arrows (subsequent panels) schematize activation of signaling cascades. (B) During plasticity-inducing stimulation, NMDAR opening causes the translocation of CaMKII to the PSD and stimulates the formation of a growth-promoting complex. This complex likely contains GKAP and SHANK as well as other proteins (symbolized by X, Y, and Z) that promote actin reorganization. The action of the growth-promoting complex requires PSD-95 since its knockdown or mutation of its N or C termini impairs spine growth. (C) CaMKII and possibly other CaMKs stabilize activity-dependent growth and are necessary for the sustained phase of spine growth. The action of CaMKII is shown downstream of PSD-95 since mutants of PSD-95 eliminate all phases of spine growth, whereas blockade of CaMKs only prevents sustained growth. (D) CaMKII phosphorylates PSD-95 at S73 and terminates spine growth by inducing the translocation of PSD-95 and SHANK out of the active spine, which we propose reflects the disassembly of the growth-promoting complex. It is possible that a PSD-95- and SHANK-containing complex is trafficked out of the active spine as a whole. The role of CaMKII in terminating growth and disassembling the complex are supported by the finding that the nonphosphorylatable mutant of PSD-95 (S73A) enhances growth and prevents the activity-dependent translocation of PSD-95 and SHANK. Furthermore, expression of a mutant that mimics phosphorylation (S73D) impairs growth and LTP. This mutant also basally destabilizes PSD-95, which we propose prevents the formation of a stable growth-promoting complex. The loss of PSD-95 and SHANK from the PSD is transient, and these proteins are rapidly replaced.