Abstract
PSD-95 is a neuronal PDZ protein that associates with receptors and cytoskeletal elements at synapses, but whose function is uncertain. We found that overexpression of PSD-95 in hippocampal neurons can drive maturation of glutamatergic synapses. PSD-95 expression enhanced postsynaptic clustering and activity of glutamate receptors. Postsynaptic expression of PSD-95 also enhanced maturation of the presynaptic terminal. These effects required synaptic clustering of PSD-95 but did not rely on its guanylate kinase domain. PSD-95 expression also increased the number and size of dendritic spines. These results demonstrate that PSD-95 can orchestrate synaptic development and are suggestive of roles for PSD-95 in synapse stabilization and plasticity.
Despite the central role for synapses in neuronal function, mechanisms underlying synapse formation remain incompletely understood. Recently, proteins containing PDZ motifs have been proposed as molecular scaffolds for receptors and cytoskeletal elements at synapses (1–4). The prototypical PDZ protein, postsynaptic density–95 (PSD-95/SAP-90), is a membrane-associated guanylate kinase (MAGUK) concentrated at glutamatergic synapses (5, 6). PSD-95 may participate in synapse development because it clusters at synapses before other postsynaptic proteins (7), and becausediscs large, a PSD-95 homolog in Drosophila, is necessary for proper development of larval neuromuscular junctions (8). Despite numerous studies it remains uncertain whether PSD-95 participates in synapse development in mammals. Targeted disruption of PSD-95 in mice does not alter synaptic structure (9), possibly because three other MAGUKs and dozens of other PDZ proteins occur at brain synapses. This molecular redundancy has obscured understanding of functions for PSD-95 and other PDZ proteins in the brain.
We overexpressed PSD-95 to help define its roles (10). Green fluorescent protein (GFP)–tagged versions of PSD-95 target faithfully to postsynaptic sites in hippocampal neurons, despite being overexpressed 5 to 10 times above endogenous levels (11, 12). To evaluate the effects of PSD-95 on synaptic development, we analyzed cultures at early developmental stages, day in vitro (DIV) 10 to 12, and noted an increase of glutamate receptor subunit–1 (GluR1) immunofluorescence at postsynaptic sites of neurons transfected for PSD-95 (Fig. 1A). This increase was detected both in permeabilized cells with an antibody to a cytosolic GluR1 epitope (Fig. 1A) and in live cells with an antibody to an extracellular site (13). When compared with neighboring untransfected cells, neurons overexpressing PSD-95 exhibited synaptic GluR1 labeling ∼250% of control. By contrast, PSD-95 overexpression did not alter synaptic clustering ofN-methyl-d-aspartate (NMDA) receptor–1 (NR1) (Fig. 1B). These results are surprising because PSD-95 directly binds to NMDA receptors (NMDARs) but not to GluRs (1,3). This suggests that PSD-95 may have an unexpected role in synaptic assembly.
Expression of PSD-95 enhances synaptic clustering of AMPA but not NMDA receptors. (A) Hippocampal neurons were transfected with PSD-95–GFP, fixed at DIV 12, and stained for GluR1 or NR1. Higher magnification micrographs of the boxed regions are shown in the panels to the right. Clusters of GluR1 are more intense in spines from the neuron transfected with PSD-95 (arrowheads) than in spines from the neighboring untransfected neuron (arrows). (B) NR1 staining is equally intense in spines from transfected (arrowheads) and untransfected (arrows) neurons. (C) GluR1 shows spiny clusters (arrowheads) in GAD-negative pyramidal cells and forms shaft clusters (arrows) in GAD-positive interneurons. (D and E) GluR1 clustering is selectively enhanced both in pyramidal cells and in interneurons overexpressing PSD-95–GFP, whereas NR1 is not. ***P < 0.001. Scale bars, 10 μm.
We wondered whether these postsynaptic effects of PSD-95 overexpression occur in both excitatory pyramidal neurons and inhibitory interneurons. Glutamic acid decarboxylase (GAD)–negative pyramidal neurons had a punctate distribution for PSD-95 and GluR1 at spiny protrusions of the dendritic membrane, whereas the GAD-positive interneurons showed more linear clustering along the dendritic shaft (Fig. 1, C and D). Quantitating the intensity of synaptic staining showed that PSD-95 transfection selectively enhanced GluR1 clustering in both pyramidal cells and interneurons (Fig. 1, D and E) and the synaptic GluR1 fluorescence intensity correlated with the PSD-95 expression (13). Overexpression of other postsynaptic proteins, including Ca2+/calmodulin-dependent protein kinase II (CaMKII) and nNOS, did not affect GluR1 clustering (13).
To determine whether this PSD-95–mediated enhancement of GluR1 clustering augments postsynaptic function, we measured miniature excitatory postsynaptic currents (mEPSCs) (14). Each recorded cell was identified as excitatory or inhibitory by injecting with Lucifer yellow and double labeling for GAD-65 (Fig. 2A). GAD-positive interneurons in our cultures had larger and more numerous mEPSCs than GAD-negative pyramidal neurons (Fig. 2, B1 and C1), consistent with differences in GluR1 clustering detected anatomically (Fig. 1C). For both pyramidal cells and interneurons, transfection with PSD-95 augmented the amplitude of mEPSCs, indicating that the additional GluRs recruited by PSD-95 are functional (Fig. 2, B and C). The frequency of mEPSCs was also enhanced by PSD-95 transfection (Fig. 2, B to E).
PSD-95 overexpression increases the amplitude and frequency of AMPA receptor–mediated miniature synaptic currents. (A) To determine neurotransmitter phenotypes for all PSD-95– GFP–transfected and –untransfected cells, Lucifer yellow was injected during recording, and cultures were later stained for GAD-65, which identifies this cell as an inhibitory interneuron. Both PSD-95–overexpressing pyramidal cells (B) (n = 16, 13) and interneurons (C) (n = 19, 10) show increased mEPSC amplitudes and frequencies (**P < 0.01, ***P < 0.001). PSD-95–transfected cells have increased mEPSC amplitude when compared with neighboring untransfected cells for both pyramidal cells (B2, n = 8 pairs) and interneurons (C2, n = 9 pairs). (D andE) Current traces from neighboring untransfected (top) and PSD-95–GFP–expressing (bottom) pyramidal cells (D) and interneurons (E). mEPSCs in untransfected pyramidal cells are marked with an asterisk.
Changes in mEPSC frequency generally reflect presynaptic effects (15); we therefore assessed whether postsynaptic expression of PSD-95 might alter the presynaptic terminal. Staining for synaptophysin and synaptic vesicle protein 2 (SV-2) was enhanced in axon terminals contacting postsynaptic sites of PSD-95–transfected pyramidal cells (synaptophysin density = 257 ± 35% of control; SV-2 = 245 ± 24% of control) [Fig. 3, A and D (13)], as well as interneurons (synaptophysin density = 203 ± 6% of control). We also labeled transfected cultures with FM4-64, which marks sites of synaptic vesicle endocytosis (16). Labeling by FM4-64 was markedly enhanced at presynaptic sites that oppose postsynaptic sites labeled by PSD-95– GFP (FM4-64 intensity = 184 ± 17% of control) (Fig. 3B), suggesting a larger presynaptic vesicle pool size and supporting our physiological data of increased presynaptic release (Fig. 2).
Postsynaptic expression of PSD-95 enhances presynaptic development. (A) PSD-95–GFP–transfected hippocampal neurons were fixed at DIV 10 and stained for GluR1 and synaptophysin. At synapses onto neurons transfected with PSD-95 (arrowheads), both GluR1 and synaptophysin (Syn) staining are more intense than at synapses onto untransfected neurons (arrows). (B) Hippocampal neurons transfected with PSD-95–GFP and incubated with 15 μM FM4-64 in the presence of 90 mM KCl for 45 s show enhanced staining of FM4-64 at sites opposing PSD-95–GFP clusters (arrowheads) than at untransfected synapses (arrows). (C) Expression of the palmitoylation-deficient mutant form of PSD-95 (PSD-95:C3,5S) reduces clustering of GluR1. (D) Quantitative analysis of synaptic changes resulting from PSD-95, PSD-95:C3,5S (C3,5S), and PSD-95ΔGK transfections (***P < 0.001). Scale bars, 10 μm.
Does the enhancement of synaptic function by PSD-95 require its targeting to synaptic sites? We transfected neurons with a PSD-95 mutant (PSD-95:C3,5S) lacking NH2-terminal palmitoylation, which is required for synaptic clustering (11). PSD-95:C3,5S occurred diffusely in hippocampal neurons (Fig. 3C). It did not enhance GluR1 clustering but instead partially disrupted GluR1 clustering in pyramidal cells (Fig. 3, C and D). Neurons transfected with this mutant also failed to display augmented mEPSC amplitude and frequency {untransfected: amplitude = 9.14 ± 0.37 pA, frequency = 1.57 ± 1.00 Hz; PSD-95:C3,5S: amplitude = 8.60 ± 0.60 pA [not significant (NS), P = 0.4], frequency = 0.79 ± 0.24 Hz [NS, P = 0.44]}. The palmitoylation-deficient mutant of PSD-95 thus failed to enhance GluR1 clustering or presynaptic maturation and may function as a partially dominant-interfering mutant.
Overexpression of PSD-95 also augmented postsynaptic clustering of a PSD-95–associated protein, guanylate kinase associated protein (GKAP) (17), but clustering of a noninteracting protein, CaMKII, was not affected (13). We wondered whether the increase in synaptic GKAP might mediate enhanced GluR clustering by PSD-95, because GKAP binds to an actin-associated postsynaptic complex containing Shank (18). However, neurons transfected with PSD-95 lacking the GK domain still showed enhanced postsynaptic clustering of GluR1 and presynaptic aggregation of synaptophysin in the absence of augmented GKAP clustering (13) (Fig. 3D). The GKAP-Shank complex does not, therefore, appear to be necessary for enhanced synaptic development by PSD-95.
Given that PSD-95 enhances a number of pre- and postsynaptic markers at early developmental stages, we wondered if there might be associated morphological changes at later stages. We therefore allowed neurons to develop in culture for 3 weeks, filled them with Lucifer yellow, and compared the morphology of transfected cells with that of their neighbors. Dendritic spines detected in PSD-95–transfected neurons were both more numerous and larger than those in untransfected neurons (Fig. 4). The modest increase in spine count may have resulted from a proliferation of spines or from the inclusion of enlarged spines that would otherwise have been undetectable. Most notable was the increased density of large spines >1 μm in diameter (Fig. 4C).
PSD-95 overexpression enhances spine maturation. Hippocampal neurons (DIV 21) expressing PSD-95–GFP were filled with Lucifer yellow, fixed, and analyzed for number and size of spines. Nearby untransfected cells were chosen randomly and filled as controls. (A) Two filled pyramidal cells from the same coverslip demonstrate the increased spine size and density in the transfected cells. Scale bars, 10 μm (upper), 2 μm (lower). (B) The density of spines was augmented in cells expressing PSD-95–GFP (n = 17 each). (C) Neurons expressing PSD-95–GFP showed an increased density of spines >1 μm in diameter. **P < 0.01, ***P < 0.001.
PSD-95 can drive maturation of synapses, not only of postsynaptic components but also of presynaptic terminals. The selective enhancement of GluR1 versus NR1 clustering correlates with previous anatomical studies showing that the number of NMDARs remains relatively constant, whereas the number of synaptic GluRs increases during development (19, 20) and the magnitude of GluR quantal response increases as synapses mature (21). It is not clear whether the increase in size of synaptic spines and the increase in GluR1 clustering induced by PSD-95 are parallel processes or if one triggers the other (22). Also unclear is the mechanism underlying the enhanced GluR1 clustering, which presumably involves an intermediary protein(s), because PSD-95 does not bind GluR1 (23).
The enhanced size of axon terminals contacting neurons transfected with PSD-95 and the increased frequency of mEPSCs can be explained by the hypothesis that PSD-95 conveys a retrograde signal for presynaptic development. The increased frequency of mEPSCs presumably reflects the increased probability of release associated with an increased vesicle pool size (24). This result may explain why the PSD-95 knockout mouse has augmented paired pulse facilitation (9), which would be consistent with a decreased probability of release in the mutant. The transsynaptic influence of PSD-95 is reminiscent of rapsyn, a nicotinic acetylcholine receptor clustering protein essential for differentiation of motor neuron terminals (25). PSD-95 may communicate with the axon through neuroligin, a PSD-95–associated cell adhesion molecule that links to the nerve terminal by means of neurexins (26,27). Very recent studies show that neuroligin expression in heterologous cells can trigger presynaptic development (28).
In addition to sculpting developing synapses, PSD-95 may contribute to synapse stabilization and remodeling in adult brain (29). Targeted disruption of PSD-95 alters activity-dependent synaptic plasticity and learning (9). Because PSD-95 clustering is controlled by its PDZ domains, palmitoylation, and intramolecular SH3/GK domain interaction (11, 30,31), it will be of interest to determine whether these sites are regulated by neural activity to modulate synaptic structure and plasticity.
