Ubiquitination regulates PSD-95 degradation and AMPA receptor surface expression

Abstract

PSD-95 is a major scaffolding protein of the postsynaptic density, tethering NMDA- and AMPA-type glutamate receptors to signaling proteins and the neuronal cytoskeleton. Here we show that PSD-95 is regulated by the ubiquitin-proteasome pathway. PSD-95 interacts with and is ubiquitinated by the E3 ligase Mdm2. In response to NMDA receptor activation, PSD-95 is ubiquitinated and rapidly removed from synaptic sites by proteasome-dependent degradation. Mutations that block PSD-95 ubiquitination prevent NMDA-induced AMPA receptor endocytosis. Likewise, proteasome inhibitors prevent NMDA-induced AMPA receptor internalization and synaptically induced long-term depression. This is consistent with the notion that PSD-95 levels are an important determinant of AMPA receptor number at the synapse. These data suggest that ubiquitination of PSD-95 through an Mdm2-mediated pathway is critical in regulating AMPA receptor surface expression during synaptic plasticity.

Figure 1. NMDA Treatment Causes Loss of PSD-95 from Synapses and a Reduction in Total Cellular PSD-95

(A) In the absence of NMDA treatment (upper panels), PSD-95 (red) colocalizes with synapsin 1 (green) at synapses. In the composite, colocalized synapsin and PSD-95 puncta appear yellow. Thirty minutes following NMDA treatment (20 μM, 3 min; bottom panels), synaptic staining for PSD-95 was reduced, without affecting synapsin immunoreactivity. Following NMDA stimulation, the overlay image of PSD-95 and synapsin immunoreactivity is dominated by green synapsin staining. (B) Quantification revealed that, following NMDA treatment, the intensity of PSD-95 was reduced to 67% ± 4% of control levels (n = 250 synapses from ten cells in four cultures). (C) NMDA treatment causes a loss of total cellular PSD-95. (Panel 1) In control conditions (DMSO), hippocampal neurons were stimulated with NMDA (20 μM, 3 min); cells were lysed 30 or 60 min later, and total levels of PSD-95 and transferrin receptor were analyzed. NMDA caused a rapid and long-lasting loss of total PSD-95 (top panel); total levels of transferrin receptor were unchanged (TfR; bottom panel). (Panel 2) Reduction in PSD-95 levels did not occur in calcium-free (Ca²⁺-free) media. (Panel 3) Treatment with the PP2B antagonist ascomycin (Asco; 2 μM) prevented loss of PSD-95. (Panel 4) The adenylyl cyclase activator forskolin (FSK; 50 μM) also prevented loss of PSD-95. (D) NMDA stimulation causes a rapid reduction in relative PSD-95 levels (PSD-95/TfR levels) that is maintained for at least 60 min (n = 5). (E) Quantification of several experiments 60 min following NMDA treatment. NMDA-induced reduction in PSD-95 levels was blocked in calcium-free media (Ca²⁺-free; n = 4). PSD-95 loss did not occur in the presence of the PP2B inhibitor ascomycin (Asco; n = 5) or the adenylyl cyclase activator forskolin (FSK; n = 5). For each condition, statistical comparisons were made to the appropriate vehicle control.

Figure 2. NMDA Treatment Causes Ubiquitination of PSD-95 that Depends on a PEST Motif

(A) NMDA stimulates ubiquitination of PSD-95 in cultured hippocampal neurons. Hippocampal neurons were treated with NMDA (20 μM, 3 min) and lysed at various times following treatment. PSD-95 was immunoprecipitated, and samples were immunoblotted with ubiquitin antibodies. NMDA treatment caused rapid ubiquitination of PSD-95 that peaked at 10 min (upper panel, lane 3). Equivalent amounts of PSD-95 were precipitated in each condition (lower panel). (B) NMDA treatment causes a loss of total PSD-95 from neuron extracts by 30 min (upper panel; consistent with Figure 1C). NMDA did not affect levels of the loading control tubulin (lower panel). (C) NMDA treatment does not stimulate ubiquitination of GluR2 AMPA receptors. The same conditions that caused ubiquitination of PSD-95 failed to stimulate ubiquitination of GluR2 (upper panel), although GluR2 was efficiently immunoprecipitated (lower panel). (D) Cartoon of PSD-95 showing its modular domain structure. PSD-95 contains three PDZ domains, an SH3 domain, and a guanylate kinase (GK)-like region. At its N terminus, PSD-95 contains a PEST motif (italicized). PSD-95ΔPEST lacks this sequence but retains the palmitoylation sites (cysteines indicated). (E) Ubiquitination of PSD-95 requires the PEST motif. HEK293 cells were transfected with myc-tagged PSD-95 alone (lane 1), together with HA-tagged ubiquitin (lane 2), or with a mutant form of PSD-95 lacking the PEST sequence motif (PSD-95ΔPEST) and HA-ubiquitin (lane 3). Wild-type PSD-95 was ubiquitinated (top panel; lane 2), whereas ubiquitination of PSD-95ΔPEST was dramatically reduced (top panel, lane 3). Reprobing the blot with antibodies against PSD-95 showed that equivalent amounts of PSD-95 were precipitated in each case (bottom panel).

Figure 3. PSD-95 Associates with Mdm2, a Ubiquitin E3 Ligase

(A) PSD-95 copurifies with Mdm2. Extracts of synaptic plasma membranes were subjected to immunoprecipitation using antibodies against several ubiquitin E3 ligases (lanes 2–6). PSD-95 copurified with Mdm2 (lane 6) but not with control nonimmune IgG, E6AP, Parkin, or Cbl. Extract (lane 1) represents 1% of amount used in immunoprecipitation. (B) Mdm2 copurifies with PSD-95. Synaptic plasma membrane extracts were immunoprecipitated with antibodies against PSD-95. Mdm2 specifically copurified in PSD-95 precipitates (lane 3) but not in control IgG precipitates (lane 2). Extract (lane 1) represents 10% of total used in immunoprecipitation. (C) Mdm2 is localized at synapses. (Upper panel) Cultured hippocampal neurons were labeled with anti-synapsin 1 antibodies (green) to mark synaptic sites and Mdm2 antibodies (red). Overlay demonstrates that Mdm2 immunoreactivity is present at a majority of synapses. (Lower panel) Higher-magnification view (corresponding to region marked by boxes in upper panel) shows that 64% ± 5% (n = 250 synapses from four cultures) of synapsin 1-positive sites are immunoreactive for Mdm2. Examples of synapses positive for both synapsin 1 and Mdm2 are indicated with arrows.

Figure 4. Mdm2 Ubiquitinates PSD-95 and Regulates Its Expression Levels

(A) Mdm2 directly ubiquitinates PSD-95. GST-PSD-95 was incubated with ubiquitin (Ub), E1 (UBA1), E2 (UBC5), and Mdm2 (E3; lanes 3–7). Ubiquitination was detected using PSD-95 antibodies. GST-PSD-95 showed a laddering of higher molecular weight ubiquitinated species when the full complement of the ubiquitination machinery was present (upper panel; lane 4). Equal amounts of GST-PSD-95 were used in each reaction (lower panel). (B) p14ARF stabilizes cellular levels of PSD-95. HEK293 cells were transfected with either myc-tagged PSD-95 (lanes 1 and 2) or PSD-95ΔPEST (lanes 3 and 4) and GFP, in the absence or presence of myc-tagged p14ARF, a cellular inhibitor of Mdm2 activity. In the presence of p14ARF, levels of PSD-95 protein in total cellular extracts were increased (upper panel; compare lanes 1 and 2). p14ARF did not affect levels of PSD-95ΔPEST (lane 4) nor of cotransfected GFP (bottom panel). (C) Quantification of p14ARF effect. p14ARF increases the levels of wild-type PSD-95 to 176% ± 23% of control levels (p < 0.05, n = 3) but does not alter levels of PSD-95ΔPEST (115% ± 17% of controls, p > 0.2, n = 3). (D) Mdm2 reduces total PSD-95 levels. HEK293 or H1299 cells were transfected with myc-tagged PSD-95 (lanes 1 and 2) or PSD-95ΔPEST (lanes 3 and 4) in the absence and presence of Mdm2. In the presence of Mdm2, levels of PSD-95 protein in total cell extracts were reduced (upper panel, lane 2). Expression of Mdm2 had no effect on levels of PSD-95ΔPEST (upper panel, lane 4). Equivalent amounts of extracts were loaded in each condition (tubulin; bottom panel). (E) Quantification of Mdm2 effect. Expression of Mdm2 decreases levels of PSD-95 to 56% ± 9% of control levels (p < 0.05, n = 3). (F) In the absence of Mdm2, PSD-95 is not ubiquitinated. Mouse embryonic fibroblasts lacking Mdm2 (Mdm2^‒/‒), were transfected with myc-tagged PSD-95 (lanes 1 and 2) or PSD-95ΔPEST (lanes 3 and 4) and HA-ubiquitin, with or without Mdm2. Cells were lysed, and PSD-95 was immunoprecipitated. Only when Mdm2 was reintroduced into the cells was PSD-95 ubiquitinated (upper panel, lane 2). Mdm2-mediated ubiquitination of PSD-95ΔPEST was greatly reduced (upper panel, lane 4). Equivalent amounts of PSD-95 were immunoprecipitated in all conditions (bottom panel).

Figure 5. The Proteasome Regulates PSD-95 Degradation and Surface AMPA Receptor Expression

(A) Under control conditions, NMDA (20 μM, 3 min) causes a loss of total PSD-95 protein in hippocampal neurons (+). (B) MG132 prevents NMDA-induced degradation of PSD-95. Cultures were treated with MG132 (50 μM, 1 hr) prior to NMDA stimulation. Inhibiting the proteasome with MG132 blocks this NMDA-induced degradation of PSD-95 (+). The proteasomal inhibitor lactacystin also blocked PSD-95 degradation (data not shown). (C) Quantification of PSD-95 levels. NMDA causes degradation of PSD-95 (60% ± 15% of controls; n = 5). MG132 prevents this effect (110% ± 14% of MG132 alone; n = 5). (D) NMDA causes loss of surface AMPA receptors. Cultures were treated with DMSO (1 hr) and then stimulated with NMDA (20 μM, 3 min). One hour following NMDA treatment, the cultures were biotinylated, and surface proteins were precipitated with neutravidin sepharose. NMDA treatment caused a loss of surface GluR1 (+). (E) MG132 prevents NMDA-induced loss of surface AMPA receptors. Cultures were incubated in MG132 (50 μM, 1 hr) prior to NMDA (20 μM, 3 min) treatment, and surface receptors were isolated 1 hr later. NMDA-induced loss of surface GluR1 was blocked by MG132 (+; right lane). (F) Quantification reveals that NMDA reduces surface GluR1 by 30% within 15 min and that this effect lasts at least 1 hr. There was no change in levels of transferrin receptor (TfR) over this time. (G) In MG132, NMDA-induced changes in surface GluR1 are blocked. There was no change in levels of transferrin receptor (TfR). (H–K) Hippocampal cultures were pretreated with MG132 (50 μM, 30 min) prior to NMDA stimulation (20 μM, 3 min). Fifteen minutes later, surface GluR2 receptors were labeled. NMDA caused a reduction in surface GluR2 staining. MG132 blocked this effect. (L) Following NMDA treatment, the intensity of GluR2 puncta was 55% ± 11% of controls (n = 10 cells from four cultures). MG132 largely blocked the NMDA-induced loss of surface GluR2 (82% ± 16% of controls; n = 10 cells from four cultures).

Figure 6. Ubiquitination of PSD-95 Regulates Endocytosis of AMPA Receptors

(A) (Top row) An antibody feeding technique was used to visualize endocytosed GluR2 in untransfected neurons (blue; left panels), neurons transfected with myc-PSD-95ΔPEST (blue; middle panels), or myc-tagged PSD-95 (blue; right panels). (Middle row) In untransfected neurons, NMDA treatment causes a loss of surface GluR2 staining (red; +; left panels). NMDA-induced loss of surface GluR2 is largely blocked in PSD-95ΔPEST-expressing neurons (red; middle panels) and blunted in neurons expressing wild-type PSD-95 (red; right panels). (Bottom row) In untransfected neurons (−; left panels), there is scant intracellular GluR2 staining (green). NMDA treatment (20 μM, 3 min) causes internalization of GluR2 (green; +; left panels). NMDA-induced GluR2 endocytosis was blocked in neurons expressing PSD-95ΔPEST (green; middle panels). NMDA-induced endocytosis of GluR2 was partially inhibited in neurons expressing wild-type PSD-95 (green; right panels). Background levels of myc antigen are faintly detected in the control neurons. (B) Higher-magnification view of representative dendritic regions shows that in untransfected neurons (left column), NMDA treatment causes a loss of surface GluR2 (red) and a concomitant increase in intracellular GluR2 (green). In contrast, in neurons expressing PSD-95ΔPEST (middle column), the NMDA-induced loss of surface GluR2 (red) and increase in intracellular GluR2 (green) is largely attenuated. Both NMDA effects are blunted in neurons expressing wild-type PSD-95 (right columns). (C) Quantification of the density of surface GluR2 puncta revealed untransfected, untreated control neurons had 2.6 ± 0.6 GluR2 puncta/10 μm (n = 16). Following NMDA treatment, untransfected neurons exhibited 1.9 ± 0.3 puncta per 10 μm (p < 0.02 versus control, n = 16). In neurons expressing PSD-95ΔPEST, GluR2 puncta density was 3.6 ± 0.4/10 μm (n = 15). NMDA treatment did not cause a significant decrease in density of surface GluR2 puncta (3.25 ± 0.2/10 μm, p > 0.09, n = 15). Neurons expressing wild-type PSD-95 had 3.5 ± 0.26 puncta/10 μm (n = 16). NMDA treatment reduced GluR2 puncta density to 2.6 ± 0.3/10 μm(p < 0.05, n = 16). Quanitatnion of data were performed using a double blind protocol. (D) In untransfected neurons, NMDA treatment increased the density of internalized GluR2 puncta from 0.44 ± 0.1 puncta/10 μm to 2.1 ± 0.15 puncta/10 μm within 30 min (n = 16, p < 0.01). This NMDA-induced increase in internal GluR2 was blocked by transfection with PSD-95ΔPEST (control: internal GluR2 puncta was 0.33 ± 0.2 puncta/10 μm; n = 15; NMDA internal GluR2 density was 0.68 ± 0.17/10 μm; n = 15, p > 0.15). In neurons expressing wild-type PSD-95, NMDA had a reduced effect compared to controls (0.36 ± 0.37 puncta/10 μm in untreated cells versus 1.44 ± 0.16 puncta/10 μm in NMDA treated cells; n = 16, p < 0.05). Quantitation of data were performed using a double blind protocol.

Figure 7. Inhibition of the Proteasome Reduces the Magnitude of Hippocampal LTD

(A and C) Sample EPSCs from a hippocampal slice illustrate the effect of MG132 (A) and lactacystin (C) on pairing-induced synaptic depression. CA1 pyramidal neurons were voltage-clamped at −65 mV. Following the pairing protocol, EPSCs were reduced by ∼50% in DMSO vehicle control recordings. Inhibition of proteasome activity by MG132 (1 μM) and lactacystin (10 μM) significantly attenuated LTD. Sweeps are the average of consecutive responses taken 5 min prior to pairing onset (1) and 25–30 min following completion of the pairing protocol (2). All recordings were performed in the presence of the GABA_A antagonist picrotoxin (50 μM). Scale bars, 50 pA, 30 ms (A); 25pA, 25 ms (C). (B and D) Overlay of grouped data time courses demonstrates that MG132 (B) and lactacystin (D) attenuate LTD magnitude. Following 15 min of stable baseline recording, LTD was induced at time zero by a pairing protocol (black bar) and assessed 25–30 min after pairing. Proteasome inhibition by MG132 (●; n = 6) or lactacystin (●; n = 7) significantly reduced the magnitude of LTD compared to DMSO control (○; n = 6, 3 for panels [B] and [D], respectively). Plotted on the leftward y axis is EPSC amplitude normalized to a 15 min baseline period. Each symbol represents data averaged into 2 min bins.

Figure 8. A Model for Ubiquitin-Mediated Regulation of Surface AMPA Receptors

(A) PSD-95 indirectly interacts with AMPA receptors through stargazin. In addition, PSD-95 also binds to Mdm2, a ubiquitin E3 ligase. (B) In response to NMDA-induced calcium influx, PP2B is activated, and PSD-95 is rapidly ubiquitinated by associated Mdm2. PP2B and PKA may be targeted to the synapse through association with AKAP79/150. (C) Following ubiquitination, PSD-95 is targeted to the proteasome where it is degraded (1). The loss of PSD-95 untethers AMPA receptors from the PSD, allowing endocytosis (2).