Specific Coupling of NMDA Receptor Activation to Nitric Oxide Neurotoxicity by PSD-95 Protein

See allHide authors and affiliations

Science  11 Jun 1999:
Vol. 284, Issue 5421, pp. 1845-1848
DOI: 10.1126/science.284.5421.1845


The efficiency with whichN-methyl-d-aspartate receptors (NMDARs) trigger intracellular signaling pathways governs neuronal plasticity, development, senescence, and disease. In cultured cortical neurons, suppressing the expression of the NMDAR scaffolding protein PSD-95 (postsynaptic density–95) selectively attenuated excitotoxicity triggered via NMDARs, but not by other glutamate or calcium ion (Ca2+) channels. NMDAR function was unaffected, because receptor expression, NMDA currents, and 45Ca2+loading were unchanged. Suppressing PSD-95 blocked Ca2+-activated nitric oxide production by NMDARs selectively, without affecting neuronal nitric oxide synthase expression or function. Thus, PSD-95 is required for efficient coupling of NMDAR activity to nitric oxide toxicity, and imparts specificity to excitotoxic Ca2+ signaling.

Calcium influx through NMDARs plays key roles in synaptic transmission, neuronal development, and plasticity (1). Excessive Ca influx triggers excitotoxicity (2), damaging neurons in diverse neurological disorders (3). Rapid Ca2+-dependent neurotoxicity is triggered most efficiently when Ca2+ influx occurs through NMDARs, and cannot be reproduced by loading neurons with equivalent quantities of Ca2+ through non-NMDARs or voltage-sensitive Ca2+ channels (VSCCs) (4). This suggests that Ca2+ influx through NMDAR channels is functionally coupled to neurotoxic signaling pathways.

We hypothesized that lethal Ca2+ signaling by NMDARs is determined by the molecules with which they interact. The NR2 NMDAR subunits bind to PSD-95/SAP90 (5), chapsyn-110/PSD-93, and other members of the membrane-associated guanylate kinase (MAGUK) family (6). NMDAR-bound MAGUKs are generally distinct from those associated with non-NMDARs (7). This raises the possibility that the preferential activation of neurotoxic Ca2+ signals by NMDARs is determined by the distinctiveness of NMDAR-bound MAGUKs, or of the intracellular proteins that they bind. We thus focused on PSD-95, an abundant scaffolding molecule that binds and clusters NMDARs preferentially and may link them to intracellular signaling molecules (8).

With the use of a 15-nucleotide oligomer phosphodiester antisense oligodeoxynucleotide (ODN), PSD-95 expression was suppressed in cultured cortical neurons to <10% of expression in controls (Fig. 1, A and B) (9, 10). Sham washes had no effect, nor did sense and missense ODNs (9). The ODNs had no effect on neuronal survivability and morphology, as gauged by viability assays (below) and phase-contrast microscopy (11).

Figure 1

Increased resilience of PSD-95–deficient neurons to NMDA toxicity despite Ca2+ loading. (A) Immunoblot showing representative effects of PSD-95 antisense (AS), sense (SE), and missense (MS) ODNs, as well as sham (SH) washes, on PSD-95 expression. PC, positive control tissue from purified rat brain cell membranes. Asterisk: nonspecific band produced by the secondary antibody, useful to control for protein loading and blot exposure times. (B) Densitometric analysis of PSD-95 expression pooled from N experiments. Asterisk: different from other groups, ANOVA, F = 102, P < 0.0001. ODNs were used at 5 μM except where indicated (AS, 2 μM). (C) Representative phase contrast and propidium iodide fluorescence images of PSD-95–deficient (AS) and control (SE) cultures 24 hours after a 60-min challenge with 30 μM NMDA. Scale bar, 100 μm. (D) Decreased NMDA toxicity at 24 hours in PSD-95–deficient neurons [with CNQX and nimodipine (NIM) added as described (12)] after selective NMDAR activation for 60 min (16 cultures per bar, pooled from four independent experiments). Asterisk: differences from SE, MS, and SH (Bonferronit test, P < 0.005). (E) No effect of suppressing PSD-95 on NMDAR-mediated Ca2+loading (12 cultures per bar from three independent experiments).

To examine the impact of PSD-95 on NMDAR-triggered excitotoxicity, we exposed ODN-treated cultures to NMDA (10 to 100 μM) for 60 min; after washing, these cultures were either used for45Ca2+ accumulation measurements or observed for a further 23 hours (12). Ca2+influx was isolated to NMDARs by adding antagonists of non-NMDARs and Ca2+ channels (4, 12). NMDA toxicity was significantly reduced in neurons deficient in PSD-95 across a range of insult severities (Fig. 1, C and D) [median effective concentration (EC50): antisense, 43.2 ± 4.3 μM; sense, 26.3 ± 3.4 μM; Bonferroni t test, P < 0.005] (13). However, PSD-95 deficiency had no effect on Ca2+ loading into identically treated sister cultures (Fig. 1E). Therefore, PSD-95 deficiency induces resilience to NMDA toxicity despite maintained Ca2+ loading.

We next examined whether the increased resilience to Ca2+loading in PSD-95–deficient neurons was specific to NMDARs. Non-NMDAR toxicity was produced using kainic acid (30 to 300 μM) (14) in the presence of NMDAR and Ca2+ channel antagonists (4, 12). Kainate toxicity was unaffected in PSD-95–deficient neurons challenged for either 60 min (11) or 24 hours (Fig. 2A). Non-NMDAR toxicity occurred without significant 45Ca2+ loading (Fig. 2B), as >92% of neurons in these cultures express Ca2+-impermeable AMPA receptors (4). However, Ca2+ loading through VSCCs, which is nontoxic (4) (Fig. 2C), was also unaffected by PSD-95 deficiency (Fig. 2D). Thus, suppressing PSD-95 expression affects neither the toxicity nor Ca2+ fluxes triggered through pathways other than NMDARs.

Figure 2

PSD-95 deficiency does not affect toxicity and Ca2+ loading produced by activating non-NMDARs and Ca2+ channels. Cultures were treated with sham washes or antisense or sense ODNs as in Fig. 1. (Aand B) Selective activation of AMPA/kainate receptors with kainate plus MK-801 and nimodipine produces toxicity over 24 hours (A) irrespective of PSD-95 deficiency, with minimal45Ca2+ loading (B). (C andD) Selective activation of VSCCs (12) produces little toxicity (C), but significant 45Ca2+ loading (D) that is also insensitive to PSD-95 deficiency. Four cultures are represented per bar in all experiments.

Immunoblot analysis (10) of PSD-95–deficient cultures revealed no alterations in the expression of the essential NMDAR subunit NR1, nor of two other NMDAR-associated MAGUKs, PSD-93 and SAP-102 (Fig. 3A). This indicated that altered expression of NMDARs and their associated proteins was unlikely to explain the reduced toxicity of NMDA in PSD-95 deficiency. Therefore, we examined whether PSD-95 modulates NMDAR function (15). NMDA currents were recorded using the whole-cell patch technique (16) (Fig. 3B). PSD-95 deficiency had no effect on passive membrane properties, including input resistance and membrane capacitance {capacitance: antisense, 55.0 ± 2.6 pF; sense, 52.7 ± 3.2 pF; sham, 48.1 ± 3.4 pF [17 to 19 neurons per group; analysis of variance (ANOVA),F = 1.29, P = 0.28]}. Whole-cell currents elicited with 3 to 300 μM NMDA were also unaffected. Peak currents were as follows: antisense, 2340 ± 255 pA; sense, 2630 ± 276 pA; sham, 2370 ± 223 pA (Fig. 3B, right; 17 to 19 neurons per group; ANOVA, F = 0.43,P = 0.65). NMDA dose-response relationships also were unchanged [Fig. 3B, left; EC50 antisense, 16.1 ± 0.8 μM, sense, 15.5 ± 2.1 μM; sham, 15.9 ± 2.9 μM (six to nine neurons per group; one-way ANOVA, F = 0.02,P = 0.98)], as were NMDA current density and desensitization (Fig. 3C).

Figure 3

Perturbing PSD-95 has no effect on NMDA receptor function. (A) Immunoblots of PSD-95 ODN-treated cultures probed for PSD-95, NR1, PSD-93, and SAP-102. PC, positive control from purified rat brain cell membranes; other abbreviations as in Figs. 1 and 2. (B) NMDA dose-response curves (left) and representative NMDA currents (right) obtained with 3 to 300 μM NMDA. (C) Measurements of NMDA (300 μM) current density (17 to 19 neurons per group; ANOVA, F = 1.10,P = 0.34) and of NMDA current desensitization.I ss, steady-state current;I peak, peak current (15 to 16 neurons per group; ANOVA, F = 0.14, P = 0.87). Time constants for current decay: antisense, 1310 ± 158 ms; sense, 1530 ± 185 ms; sham, 1190 ± 124 ms (ANOVA,F = 1.22, P = 0.31). (D) Currents elicited with 300 μM NMDA in neurons dialyzed with Lucifer Yellow (inset) and 1 mM tSXV or control peptide.

To further examine the effect of PSD-95 binding on NMDAR function, we injected a nine–amino acid peptide (Lys-Leu-Ser-Ser-Ile-Glu-Ser-Asp-Val) corresponding to the COOH-terminal domain of the NR2B subunit characterized by the tSXV motif (6) into the neurons. At 0.5 mM, this peptide competitively inhibited the binding of PSD-95 to GST-NR2B fusion proteins (6) and was therefore predicted to uncouple NMDARs from PSD-95. Intracellular dialysis of 1 mM tSXV or control peptide (Cys-Ser-Lys-Asp-Thr-Met-Glu-Lys-Ser-Glu-Ser-Leu) (6) was achieved through patch pipettes (3 to 5 megohms) also containing the fluorescent tracer Lucifer Yellow. This had no effect on NMDA currents over 30 min, despite extensive dialysis of Lucifer Yellow into the cell soma and dendrites (Fig. 3D). Peak current amplitudes were as follows: tSXV, 2660 ± 257 pA (n = 9); control, 2540 ± 281 pA (n = 10; t (17) = 0.31,P = 0.76).

Consistent with data obtained from recently generated mutant mice expressing a truncated 40K PSD-95 protein (17), we found no effects of PSD-95 deficiency on NMDAR expression, on other NMDAR-associated MAGUKs, or on NMDA-evoked currents. In addition, NMDAR function gauged by measuring NMDA-evoked 45Ca2+accumulation was unaffected. Thus, the neuroprotective consequences of PSD-95 deficiency must be due to events downstream from NMDAR activation.

The second PDZ domain of PSD-95 binds to the COOH-terminus of NR2 subunits and to other intracellular proteins (8). Among these is neuronal nitric oxide synthase (nNOS) (18), an enzyme that catalyzes nitric oxide (NO) production (19). Although never demonstrated experimentally, the NMDAR–PSD-95–nNOS complex was postulated to account for the preferential production of NO by NMDARs over other pathways (8). To determine whether NO signaling plays a role in NMDA toxicity in the present cultures, we treated the cells with N-nitro-l-arginine methyl ester (l-NAME), a NOS inhibitor (19).Because l-NAME protected the neurons against NMDA toxicity (Fig. 4A), it is possible that suppression of PSD-95 might perturb this toxic signaling pathway.

Figure 4

Coupling of NMDAR activation to nitric oxide signaling by PSD-95. (A) l-NAME protects against NMDA toxicity (eight cultures per bar). Asterisk: difference from 0 μM l-NAME (Bonferronit test, P < 0.05). (B) No effect of sham washes and of PSD-95 antisense and missense ODNs on nNOS expression (immunoblot) and of antisense and sense ODNs on NADPH diaphorase staining. (C) Effect of selective NMDAR activation (12) on cGMP formation (12 cultures per bar). (D and E) Effects of VSCC activation (eight cultures per bar) and selective AMPA/kainate receptor activation (12) (four cultures per bar, one experiment) on cGMP formation. Data in (C) to (E) are expressed as the fraction of cGMP produced in sense-treated cultures by 100 μM NMDA. Asterisk: differences from both sham and sense controls (Bonferroni t test, P< 0.0001). (F) Sodium nitroprusside toxicity is similar in PSD-95 antisense-, sense-, and sham-treated cultures. Data in all bars in (A), (C), (D), and (F) were pooled from two or three independent experiments.

The effect of suppressing PSD-95 expression on NO signaling and toxicity was examined using guanosine 3′,5′-monophosphate (cGMP) formation as a surrogate measure of NO production by Ca2+-activated nNOS (20, 21). PSD-95 deficiency had no impact on nNOS expression (Fig. 4B), nor on the morphology (Fig. 4B) or counts of the reduced form of nicotinamide adenine dinucleotide phosphate (NADPH) diaphorase–staining (19) neurons (sham, 361 ± 60; sense, 354 ± 54; antisense, 332 ± 42 staining neurons per 10-mm cover slip, three cover slips per group). However, in neurons lacking PSD-95 challenged with NMDA under conditions that isolated Ca2+influx to NMDARs (4), cGMP production was markedly attenuated (>60%; ANOVA, P < 0.0001) (Fig. 4C). Like inhibited toxicity, inhibited cGMP formation in neurons lacking PSD-95 was only observed in response to NMDA. It was unaffected in neurons loaded with Ca2+ through VSCCs (Fig. 4D), even under high neuronal Ca2+ loads matching those attained by activating NMDARs (compare Figs. 1E and 2D) (4). Therefore, nNOS function was unaffected by PSD-95 deficiency. AMPA/kainate receptor activation failed to load the cells with Ca2+(Fig. 2B) and thus failed to increase cGMP concentrations (Fig. 4E). Suppressing PSD-95 thus selectively reduces NO production efficiency by NMDAR-mediated Ca2+ influx, but preserves NO production by Ca2+ influx through other pathways.

Bypassing nNOS activation with NO donors restored toxicity in neurons lacking PSD-95. The NO donors sodium nitroprusside (19) (Fig. 4F; EC50 = 300 μM) andS-nitrosocysteine (11, 22) were highly toxic, irrespective of PSD-95 deficiency. Thus, reduced NMDA toxicity in PSD-95–deficient cells was unlikely to be caused by altered signaling events downstream from NO formation.

Suppressing PSD-95 expression uncoupled NO formation from NMDAR activation (Fig. 4C), and also protected neurons against NMDAR toxicity (Fig. 1, C and D) without affecting receptor function (Fig. 1E and Fig. 3, A to D), by mechanisms downstream from NMDAR activation and upstream from NO-mediated toxic events (Fig. 4F). Therefore, PSD-95 imparts signaling and neurotoxic specificity to NMDARs through the coupling of receptor activity to critical second messenger pathways. Our results have broader consequences, in that NMDAR activation and NO signaling are also critical to neuronal plasticity, learning, memory, and behavior (1, 20, 23). Thus, our report provides evidence for a potential mechanism by which PSD-95 may govern important physiological and pathological aspects of neuronal functioning.

  • * To whom correspondence should be addressed. E-mail: mike_t{at}


Stay Connected to Science

Navigate This Article