AMPA receptor subunit GluR1 (GluA1) serine-845 site is involved in synaptic depression but not in spine shrinkage associated with chemical long-term depression

Abstract

The structure of dendritic spines is highly plastic and can be modified by neuronal activity. In addition, there is evidence that spine head size correlates with the synaptic α-amino-3-hydroxy-5-methylisoxazole propionic acid (AMPA) receptor (AMPAR) content, which suggests that they may be coregulated. Although there is evidence that there are overlapping mechanisms for structural and functional plasticity, the extent of the overlap needs further investigation. Specifically, it is unknown whether AMPAR levels determine spine size or whether both are regulated via parallel pathways. We studied the correlation between spine structural plasticity and long-term synaptic plasticity following chemical-induced long-term depression (chemLTD). In particular, we examined whether the regulation of AMPARs, which is implicated in LTD, is critical for spine morphological plasticity. We used mutant mice specifically lacking the serine-845 site on the type 1 glutamate receptor (GluR1, or GluA1) subunit of AMPARs (mutants). These mice specifically lack N-methyl-D-aspartate (NMDA) receptor (NMDAR)-dependent LTD and NMDAR activation-induced AMPAR endocytosis. We found that chemLTD causes a rapid and persistent shrinkage in spine head volume of hippocampal CA1 pyramidal neurons in wild types similar to that reported in other studies using low-frequency stimulation (LFS)-induced LTD. Surprisingly, we found that although S845A mutant mice display impaired chemLTD, the shrinkage of spine head volume occurred to a similar magnitude to that observed in wild types. Our results suggest that there is dissociation in the molecular mechanisms underlying functional LTD and spine shrinkage and that GluR1-S845 regulation is not necessary for spine morphological plasticity.

Fig. 1.

Chemical long-term depression (chemLTD) induction is associated with a rapid and persistent shrinkage of spine head volume. A: chemLTD induced by 3-min infusion of 20 μM N-methyl-d-aspartate (NMDA) caused a long-lasting synaptic depression as measured by extracellular field potential (FP) recordings. P < 0.0005, paired t-test. B: 2 sets of representative spine images taken from control (CTL) and chemLTD-induced slices. ChemLTD-induced slices were fixed at specified times as noted. Each projected 3-dimensional image was obtained by reconstruction of 15–30 sections taken at 0.27-μm z-axis intervals. C₁: average spine head volume was immediately reduced after chemLTD induction and persisted up to 1 h. Normalized spine head volume (normalized to the average of control spines from each mouse) shows about 30% reduction after chemLTD induction [CTL: 99.18 ± 5.01% (0.18 ± 0.01 μm³), 578 spines from 15 dendritic segments; 0 min post-NMDA: 73.46 ± 5.30% (0.14 ± 0.01 μm³), 568 spines from 12 dendrites; 30 min post-NMDA: 67.69 ± 4.06% (0.12 ± 0.01 μm³), 636 spines from 18 dendrites; 60 min post-NMDA: 69.32 ± 3.60% (0.13 ± 0.01 μm³), 589 spines from 12 dendrites; experiments repeated on hippocampal slices from 3 mice]. ANOVA: F(3,59) = 11.115, P < 0.0001. *P < 0.001. C₂: cumulative probability curves of spine head volume for the chemLTD-induced groups (shaded lines) are all shifted to smaller values compared with the CTL curve (solid line). P < 0.0001, Kolmogorov-Smirnov test. D: density of spines was not significantly altered up to 1 h post-chemLTD induction (spine number per 10-μm segment of dendrite: CTL: 14.22 ± 0.62; 0 min post-NMDA: 15.22 ± 1.06; 30 min post-NMDA: 13.55 ± 0.59; 60 min post-NMDA: 13.44 ± 0.68; n = same as in C₁).

Fig. 2.

S845A mutants display higher spine density but similar spine head volume compared with wild types (WT). A: generation of S845A-2J mice. Representative immunoblots were probed with phosphorylation site-specific antibody against the serine-845 site on type 1 glutamate receptor (GluR1-S845; top blot), GluR1-specific antibody (middle blot), and antibody against green fluorescent protein (GFP), which is known to cross-react with yellow fluorescent protein (YFP; bottom blot). S845A and YFP-2J mice were crossed to produce F1s (S845A+/−;YFP+/−). Hippocampal samples were obtained from the F2 offspring resulting from the F1 crosses. As expected from a Mendelian inheritance, we obtained 4 types of F2s: WT (carrying WT GluR1 but not the YFP transgene), WT-2J (carrying WT GluR1 and the YFP transgene), S845A-2J (carrying GluR1-S845A and the YFP transgene), and S845A (carrying GluR1-S845A but not YFP). B₁: 2 sets of representative spine images from WT-2J and S845A-2J mice. B₂: S845A-2J mice had higher spine density than WT-2J mice (spine number per 10-μm dendritic segment: WT-2J: 14.22 ± 0.62, 578 spines from 15 dendritic segments, 3 mice; S845A-2J: 17.57 ± 0.94, 1,038 spines from 18 dendritic segments, 3 mice). *P < 0.007, Student's t-test. B₃: neither the average spine head volume (left; WT-2J: 0.18 ± 0.01 μm³; S845A-2J: 0.19 ± 0.01 μm³) nor the distribution of spine head volume (right) was different between WT-2J and S845A-2J mice [small (<0.3 μm³): WT-2J: 66.27 ± 3.05%; S845A-2J: 67.65 ± 2.35%; medium (≥0.3 and <0.6 μm³): WT-2J: 32.40 ± 3.00%; S845A-2J, 28.31 ± 2.02%; large (≥0.6 μm³): WT-2J: 1.23 ± 0.49%; S845A-2J: 3.96 ± 0.97%]. Two-way ANOVA, P = 0.4.

Fig. 3.

S845A mutants have larger α-amino-3-hydroxy-5-methylisoxazole propionic acid (AMPA) receptor (AMPAR)-mediated miniature excitatory postsynaptic currents (mEPSCs). A₁: cumulative probability curve of mEPSC amplitude of S845A homozygous (HM) mice was shifted to the right compared to WT. P < 0.0001, Kolmogorov-Smirnov test. Inset: average AMPAR mEPSC amplitude was significantly larger in S845A HM. *P < 0.01, Student's t-test. A₂: average frequency of AMPAR mEPSCs was normal in S845A HM. A₃, top: average mEPSC traces from WT and HM. Scale bars, 10 ms and 3 pA. Bottom, representative recordings (5-s traces) from WT and S845A HM. B: presynaptic release probability was normal in S845A HM as measured by paired-pulse facilitation (PPF) ratio (slope of second FP/slope of first FP). Top: representative traces. Paired-pulse stimulation was given at 50-ms interstimulus interval (ISI). Scale bars, 1 mV and 10 ms. C: AMPAR current-to-NMDA receptor current ratios (I_AMPARI_NMDAR) were not different between WT and S845A HM. Top: superimposed EPSC traces recorded at −60 mV (solid trace, I_AMPAR) and at +40 mV in the absence (solid trace, compound EPSC of I_AMPAR and I_NMDAR) or presence of 100 μM d,l-2-amino-5-phosphonovalerate (APV; shaded trace, I_AMPAR only). Scale bars, 20 pA and 20 ms. D: S845A mutants displayed an increase in the level of NMDAR subunit NR1 at the postsynaptic densities (left; P = 0.05, Student's t-test) but not in the total homogenates (right).

Fig. 4.

ChemLTD induction causes spine morphological changes in S845A-2J mice. A: NMDA (20 μM, 3 min) application produced synaptic depression in S845A-2J mice (filled circles; P < 0.02 with 2-tailed paired t-test of averages of 10-min baseline before chemLTD and the last 10 min of recording), which was significantly less (P < 0.05, t-test) than that observed in WT (open circles: data duplicated from Fig. 1A for comparison). B: 2 sets of representative spine images taken from CTL and chemLTD slices of S845A-2J. Slices were fixed at specified times after NMDA application. C₁: chemLTD induction triggered a rapid spine shrinkage that lasted up to 1 h [CTL: 100 ± 4.88% (0.19 ± 0.01 μm³), 1,038 spines from 18 dendritic segments; 0 min post-NMDA: 68.70 ± 4.02% (0.13 ± 0.01 μm³), 849 spines from 18 dendrites; 30 min post-NMDA: 72.85 ± 4.80% (0.14 ± 0.01 μm³), 642 spines from 17 dendrites; 60 min post-NMDA: 78.13 ± 4.94% (0.14 ± 0.01 μm³), 749 spines from 19 dendrite; data from 3 mice]. ANOVA: F(3,68) = 8.632, *P < 0.0001. C₂: cumulative probability curves of spine head volume in chemLTD groups (shaded lines) were all shifted to the left of CTL (solid line). D: spine density in S845A-2J mice was significantly reduced 1 h after chemLTD induction [spine number per 10-μm segment of dendrite: CTL: 100 ± 8.17% (17.57 ± 0.94); 0 min post-NMDA: 101.56 ± 8.85% (17.88 ± 1.00); 30 min post-NMDA: 93.36 ± 13.12% (15.79 ± 0.94); 60 min post-NMDA: 79.52 ± 6.80% (14.22 ± 0.88)]. ANOVA: F(3, 68) = 3.354, P = 0.024. *P < 0.02, Fisher's protected least significant difference post hoc test.

Fig. 5.

Rapid shrinkage of individual spines following chemLTD induction observed by 2-photon time-lapse imaging of live slices from both WT-2J and S845A-2J mice. A₁: representative spine images from WT-2J mice before and after NMDA (20 μM, 3 min) application at indicated time points. Arrows point to the spines that were stable for at least 15 min before chemLTD induction and used for data analysis. A₂: chemLTD induction led to about 40% reduction in average spine head volume in WT-2J mice. B₁: representative spine images from S845A-2J mice. Arrows point to spines that were stable for at least 15 min during baseline observation and used for data analysis. B₂: chemLTD induction led to about 40% reduction in average spine head size in S845A-2J mice.

Fig. 6.

Spine morphological changes after chemLTD induction depend on NMDAR activation and were verified using immunohistochemical staining of YFP. A₁: representative spine images from WT-2J slices treated with 100 μM APV before and after NMDA (20 μM, 3 min) application at indicated time points. A₂: in the presence of 100 μM APV, spine head volume did not change with chemLTD induction. B₁: representative confocal images of WT-2J dendritic spines in CTL slices and chemLTD-induced slices (60 min). Top: green channel shows YFP signal. Middle: red channel (magenta) shows immunohistochemical labeling of YFP using an anti-GFP antibody (Alexa633-linked 2nd antibody). Bottom: overlay between the 2 channels (overlap of green and magenta is shown as white). B₂: a similar level of spine shrinkage was observed 60 min after chemLTD induction in both green (YFP: CTL: 0.16 ± 0.01 μm³; 60 min post-NMDA: 0.13 ± 0.01 μm³; P < 0.05, Student's t-test) and red channels (anti-GFP labeling: CTL: 0.18 ± 0.02 μm³; 60 min post-NMDA: 0.13 ± 0.01 μm³; P < 0.02, Student's t-test). CTL: 225 spines from 16 dendritic segments; 60 min post-NMDA: 175 spines from 16 dendritic segments.

Fig. 7.

Proposed signaling cascades for synaptic plasticity and spine structural plasticity after chemLTD induction. ChemLTD induction requires the activation of NMDARs, which recruit protein phosphatases (PPases, such as PP2B) and PLC. These 2 signaling pathways lead to long-term synaptic depression and spine head shrinkage. Dephosphorylation of GluR1-S845 is involved in synaptic depression but not in structural changes of dendritic spines. The residual chemLTD in S845A mutants may be PLC mediated and/or GluR2 dependent.