AMPAR removal underlies Abeta-induced synaptic depression and dendritic spine loss

Abstract

Beta amyloid (Abeta), a peptide generated from the amyloid precursor protein (APP) by neurons, is widely believed to underlie the pathophysiology of Alzheimer's disease. Recent studies indicate that this peptide can drive loss of surface AMPA and NMDA type glutamate receptors. We now show that Abeta employs signaling pathways of long-term depression (LTD) to drive endocytosis of synaptic AMPA receptors. Synaptic removal of AMPA receptors is necessary and sufficient to produce loss of dendritic spines and synaptic NMDA responses. Our studies indicate the central role played by AMPA receptor trafficking in Abeta-induced modification of synaptic structure and function.

Figure 1. Aβ reduces spine density

(A) CA1 pyramidal cells co-transfected with EGFP and either APP or APP(MV) as a control were imaged with 2 photon laser scanning microscopy (2PLSM) Scale: 10μm. (B) CA1 pyramidal cells transfected with EGFP and incubated for 7 days in either vehicle (0.2% DMSO,0.5mM Tris), Aβ(1–42)(1μM) or reverse Aβ(42-1)(1μM). (C) Over-expression of Aβ decreases spine density in comparison to EGFP expression alone or with APP(MV) (Spine Density (sp/μm), p values are in comparison to EGFP: EGFP: 0.66±0.03, 53 dendrites, 5 cells; APP: 0.42±0.02, 106 dendrites, 10 cells, p<0.01; APP(MV): 0.60±0.02, 114 dendrites, 10 cells, p=0.15; APP vs. APP(MV), p<0.01). (D) Aβ(1–42) incubation for 7 days decreases spine density (Spine Density (sp/μm): Control (vehicle and Aβ(42-1) combined: 0.79±0.04, 66 dendrites, 8 cells, Aβ(1-42): 0.59± 0.05, 24 dendrites, 4 cells, p<0.01) (E) Aβ over-expression depresses glutamatergic synaptic transmission (normalized to control(pA): AMPA: control: 1.0±0.07(40.4±2.7), APP: 0.56±0.08(22.7±3.2), n=13, p<0.01; NMDA: control: 1.0±0.24(32.3.0±7.7), APP: 0.53±0.16(17.2±5.0), n=11, p<0.01), while EGFP alone (AMPA: control: 1.0±0.15(33.3±5.0), EGFP: 1.25±0.14(41.7±4.7), n=14 pairs, p=0.4; NMDA: control: 1.0±0.20(24.3±4.8), EGFP: 1.03± 0.29(24.9±7.0), n=12, p=0.4) or with APP(MV) do not depress transmission (AMPA: control: 1.0±0.07(30.7±2.2), APP(MV): 0.95±0.09(29.3±2.6), n=20, p=0.6; NMDA: control: 1.0±0.25(12.2±3.0); APP(MV): 1.10±0.37(13.3±4.5), n=14, p=0.7). Scale: 20pA, 20ms. p values from t-test (spine density) and Wilcoxon test (EPSC). All data are reported as mean ± standard error of the mean (SEM). **=p<0.01.

Figure 2. Aβ-induced synaptic depression partially mimics and occludes mGluR long-term depression (LTD)

(A) Normalized EPSC amplitudes plotted against time from simultaneously patched pairs of uninfected (○) and β-CTF (♦) infected CA1 pyramidal neurons during baseline and after mGluR LTD induction (100μM DHPG, 5min) are shown on the left. Baseline and 35–40min data are summarized on the right. During baseline, the β-CTF infected neurons show depressed AMPA currents (normalized to control(pA): control: 1.0±0.08(82.7±7.0), β-CTF: 0.65±0.12(52.5±9.7), n=14, p<0.01). After DHPG, there is no difference in amplitude between control and infected neurons (normalized to control baseline(pA): control: 0.29±0.04(22.5±2.7), β-CTF: 0.28±0.06(19.6±2.4), n=9, p=0.6). Scale: 10pA, 20ms. (B) Aβ-induced depression of NMDA currents occludes mGluR-dependent LTD. Without DHPG, AMPA and NMDA currents are depressed in β-CTF infected cells (AMPA: control: 1.0±0.15(43.9±6.4), β-CTF: 0.51±0.10(22.4±4.3), n=16, p<0.01; NMDA: control: 1.0±0.10(29.1±3.0), β-CTF: 0.78 ± 0.09(22.7±2.6), n=15, p<0.01). After 5 min of 100μM DHPG, AMPA and NMDA currents are equivalent between control and infected neurons (AMPA: control: 1.0± 0.13(26.2±3.5), β-CTF: 1.12±0.18(29.4±4.7), n=17, p=0.8; NMDA: control 1.0±0.3(5.4±1.7), β-CTF: 1.3±0.3(6.8±1.5), n=16, p=0.2). (C) The APLP2/APP construct does not affect mGluR LTD induction. There is no difference in amplitude between control and infected cells before (AMPA: control: 1.0±0.13(47.0±6.3), APLP2/APP: 0.87±0.10(41.4±4.7), n=17 , p=0.3; NMDA: control: 1.0±0.16(24.9±3.9), APLP2/APP: 0.90±0.13(22.4±3.3), n=13, p=0.08) and after 5 min of 100μM DHPG (AMPA: control: 1.0±0.16(36.8±5.9), APLP2/APP: 0.92±0.14(33.9±5.0), n=13, p=0.3; NMDA: control: 1.0±0.25(35.8±8.8), APLP2/APP: 1.1±0.30(39.1±10.6), n=12, p=0.8). All p values from Wilcoxon test. All data are reported as mean ± SEM. **=p<0.01. Scale: 20pA, 20ms.

Figure 3. Aβ-induced synaptic depression requires p38 MAPK and calcineurin activity

(A) AMPA and NMDA EPSCs recorded from pairs of non-infected or APP infected CA1 pyramidal neurons. Slices maintained in culture with vehicle (normalized to control (pA): AMPA: control: 1.0±0.06(42.5±2.7), APP: 0.63±0.06(26.9±2.4), n=22, p<0.01 ; NMDA: control: 1.0±0.14(42.1±6.0), APP: 0.64±0.11(27.1±4.5), n=19, p<0.01), p38 MAPK inhibitor, SB203580(2μM) (AMPA: control: 1.0±0.09(37.4±3.3), APP: 0.87±0.07(32.5±2.6), n=19, p=0.14; NMDA: control: 1.0±0.14(15.5± 2.2), APP: 1.02±0.18(15.9±2.7), n=19, p=0.9), calcineurin inhibitor, FK506(50μM) (AMPA: control: 1.0±0.16(37.3±6.1), APP: 1.2±0.11(45.1±4.0) n=9, p=0.2; NMDA: control: 1.0±0.11(25.0±2.8), APP: 0.99±0.17(24.6±4.2), n=9, p=0.95) or JNK inhibitor, SP600125(5μM) (AMPA: control: 1.0±0.2(57.6±11.7), APP: 0.52±0.08(29.7±4.6), n=11, p=0.01; NMDA: control: 1.0±0.16(47.8±7.7), APP: 0.64±0.11(30.6±5.6), n=11, p<0.01). p values from Wilcoxon test. Scale: 20pA, 20ms (B) Results with cells expressing β-CTF. (Vehicle: AMPA: control: 1.0±0.15(43.9±6.4), β-CTF: 0.51±0.10(22.4±4.3), n=16, p<0.01; NMDA: control: 1.0±0.10(29.1±3.0), β-CTF: 0.78 ± 0.09(22.7±2.6), n=15, p<0.01; SB203580: AMPA: control: 1.0±0.14(29.4±4.1), β-CTF: 0.93±0.15(27.5±4.5), n=16, p=0.4; NMDA: control: 1.0±0.14(11.1±5.9), β-CTF: 1.05±0.15(11.6±1.6), n=15, p=0.6, FK506: AMPA: control: 1.0±0.17(32.4±5.7), β-CTF: 1.01±0.14(32.8±4.5), n=11, p=0.6; NMDA: control: 1.0±0.29(23.4±6.9), β-CTF: 0.98±0.22(23.1±5.2), n=10, p=0.8). p values from Wilcoxon test. Scale: 20pA, 20ms (C,D) Incubation of transgenic APP_Swe mouse hippocampal slices with p38 MAPK inhibitor (SB203580) does not affect Aβ40 or Aβ42 secretion (in pM: Vehicle: Aβ40: 2859±158, Aβ42: 194±12.5; SB203580: Aβ40: 3135±122, p=0.2, Aβ42: 211±9.0, p=0.3, n=4 for each condition) while calcineurin/PP2B inhibitor (FK506) slightly decreases Aβ40 and increases Aβ42 secretion (in pM: Vehicle: Aβ40: 3285±203, Aβ42: 162±10.0; FK506: Aβ40: 2626±216, p=0.04, Aβ42: 287±18.3, p<0.01, n=9 for each condition). A gamma secretase inhibitor (L-685,458, 2μM) decreases Aβ40 levels (in pM: for SB203580 trial: Aβ40: 1767±130, p<0.01, Aβ42: 175±5.6, p=0.2; for FK506 trial: Aβ40: 2049±208, p<0.001, Aβ42: 174±9.3, p=0.5). p values from t-test. (E) Sample traces of mEPSCs recorded in vehicle, p38 MAPK inhibitor (SB203580) and calcineurin/PP2B inhibitor (FK506). Scale: 20pA, 0.25s. (F) p38 MAPK inhibitor (SB203580) increases mEPSC frequency, while calcineurin/PP2B inhibitor (FK506) increases mEPSC frequency and amplitude (Frequency (Hz): Vehicle: 0.18±0.02, n=16, SB203580: 0.28±0.03, n=16, p=0.01, FK506: 0.29±0.03, n=16, p<0.01)(mEPSC amplitude (pA): Vehicle: 14.0±0.5, n=16, SB203580: 15.7±0.7, n=16, p=0.07, FK506: 17.3±0.9, n=16, p<0.01). p values from t-test. All data are reported as mean ± SEM. *=p<0.05. **=p<0.01.

Figure 4. Aβ reduces surface and synaptic GluR2

(A) 2PLSM images of neurons co-expressing GluR2 tagged with Super-ecliptic pHluorin (SEP-GluR2), T1dimer and APP or APP(MV). Note reduced surface SEP-GluR2 in cells expressing APP. Scale: 1μm. (B) 2PLSM images of neurons co-expressing GluR1 tagged with Super-ecliptic pHluorin (SEP-GluR1), T1dimer and APP or APP(MV). Note reduced surface SEP-GluR1 in cells expressing APP. Scale: 1μm. (C) The surface SEP-GluR2 (green signal) normalized by volume (red signal) (Green/Normalized Red) in spines and dendrites is decreased in APP co-expressing neurons compared to SEP-GluR2 expression alone or with APP(MV)(Spines: SEP-GluR2: 45996.2±926.7, n=1046 spines, APP/SEP-GluR2: 17144.3±402.0, n=482 spines, compare to SEP-GluR2: p<0.01; Dendrite: SEP-GluR2: 14760.8±463.8, n=1046 boxes, 4 cells, APP/SEP-GluR2: 4371.3±162.3, n=482 boxes, compare to SEP-GluR2: p<0.01, 5 cells; APP(MV)/SEP-GluR2: Spines: 36716.6±1517.3, n=554 spines, compare to APP/SEP-GluR2: p<0.01, Dendrites: 14378.2±745.4, n=554 boxes, compare to APP/SEP-GluR2: p<0.01, 5 cells). (D) The surface SEP-GluR1 normalized by volume (Green/Normalized Red) in spines and dendrites is decreased in APP co-expressing neurons compared to SEP-GluR1 expression alone or with APP(MV)(Spines: SEP-GluR1: 37683.5±1022.6, n=512 spines, APP/SEP-GluR1: 27045.5±1224.4, n=315 spines, compare to SEP-GluR1: p<0.01; Dendrite: SEP-GluR1: 14259.5.8±710.8, n=512 boxes, 4 cells, APP/SEP-GluR1: 10893.6±533.5, n=315 boxes, compare to SEP-GluR1: p<0.01, 5 cells; APP(MV)/SEP-GluR1: Spines: 35050.6±1060.5, n=505 spines, compare to APP/SEP-GluR1: p<0.01, Dendrites: 16184.8±785.4, n=505 boxes, compare to APP/SEP-GluR1: p<0.01). All data are reported as mean ± SEM. p values from t-test. **=p<0.01.

Figure 5. Aβ over-expression decreases synaptic GluR2 and depresses synaptic transmission. GluR2(R845A) blocks the Aβ-induced decrease in synaptic GluR2 and transmission

(A) The GluR2(R607Q) mutant increases rectification (normalized to control: control: 1.0±0.12, GluR2(R607Q): 2.17±0.33, n=12, p<0.01 ), which is not different from APP(MV)/GluR2(R607Q) expressing neurons (control: 1.0±0.04, APP(MV)/GluR2(R607Q): 2.03±0.22, n=18, p<0.01; compare to GluR2(R607Q): p=0.9). APP co-expression with GluR2(R607Q) decreases rectification (control: 1.0±0.06, APP/GluR2(R607Q): 1.53±0.12, n=20, p<0.01; compare to GluR2(R607Q): p=0.035; compare to APP(MV)/GluR2(R607Q): p<0.05 ). Over-expression of GluR2(R607Q;R845A) results in slightly increased rectification (control: 1.0±0.05, GluR2(R607Q;R845A): 2.83±0.31, n=13, p<0.01). Rectification of cells co-expressing GluR2(R607Q;R845A) with APP or APP(MV) is not statistically different from cells expressing receptor alone (control: 1.0±0.05, APP/GluR2(R607Q;R845A): 3.18±0.31, n=13, p<0.01, compare to GluR2(R607Q;R856A), p=0.4, compare to APP(MV)/GluR2(R607Q;R845A), p=0.6)(control: 1.0±0.18, APP(MV)/GluR2(R607Q;R845A): 3.05±0.33, n=15, p<0.01, compare to GluR2(R607Q;R845A), p=0.8). p values from t-test. (B) AMPA and NMDA EPSCs recorded simultaneously from pairs of untransfected and transfected neurons. Over-expression of GluR2(R607Q) alone or with APP(MV) does not affect synaptic transmission. (Normalized to control(pA): AMPA: control: 1.0±0.19(33.2±6.5), GluR2(R607Q): 1.03±0.13(34.2±4.3), n=10, p=0.6; NMDA: control: 1.0±0.28(22.6±6.4), GluR2(R607Q): 0.84±0.17(18.9±3.8), n=9, p=0.8)(AMPA: control: 1.0±0.10(39.9±4.1), APP(MV)/GluR2(R607Q): 1.17±0.09(46.6±3.6), n=24, p=0.6; NMDA: control: 1.0±0.18(30.4±5.4), APP(MV)/GluR2(R607Q): 0.85±0.15(25.9±4.7), n=14, p=0.06). APP co-expression with GluR2(R607Q) depresses synaptic transmission (AMPA: control: 1.0±0.06(48.5±3.1), APP/GluR2(R607Q): 0.64±0.05(31.1±2.6), n=26, p<0.01; NMDA: control: 1.0±0.18(19.4±3.6), APP/GluR2(R607Q): 0.57±0.09(11.0±1.83), n=13, p<0.01). GluR2(R607Q;R845A) expression alone potentiates AMPA, while not affecting NMDA EPSCs (AMPA: control: 1.0±0.10(37.7±3.8), GluR2(R607Q;R845A): 1.34±0.14(50.7±5.2), n=24, p<0.01; NMDA: control: 1.0±0.18(29.5±5.4), GluR2(R607Q;R845A): 0.96±0.18(28.2±5.5), n=13, p=0.4). Co-expression with APP or APP(MV) does not significantly change synaptic transmission when compared to receptor alone (AMPA: control: 1.0±0.08(35.6±2.9), APP/GluR2(R607Q;R845A): 1.39±0.11(49.5±4.1), n=28, p=0.01; NMDA: control: 1.0±0.16(27.9±4.4), APP/GluR2(R607Q;R845A): 0.90±0.08(25.2±2.1), n=14, p=0.8) (AMPA: control: 1.0±0.09(37.7±3.3), APP(MV)/GluR2(R607Q;R845A): 1.28±0.12(48.2±4.5), n=21, p=0.03; NMDA: control: 1.0±0.37(14.7±5.4), APP(MV)/GluR2(R607Q;R845A): 0.87±0.26(12.7±3.8), n=6, p=0.5). (C) Aβ application reduces synapticGluR2(R607Q). Hippocampal CA1 neurons were infected with GluR2(R607Q) and slices subsequently treated with 2 μM Aβ(1–42) or Aβ(42-1). Rectification was measured from neighboring infected and non-infected cells. (normalized rectification: Aβ (42-1): control: 1.0±0.09, GluR2(R607Q): 1.37±0.11; Aβ (1–42): control: 1.0±0.08, GluR2(R607Q): 1.09±0.08; each n=10; p<0.05). All data are reported as mean ± SEM. p values calculated with Wilcoxon test. ** =p<0.01. * =p<0.05. Scale: 20pA, 20ms.

Figure 6. Expression of S880E mutant of GluR2, which drives synaptic AMPA receptor endocytosis, results in depressed AMPA and NMDA mediated currents and spine loss

(A) Apical dendrites expressing GluR2(R607Q) or GluR2(R607Q;S880E). Scale: 10μm. (B) Over-expression of GluR2(R607Q;S880E) results in a decrease in spine density (GluR2(R607Q) (spines/μm): 0.78±0.03, 43 dendrites, 6 cells, GluR2(R607Q;S880E): 0.66±0.03, n=63 dendrites, 6 cells, p=0.01). p value calculated with t-test (C) Over-expression of GluR2(R607Q;S880E) depresses AMPA and NMDA EPSCs (normalized to control(pA): AMPA: control: 1.0±0.11(47.1±5.2), GluR2(R607Q;S880E): 0.80± 0.10(37.6±4.8), n=14, p=0.01; NMDA: control: 1.0±0.14(28.7±4.0), GluR2(R607Q:S880E): 0.61 ± 0.07(17.4±2.1), n=13, p<0.01). p values calculated with Wilcoxon test. Scale: 20pA, 20ms. **=p<0.01. *=p<0.05.

Figure 7. GluR2(R845A) blocks Aβ-induced loss in spine number

(A) Apical dendrites expressing indicated constructs (along with EGFP) were imaged with 2PLSM. Scale: 10μm. (B) Co-expression of APP with GluR2(R607Q) decreases spine density when compared to GluR2(R607Q) expression alone (Spine Density (sp/μm): GluR2(R607Q): 0.65±0.04, 32 dendrites; APP/GluR2(R607Q): 0.51±0.05, 33 dendrites, 4 cells/construct, p=0.036). When GluR2(R607Q;R845A) is co-expressed with APP, there is no difference in spine density when compared to GluR2(R607Q;R845A) expression alone (Spine Density (sp/μm): GluR2(R607Q;R845A): 0.73±0.06, 27 dendrites; APP/GluR2(R607Q;R845A): 0.77±0.05, 29 dendrites; 4 cells/construct, p=0.6). All data are reported as mean ± SEM. p values calculated with t-test. *=p<0.05.