Aβ-dependent reduction of NCAM2-mediated synaptic adhesion contributes to synapse loss in Alzheimer's disease

Abstract

Alzheimer's disease (AD) is characterized by synapse loss due to mechanisms that remain poorly understood. We show that the neural cell adhesion molecule 2 (NCAM2) is enriched in synapses in the human hippocampus. This enrichment is abolished in the hippocampus of AD patients and in brains of mice overexpressing the human amyloid-β (Aβ) precursor protein carrying the pathogenic Swedish mutation. Aβ binds to NCAM2 at the cell surface of cultured hippocampal neurons and induces removal of NCAM2 from synapses. In AD hippocampus, cleavage of the membrane proximal external region of NCAM2 is increased and soluble extracellular fragments of NCAM2 (NCAM2-ED) accumulate. Knockdown of NCAM2 expression or incubation with NCAM2-ED induces disassembly of GluR1-containing glutamatergic synapses in cultured hippocampal neurons. Aβ-dependent disassembly of GluR1-containing synapses is inhibited in neurons overexpressing a cleavage-resistant mutant of NCAM2. Our data indicate that Aβ-dependent disruption of NCAM2 functions in AD hippocampus contributes to synapse loss.

Figures

Figure 1. Synaptic accumulation of NCAM2 is reduced in the hippocampus of AD-affected individuals.

(a) Western blot analysis of homogenates and synaptosomes prepared using brain tissue from cerebellum, temporal cortex and hippocampus of control and AD individuals. Note enrichment of synaptophysin, PSD95, actin, VGLUT and VGAT in synaptosomes indicating efficient synaptosome isolation. NCAM2 is highly enriched in synaptosomes versus homogenates from the hippocampus in control but not in AD individuals. Full-length versions of the western blots are shown in Supplementary Figs 6 and 7. (b) Graphs show the ratio of the respective protein levels in synaptosomes to homogenates for individual cases and mean±s.e.m. (n=10 control and n=10 AD cases were analysed). *P=0.0039, Mann–Whitney test. (c) Western blot analysis of the soluble protein fractions and synaptosomes prepared using brain tissue from the cerebellum, temporal cortex and hippocampus of control and AD individuals. Total protein concentration in synaptosomes was kept at 25% of that in the soluble protein fraction to improve visualization of the protein bands in both fractions on one blot. Probes were analysed with antibodies against the extracellular domain of NCAM2 and GAPDH, which served as a loading control. Note increased levels of soluble ∼100-kDa NCAM2 fragments in the soluble protein fraction from the hippocampus and temporal cortex of the AD individuals. Full-length versions of the western blots are shown in Supplementary Fig. 8. (d,e) Graphs show NCAM2 levels in the soluble protein fraction normalized to NCAM2 levels in synaptosomes (d) and NCAM2 levels in the soluble protein fraction normalized to GAPDH levels in homogenates (e) for individual cases and mean±s.e.m. (n=10 control and n=10 AD cases were analysed). *P<0.05, Mann–Whitney test.

Figure 2. Cleavage of the membrane-adjacent extracellular fragment of NCAM2 is increased in AD brains.

(a) Diagram showing the structure of NCAM2. Ig, immunoglobulin-like domain; Fn, fibronectin type III domain. Peptides used in the cleavage assay and corresponding to aa682-701 (NCAM2aa682-701) and aa666-685 (NCAM2aa666-685) of human NCAM2 are shown below. Asparagine 689 and aspartic acid 693 exchanged to alanine in the cleavage assay shown in c are highlighted in bold. (b) Scheme of the peptide cleavage assay. Peptides labelled with FITC and biotin at the N- and C-terminus, respectively, were incubated either with the total brain lysate or lysate of synaptosomes. Non-cleaved peptides and biotin-containing fragments of the cleaved peptides were removed using streptavidin-coated beads. The remaining FITC fluorescence was used as an estimate of peptide cleavage. (c) Graphs show the efficiency of the peptide cleavage (mean+s.e.m.) with the fluorescence signals for NCAM2aa682-701 in controls set to 100%. Lysates from eight controls and eight AD patients were analysed. Note that the efficiency of NCAM2aa682-701 cleavage is higher in AD cases and particularly in synaptosomes. NCAM2aa682-701 cleavage is reduced by mutating aspartic acid 693. *P<0.05 (compared as indicated), ^P<0.05 (compared with NCAM2aa682-701 cleavage in AD) analysis of variance with Tukey's multiple comparisons test.

Figure 3. The extracellular domain of NCAM2 binds to Aβ.

(a) Recombinant extracellular domains of NCAM2 (NCAM2-ED) analysed by silver staining and western blot with antibodies against the extracellular domain of NCAM2. (b) NCAM2-ED immobilized on the plastic surface of 96-well plates was assessed by ELISA for its ability to bind increasing concentrations of Aβ1-42. Mean+s.e.m. (n=3) OD values from a representative experiment are shown. Note that Aβ1-42 binds to NCAM2-ED but not to BSA in a concentration-dependent manner. The experiment was performed five times with the same effect. (c) DLS analysis of the hydrodynamic diameters of the protein particles in the solutions containing Aβ1-42 oligomers alone, NCAM2-ED alone or a mixture of Aβ1-42 oligomers and NCAM2-ED. Note that the particle size peaks (marked by black dashed lines) are shifted to a larger hydrodynamic diameter (grey dashed line) in the solution containing a mixture of Aβ1-42 oligomers and NCAM2-ED when compared with solutions containing Aβ1-42 oligomers or NCAM2-ED alone. (d) DLS analysis of the hydrodynamic diameters of the protein particles in the solutions containing Aβ1-42 oligomers alone, BSA alone or a mixture of Aβ1-42 oligomers and BSA. Note that the particle size peaks (marked by dashed lines) are not shifted when Aβ1-42 oligomers are incubated with BSA. The experiment in c,d was performed twice with the same effect and a representative experiment is shown. (e) Western blot analysis of the Aβ1–42 oligomer preparation used in this study. (f,g) PAGE (f) and western blot (g) analyses of the probes containing NCAM2-ED, NCAM2-ED incubated with Aβ1-42, or Aβ1-42 performed under non-reducing conditions. Dashed lines and arrows indicate the position of the peaks in the labelling density determined by the densitogram analysis. A fragment of the densitogram covering 75–150 kDa range is shown in f. Note that the density peak in the NCAM2 band in the probes incubated with Aβ1-42 is shifted to the higher molecular weight (grey arrows) when compared with the NCAM2 band in the absence of Aβ1-42 (black arrows). Full-length versions of the western blots are shown in Supplementary Fig. 9. OD, optical density. AU, arbitrary units.

Figure 4. NCAM2 accumulates in excitatory synapses of cultured hippocampal neurons.

(a) Low-magnification image of a cultured hippocampal neuron labelled by indirect immunofluorescence with antibodies against NCAM2, synaptophysin and MAP2. Note expression of NCAM2 along MAP2 positive dendrites. NCAM2 is also expressed in astrocytes (marked a) which are present in these cultures. Scale bar, 20 μm. (b) High-magnification image of dendrites of neurons co-labelled with antibodies against NCAM2, synaptophysin and MAP2. Arrows show clusters of NCAM2 partially overlapping with synaptophysin accumulations. NCAM2-negative synapses are also observed (arrowheads). Scale bar, 10 μm. (c) High-magnification image of a dendrite of a cultured hippocampal neuron labelled with antibodies against NCAM2, synaptophysin and PSD95. NCAM2 clusters partially overlap with accumulations of PSD95 and synaptophysin (arrows). Scale bar, 10 μm. Three-dimensional analysis of the co-localization within the outlined area is on the right. Z-stack has been acquired with 0.15 μm steps. The xz and yz sections along the dashed lines on the xy image are shown. Note co-localization of the NCAM2 cluster with synaptic markers. (d) Negative control, that is, labelling performed without primary antibodies, is shown. Scale bar, 10 μm.

Figure 5. Aβ1–42 oligomers bind to NCAM2 at the cell surface of neurons.

Representative images of cultured hippocampal neurons, which were either mock-treated or incubated with Aβ1-42 oligomers for 30 min or 24 h. Aβ1-42/NCAM2 complexes at the cell surface were detected by PL using antibodies against Aβ1-42 and the extracellular domain of NCAM2 applied to detergent non-permeabilized neurons. Neurons were then co-labelled with antibodies against Aβ1-42 and synaptophysin used to visualize neurons. Differential interference contrast images (DIC) and inverted grey scale fluorescence images are shown. Note endogenous APP labelling in mock-treated neurons and increased levels of Aβ1-42 immunoreactivity and NCAM2/Aβ1-42 proximity ligation reaction products along neurites of neurons treated with Aβ1-42 oligomers for 30 min. NCAM2/Aβ1-42 proximity ligation reaction products and Aβ1-42 oligomers were observed along neurites and also adsorbed to the substrate around neurons treated with Aβ1-42 oligomers for 24 h. A dead neuron with disrupted morphology of the soma, fragmented dendrites and highly positive for Aβ1-42 is marked with an arrow. Graphs show levels of Aβ1-42 and synaptophysin immunoreactivity and Aβ1-42/NCAM2 proximity ligation products measured along neurites and randomly sampled areas around neurons (mean+s.e.m., n>50 neurons analysed in each group). P<0.05, analysis of variance with uncorrected Fisher's least square difference test. The experiment was performed twice with the same effect. Scale bar, 40 μm.

Figure 6. Levels of NCAM2 are reduced in synaptosomes of cultured hippocampal neurons treated with Aβ1-42 oligomers.

(a) Western blot analysis of homogenates and synaptosomes isolated from cultured hippocampal or cortical neurons either mock treated or incubated with Aβ1-42 oligomers for 24 h. Blots were probed with antibodies against NCAM2 and synaptophysin used as a loading control. Note that NCAM2 is enriched in synaptosomes versus homogenates in hippocampal neurons and this enrichment is reduced in hippocampal neurons treated with Aβ1-42. Levels of NCAM2 in synaptosomes from cortical neurons are not reduced in response to Aβ1-42. NCAM2 levels are increased in homogenates of neurons treated with Aβ1-42 versus mock-treated neurons. Graphs show mean+s.e.m. fold change in NCAM2 levels in cell culture homogenates with levels in mock-treated neurons set to 1 (left panels), and mean+s.e.m. of the synaptic enrichment of NCAM2 defined as a ratio of NCAM2 levels in synaptosomes to homogenates (right panels) from n=5 independent experiments. *P<0.05 paired t-test. (b) Western blot analysis of the cell culture media collected from the hippocampal and cortical neurons either mock-treated or incubated with Aβ1-42 oligomers for 24 h. Blots were probed with antibodies against the extracellular domain of NCAM2. Note that levels of the soluble ∼100 kDa extracellular fragments of NCAM2 are increased in the culture media from cultured hippocampal neurons treated with Aβ1-42. Graphs show quantification of the blots (mean+s.e.m., n=3 independent experiments for hippocampal neurons, n=6 independent experiments for cortical neurons) with levels in mock-treated cultures set to 1. *P<0.05 paired t-test. For a,b, full-length versions of the blots are shown in Supplementary Fig. 10.

Figure 7. NCAM2 co-localizes with Aβ1-42 in brains of APP23 transgenic mice.

Low-magnification confocal images of the hippocampus (a) and high-magnification images of the CA1 region of the hippocampus (d) of 9-month-old wild-type (wt) and APP23 mice (APP23) are shown. Brain sections were immunolabelled with antibodies to NCAM2 and Aβ and counter stained with DAPI. Lower panel in d shows a blow-up of the area outlined with dashed lines. Note accumulations of Aβ co-localizing with clusters of NCAM2 along dendrites of neurons. Negative controls (labelling performed without primary antibodies) are shown in b,c. Scale bar=200 μm (a–c), 40 μm (d), 10 μm (d, blow-up image).

Figure 8. NCAM2 binds to Aβ and its synaptic accumulation is reduced in the hippocampus of APP23 transgenic mice.

(a) Western blot analysis of the homogenates (hom.) and synaptosomes (synapt.) prepared using hippocampus, cortex and cerebellum from wild-type and APP23 transgenic (APPtg) mice. Probes were analysed with antibodies against NCAM2 and synaptophysin used as a loading control. Note reduced synaptic enrichment of NCAM2 in the hippocampus of APP23 transgenic mice. Full-length versions of the western blots are shown in Supplementary Fig. 11. (b–d) Graphs show synaptic enrichment of NCAM2 defined as a ratio of NCAM2 levels in synaptosomes and homogenates (mean+s.e.m., samples from four 5–9-month-old, four 12-month-old and two 15-month-old pairs of littermates were analysed twice by western blot). *P<0.05 paired t-test. (e) Western blot analysis of the soluble protein fraction and synaptosomes prepared using hippocampus, cortex and cerebellum from wild-type and APP23 transgenic mice. Probes were analysed with antibodies against NCAM2. Note reduced synaptic enrichment of NCAM2 in the hippocampus of APP23 mice (short exposure) and higher levels of soluble NCAM2 fragments in the hippocampus of APP23 mice when compared with wild-type hippocampus revealed after prolonged exposure of the blot (long exposure). Full-length versions of the western blots are shown in Supplementary Fig. 12. (f) Western blot analysis of NCAM2 immunoprecipitates (IP) from the cortex and hippocampus of wild-type and APP23 mice. Mock IP with non-immune immunoglobulins (Ig) served as control. Membranes were analysed with antibodies to NCAM2, human APP and Aβ (6E10, Covance), or Aβ1-42 (D3E10, Cell Signaling). Note full-length APP and Aβ immunoreactivity in NCAM2 immunoprecipitates from APP23 mice. A band representing the heavy chain of the immunoglobulins (Ig-HC) used for immunoprecipitation and recognized by the secondary antibodies is also observed. Labelling with antibodies to NCAM2 shows NCAM2 immunoprecipitation efficiency. Full-length versions of the western blots are shown in Supplementary Fig. 13.

Figure 9. Disruption of NCAM2 functions at the neuronal cell surface promotes glutamatergic synapse disassembly.

(a–e) Cultured hippocampal neurons were either mock-treated or incubated with the recombinant soluble extracellular domains of NCAM2 (NCAM2-ED), antibodies against the extracellular domain of NCAM2 (NCAM2mAb), or Aβ1-42 oligomers. In a,b, neurons were labelled with antibodies against the extracellular domain of GluR1 before permeabilization of membranes with detergent, and co-labelled with antibodies against synaptophysin after permeabilization of membranes with detergent. Representative images of dendrites are shown (a). Note co-localization of cell surface GluR1 accumulations with synaptophysin clusters in mock-treated neurons, and increased levels of non-synaptic cell surface GluR1 accumulations in neurons treated with NCAM2-ED, NCAM2mAb or Aβ1-42. Graphs (b) show the percentage of synaptic and non-synaptic GluR1 clusters relative to total number of GluR1 clusters along dendrites and numbers of synaptophysin accumulations per dendrite length (mean+s.e.m.). *P<0.0001 (analysis of variance with Dunnett's multiple comparison test, n>80 dendrites from 20 neurons were analysed in each group). In c, neurons were labelled with antibodies against the extracellular domain of NR1 before permeabilization of membranes with detergent, and co-labelled with antibodies against synaptophysin after permeabilization of membranes with detergent. Graphs show the percentage of synaptic and non-synaptic NR1 clusters relative to total number of NR1 clusters along dendrites (mean+s.e.m.). *P<0.0001 (analysis of variance with Dunnett's multiple comparison test, n>85 dendrites from 20 neurons were analysed). In d,e, neurons were co-labelled with fluorescent phalloidin and synaptophysin antibodies. Representative images of dendrites are shown in d. Note higher labelling intensity and co-localization with synaptophysin of the phalloidin-labelled polymerized actin accumulations in control neurons versus neurons treated with Aβ1-42, NCAM2-ED or NCAM2mAb. Note increased numbers of filopodia and lamellipodia in neurons treated with Aβ1-42, NCAM2-ED or NCAM2 mAb. Graphs (e) show ratio of the dendrite area-to-length and phalloidin labelling intensity of dendrites of neurons. Mean values+s.e.m. are shown. *P<0.0001 (analysis of variance with Dunnett's multiple comparison test, n=50 dendrites from 20 neurons were analysed in each group). Scale bar, 10 μm (in a,d).

Figure 10. Aβ1-42 reduces the number of GluR1-containing synapses in the NCAM2-dependent manner.

(a) Representative images of dendrites of cultured hippocampal neurons transfected either with control negative miRNA (negative miR) or NCAM2miR and either mock-treated or incubated with Aβ1-42. Transfected neurons were identified by fluorescence of GFP, which is co-expressed together with miRNA. Neurons were co-labelled with antibodies against cell surface GluR1 and synaptophysin. Note that the number of synaptic GluR1 clusters is reduced and the number of non-synaptic GluR1 clusters is increased in neurons transfected with NCAM2miR. Scale bar, 10 μm. (b,c) Graphs show mean+s.e.m. percentage of synaptic and non-synaptic GluR1 clusters relative to the total number of GluR1 clusters along dendrites (b) and numbers of synaptophysin accumulations per dendrite length normalized to the mean number in mock-treated neurons (c) for neurons described in (a). (d–f) Graphs show mean+s.e.m. percentage of synaptic and non-synaptic GluR1 clusters relative to the total number of GluR1 clusters along dendrites (d), number of synaptophysin accumulations per dendrite length normalized to the mean number in mock-treated neurons (e), and area/length ratio (f) in cultured hippocampal neurons transfected either with GFP alone or co-transfected with GFP and non-mutated NCAM2 (NCAM2WT) or NCAM2D693A mutant and either mock-treated or incubated with Aβ1-42. (g,h) Graphs show mean+s.e.m. percentage of non-synaptic GluR1 clusters relative to the total number of GluR1 clusters along dendrites (g) and area/length ratio (h) in cultured hippocampal neurons co-transfected with NCAM2 miR and either GFP, non-mutated NCAM2 (WT) or NCAM2D693A mutant (D693A) and either mock-treated or incubated with Aβ1-42. In b–h, *P<0.01 (compared as indicated), ^P<0.01 (compared with mock-treated neurons transfected with negative miR (b), GFP (d–f) or co-transfected with NCAM2miR and GFP (g–h)), analysis of variance with Tukey's multiple comparison test, n>50 dendrites from 20 neurons were analysed in each group.