Inactivation of γ-secretases leads to accumulation of substrates and non-Alzheimer neurodegeneration

Abstract

γ-Secretases are a family of intramembrane cleaving aspartyl proteases and important drug targets in Alzheimer's disease. Here, we generated mice deficient for all γ-secretases in the pyramidal neurons of the postnatal forebrain by deleting the three anterior pharynx defective 1 (Aph1) subunits (Aph1abc cKO Cre⁺). The mice show progressive cortical atrophy, neuronal loss, and gliosis. Interestingly, this is associated with more than 10-fold accumulation of membrane-bound fragments of App, Aplp1, Nrg1, and Dcc, while other known substrates of γ-secretase such as Aplp2, Lrp1, and Sdc3 accumulate to lesser extents. Despite numerous reports linking neurodegeneration to accumulation of membrane-bound App fragments, deletion of App expression in the combined Aph1 knockout does not rescue this phenotype. Importantly, knockout of only Aph1a- or Aph1bc-secretases causes limited and differential accumulation of substrates. This was not associated with neurodegeneration. Further development of selective Aph1-γ-secretase inhibitors should be considered for treatment of Alzheimer's disease.

Keywords: Alzheimer's disease; Aph1 subunit; selective inhibition; side effects; γ‐Secretase.

Figure EV1. Western blot analysis of cortical lysates of the different mouse strains used in the experiments

Cortical lysates from six wild‐type, six Aph1abc cKO Cre ⁻, six Aph1abc cKO Cre ⁺, six Aph1a cKO Cre ⁻, six Aph1a cKO Cre ⁺, six Aph1bc cKO Cre ⁻, and six Aph1bc cKO Cre ⁺ mice were analyzed by immunoblotting using antibodies against Nct, Pen‐2, Aph1a, Aph1bc, Psen1‐NTF, and App C‐terminus. Two samples per genotype are shown. The asterisk indicates a non‐specific band.

Figure EV2. Only complete loss of the Aph1 subunits in CaMKIIa neurons leads to decreased complex formation and impaired APP processing

1. Cortical lysates from wild‐type, Aph1abc cKO Cre ⁻, and Aph1abc cKO Cre ⁺ were analyzed by immunoblotting using antibodies against Nct, Pen‐2, and Psen1 NTF. Protein levels were standardized with β‐actin and normalized to wild types. Mean, SEM, and P‐values of 6 animals per genotype are shown. Differences in protein expression between wild‐type and Aph1abc cKO Cre ⁻ or Aph1abc cKO Cre ⁺ were computed using one‐way ANOVA per protein, followed by FDR P‐value adjustment of a Dunnett's post hoc test.

2. Cortical lysates from wild‐type, Aph1abc cKO Cre ⁻, and Aph1abc cKO Cre ⁺ were analyzed by immunoblotting using antibodies against App C‐terminus. Protein levels were quantified and FL/CTF ratios are plotted normalized to wild‐type controls. Mean, SEM, and P‐values of six animals per genotype are shown. One‐way ANOVA per protein, followed by FDR P‐value adjustment of the P‐values obtained from a Dunnett's post hoc test.

3. Cortical lysates from seven wild‐type, seven Aph1abc cKO Cre ⁻, and seven Aph1abc cKO Cre ⁺ mice were analyzed by Aβ ELISA. Aβ levels were normalized to wild types. Mean, SEM, and P‐values of seven animals are shown. One‐way ANOVA per protein, followed by FDR P‐value adjustment of a Dunnett's post hoc test. ns = not statistically significant.

Source data are available online for this figure.

Figure 1. Loss of complete γ‐secretase activity in CaMKIIa‐expressing neurons leads to neurodegeneration

1. Hematoxylin/eosin staining of sagittal brain slices of 9‐month‐old mice. Upper panel scale bar = 1,100 μm; lower panel scale bar = 195 μm. Cortical atrophy is present in Aph1abc cKO Cre ⁺ mice, whereas absent in Aph1abc cKO Cre ⁻ mice.

2. NeuN, Iba1 and Gfap immunofluorescent labeling of parietal cortex of 9‐month‐old mice. Scale bar = 115 μm. Full cortical thickness counts of NeuN‐positive cells, as determined via NeuN staining, was decreased in Aph1abc cKO Cre ⁺ mice. Reactive gliosis, as determined via Gfap and Iba1 staining, is present in Aph1abc cKO Cre ⁺ mice, whereas absent in Aph1abc cKO Cre ⁻ mice.

Data information: Mean, standard deviation, and P‐values are shown. One‐way ANOVA and Dunnett's post hoc test. CTX = cortex; cc = corpus callosum; ns = not statistically significant. N = 3.

Figure EV3. Aph1bc is functionally more prominent with regard to reconstitution of mature complex

1. Cortical lysates from six wild‐type, six Aph1a cKO Cre ⁺, and six Aph1bc cKO Cre ⁺ mice were analyzed by immunoblotting using antibodies against Nct, Pen‐2 and Psen1‐NTF. Protein levels were standardized with β‐actin and normalized to wild type.

2. Cortical lysates from six wild‐type, six Aph1a cKO Cre ⁺, and six Aph1bc cKO Cre ⁺ mice were analyzed by immunoblotting using antibodies against App C‐terminus. Protein levels were quantified and FL/CTF ratios are plotted normalized to wild‐type control.

3. Cortical lysates from seven wild‐type, seven Aph1a cKO Cre ⁺, and seven Aph1bc cKO Cre ⁺ mice were analyzed by Aβ ELISA. Aβ levels were normalized to wild type.

Data information: Mean, SEM, and P‐values are shown. One‐way ANOVA per protein, followed by FDR P‐value adjustment of a Dunnett's post hoc test. ns = not statistically significant. Source data are available online for this figure.

Figure 2. No neurodegenerative phenotype observed in single Aph1a and Aph1bc cKO Cre ⁺ mice

1. Hematoxylin/eosin staining of sagittal brain slices of 6‐month‐old mice. Scale bar = 195 μm. Cortical atrophy is absent in Aph1a cKO Cre ⁺ and Aph1bc cKO Cre ⁺ brains.

2. NeuN, Iba1, and Gfap immunofluorescent labeling of parietal cortex. Scale bar = 115 μm. Full cortical thickness counts of NeuN‐positive cells, as determined via NeuN staining, was comparable in control, Aph1a cKO Cre ⁺, and Aph1bc cKO Cre ⁺ mice. Reactive gliosis, as determined via Gfap and Iba1 staining, is absent in Aph1a cKO Cre ⁺ and Aph1bc cKO Cre ⁺ mice.

Data information: Mean and standard deviation are shown. One‐way ANOVA and Dunnett's post hoc test. ns = not statistically significant. N = 3.

Figure 3. Differential accumulation of substrates in vivo

Hippocampal lysates were analyzed by immunoblotting using antibodies against the C‐terminus of the different γ‐secretase substrates. For every substrate, full‐length (FL) or the 80‐kDa furin‐cleaved LRP fragment and C‐terminal fragments (CTF) are shown. β‐Actin was used as loading control to normalize the data. Molecular weight markers are depicted in kDa.

1. Western blot analysis of hippocampal lysates from Aph1abc cKO Cre ⁻ and Aph1abc cKO Cre ⁺ mice, three representative samples are shown. Five mice per genotype were analyzed.

2. Western blot analysis of hippocampal lysates from wild‐type, Aph1a cKO Cre ⁺ , and Aph1bc cKO Cre ⁺, three representative samples are shown. Ten mice per genotype were analyzed.

3. FL and CTF protein levels in panel (A) were quantified and FL/CTF ratios are plotted, normalized to Aph1abc cKO Cre ⁻ controls. Significant differences in protein expression between controls and Aph1abc cKO Cre ⁺ mice were assessed using Student's t‐test per protein (GraphPad), followed by false discovery rate (FDR) P‐value adjustments (R, version 3.3.1) to correct for multiple protein testing. Mean, SEM, and P‐values are shown, ns = not statistically significant, N = 5.

4. FL and CTF protein levels in panel (B) were quantified and FL/CTF ratios are plotted, normalized to wild‐type controls. Differences in protein expression between wild‐type, Aph1a cKO Cre ⁺, and Aph1bc cKO Cre ⁺ were computed using one‐way ANOVA per protein, followed by FDR P‐value adjustment of the Tukey's post hoc test. Mean, SEM, and P‐values are shown, ns = not statistically significant, N = 10.

Source data are available online for this figure.

Figure 4. App‐CTFs accumulate in the neurites and synaptic compartments in Aph1abc cKO Cre ⁺ brains

Immunostaining with an antibody recognizing the C‐terminal part of App. Sagittal brain slices from App KO mice were used to show the specificity of the antibody. Conditional deletion of the Aph1 genes in the pyramidal neurons causes accumulation of App‐CTFs in the neurites in the hippocampal and cortical areas. Zoom‐ins on the dentate gyrus (DG) and CA3 region of the hippocampus and the parietal cortex overlying the hippocampus show that expression of App and App‐CTFs is mainly confined to the neuronal somata in control brains, whereas App‐CTFs accumulate in the neurites and synaptic compartments in Aph1abc cKO Cre ⁺ brains. Scale bar for left panels = 390 μm; scale bar for central panels = 150 μm; scale bar for right panels = 110 μm.

Figure 5. App depletion does not rescue cortical atrophy and gliosis in Aph1abc cKO Cre ⁺ mice

1. Hematoxylin/eosin staining of sagittal brain slices of 9‐month‐old mice. Scale bar = 195 μm. Progressive cortical atrophy is present to the same extent in Aph1abc cKO Cre ⁺ and Aph1abc cKO Cre ⁺ × App KO mice, whereas absent in control and App KO mice.

2. NeuN, Iba1, and Gfap immunofluorescent labelings of parietal cortex. Scale bar = 115 μm. Full cortical thickness counts of NeuN‐positive cells, as determined via NeuN staining, was comparable in control and App KO mice, whereas decreased to the same extent in Aph1abc cKO Cre ⁺ and Aph1abc cKO Cre ⁺ × App KO mice. Reactive gliosis, as determined via Gfap and Iba1 staining, is present to the same extent in Aph1abc cKO Cre ⁺ and Aph1abc cKO Cre ⁺ × App KO mice, whereas absent in control and App KO mice.

Data information: Mean, standard deviation, and P‐values are shown, ns = not statistically significant. One‐way ANOVA and Dunnett's post hoc test. N = 3.

Figure EV4. Progressive neurodegenerative phenotype in Aph1abc cKO Cre ⁺ mice

Hematoxylin/eosin staining of sagittal brain slices of 3‐, 6‐, and 9‐month‐old mice. Scale bar = 195 μm. Absolute numbers of cortical thickness are normalized to Cre ⁻ littermates for each age. Progressive cortical atrophy is present in Aph1abc cKO Cre ⁺ mice. Representative pictures for the 9‐month‐old mice have been reported also in Fig 1. Mean, standard deviation, and P‐values are shown. Two‐way ANOVA and Tukey's post hoc test. CTX = cortex; cc = corpus callosum, ns = not statistically significant. N = 3.