Intraneuronal aggregation of the β-CTF fragment of APP (C99) induces Aβ-independent lysosomal-autophagic pathology

Virus production was performed following a protocol previously described [4]. Briefly, HEK293 cells were transfected with the adenovirus helper plasmid (pXX6), the AAV packaging plasmid (rAAV2-rh10), and the AAV10 plasmid empty vector or encoding either human C99 or GFP under control of the synapsin-1 promoter (AAV-empty, AAV-synC99 and AAV-synGFP). Viruses were produced, purified and vector titers were determined by real-time PCR and expressed as viral genomes per ml (vg/ml) [4].

3xTgAD (harboring PS1_M146V, βAPP_swe, and Tau_P301L transgenes) and non-transgenic (nonTg) mice [37] were generated from breeding pairs provided by Dr. LaFerla (Irvine, USA). 2xTgAD (PS1wt, βAPP_swe and Tau_P301L) were produced by crossing 3xTgAD with nonTg mice, as described [38]. For AAV-mediated in vivo delivery, 1-day-old C57BL6 mice (Janvier Labs., France) were injected with 4 µl of AAV virus (5.5 × 10¹² vg/ml (viral genomes per ml)) into the left lateral ventricle, as described [21] and mice were analyzed at 2 months post-AAV delivery. 5-month-old nonTg and 3xTgAD males or 2-month-old AAV-infected mice (males and females) were treated daily for 12 or 30 days with the γ-secretase inhibitor ELND006, referred to as D6 hereafter (30 mg/kg, Elan Pharmaceuticals, San Francisco, USA) [9, 45] or with vehicle alone (methylcellulose/polysorbate 80, Sigma) via oral gavage, as described [24]. All experimental procedures were in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC) and local French legislation.

Animals were deeply anesthetized with a lethal dose of pentobarbital and perfused transcardially with cold phosphate-buffered saline (PBS). Brains were fixed in 4 % Paraformaldehyde/PBS then embedded in paraffin and sliced (8 μm) or cut on a vibratome (50 μm). For FCA18, 82E1, NU1, 4G8 and 6E10 staining, sections were treated with formic acid (90 % or 50 %/5 min for paraffin and vibratome sections, respectively). Sections were then incubated at 4 °C overnight with primary antibodies (see supplementary table) followed by Alexa Fluor-conjugated antibodies (Molecular Probes, 1:1000). Cathepsin B labeling was amplified using the Vectorstain ABC kit (Vector) and streptavidin-Alexa594 (Molecular Probes, 1:1000). Nuclei were stained with DAPI (Roche, 1:20,000) and fluorescence was visualized using a confocal microscope (Fluoview10, Olympus). For DAB development, sections were incubated with HRP-conjugated secondary antibodies (Jackson ImmunoResearch, 1/1000) followed by DAB substrate (DAB impact, Vector). Neuronal nuclei were then stained with cresyl violet. For quantitative analysis of lamp1 staining in AAV-empty and AAV-C99 brain slices, images were thresholded, converted to mask and analyzed using Particle Analysis ImageJ plugin.

The human neuroblastoma cell line SH-SY5Y, naive or stably expressing APPswe or pcDNA3 (mock), previously described [40] was exposed to the following drugs for 16–20 h: D6 (50 nM to 5 μM, Elan Pharmaceuticals, San Francisco) in vehicle (methylcellulose/polysorbate 80, Sigma), DAPT (5 μM in DMSO, Sigma), leupeptin (10 μM in H₂O, Sigma), PADK (5 μM in DMSO, Bachem), NH₄Cl (10 mM in H₂O, Sigma), pepstatin A (10 μM in EtOH, Enzo Life sciences), smer28 (50 μM in DMSO, Sigma) and bafilomycin A1 (50 nM in H20, Sigma). Some cells were transfected with GFP-LC3 (Addgene) using Lipofectamine 2000 according to standard protocols. Other cells were infected with lentiviruses (LV) encoding mouse cathepsin B (LV-mCatB-FUGW2) or control (LV-FUGW2) [32] and polyclonal cell lines stably expressing mCatB (naive SH-SY5Y or APPswe-SH-SY5Y cells) were obtained. For immunocytochemical experiments of C99 expression, COS-7 cells were transfected with C99 using Lipofectamine 2000.

Cells were immunostained with α-APPct (1:5000) and α-lamp-1 (E-5, 1:100) or α-cis-golgi (GM130, Cell signaling, 1:400) followed by Alexa Fluor-488 or Alexa Fluor-594 conjugated antibodies (1:1000, Molecular Probes). Nuclei were stained with DAPI (1:20,000, Molecular probes). Other cells were incubated 30 min with Lysotracker-Red DND-99 (1:20,000, Invitrogen) at 37 °C, fixed with PFA 4 % and stained with DAPI. Images were acquired using an inverted confocal microscope (Fluoview10, Olympus, France).

Recombinant C99 (C100-flag), previously described [49], was incubated in the absence or presence of purified cathepsin B from human placenta (5, 10 or 50 ng, Sigma) in 100 mM sodium acetate pH 5 buffer containing 1 mM EDTA and 8 mM cysteine at 37 °C during 60 min, then analyzed on Tris-Tricine 16 % acrylamide gels as described below. C100flag was revealed using an anti-flag antibody (M2, 1:5000, Sigma).

Cells were fixed 20 min with 2.5 % glutaraldehyde/phosphate buffer 0.1 N (pH 7.4). Mice were anesthetized with a lethal dose of pentobarbital and perfused transcardially with ice cold physiological saline followed by 2 % glutaraldehyde/cacodylate buffer 0.1 M (pH 7.4). The brain was sliced (200 μm) on a vibratome and 2 mm cubes from the subiculum were microdissected under binoculars and post-fixed in osmium tetroxide (1 % in cacodylate buffer 0.1 M). The tissue was embedded in Epon resin (EMS) and 80 nm ultrathin sections were contrasted with uranyl acetate and lead citrate and visualized using a JEM 1400 electron microscope operating at 100 kV equipped with a Morada SIS camera.

Dissected hippocampi of vehicle or D6-treated 3xTgAD or nonTg mice, or hemispheres from AAV-injected mice, were homogenized in RIPA buffer (Tris 50 mM; pH 7.4 containing NaCl (150 mM), EDTA (1 mM), Triton X100 (1 %), deoxycholate (0.5 %), SDS (0.1 %) and protease inhibitor cocktail (Sigma) and soluble and insoluble fractions were prepared as described [24]. Synaptosomal and microsomal enriched fractions were prepared from hippocampi of vehicle or D6-treated 3xTgAD mice. Tissue was homogenized in 0.32 M sucrose containing protease inhibitor cocktail (Sigma), centrifuged twice at 850×g/5 min and the collected supernatants were centrifuged twice at 12,000×g/20 min. The obtained pellet (“synaptomal fraction”) was resuspended in Hepes 4 mM/EDTA. The supernatant was centrifuged at 20,000×g/1 h to obtain the “microsomal fraction”.

Proteins were separated on 16 % Tris-Tricine gels (for APP-CTFs, cathepsin B and LC3) or 10 % Tris–glycine gels (for βAPP p62 and actin) and transferred to nitrocellulose membranes. After probing with primary antibodies (see supplementary table), immunological complexes were revealed with HRP-conjugated antibodies (Jackson ImmunoResearch, 1/10,000) followed by electrochemiluminescence (Supersignal West Pico/West Maximum Sensitivity chemiluminescence substrate, Thermo Scientific, France). Peak height of signal intensities from protein bands were quantified with ImageJ software.

Aβ40 and Aβ42 levels were quantitatively detected in soluble and insoluble fractions (see above) of mice hippocampi (for 3xTgAD mice) or hemispheres (for AAV-synC99 mice) using sandwich ELISA kits detecting human Aβ40 and Aβ42, respectively (Biosource, Invitrogen, France). Absorbance was read at 450 nm using a spectrophotometer.

Cells were lysed mechanically in homogenization buffer (250 mM sucrose, 1 mM EDTA, 5 mM Hepes pH 7.4). The cell suspension was centrifuged at 850×g/5 min and the supernatant was centrifuged at 20,000×g/90 min. The pellet (membrane enriched fraction) was resuspended in Tris–HCl (10 mM, pH 7.5). To monitor cathepsin B (CatB) activity, 60 µg of protein extracts were incubated in acetate buffer (25 mM, pH 5.5, 100 µl and 8 mM l-cysteine HCl) containing CatB substrate (CBZ-Arg-Arg-7-Amido-4-methylcoumarin, 100 µM, Sigma) in the absence or presence of leupeptin (10 µM, Sigma). Specific CatB activity corresponds to the leupeptin-sensitive fluorescence recorded at 320 nm (excitation) and 420 nm (emission) using a fluorescence plate reader (FLUOstar Omega, BMG Labtech, France). CatB activity was calculated as the slope in the linear range corresponding to the initial 30 min.

LTP measurements were performed by E-Phy-Science, Sophia Antipolis, France. 2-month-old AAV-empty, vehicle or D6-treated AAV-C99 females, were anesthetized with isoflurane and decapitated. Stimulations were delivered in CA1 fibers in the alveus and field excitatory postsynaptic potential (f-EPSP) were recorded from the subicular pyramidal neurons in the middle portion of the subiculum. The LTP-induction protocol consisted of four trains of 100 stimuli at 100 Hz repeated every 20 s, as described [41].

Statistical analysis was performed with PRISM software (Graph-Pad Software, San Diego, CA, USA) by using a non-parametric two-tailed Mann–Whitney U test for pairwise comparisons or the two-way ANOVA test followed by either the Tukey’s post hoc test for multiple comparisons or the Dunnett post hoc test when comparison to controls. All data are expressed as the mean ± SEM. Differences were considered significant at *p < 0.05, **p < 0.01, ***p < 0.001.

We previously demonstrated that the β-secretase-derived APP C-terminal stub C99 corresponded to the early appearing and accumulating intracellular label in the subiculum of 3xTgAD mice [24]. This label could be detected with N-terminal directed antibodies including the β-secretase-mediated cleavage-specific antibodies FCA18 [3] and 82E1 [19] known to recognize both Aβ and C99. A similar accumulating intraneuronal FCA18-immunostaining was observed in double transgenic (2xTgAD) mice harboring APP_swe and Tau_P301L but wild-type presenilin1 [24] (Suppl. Figure 1a). Interestingly, in the two strains this intracellular staining (Suppl. Figure 1a) and C99 levels (Suppl. Figure 1c, d) increased similarly with age, while only 3xTgAD mice produced extracellular plaques at late ages that were stained with both FCA18 (Suppl. Figure 1a) and anti-Aβ42 (α-Aβ42) (Suppl. Figure 1b). The detection of very few extracellular Aβ plaques in situ in aged 2xTgAD mice was corroborated by the absence or very low detectable levels of both Aβ40 and Aβ42 measured by ELISA (Suppl. Figure 1e). Overall, these data indicated that C99 accumulation did not correlate with Aβ load and therefore could not be accounted for altered γ-secretase processing alone. In situ immunolabeling with 82E1 revealed that C99 accumulated in intraneuronal organelles positive for the lysosomal enzyme cathepsin B (Fig. 1a) and suggested a link between C99 accumulation and defective lysosomal clearance. In young 3xTgAD mice (3 months of age), the majority of the neurons displayed numerous small-sized cathepsin B puncta (Fig. 1a, arrows), but some neurons also presented enlarged cathepsin B structures which were positive for C99 (Fig. 1a, arrowheads). Similar results were obtained with an antibody directed against the endosomal-lysosomal membrane-associated protein lamp1, and the number and size of these structures were high in both 14-month-old 2AD and 3AD mice (Fig. 1b). The large size of these structures proposed that many of them corresponded to endo- or autolysosomes rather than to lysosomes and would support impaired lysosomal-autophagic-mediated C99 degradation in both 2xTgAD and 3xTgAD mice. In agreement, electron microscopy revealed an increased number of dense autophagic vesicles in brains from both of these mice, as compared to non-transgenic mice (Fig. 1c, red arrowheads). These autophagic vacuoles corresponded to both small and very dark autolysosomes (<500 nm, lower left image) and giant autolysosomes (>1.5 µm, examples in lower middle and right images) containing various compositions of densely packed undigested material and lipid droplets (clear round structures).

The above data suggested a role of lysosomal-autophagic degradation in C99 accumulation and would agree with earlier data proposing a major contribution of this degradation pathway in APP-CTF clearance [1, 8, 54, 57]. To confirm the role of lysosomal enzymes in the fate of C99, we used SH-SY5Y neuroblastoma overexpressing APPswe (known to potentiate β-secretase-mediated C99 and Aβ production [29] and harbored by the βAPP transgene in both 2xTgAD and 3xTgAD mice [37]) (SH-APPswe). Clearly, treatments of SH-APPswe with either the lysosomotrophic weak base NH₄Cl or specific lysosomal protease inhibitors, such as the cathepsin B inhibitors leupeptin and PADK or the cathepsin D inhibitor pepstatin A, all led to large increases in APP-CTFs, including the most abundantly expressed C83, C99 and AICD (Fig. 2a, b). These treatments also led to the accumulation of other higher molecular mass APP-CTFs, notably one about a size of 25 kDa (Fig. 2a, arrows), whose levels were very low in non-treated cells, showing its particularly high sensitivity to lysosomal degradation. Although, we did not determine the exact identity of this fragment, it could possibly correspond to the recently described ƞ-secretase-derived APP-CTF [2, 59]. Immunostaining showed that these APP-CTFs strongly accumulated in lamp1-positive structures (Fig. 2c) and similarly to lamp1, they localized at their membranes (Fig. 2c, arrow). The fact that they were more numerous and much larger in NH₄Cl-treated cells than in control cells suggested that many of them could represent endo- or autolysosomes instead of simple lysosomes (Fig. 2c). If C99 load is genuinely increased by a defect of the enzymatic machinery, one should conversely expect to reduce C99 levels by enhancing lysosomal enzyme expression. Indeed, reduced APP-CTF levels were observed in both lentiviral-induced cathepsin B overexpressing cells (Fig. 2d) and in cells stably expressing this enzyme (Fig. 2e, f), as compared to mock-transfected cells. Furthermore, to demonstrate the ability of cathepsin B to catabolize C99 directly, we designed a bimolecular enzymatic assay in which various amounts of recombinant cathepsin B were incubated with recombinant C99 (C100-flag). This experiment confirmed the direct degradation of C99 by cathepsin B (Fig. 2g). Finally, we used a pharmacological approach to either activate or inhibit autophagy by using the small molecule enhancer of rapamycin, smer-28 [54] or the vacuolar-type H+ ATPase inhibitor, bafilomycin A1, respectively. As expected, APP-CTFs levels were reduced by smer-28 and enhanced by both bafilomycin A1 and the cathepsin B inhibitor PADK (Fig. 2h, i). All together, these findings highlighted a critical role of lysosomal-autophagic function in APP-CTF clearance and clearly supported our hypothesis connecting impaired lysosomal-autophagic degradation and C99 accumulation.

We then questioned whether C99 per se could contribute to lysosomal-autophagic dysfunction. To this end, we took advantage of a pharmacological approach in which we used the γ-secretase inhibitor ELND006 (D6) found to induce enhanced C99 (and C99-derived C83) levels [24]. Young mice were treated daily for 1 month with D6 and analyzed for C99 expression by immunohistochemistry (Fig. 3a, f) or western blot (Fig. 3b, c) as well as for lysosomal function (Fig. 3b, c). As expected, FCA18-associated immunolabeling showed a remarkable increase in lysosomal-associated C99 staining in brains of D6-treated mice (AD-D6) as compared to those of vehicle-treated mice (AD-CT) (Fig. 3a). This pharmacological treatment also drastically enhanced the number of enlarged cathepsin B-positive structures in C99-containing neurons indicating concomitant C99 accumulation and exacerbated lysosomal pathology (Fig. 3f). These observations were confirmed by western blot analysis revealing significant increases in the autophagic marker LC3-II and the autophagic substrate p62/SQSTM1 (Fig. 3b, c), thus indicating autophagic impairment [6]. In wild-type (nonTg) mice, the treatment with the inhibitor also led to increased levels of C83, which however remained very low as compared to those observed in AD-D6 mice (right lane in the western blot) (Fig. 3d, e). In both vehicle and D6-treated nonTg mice, LC3-II was not detected and p62 levels were unchanged (Fig. 3d, e) indicating the absence of effect of D6 on autophagic function in these mice.

In light of these observations, electron microscopy analysis (EM) was performed to confirm the above-described autophagic alterations. The subiculum of young AD vehicle-treated mice (AD-CT) displayed some small dense bodies (Fig. 4a, blue arrows) and various-sized dense lipid-containing autophagic vacuoles (Fig. 4a, red arrows) localized to perikarya, while age-matched vehicle-treated (nonTg-CT) (Suppl. Figure 2a) or D6-treated nonTg (nonTg-D6) neurons (Suppl. Figure 2b) presented very few of such structures. In D6-treated AD animals (AD-D6), EM revealed a strong autophagy-related pathology and intraneuronal damage (Fig. 4b, d–i and Suppl. Figure 2e–j). The number of small dense structures (<0.5 μm, blue arrows “1” in Fig. 4i) was identical in AD-CT and AD-D6 mice, but the number of larger sized autophagic vesicles, many containing lipids (>0.5 μm, red arrows “2” in Fig. 4i), was significantly increased in AD-D6 mice. However, the most striking feature of the autophagy-related pathology was the presence of multiple large vacuoles filled with heterogenous non-digested cargoes and multilamellar bodies (blue ML) (Fig. 4d, e–i and Suppl. Figure 2e–j—red arrowheads), (red arrowheads “3” in Fig. 4i), probably corresponding to immature autophagic vesicles [62]. Moreover, many AD-D6-treated neurons exhibited large electron-lucent areas of the cytoplasm localizing close to autophagic vesicles, suggesting a link between autophagic leakage and erosive destruction (Fig. 4h, and Suppl. Figure 2f, j—blue star). D6-treated neurons also presented a high number of damaged mitochondria (see examples in Fig. 4d, g and Suppl. Figure 2g, i—black arrows) and reduced synaptic contacts within the neuropil (Fig. 4d, yellow arrows), as compared to control mice (Fig. 4c—yellow arrows).

We next compared the effect of D6-treatment on APP-CTF accumulation in 2xTgAD and 3xTgAD mice. This treatment led to same increases in C99 (and C99-derived C83) (Suppl. Figure 3a, b) and in punctiform staining (Suppl. Figure 3d) in the two strains, whereas the effect of D6 on Aβ load was detected only in 3xTgAD mice displaying enough Aβ do detect a decrease (Suppl. Figure 3c). These data fully agreed with our proposal of a dual effect of D6 on APP-CTF accumulation, one linked to its canonical inhibitory effect of γ-secretase and one to its inhibitory effect on lysosomal degradation. In agreement with defective lysosomal degradation, both D6-treated 2xTgAD and 3xTgAD animals displayed enhanced, rather than lowered levels of the other γ-secretase cleavage-derived metabolite AICD (Suppl. Figure 3a, b) another lysosomal substrate (Fig. 2a) (see also [56]).

To more precisely dissect the C99-mediated molecular mechanisms, we studied the effect of D6 treatment on lysosomal function in SH-APPswe cells. Firstly as expected, D6 triggered a drastic decrease in extracellular Aβ (not shown) and concomitant increase in C99 and C83 (Fig. 5a). Immunocytochemical analysis showed that these APP-CTFs were particularly abundant in lamp1-positive structures (Fig. 5b), which were enlarged and more numerous than those observed in control cells, similarly to cells treated with lysosomal inhibitors (Fig. 2c). These data were confirmed by Lysotracker-red, which labeled many and mostly large-sized puncta in D6-treated cells as compared to vehicle-treated cells (Fig. 5c). No increase in Lysotracker-red staining was seen in D6-treated SH-mock cells (i.e., cells only expressing endogenous levels of wild-type APP), suggesting a strong link between lysosomal dysfunction and APP-CTF accumulation. In addition, four independent lines of data supported our hypothesis, when we compared the effects of D6 and DAPT (another well-known γ-secretase inhibitor [58]) in SH-APPswe cells to SH-mock cells. Firstly, D6 and DAPT reduced in vitro cathepsin B activity (Fig. 5d) and lowered the levels of mature cathepsin B (Fig. 5e, f) in SH-APPswe but not in SH-mock cells (Fig. 5e), while NH₄Cl strongly inhibited cathepsin B in both cell types (Fig. 5d–f). Moreover, in SH-APPswe cells both γ-secretase inhibitors led to the accumulation of a low-molecular weight (12 kDa) cathepsin B immunopositive product, probably corresponding to a degradation product of this enzyme (CatBdg) and supporting defective lysosomal clearance (Fig. 5e, f). Secondly, D6 and DAPT increased the levels of the autophagic markers LC3-II and p62 in only APPswe cells and the levels of these markers correlated with those of APP-CTFs (Fig. 5e, f). Thirdly, D6 treatment significantly increased LC3-positive vesicular profiles in SH-APPswe cells, whereas little if any LC3-GFP puncta were found in D6-treated SH-mock cells (Fig. 5g). Fourth, electron microscopy revealed the presence of large (0.5–2 μm) abnormal dense autolysosomal structures in D6-treated SH-APPswe cells, which structurally looked similar to those observed in leupeptin-treated cells (Fig. 5h).

In order to definitely demonstrate that C99 per se is sufficient to induce lysosomal pathology, we used an AAV-mediated strategy to express C99 under the transcriptional control of the neuron-specific synapsin I in wild-type animals. AAV-C99, AAV-GFP- or AAV-empty was injected into the ventricle of newborn mice and C99 expression was analyzed at 2 months post-AAV delivery. Western blot revealed a significant and specific increase of C99 in AAV-C99 injected brains (Fig. 6a). As compared to AAV-GFP mice, in which GFP was observed in many neurons and throughout the brain (Fig. 6b, left panels), immunohistochemical analysis of C99 in AAV-C99 mice using α-APPct, showed a more restricted expression that was particularly high in the hippocampus and neocortex (Fig. 6b, right panels). Within the neurons of these areas, the staining was intraneuronal and membrane-associated giving rise to a distinct immunostaining when compared to that of the cytoplasmic protein GFP (Fig. 6b). To discriminate between overexpressed C99 and endogenous APP and APP-CTFs that are also detected with the α-APPct antibody (recognizing both human and mouse APP, see supplementary table), we compared the immunostainings in AAV-C99 and AAV-empty mice. As expected, α-APPct detected membranous staining in many neurons and throughout the brain in AAV-empty mice; however, the intensity of this staining was much lower than that obtained in the AAV-C99 mice, indicating that the largest part of this staining corresponded to exogenously expressed C99. To determine the exact intraneuronal staining of C99, we performed immunocytochemical staining of C99 in cultured transfected cells that revealed that C99 was localized almost exclusively to the Golgi apparatus (Suppl. Figure 4a), whereas plasma membrane-associated C99 was also evident in few cells (Suppl. Figure 4b). As observed in SH-APPswe cells, the treatment with NH₄Cl or D6 both led to the redistribution of APP-CTFs into EAL-associated structures and very little or no co-localization with GM130 was seen in these cells (Suppl. Figure 4c, d). In AAV-C99 brains, α-APPct did not detect EAL-associated staining, but interestingly when we used FCA18 or 82E1, the staining was punctiform and looked similar to that observed in the 3xTgAD mice. This labeling was present in the same brain regions labeled with α-APPct, but was localized to a much more restricted number of neurons and most strongly in the subiculum (Fig. 6d) and in some layers of the cortex. Interestingly, the other N-terminal-directed antibodies 6E10 and 4G8 recognized both the membrane-associated labeling and puncta (Fig. 6d, e—arrow and arrowhead, respectively) suggesting the presence of two distinct C99 species. Since, the lysosomal-associated staining required formic acid retrieval and since C99 has been shown to have a high tendency to self-aggregate and form amyloid-like fibrils in vitro [30, 60], we hypothesized that the C99 accumulating in lysosomes could correspond to aggregated or misfolded forms that would not be detected with C-terminal-directed antibodies because of aggregation-dependent structural or conformational changes of the C-terminal moiety. In agreement with this hypothesis, we found that the punctiform staining was also detected with the aggregate-specific antibody NU1 [23] (Fig. 6d). As observed in the 3xTgAD mice, these aggregates were localized to enlarged cathepsin B- and lamp1-positive structures (Fig. 6f, g) and the number and size of such enlarged structures was enhanced in the AAV-C99 mice, as compared to AAV-empty mice (Fig. 6h).

ELISA showed that C99-expressing mice do produce Aβ, corresponding mainly to Aβ40 but only very low levels of the aggregation-prone Aβ42 (Fig. 6c). Therefore, one could not exclude the possibility, that in this model, the punctiform staining could correspond to aggregated Aβ, instead of C99. Thus, to rule out this possibility, the C99-expressing mice were treated with D6. As described for 3xTgAD mice, γ-secretase inhibition enhanced the levels of C99 and C99-derived C83 [16] (Fig. 7a, but see also Fig. 8b) and strongly reduced Aβ levels (Fig. 7b). Immunohistochemical analysis using NU1 showed an increased rather than decreased punctiform immunostaining in D6-treated mice, clearly indicating that this staining should be accounted for C99 and not Aβ (Fig. 7c). Similar enhanced punctiform staining was observed with other N-terminal directed antibodies, 82E1 (Fig. 8a), FCA18, 6E10 (not shown) and 4G8 (Fig. 7d). In D6-treated mice, in areas expressing particular high punctiform staining, NU1 and 4G8 also gave rise to an extracellular diffuse staining surrounding the neurons (Fig. 7c, d, respectively), probably corresponding to the release of intraneuronal material from dying neurons (see also below). We also used α-APPct to analyze the effect of D6-treatment (Fig. 7c, e) and found that, within the subiculum, D6 also led to a strong APP-CTF associated staining localizing to synaptic regions (note the high overlap of α-APPct and the presynaptic marker synaptophysin in D6 treated AAV-C99 mice (Fig. 7e). The low intensity of α-APPct labeling in D6-treated AAV-empty mice showed that these synaptic localized APP-CTFs were mostly linked to viral-mediated C99. These observations confirmed the presence of two species of C99 (APP-CTFs), one non-aggregated and recognized by α-APPct and one aggregated and detected with N-terminal- and aggregate-specific antibodies. D6 treatment led to increased levels of both non-aggregated APP-CTFs localizing in presynaptic regions and of aggregated APP-CTFs accumulating in lysosomal-derived structures. It is noteworthy that similar D6-induced increases in both synaptophysin-positive APP-CTFs (Suppl. Figure 5a, d–e) and NU1-labeled aggregated APP-CTFs (Suppl. Figure 5b) was found in 3xTgAD mice, thus ruling out the possibility of artificial routing of C99 in the C99-expressing mice model. Moreover, western blot analysis confirmed the accumulation of C99 and C83 in both microsome- and synaptosome-enriched fractions of hippocampi from D6 treated 3xTgAD mice (suppl. Figure 5c).

Furthermore, and more surprisingly, in the C99-expressing mouse, D6 not only led to an enhanced staining in the cortex and subiculum, the two main regions expressing C99, but it also induced a strong immunostaining in various other brain regions, in which very low or no C99 staining was seen in vehicle-treated AAV-C99 mice (Fig. 7f), but in which GFP expression was high in AAV-GFP mice (Fig. 6b, left panels). Again, the labeling obtained with N-terminal antibodies was punctiform and in some regions also diffuse and extracellular. Taken together, these findings clearly demonstrated that this labeling should be ascribed to C99 and not to Aβ. Ηowever, they also suggested that these regions display either high γ-secretase activity or alternative D6-sensitive C99 degradation, explaining the absence of C99 labeling in vehicle-treated C99-expressing mice.

As described above, C99 accumulated in enlarged cathepsin B- and lamp1-positive structures and particularly within the subiculum in C99-expressing mice. In brains from D6-treated AAV-C99 mice, the number of neurons displaying such enlarged structures was increased rather than decreased (Fig. 8a). Furthermore, these brains displayed elevated levels of the autophagic marker LC3-II (Fig. 8b, c), while this was not the case in those from D6-treated AAV-GFP mice (Fig. 8d), again indicating that these lysosomal alterations were linked to C99 expression. Moreover, as described above, in some regions in which lysosomal-associated C99 was abundant, 4G8 also unraveled an important and diffuse extracellular staining likely corresponding to the release of intracellular material from dying neurons (Fig. 8e, f, but see also Fig. 7c, d). In agreement with necrotic cell death, nuclear staining with DAPI or Cresyl violet revealed the presence of many pycnotic nuclei (Fig. 8e, f). The same regions also displayed an increased number of both Iba-1-positive microglia (Fig. 8g) and GFAP-positive astrocytes (Fig. 8h). Iba-1 labeling was enhanced by the sole expression of C99 (compare AAV-C99 and AAV-empty) and further enhanced by D6 treatment, indicating concomitant C99 accumulation and inflammatory responses (Fig. 8g). Importantly, microglial and astrocytic activation (Fig. 8h) were not induced by D6 in AAV-empty mice. Finally, we assessed the effects of C99 expression on synaptic function by measuring long-term potentiation (LTP) at the CA1 pyramidal neurons to subicular neurons in AAV-empty or AAV-C99 vehicle or D6-treated mice (Fig. 8i–k). Whereas no differences were seen in basal synaptic transmission between AAV-empty, AAV-C99 vehicle or D6-treated mice (Fig. 8i), only the AAV-empty mice displayed a robust induction in LTP (Fig. 8j, k). D6 did not worsen LTP alterations, but this could be due to the drastic floor effect of C99 over-expression alone. However, the fact that D6 did not reverse synaptic dysfunction indicated that it was not caused by Aβ, but should be directly linked to C99.