NSCs were isolated from mouse embryos (E14.5) as previously described (Zhang et al., 2015; Jiao et al., 2016; Ghazale et al., 2018). Epidermal growth factor, fibroblast growth factor, B27, and N2 were added to Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 as the NSC culture medium. NSCs from the P2 generation were plated at a density of 9 × 10⁷ cells/well. NSCs were transfected with Gpc4 siRNA adenovirus (AdV), according to the manufacturer’s instructions (GeneChem, Shanghai, China). NSCs were treated with AdV for 24 h and then the medium with AdV was changed to complete medium without AdV. After 72 h of infection, Aβ was added. Internalization was detected after 6 h, and toxic effects were detected after 12 h.

Cell viability was evaluated using the Cell Counting Kit-8 (CCK-8, Sigma-Aldrich, St. Louis, MO, USA) assay. Cells were plated in 96-well plates at 20,000 cells/well density, 24 h before the experiments. The cells were then treated with different concentrations of Aβ (0, 200 nM, 1 μM, 2 μM, 5 μM, 10 μM, and 20 μM) for 12 h. Next, 20 μl of CCK-8 were added to each well and the cells were incubated for 2 h at 37°C. Absorbance was measured at 490 nm using a microplate spectrophotometer (BioTek, Winooski, VT, USA). Triplicate parallel wells were examined in each experiment, and data were collected from three independent experiments. The results are presented as a percentage of the absorbance compared to the control group.

To investigate the toxic effect of Aβ on NSC apoptosis mechanism, a terminal deoxynucleotidyl transferase dUPT nick end labeling (TUNEL) assay was performed according to the manufacturer’s instructions (Roche Diagnostics, Basel, Switzerland). Cells were fixed on cover slides in 4% paraformaldehyde for 15 min. The cells were then permeabilized in 0.3% Triton X-100 in phosphate-buffered saline for 5 min on ice. Cells were incubated in the TUNEL reaction mixture for 1 h at 37°C. Subsequently, the cells were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 1:3,000, Sigma-Aldrich) before mounting. Images were taken within 2 h. Nine random fields were counted for each group with a 40× objective. The ratio was calculated as the percentage of TUNEL-positive cells compared to that of DAPI-positive cells.

Flow cytometry analysis was used for the apoptosis tests in this study. The propidium iodide (PI)/Annexin V apoptosis detection kit (BD Biosciences, Franklin Lakes, NJ, USA) was used according to the manufacturer’s instructions. Cells were dissociated, washed in cold phosphate-buffered saline, and resuspended in 1× binding buffer to obtain a concentration of 1 × 10⁶ cells/ml. The cells were incubated with PI/Annexin V for 15 min in the dark. Subsequently, the cell solution was analyzed using FACSCalibur (BD Biosciences) for PI/Annexin V. The percentage of apoptotic cells in this experiment was calculated as the ratio of Annexin V-positive and PI-negative cells from the lower right quarter.

For immunohistochemistry performed in APP/PS1 mice, mouse brains were fixed with 4% paraformaldehyde and frozen coronal slices (30 μm) were obtained using a frozen section machine (Leica, Wetzlar, Germany). NSCs were fixed in 4% paraformaldehyde. The sections were treated with 0.3% Triton for 30 min. The cells were then blocked with 5% bovine serum albumin for 30 min. The sections were incubated with primary antibodies overnight at 4°C. The cells were then incubated with the secondary antibodies, goat-anti-rabbit-IgG-488 and goat-anti-mouse-IgG-594 (Invitrogen, Carlsbad, CA, USA), for 2 h at room temperature. Slides were cover-slipped using a mounting medium containing DAPI (blue). The primary antibodies used in this experiment were: mouse polyclonal anti-nestin antibody (1:200, Novus, St. Louis, MO, USA), rabbit polyclonal anti-Gpc4 antibody (1:100, Abcam, Cambridge, UK), rabbit polyclonal anti-sox2 antibody (1:200, Abcam), rabbit polyclonal anti-Ki67 antibody (1:200, Abcam, Cambridge, UK), mouse polyclonal anti-sox2 antibody (1:200, Abcam, Cambridge, UK), mouse polyclonal anti-LRP1 antibody (1:1,000, Abcam, Cambridge, UK), goat-anti-rabbit-IgG-488, goat-anti-mouse-IgG-594 (Invitrogen, Carlsbad, CA, USA), and DAPI (1:3,000, Sigma-Aldrich, USA). For 3,3’-diaminobenzidine staining, we followed the manufacturer’s instructions using a secondary antibody staining kit (ZSGB-BIO Inc., Beijing, China).

NSCs were homogenized in radioimmunoprecipitation assay buffer supplemented with a protease inhibitor and centrifuged (15,000× g) for 30 min at 4°C. The protein concentration was quantified using the BCA assay (Beyotime, Nanjing, China). Protein (10 μg) was added to each sample, electrophoresed on 10% sodium dodecyl sulfate polyacrylamide gels, and transferred to a polyvinylidene fluoride membrane (Millipore, Burlington, MA, USA). The membranes were then blocked in 10% skimmed milk for 2 h and incubated with the following primary antibodies: rabbit anti-Gpc4 (1:2,000), rabbit anti-Cyt c (1:500), rabbit anti-Bax (1:500), rabbit anti-Bcl 2 (1:500), rabbit anti-GAPDH (1:10,000), or β-actin (1:10,000) (Abcam, Cambridge, UK). The primary antibody was dissolved in 10% skimmed milk, and the membrane was incubated at 4°C overnight. The membrane was then washed and incubated with horseradish peroxidase-conjugated secondary antibody. Membranes were developed using enhanced chemiluminescence and exposed to photographic films. All results were obtained from three independent experiments. The results were analyzed using ImageJ software (Yang et al., 2015).

siGpc4, overexpressed Gpc4 (overGpc4), and siLRP1 were coated with AdV (4 × 10¹² μg/ml) to infect NSC cell cultures. The overGpc4 sequence was obtained from Gene_ID:14735, GenBank: NM_008150. The forward siGpc4 sequence was 5-GGAUGGCAGU GGAUGACUUTT-3 and the reverse was 5-AAGU CAUCCACUGCCAUCCTT-3. The forward and reverse siLRP1 sequences were 5-GCCCAUUGGAUGAGUUUCATT-3 and 5-UGAAACUCAUCCAAUGGGCTT-3, respectively. The lentivirus vector for Gpc4 overexpression was GV367 Ubi-MCS-SV40-EGFP-IRES-pruromycin (GENE, Shanghai, China), and the lentivirus vector for siGpc4 was hU6-MCS-Ubiquitin-FGFP-IRES-puromycin (GENE, Shanghai, China). AdV was plated into 6-well plates at a density of 2 × 10⁶ cells/ml; 10 μl of 2 × 10¹² μg/ml AdV was added to each well in NSC complete medium (GENE, Shanghai, China). Twenty-four hours after infection, the medium was replaced with a complete medium without AdV. The cells were used for the experiment 72 h after infection.

10 μl of 2 × 10¹² μg/ml AdV (overGpc4 and si Gpc4) was plated into 6-well plates at a density of 2 × 10⁶ cells/ml (GENE, Shanghai, China). Twenty-four hours after infection, the medium was replaced with a complete medium without AdV. Aβ was added into each 6-well plates cells at a working concentration of 200 nM or 2 μM, 72 h after Adv infection. NSC samples were collected 12 h after Aβ treatment. Mouse Aβ enzyme-linked immunosorbent assay (ELISA) Kit (R&D Systems, Minneapolis, MN, USA; Ma et al., 2016) and quantification experiments were conducted according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN, USA). Protein samples in each group were detected three times in the 96T ELISA kit. Aβ concentrations were calculated by the standard curve. This experiment was repeated three times.

The APP/PS1 mice used in this study were double transgenic mice bred by crossing two lines of commercial simple transgenic mice: APPSWE (Tg2576 on a B6J background) and PS1dE9 on a C57BL6SJL background. Acute crossing of these two lines produces an accelerated mouse model of AD on a B6J/B6SJL background combining cognitive and amyloid pathologies starting at 4 months old. Genotypes were confirmed by a polymerase chain reaction from a tail biopsy. All wild type mice in this study are of B6J/B6SJL littermates. All APP/PS1 mice were APP-and PS1 positive. The animals were obtained by Yang Wei-Na from Xi’an Jiaotong University, housed in an animal facility with two mice per cage in temperature- (22 ± 1°C) and humidity-controlled (50 ± 10%) conditions under a 12 h light/dark cycle. The mice had access to food and water ad libitum for 4 months before being used in the experiment.

Gpc4 siRNA lentivirus was injected into each mouse’s brain via stereotaxic surgery under deep isoflurane anesthesia (n = 6). Gpc4 siRNA lentivirus was injected using stereotaxic coordinates and a microsyringe. Virus was injected into the lateral ventricles and the CA1 hippocampal subfield (rostrocaudal: 1.0 mm, mediolateral: 1.30 mm, dorsoventral: 1.8 mm vs. Bregma) of both sides (rostrocaudal: 1.80 mm, mediolateral: 1.00 mm, dorsoventral: 3.10 mm vs. Bregma). The injection speed was 0.2 μl/min, and the total volume/side was 1 μl, 2 × 10⁸/μl. The control virus was also injected into APP/PS1 mice (n = 5). After surgery, the animals were given 2 months for recovery and lentivirus expression.

After the brain was perfusion-fixed and sectioned at 30 μm on a freezing microtome, the selected series were mounted onto gelatine-coated glass slides and air-dried. The tissue was hydrated by passing it through a series of ethyl alcohol solutions (100%, 95%, 80%, and 70%, 1–2 min in each) before staining with 1% Thioflavin-S for 30 min (Oakley et al., 2006). The slides were subsequently rinsed in distilled water and transferred through 80% and 95%, and two sets of 100%, ethanol for 30 s before cover-slipping with mounting media and storage under refrigeration. Thioflavin-S staining was viewed under a green fluorescence microscope (480/525 nm).

Immunoprecipitation (IP; Proteintech, Chicago, IL, USA) was performed as previously described (Ma et al., 2018). Eluted proteins that bind to any protein complex containing Gpc4 were separated via sodium dodecyl-sulfate polyacrylamide gel electrophoresis. Gpc4, LRP1, and Aβ were detected using western blotting. Briefly, APP/PS1 mouse brain tissue supernatants containing 4 mg protein (in approximately 1 ml lysate) were pre-cleared for 1–4 h at 4°C with 100 μl of 80–100 protein G. Precleared supernatants were incubated overnight at 4°C with the desired IP antibody (MOAB2, 1:200; Gpc4, 1:100; goat anti-rabbit IgG, 1:100) and 200 μl of protein G beads. IP complexes were washed five times with wash buffer and eluted twice with 40 μl elution buffer. Protein complexes (80 μl) were obtained and used for western blotting.

NSCs were stained with JC-1 to determine the mitochondrial membrane potential (MMP). JC-1 is a membrane-permeable lipophilic dye that exists as JC-1 aggregates in the mitochondrial matrix (red) or as monomers in the cytoplasm (green). During mitochondrial depolarization, the red JC-1 aggregates form green monomers due to a change in proton motive force (ΔΨ). Thus, depolarization can be measured as an increase in the green fluorescent/red fluorescent intensity ratio. The JC-1 assay was performed as follows. Solutions containing 5 μg/ml JC-1 were added for AdV and Aβ treated NSCs for 30 min. The stained cells were rinsed twice with the culture medium. After the fresh media was added, the cells were analyzed under a green fluorescence microscope (480/525 nm).

All data were analyzed with GraphPad Prism version 6.0 (GraphPad Software, San Diego, CA, USA). For normal distribution data, one-way analysis of variance (ANOVA) followed with Tukey’s multiple comparisons test were performed. P-value <0.05 was regarded as significant. t-test was used to compare between two groups. All in vitro experiments have been independently repeated at least three times. For in vivo experiments, in each group at least five mice were used for statistical analysis.

Neural stem cells derived from the cortex of embryonic 14.5 (E14.5) mice were observed. Approximately 5 days after primary culture (P0), the isolated NSCs were passed for the next generation (P1) and then for P2 (Figure 1A). We observed that neural spheres from P0-P2 stably expressed Gpc4 and nestin (Figure 1B). After NSC differentiation, we also detected high colocalization of Gpc4 and both neurons and astrocytes (Figures 1C,D).

This experiment aimed to determine the toxic dose of Aβ in NSCs. We found that 200 nM Aβ induced cell viability of NSCs, but that 1–20 μM Aβ was toxic to NSCs (Figure 2A). Therefore, 200 nM Aβ and 2 μM Aβ were used in this study. TUNEL staining showed that the number of apoptotic cells increased in the group treated with 2 μM Aβ compared to the 0 μM Aβ group (Figures 2B,C). We then used flow cytometry to detect PI/AnnexinV stained cells. The percentage of living cells in the group treated with 200 nM Aβ was significantly higher than that in the control group (0 μM Aβ). The percentage of living cells in the 2 μM Aβ group was significantly lower than that in the control group (0 μM Aβ). The apoptotic and living cell percentage in different concentrations of Aβ on NSCs is contrasting (Figures 2D,E). We used ELISA to detect the intracellular Aβ level. NSCs in the 200 nM group absorbed more Aβ than those in the control group. Cells treated with 2 μM Aβ showed a higher uptake of Aβ compared to the other groups (Figure 2F).

To identify the role of Gpc4 in NSCs in AD, we examined Gpc4 expression patterns in NSCs. Gpc4 (green) colocalized with cells positive for Ki67 or nestin (red), which proves the expression of Gpc4 in NSCs. Gpc4 was widely expressed in NSCs. Double-positive Gpc4/Ki67 represents NSCs undergoing cell division (Figure 3A, yellow arrow). We also observed some Gpc4-negative Ki67-positive NSCs (red arrow) and Gpc4-positive Ki67-negative NSCs (green arrow). Furthermore, we assessed whether the expression of Gpc4 was regulated by Aβ treatment. NSCs were incubated with 200 nM and 2 μM Aβ for 12 h. Our results from immunofluorescence staining and western blotting illustrated that Gpc4 expression increased with Aβ concentration (Figures 3B–D).

Co-IP was performed to detect the relationship between Gpc4 and Aβ. We found that Gpc4 forms a protein complex with Aβ in the APP/PS1 mouse brain (Figure 4A). To further determine the role of Gpc4 in Aβ internalization, AdV and lentivirus were used to either overexpress or attenuate Gpc4 expression in NSCs and APP/PS1 mice, respectively. The AdV transfection ratio was 72% for siGpc4-AdV and 70% for overGpc4-AdV (Figure 4B). The timing for the experimental procedure of Gpc4 and LRP1 AdV and Aβ treatment is laid out in Figure 4C. Furthermore, cy3-Aβ (200 nM and 2 μM) was added to siGpc4-AdV-, Gpc4-AdV-, or control-AdV-infected NSCs for 6 h. Subsequently, the Aβ particles per cell were counted. Attenuating Gpc4 decreased the number of Aβ particles in the 200 nM and 2 μM groups. Overexpression of Gpc4 was able to upregulate the Aβ particle number in the 200 nM group but had no significant effect in the 2 μM group (Figures 4D,E).

Next, cells were monitored to detect whether Gpc4 induced Aβ production. No significant differences were observed between the siGpc4 and overGpc4 adenovirus groups (Figure 4F). overGpc4 had no significant effect on 2 μM Aβ internalization, but siGpc4 significantly attenuated the internalized Aβ level (Figure 4G). In the 200 nM Aβ-treated group, overGpc4-AdV increased NSC Aβ internalization, but siGpc4-AdV had no significant effect on blocking Aβ internalization (Figure 4H). siGpc4/overGpc4 lentivirus were injected into the hippocampus of 4-month old (APP/PS1) mice brains, and 2 months after injection, mice were sacrificed for S-plaque detection. Aβ plaques were stained using Thioflavin staining. Plaque numbers in the siGpc4 and overGpc4 groups did not change at this stage in APP/PS1 mice (Figures 4I,J). These results suggest an important dose-dependent effect of Gpc4 on Aβ internalization in NSCs at the early stage of AD.

LRP1 is an Aβ receptor that plays a role in Gpc4-induced Aβ internalization. We first identified the expression of LRP1 in NSCs by double staining NSCs with Ki67 and LRP1. LRP1 colocalized with Ki67-positive cells (Figure 5A). Double-positive LRP1/Ki67 (yellow arrow), LRP1-negative Ki67-positive (red arrow), and LRP1-positive Ki67-negative NSCs (green arrow) were all observed (Figure 5A). Moreover, the transfection ratio of siLRP1 was 81% (Figure 5B. The effect of LRP1 on Aβ internalization was analyzed by double staining of LRP1 with the Aβ antibody MOAB2 in AdV control and AdV-siLRP1-treated groups. siLRP1 remarkably decreased Aβ levels in NSCs (Figures 5C,D), suggesting LRP1 is responsible for Aβ transportation in NSCs. However, the relationship between LRP1 and Gpc4 remains unclear.

To more precisely define the relationship between Gpc4 and LRP1, co-IP was performed on the cell lysis of APP/PS1 mice. Gpc4 was able to bind to LRP1, forming a protein complex in the brains of AD mice (Figures 6A,B). These observations suggest that LRP1 plays a role in Gpc4-induced Aβ internalization.

We next aimed to reveal the action of Gpc4 and LRP1 in Aβ internalization. NSCs were transfected with control, siLRP1, siLRP1+siGpc4, and siLRP1+overGpc4 AdV for 3 days, and then Aβ was added for 6 h before cells were harvested for ELISA. siLRP1 remarkably decreased Aβ uptake by NSCs. In the siLRP1 and siGpc4 groups, the amount of internalized Aβ was much less compared to the siLRP1-treated group. Furthermore, in the siLRP1+overGpc4-treated group, the internalized Aβ level was increased compared to the siLRP1+siGpc4 group (Figure 6C).

Control, siLRP1, overGpc4, and overGpc4+siLRP1 AdV were transfected into NSCs for 3 days, and then Aβ was added for 12 h. The JC-1 kit was used to detect the MMP in NSCs. JC-1 monomers (green) and aggregates (red) were compared and normalized to those of the control group. In 2 μM Aβ experiments, JC-1, TUNEL, and western blotting were used to determine whether Gpc4/LRP1 regulates the toxic effect of Aβ on NSCs. siLRP1 significantly decreased the ratio of JC-1 monomers/aggregates compared to the control group. Moreover, overGpc4 increased MMP damage compared to siLRP1. overGpc4+siLRP1 significantly increased the MMP compared to siLRP1; however, the MMP damage in the overGpc4+siLRP1 group was lower than that in the overGpc4 group (Figures 7A,B). Moreover, the level of Cytc was decreased in the siLRP1 and siLRP1+siGpc4 groups, and higher in the siLRP1+overGpc4 group (Figure 7C). Furthermore, TUNEL staining was used to detect whether Gpc4 and LRP1 regulate Aβ-induced cell death. Compared to the control group, siLRP1 attenuated NSC apoptosis while overGpc4 increased it. siLRP1 decreased the effect caused by overGpc4 (Figures 7D,E). The expression of Bax was also detected using western blotting, which is consistent with the results of TUNEL staining (Figures 7D–F).

In 200 nM Aβ experiments, JC-1 and TUNEL assays were used to determine whether Gpc4 induces toxic effects by enriching the nontoxic effect of Aβ on NSCs. Initially, 200 nM Aβ had no toxic effect on NSCs, but overGpc4 in 200 nM Aβ-treated NSCs increased MMP damage (Figures 8A,B). Furthermore, TUNEL staining was used to detect whether Gpc4/LRP1 regulated Aβ-induced cell death. No significant changes were observed among the four groups (Figures 8C,D).