Full-length amyloid precursor protein regulates lipoprotein metabolism and amyloid-β clearance in human astrocytes

Mounting evidence suggests that alterations in cholesterol homeostasis are involved in Alzheimer's disease (AD) pathogenesis. Amyloid precursor protein (APP) or multiple fragments generated by proteolytic processing of APP have previously been implicated in the regulation of cholesterol metabolism. However, the physiological function of APP in regulating lipoprotein homeostasis in astrocytes, which are responsible for de novo cholesterol biosynthesis and regulation in the brain, remains unclear. To address this, here we used CRISPR/Cas9 genome editing to generate isogenic APP-knockout (KO) human induced pluripotent stem cells (hiPSCs) and differentiated them into human astrocytes. We found that APP-KO astrocytes have reduced cholesterol and elevated levels of sterol regulatory element-binding protein (SREBP) target gene transcripts and proteins, which were both downstream consequences of reduced lipoprotein endocytosis. To elucidate which APP fragments regulate cholesterol homeostasis and to examine whether familial AD mutations in APP affect lipoprotein metabolism, we analyzed an isogenic allelic series harboring the APP Swedish and APP V717F variants. Only astrocytes homozygous for the APP Swedish (APPSwe/Swe) mutation, which had reduced full-length APP (FL APP) due to increased β-secretase cleavage, recapitulated the APP-KO phenotypes. Astrocytic internalization of β-amyloid (Aβ), another ligand for low-density lipoprotein (LDL) receptors, was also impaired in APP-KO and APPSwe/Swe astrocytes. Finally, impairing cleavage of FL APP through β-secretase inhibition in APPSwe/Swe astrocytes reversed the LDL and Aβ endocytosis defects. In conclusion, FL APP is involved in the endocytosis of LDL receptor ligands and is required for proper cholesterol homeostasis and Aβ clearance in human astrocytes.

Alzheimer's disease (AD) 2 is a progressive and irreversible neurodegenerative disease that is the most common form of dementia in the elderly (1). Genetically, AD can be subdivided into two subgroups: early onset, familial Alzheimer's disease (FAD) and late onset, sporadic Alzheimer's disease (SAD) (2). Although FAD can be attributed to rare and highly penetrant mutations in amyloid precursor protein (APP), presenilin-1 or presenilin-2, the precise etiology of SAD is unknown (3). Recently, several lines of evidence have implicated cholesterol metabolism as a common biological pathway involved in FAD and SAD. The ⑀4 allele of APOE, the major cholesterol carrier of the brain, is the strongest known genetic risk factor in AD (4,5). Genome-wide association studies have identified additional mutations in lipid metabolism-related proteins like APOJ/ Clusterin and ABCA7 as other highly-associated risk factors (6). Finally, retrospective studies, although controversial, have suggested that use of the cholesterol-lowering statins lowers the risk for AD (7)(8)(9).
APP is a ubiquitously expressed, single-pass, type I transmembrane protein that is thought to have multiple putative biological functions (10). The functional heterogeneity of APP may stem from its multiple proteolytic fragments, which are generated by two major pathways: nonamyloidogenic cleavage via sequential proteolysis by ␣and then ␥-secretase or amyloidogenic cleavage by ␤and then ␥-secretase. APP or various proteolytic fragments of APP have recently been implicated in the control of brain cholesterol metabolism via regulation of LDL receptor-family mRNA and protein (11)(12)(13)(14)(15). Normally, low intracellular cholesterol levels induce cholesterol biosynthesis via intricate transcriptional and posttranslational mechanisms involving increased proteolytic processing of SREBPs, which result in up-regulated cholesterol biosynthesis enzymes and increased internalization of extracellular cholesterol via LDL receptors. In contrast, high-intracellular cholesterol levels inhibit the proteolytic processing of SREBP and thus turn off cholesterol biosynthesis and uptake (16). Regulation of LDL receptor mRNA and protein is also relevant to mechanisms of A␤ clearance in the brain because LDL receptors like low-density lipoprotein receptor (LDLR) and lipoprotein receptor-related protein 1 (LRP1) mediate the internalization of A␤ by binding to A␤ directly or via apoE (17)(18)(19)(20)(21).
Despite our understanding of the sensitive and complex mechanisms regulating intracellular cholesterol, there is little consensus as to how APP regulates this pathway in the brain. Multiple studies on these subjects have yielded mixed results, likely due to the use of nonendogenous APP levels, the study of nonneural cell types, or the use of whole brain tissue, which masks the unique phenotypes of individual cell types. To begin to address these issues, we utilized CRISPR/Cas9 genome editing to generate an isogenic series of APP-KO and FAD mutant hiPSCs. We further differentiated these hiPSCs into astrocytes, the cell type primarily responsible for the brain's de novo cholesterol synthesis and regulation. Using APP-KO, APP Swedish, and APP V717F astrocytes, we identify a role for FL APP in the uptake of LDL receptor ligands and demonstrate that proper levels of FL APP in human astrocytes are essential for lipoprotein regulation and A␤ clearance. Our data shed light onto the elusive function of FL APP, establish a linkage between APP and biological pathways implicated in SAD, and finally highlight the utility of using hiPSC technology to study the physiological function of endogenous proteins in specific cell types.

Generation of isogenic APP-knockout hiPSCs using CRISPR/Cas9
To study the role of endogenous human amyloid precursor protein (APP) in regulating astrocytic cholesterol metabolism, we utilized CRISPR/Cas9 genome editing to knock out (KO) APP in hiPSC (Fig. 1). To induce gene disruption, we utilized a guide RNA targeting exon 16 of APP. In both APP-KO clones used in our experiments, the CRISPR/Cas9 -guide RNA complexes generated unique indels at the predicted cut site in each allele. All of these indels were predicted to generate premature stop codons in either exon 16 or exon 17 of APP ( Fig. 2A). For analysis, we compared these APP-KO hiPSCs to WT, unedited subclones of the original hiPSC line that also underwent the genome editing process but were not modified.
To test whether premature stop codon formation had induced nonsense-mediated decay of APP mRNA and thus loss of APP protein, we first examined APP transcript using primer sets targeting regions upstream and downstream of the CRISPR/Cas9 cut site. There was little detectable APP mRNA in APP-KO neural precursor cells (NPCs), neurons, and astrocytes (Fig. 2, B-D). Consistent with these results, two N-terminal full-length APP (FL APP) antibodies, which recognized epitopes either upstream (22C11) or downstream (6E10) of the CRISPR/Cas9 cut site, and one C-terminal APP antibody (APP CTF) failed to detect any APP protein in APP-KO NPCs (Figs. 2E and Fig. S1). Furthermore, the medium from APP-KO neurons contained no detectable cleaved APP fragments like A␤ or the secreted soluble APP (sAPP) fragments (data not shown).

Astrocytic APP controls LDL receptor function
Together, these data suggested that CRISPR/Cas9-induced nonsense-mediated decay of APP transcripts prevented translation of APP protein in APP-KO NPCs, neurons, and astrocytes.
We also examined the expression of APP family members, amyloid precursor-like protein 1 and 2 (APLP1 and APLP2), which have been shown to exert some functional redundancy A, insertions or deletions induce premature stop codon formation in both alleles of two independently derived APP-KO hiPSC clones, IA1 and IB6. The qRT-PCR analysis of APP, APLP1, and APLP2 mRNA levels in NPCs (n Ն 8 from three independent experiments) (B), purified neurons (n Ն 8 from two independent experiments) (C), and astrocytes (n Ն 6 from at least two independent experiments) (D) shows little detectable APP mRNA (***, p Ͻ 0.001; ****, p Ͻ 0.0001) and no significant differences in APLP1 or APLP2 mRNA levels in neurons or astrocytes. There was a significant increase in APLP1 (**, p ϭ 0.0082) and APLP2 (**, p ϭ 0.0035) mRNA in NPCs. All qRT-PCR data were normalized to RPL13A, RPL27, and TBP. E, representative Western blots from WT and APP-KO NPCs using antibodies for APP and APP family members, APLP1 and APLP2, show no detectable APP protein but a slight elevation in APLP1 and APLP2. F, percentage of neurons identified by flow cytometry using a CD44Ϫ/CD184Ϫ/CD24ϩ cell-surface signature after 3 weeks of neural differentiation shows that loss of APP does not affect neuronal differentiation (n Ն7 from four independent experiments). G, percentage of astrocytes positive for CD44, CD184, and GFAP by flow cytometry is not different between WT and APP-KO astrocytes (n ϭ 12 from three independent experiments). Data are depicted with bar graphs of the mean Ϯ S.D. NS is nonsignificant.

Astrocytic APP controls LDL receptor function
with APP. Interestingly, in APP-KO NPCs, we saw a modest up-regulation of APLP1 and APLP2 mRNA levels (Fig. 2B) and protein (Fig. 2E). However, when these APP-KO NPCs were further differentiated to neurons (Fig. 2C) or astrocytes (Fig.  2D), this up-regulation of APLP1 and APLP2 mRNA was no longer detected. These data indicated that there were no offtarget effects on homologous APP family members, and any phenotypes observed in APP-KO neurons or astrocytes were not due to loss or overexpression of APLP1 or APLP2.
Because APP is a key protein in neural development (22), we next tested whether the loss of APP influenced the capability of APP-KO NPCs to differentiate into neurons or astrocytes. Using flow cytometry for a neuronal cell-surface signature of CD184Ϫ, CD44Ϫ, and CD24ϩ (23), we found no difference in the percentage of neurons generated from WT or APP-KO NPCs following multiple rounds of neuronal differentiation (Fig. 2F). Similarly, using the glial markers, CD44, CD184, and GFAP, we found no difference in the differentiation capability of WT or APP-KO NPCs to generate astrocytes (Fig. 2G).

hiPSC-derived astrocytes produce high levels of de novo synthesized cholesterol in vitro
Given the relative isolation of the brain from the periphery because of the blood-brain barrier, cholesterol is synthesized locally in the brain (24). Previous studies examining sterol synthesis in brain-specific cell types have identified astrocytes as the brain's primary source of cholesterol (25). To test whether our hiPSC-derived system recapitulated this essential function of astrocytes in vivo, we performed a comprehensive analysis of free sterols in astrocytes, neurons, and NPCs using LC-MS. We found that, relative to neurons and NPCs, astrocytes had significantly elevated levels of the immediate cholesterol precursors, desmosterol ( Fig. 3A) and 7-dehydrocholesterol (Fig. 3B). Consistent with previous work reporting that astrocytes predominantly contain sterols from the Bloch pathway of cholesterol biosynthesis (26), we observed higher concentrations of astrocyte-derived desmosterol relative to 7-dehydrocholesterol. These data suggest that hiPSC-derived astrocytes recapitulate the ability to synthesize cholesterol, similar to their native function in vivo.

APP-KO astrocytes have altered cholesterol metabolism
We next sought to determine whether loss of APP affected cholesterol levels in hiPSC-derived astrocytes. Compared with WT, APP-KO astrocytes had decreased cholesterol (Fig. 3C). Given that levels of intracellular cholesterol regulate the activity of the sterol regulatory element-binding protein (SREBP) family of transcription factors (27,28), we next analyzed SREBP-target genes. These included HMG-CoA reductase (HMGCR), the ratelimiting enzyme in cholesterol biosynthesis, and the low-density lipoprotein receptor (LDLR). We also analyzed expression of lowdensity lipoprotein receptor-related protein 1 (LRP1), another highly expressed lipoprotein receptor in human astrocytes (29), whose expression has been reported to be regulated by the APP ; B, 7-dehydrocholesterol (****, p Ͻ 0.0001; n Ն 4) in astrocytes, neurons, and NPCs shows much higher levels of cholesterol precursors in astrocytes. C, quantification of total cellular cholesterol levels by Amplex Red cholesterol assay kit shows decreased cholesterol in APP-KO astrocytes (*, p ϭ 0.0226; n Ն 13 from three independent experiments). D, qRT-PCR analysis of mRNA levels from astrocytes grown in medium containing 3% FBS shows up-regulated HMGCR (*, p ϭ 0.0382), LDLR (*, p ϭ 0.0427), and LRP1 (*, p ϭ 0.0157) gene expression in APP-KO astrocytes (n Ն 10 from at least two independent experiments). E, qRT-PCR analysis of HMGCR, LDLR, and LRP1 mRNA levels from astrocytes grown with lipoprotein-depleted serum (LDS) shows no significant differences in expression between WT and APP-KO astrocytes (n Ն 8 from at least two independent experiments). F, quantification of mRNA induction of HMGCR, LDLR, and LRP1 mRNA by qRT-PCR 24 h after changing growth medium from 3% FBS-containing medium to lipoprotein-depleted medium shows no significant differences in the fold induction of SREBP-target genes in WT and APP-KO astrocytes (n Ն 4 from at least two independent experiments). Data are depicted with bar graphs of the mean Ϯ S.D.

Astrocytic APP controls LDL receptor function
intracellular domain (12) in addition to SREBP (30). mRNA expression of HMGCR, LDLR, and LRP1 was up-regulated in APP-KO astrocytes when grown in normal growth medium with 3% serum (Fig. 3D). To test whether this result in APP-KO astrocytes was a consequence of low intracellular cholesterol and not simply aberrant SREBP function, we examined whether APP-KO astrocytes could further up-regulate transcript in response to prolonged lipoprotein depletion. Indeed, after culture under lipoprotein-free conditions, differences in SREBPtarget gene expression were no longer observed between WT and APP-KO astrocytes (Fig. 3E). Furthermore, there was no difference between WT and APP-KO astrocytes in the fold induction of either HMGCR, LDLR, or LRP1 mRNA after 24 h of cholesterol withdrawal (Fig. 3F). These data suggested that because APP-KO astrocytes could still modulate SREBP-target gene expression, low intracellular cholesterol resulted in up-regulation of cholesterol synthesis and internalization genes.

APP-KO astrocytes have decreased lipoprotein endocytosis but are not defective in bulk endocytosis, bulk receptor recycling, or expression of LDL receptors at the cell surface
Because APP-KO astrocytes exhibited a cholesterol starvation phenotype, but did not demonstrate impaired SREBP signaling in response to cholesterol withdrawal, we hypothesized that APP-KO astrocytes had reduced endocytosis of extracellular lipoproteins from the culture media. To test this, WT and APP-KO cells were treated with fluorescently-labeled LDL (Fig.  4A). After 1 h of continuous LDL treatment, APP-KO astrocytes demonstrated a modest but significant reduction in lipoprotein endocytosis as reflected by a reduction in intensity of intracellular LDL fluorescence. To test whether this reduction in lipoprotein endocytosis was a result of reduced bulk endocytosis, we treated cells with fluorescently-labeled dextran (Fig.  4B). We found no difference in dextran internalization between WT and APP-KO astrocytes, suggesting that impaired endocytosis was specific for lipoproteins.
Next, to understand the mechanism of reduced lipoprotein endocytosis in APP-KO astrocytes, we also tested bulk receptor recycling by flow cytometry using fluorescently-labeled transferrin (Tfn). Tfn marks recycled cargo and allows for characterization of recycling compartments (31), and APP is known to recycle back to the cell surface in Tfn receptor-positive vesicles. To test the endocytic recycling of receptors at the cell surface, astrocytes were incubated with Tfn at 37°C for 10 min to allow Tfn uptake. Cells were then acid-washed to remove surface-bound Tfn before being chased with growth medium at fixed time points to allow Tfn to be recycled back to the cell surface. We observed no difference in the rate of Tfn recycling between WT and APP-KO astrocytes over time (Fig. 4C), suggesting that general recycling of receptors in endosomes, which normally contain APP, was not impaired in APP-KO astrocytes.
Given that both bulk endocytosis and Tfn receptor-marked recycling pathways were not defective in APP-KO astrocytes, we next examined whether newly synthesized lipoprotein receptors could be shuttled to the cell surface via the secretory pathway. To do this, we measured cell-surface LDLR by flow cytometry after treatment with berberine (BBR), which stimu-lates LDLR mRNA expression and up-regulates cell-surface LDLR (32). Although BBR treatment up-regulated surface LDLR levels compared with DMSO-treated astrocytes (Fig.  4D), there were no differences in cell-surface LDLR between WT or APP-KO astrocytes in either condition. To further verify that the loss of APP does not affect cell-surface LDL receptor levels, we used a cell-surface biotinylation assay to label plasma membrane-bound proteins with a cleavable biotin and then used streptavidin beads to pull down surface proteins. This allowed us to measure cell surface (pulldown fraction), total (input fraction), and intracellular (supernatant fraction) LDLR and LRP1 levels via Western blotting (Fig. 4, E-P). Confirming our previous results, we saw no difference in the amount of cell-surface LDLR or LRP1 protein (Fig. 4, E-H). However, in line with up-regulated LDLR and LRP1 transcript in APP-KO astrocytes, we observed increased total LDLR and LRP1 protein (Fig. 4, I-L). Consistent with these results, intracellular LDLR and LRP1 protein were also elevated ( Fig. 4, M-P). Finally, examination of APP-KO astrocytes by immunofluorescence revealed the presence of enlarged LRP1 puncta (Fig. 4Q).
Together, these data revealed no defects in regulation of transcription, Tfn-receptor recycling pathways, or the shuttling of newly synthesized receptors to the cell surface via the secretory pathway in APP-KO astrocytes. These data suggest that the loss of APP attenuated lipoprotein endocytosis and contributed to decreased cholesterol and increases in SREBP-target genes.

FAD astrocytes exhibit alterations in APP processing
In light of previous work implicating multiple APP fragments in the regulation of cholesterol homeostasis, we aimed to determine which APP fragments are required for proper lipoprotein metabolism. We also sought to examine whether FAD mutations in APP affect cholesterol homeostasis in human astrocytes. To do this, we analyzed an isogenic allelic series of astrocytes either heterozygous or homozygous for the APP Swedish mutation (APP Swe/WT and APP Swe/Swe ) or APP V717F mutation (APP V717F/WT and APP V717F/V717F ) (33). The Swedish and V717F mutations are thought to have different defects in APP processing, which would allow us to make distinct predictions about which APP fragments are relevant for lipoprotein regulation (34,35).
To characterize the APP processing alterations in FAD mutant astrocytes, we quantified protein levels of FL APP (Fig.  5A). Of all FAD genotypes, only APP Swe/Swe astrocytes exhibited a reduction of FL APP (Fig. 5B). However, given that ␤-secretase processing of FL APP is favored in the APP Swedish mutation (36,37), we hypothesized that this loss of FL APP in APP Swe/Swe astrocytes would coincide with increases in A␤ and soluble APP (sAPP) ␤ along with a decrease in sAPP␣. As predicted, APP Swedish astrocytes secreted high levels of A␤40, A␤42, and A␤38 (Fig. 5, F-H), with no change in the A␤42/ A␤40 ratio compared with WT (Fig. 5E). Additionally, APP Swedish astrocytes exhibited decreases in sAPP␣ (Fig. 5I), little detectable WT sAPP␤ (Fig. 5J), and increased Swedish sAPP␤ (Fig. 5K), which was recognized by an antibody specific for the APP Swedish mutation and only detectable in APP Swedish astrocytes. To detect APP C-terminal fragments (APP CTFs),  ). B, quantification of bulk endocytosis of dextran by flow cytometry shows no significant differences between WT and APP-KO astrocytes (n ϭ 24 from four independent experiments). C, quantification of transferrin recycling over time by flow cytometry shows no difference in the rate of recycling between WT and APP-KO astrocytes (n Ն 10 from three independent experiments). D, quantification of cell-surface LDLR protein of WT and APP-KO astrocytes upon DMSO or berberine (BBR) treatment by flow cytometry. Although BBR treatment did up-regulate cell-surface LDLR (****, p Ͻ 0.0001), there were no significant differences between WT and APP astrocytes (n Ն 2 from two independent experiments). E-H, cell-surface biotinylation and Western blot analysis of biotinylated LRP1 (G) or LDLR (H) at the cell surface (50% pulldown), quantified in E and F in WT or APP-KO astrocytes, show no significant differences in surface receptor levels between WT and APP-KO astrocytes (n ϭ 6 from three independent experiments). I-L, Western blot analysis of 5% input LRP1 or LDLR shows elevated total LRP1 (**, p ϭ 0.0013) and LDLR (**, p ϭ 0.0013) protein levels in APP-KO astrocytes (n ϭ 12 from three independent experiments). M-P, Western blot analysis and quantification of 5% supernatant or intracellular lipoprotein receptor levels showed elevated intracellular LRP1 (*, p ϭ 0.0452) and LDLR (**, p ϭ 0.0083; n ϭ 12 from three independent experiments). Q, immunofluorescence images of enlarged LRP1 puncta in APP-KO astrocytes compared with WT (**, p ϭ 0.0050; n Ն 19 from two independent experiments). Data are depicted with bar graphs of the mean Ϯ S.D.

Astrocytic APP controls LDL receptor function
astrocytes were treated with a ␥-secretase inhibitor for 48 h. Only APP Swedish astrocytes exhibited increased APP ␤-CTF fragments (Fig. 5D) and reductions in the APP ␣-CTF fragment (Fig. 5C). In APP V717F astrocytes, by contrast, we observed a dose-dependent increase in the A␤42/A␤40 ratio (Fig. 5E) as a result of increased A␤42 (Fig. 5G) and no significant change in the amount of APP CTFs compared with WT after 48 h of ␥-secretase inhibitor treatment (Fig. 5, C and D). APP V717F astrocytes also showed no changes in the levels of the soluble fragments, sAPP␣ (Fig. 5I) or sAPP␤ (Fig. 5J). Collectively, these data are consistent with previous data reporting that the APP Swedish mutation enhances ␤-secretase cleavage of APP (36,37), and the APP V717F mutation promotes an increased A␤42/A␤40 ratio (38,39). We hypothesized that these distinct alterations in APP processing could help us elucidate which APP fragment is most important in regulating lipoprotein metabolism in FAD astrocytes.

APP Swe/Swe astrocytes recapitulate APP-KO phenotypes of impaired lipoprotein endocytosis and altered cholesterol metabolism
To examine cholesterol homeostasis in APP mutant FAD astrocytes, we tested lipoprotein endocytosis in which we had previously observed a defect in APP-KO astrocytes. Of all APP mutant genotypes, only APP Swe/Swe astrocytes phenocopied APP-KO astrocytes in reduced lipoprotein endocytosis (Fig.  6A) without a concomitant reduction in bulk endocytosis (Fig.  6B). APP-KO, APP Swe/WT , and APP Swe/Swe astrocytes all exhibited reductions in the APP fragments generated by ␣-secretase cleavage (Fig. 5, C and I). However, given that only APP Swe/Swe , but not APP Swe/WT , astrocytes mimicked APP-KO phenotypes, we hypothesized that FL APP might be crucial in regulating lipoprotein metabolism.
To determine whether reduced lipoprotein endocytosis in FAD astrocytes with reduced FL APP levels also led to downstream alterations in cholesterol metabolism, we looked at the expression of multiple proteins involved in lipoprotein regulation (Fig. 6C). We first examined the transcription factor SREBP1, which regulates intracellular cholesterol levels. SREBP function is controlled by multiple mechanisms, including selfregulation by transcriptional positive feedback and activation via sequential proteolysis and translocation of its mature, cleaved fragment to the nucleus. Protein levels of both fulllength (Fig. 6D) and cleaved SREBP1 (Fig. 6E) were elevated in APP-KO and APP Swe/Swe astrocytes. In response to attenuated lipoprotein endocytosis, we observed that the ratio of cleaved/ FL-SREBP protein was also significantly increased in APP-KO and APP Swe/Swe astrocytes (Fig. 6F). Because mature SREBP also up-regulates lipoprotein receptor-mediated uptake of extracellular lipoproteins, we further examined LDLR protein in FAD astrocytes. As observed previously in APP-KO astrocytes, LDLR protein was elevated in APP Swe/Swe astrocytes (Fig. 6G). Together, these data suggest that both APP-KO and APP Swe/Swe astrocytes have impaired lipoprotein endocytosis and exhibit downstream biochemical changes expected in cholesterol-deficient cells.

APP-KO astrocytes and APP Swe/Swe astrocytes also have impaired uptake of A␤, another LDL-receptor ligand
Given that reduction of FL APP coincided with impaired lipoprotein endocytosis and cholesterol homeostasis, we speculated that FL APP might also be required for other astrocytespecific functions related to lipoprotein receptor function. We hypothesized that astrocytes with reduced FL APP would also be defective in lipoprotein receptor-mediated internalization of the A␤ peptide.
To determine whether our hiPSC-derived astrocytes could internalize A␤, WT astrocytes were treated with FITC-conjugated A␤ for 15 min, washed, and given fresh medium (Fig. 7A). Over the course of 72 h, we examined the presence of A␤-FITC, early endosome marker EEA1, and M6PR, which tags vesicles destined for transport to the lysosome. Although the amount of EEA1 and M6PR puncta remained constant, the number of A␤-FITC puncta decreased over time (Fig. 7B). Further analysis demonstrated that A␤ colocalization with M6PR increased over time (Fig. 7C), suggesting that A␤ was being targeted for lysosomal degradation. We further verified this observation using flow cytometry of WT astrocytes treated with the pHsensitive A␤-FITC or the pH-insensitive A␤-HiLyte Fluor 647 (Fig. 7D). Both probes were utilized because a pH-sensitive signal will decrease in fluorescence intensity when it reaches a more acidic compartment like the lysosome (40). Over the course of 48 h, the intensity of both A␤-FITC and A␤-HiLyte Fluor 647 increased over time. However, at 72 h we observed a reduction in the pH-sensitive A␤-FITC, but not A␤-HiLyte Fluor 647, suggesting that A␤ was being targeted to an acidic compartment following internalization. To exclude the possibility that the reduction of A␤-FITC simply reflected an inability to detect the probe, but not actual A␤ degradation, we supplemented astrocyte culture medium with A␤ for 24 h and measured the concentration of A␤ in astrocytes over time (Fig.  7E). Over the course of 48 h, we observed a 90% reduction in A␤. Together, these data indicate that WT hiPSC-astrocytes could both internalize and degrade A␤. To test whether APP-KO and APP Swe/Swe astrocytes are defective in A␤ internalization in addition to lipoprotein endocytosis, we treated astrocytes with A␤-HiLyte Fluor 647 for 24 h of continuous uptake (Fig. 7F). We observed reduced internalization of A␤ in both APP-KO and APP Swe/Swe astrocytes but not in APP V717F astrocytes, indicating that normal levels of FL APP are required for proper LDL receptor function in the endocytosis of both extracellular lipoproteins and A␤.

␤-Secretase inhibitor treatment reverses impairments in lipoprotein and A␤ endocytosis in APP Swe/Swe astrocytes
Because APP Swe/Swe astrocytes recapitulated defects observed in APP-KO cells, we hypothesized that this was a consequence of increased cleavage and loss of FL APP protein by ␤-secretase. Because FL APP is transported away from the plasma membrane in an endocytic compartment for ␤-secretase cleavage (41, 42), we predicted that APP Swe/Swe astrocytes also had reduced APP at the cell surface. Using cell-surface biotinylation and streptavidin beads to pull down surface proteins (Fig. 8A), we find that ϳ10% of total cellular APP is present at the cell Astrocytic APP controls LDL receptor function surface in WT astrocytes (Fig. 8B). However, in APP Swe/Swe astrocytes, we observed a 50% reduction of cell-surface APP compared with WT.

Astrocytic APP controls LDL receptor function
APP Swe/Swe astrocytes. Because BSIs are sometimes reported to have low potency in cells expressing the APP Swedish mutation (43), we measured whether 24 h of BSI treatment could reduce A␤ secretion in WT and APP Swe/Swe astrocytes. We observed decreased A␤40 peptides in BSI-treated WT astrocytes and a marked reduction of A␤40 in BSI-treated APP Swe/Swe astrocytes near WT levels (Fig. 8C). To test whether BSI treatment could rescue the defects we observed in APP Swe/Swe astrocytes, we treated WT, APP-KO, and APP Swe/Swe astrocytes with a BSI for 24 h and measured LDL and A␤ endocytosis. Upon pharmacological inhibition of ␤-secretase in APP Swe/Swe astrocytes, we observed a reversal of defects in both LDL endocytosis (Fig.  8D) and A␤ endocytosis (Fig. 8E). Significant increases in LDL or A␤ endocytosis were not observed in the absence of APP, Figure 6. APP Swe/Swe astrocytes recapitulate APP-KO phenotypes. A, quantification of LDL endocytosis in FAD astrocytes by flow cytometry demonstrates that APP Swe/Swe astrocytes also exhibit reduced lipoprotein endocytosis (p ϭ 0.0242; n Ն 4 from three independent experiments). B, quantification of bulk endocytosis in FAD astrocytes by flow cytometry for fluorescently-tagged dextran shows no significant differences between WT and all genotypes (n Ն 6 from three independent experiments). C-F, Western blot analysis of full-length SREBP1 protein, cleaved or mature SREBP1 protein, LDLR, and actin show increased levels of FL SREBP1 protein (**, p ϭ 0.0071; ***; p ϭ 0.0004; n Ն 4 from four independent experiments) (D), cleaved SREBP1 protein (**, p ϭ 0.0077; ***, p ϭ 0.0005; n Ն 4 from four independent experiments) (E), and an increased ratio of cleaved/FL SREBP1 protein (*, p ϭ 0.0204; **, p ϭ 0.0052; n Ն 4 from four independent experiments) (F) in APP-KO and APP Swe/Swe astrocytes. G, APP-KO and APP Swe/Swe astrocytes also demonstrate increased total LDLR protein (*, p ϭ 0.0326; **, p ϭ 0.0022; n Ն 5 from five independent experiments). Data are depicted with bar graphs of the mean Ϯ S.D.

Astrocytic APP controls LDL receptor function
indicating that rescue of impaired endocytosis relied on impairing ␤-secretase cleavage of FL APP.

Discussion
APP is a transmembrane protein that is highly expressed in the central nervous system. It has been shown to have many varied biological functions, likely due to multilayered mechanisms of regulation resulting in multiple proteolytic products and alternatively spliced isoforms. To determine whether APP or any of its proteolytic cleavage products are involved in lipoprotein regulation, we employed CRISPR/Cas9-genome editing to generate APP-KO, APP Swedish, and APP V717F hiPSCs , quantification of puncta count shows relatively equal numbers of EEA1 and M6PR over time, but decreased A␤-FITC puncta in WT astrocytes over 72 h. C, Colocalization analysis shows that the percent of A␤ with M6PR increased over time, but with a slight reduction from 48 to 72 h. Data are representative of two independent experiments. D, flow cytometry analysis of continuous A␤ internalization in WT astrocytes treated with A␤-FITC or A␤-HiLyte Fluor 647 shows increased A␤ over the course of 48 h, but a reduction of intracellular A␤ between 48 and 72 h in astrocytes treated with the pH-sensitive A␤-FITC, but not pH-insensitive A␤-HiLyte Fluor 647 (n Ն 10 from two independent experiments). E, quantification of cellular A␤42 by MSD immunoassay in WT astrocytes over time after pre-treatment with A␤42 for 24 h (n ϭ 6 from two independent experiments). F, quantification of A␤ internalization in APP-KO and FAD astrocytes by flow cytometry demonstrates that APP-KO (****, p Ͻ 0.0001) and APP Swe/Swe (**, p ϭ 0.0012) astrocytes have impaired A␤ uptake (n Ն 6 from two independent experiments). Data are depicted with bar graphs of the mean Ϯ S.D.

Astrocytic APP controls LDL receptor function
and differentiated them into human astrocytes, the source of de novo cholesterol in the brain. Here, we show that FL APP regulates LDL receptor function. Loss of FL APP resulted in impaired lipoprotein and A␤ endocytosis, reduced intracellular cholesterol, and aberrant elevations of transcripts and proteins related to cholesterol synthesis and internalization. Finally, we show that inhibiting cleavage of FL APP by ␤-secretase can reverse LDL and A␤ endocytosis defects, but only in the presence of APP. Thus, in addition to having a critical role in mammalian brain development (44), normal levels of FL APP are also critical in the maintenance of homeostatic brain functions. We propose that pathological alterations of FL APP levels could contribute to glial dysfunction in multiple neurodegenerative disorders.
Mechanistically, we attribute defective lipoprotein and A␤ endocytosis in APP-KO and APP Swe/Swe astrocytes to the loss of FL APP in each of these genotypes. APP isoforms containing the KPI domain have been shown to bind to LRP1 at the N terminus (45,46). Intracellularly, the cytoplasmic adaptor protein FE65 has been shown to link the C-terminal NPXY endocytosis motifs of APP and multiple LDL receptors (47)(48)(49). Thus, astrocytic APP isoforms, which include the KPI domain, may have a dual linkage with LDL receptors at both the N and C termini. In light of this prior work demonstrating interactions between APP and LDL receptors, it is possible that FL APP acts as a coreceptor for LDL receptor ligands in human astrocytes. A further explanation for dysregulated cholesterol metabolism in APP-KO and Swedish astrocytes is that these genotypes also have a high rate of ligand-independent receptor endocytosis. It is feasible that both decreased endocytosis of lipoproteins and increased ligand-independent endocytosis of the LDL recep-tors combine in an additive manner to produce the results we observed. Overall, it is consistent with our data that without proper FL APP levels the function of LDL receptors is impaired. This dovetails nicely with previous studies demonstrating that APP interacts with diverse binding partners, including APOE (50), kinesin (51,52), SORLA (53), and cholesterol (54). Interestingly, all of these interacting partners modulate APP metabolism, and alterations in these interactions are hypothesized to contribute to AD pathology.

FL APP regulates brain cholesterol metabolism
In human astrocytes, we find that normal levels of FL APP are essential for proper regulation of cholesterol homeostasis. Loss of FL APP led to impairments in lipoprotein endocytosis (Fig.  4A), resulting in decreased intracellular cholesterol (Fig. 3C) and activation of SREBP-target gene transcripts and protein (Figs. 3D and 4, I-L) in both APP-KO and Swedish astrocytes. Interestingly, we did not observe a dose-dependent effect of FL-APP levels in our LDL endocytosis experiments as we did in the A␤ endocytosis experiments. Although both APP-KO and Swedish astrocytes exhibited differential handling of lipoproteins compared with WT astrocytes, it is possible that more subtle changes in lipoprotein endocytosis correlating with FL APP levels are below the level of detection in the assay.
Previous work done by our lab using the same isogenic FAD hiPSC lines to study human neurons revealed that accumulation of a different APP fragment, ␤-CTF, caused impairments in lipoprotein endocytosis and a neuron-specific transcytotic trafficking pathway via defects in recycling (33). Although we cannot rule out that the overabundance of APP ␤-CTF in Figure 8. Pharmacological inhibition of ␤-secretase reverses LDL and A␤ endocytosis defects in APP Swe/Swe astrocytes. A and B, quantification of cell-surface biotinylation experiments shows that APP Swe/Swe astrocytes have reduced cell-surface APP levels, identified in the pulldown (PD) lanes, compared with WT astrocytes (**, p ϭ 0. 0016). Western blots were run with 5% of input, 5% of supernatant (sup.), and 50% of pulldown (n Ն 5 from three independent experiments). C, secreted A␤40 peptide levels, measured by MSD immunoassay, in WT and APP Swe/Swe astrocytes after 24 h of treatment Ϯ a ␤-secretase inhibitor (BSI) exhibit significant reductions in secreted A␤40 peptides upon ␤-secretase inhibitor treatment (*, p ϭ 0.0154; **, p ϭ 0.0020; n ϭ 6 from two independent experiments). D and E, flow cytometry analysis of LDL (n Ն 4 from two independent experiments) (D) and A␤ endocytosis (E) (n Ն 4 from two independent experiments) in WT and APP Swe/Swe astrocytes Ϯ ␤-secretase inhibitor treatment shows that ␤-secretase inhibition reverses defects in APP Swe/Swe (LDL, p ϭ 0.0120; A␤, p ϭ 0.0092) but not APP-KO astrocytes. ****, p Յ 0.0001. Data are depicted with bar graphs of the mean Ϯ S.D.

Astrocytic APP controls LDL receptor function
APP Swe/Swe astrocytes outcompetes FL APP for binding to LDL receptors and thus impairs recycling of LDL receptors to the surface, CTFs were difficult to detect in our hiPSC-derived astrocytes without ␥-secretase inhibition. This suggests that steady-state levels of the rapidly processed or degraded ␤-CTF might be quite low relative to FL APP or A␤ in APP Swe/Swe astrocytes. Given the low levels of astrocytic ␤-CTF we observed, we hypothesize that we might not observe cholesterol phenotypes in FAD astrocytes that are characterized by accumulation of ␤-CTF protein, like presenilin mutant astrocytes. By contrast, in mutant astrocytes characterized by increased APP expression (e.g. APP duplication or Down syndrome), we hypothesize that there will be alterations in cholesterol metabolism. Altering APP gene dosage in astrocytes is likely to alter its normal physiological processing as well as the processing of other proteins, like LRP1, with which APP may be competitive substrate. Collectively, these data shed light on the importance of studying the role of endogenous proteins in specific cell types as we find that alterations in APP processing can affect either astrocyte or neuron-specific functions via different mechanisms. Increased ␤-CTF in human neurons and loss of FL APP in human astrocytes may impair lipoprotein endocytosis independently in a cell-specific context. Additional work comparing astrocytes and neurons from different FAD mutations would be revealing.
In this study, we demonstrate a linkage between FL APP levels and lipoprotein endocytosis. However, further study of how FL APP levels in astrocytes influence lipoprotein export and thus neuronal health is also needed, given the reliance of neurons on astrocyte-derived lipoproteins (25,56). Future work to investigate how aberrant lipoprotein metabolism in astrocytes contributes to AD phenotypes could provide mechanistic insight into the development of new AD therapeutics.

FL APP regulates A␤ clearance
Here, we report that loss of FL APP in APP-KO and APP Swe/Swe astrocytes impairs A␤ internalization (Fig. 7F). Interestingly, we do not observe A␤ clearance defects in the other FAD mutations. This not only reflects the high degree of clinical and pathological heterogeneity within AD, but also heterogeneity within subgroups of FAD patients harboring different FAD mutations (57,58). Although it is possible that some FAD mutant astrocytes could have small alterations in FL APP levels that are also below our level of detection, these phenotypic differences are also in agreement with the notion that the accumulation of A␤ in FAD is primarily due to neuronal overproduction of A␤. In line with this idea, FAD mutations in APP are often considered to be gain-of-function mutations due to the generation and accumulation of some toxic proteolytic product (3). However, here we find in APP Swe/Swe astrocytes that the mutation confers a loss-of-function phenotype associated with the loss of FL APP to ␤-secretase cleavage. We postulate that the impairment in A␤ clearance in APP Swe/Swe astrocytes could also contribute to the greatly increased A␤ plaque load observed in mouse models overexpressing the APP Swedish mutation (59 -61).
In light of data suggesting that the increased deposition of A␤ in SAD is primarily a result of impaired A␤ clearance rather than increased A␤ generation (62), further work in this system could examine whether astrocytes derived from SAD patients are defective in A␤ clearance. Also, given the observation that both reactive astrocytes and cells undergoing a cellular stress response alter APP expression and APP processing (63)(64)(65), it would be revealing to study how FL APP levels under these pathological conditions affect A␤ clearance mechanisms.

Linking APP to mechanisms of glial dysfunction in SAD
In this study, we find that loss of FL APP impairs LDL receptor function. We provide a novel linkage between FL APP levels and two potential mechanisms of glial dysfunction in AD: dysregulation of cholesterol metabolism and A␤ clearance. Although several studies have addressed the effect of the loss of LDL receptors on APP processing (66 -69), little has been done to elucidate how the loss of APP affects LDL receptor function. The concept that FL APP levels directly affect the ability of LDL receptors to endocytose lipoproteins or A␤ is in agreement with the known stoichiometry of the LRP1-FE65-APP trimeric complex (47). Loss or overproduction of any member of this multimeric complex could alter the various functions of the LDL receptors by abrogating complex formation. Together, using endogenous protein levels in an isogenic series of iPSCderived human astrocytes, our data shed light on the novel function of FL APP in controlling LDL receptor-mediated cholesterol metabolism and A␤ clearance in human astrocytes. These findings suggest that FL APP may have a more central role in the etiology of AD than previously suspected.

Genome editing
All isogenic iPSCs were derived using CRISPR/Cas9 as described previously from the CVB iPSC line (33,71) derived from J. Craig Venter (73), whose diploid genome sequence is publicly available (74). In brief, iPSCs were pretreated with 10 M RI prior to nucleofection. To obtain single cells, iPSCs were dissociated with Accutase and filtered twice through 100-m filters. To generate APP-KO and APP V717F clones, 2 ϫ 10 6 iPSC were nucleofected using the Amaxa Human Stem Cell Nucleofector Kit I (Lonza) with 6 g of CMV::Cas9 -2A-eGFP vector and 3 g of U6::gRNA vector, which was generated using the gRNA synthesis protocol as described previously (75). For APP Swedish clones, 8 ϫ 10 5 iPSC were nucleofected with 5 g of pSpCas9(BB)-2A-GFP (PX458) vector, which was generated as described previously (76). The following guide RNA sequences were used, targeting exon 16 for APP-KO and APP Swedish: GGA GAT CTC TGA AGT GAA GAT GG, and exon 17 for APP V717F: GAC AGT GAT CGT CAT CAC CTT GG. To introduce the APP Swedish or APP V717F point mutations in our WT CVB hiPSC line using CRISPR/Cas9-mediated homology directed repair, 100 M single-stranded DNA oligonucleotides were also included during nucleofection. After culturing the iPSCs in the presence of RI for 48 -72 h, 1 ϫ 10 4 GFPϩ iPSCs were FACS sorted (FACS Aria IIu, BD Biosciences) and plated on 10-cm MEF feeder plates in the presence of RI. After 1 week, single colonies were manually picked, cultured in 96-well plates, and expanded. DNA from single clones was harvested using QuickExtract DNA Extraction Solution (Epicenter) and PCR-amplified using the following PCR primers: APPex16-F, CCC GTA AGC CAA GCC AAC AT, and APPex16-R, CAT GCA CGA ACT TTG CTG CC; or APPex17-F, TGT TCC ACC TGT CAA AGG GT, and APPex17-R, AGT TGA GAT AAC AAC ACA CAC TCT. PCR products were purified using the QIAquick PCR purification kit (Qiagen) or Exo-SAP-IT PCR Product Cleanup Reagent (ThermoFisher Scientific) as directed by the manufacturer and Sanger sequenced. Clones in which disruptions at the predicted gRNA/Cas9 cut site were observed were further sequenced after cloning using the Zero Blunt TOPO PCR cloning kit (Invitrogen). All previously unpublished isogenic iPSCs were digitally karyotyped by hybridization to the Infinium CoreExome-24 BeadChip (Illumina) as described previously (77) and determined to be euploid.

RNA expression
For mRNA expression analysis, RNA was isolated using the RNeasy mini kit (Qiagen) and DNase-treated using TURBO DNase (Ambion) for 1 h at 37°C. cDNA was synthesized from RNA primed with oligo(dT) using the SuperScript First-Strand synthesis system (Invitrogen). qRT-PCR was performed using FastStart SYBR Green (Roche Applied Science), and samples were run in triplicate on an Applied Biosystems 7300 RT-PCR system. Data were analyzed using the ⌬⌬Ct method, and target genes were normalized to the geometric mean of three housekeeping genes: RPL13A, RPL27, and TBP. The

Protein expression
Cells were lysed in RIPA Lysis Buffer (Millipore) with protease (Calbiochem) and phosphatase inhibitors (Halt). Protein concentrations were determined using the Pierce BCA protein assay kit (ThermoFisher Scientific). Equal amounts of protein lysates were run on NuPAGE 4 -12% BisTris gels (Invitrogen), transferred to nitrocellulose or polyvinylidene difluoride membranes, and blocked for 1 h at room temperature using either 5% milk or Odyssey Blocking Buffer (LI-COR). Blots were probed overnight at 4°C using the corresponding primary antibodies followed by HRP-conjugated (Vector Laboratories) or IRDye secondary antibodies (LI-COR) at 1:5000. Bands were quantified using ImageJ software or the Odyssey Imaging System follow-

Sterol analysis
For free sterol analysis, 1 ϫ 10 6 cells were pelleted and stored at Ϫ80°C before sterols were extracted and analyzed as described previously at the UCSD Lipidomics Core (55,78). Briefly, sterols were extracted by dichloromethane/methanol (50:50; v/v), hydrolyzed, and separated using reverse-phase LC using a 1.7-m 2.1 ϫ 150-mm Kinetex C18 column (Phenomenex) on an ACQUITY UPLC system (Waters) followed by analysis on QTRAP 6500 mass spectrometer (Ab Sciex). A mixture of deuterated standards (Avanti Polar Lipids) was used for internal standards. For intracellular cholesterol measurements, 1 ϫ 10 6 subconfluent astrocytes were pelleted and stored at Ϫ80°C. Cells were resuspended in PBS, mixed with chloroform/methanol (2:1 v/v), vortexed, and rotated. Following centrifugation, the chloroform and lipid-containing layer was transferred to a new tube, vacuum-dried, and resuspended in Reaction Buffer E from the Amplex Red Cholesterol Assay (Invitrogen). Cholesterol was measured from the chloroform/methanol-extracted samples using the Amplex Red Cholesterol Assay as directed by the manufacturer.

Cell-surface biotinylation
For surface biotinylation of lipoprotein receptors, astrocytes were seeded in a 10-cm plate at a density of 2 ϫ 10 6 cells per plate in duplicate. After 48 h, sub-confluent astrocytes were washed three times with ice-cold PBS and then incubated with either PBS or 2 mM EZ-Link Sulfo-NHS-SS-Biotin (Thermo-Fisher Scientific) for 30 min at 4°C. Cells were washed three times with TBS, pH 7.4, and lysed in RIPA buffer with protease and phosphatase inhibitors. For streptavidin pulldown, 250 g of protein lysate at 0.5 g/l was incubated with 100 l of pre-washed PureProteome Streptavidin Magnetic Beads (EMD Millipore) by rotating overnight at 4°C. The next day, beads were immobilized, and a sample of supernatant was saved to measure nonbiotinylated intracellular protein. The beads were washed five times in cold PBS containing 1% Triton X-100, and biotinylated proteins were released from the streptavidin beads by boiling the samples in 2ϫ NuPAGE LDS Sample Buffer (Invitrogen) at 100°C. Western blots were run with 5% of input, 5% of supernatant, and 50% of pulldown.

Preparation of A␤ peptide
Lyophilized FITC-labeled ␤-amyloid(1-42) (American Peptide) or ␤-amyloid(1-42) HiLyte Fluor 647 (Anaspec) was solubilized following the manufacturer's instructions using a minimal amount of alkaline 1.0% NH 4 OH immediately followed by 1ϫ PBS to a working concentration of 100 M. Small aliquots were immediately stored at Ϫ80°C and only used once to eliminate variability due to freeze-thawing.

Flow cytometry
For lipoprotein and bulk endocytosis assays, astrocytes were seeded in a 24-well plate at a density of 8 ϫ 10 4 cells per well. Two days later, cells were treated with 20 g/ml BODIPY FL LDL or DiI LDL for 1 h at 37°C or 50 g/ml dextran-fluorescein or dextran-Alexa Fluor 647 (Life Technologies, Inc.). Following incubation with labeled substrates, cells were washed with cold PBS and dissociated with trypsin (Invitrogen) for 5-10 min at 37°C to remove any ligand bound to the cell surface. For the transferrin recycling assay, astrocytes were seeded in a 24-well plate at a density of 8 ϫ 10 4 cells per well. Two days later, cells were treated with 100 g/ml transferrin-Alexa Fluor 647 (Life Technologies, Inc.) for 10 min at 37°C. Following the 10-min incubation, cells were washed with cold PBS and cold acid wash buffer to remove transferrin bound to the cell surface. Cells at the "0" time point were dissociated using Accutase, filtered, and stored on ice until analysis. The remaining conditions were "chased" with culture medium and harvested at the indicated time points. For analysis of cell-surface LDLR protein, astrocytes were seeded in a 24-well plate at a density of 8 ϫ 10 4 cells per well. After 24 h, astrocytes were treated with either DMSO or berberine (10 g/ml, Selleck Chemicals) for 24 h. Next, cells were washed with PBS and dissociated using an EDTA dissociation buffer (50 mM HEPES, pH 7.4, 1 mM EDTA, 5 mM glucose, 5 mM KCl, 125 mM NaCl, and 2 mg/ml BSA) for 10 min at 37°C. Cells were then incubated with (R)-phycoerythrin mouse anti-human LDLR (BD Biosciences 565653) at a final concentration of 8 g/ml on ice for 30 min. For continuous A␤ uptake assays, astrocytes were incubated with 500 nM FITC-labeled ␤-amyloid(1-42) (American Peptide) or 500 nM ␤-amyloid(1-42) HiLyte Fluor 647 (Anaspec). At the indicated time points, cells were harvested by trypsinization to remove any surface-bound ligand and fixed with 4% paraformaldehyde at room temperature for 15 min. Fixed cells were stored in PBS at 4°C until analysis. All experiments were analyzed on an Accuri C6 flow cytometer (BD Biosciences). 10,000 -20,000 events were recorded per sample, and the median fluorescence intensity was quantified.

A␤ and sAPP measurements
For secreted A␤ and sAPP measurements, astrocytes were seeded at 2.5 ϫ 10 5 cells per well of a 24-well plate. The follow-

Astrocytic APP controls LDL receptor function
ing day, the medium was changed. After 5 days in culture, medium was harvested and run on a V-PLEX A␤ Peptide Panel 1 (6E10) kit, sAPP␣/sAPP␤ kit, and/or Swedish sAPP␤ kit (Meso Scale Discovery). For cellular A␤ measurements, 500 nM A␤(1-42) (American Peptide) was supplemented to astrocyte culture medium. After 24 h, cells were washed with PBS, and fresh medium was added. At the indicated times, cell lysates were harvested using MSD Lysis Buffer with protease and phosphatase inhibitors and stored at Ϫ80°C until they were run on a V-PLEX A␤ Peptide Panel 1 (6E10) kit (Meso Scale Discovery). These measurements were normalized to protein content using the Pierce BCA Protein Assay kit (ThermoFisher Scientific). ␥-Secretase inhibitor treatment was performed using 200 nM Compound E (EMD Chemicals) for 48 h. ␤-Secretase inhibitor treatment was performed using 4 M ␤-Secretase Inhibitor IV (Calbiochem) for 24 h.

Statistics
All data were analyzed using GraphPad Prism Software. Statistical analysis comparing the two groups was performed using Student's t test. Statistical analysis comparing different genotypes to WT controls was performed by Dunnett's multiple comparisons test. Data are depicted with bar graphs of the mean Ϯ S.D.