Altered APP Processing in Insulin-Resistant Conditions Is Mediated by Autophagosome Accumulation via the Inhibition of Mammalian Target of Rapamycin Pathway

C57BL/6 male mice were used for these experiments because this mouse strain is a good model of human obesity and T2DM (21). The mice were fed for 22 months with a Western diet (Harlan, TD.88137), comprising 21% (w/w) total lipid (42% calories from fat) and 0.2% (w/w) cholesterol, or a control diet (no. 5001, Test Diet), comprising 4% (w/w) total lipid (<12% calories from fat) and <0.04% (w/w) cholesterol. Body weight was recorded weekly. Food intake was measured for a 2- to 3-day period three times during the Western diet regimen. Mice were fasted overnight on the day before they were killed (24-month-old mice), and body fat contents were measured by dual-energy X-ray absorptiometry (PIXImus and Lunar Piximus 2 2.0 software, GE Lunar) and blood samples were collected.

Mice were maintained in the accredited pathogen-free facility at Seoul National University Bundang Hospital on a 12-h light/dark cycle and fed the indicated diet and water ad libitum. All procedures involving the use of laboratory animals were in accordance with the Guide for Standard Operation Procedures and approved by institutional animal care and use committee of the Clinical Research Institute, Seoul National University Bundang Hospital. Male C57BL/KsJ db/db mice (n = 10), a rodent model of T2DM (19), and their nondiabetic db/+ littermates (n = 9) were purchased from Jung Ang Laboratory Animal Inc. (Korea). The animals had free access to tap water and standard chow. Body weights and nonfasting blood glucose levels were determined at 14 weeks of age.

Serum cholesterol and triglyceride levels were determined by Beckman Coulter AU480 automatic biochemistry analysis system (Japan), with reagent kits provided by the manufacturer. Insulin concentration was measured using a mouse insulin radioimmunoassay kit (Millipore-Linco).

Mouse primary cortical neurons were prepared as described previously (11). Experiments were conducted at days in vitro (DIV) 14. The human neuroblastoma cell line SH-SY5Y was cultured as described previously (11). The cells were transfected with several expression vectors using Lipofectamine Plus (Invitrogen) and small interfering (si)RNAs (Santa Cruz Biotechnology) using RNAimax (Invitrogen) according to the manufacturer’s instructions.

Harvested cell pellets and mouse brain tissues were prepared as described (22). The antibodies for the Western blot analysis were anti-tumor necrosis factor-α converting enzyme (TACE), anti-mTOR, and anti-insulin receptor substrate 1 (IRS1) (Santa Cruz Biotechnology); anti-p62/sequestosome 1 (SQSTM1) and anti–β-actin (Sigma-Aldrich); anti-LC3B (microtubule-associated protein 1 light chain 3β), anti–phosphorylated (p)-AMPK (Thr172), anti-AMPK, anti–p-Akt (Thr308), anti-Akt, p-extracellular signal–related kinase (ERK)1/2, anti-ERK1/2, anti–BACE1, anti-matrix metalloproteinase (MMP)-9, anti­–p-mTOR (S2448), anti–p-p70S6 kinase (p70S6K), and anti–p70S6K (Cell Signaling Technology); anti-phosphatidylinositol 3-kinases (PI3K) (p85α) and anti–p-IRS1 (Ser307) (Upstate Biotechnology); anti-APP (6E10) and anti-sAPPβ (Signet); anti-nicastrin and anti-PS1 loop (Millipore); and anti-beclin 1 and anti-IDE (Abcam). The immunoreactive bands were photographed and quantified on an LAS-3000 with MultiGauge (FujiFilm Inc.) and ImageJ software.

For measuring cell viability, we performed calcein-acetoxymethyl ester assay and MTT assay as previously described (11).

Sandwich enzyme-linked immunosorbent assay (ELISA) was performed as described for Aβ quantification (23).

In vitro peptide cleavage assay was performed as described for the measurement of β- or γ-secretases activity (22).

To analyze green fluorescent protein (GFP)-LC3, red fluorescent protein (RFP)-LC3, and tandem-fluorescence LC3 (TfLC3), plasmid encoding GFP-LC3, RFP-LC3, and TfLC3 (a gift of Dr. M.S. Lee, Sungkyunkwan University) was transfected into SH-SY5Y cells. The appearance of RFP-LC3 and GFP-LC3 puncta was visualized on a confocal laser scanning microscope (FV10i-w; Olympus).

Immunocytochemical staining was performed as described (24). Images were taken on a confocal laser scanning microscope (FV10i-w; Olympus).

Cell medium was centrifuged at 4,000 rpm for 5 min to remove cell debris and subjected to trichloroacetic acid precipitation.

For electron microscopy (EM) analysis, brain tissue and SH-SY5Y cells were prepared as described (11).

Immunogold electron microscopy (ImmunoEM) was performed as described previously (25).

All data are expressed as mean ± SEM. Differences between groups were analyzed by Tukey-Kramer multiple comparison test and t test. P values < 0.05 were considered to be statistically significant.

HFD-induced insulin resistance was confirmed in 24-month-old C57BL/6 mice on the basis of elevated circulating serum insulin, obesity (increased body weight and ependymal fat pad weight), and elevated levels of blood glucose and cholesterol relative to those of age- and sex-matched control mice (mice fed regular chow; Supplementary Fig. 1).

Abnormalities in insulin metabolism are among the central factors thought to mechanistically influence the onset of AD (5) via their influence on the synthesis and degradation of Aβ, held central to AD neuropathology. To identify the alteration in APP processing under insulin-resistant conditions, Aβ in the cortical areas of brains of HFD mice was quantified by ELISA. Levels of Aβ40 and Aβ42 were increased in HFD mice compared with levels in control mice (Fig. 1A), indicating that the insulin resistance directed APP processing toward the amyloidogenic pathway. To determine changes in the Aβ-generating pathway, we performed in vitro peptide cleavage assay to measure β- and γ-secretase activities. Their activities were increased in HFD mice compared with control mice (Fig. 1B and C). Western blot analysis revealed no difference in expression of a disintegrin and metallopeptidase domain 17 (ADAM17) (α-secretase), BACE1 (β-secretase), or presenilin 1 (γ-secretase component) (Fig. 1D), indicating that the brains of HFD mice are subject to increased β- and γ-secretase activities and Aβ expression without a change in protein expression.

To explore the mechanisms by which insulin resistance promoted APP processing directed toward the amyloidogenic pathway, we looked for indices of altered insulin signaling in the brains of HFD mice relative to that of control mice. Previous studies have indicated that increased Ser/Thr phosphorylation of IRS proteins leads to an insulin-resistant state by various mechanisms (26,27). According to Western blot analysis, the phosphorylation level of IRS1 (Ser307) was increased in HFD mice compared with control mice (Fig. 2A). Also, the alteration in insulin signaling in the brain of HFD mice was examined. We found that the dietary condition leading to insulin resistance results in a significant reduction of PI3K p85 subunit expression (indicative of reduced PI3K signaling) (28) and phosphorylation of downstream Akt/protein kinase B in the cortex of HFD mice significantly (Fig. 2A). Considering that p-IRS1 and p-Akt were modified by diet, we evaluated other proteins involved in the insulin-signaling pathway (i.e., ERK1/2 and AMPK). The total level of ERK1/2 did not change, but the HFD significantly decreased p-ERK1/2 but not p-AMPK (Fig. 2A). These data show that the brains of HFD mice undergo altered insulin signaling.

A greatly increased level of autophagosomes was reported in the pancreatic β-cells of diabetic db/db mouse and HFD mice (15). Another study demonstrated that autophagic vacuoles accumulate pathologically in the brains of AD patients, and autophagy acts as a major Aβ-generating factor (10,12). To explore the mechanisms of altered APP processing in the HFD condition, we examined whether HFD-induced insulin resistance increases autophagosome formation in mouse brains. Western blot analysis showed that the ratio of LC3-II to LC3-I, protein markers of autophagy (29), increased in cortical areas of the brains of HFD mice (Fig. 2B). EM of the cortex of HFD mice revealed the abnormal accumulation of large autophagosomes (Fig. 2C); the brains of control mice did not harbor autophagosomes. Previous studies have shown that autophagosome formation is induced by inhibition of mTOR (30). The brains of HFD mice showed decreased phosphorylation of p70S6K (Fig. 2B), a downstream component of the mTOR pathway (31), indicating that the HFD induced accumulation of autophagosomes via the mTOR pathway.

We further investigated whether insulin resistance induces autophagosome accumulation and altered APP processing in db/db mice, which showed typical characteristics of T2DM (19). At 14 weeks, body weights and blood glucose levels of db/db mice were increased significantly compared with those of their db/+ littermates (Fig. 3A). By ELISA, Aβ40 and Aβ42 levels were increased in db/db mice compared with levels in littermates (Fig. 3B). The phosphorylation level of IRS1 (Ser307) and the LC3-II–to–LC3-I ratio were increased in the brains of db/db mice (Fig. 3C), indicating defective insulin signaling and accumulation of autophagosomes in the db/db mice. Thus, as with HFD mice, db/db mice show insulin resistance, autophagosome accumulation, and increased Aβ levels in the brain.

To determine the relationship between HFD-induced autophagosome accumulation and alteration in APP processing, we established a cell culture model of insulin resistance using SH-SY5Y cells. To make an in vitro model of insulin resistance based on previous study in skeletal muscle and neurons (20,32), we cultured SH-SY5Y cells in serum-free media in the absence or presence of chronic low insulin (LI; 10 nmol/L insulin) or chronic high insulin (HI; 1 μmol/L insulin) for 48 h, SH-SY5Y cells in the HI condition showed comparable attenuation of insulin-stimulated phosphorylation of Akt (Fig. 4A and B). Acute insulin treatment induced Akt phosphorylation in a concentration-dependent manner (5–20 nmol/L), whereas cells in the HI condition had no effect on Akt phosphorylation in response to acute insulin treatment (Fig. 4B).

We next tested the effect of the HI condition on other insulin-signaling and autophagy-related proteins. We observed a marked reduction in the phosphorylation level of p70S6K and ERK and an increased LC3-II–to–LC3-I ratio (Fig. 4C), indicating that the HI condition induced altered insulin signaling and increased autophagosome formation. To determine whether the alteration in insulin signaling increased the LC3-II–to–LC3-I ratio, rosiglitazone, an insulin-sensitizing agent (33), was administered to SH-SY5Y cells in the HI condition. Rosiglitazone reversed changes in the phosphorylation states of p70S6K, ERK, and the LC3-II–to–LC3-I ratio (Fig. 4C). To visualize autophagosome formation, SH-SY5Y cells were transiently transfected with RFP-LC3. Treatment of cells with 1 μmol/L insulin increased the number of RFP-positive puncta (Fig. 4D). Conversely, rosiglitazone reversed the changes in RFP-LC3 puncta (Fig. 4D). The agents had no effect on cell survival (Supplementary Fig. 2). EM revealed an abnormal accumulation of autophagosomes in SH-SY5Y cells under the HI condition (Fig. 4E), indicating that altered insulin signaling induces upregulation of autophagosome formation.

It is reported that the reduced insulin signaling can induce neuronal autophagy activation under nutrient-deprived conditions (16). To investigate how altered insulin signaling induces upregulation of autophagosome formation, Akt inhibitor and rapamycin, the mTOR inhibitor, were used to treat SH-SY5Y cells. Both inhibitors induced autophagosome accumulation under the normal condition but not in the HI condition (Fig. 5A). These data indicate that insulin-induced accumulation of autophagosome is mediated by the Akt-mTOR pathway in SH-SY5Y cells.

To determine whether the insulin-induced accumulation of autophagosomes was caused by activation of the induction for autophagosome formation or by defects in autolysosome maturation, we used LysoTracker probes to label lysosomes on GFP-LC3–transfected SH-SY5Y cells. Double-labeling with GFP-LC3 and LysoTracker Red revealed that treatment with 1 μmol/L insulin induced LC3 degradation in LysoTracker-positive vesicles (Fig. 5B; arrowhead). To assess the defects in autolysosome maturation, we incubated cells with leupeptin, a lysosomal inhibitor, to halt the digestion of autolysosomal contents before the insulin treatment. We found insulin-induced accumulation of LC3 in LysoTracker-labeled vesicles after pre-exposure to leupeptin (Fig. 5B; asterisk). To examine autophagosome clearance in greater detail, we used mRFP-GFP TfLC3 (18). The application of the TfLC3 construct in SH-SY5Y cells with 1 μmol/L insulin showed a significant accumulation of LC3 dots that were only RFP-positive signals but not GFP signals (Fig. 5C), suggesting that the cells were localized in the autolysosomes because the GFP signals were preferentially quenched in the acidic condition (18). When chloroquine (CQ), a lysosomal inhibiting agent, was used to treat SH-SY5Y cells with 1 μmol/L insulin, GFP and RFP signals were colocalized in most cells (Fig. 5C). Western blot analysis also showed that SH-SY5Y cells with 1 μmol/L insulin in the presence of CQ resulted in a significantly higher level of LC3-II–to–LC3-I ratio than with 1 μmol/L insulin alone (Fig. 5D). In an alternative approach, p62/SQSTM1 degradation was examined to evaluate impairment in autophagic protein degradation (34). Treatment with 1 μmol/L insulin induced p62 degradation (Fig. 5E), suggesting that the HI condition did not impair autolysosome maturation. We also checked levels of beclin1, a component of the autophagosome initiation complex. Under the HI condition, beclin1 expression was increased (Fig. 5F). Each of these findings supports the notion that the HI condition can induce autophagosome formation without defects in autolysosome maturation.

To determine whether HI condition–induced insulin resistance alters APP processing through autophagosome accumulation, we analyzed alteration of secretase activity by measuring levels of secreted forms of APP (sAPPα and sAPPβ) in conditioned medium. With 1 μmol/L insulin, the membrane-associated APP (full-length APP) level did not change, whereas the sAPPα level decreased but the sAPPβ level increased (Fig. 6A), indicating that the HI condition induced β-secretase activation. Cotreatment of 1 μmol/L insulin with rosiglitazone or 3-methyladenine (3MA) prevented the insulin-induced β-secretase activation (Fig. 6A). To assess the activation of β- and γ-secretase under the HI condition, we measured β- and γ-secretase activities by means of in vitro peptide cleavage assay under the HI condition (Fig. 6B and C). Both β- and γ-secretase activities were increased markedly, and cotreatment with 1 µmol/L of insulin and rosiglitazone or 3MA attenuated the β- and γ-secretase activation (Fig. 6B and C). Consistent with activation of β- and γ-secretases under the HI condition, the condition increased Aβ generation intra- and extracellularly (Fig. 6D). Conversely, rosiglitazone and 3MA prevented these changes in Aβ generation (Fig. 6D).

To confirm that insulin-induced autophagosome accumulation induces increased Aβ generation, we transfected SH-SY5Y cells with beclin1-specific siRNA for 48 h. Beclin1 siRNA reduced total beclin1 levels by up to ∼60% without altering β-actin expression, and p62/SQSTM1 accumulated in the beclin1 knockdown condition (Fig. 6E). Consistent with the effects of 3MA, beclin1 siRNA blocked the significant insulin-induced increased Aβ generation (Fig. 6F). These data indicate that HI condition–induced insulin resistance promotes amyloidogenic pathway through accumulation of autophagosome.

To explore the mechanisms through which insulin resistance–induced autophagosome accumulation alters APP processing, we analyzed alteration in secretase protein expression by treatment with 1 μmol/L insulin. Western blot analysis revealed no difference in the expression levels of ADAM17 (α-secretase), BACE1 (β-secretase), and nicastrin (γ-secretase component) between vehicle-treated and insulin-treated cells (Fig. 7A). Previous studies have indicated that autophagosome induction with rapamycin or Leu/His deprivation increases Aβ generation due to enrichment of APP processing-related proteins in autophagosomes (10). On the basis of this consideration, we explored whether secretase proteins enriched predominantly in insulin resistance–induced autophagosomes. Using ImmunoEM labeling of insulin-treated SH-SY5Y cells, we observed that BACE1 was preferentially localized to autophagosomes (Fig. 7B). Also, by immunostaining LC3-RFP–labeled cells with nicastrin antibody, the HI condition was shown to increase colocalization on autophagosomes and nicastrin (Fig. 7C). These data demonstrate that insulin-induced autophagosome induction activates β- and γ-secretases through enrichment of secretase proteins in autophagosomes, thus increasing Aβ generation.

To confirm that the insulin-induced insulin resistance alters APP processing through autophagosome accumulation in primary cortical neurons, we established insulin resistance in primary cortical neurons (Fig. 8A). As in the SH-SY5Y cells, we cultured primary cortical neurons (DIV 14) in serum-free media in the absence or presence of insulin (1 μmol/L) for 24 h (HI condition) and found that the HI condition induced insulin resistance in primary neurons (Fig. 8A). We then observed a markedly increased autophagosome formation on primary neurons with 1 μmol/L insulin (Fig. 8B and C), indicating that the HI condition induced altered insulin signaling and increased autophagosome formation in primary neurons. Furthermore, the HI condition increased Aβ generation in primary neurons (Fig. 8D). Rosiglitazone and 3MA prevented these changes in autophagosome formation and Aβ generation (Fig. 8B and D). These data indicate that altered insulin signaling induces upregulation of autophagosome formation and that accumulated autophagosomes increase Aβ generation in primary cortical neurons.

Although several epidemiological studies have identified a relationship between AD and T2DM in the central nervous system (1,2,4), the molecular mechanisms underlying this comorbidity are not fully understood (35). In this study, we found altered insulin signaling, autophagosome accumulation, and an enhanced amyloidogenic pathway in HFD mice and in db/db mice, two widely used animal models of diabetes. The db/db mice carry a homozygous mutation in the leptin receptor and suffer severe glucose intolerance, obesity, hyperglycemia, and hyperinsulinemia (19). In contrast, HFD mice have milder metabolic symptoms, which is reversible with a low-fat diet (36). To investigate the relationship among diet-induced or genetically induced insulin resistance, autophagosome accumulation, and alteration in APP processing, we established cell culture models of insulin resistance using SH-SY5Y cells and primary cortical neurons. In insulin-resistant SH-SY5Y cells and primary cortical neurons, insulin resistance induced autophagosome accumulation via inhibition of the Akt-mTOR pathway, and this accumulation provided for greater Aβ peptide generation, resulting in accelerated Aβ generation by modulation of β- and γ-secretase activities. Finally, we found that secretase proteins enriched predominantly in insulin resistance–induced autophagosomes, thus increasing Aβ generation.

These findings point to autophagosome accumulation as at least one link between AD and T2DM. We also found that the level of IDE, an Aβ-degrading protease (37), was decreased in the brains of HFD mice, resulting in accumulation of Aβ in the brains, whereas MMP-9, another Aβ-degrading enzyme (38), was not changed (Supplementary Fig. 3). Increased Aβ levels in the brains of HFD mice may be affected by enhanced APP processing due to autophagosome accumulation and by partially decreased IDE levels. We chose SH-SY5Y cells for an in vitro study based on previous reports that showed effects of insulin on AD pathology and alteration in brain metabolism (AD-like phenotype) under diabetic conditions (39,40). When we repeated several experiments on autophagosome formation and altered APP processing with primary cortical neurons, similar results were obtained (Figs. 4 and 6 vs. 8).

An HFD-induced increase in cholesterol is reported to enhance the levels of membrane lipid rafts, which are implicated in Aβ production (41). With insulin-resistant SH-SY5Y cells, cotreatment of simvastatin or methyl-β-cyclodextrin (MβCD), cholesterol-lowering drugs, with 1 μmol/L insulin had no effect on Aβ generation (Supplementary Fig. 4). These results indicate autophagosome accumulation, not cholesterol or lipid raft, is a major factor in Aβ generation, at least in our cell culture system.

There are many reports about the role of autophagosomes in the pathology of AD and T2DM. Yu et al. (10) reported that AVs accumulated in the brains of AD model mice and that AVs were highly enriched in the components and activities of γ-secrease complexes, APP and βCTF, and were one of major sources of intracellular Aβ in the AD brain. A recent study reported that the brains of mice with streptozotocin-induced diabetes showed upregulated autophagosome and increased Aβ levels (42). Although a relation between increased Aβ levels in the brains of diabetic mice and autophagosome accumulation has not been established, there is indirect evidence that increased Aβ levels in diabetic mice are associated with autophagosome accumulation. Previous studies showed that insulin resistance–induced Aβ accumulation is caused by alteration in Aβ generation via certain insulin-signaling pathways, such as via glycogen synthase kinase-3 (GSK-3) (43), and inhibition of GSK-3 signaling decreased Aβ generation and reduced autophagosome formation (44).

There is also evidence that insulin resistance mediates pathological conditions such as oxidative stress and inflammation (5,45). These pathological conditions have been shown to induce both abnormal autophagosome accumulation and Aβ over-production (46–48). Moreover, rosiglitazone reduces the Aβ level and thus rescues memory impairment in AD model mice (49). We found that rosiglitazone reduced autophagosome accumulation by sensitizing the insulin-signaling pathway (Fig. 4C and D); thus, it is possible that decreased Aβ level by rosiglitazone is associated with reduced autophagosome accumulation. However, these reports do not provide sufficient direct evidence that increased Aβ levels in the insulin-resistant conditions are associated with autophagy activation through altered insulin-signaling pathways. We found a direct relation between increased Aβ levels and autophagosome accumulation in diabetic mouse brains and insulin-resistant cells. In particular, we established in vitro models of insulin resistance and found that insulin-resistant neurons show autophagosome accumulation and increased Aβ generation. These effects were blocked by treatment with autophagy inhibitor and by beclin1 knockdown, indicating that exacerbation of Aβ generation under insulin-resistant conditions might be associated with autophagosome accumulation.

In conclusion, this study provides in vitro and in vivo evidence that hyperinsulinemia and insulin resistance induced by genetic or dietary factors enhanced Aβ generation through autophagosome accumulation. Our findings suggest that insulin resistance–induced autophagosome accumulation might be a potential link between AD and diabetes.

This work was supported by grants from the National Research Foundation (2012R1A2A1A01002881, 2008-05943), the Medical Research Council (2011-0030738), the World Class University (R32-10084), and the Korea National Institute of Health ROAD R&D Program Project (A092058) to I.M.-J., and by the Hi Seoul Science/Humanities Fellowship from Seoul Scholarship Foundation to S.M.S.

No potential conflicts of interest relevant to this article were reported.

S.M.S. wrote the manuscript and researched data. H.S. researched data. J.B. researched data and contributed to discussion. K.S.P, H.C.J., and Y.J.P. researched animal data. I.M.-J. supervised the study and reviewed and edited the manuscript. I.M.-J. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.