Abstract
The evidence of strong pathological associations between type 2 diabetes and Alzheimer’s disease (AD) has increased in recent years. Contrary to suggestions that anti-diabetes drugs may have potential for treating AD, we demonstrate here that the insulin sensitizing anti-diabetes drug metformin (Glucophage®) increased the generation of amyloid-β (Aβ), one of the major pathological hallmarks of AD, by promoting β- and γ-secretase-mediated cleavage of amyloid-β protein precursor (AβPP) in SH-SY5Y cells. In addition, we show that metformin caused autophagosome accumulation in Tg6799 AD model mice. Extremely high γ-secretase activity was also detected in autophagic vacuoles, apparently a novel site of Aβ peptide generation. Together, these data suggest that metformin-induced accumulation of autophagosomes resulted in increased γ-secretase activity and Aβ generation. Additional experiments indicated that metformin increased phosphorylation of AMP-activated protein kinase, which activates autophagy by suppressing mammalian target of rapamycin (mTOR). The suppression of mTOR then induces the abnormal accumulation of autophagosomes. We conclude that metformin, an anti-diabetes drug, may exacerbate AD pathogenesis by promoting amyloidogenic AβPP processing in autophagosomes.
INTRODUCTION
Alzheimer’s disease (AD), a devastating form of dementia, is the most common age-related neurological disorder [1]. It is pathologically characterized by abnormal protein accumulation, including extracellular and intracellular aggregation of amyloid-β (Aβ) protein and hyperphosphorylated tau protein [1]. A large amount of literature substantiates the paradigm that most cases of AD are sporadic, and that aging is the principal risk factor [1, 2]. Although there are other potential risk factors, several studies have shown that the incidence of AD is closely associated with that of type 2 diabetes [3, 4]. For example, AD patients have an increased risk of developing type 2 diabetes [3] or symptoms of metabolic syndrome such as hypertension, hyperglycemia, low HDL level, and high triglyceride level [4, 5]. Conversely, type 2 diabetes patients show reduced cognitive function and have almost double the risk of developing AD [4, 6].
The putative major toxins in AD pathogenesis, Aβ peptides originate from amyloid-β protein precursor (AβPP), a type I transmembrane protein, via sequential enzymatic cleavages. Amyloidogenic processing occurs when AβPP is cleaved first by β-secretase (β-site amyloid-β protein precursor-cleaving enzyme 1, BACE 1), then γ-secretase (a complex including presenilin-1 (PS1), presenilin-2 (PSEN2), nicastrin, anterior pharynx-defective 1 (APH-1), and presenilin enhancer-2 (PEN-2). A second sequence, non-amyloidogenic processing, occurs when α-secretase cleavage precedes γ-secretase cleavage [7]. The regulation of AβPP processing is now a major therapeutic target for AD [8]. The β- and γ-secretases reside in one or more compartments of the vacuolar apparatus, including the endoplasmic reticulum (ER), endosomes, lysosomes, and autophagic vacuoles (AVs) [9–11]. Nixon et al. reported that AVs accumulate voluminously in the brains of patients with AD, and are co-localized with AβPP, β-secretase-derived C-terminal fragments (β-CTFs), and γ-secretase [11, 12].
The biguanide-class anti-diabetes drug metformin (Glucophage®; 1, 2-dimethylbiguanide hydrochloride) is one of the most commonly used anti-hyperglycemic agents. It seems to act by triggering AMP-activated protein kinase (AMPK), an enzyme important in metabolic regulation. Metformin enhances peripheral glucose uptake, decreases fatty acid oxidation, and activates the AMPK signal pathway [13, 14]. Metformin was expected to have a beneficial role in AD pathogenesis because of its glucose-utilizing effects. However, a previous report showed that metformin can accelerate AD progression by increasing BACE 1 activity and subsequent Aβ generation [15]. Recently, Picone et al. reported that metformin increased AβPP and presenilin expression in neuroblastoma cells, and then induced Aβ accumulation, oxidative stress, and mitochondrial dysfunction [16]. However, the detailed mechanism on how metformin can affect the risk factors of AD has not been fully investigated.
Autophagy is an intracellular degradation process that involves lysosomal fusion [17], and is induced as needed to degrade cellular compartments, protein aggregates, or long-lived organelles for recycling, as well as to protect against apoptosis during nutritional starvation or other types of cellular stress [17, 18]. Autophagosome formation is induced by the inhibition of mammalian target of rapamycin (mTOR). Autophagosomes and their contents undergo clearance by fusing with protease-enriched lysosomes. This process, named autophagic flux, is disrupted in AD brains [19, 20]. Nixon et al. found notably impaired autophagy with accumulation of various undigested or partially digested substrates in AD brains [20]. Toxic accumulation of proteins in the cytoplasm is known to be involved in neuronal death [21], and accumulated Aβ peptides are particularly neurotoxic because they impair lysosomal function and autophagy. In fact, the AVs that accumulate due to impaired lysosomal proteolysis become novel sites of Aβ formation with extremely high γ-secretase activity [11, 21].
In this study, we first aimed to elucidate the effects of metformin on γ-secretase activity, to reinforce previous work indicating that metformin promotes amyloidogenic AβPP processing in vivo [15]. We then sought to determine the cause of the high γ-secretase activity, and determine the mechanism through which metformin increases Aβ production. Finally, we tested the hypothesis that metformin activates the autophagic pathway via phosphorylation of AMPK and subsequent suppression of mTOR.
MATERIALS AND METHODS
Mice
Transgenic mice (Tg6799) that overexpress both human AβPP 695 with the Swedish (K670N, M671L), Florida (I716V), and London (V717I) mutations; and mutated human α (M146L and L286V), were generated. Heterozygous Tg6799 mice were intercrossed with B6/SJL mice (Taconic Farms Inc., NY, USA) to generate the desired genotypes. Mice were genotyped by PCR analysis of tail DNA; only females were used in the experiments. Age-matched female negative mice from similar crosses were used as wild-type controls. Animal care and procedures were performed in accordance with the Laboratory Animal Care Guidelines approved by Seoul National University.
Drug injection/sample collection
Metformin (final dose 200 mg/kg) was dissolved in isotonic sodium chloride solution (200μl). Fourteen-week-old female Tg6799 (n = 7 each group) were injected daily with 200μl intraperitoneal metformin or vehicle (isotonic sodium chloride) for 9 days. Six hours after the final injection, the mice were sacrificed and their brains were quickly removed; the left hemisphere was micro-dissected and then frozen in liquid nitrogen, while the right hemisphere was fixed with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS).
Immunohistochemistry
For histological analysis, mice were sacrificed 9 days after intraperitoneal injection of metformin. Mouse brain tissue sections were prepared as previously described [22, 23] for the detection of Aβ labeling. Six brain sections were taken from each mouse, mostly around 0.14 mm anterior to bregma (frontal cortex).
Cell culture and drug treatment
The human neuroblastoma cell lines SH-SY5Y and SY5Y-C99 were cultured as previously described [19]. Cells were washed twice in warm DMEM before treatment with each drug at the indicated concentrations/for the indicated periods; only metformin was dissolved in DMEM supplemented with 10% fetal bovine serum (HyClone), 100 U/ml penicillin, and 100μg/ml streptomycin. Doses and durations were: 2, 5, or 10 mM metformin (Sigma-Aldrich, St. Louis, MO, USA) for 12 or 24 h; 20μM Compound C (6-[4-(2-Piperidin-1-ylethoxy)phenyl]-3-pyridin-4-ylpyrazolo[1,5-a]pyrimidine) (Sigma-Aldrich, St. Louis, MO, USA) for 1 h (pre)treatment; 5 mM 3MA (3-methyladenine) (Sigma-Aldrich, St. Louis, MO, USA) for 1 h (pre)treatment. For cell starvation, cells were washed twice in warm DMEM and then incubated for 0.5, 1, 2, 4, or 6 h.
Antibodies
Mouse anti-human Aβ1 - 17 monoclonal antibody (6E10; Signet Labs., USA), mouse monoclonal anti-AβPP antibody, mouse monoclonal anti-PS1 loop antibody, mouse anti- Alzheimer precursor protein A4 monoclonal 22C11 antibody, mouse monoclonal anti-Alzheimer precursor protein AβPP 643-695 C-terminal fragment antibody (all from EMD-Millipore, Temecula, CA, USA), rabbit anti-Pen-1 polyclonal antibody (Oncogene, La Jolla, CA, USA), rabbit anti-nicastrin polyclonal antibody (Affinity Bioreagents, Golden, CO, USA), rabbit anti-aph-1a polyclonal antibody (Covance Research Product, Denver, PA, USA), anti-sAβPPα antibody (Signet), rabbit anti-AMPK polyclonal antibody, rabbit anti-calreticulin polycolonal antibody (Cell Signaling, Beverly, MA, USA), rabbit anti-LC3B polyclonal antibody (Cell Signaling, Beverly, MA, USA), and mouse monoclonal anti-actin antibody (Sigma-Aldrich, MO, USA) were used.
In vitro peptide cleavage assay
For measurement of γ-secretase activity, an in vitro peptide cleavage assay described previously [24] was utilized.
Luciferase reporter gene assay
Drug-treated cells were lysed in 1×Passive Lysis Buffer (Promega, Schildkrotstr, Mannheim, Germany), quantified using the BCA Protein Assay Kit (Thermo Scientific, Rockford/IL, USA), and the lysates allowed to react with added γ-secretase substrate. Cleavage by γ-secretase causes AICDGVP to translocate into the nucleus and activate luciferase, resulting in light emission. Signals were detected using a luminometer (Infinite® 200, Tecan, Austria).
BACE 1 promoter activity assay
The plasmid uBACE-2K, containing the human BACE 1 promoter region (+50 to –2100 bp) in the pGL3-Basic vector, was used to detect BACE 1 promoter activity (38). For the BACE 1 promoter assay, cells were co-transfected using Lipofectamine Plus (Invitrogen, Carlsbad, CA, USA) with the uBACE-2K and pRL-TK (Promega, Madison, WI, USA) vectors to maximize transfection efficiency. Twenty-four hours post-transfection, cells were treated with metformin for 24 h, and then lysed with 1×Passive Lysis Buffer (Promega, Schildkrotstr, Mannheim, Germany). Luciferase activity was measured using a luciferase kit (Promega) as described above.
Western blotting
Cell lysate proteins were analyzed using Tris-Glycine and Tris-Tricine SDS-PAGE protocols. Samples were prepared as previously described [25].
Aβ ELISA
Aβ concentrations were measured by sandwich ELISA with brain tissues or cell lysates as previously described [22].
Trichloroacetic acid (TCA) precipitation
Cell culture medium was centrifuged at 4,290 rcf for 5 min to remove cell debris, and then subjected to TCA precipitation (up to 10% (T6399, Sigma-Aldrich))
Electron microscopy
SH-SY5Y cells were prepared as previously described [19] for electron microscopic analysis.
Partial purification of lysosomes
Partial isolation of lysosomes was performed as previously described. Briefly, cells were resuspended in a hypotonic buffer (10 mM KCl, 30 mM Tris, pH 7.5, 5 mM MgOAc, 1 mM β-mercaptoethanol) and dissociated by piston strokes. Homogenates were centrifuged at 1,000 X g to precipitate the nuclei. Supernatants were collected and centrifuged at 100,000 X g for 1 h at 4°C.
Statistics
All data were expressed as mean±SEM. Differences between groups were examined for statistical significance by unpaired t-test for a single comparison, or by 1-way ANOVA followed by Dunnett’s test for multiple comparisons (for more than 2 conditions). All statistical analyses were performed using GraphPad Instat 3.0 (GraphPad Software Inc., San Diego, CA). n.s., p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001.
RESULTS
Metformin increases Aβ plaque formation and Aβ42 generation in Tg6799 mice
As metformin has been reported to increase Aβ levels [15], we investigated whether metformin increases the formation of Aβ plaques in vivo. In Tg6799 mice, Aβ starts to deposit at 2 months of age, and is initially localized to the frontal cortex in the younger mice [26]. Here, metformin was administered to the Tg6799 mice beginning at 14 weeks of age (n = 7 each group). We examined and compared biotin-4G8-stained Tg6799 brain sections from the cortices of metformin- and vehicle-treated mice (Fig. 1A, B). We found that more Aβ is accumulated in the frontal cortices of metformin-treated Tg6799 mice than in those of vehicle-treated mice (Fig. 1A). In addition, metformin-treated mice had significantly more intraneuronal Aβ and extracellular plaques in terms of the mean percentage of labeled area (Fig. 1B). We next performed ELISA to measure the levels of Aβ42 in whole mouse brains (Fig. 1C, D). Metformin treatment increased both formic acid- and detergent-soluble Aβ42 by about 1.5 fold relative to those of vehicle-treated mice, respectively (Fig. 1C, D). Taken together, these data demonstrate that metformin increased not only Aβ plaque formation in the frontal cortex of Tg6799 mice but also Aβ42 in the whole brain.
Metformin increases both β- and γ-secretase cleavage of AβPP in SH-SY5Y cells
Since metformin elevated Aβ levels in the SH-SY5Y human neuroblastoma cells (Fig. 2A), to investigate the mechanism of metformin induced Aβ increase, we checked the difference in amyloidogenic processing related activities on both β- and γ-secretases in metformin treated group. By using the β-promoter assay and measuring sAβPPβ (soluble extracellular fragment of AβPP generated by β-secretase) levels, we found that metformin increased both β-secretase activity (Fig. 2B) and sAβPPβ levels in the media in a dose dependent manner, relative to vehicle-treated cells (Fig. 2C). To investigate the effect of metformin on γ -secretase activity, we used an in vitro peptide cleavage assay and luciferase reporter assay (Fig. 2D, E). We found that metformin also activated γ-secretase in addition to β-secretase. However, full length AβPP was relatively unchanged while AβPP-CTF levels showed the decreased tendency, indicating that metformin elevated Aβ by facilitating amyloidogenic cleavage, not by changing AβPP protein levels (Fig. 2F). Next, we performed immunoblotting to examine whether this effect of metformin was due to any changes in the protein expression of the γ-secretase complex (Fig. 2G), and found no changes in the expression levels of the γ-secretase complex components (Nicastrin, APH-1, PS1, and Pen-2). Taken together, these findings indicate that metformin increases β- and γ-secretase cleavage of AβPP in SH-SY5Y cells, without affecting the production of either the secretases or the precursor protein.
Metformin activates autophagy in SH-SY5Y cells
Because Aβ production has been linked to AVs, and metformin has been linked to increased autophagy [11, 27], we sought to determine the effect of metformin on autophagy in our SH-SY5Y cells. First, we checked the level of microtubule-associated protein 1A/1B-light chain 3 (LC3) I and II with immunoblotting, to identify the degree of autophagic induction. When autophagy becomes autophagic vesicles, LC3 I is cleaved and then, converted to LC3 II through lipidation of LC3 I. Metformin increased LC3 II relative to vehicle in SH-SY5Y cells, whereas the 3MA (inhibitor of autophagic induction) induced no changes in either LC3 I or II levels compared to vehicle (Fig. 3A). Electron microscopy was used to visualize metformin-induced autophagic activation. Autophagosomes accumulated significantly more in the metformin treated group than in the vehicle treated group (Fig. 3B). Taken together, these data suggest that metformin is involved in Aβ production by excessively increasing autophagy, a mechanism which may at least partly underlay the noted enhancement of β- and γ-secretase activity.
Metformin increases autophagy and promotes amyloidogenic processing via AMPK signaling
We hypothesized that Aβ production induced by metformin utilized AMPK activation, given the previous work indicating AMPK activation and subsequent mTOR inhibition following metformin treatment [27, 28]. To test this hypothesis, we treated SH-SY5Y cells with metformin plus compound C (CC; a specific AMPK inhibitor), or metformin plus 3-MA (an autophagy inhibitor), and used immunoblotting to check the levels of phosphorylated AMPK. Metformin alone induced AMPK activation, whereas co-treatment with CC blocked the effect (Fig. 4A). However, treatment with 3MA and metformin did not recover AMPK activation, implying that metformin-induced changes in AMPK signaling and autophagy might in part be downstream of the early AMPK signaling pathway steps. In addition, CC treatment blocked the metformin-induced sAβPPβ generation (Fig. 4B) described above. On the other hand, treatment with 3-MA incompletely blocked metformin-induced sAβPPβ production (Fig. 4B). The elevation of γ-secretase activity and Aβ levels were also blocked by co-treatment of metformin and CC (Fig. 4C, D). To summarize, these data indicate that metformin-induced autophagy requires AMPK activation in SH-SY5Y cells. Finally, we investigated whether the accumulated AVs contribute to the additional Aβ generation. The levels of AβPP, BACE 1, PS1-CTF, and Nicastrin in the autophagosome- and lysosome-rich fraction (indicated as “AP & LY enrich” in Fig. 4E) were analyzed with immunoblotting. After metformin treatment, only BACE 1 protein was increased in the total fraction, whereas AβPP, BACE 1, and nicastrin were all increased in the AP- & LY-enriched fraction. The increase in these components indicates that AVs increased by metformin provide a potent site for AβPP processing and subsequent Aβ generation.
DISCUSSION
BACE 1 and γ-secretase are two key enzymes in the amyloidogenic AβPP processing pathway, in which potentially neurotoxic Aβ is produced. Accumulation of this neurotoxic Aβ species is a key pathological feature, and possibly the cause, of AD. BACE 1 and γ-secretase are upregulated in neurons in response to various kinds of cell stressors, including oxidative stress, ischemic stress, and inflammation. Since these stressors tend to increase with age, it is not surprising that age is a highly significant risk factor for AD [29–32].
The association between type 2 diabetes and dementia including AD condition has been persistently suggested despite several confounding factors such as, obesity and vascular diseases. For instance, streptozotocin induced insulin deficiency in Tg6799 mouse significantly increased AβPP and BACE1 activity thereby enhancing the concentration of cerebral Aβ peptides [33]. In addition, 15 weeks of alloxan treatment induced not only symptoms of diabetes including high glucose level but also AD-like pathology in both cortex and hippocampus of rabbits [34]. These results show the direct association between diabetes and AD by suggesting the evidences of strong correlation between insulin deficiency and amyloidogenic processing. Based on clinical and experimental evidence linking type 2 diabetes and AD, metformin has been highlighted as a potential therapeutic agent for AD in that these two diseases share common causes and symptoms. Research has shown that type 2 diabetes and AD have common abnormalities such as disrupted glucose metabolism and altered insulin signaling. [35], and patients with either disease are at significantly higher risk for the other [3]. Since metformin can cross the blood-brain barrier, its function in the central nervous system (CNS) has garnered interest. However, reports conflict on whether metformin is beneficial or detrimental to CNS function, especially where AD is concerned. For example, Kickstein et al. showed that metformin significantly reduced tau phosphorylation blocking the mTOR signaling pathway and upregulating phosphatase 2A activity in primary neuronal cell lines from mouse and human tau transgenic mice [36]. In addition, another report indicated that chronic metformin treatment in a diabetes model mouse improved AD-like pathology in the mouse, reducing phosphorylation of tau and the Aβ plaque burden [37]. However, metformin did not attenuate the impairment of spatial memory and learning, or long-term hyperglycemia. Furthermore, several recent studies have also suggested metformin exacerbates AD pathology. For instance, Chen et al. showed that metformin upregulates Aβ production in primary neuronal cultures and N2a695 cells, which overexpress AβPP [15]. Although this has yet to be replicated in other animal models, these data warn us to consider the risk of facilitating AD progression in an already high-risk group (patients with type 2 diabetes). However, the underlying mechanism through which metformin induces Aβ overproduction is still not clear. Although Chen et al. suggested metformin enhanced BACE 1 activity via AMPK signaling [15], the connection between AMPK and BACE 1 had not been identified. Thus, here we tried to reveal the specific mechanism linking AMPK signaling and increased amyloidogenesis (Fig. 5). Our hypothesis, that upregulation of AMPK in AVs caused increased amyloidogenesis, was based on two previous reports. First, it was noted that metformin activated AMPK signaling and promoted autophagy [38]. Second, it was found that AVs are a previously unidentified site of Aβ production, when Son et al. reported that insulin resistance-induced autophagosome accumulation was correlated with enrichment of AβPP and amyloidogenic secretases [39]. Here, we first confirmed that Aβ plaque formation and Aβ42 generation are increased in 3.5-month-old, Tg6799 transgenic AD model mice (Fig. 1). Then, we demonstrated that metformin causes abnormal AV accumulation (Fig. 3). Next, we showed that inhibition of AMPK blocked the metformin-induced amyloidogenesis and its associated features (e.g., autophagosome accumulation, β- and γ-secretase activation) (Fig. 4). Since the increase in autophagosomes appeared to mediate the increase of both β- and γ- secretase activities, we expected that cotreatment with metformin and 3MA would fully inhibit the metformin-induced sAβPPβ increase (Fig. 4B). However, there was still a slight increase of sAβPPβ in the 3MA-treated group relative to the vehicle-treated group (Fig. 4B), despite Aβ levels being significantly reduced in the same group (Fig. 4D). Thus, it is possible that 3MA alone might be not enough to block the upregulation of autophagy, but could be enough to reduce AV accumulation and stabilize γ-secretase activity at a level similar to that of the vehicle group.
Finally, we demonstrated that AβPP, BACE 1, and Nicastrin levels were increased in the autophagosome- and lysosome-enriched fraction of metformin-treated cell lysates (Fig. 4E). However, total AβPP and γ-secretase levels were not changed by metformin administration, suggesting that metformin alters Aβ via abnormally increased autophagy, and not by a direct effect on expression of AβPP or γ-secretase, which was also shown in Chen et al.’s previous study. Metformin did increase BACE 1 transcription and total BACE 1 levels (Figs. 2A, 4E), though. If this was indeed even partially mediated by AMPK activation [15], it would support our hypothesis of a link between metformin, AMPK, and Aβ production. However, there have been many studies reporting the beneficial effects of activation of AMPK and autophagy in AD. Spilman et al. reported that since most Aβ is cleared by the autophagy-lysosomal system, Aβ clearance can be accelerated and Aβ plaque can be decreased when activation of AMPK and following inhibition of mTOR activity can induce autophagy [40]. In addition, Yang et al. also suggested that autophagy activation by AMPK activator or mTOR inhibitor could attenuate pathogenesis of AD transgenic mice [41]. Therefore, the effects of AMPK and autophagy activation on AD are still controversial and needed to be further investigated.
Despite these controversies, our findings still suggest a modicum of caution in the use of metformin, which is currently the most widely prescribed drug for type 2 diabetes. In fact, recent clinical studies have reported on the possible risk involved. For example, Imfeld et al. reported that long-term metformin use increased the risk of AD among 14,172 patients over 65 years old [42]. In addition, another case-control study suggested that patients with type 2 diabetes taking metformin have over two to three times more severely impaired cognitive function than similar patients not taking metformin [43]. Interestingly, the association between metformin and impaired cognitive function was weakened when patients were also given vitamin B12 supplements, indicating that a metformin-induced B12 deficiency might contribute to the effects of metformin on cognitive function, at least in part. Thus, it is possible that vitamin B12 supplementation in patients with type 2 diabetes would reduce the risk of metformin monotherapy, allowing patients to maintain intact cognitive function. Notably, calcium supplements also improved cognitive performance in the above patient group. Another cohort study reported that co-treatment with sulfonylureas and metformin alleviated the risk of dementia by up to 35 % over 8 years among patients with diabetes [44], but like many cohort studies, it suffered from drawbacks including insufficient information on other patient medications, different durations or dosages of metformin, and (specific to this case) a lack of APOE gene polymorphism information. Finally, a recent report revealed that metformin’s action depends on individual biology – that is, there is a genetic component to the glycemic response to metformin [45]. Therefore, a patient’s genetic background should also be considered when seeking to identify the presence of metformin-induced cognitive impairment.
In conclusion, too many confounds exist at present to definitively establish the relationship between metformin and AD, and further investigation will be required to solidify or refute this relationship in humans. However, if the previously-mentioned variables are carefully considered during the analysis of metformin’s effect on AD, a better picture of both its positive and negative effects could be obtained that would allow clinicians to more efficiently and safely treat patients with type 2 diabetes with metformin, while preventing or mitigating the loss of cognitive functions.
