Autophagy induction by Bexarotene promotes mitophagy in Presenilin 1 familial Alzheimer’s disease iPSC-derived neural stem cells

Abstract

Adult neurogenesis defects have been demonstrated in the brains of Alzheimer’s disease (AD) patients. The neurogenesis impairment is an early critical event in the course of familiar AD (FAD) associated with neuronal loss. It was suggested that neurologic dysfunction in AD may be caused by impaired functioning of hippocampal neural stem cells (NSCs). Multiple metabolic and structural abnormalities in neural mitochondria have long been suspected to play a critical role in AD pathophysiology. We hypothesize that the cause of such abnormalities could be defective elimination of damaged mitochondria. In the present study, we evaluated mitophagy efficacy in a cellular AD model, hiPSC-derived NSCs harboring the FAD-associated PS1 M146L mutation. We found several mitochondria respiratory chain defects such as lower expression levels of cytochrome c oxidase (Complex IV), cytochrome c reductase (Complex III), succinate dehydrogenase (Complex II), NADH:CoQ reductase (Complex I) and also ATP-synthase (Complex V), most of which had been previously associated with AD. The mitochondria network morphology and abundance in these cells was aberrant. This was associated with a marked mitophagy failure stemming from autophagy induction blockage, and deregulation of the expression of proteins involved in mitochondrial dynamics. We show that treating these cells with autophagy-stimulating drug bexarotene restored autophagy and compensated mitochondrial anomalies in PS1 M146L NSCs, by enhancing the clearance of mitochondria. Our data support the hypothesis that pharmacologically-induced mitophagy enhancement is a relevant and novel therapeutic strategy for the treatment of AD.

Background

Alzheimer’s disease (AD) is the most prevalent neurodegenerative disease and the leading cause of dementia characterized by several histopathological markers including extracellular-amyloid plaques, neurofibrillary tangles within neurons, the loss of synaptic connections manifested as brain atrophy and neuroinflammation [1, 2]. Early-onset, dominantly inherited forms of AD (familiar AD, abbreviated as “FAD”) are associated with more than 200 mutations in the amyloid precursor protein [3] or the presenilin 1 and 2 genes (PS1/2) [4]. Of those, mutations in the PS1 gene, located on chromosome 14, occur most frequently in FAD. Defective PS1 interferes with the function of the γ-secretase complex, which alters the processing of the APP and leads to the overproduction of a longer, toxic version of Aβ peptide (Aβ42) that is more oligomerogenic [5]. Generation of toxic Aβ oligomers likely lead to synaptic dysfunction and ultimately the death of neurons.

Adult neurogenesis defects have been described in the brain of patients and in animal models of AD [6, 7] and the neurogenesis impairment is an early critical event in the course of FAD than the onset of hallmark lesions or neuronal loss, at least in mouse models of FAD [8]. It is hypothesized that neurologic dysfunction associated with AD may be caused by impaired functioning of hippocampal NSCs [9]. Reduction or dysfunction of PS1 may play a role in neurogenesis [10–12]. It was shown that the deposition of Aβ itself may disrupt this process [13, 14] suggesting, at least partially, a γ-secretase-dependent role of PS1 in neurogenesis [15]. Therefore, strategies to improve NSCs survival and function may be beneficial for the treatment of AD.

Mitochondria play multiple important roles in maintaining cell survival, cell death and cellular metabolic homeostasis. Several studies have demonstrated the crucial role of mitochondrial function in NSC survival and differentiation [16–19]. It was reported that mitochondrial damage increases the levels of mitochondrial ROS production, which then suppress neurogenesis through mechanisms which include Sirt1 oxidation as well as cell death of NSCs via necrotic or apoptotic signaling activation [20]. Furthermore, mitochondrial dynamics can act as a regulatory factor in the developmental program of NSCs [21]. Mitochondrial dysfunction is believed to play a major role in development of AD [22]. However, little is known about the role of mitochondria quality control pathways, such as mitophagy in connecting NSCs mitochondria abnormalities and neurogenesis. Mitophagy is the selective degradation of mitochondria by autophagy. It promotes turnover of mitochondria and prevents accumulation of dysfunctional mitochondria which can lead to cellular degeneration [23]. Parkin-dependent mitophagy is one of the best studied mechanisms for mitophagy in mammalian cells [24]. After mitochondrial damage, PTEN-induced putative kinase 1 (PINK1) is stabilized in the mitochondrial membrane. In turn, Parkin is recruited to the mitochondria [25, 26], where Parkin mediates the ubiquitination of mitochondrial proteins VDAC1 [27] and mitofusins [28]. This results in the recruitment of autophagy adapter proteins, such as sequestosome 1 (SQSTM1/p62), optineurin (OPTN) or NDP52 that mediate the cargo engulfment into autophagosomes by the interaction with MAP1LC3/LC3 (microtubule-associated protein 1 light chain 3) to finally fuse with the lysosomes to degrade their content [29]. The AD-related mitophagy impairment was previously suggested by the accumulation of mitochondria in AD patients [30–33].

Retinoids regulate the expression of numerous genes involved in different cell processes mediated by two classes of receptors, retinoic acid receptors (RARs) and retinoid X receptors (RXRs). Defective retinoid signaling is well demonstrated in the pathology of AD [34]. Retinoids appear to be promising therapeutic targets for AD treatment through the potential benefits in amyloid modulation, inflammation’s suppression, neurotransmission and neurogenesis. Bexarotene, a synthetic product structurally similar to retinoic acid compounds, is a retinoid X receptor (RXR) agonist approved by the FDA for all stages of cutaneous lymphoma [35]. In addition to targeting the cancer related genes involved in different cellular mechanisms, this antineoplastic agent stimulated protein biosynthesis and improved the expression of genes related to mitochondrial bioenergetics [36]. Bexarotene has been shown to be beneficial in several models of neurodegenerative diseases including Parkinson’s disease, Hungtinton’s disease, amyotrophic lateral sclerosis and epilepsy [37–39]. In case of AD, several laboratories demonstrated the clearance of soluble Aβ by apolipoprotein E accompanied by reversing memory deficits in a mouse model [40, 41] and AD patients [42, 43]. Furthermore, bexarotene also counteracted the effect of apoE4 on the levels of hyperphosphorylated tau in AD mouse model [44].

In the present work, we evaluated mitophagy failure in iPSC-derived NSCs harboring the FAD-associated PS1 M146L mutation that had been knocked into both alleles of a control iPSC line, as compared to its isogenic control. PS1 M146L NSCs demonstrated mitochondrial accumulation due to a mitochondrial recycling failure stemming from autophagy induction blockage as well as an aberrant mitochondrial network morphology and a deregulation of the proteins involved in mitochondrial dynamics. We found that bexarotene treatment enhanced autophagy induction and restored mitochondrial network morphology in PS1 M146L NSCs, thereby increasing the removal of accumulated mitochondria. Overall, our data suggest that pharmacological mitophagy enhancement may be a promising therapeutic strategy for treating AD.

Methods

Generation of PS1 M146L Knockin iPSCs and NSC Differentiation

Isogenic iPSC lines used in this study were generated and validated by Andrew Sproul and Scott Noggle from New York Stem Cell Foundation. The 7889(S)B parent control iPSC line was reprogrammed from a control PS1 M146L family fibroblast (PS1 WT/WT; Coriell) using Sendai viral expression of the four canonical Yamanaka factors - Oct4, KLF4, SOX2 and c-Myc (Cytotune, Life), as had been done previously for the same fibroblast line with retroviral reprogramming [45], and has been described in a recent publication [46]. iPSC colonies were initially selected by morphology, passaged several times to remove transformed cells, and expanded before characterization.

The PS1 M146L was knocked into of 7889(S)B using TALENs targeting exon 5 of PS1. TALENs were designed by Dr. Dominik Paquet (Rockefeller University) and generated via FLASH assembly (Rockefeller University Core Facility and Siobhan Gregg). Multiple TALEN pairs were tested for efficacy in 293 cells by transfecting left and right TALENs in different combinations, subsequently amplifying the gDNA region surrounding PS1 exon 5, and testing for loss of the BspH1 restriction site by NHEJ repair. One of the best TALEN pairs (L3R1) was electroporated into into 7889(S)B (15 ug/TALEN; Biorad X Cell) along with 30 ug of a ssDNA template (67 bps; Integrated DNA technologies) that incorporated the PS1 M146L mutation and was plated at clonal density. After splitting, one well for each clone was used to isolate gDNA (Qiagen), amplified for PS1 exon 5, and Sanger sequenced for introduction of the mutation. Homozygous knockin clone G14 was used for further studies along with its isogenic control G4, which remained WT after undergoing targeting.

Undifferentiated iPSCs were cultured on Cultrex (Trevigen, 3432–005-01)-coated dishes and grown in feeder-free maintenance basal medium for hESCs and hiPSCs mTesR1 (StemCell Technologies, 05851) supplemented with mTeSR1 5× Supplement (StemCell Technologies, 05852) and penicillin-streptomycin (100 U/mL-0.1 mg/mL) in a 5% CO2 in a humid incubator at 37°C.For neural stem cell differentiation, iPSC colonies were dissociated into single cells by washing with PBS and adding 1 ml Accutase (Life Technologies, A1110501) and then plated onto Matrigel-coated dishes in in mTesR1 medium containing 10 gM ROCK inhibitor (Y-27632, Stemgent, 04–0012). Cells were plated at a density of 3 × 10⁵ cells/well in six well-plate and allowed to recover for 24 h. Then, media was change to PSC Neural Induction medium (Gibco, A1647801) for neural induction for 7 days and the media was change every other day. We confirmed the expression of canonical pluripotency markers of iPSC colonies by immunofluorescence (Additional file Figure 1B) and neural markers expression by qPCR and immunofluorescence analysis of NSCs (Additional file Figure 1D–E respectively).

Antibodies

The primary antibodies used were: OXPHOS (ab110413, Abcam), TOM20 (sc-11415, Santa Cruz), LC3 (B7931, Sigma for WB), LC3 (M152–3, MBL for immunostaining), pULK1 S757 (6888, Cell Signaling), p62 (610832, BD Bioscience), Ubiquitin (sc-8017, Santa Cruz), LAMP1 (sc-5570, Santa Cruz), TFEB (mbs120432, MyBiosource), Histone H3 (9715, Cell Signalling), GAPDH (ab8245, Abcam), Parkin (sc-32282, Santa Cruz), PINK1 (BC100–494, Novus), MTF2 (M6444, Sigma), OPA1 (612606, BD Bioscience), DLP1 (611113, BD Bioscience), Oct4 (09–0023, Stemgent), Nanog (4903, Cell Signaling Technologies), Sox2 (09–0024, Stemgent), SSEA4 (ab16287, Abcam), TRA 1–81 (09–0011, Stemgent), TRA 1–60 (09–0010, Stemgent) and Nestin (ab22035, Abcam). The secondary antibodies for immunoblot studies were horseradish peroxidase-conjugated anti-mouse (Jackson ImmunoResearch, West Grove, PA) or anti-rabbit (Thermo Fisher Scientific) and for immunofluorescence were anti-mouse or anti-rabbit Cyanine Cy2 or Cy3 labeled (Jackson ImmunoResearch) and anti-mouse or anti-rabbit Alexa Fluor 488 or 555 (Thermo Fisher Scientific).

Western blot analysis

The cells were lysed in RIPA buffer (Thermo Fisher Scientific) and centrifuged at 14,000 rpm at 4 °C for 10 min. Protein concentration was measured using Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific). Proteins were separated on 10 or 4–20% gradient Mini-PROTEAN TGX gels (Bio-Rad) and transferred to nitrocellulose blotting membranes (Bio-Rad). Western blot and immunoreactive proteins were developed using an enhanced chemiluminescence detection kit (Bio-Rad).

Autophagy Flux Study

NSCs were treated with 10 μM carbonyl cyanide m-chlorophenylhydrazone (CCCP; C2759, Sigma-Aldrich) for 24 h followed by an additional treatment of PBS or Baf A1 (0.1 μM) for 4 h in the presence of CCCP [30]. After the treatment, cells were lysed in Western blot buffer and immunodetection of autophagy-involved proteins was performed as described. Quantification of the autophagy flux was analyzed according to autophagy standard guidelines [3, 47]. Briefly, autophagic vesicles synthesis was calculated as the ratio between the treatment with CCCP and Baf A1 with respect to the condition without CCCP but maintaining Baf A1 treatment. Quantification of autophagic degradation ratio was obtained by the relation between the treatment with CCCP and Baf A1 and the ones without Baf A1 but maintaining CCCP treatment.

mtDNA copy number

DNA was isolated from cells with DNeasy Blood & Tissue Kit (69504, Qiagen). mtDNA content was assessed as described before [48] by quantification of a unique mitochondrial fragment relative to a single copy region of the nuclear gene RNaseP. Quantitative PCR was performed in a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) by using Thermo Scientific Maxima SYBR Green/Rox qPCR Master Mix (K0222, Thermo Scientific). MtDNA fragment was amplified using the primers: Mito-F: CACTTTCCACACAGACATCA and Mito-R: TGGTTAGGCTGGTGTTAGGG. Nuclear content was amplified using the primers: RNaseP-F: CCCCGTTCTCTGGGAACTC and RNaseP-R: TGTATGAGACCACTCTTTCCCATA.

Complex I activity measurement

The NADH:HAR (hexammineruthenium (III)) oxidoreductase activity of Complex I [49] was determined spectrophotometrically using a plate reader (SpectraMax M5, Molecular Devices, USA). Oxidation of 0.160 mM NADH (ε_346nm 6.22 mM⁻¹cm⁻¹) by 1 mM HAR was followed at 346 nm in 0.2 ml of KCl-tween standard buffer (KCl 125 mM, HEPES 20 mM, MgCl₂ 1 ml/l, 0.5 ml/l Tween20; pH 7.8) supplemented with 40 μg/ml alamethicin (to permeabilize mitochondria), 1 mM MgCl₂, 1mM cyanide, and 30 μg protein/ml mitochondria.

Isolation of nuclear and cytoplasmic extract

The nuclear extraction was prepared using an NE-PER Nuclear Cytoplasmic Extraction Reagent kit (78833, Thermo Fisher Scientific) according to the manufacturer’s instruction. Briefly, the treated cells were washed twice with cold PBS and centrifuged at 500 g for 3 min. The cell pellet was suspended in 200 μl of cytoplasmic extraction reagent I by vortexing. The suspension was incubated on ice for 10 min followed by the addition of 11 μl of a second cytoplasmic extraction reagent II, vortexed for 5 s, incubated on ice for 1 min and centrifuged for 5 min at 16 000 g. The supernatant fraction (cytoplasmic extract) was transferred to a pre-chilled tube. The insoluble pellet fraction, which contains crude nuclei, was resuspended in 100 μl of nuclear extraction reagent by vortexing during 15 s and incubated on ice for 10 min, then centrifuged for 10 min at 16 000 g. The resulting supernatant, constituting the nuclear extract, was used for the subsequent experiments.

Immunocytochemistry

NSCs were grown on sterile glass coverslips, treated as described for each experiment, followed by washing with PBS and fixing with 4% paraformaldehyde in PBS for 20 min at room temperature. After blocking with PBS containing 1% BSA, permeabilizing with 0.1% Triton X-100 and Glycine 1 M for 30 min, cells were washed with PBS and stained by indirect immunofluorescence using the antibodies described before. Samples were mounted with Prolong Gold Antifade (P-36930, Life Technologies) and randomly chosen field images were obtained in an Inverted Confocal DMI6000CS (Leica, Wetzlar, Germany) fluorescence microscope.

Quantification of Colocalization and Vesicles Average Size

Colocalization analysis was performed with ImageJ software (Bethesda, MD, USA) as it was previously described [30]. Briefly, by a threshold intensity, binary images were obtained and Image Calculator tool was used to generate an image harboring only overlapping structures. Colocalization measurement was obtained by quantifying the area occupied by the overlapping elements per cell. At least 200 cells were measured for each cell line. For the lysosomal size analysis, cells were thresholded and the average size was calculated using the particle analyzer of ImageJ software.

Mitochondrial Morphology Study

Cells were fixed with 4% paraformaldehyde and mitochondrial pattern was observed by TOM20 staining. Images were acquired with Inverted Confocal DMI6000CS (Leica, Wetzlar, Germany) fluorescence microscope. For analysis, we used the Mitochondrial Network Analysis [3] toolset, a combination of different ImageJ macros that allows the semi-automated analysis of mitochondrial networks in cultured mammalian cells [30]. Briefly, the image was converted to binary by thresholding following the conversion to a skeleton that represents the features in the original image using a wireframe of lines one pixel wide. All pixels within a skeleton were then grouped into three categories: end point pixels, slab pixels and junction pixels. The plugin analyzes how the pixels are spatially related and defined to measure the length of each branch and the number of branches in each skeletonized feature as well as the mitochondrial network morphology. The parameters used in the study were 1) Individuals, punctate, rods and large/round mitochondrial structures; 2) Networks, mitochondrial structures with at least a single node and three branches; 3) the mean number of Branches per Network; and 4) the average of length of rods/branches. 100 cells per each independent experiment were used to quantify the pattern of mitochondria.

Quantitative real-time PCR assays

RNA was isolated by using SV Total RNA Isolation System (Promega, Madison, WI, USA) following provider’s guidelines. RNA was reverse transcribed into cDNA using ImProm-II reverse transcription systems (Promega, Madison, WI, USA), which was then amplified by quantitative PCR performed in a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) by using Thermo Scientific Maxima SYBR Green/Rox qPCR Master Mix (K0222, Thermo Scientific). The primers used for the detection were TFEB-F: CCAGAAGCGAGAGCTCACAGAT, TFEB-R: TGTGATTGTCTTTCTTCTGCCG, LC3-F: GAGAAGCAGCTTCCTGTTCTGG, LC3-R: GTGTCCGTTCACCAACAGGAAG, LAMP1-F: ACGTTACAGCGTCCAGCTCAT, LAMP1-R: TCTTTGGAGCTCGCATTGG, SOX1-F: GCGGAAAGCGTTTTCTTG, SOX1-R: TAATCTGACTTCTCCTCCC, SOX2: ATGCACCGCTACGACGTGA, SOX2-R: CTTTTGCACCCCTCCCATTT, Nestin-F: CAGCGTTGGAACAGAGGTTGG and Nestin-R: TGGCACAGGTGTCTCAAGGGTAG Gene expression were normalized to human GAPDH using the primers GAPDH-F: GTCGGAGTCAACGGATT, GAPDH-R: AAGCTTCCCGTTCTCAG.

Statistical analysis

Graphs represent means and standard errors of the values obtained from, at least, experimental triplicates using control and PS1 M146L samples, which are specified in each legend to the figure. When necessary, values represented in the graphs were obtained by normalizing every PS1 M146L sample data with the correspondent control sample. Student’s t test was performed for statistical comparison of the data sets. To examine the differences between experimental factors and their interaction, two-way analysis of variance (ANOVA) test was performed. When more than two experimental groups were compared, a post hoc Bonferroni test was used. Non-parametric Mann-Whitney U test was used when the distribution of the data was not normal. The differences are given with their corresponding statistical significance or p value, which is the probability that the difference occurred merely by chance under the null hypothesis.

Results

Mitochondrial anomalies in PS1 M146L NSCs

Although, previous results demonstrate mitochondria dysfunction and oxidative stress in human NSCs related to Aβ pathology [50], further investigations are needed to understand the involvement of mitochondrial anomalies in the pathogenesis of AD. We elected to use a PS1 M146L homozygous knockin model for our study (see Methods for TALEN-mediated generation), as this would facilitate more robust AD phenotypes while still keeping PS1 under physiological expression and allowing cross comparisons with the isogenic control backbone (Additional file Figure 1). Thus, we first assayed the expression levels of the oxidative phosphorylation-related proteins (OXPHOS) by immunoblot. Decreased expression levels were detected for the subunits of ATP-synthase (Complex V), cytochrome c oxidase (Complex IV), cytochrome c reductase (Complex III), succinate dehydrogenase (Complex II) and NADH:CoQ reductase (Complex I) OXPHOS complexes in PS1 M146L NSCs, as compared to control cells (Fig. 1A–B). A downregulation of expression of mitochondrial Complex I subunits had been described in brains of early and severe AD patients [51], as well as a decrease in Complex I activity in several AD mouse models [52]. Therefore, we analyzed the activity of Complex I [53, 54] in our model and found that it was decreased in PS1 M146L NSCs (Fig. 1C). Moreover, we observed that mitochondrial DNA (mtDNA) copy number in PS1 M146L NSCs was significantly increased as compared to the control cells (Fig. 1D). In parallel, PS1 M146L cells exhibited higher levels of TOM20, a constitutive mitochondrial marker (Fig. 1E). One of the reasons to this mitochondrial accumulation could be a failure in the recycling process of mitochondria.

Figure 1. Mitochondrial anomalies in PS1 M146L NSCs.

(A-B) Representative immunoblot and quantification of mitochondrial OXPHOS respiratory complex protein levels under basal conditions. (C) Complex I NADH:hexaammineruthenium (HAR) reductase activities were determined spectrophotometrically in whole cell homogenates as described in Methods. (D) Expression levels of mtDNA copy number by qPCR of a unique mitochondrial DNA fragment relatively to a single copy region of the nuclear gene RNaseP. (E) Representative immunoblot of TOM20 protein levels of cells incubated in the presence of CCCP (10 μM) for 24 h. Quantification of TOM20 expression under basal conditions. Graphs showed means and SEM (n=5 independent experiments; * p<0.05; ** p<0.01; *** p<0.001).

Impaired autophagy in PS1 M146L NSCs

The transcription factor EB (TFEB) is a master regulator of transcription of target genes involved in autophagy and lysosome biogenesis [55]. To identify whether autophagy may be altered in our PS1 M146L NSCs, we analyzed the expression of autophagy-related markers measured by RT-qPCR. PS1 M146L cells exhibited a decreased TFEB expression and a reduction in the expression of the autophagy marker LC3 and the lysosomal marker LAMP1 (Fig. 2A). Then, we analyzed the TFEB protein levels under basal conditions in the whole lysate and we observed a significant decreased in PS1 M146L NSCs compared to control cells (Additional file Figure 2A). Due to the cellular localization and activity of TFEB are mainly controlled by its phosphorylation status [56], we studied the TFEB subcellular localization in control and PS1 M146L NSCs by the isolation of cytosolic and nuclear fraction. The results showed that TFEB localizes predominantly in the cytoplasm of PS1 M146L NSCs suggesting that a substantial fraction of TFEB is phosphorylated correlating with a reduction of dephosphorylation and activation of TFEB monitored by its significant decreased translocation to the nuclear compartment under basal conditions (Additional file Figure 2B). To determine autophagy flux, NSCs were treated with CCCP followed by a treatment of Baf A1 in the presence of CCCP to block the degradation phase of autophagy. Immunoblot analysis revealed a significantly decreased amount of autophagic vesicles (AVs) in PS1 M146L cells exhibiting a decrease in LC3II protein levels and a decreased LC3II/LC3I ratio (Fig. 2B–C) under basal conditions. Although CCCP treatment showed induction of AVs, a reduction of autophagosome synthesis in PS1 M146L cells was observed. This observation correlates with significantly smaller LC3II accumulation compared to control samples after CCCP treatment when vacuole degradation was blocked by Baf A1 (Fig. 2D). Moreover, we observed a decreased degradation of AVs in PS1 M146L NSCs (Fig. 2E), suggesting a decreased autophagy flux stemming from a deficient autophagic vacuole formation. This data correlates with the significant increase in phosphorylated ULK1 at Ser758 levels in PS1 M146L NSCs (Fig. 2F). The ULK1 kinase is the key initiator of the autophagic process in mammalian cells and its phosphorylation inactivates the kinase resulting in autophagy down-regulation [57]. To confirm this hypothesis, we examined the turnover of p62, an adaptor protein that is degraded by autophagy, under the same conditions. We observed that basal levels of p62 were notably elevated in PS1 M146L cells compared with control samples (Fig. 2G–H) as had been previously reported [58]. Moreover, these cells exhibited a remarkably lower degradation of p62 (Fig. 2I). Similar to p62, other autophagy adaptors such as optineurin (OPTN) and NDP52 also directly bind ubiquitin and LC3 to coordinate the autophagosome-mediated engulfment of cargo [59]. We found that levels of OPTN were increased, and levels of NDP52 were not altered in PS1 M146L NSCs as compared to control cells (Additional file Figure 2B–C). Consistent with these autophagy alterations, we also detected a significant accumulation of ubiquitinated proteins in PS1 M146L cells (Additional file Figure 1D). To address a possible lysosomal dysfunction, we studied the levels of lysosomal markers in PS1 M146L NSCs. We observed a decrease in the levels of LAMP1 (Fig. 2J) in parallel with a clear reduction in the average size of lysosomes in PS1 M146L cells, as compared to control cells (Fig. 2K–L).

Figure 2. Autophagy failure in PS1 M146L NSCs.

(A) Expression levels of TFEB, LC3 and LAMP1 measured by RT-qPCR in control and PS1 M146L NSCs under basal conditions. Results were normalized to GAPDH and are presented relative to control. (B) Representative immunoblot of LC3 expression for the study of autophagy flux. NSCs were treated with 10 μM CCCP for 24 h followed by an additional treatment of PBS or Baf A1 (0.1 μM) for 4 h in the presence of CCCP. (C) Quantification of LC3II levels and LC3II/LC3I ratio in the PS1 M146L samples with respect to the controls under basal conditions. (D) Quantification of LC3II synthesis as the ratio between the treatment with CCCP and Baf A1 with respect to the condition without CCCP but maintaining Baf A1 treatment. (E) Quantification of autophagic degradation ratio was obtained by the relation between the treatment with CCCP and Baf A1 and the ones without Baf A1 but maintaining CCCP treatment. (F) Representative immunoblot and quantification of pULKl S757 expression treated with CCCP (10 μM) for 24 h. (G-H). Representative immunoblot of p62 expression after the treatment as in (B) and quantification of basal levels. (I) Quantification of p62 degradation ratio as defined in (E). (J) Representative immunoblot and quantification of LAMP1 under basal conditions. (K-L) Representative confocal microscopy immunofluorescence images showing the lysosomal markers LAMP2 and quantification of lysosomal average size of control and PS1 M146L NSCs in basal conditions. Images on the right show enlargement of the squared area. Graphs showed means and SEM (n=5 independent experiments; * p<0.05; ** p<0.01). Scale bar: 20 μm.

Mitophagy alterations in PS1 M146L NSCs

The obtained data indicate a defect in autophagy in PS1 M146L NSCs. Hence, we investigated whether mitochondrial recycling could be deregulated by the PS1 mutation. PS1 M146L cells showed elevated levels of Parkin protein (Fig. 3A). On the other hand, PINK1 protein levels were highly increased in PS1 M146L NSCs compared to the controls, after the treatment with CCCP for 24 h (Fig. 3B). Parkin is recruited to impaired mitochondria with low membrane potential in mammalian cells and promotes the engulfment of mitochondria by autophagosomes and the selective elimination of impaired mitochondria [60].Thus, we studied the recruitment of Parkin to the mitochondria by immunofluorescence after the treatment with CCCP for 1 h (Fig. 3C). We observed that mitochondrial localization of Parkin was increased in PS1 M146L cells (Fig. 3D) correlating with a decreased ubiquitin phosphorylation at Ser65 and Parkin activation (Additional file Figure 3) suggesting an accumulation of damaged mitochondria that cannot be recycled by mitophagy. Accordingly, we found a reduction of mitochondria content in autophagic vesicles after CCCP treatment, as indicated by a decrease in colocalization of TOM20 and LC3II (Fig. 3E–F). Overall, these data suggest mitophagy impairment due to autophagy blockage.

Figure 3. Mitophagy impairment in PS1 M146L NSCs.

(A) Representative immunoblot and quantification of Parkin levels under basal conditions. (B) Representative immunoblot and quantification of PINK1 levels in control and PS1 M146L NSCs treated with CCCP (10 μM) for 24 h. (C-D) Representative confocal microscopy immunofluorescence images showing Parkin in red and TOM20 as a mitochondrial constitutive marker in green in the same cells treated with CCCP (10 μM) for 1 h. On the right, binary image representing the colocalization of both labels and dotted line delimits cytoplasm of each cell. (E) Representative confocal microscopy immunofluorescence images showing LC3 in red and TOMM20 in green of control and PS1 M146L NSCs treated with CCCP (10 μM) for 24 h. (F) Quantification of the colocalization between LC3 and TOMM20 expressed as area occupied by the overlapping elements per cell. Graphs showed means and SEM (n=5 independent experiments; * p<0.05; *** p<0.001). Scale bar: 20 μm.

Changes in mitochondrial dynamics in PS1 M146L NSCs

Mitochondrial dynamics is regulated by a machinery involving large dynamin-related GTPases with mitofusin 1 (MFN1), mitofusin 2 (MFN2), and optic atrophy 1 (OPA1) proteins being involved in mitochondrial fusion [61]. Mitochondrial fission is mediated by dynamin-like protein 1 (DLP1) through its interaction with four mitochondrial receptor proteins [61–63]. Previous studies revealed an impaired balance of mitochondrial fission and fusion proteins in brains of AD patients [64]. Therefore, we examined whether proteins that control mitochondrial dynamics could be altered in PS1 M146L NSCs. We observed diminished levels of fusion protein MTF2 and OPA1 in PS1 M146L cells (Fig. 4A–B). Several studies demonstrated reduced levels of DLP1 in neurons and fibroblasts obtained from AD patients [65, 66]. We observed a decrease in DLP1 fission protein level in PS1 M146L cells (Fig. 4C). DLP1 function depends on its subcellular localization [67]; therefore, we studied its distribution by immunofluorescence analysis of mitochondria labeled with TOM20. Quantitative analysis revealed an increase in DLP1 abundance in mitochondria in PS1 M146L NSCs, as compared to control cells (Fig. 4D). An aberrant mitochondrial morphology and distribution in AD had been reported earlier [65]. These data and the previously reported downregulation of mitochondrial dynamics proteins prompted us to evaluate morphological features of mitochondria. We examined the mitochondrial network skeleton in representative images of control and PS1 M146L NSCs. The network parameters were calculated by the Mitochondrial Network Analysis toolset [68] (Fig. 4E). Although, there were no differences in the number of individuals puncta and rods mitochondrial structures (Fig. 4F) and networks (mitochondrial structures with at least a single node and three branches, Fig. 4G) between control and PS1 M146L cells, we observed a remarkably diminished amount of the branches per network and the shorter length of the branches in PS1 M146L NSCs (Fig. 4H–I), which correlated with an increase in the number of fragmented mitochondria, likely as a consequence of excessive mitochondrial fission.

Figure 4. Alterations in mitochondrial network morphology in PS1 M146L NSCs.

Representative immunoblots and quantification of mitochondrial dynamics proteins MTFN2 (A) and OPA1 (B) and DLP1 (C) under basal conditions. (D) Representative confocal microscopy immunofluorescence images showing DLP1 in red and TOM20 in green and colocalization quantification of control and PS1 M146L NSCs under basal conditions. (E) Representative images showing mitochondrial morphological skeleton generated by Mitochondrial Network Analysis [3] toolset of control and PS1 M146L NSCs under basal conditions. Quantification of number of individuals (F), networks (G), branches per network (H) or length of the branches (I). Graphs showed means and SEM (n=3 independent experiments; * p<0.05; *** p<0.001). Scale bar: 20 μm.

Bexarotene treatment compensates autophagy deficiency and restores mitophagy

Recent studies have demonstrated neuroprotective potential of bexarotene in different neurodegenerative diseases including AD [37, 38, 40]. To evaluate whether bexarotene treatment of PS1 M146L NSCs promotes autophagy activation, we measured the RNA expression of autophagy-related markers TFEB, LC3 and LAMP1. We found that bexarotene-treated cells demonstrate an induced expression of these autophagy target genes (Fig. 5A). Moreover, the autophagy flux study revealed a diminished AVs content (Fig. 5B–C) correlating with an increased autophagosome synthesis and degradation in PS1 M146L cells (Fig. 5D–E; previously decreased in PS1 mutant cells (Fig. 2D–E)) due to an autophagy induction after the treatment with bexarotene. These data correlate with a reduction in p62 accumulation caused by p62 degradation induction (Fig. 5 F–H; previously increased in PS1 mutant cells (Fig. 2H)) as well as with an increase in LAMP1 protein abundance (Fig. 5 I) in PS1 M146L cells after the treatment. Together, these data suggest that bexarotene activates the autophagy flux.

Figure 5. Autophagy induction in PS1 M146L cells after bexarotene treatment.

(A) Expression levels of TFEB, LC3 and LAMP1 measured by RT-qPCR in control and PS1 M146L NSCs after the treatment with bexarotene (0.5 μM) for 24 h. (B) Representative immunoblot of LC3 expression for the study of autophagy flux after the treatment with bexarotene. After the treatment with Bexarotene (0.5 μM) for 24 h, NSCs were treated with 10 μM CCCP for 24 h followed by an additional treatment of PBS or Baf A1 (0.1 μM) for 4 h in the presence of CCCP. (C) Quantification of LC3II basal levels in the PS1 M146L samples with respect to the controls before and after bexarotene treatment. (D) Quantification of LC3II synthesis as the ratio between the treatment with CCCP and Baf A1 with respect to the condition without CCCP but maintaining Baf A1 treatment. (E) Quantification of autophagic degradation ratio was obtained by the relation between the treatment with CCCP and Baf A1 and the ones without Baf A1 but maintaining CCCP treatment. (F) Representative immunoblot of p62 expression after the treatment as in (B) and quantification of basal levels (G) before and after bexarotene treatment. (H) Quantification of p62 degradation ratio as defined in (E). (I) Representative immunoblot and quantification of LAMP1 basal levels before and after the treatment with bexarotene. Graphs showed means and SEM (n=5 independent experiments; * p<0.05; ** p<0.01).

Further, we examined whether the bexarotene treatment improves the mitochondrial recycling process. We observed that Parkin levels remained increased in PS1 M146L cells (Fig. 6A), and its targeting to mitochondria after CCCP insult and bexarotene treatment was not disturbed (Fig. 6B). Finally, bexarotene-treated PS1 M146L cells demonstrated an enhanced mitophagy, as can be judged from increased ubiquitin phosphorylation at Ser65 and Parkin activation (Fig. 6C. As a result, mitochondria recycling was increased, which was reflected by a decrease in the mtDNA copy number as well as the levels of TOM20 in bexarotene-treated cells (Fig. 6D–E). On the other hand, the treatment with bexarotene recovered the protein levels of the oxidative phosphorylation system Complex I as well as its activity suggesting an improvement of the respiratory chain function (Fig. 6F–G).

Figure 6. Bexarotene treatment compensates mitophagy deficiency in PS1 M146L cells.

(A) Representative immunoblot and quantification of Parkin basal levels after the treatment with bexarotene. (B) Representative confocal microscopy immunofluorescence images showing Parkin in red and TOM20 as a mitochondrial constitutive marker in green in control and PS1 M146L cells treated with bexarotene (0.5 μM) for 24 h followed by an additional treatment of CCCP (10 μM) for 1 h. On the right, binary image representing the colocalization of both labels and dotted line delimits cytoplasm of each cell. (C) Representative confocal microscopy immunofluorescence images showing pUB-Ser65 in green and TOM20 in red in control and PS1 M146L cells treated with bexarotene (0.5 μM) for 24 h followed by an additional treatment of CCCP (10 μM) for 1 h. (D) Expression levels of mtDNA copy number by RT-qPCR of a unique mitochondrial DNA fragment relatively to a single copy region of the nuclear gene RNaseP after the treatment with bexarotene. (E) Representative immunoblot and quantification of TOM20 basal levels after the treatment with bexarotene. (F) Representative immunoblot image and quantification of Complex I protein levels after the treatment with bexarotene. (G) Complex I NADH:hexaammineruthenium (HAR) reductase activities were determined spectrophotometrically in whole cell homogenates after the treatment of bexarotene for 24 h as described in Methods. Graphs showed means and SEM (n=5 independent experiments; * p<0.01, ** p<0.01; *** p<0.001). Scale bar: 20 μm.

Bexarotene treatment rescues altered mitochondrial network morphology

Next, we examined whether bexarotene treatment could improve the mitochondrial morphology. First, we analyzed the mitochondrial dynamics proteins levels in bexarotene-treated PS1 M146L cells. We found that MTFN2, OPA1 and DLP1 levels were restored to the levels observed in control cells (Fig. 7A–C). In order to analyze whether these changes could reflect a recovery of the mitochondrial morphology, we examined the mitochondrial network skeleton by MiNa after the treatment with bexarotene (Fig. 7D). The treatment increased the length of the branches in PS1 M146L cells (Fig. 7 F–I). These data strongly suggest that bexarotene treatment rescued mitochondrial network morphology and distribution.

Figure 7. Mitochondrial morphology restoration in PS1 M146L NSCs after bexarotene treatment.

[3] Representative immunoblots and quantification of mitochondrial dynamics proteins MTFN2 (A) and OPA1 (B) and DLP1 (C) levels after the treatment with bexarotene. (D) Representative images showing mitochondrial morphological skeleton generated by MiNa toolset of control and PS1 M146L NSCs after the treatment with bexarotene. Analysis of number of individuals (F), networks (G), branches per network (H) or length of the branches (I). Graphs showed means and SEM (n=5 independent experiments; * p<0.05; ** p<0.01). Scale bar: 20 μm.

Discussion

Large numbers of PS1/PS2 mutations are linked to FAD [69]. These PS1/PS2 mutations compromise the essential physiological roles of presenilin in memory, synaptic function and neuronal survival, eventually leading to neurodegeneration and dementia [70, 71]. Prior studies of AD pathogenesis had been focused on the degeneration of existing neuronal circuits. However, deciphering the physiological state of NSCs is the key for the understanding AD development. It was shown that physiological levels of expression of a single PS1 M146V allele is sufficient to impair environmental enrichment-induced adult hippocampal NSCs proliferation, survival and neuronal differentiation, in vivo [72], The PS1 M146L mutation is a prevalent and most characterized FAD mutation that shows typical phenotypes of AD [73–75]. It appears that PS1-mutated isogenic hiPSC-derived NSCs that we employed in this study are a useful tool to examine the alterations in FAD progenitors because it is a consistent and fully controllable cell model relevant to the study of AD with the benefit to proliferate the cells in the amounts sufficient for a screening of new drugs.

In the present work, we have analyzed the mitochondrial recycling process in a human neural cell model, generated from iPSC harboring the PS1 M146L mutation. We have demonstrated a deregulation of autophagy as a result of a deficiency in autophagy induction which is manifested by the decreased expression of autophagy-related markers, such as TFEB expression and its nuclear translocation as well as lower levels of autophagic vesicle (LC3) formation (Fig. 2 A–E). This data correlates with the accumulation of ubiquitinated proteins and the proteins degraded by autophagy (p62) (Fig. 2G–H). Earlier, an accumulation of p62 has been observed in neuronal and glial ubiquitin-containing inclusions [76] as well as in neurofibrillary tangles [77] in AD brains. We also found lysosomal anomalies in PS1 M146L NSCs manifested by an abnormally decreased size of lysosomes (Fig. 2K–L). This data is in line with previous findings where lysosomal function was compromised by a deficit in lysosomal acidification [58]. In that study, PS1 was demonstrated to act as chaperone protein that facilitates the N-glycosylation of V-ATPase subunit V0a1, which help V-ATPase traffic to lysosome and complete lysosome acidification. These results combined with the reduced expression of subunits of respiratory chain complexes and ATP synthase as well as reduced activity of Complex I in PS1 M146L cells correlates with recent report where chronic mitochondrial respiratory chain dysfunction results in a repression of TFEB/MITF transcriptional activity and lysosomal biogenesis [78]. Overall, these data suggested that PS1 mutation blocks autophagy induction in PS1 M146L NSCs. The data obtained in our study confirms this hypothesis.

Our findings also demonstrated that mitophagy triggered by mitochondrial insult is severely affected resulting in mitochondrial accumulation in our PS1 M146L cells (Fig. 1D–E). These cells showed marked accumulation of Parkin and PINK1 (Fig. 3A–B). This had been previously attributed to an accumulation of depolarized mitochondria unable to be degraded by autophagy [25, 79]. The Parkin accumulation that we have observed in PS1 mutant cells is in line with earlier reports demonstrating that an accumulation of insoluble Parkin, intraneuronal Aβ, and phospho-MAPT was detected in AD brain autophagosomes, indicating autophagy failure [80]. Additionally, the upregulation of Parkin expression had been suggested to serve as a defense mechanism counteracting AD –associated stress-induced damage [81]. Further, upon mitochondrial depolarization, Parkin (despite its correct translocation to the mitochondria) does not correctly activate and properly target mitochondria to autophagic vesicles, correlating with the autophagy blockage in PS1 M146L NSCs. To note, our finding agrees well with published results obtained with 5xFAD mice exhibiting drastically increased, age-depedent Parkin translocation to synaptosomal mitochondria [82]. More recently, Bohr’s group showed alterations in mitochondrial function and changes in the immunoreactivity of mitophagy-associated proteins in both brain biopsies and iPSC-derived neurons from FAD AD patients [83]. They also showed that Aβ inhibits mitophagy while mitophagy induction reduces Aβ in the hippocampus of transgenic APP/PS1 AD mice. In correlation with our findings, the authors suggested an impaired initiation of selective mitochondrial autophagy, due to decreased levels of activated mitophagic proteins, resulting in the accumulation of dysfunctional mitochondria and impaired cellular energy metabolism. They demonstrated that defective mitophagy, which can be caused by both Aβ and p-tau, is a major element of AD progression and memory loss in a manner that is conserved from C. elegans and mice to humans. Accumulation of Parkin in mitochondria, increased LC3II levels and autophagic vesicles containing altered mitochondrial structures had also been found in hippocampus of AD patients [30, 84] and in some cellular models [31] thereby confirming the AD relevance of our findings in PS1 M146L NSCs.

Mitochondrial dynamics can act as a functional regulatory point in coordinating a cellular redox state-attributed nuclear developmental program in NSCs [21]. It is known that a deregulation of mitochondrial dynamics and its imbalance toward a chronically fragmented state, such as in AD [85], impairs the self-renewal capacity of stem cells and leads to a depletion of the stem cell pool. In the present work, we describe an impaired balance of mitochondrial fission and fusion proteins and mitochondrial morphology in PS1 M146L NSCs (Fig. 6) as it was previously described in AD patients [64]. Reduction in mitochondrial length is consistent with unopposed fission suggesting an altered mitochondrial dynamics in AD neurons [65, 86]. We should also note that in experiments with primary hippocampal neurons treated with Aβ-derived diffusible ligands (ADDLs), it was found that mitochondria in neurons are shortened and exhibit altered levels of fission and fusion proteins [64]. Similarly, mitochondria in pyramidal neurons of AD patients where abnormally redistributed away from axons [64]. We observed that PS1 M146L NSCs exhibited an abnormal mitochondrial morphology resulting in an increased mitochondrial fragmentation, with a prominent shift toward mitochondria fission as a characteristic AD feature [85]. An increase in mitochondrial network fission may result in an increased abundance of dysfunctional mitochondria and increased ROS production [87]. It is also known that increased mitochondrial fragmentation alters the metabolic and respiratory profileof NSCs due to its upstream regulation by mitochondrial dynamics [21].

Considering the above, we evaluated whether the treatment with bexarotene would be able to reverse the observed mitophagy and mitochondrial morphology abnormality in NSCs. Bexarotene is an FDA-approved retinoid X receptor (RXR) agonist for all stages of cutaneous T-cell lymphoma [35]. Currently, it is in clinical trials in AD patients based on previous preclinical trial works in an AD mouse model [40, 88, 89]. These studies indicate the potential therapeutic efficacy of bexarotene stemming from increased Aβ clearance by PPARy activation [90], an improved neural network function, and a reversal of behavioral deficits in mice [44, 91, 92]. In this study, we have evaluated the effect of bexarotene on the recovery of the autophagy impairment in our model of PS1 M146L NSCs. Bexarotene improved autophagic induction consistent with the upregulation of the TFEB, master regulator of autophagy, and the increased LC3 synthesis and degradation (Fig. 5A–E). These data correlated with the reduction of p62 levels as a result of the enhanced autophagy flux (Fig. 5F–H). Previous findings demonstrated that bexarotene enhanced autophagy and protected against thromboembolic stroke in aged transgenic mice expressing the human P301L-Tau mutation [93]. Moreover, a recent study demonstrated that bexarotene promoted neuroprotection by the regulation of PPARδ restoring affected mitochondria respiration and quality control homeostasis in a model of Hungtintons disease [94]. It is shown that activation of PPARδ is also neuroprotective in a mouse model of AD through inhibition of inflammation [95]. We observed that bexarotene improves mitophagy as reflected by the recruitment of Parkin to mitochondria and its activation, to restore mitophagy (Fig. 6B–C). This decreased the abnormal accumulation of mitochondria (TOM20) (Fig. 6D–E) under basal conditions and after CCCP treatment, by the increased of autophagy vesicle synthesis. We also observed that bexarotene restored the expression of some Complex I subunit and its activity in PS1 M146L cells (Fig. 6F–G). Moreover, bexarotene treatment induced an upregulation of the proteins involved in mitochondrial dynamics, which correlated with mitochondrial network morphology rescue and distribution (Fig. 7). It has been postulated that bexarotene acts by stimulating both LXR:RXR and PPARγ:RXR transcriptional activity [96, 97]. However, further studies are needed to determine the molecular pathway involved in the upregulation of autophagy and energetic changes in AD models.

Conclusions

In summary, our findings indicate that iPSC-derived PS1 M146L NSCs harboring PS1 M146L mutation demonstrate a marked mitophagy failure associated with an autophagy induction blockage. We describe a deregulation of the expression of proteins involved in mitochondrial dynamics as well as an aberrant mitochondrial network morphology, suggesting an impairment in mitochondria distribution and degradation. We also show that bexarotene treatment restores autophagy and can compensate mitochondrial anomalies in PS1 M146L NSCs, by enhancing the clearance of mitochondria. Our data support the hypothesis that pharmacologically-induced mitophagy enhancement is a relevant and novel therapeutic strategy for the treatment of AD.

Abbreviations

AD

Alzheimer’s disease

Aβ

Amyloid-beta

APP

Amyloid beta (A4) precursor protein

AVs

Autophagic vacuoles

Baf A1

Bafilomycin A1

CCCP

Carbonyl cyanide m-chlorophenylhydrazone

DLP1

Dynamin-like protein 1

FAD

Familial Alzheimer’s disease

GAPDH

Glyceraldehyde-3-phosphate dehydrogenase

iPSC

Induced pluripotent stem cells

HAR

hexammineruthenium (III)

LAMP1

Lysosomal-associated membrane protein 1

MAP1LC3/LC3

Microtubule-associated protein 1 light chain 3

mtDNA

Mitochondrial DNA

MFN1

Mitofusin 1

MFN2

Mitofusin 2

NPCs

Neural progenitor cells

OPA1

Optic atrophy 1

OPTN

Optineurin

OXPHOS

Oxidative phosphorylation system

PSEN

Presenilin

PINK1

PTEN-induced putative kinase 1

ROS

Reactive oxygen species

RXR

Retinoid X receptor

TFEB

Transcription factor EB

TOMM20

Translocase of outer mitochondrial membrane 20 homolog