Background
There are nearly 47 million people living with dementia worldwide, which is predicted to double every 20 years, increasing to more than 131 million by 2050. Alzheimer’s disease (AD) is the best characterized among them, and it accounts for 50–60% of all dementia cases [
1]. This common neurodegenerative disease is clinically characterized by a progressive and gradual cognitive impairment, synapse loss, and substantial loss of neurons in later stages. Owing to the disease’s heterogeneity, the etiology of AD is still not very well understood. Most cases of early-onset AD are linked to autosomal dominant inherited mutations in the genes encoding amyloid precursor protein (
APP), presenilin 1 (
PSEN1), and presenilin 2 (
PSEN2). These cases are referred to as familial Alzheimer’s disease (fAD) and are well characterized. In contrast, the etiology of the remaining 95% cases of late-onset AD, often referred to as sporadic Alzheimer’s disease (sAD), requires further investigation owing to the various factors involved in the pathology, including genetic and environmental exposures [
2]. Moreover, cellular changes in the brain precede the first clinical symptoms by 10–15 years, and there is a lack of early diagnostic biomarkers for the prodromal stages of AD. Currently, there is no cure for AD, and the available medications can only slow down the progression of dementia and slightly improve the quality of life of the patients [
3].
Two well-defined pathological hallmarks of AD have been described: the formation of extracellular amyloid plaques and the development of intracellular neurofibrillary tangles (NFTs) formed by aggregated hyperphosphorylated TAU protein [
4]. The amyloid plaques are composed mainly of Aβ and form aggregates of regularly ordered amyloid fibers [
5,
6]. Aβ peptides are derived from the sequential cleavage of APP by the proteolytic enzymes β- and γ-secretase. The cleavage of APP by β-secretase results in production of a large, secreted, derivative soluble amyloid precursor protein β protein and a 99-amino acid membrane-associated fragment (CT99) [
7]. The γ-secretase complex cleaves within the transmembrane fragment of CT99 at various positions and generates peptides with differing lengths of the C-terminus (Aβ
1–37, Aβ
1–38, Aβ
1–40, Aβ
1–42, Aβ
1–43; reviewed in [
8]). Among these, Aβ
1–40 is the most abundant isoform (~80–90% of the total Aβ peptide) [
9], whereas the Aβ
1–42 variant is the main isoform of the amyloid deposits found in AD [
10,
11]. Aβ
1–42 peptides have higher self-aggregating properties than the less hydrophobic, smaller Aβ species, and they have the ability to form oligomers. These diffusible, nonfibrillar, soluble, low-molecular-weight oligomers have neurotoxic effects [
9]. It has been shown that soluble Aβ oligomers disrupt axonal transport [
12] and cause synapse deterioration, reversal of spines, reduction of long-term potentiation (LTP), and induction of long-term depression (LTD), as well as altered mitochondrial distribution and dynamics in neurons [
13]. Inhibition of LTP [
14] and reversal of LTD [
15] observed in the rat dentate gyrus are associated with impairment of cognitive function. Binding of Aβ oligomers to specific binding sites on neurons (reviewed in [
16]) leads to redistribution of synaptic proteins and activation of ionotropic glutamate receptors [
17]. Aβ oligomers inhibit LTP by increasing
N-methyl-
d-aspartate (NMDA) response through glutamate ionotropic receptor NMDA type subunit 2B. Furthermore, these molecules may cause a rapid depolarization of neurons and release of glutamate due to Ca
2+ entry, leading to excitotoxicity and synapse loss [
17].
Additionally, mutations in
PSEN1,
PSEN2, and
APP cause imbalances between synthesis and degradation of Aβ. For instance,
PSEN mutations lead to a partial loss of function in the γ-secretase complex and consequently to incomplete Aβ digestion [
18], whereas most of the mutations in
APP have been found to increase production of Aβ peptides [
19]. The above-described gene aberrations result in many abnormal cellular responses, such as mitochondrial dysfunction, inflammation, activation of microglia, and neuronal loss [
20].
The main mechanism in amyloid pathology in fAD is the increased production of Aβ species, whereas decreased Aβ clearance is postulated in sAD, which is modulated by the apolipoprotein E (
APOE) genotype. APOE isoforms differentially regulate Aβ aggregation and clearance in the brain and have distinct functions in regulating brain lipid transport, glucose metabolism, neuronal signaling, neuroinflammation, and mitochondrial function (reviewed in [
21]). There are at least three alleles of the
APOE gene—ε2, ε3, and ε4—coding different isoforms of the protein. The most common allele (~60% in the population) is ε3. Individuals carrying the ε4 allele are at increased risk of AD compared with those carrying the ε3 allele, whereas the ε2 allele is known to decrease the risk.
The second characteristic of AD is TAU-based neurofibrillary pathology. TAU is a microtubule-associated protein (MAPT) required for stabilizing microtubules, the major component of the neuronal cytoskeleton. This protein is involved in neurite outgrowth, maintenance of neuronal polarity, and axonal transport (reviewed in [
22]). Phosphorylation of TAU is regulated by kinases and phosphatases, and their actions on TAU molecules are required for proper neuronal growth. Site-specific phosphorylation of TAU plays a crucial role in microtubule stabilization, dynamic behavior, and spatial organization of microtubules in neurons and axonal transport regulation (reviewed in [
23]).
In AD, hyperphosphorylated TAU (AD pTAU) spontaneously aggregates into paired helical filaments (PHFs) and form NFTs. Abnormal phosphorylation of TAU impairs microtubule-binding capacity and leads to microtubule destabilization [
24]. These disturbances may cause improper anterograde and retrograde axonal transport and perturb intraneuronal signaling, including synaptic transmission (reviewed in [
25]).
TAU dysfunction is also related to other forms of dementia, including frontotemporal lobar degeneration, commonly referred as
frontotemporal dementia (FTD), which affects people before the age of 65 [
26]. Mutations in several different genes can cause FTD; however, the granulin precursor (
GRN) and
MAPT genes are the most frequent harborers of such mutations. This genetic alteration induces TAU hyperphosphorylation and aberrant NFT formation [
27,
28].
Studies on normal human brain tissue have revealed phosphorylation of TAU at several serine and threonine residues, whereas in AD more than 40 different phosphorylation sites have been identified [
29]. Most of the sensitive TAU hyperphosphorylation sites are located at the microtubule-binding repeat domain in the proline-rich regions [
30]. These regions include serine and threonine residues in serine-proline and threonine-proline motifs, and therefore they are targets of proline-directed protein kinases such as glycogen synthase kinase 3β (GSK3B) [
31]. Studies have clearly demonstrated that phosphorylation of TAU by GSK3B and cyclin-dependent kinase 5 reduces the affinity of TAU to bind to microtubules [
31], whereas phosphorylation of the serines within the KXGS motifs destructively affects TAU-microtubule interactions [
32]. Recent data have shown that Aβ accumulated in the AD brain can activate kinases that promote TAU phosphorylation, including GSK3B [
33]. GSK3B phosphorylates serine and threonine residues in TAU PHFs, and its activity corresponds with increased Aβ expression and Aβ-mediated neuronal death. Additionally, active forms of GSK3B were observed in cell lines with presenilin mutations [
34].
Current understanding of AD pathogenesis is limited owing to the difficulties in obtaining and culturing human brain tissue and live neurons and inability to model the disease. The recently developed induced pluripotent stem cell (iPSC) technology may provide a new platform to create reliable human disease models for better understanding the pathological mechanisms of AD and for establishment new therapeutic strategies. An increasing number of studies have shown that neurons derived from patient-specific iPSCs can recapitulate the major cellular phenotypes of AD and hence may offer themselves for disease modeling or drug development [
35]. In regard to AD, a few research groups have reported elevated levels of Aβ
1–42 and increased Aβ
1–42/Aβ
1–40 ratio in neurons differentiated from iPSCs of patients with fAD [
36‐
39], whereas authors of one report detected an increased secretion of Aβ
1–42 in only one patient with sAD [
40]. Additionally, AD neurons revealed elevated phosphorylation of TAU at Thr231 as well as an increased level of active GSK3B, and they accumulated large RAB5A-positive early endosomes [
40].
We report the generation and neural differentiation of iPSCs derived from several patients with fAD and sAD, as well as from control individuals without dementia. The aim of our present study was to establish an in vitro cellular model that reveals major phenotypes of AD and enables investigation of the pathomechanisms of late-onset AD. Amyloid-β (Aβ) production, TAU phosphorylation, and GSK3B activation were investigated in all lines. Moreover, cellular responses to extracellular hydrogen peroxide and Aβ1–42 oligomers were analyzed in control and AD-derived neural cells. We demonstrate that both fAD and sAD pathology can be recapitulated and that TAU pathology can be investigated in an iPSC-based in vitro human cellular model.
Methods
The chemicals used were purchased from Sigma-Aldrich (St. Louis, MO, USA), and the cell culture reagents and culture plates were purchased from Thermo Fisher Scientific (Waltham, MA, USA), unless specified otherwise.
iPSC lines
iPSC lines derived from patients with AD used in this study were characterized and published earlier [
41‐
46], as detailed in Table
1. The patients were clinically diagnosed and characterized by the Institute of Genomic Medicine and Rare Disorders, Semmelweis University, Budapest, Hungary, or at the Danish Dementia Research Centre, Rigshospitalet, University of Copenhagen, as described previously. Volunteers without dementia (assessed by clinical evaluation) were used as control subjects, from whom iPSC lines were established, characterized, and maintained under identical conditions as the AD iPSC lines. The hiPSC lines were maintained on Matrigel (BD Matrigel; STEMCELL Technologies, Vancouver, BC, Canada) in mTESR1 (STEMCELL Technologies) culture media. The media were changed daily, and the cells were passaged every 5–7 days using Gentle Cell Dissociation Reagent (STEMCELL Technologies) according to the manufacturer’s instructions.
Table 1
Cell lines used in this study
BIOT-7183-PSEN1 | S1 | fAD-1 | Early-onset fAD |
PSEN1 c.265G>C, p.V89L | F | |
S2 | fAD-2 |
H234 | C5 | fAD-3 | Early-onset fAD |
PSEN1 c.449T>C, p.L150P | M | |
H235 | C6 | fAD-4 | Early-onset fAD |
PSEN1 c.449T>C, p.L150P | M | – |
BIOT-0904-LOAD | S2 | sAD-1 | Late-onset sAD | Unknown | M | |
S3 | sAD-2 |
BIOT-0630-LOAD | S4 | sAD-3 | Late-onset sAD | Unknown | F | |
BIOT-4828-LOAD | S1 | sAD-4 | Late-onset sAD | Unknown | F | |
S6 | sAD-5 |
BIOT-0726-LOAD | S3 | sAD-6 | Late-onset sAD | Unknown | F | |
CTRL1 | S9 | Ctrl-1 | Healthy | – | F | – |
S11 | Ctrl-2 |
H250 | C16 | Ctrl-3 | Healthy | – | F | – |
H256 | C6 | Ctrl-4 | Healthy | – | M | – |
Neural induction of iPSCs
Neural progenitor cells (NPCs) were generated from each of the human iPSCs by dual inhibition of SMAD signaling pathway using LDN193189 and SB431542 [
47]. Neural induction was initiated upon reaching a desired confluence of iPSCs on Matrigel-coated dishes by addition of neural induction medium (NIM) (1:1 vol/vol mixture of DMEM/F12 and neurobasal medium, 1× N-2 supplement, 1× B-27 supplement, 1× nonessential amino acids [NEAA], 2 mM
l-glutamine, 50 U/ml penicillin/streptomycin, 100 μM β-mercaptoethanol, 5 μg/ml insulin), which was supplemented with 5 ng/ml basic fibroblast growth factor (bFGF), 0.2 μM LDN193189 (Selleckchem, Houston, TX, USA), and 10 μM SB431542. The NIM was changed every day. At day 10, neural rosettes were picked manually, replated on poly-
l-ornithine/laminin (POL/L) (Sigma-Aldrich)-coated dishes, and expanded in neural maintenance medium (NMM) (1:1 vol/vol mixture of DMEM/F12 and neurobasal medium, 1× N-2 supplement, 1× B-27 supplement, 1× NEAA, 2 mM
l-glutamine, 50 U/ml penicillin/streptomycin), and supplemented with 10 ng/ml epidermal growth factor and 10 ng/ml bFGF.
Neural differentiation of NPCs
To generate human neurons, NPCs were plated on the POL/L-coated dishes and cultured in neural differentiation medium (1:1 vol/vol mixture of DMEM/F12 and neurobasal-A medium, 1× N-2 supplement, 1× B-27 supplement, 1× NEAA, 2 mM l-glutamine, 50 U/ml penicillin/streptomycin) supplemented with 0.2 mM ascorbic acid and 25 μM β-mercaptoethanol. For terminal differentiation (TD) into cortical neurons, the cells were plated on POL/L (0.002%/2 μg/cm2) at a seeding density of 40,000 cells/cm2 for immunocytochemistry (ICC) and 100,000 cells/cm2 for enzyme-linked immunosorbent assay (ELISA) and Western blotting experiments with NMM. The medium was changed every 3–4 days during the course of TD. The efficiency of TD was monitored by ICC staining and RT-qPCR for tubulin-β 3 class III (TUBB3) and microtubule-associated protein 2 (MAP2) expression at week 10. In the present study, NPCs from passage 9 to passage 10 were differentiated up to 10 weeks for ELISA and Western blotting experiments, whereas samples were collected at weekly intervals.
Immunocytochemistry
Cells were fixed in 4% paraformaldehyde (PFA) for 20 minutes at room temperature (RT), washed twice with PBS, and permeabilized with 0.2% Triton X-100 in PBS for 20 minutes. Then, cells were blocked with 3% bovine serum albumin (BSA) in the presence of 0.2% Triton X-100 in PBS for 60 minutes at RT. The respective primary antibodies were applied overnight at 4 °C (Additional file
1: Table S1). To detect the signal, cells were incubated for 60 minutes at RT with the appropriate secondary antibodies (Alexa Fluor 488 donkey anti-rabbit immunoglobulin G [IgG] [H + L], Alexa Fluor 594 donkey anti-mouse IgG [H + L], Alexa Fluor 488 donkey anti-mouse IgG [H + L], or Alexa Fluor 594 donkey anti-rabbit IgG [H + L]; Life Technologies, Carlsbad, CA, USA). Cell nuclei were visualized using VECTASHIELD Mounting Medium with 4′,6-diamidino-2-phenylindole (1.5 μg/ml; Vector Laboratories, Burlingame, CA, USA). Cells were analyzed under a fluorescence microscope equipped with a 3D imaging module (Axio Imager system with ApoTome; Carl Zeiss MicroImaging GmbH, Göttingen, Germany) controlled using AxioVision 4.8.1 software (Carl Zeiss MicroImaging GmbH).
Electron microscopy
To evaluate the neuronal cultures, a monolayer of 5-week-old neurons grown on POL/L-treated glass coverslips was fixed with a fixative solution containing 3.2% PFA, 0.2% glutaraldehyde, 1% sucrose, and 40 mM CaCl
2 in 0.1 M cacodylate buffer for 24 h at 4 °C. Samples were rinsed for 2 days in cacodylate buffer, then postfixed in 1% ferrocyanide-reduced osmium tetroxide [
48] for 1 h at RT. The samples were then treated with aqueous 1% uranyl-acetate for 30 minutes and embedded in Spurr low-viscosity epoxy resin medium (Sigma-Aldrich) according to the manufacturer’s instructions and cured for 24 h at 80 °C. Ultrathin sections were stained with Reynolds lead citrate for 2 minutes and examined under a JEOL JEM 1011 transmission electron microscope (JEOL, Peabody, MA, USA) operating at 60 kV. Photographs were taken using an Olympus Morada 11-megapixel camera and iTEM software (Olympus, Center Valley, PA, USA).
Flow cytometry
NPCs growing on POL/L-coated dishes were dissociated into single cells with Accutase (Sigma-Aldrich) and fixed with 4% PFA for 20 minutes at RT. Cells were permeabilized with 0.2% Triton X-100 in PBS for 5 minutes at RT and blocked with 10% FBS in PBS for 20 minutes at RT. Cells were stained for 1 h at RT with Alexa Fluor 647 mouse anti-Nestin and phycoerythrin mouse anti-human paired box 6 (PAX6) antibodies (BD Pharmingen, San Jose, CA, USA). Flow cytometric analysis was performed using a Cytomics FC 500 flow cytometer (Beckman Coulter, Brea, CA, USA). To detect Nestin (NES) and PAX6 expression in NPCs, an argon laser (488 nm) and a red solid laser (635 nm), respectively, were used. Flow cytometric data analysis was performed using FlowJo software (version 7.6.5; FlowJo, LLC, Ashland, OR, USA).
Immunoblotting
The cell cultures were lysed with RIPA Lysis and Extraction Buffer supplemented with Halt™ Protease and Phosphatase Inhibitor Cocktail and Pierce™ Universal Nuclease for Cell Lysis (all from Thermo Fisher Scientific). The protein extracts were derived from one well of a six-well plate of a single neuronal differentiation. Lysed samples were sonicated and incubated for 60 minutes on ice. Total protein concentration was determined using a Pierce BCA Protein Assay Kit. Cell lysates (2 μg) were separated on 10% sodium dodecyl sulfate-polyacrylamide gel and transferred to Immun-Blot® PVDF Membrane (Bio-Rad Laboratories, Hercules, CA, USA). The membranes were blocked with Tris-buffered saline with Tween 20 (TBST; 20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Tween 20) containing 5% BSA and then incubated with the respective primary antibody solution overnight at 4 °C (Additional file
1: Table S1). After being washed with TBST, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h at RT. All the secondary antibodies were either goat anti-mouse or goat anti-rabbit (Sigma-Aldrich). All samples were analyzed at days 14, 28, 42, 56, and 70 of TD, and the amount of pTAU and pSer9GSK3B relative to total TAU and GSK3B in lysates was measured. Signals were detected with SuperSignal™ West Dura Extended Duration Substrate by KODAK Gel Logic 1500 Imaging System (Bruker, Billerica, MA, USA) and Kodak MI SE imaging software (Carestream, Atlanta, GA, USA). Densitometric measurement of protein band intensity was carried out using Image Studio™ Lite software (LI-COR Biosciences, Lincoln, NE, USA). Phosphorylation of TAU and GSK3B protein was calculated from densitometric data of phosphoproteins divided by densitometric data of nonphosphorylated proteins.
Measurement of Aβ1–40 and Aβ1–42 by ELISA
Conditioned medium was collected after 4 days of culture (without media change) at every week from one well of a six-well plate. To prevent protein degradation, 4-(2-aminomethyl)benzenesulfonyl fluoride hydrochloride was added to the medium. Extracellular Aβ1–40 and Aβ1–42 levels were measured using the Human β-Amyloid (1-40) ELISA Kit and Human β-Amyloid (1-42) ELISA Kit (Wako Chemicals, Neuss, Germany) according to the manufacturer’s instructions. The secreted Aβ levels determined (in picomolar concentrations) were normalized to total protein content of cell lysate. The signal was detected using a Varioskan Flash multimode reader (Thermo Fisher Scientific).
As a control value we used the average value (±SEM) of the four clones derived from healthy individuals (ctrl-1, ctrl-2, ctrl-3, ctrl-4) in all experiments. The individual expression levels are presented in Additional file
2: Figure S1.
RT-qPCR
Total RNA was isolated from differentiated neurons at different time points using the RNeasy Plus Mini Kit (QIAGEN, Valencia, CA, USA) according to the manufacturer’s protocol. One microgram of RNA was transcribed using the SuperScript™ III VILO™ cDNA Synthesis Kit (Thermo Fisher Scientific). The PCR conditions were subjected to 94 °C for 3 minutes as an initial denaturation step, followed by 40 cycles of 95 °C for 5 seconds for denaturation, 60 °C for 15 seconds for annealing, and 72 °C for 30 seconds for elongation. The amplification reactions were carried out in a total volume of 15 μl using SYBR Green JumpStart Taq ReadyMix (Sigma-Aldrich). RT-qPCR was run on the Rotor-Gene Q 5plex Platform (QIAGEN) using oligonucleotide primers detailed in Additional file
3: Table S2. Human glyceraldehyde 3-phosphate dehydrogenase
GAPDH) was used as a reference gene. The data were analyzed using REST software (2009; version 2.0.13).
Library preparation and next-generation sequencing
Genomic DNA (gDNA) was isolated from peripheral blood mononuclear cells using the GenElute Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich). An ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) was used to determine the concentration and purity of the isolated gDNA. The gDNA integrity was evaluated by 0.8% agarose gel electrophoresis.
The multiplexed library preparation was performed by using the SureSelect
QXT and SureSelect
XT Human All Exon V6 Target Enrichment systems according to the manufacturer’s instructions (Agilent Technologies, Santa Clara, CA, USA). The probe set targeted > 99% of RefSeq (~20,000) genes with a total target size of 60 Mb. Pooled libraries were sequenced on the NextSeq 500 platform (Illumina, San Diego, CA, USA) in a 2 × 80-bp paired-end run. The average throughput was > 80 million reads (4 Gbp per sample). Primary data analysis, filtering, and adapter trimming were performed using NextSeq Control Software (version 1.4.1; Illumina). Alignment was done using the Burrows-Wheeler Aligner (Illumina) [
49] to the GRCh37/hg19 reference genome, and variant calling was performed with the Genome Analysis Toolkit (GATK) [
50]. Variants were annotated using the wANNOVAR web server [
51,
52], and their phenotypic effects were evaluated using the American College of Medical Genetics and Genomics standards and guidelines [
53].
Cell viability assay after H2O2 and Aβ treatment
Neurons growing on a 96-well plate at days 28 and 56 of TD were treated with 30 μM and 60 μM of hydrogen peroxide (H
2O
2) (Sigma-Aldrich) or 5 μM Aβ
1–42 oligomer solution. For preparation of oligomer solution, iso-Aβ
1–42 peptides were used as detailed elsewhere [
54] (provided by University of Szeged, Hungary). Viability of the cultures following H
2O
2 and Aβ treatment was assessed at 24 h of stimulation using a CellTiter-Glo® Luminescent Cell Viability Assay (Promega, Madison, WI, USA). The luminescent signal was detected using the Varioskan Flash Multimode Reader. Neuronal survival was represented as a percentage of control.
Statistical analysis
All results were analyzed using Prism 5 (GraphPad Software, La Jolla, CA, USA) and Microsoft Office 2010 (Microsoft, Redmond, WA, USA) software. All data on the graphs represent the average of the triplicate measurements (n = 3). The “n” value corresponds to the number of replicates for each cell line. Analysis of data is presented as the mean ± SEM. Dunnett’s method was used to compare the individual groups with control subjects. In all cases, significance was noted at p < 0.05.
Discussion
The present study shows that particular key disease phenotypes of the most common age-related neurological disorder, AD, can be modeled using patient-specific, iPSC-derived neural cells. Over the past few decades, primary neurons from animal models and immortalized neuronal cell lines have been used to study neurodegenerative diseases. Owing to many limitations, such as genetic and physiological differences between human and rodent brains, such studies of AD mechanisms have been controversial. Most animal models of AD are generated on the basis of modifications in three genes related to fAD (APP, PSEN1, and PSEN2), whereas modeling of sAD in rodents is currently not well examined. Therefore, there are huge expectations regarding iPSC technology, which allows for the generation of pluripotent cells from any individuals in the context of their own genetic identity, to provide new patient-specific in vitro disease models to study neurological disorders.
Modeling AD using iPSCs was initiated with fAD carrying mutations in
PSEN1,
PSEN2, and
APP. Until now, only a few groups have reported generation of iPSC-derived neurons from patients with fAD, and sAD has been studied even less [
36,
37,
40,
67]. In our present study, we analyzed samples and generated neurons from patients with fAD carrying pathogenic mutations in the
PSEN1 gene (V89L and L150P; these mutations were first identified by Queralt et al. [
68] and Wallon et al. [
69], respectively) and patients with sAD (Table
1). Our findings demonstrate that iPSCs derived from patients with fAD and patients with sAD can be successfully induced into NPCs with very uniform expression of NES and PAX6 and that they can be further differentiated into neurons and glial cells. Gene expression and ICC analysis revealed the presence of various neuronal subtypes, including GABAergic, glutamatergic, cholinergic, and dopaminergic neurons and progenitor cells of astrocytes and oligodendrocytes in control and AD cultures. We therefore conclude that there is no prominent difference in the differentiation and maturation propensity or in marker gene expression between control and AD neural cells, which is in accordance with a previous report [
37].
Initially, our main question was to evaluate if sAD can be modeled through iPSC-derived neural cultures and represent a suitable model system for neuropathological investigations and drug development studies. Therefore, we first analyzed the relevant in vivo pathological hallmarks of the disease in our in vitro system. The accumulation of Aβ into extracellular aggregates is one of the pathological signs of AD in the human brain. Alterations in the level of Aβ peptides are often presented as the ratios between different isoforms. In
PSEN mutant mouse models, elevated Aβ production can be detected [
70‐
72]. In humans, elevated Aβ production was revealed in iPSC-derived neural lines from patients with fAD [
36]. However, another study demonstrated that Aβ secretion in sAD-derived neurons is not consistently altered [
37]. Our study demonstrated increased extracellular Aβ
1–40 and Aβ
1–42 levels in neurons derived from all fAD and sAD lines in a maturation-dependent manner. Interestingly, Aβ
1–40 secretion was approximately twofold higher in sAD neurons than in fAD, whereas Aβ
1–42 levels were similar. Moreover, we observed an elevated ratio of Aβ
1–42 to Aβ
1–40, one of the AD hallmarks, in fAD-derived neurons, whereas the Aβ
1–42/Aβ
1–40 ratio in sAD lines remained unchanged and comparable to that of non-AD control lines. Additionally, we observed upregulated expression of APP and APP-CTF in all AD-derived cell lines, which is in line with the increased Aβ levels we measured. Some groups also detected an increased level of Aβ
1–40, Aβ
1–42, and Aβ
1–42/Aβ
1–40 ratio in fAD cell lines with mutations in
PSEN1,
PSEN2, and
APP [
36,
37,
40]. On the basis of the above results, we can conclude that mutations in
PSEN1 may change the metabolism of Aβ peptides and drive amyloidosis in patients with fAD. Our findings also indicate the possible heterogeneity of fAD and sAD. AD-iPSC lines with
PSEN1 mutations and sAD do not always recapitulate the same phenotypes [
37]. In fAD, genetic factors modify the clinical phenotype of the disease, whereas mechanisms underlying the pathogenesis of sAD are still not well understood and combine multiple genetic and environmental risk factors. It is possible that underlying mutations that have not yet been discovered may play an important role in sAD, reflecting the inherent variability of iPSCs. Thus, more cell lines have to be analyzed to reveal the broad heterogeneity of AD phenotypes.
We have demonstrated that another pathological hallmark characteristic of AD, TAU hyperphosphorylation, could be detected in AD neurons. Conformational changes and misfolded protein structure result in aberrant aggregation of TAU into neurofibrillary structures, NFTs [
73]. TAU phosphorylation at various sites affects TAU activity, as well as its biological function and pathogenic role. Studies on the physiological properties of TAU have revealed that phosphorylation of Ser262 significantly diminishes the ability of TAU to bind microtubules [
74]. Others have reported that phosphorylation of few KXGS motifs, including Ser262 and Ser356, reduces TAU binding capacity to microtubules and thus increases the dynamics of microtubules, which plays an important role in neurite growth and the development of neuronal polarity [
75]. Phosphorylation at Thr231 leads to decreased ability of TAU to bind microtubules, reduces the level of acetylated tubulin, and consequently leads to microtubule destabilization [
33]. Moreover, the increased TAU phosphorylation at Ser396 and Ser404 impairs microtubule assembly by detachment of TAU molecules from microtubules [
76]. Elevated phosphorylation on Ser262, Thr231, and Ser396 residues can be detected early in the AD disease process [
77,
78].
Quantitative in vitro data have revealed a negative impact of TAU phosphorylation at many epitopes on TAU activity and microtubule stability [
31,
32,
79]. Previous studies on iPSC-derived, patient-specific fAD and sAD neurons have been limited to detecting mostly only one TAU phosphorylation site at Thr231 [
40,
67]. We analyzed TAU phosphorylation in iPSC-derived neurons from patients with fAD and patients with sAD at six different phosphorylation sites. As a novel finding, we demonstrated increased TAU phosphorylation at all examined epitopes (Fig.
5a): Ser262, Ser202/Thr205, Ser396, Ser400/Thr403/Ser404, Thr181, and Thr231 in both fAD and sAD neurons. According to the literature, phosphorylation of TAU at Ser262 and Thr231 greatly diminishes its ability to bind microtubules, by 35% and 25%, respectively [
74], whereas Ser396 and Ser404 phosphorylation generates more fibrillogenic TAU in vitro [
78], which shows an increased propensity to aggregate [
80]. Moreover, phosphorylation of the AT8 epitope results in a decrease in TAU microtubule nucleation activity, leading to microtubule depolymerization and destabilization [
81]. TAU phosphorylation increased during neural differentiation, which is in agreement with in vivo studies on mice with
APP mutations [
82].
Furthermore, we have shown higher levels of active GSK3B in our AD cultures. This observation correlates with previously described findings for which researchers showed increased levels of active GSK3B measured in fAD and sAD neurons in vitro [
40] and in transgenic animal models [
83]. Pathological activation of GSK3B establishes a feedforward loop that contributes to abnormal APP processing [
84], enhanced apoptosis in hippocampal neurons, TAU hyperphosphorylation, and synaptic failure in rodent models of AD [
85,
86]. Our findings revealed that activation of GSK3B might contribute to TAU misregulation and abnormal phosphorylation. This is in accordance with previous reports confirming the important role of GSK3B in regulating TAU phosphorylation mostly on Thr231 and Ser199, Ser396, Ser400, Ser404, and Ser413 [
87]. Consistent with this, restoring the normal level of GSK3B has been shown to reduce TAU hyperphosphorylation, decrease Aβ production and neuronal death in AD murine models [
88], and decrease Aβ-induced neurotoxicity in cultured mouse primary neurons in vitro [
89].
In our study, we evaluated cell viability after hydrogen peroxide and extracellular Aβ
1–42 exposure in neurons at different maturation stages. Exposure to stress agents such as H
2O
2 induces ROS production and toxicity in many different cell types [
90‐
92]. ROS show high reactivity with macromolecules and play an important role as signaling molecules (reviewed in [
93]). Oxidative damage is linked with mitochondrial abnormalities and is catalyzed by the presence of Fe and Cu ions. In our results, we demonstrated a significant H
2O
2 dose-related decrease in the survival of fAD and sAD neurons. More mature neurons (TD56) showed a greater sensitivity to H
2O
2 than younger neuronal cultures (TD28). These observations indicate that H
2O
2 may provoke an antioxidant stress response resulting in increased level of ROS and may lead to subsequent cell death. Additionally, we observed that treatment with Aβ
1–42 oligomers induced cell death in both fAD and sAD neurons. According to the literature, Aβ treatment may lead to activation of glutamate receptors and inhibition of glutamate transporters that leads to abnormal release of glutamate and disturbances in glutamatergic neurotransmission. Aβ treatment induces NMDA receptor-mediated cellular events in neurons and astrocytes, leading to synaptic damage and spine loss [
94]. Chronic stimulation of NMDA receptor results in Ca
2+ influx, which activates apoptotic pathways and increased glutamate excitotoxicity, leading to generation of ROS and to neuronal damage and cell death [
95]. In cortical neurons, accumulation of glutamate and NMDA receptors promotes H
2O
2-mediated neurotoxicity and oxidative damage in DNA.
Furthermore, increased influx of Ca
2+ mediated by Aβ treatment activates mitochondrial permeability transition pore, leading to deregulation of respiratory chain enzymes and ROS overproduction, and consequently to neurotoxicity [
96]. On the basis of the above data, we can speculate that the neuronal death observed in our cultures upon synthetic Aβ treatment may be a consequence of mitochondrial stress and higher ROS production.
It has previously been reported that accumulated Aβ oligomers induce endoplasmic reticulum (ER) stress and ROS production [
97], which lead to membrane lipid peroxidation and impairment of membrane protein function (reviewed in [
98]). Furthermore, gene analysis of APP E693Δ neurons revealed upregulation of the oxidative stress-related markers [
37]. Additionally, in vivo studies showed increased protein oxidation and lipid peroxidation in
PSEN1 mutant brains [
99,
100], leading to destruction of spine morphology and impaired synaptic plasticity [
101]. Thus, we speculate that our fAD and sAD neurons may also exhibit increased levels of stress-related gene expression, suggesting ER and Golgi abnormalities. It is worth considering that mutations in
PSEN1 and pathological changes in sAD, combined with neuronal aging, can upregulate ROS production, leading to mitochondrial damage that may contribute to neurodegenerative processes and AD progression.
Our results provide insight into the molecular basis of AD and provide patient cell models that are relevant for further AD investigations. In all fAD and sAD lines, we observed higher Aβ1–40 and Aβ1–42 secretion, increased activation of GSK3B, and hyperphosphorylation of TAU at six different epitopes. We showed that both fAD and sAD neurons revealed AD phenotypes, suggesting that sAD, with an unknown disease etiology, resembles the fAD phenotype.
The results of this study provide strong evidence that iPSC technology can be used in modeling the most common age-related neurodegenerative disorders, such as AD. This new strategy will make it possible to study pathological mechanisms on the molecular level and analyze patient-derived neurons in vitro from individuals who are still alive. Potentially, iPSCs from complex, unknown genetic background diseases, such as sAD, may be equally useful and able to provide suitable models such as the monogenic forms of AD carrying mutations in PSEN1, PSEN2, and APP.
Although iPSC-derived models afford the opportunity to identify factors associated with disease phenotypes, there are still many uncertainties that should be resolved in the near future. The most crucial limitations that must be addressed in future studies include development of reliable protocols for more rapid neuronal differentiation and optimization of culture conditions as well as current differentiation methods to eliminate clonal variability and heterogeneity of differentiated neuronal cells.
Several preclinical studies using fAD and sAD models have established new methods to gather insight into the molecular mechanisms, the role of different cell types, and patient-specific drug response. We believe that iPSCs have great potential for improving the model systems used in preclinical and clinical applications in neurodegenerative diseases.