Neurons derived from sporadic Alzheimer’s disease iPSCs reveal elevated TAU hyperphosphorylation, increased amyloid levels, and GSK3B activation

Alzheimer’s disease (AD) is the most common type of dementia, affecting one in eight adults over 65 years of age. The majority of AD cases are sporadic, with unknown etiology, and only 5% of all patients with AD present the familial monogenic form of the disease. In the present study, our aim was to establish an in vitro cell model based on patient-specific human neurons to study the pathomechanism of sporadic AD. We compared neurons derived from induced pluripotent stem cell (iPSC) lines of patients with early-onset familial Alzheimer’s disease (fAD), all caused by mutations in the PSEN1 gene; patients with late-onset sporadic Alzheimer’s disease (sAD); and three control individuals without dementia. The iPSC lines were differentiated toward mature cortical neurons, and AD pathological hallmarks were analyzed by RT-qPCR, enzyme-linked immunosorbent assay, and Western blotting methods. Neurons from patients with fAD and patients with sAD showed increased phosphorylation of TAU protein at all investigated phosphorylation sites. Relative to the control neurons, neurons derived from patients with fAD and patients with sAD exhibited higher levels of extracellular amyloid-β 1–40 (Aβ1–40) and amyloid-β 1–42 (Aβ1–42). However, significantly increased Aβ1–42/Aβ1–40 ratios, which is one of the pathological markers of fAD, were observed only in samples of patients with fAD. Additionally, we detected increased levels of active glycogen synthase kinase 3 β, a physiological kinase of TAU, in neurons derived from AD iPSCs, as well as significant upregulation of amyloid precursor protein (APP) synthesis and APP carboxy-terminal fragment cleavage. Moreover, elevated sensitivity to oxidative stress, as induced by amyloid oligomers or peroxide, was detected in both fAD- and sAD-derived neurons. On the basis of the experiments we performed, we can conclude there is no evident difference except secreted Aβ1–40 levels in phenotype between fAD and sAD samples. To our knowledge, this is the first study in which the hyperphosphorylation of TAU protein has been compared in fAD and sAD iPSC-derived neurons. Our findings demonstrate that iPSC technology is suitable to model both fAD and sAD and may provide a platform for developing new treatment strategies for these conditions.


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 βand 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  ; 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, lowmolecular-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 prolinerich regions [30]. These regions include serine and threonine residues in serine-proline and threonineproline motifs, and therefore they are targets of prolinedirected 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][37][38][39], whereas authors of one report detected an increased secretion of Aβ  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 lateonset AD. Amyloid-β (Aβ) production, TAU phosphorylation, and GSK3B activation were investigated in all lines. Moreover, cellular responses to extracellular hydrogen peroxide and Aβ 1-42 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 iPSCbased 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][42][43][44][45][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.

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/cm 2 ) at a seeding density of 40,000 cells/cm 2 for immunocytochemistry (ICC) and 100,000 cells/cm 2 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:

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  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 Aβ 1-40 and Aβ 1-42 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 Aβ 1-40 and Aβ 1-42 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 pairedend 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 H 2 O 2 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 2 O 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 2 O 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.

Neuronal differentiation capacity of iPSCs derived from patients with fAD and patients with sAD is similar
Samples from patients with fAD and patients with sAD were reprogrammed, and iPSC lines were generated and characterized by our teams as previously reported [41][42][43][44][45][46]. In the present study, we studied iPSC clones derived from three patients with fAD with mutation in PSEN1 (contributing with in total four lines: fAD-1, fAD-2, fAD-3, fAD-4), four patients with late-onset sAD (contributing with in total six lines: sAD-1, sAD-2, sAD-3, sAD-4, sAD-5, sAD-6), and three individuals without dementia as control subjects (contributing with in total four lines: ctrl-1, ctrl-2, ctrl-3, ctrl-4) (see also Table 1). In all sAD samples, next generation sequencing analysis revealed a lack of pathogenic mutations in APP, PSEN1, and PSEN2, indicating late-onset AD. Moreover, sequencing of MAPT and GRN excluded FTD as well. The APOE status of the cell lines included in the study is shown in Additional file 4: Table S3.
All AD iPSCs and control iPSCs were successfully converted into neuronal progenitor cells (NPCs) and expressed neuroepithelial markers PAX6 and NES (Fig. 1a).
To quantify the efficiency of NPC formation, flow cytometry was applied, which showed that there were no significant differences between any of the cell lines with respect to generating NPCs including > 70% copositivity for PAX6 and NES (Fig. 1b).
Next, the neuronal differentiation of fAD and sAD NPCs was investigated and compared with that of control lines at different time points (TD14, TD28, TD42, TD56, TD70). ICC labeling showed that the increasing expression of neuronal markers TUBB3 and MAP2 during neuronal differentiation was comparable among AD and control cell lines (Fig. 2a). More mature markers-TAU and neurofilament, heavy polypeptide 200 kDa (NF200)-appeared later during differentiation than MAP2 and TUBB3 (Fig. 2a) (Fig. 2b). In order to compare gene expression levels, RT-qPCR was performed on day 70 neurons. The analysis revealed that the AD and control lines expressed comparable levels of MAP2, sodium-dependent serotonin transporter (SLC6A4), NMDA receptor subunit NR1 (glutamate ionotropic receptor N-methyl-D-aspartate type subunit 1 [GRIN1]), marker of maturing neurons -RNA binding protein, fox-1 homolog 3 (RBFOX3), and GAD1 (Fig. 2c). Moreover, they also showed high expression levels of dopaminergic neuron marker (TH), astrocyte marker (glial fibrillary acidic protein [GFAP]), and an early oligodendrocyte marker (claudin 11 [CLDN11]) (Fig. 2c). Furthermore, the cells expressed TUBB3, choline O-acetyltransferase (CHAT), and postsynaptic density protein 95 (discs large MAGUK scaffold protein 4 [DLG4]). Importantly, no evident differences in gene expression levels were observed between control, fAD, and sAD lines. Additionally, electron microscopy of neural cell cultures at day 35 revealed the presence of synapses with docking of synaptic vesicles and synaptic clefts (Fig. 2d). These findings indicate that AD has no significant effect on the overall neural differentiation of iPSCs in the used system, as well as that NPCs can be differentiated into various neuronal and glial cell types with similar efficiency independent of their origin from patients with AD or healthy individuals.

Elevated level of Aβ was secreted by fAD and sAD neurons
To determine Aβ production from iPSC lines of patients with fAD and patients with sAD, secreted Aβ 1-40 and Aβ 1-42 levels were measured weekly by ELISA in the conditioned media of differentiated neural cell lines from the second to tenth weeks of differentiation. Elevated secreted Aβ 1-40 and highly increased Aβ 1-42 levels were observed in all fAD neural cultures compared with the mean level in the control lines at almost all time points (Fig. 3). The Aβ 1-42 /Aβ 1-40 ratio was also increased up to twofold for the fAD-iPSC-derived neurons. The extracellular level of Aβ 1-40 in all sAD neurons was significantly higher than in control and fAD neurons. The Aβ 1-42 secretion in sAD neurons was also elevated compared with the average values of the control neurons; however, the levels did not reach the same levels as in fAD neurons. Interestingly, the measured Aβ 1-42 /Aβ 1-40 ratio in all sAD-iPSC-derived neurons was similar to that in control neurons (Fig. 3c). In nonmature neurons up to day 14 (TD14), Aβ production was comparable among AD and control lines (Additional file 2: Figure S1). During neural differentiation, Aβ 1-40 and Aβ 1-42 levels gradually increased to reach the investigated maximum at day 70 of c Gene expression of neuronal markers from ctr-1, early-onset familial Alzheimer's disease (fAD-1), and sporadic Alzheimer's disease (sAD-1) lines at TD70 obtained by qPCR. The expression values were normalized to GAPDH and calculated as a relative amount of messenger RNA versus expression value of neural progenitor cells, which was set to 1. Data are reported as mean ± SEM of three independent measurements. AD-iPSC neurons demonstrate expression pattern similar to that of neuronal cells derived from control iPSCs. d Ultrastructure of a synapse at TD35 with synaptic vesicles (Sv) and synaptic cleft with synaptic junctions (tight junctions, black arrowheads). Docking synaptic vesicles are also observable (white arrowheads). Scale bar = 100 nm. DAPI 4′,6-Diamidino-2-phenylindole TD (endpoint of experiment; TD70), suggesting a maturation-dependent secretion of Aβ (Fig. 3, Additional file 2: Figure S1). The results at TD14 and TD28 indicated a transition period in which not all AD samples showed significant differences from the control samples, but from TD42, all AD samples displayed elevated amounts of Aβ. Therefore, we concentrated our further experimental analyses on the samples matured for a minimum of 6 weeks.
To evaluate the level of APP, the direct precursor of Aβ in neurons, we performed Western blot analysis. The results revealed that APP expression increased during neuronal differentiation and reached a maximum at day 70 (TD70) in all fAD and sAD neural cultures (Fig. 4, Additional file 5: Figure S2a). Expression of this protein was significantly increased in all AD cell lines from TD42. Subsequently, we quantified the amount of amyloid precursor protein carboxy-terminal fragment (APP-CTF) generated through APP cleavage by β-secretase. As shown in Fig. 4 and Additional file 5: Figure S2b, remarkably higher APP-CTF protein levels were detected in fAD and sAD neurons. No significant differences in the expression of APP and APP-CTF were observed between fAD and sAD neurons. In summary, both Aβ and APP levels were significantly elevated in fAD and sAD neural cultures compared with control cultures. These differences were detectable after 6 weeks (TD42) but were prominent with all sAD lines at week 10 (TD70).

Epitope-specific TAU hyperphosphorylation was detected in fAD and sAD lines
One of the proposed mechanisms for the TAU protein pathomechanism in AD are posttranslational modifications through abnormal phosphorylation (hyperphosphorylation) (reviewed in [55]). The appearance and accumulation of abnormally phosphorylated TAU leads to mislocalization and aggregation of TAU as well as formation of NFTs [56]. Therefore, we focused on the alterations of TAU phosphorylation in our fAD and sAD neural cultures. TAU protein contains several phosphorylation sites (reviewed in [57]). We examined six different phosphorylation sites (Fig. 5a) in fAD, sAD, and control neural cultures at different time points of TD. We observed significantly higher phosphorylation of TAU at Ser262 (12E8 epitope) in all our AD neurons than in control neurons (Fig. 5b, Additional file 6: Figure S3a). Elevated phosphorylation was detected already at TD42 and reached the highest level between weeks 8 and 10 (TD56 and TD70) of TD. Further analysis revealed upregulated TAU phosphorylation at the Ser396 epitope in all fAD and sAD neural cultures (Fig. 5c, Additional file 6: Figure S3b). Western blot quantification showed an increase in TAU phosphorylation at Ser202/Thr205 (AT8 epitope) in the AD group (Fig. 5d, Additional file 6: Figure S3c). Furthermore, we observed higher phosphorylation of TAU at Thr181 and Ser400/Thr403/ Ser404 in both fAD and sAD neurons ( Fig. 5e and f, Additional file 6: Figure S3d and e). On the basis of these results, we can conclude that, although the phosphorylation level showed some clonal variations, the ratio of pTAU/TAU was significantly elevated in all AD neurons and that no major differences were observed between fAD-and sAD-derived neurons.

GSK3B activation in AD-derived neurons
Scientific evidence suggests that GSK3B is involved in many pathological hallmarks of AD, including hyperphosphorylation of TAU [58], increased Aβ production [59], memory impairment, and neuronal loss [60]. To verify if AD neurons with elevated TAU phosphorylation have increased GSK3B activity, the percentage of the active form of GSK3B was calculated by assessing the ratio of inhibitory phosphorylation on Ser9 by immunoblotting. Our results demonstrate that fAD and sAD neurons in most cases exhibited significantly higher levels of active GSK3B than control neurons (Fig. 6a-c, Additional file 7: Figure S4a). During neuronal differentiation, the level of active GSK3B in control neurons remained constant, whereas in fAD and sAD neurons, the active GSK3B amount increased to reach its maximum at day 56 (sAD-2) or day 70 of TD for all remaining cell lines (Fig. 6c, Additional file 7: Figure S4a). Furthermore, higher phosphorylation of TAU at Thr231 (AT180 epitope), which can be catalyzed by GSK3B in vitro [61], was observed in all fAD and sAD neurons (Fig. 6d, Additional file 7: Figure S4b). These results indicate that activation of the GSK3B signaling in both fAD and sAD neurons contributes to the abnormal TAU phosphorylation pattern observed in these cell lines.

Oxidative stress response was strongly affected in fAD and sAD neuronal cultures
Oxidative stress plays an important role in the pathogenesis of neurodegenerative disorders. The molecular mechanism of reactive oxygen species (ROS) action in the nervous system has been studied ( [62,63], reviewed in [64]). Despite this, the effect of oxidative stress on human iPSC-derived AD neurons is still not well studied.
In our experiments, we used H 2 O 2 in two different concentrations to examine neuronal response to a stressor. In AD and control neurons, treatment with increasing doses of H 2 O 2 engendered a dose-dependent loss of cell viability; however, these two cell groups presented differential susceptibility to the stressor. The fAD and sAD neurons were more sensitive than control neurons to  reached about 50% and 70%, respectively, in the AD cultures (Fig. 7a). Evidence of the toxicity of Aβ to neurons has been demonstrated in rodent models and neuronal cell lines (reviewed in [65,66]). Owing to the limited studies on neurons generated from iPSCs, we examined the effect of synthetic Aβ 1-42 oligomer solution on iPSC-derived neurons. Cells treated at days 28 and 56 of neural differentiation for 24 h with 5 μM Aβ 1-42 oligomers showed reduced neuronal survival, depending on the neuronal maturation state (Fig. 7b). Relative to control samples, AD cultures had enhanced sensitivity to cell toxicity induced by Aβ 1-42 . The 28-day-old AD neurons displayed about 15-20% cell death compared with 10% cell loss in control individuals. Mature neurons at TD56 were more susceptible to cell death; almost 40% of neurons died in AD cultures, whereas cell loss in control neurons increased slightly to 15% (Fig. 7b). On the basis of our results, we can conclude that fAD and sAD neural cultures are more susceptible to Aβ 1-42 oligomerinduced cell death than the control cultures.

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 iPSCderived 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][71][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 2 O 2 induces ROS production and toxicity in many different cell types [90][91][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 2 O 2 dose-related decrease in the survival of fAD and sAD neurons. More mature neurons (TD56) showed a greater sensitivity to H 2 O 2 than younger neuronal cultures (TD28). These observations indicate that H 2 O 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 2 O 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 stressrelated 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 Aβ 1-40 and Aβ 1-42 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.

Conclusions
Our findings demonstrate that increased TAU phosphorylation occurs in fAD and sAD iPSC-derived mature neurons. To our knowledge, this is the first study involving comparison of TAU protein hyperphosphorylation in an iPSC-based fAD and sAD model. We also present evidence of elevated Aβ 1-40 and Aβ 1-42 levels in fAD and sAD samples; however, the Aβ 1-42 /Aβ 1-40 ratios are significantly different from control cell lines in fAD cases, whereas the ratio of sAD cell lines not differ significantly. Moreover, we have shown the more sensitive oxidative stress response and higher susceptibility to exogenously added synthetic Aβ 1-42 peptide solution of AD lines. Our findings demonstrate that the iPSC technology is suitable to model both fAD and sAD and may provide a platform for developing new treatment strategies for these conditions.

Additional files
Additional file 1: Table S1. Antibodies used in this study. (TIF 141 kb) Additional file 2: Figure S1. Characterization of Aβ secretion in control and AD iPSC-derived neurons at TD14 and TD28. a The amount of secreted Aβ 1-40 and (b) Aβ 1-42 from control-, fAD-(fAD-1-fAD-4), and sAD-(sAD-1-sAD-6)-iPSC-derived neurons. c The ratio of Aβ 1-42 /Aβ 1-40 from neurons derived from control and AD lines. Aβ 1-40 and Aβ 1-42 secreted from neural cells into the medium were measured at day 4 after the last medium change. The extracellular Aβ levels determined (in picomolar concentrations) were normalized to total protein content. Data represent mean ± SEM (n = 3). Because the detected levels of Aβ  and Aβ  were not significantly different between the four healthy individual-derived clones (ctrl-1, ctrl-2, ctrl-3, ctrl-4), the average value (±SEM) as a control value was used in all graphs. Dunnett's test was performed to evaluate the significance of groups compared with control (*p < 0.05). (TIF 396 kb) Additional file 3: Table S2. PCR primers used in this study. (TIF 109 kb) Additional file 4: Table S3. Polymorphism of AD-and FTD-related genes and APOE status of cell lines used in this study. Footnote: Cell lines not detailed in the table were not permitted to be involved in whole exome sequencing; therefore, only clinical mutation identification was available. (TIF 130 kb) Additional file 5: Figure S2. Characterization of APP and APP-CTF expression during neuronal differentiation. Densitometric analysis of (a) APP and (b) APP-CTF expression in control neurons (ctrl-1), fAD-derived neurons (fAD-1-fAD-4), and sAD-derived neurons (sAD-1-sAD-6) at TD42, TD56, and TD70. Quantification of APP and APP-CTF signals was normalized to GAPDH. Data are presented as mean ± SEM (n = 3). Dunnett's test was performed to evaluate the significance of groups compared with control (*p < 0.05). (TIF 1295 kb) Additional file 6: Figure S3. Western blot analysis of total TAU and pTAU protein. a-e Densitometric analysis of TAU phosphorylated at different epitopes: S262, S396, S202/T205, T181, and S400/T403/S404. All samples were analyzed at days 42, 56, and 70 of terminal differentiation. The amount of pTAU relative to total TAU levels in the lysates was measured. GAPDH as the loading control was used to normalize the data. All values are the mean ± SEM (n = 3). Dunnett's test was performed to evaluate the significance of groups compared with control (*p < 0.05). (TIF 804 kb) Additional file 7: Figure S4. Analysis of GSK3B and active form of GSK3B in neuronal culture. a Quantification of active GSK3B form in control neurons (ctrl-1), fAD neurons (fAD-1-fAD-4), and sAD neurons (sAD-1-sAD-6) at days 42, 56, and 70 of terminal differentiation was presented as a percentage of nonphosphorylated GSK3B at Ser9 (inactive form of the kinase [102]). b Densitometric analysis of TAU phosphorylated at T231 epitope at days 42, 56, and 70 of terminal differentiation. All values were normalized to GAPDH and are presented as mean ± SEM (n = 3). Dunnett's test was performed to evaluate the significance of groups compared with control (*p < 0.05). (TIF 384 kb)