Cholinergic-like neurons carrying PSEN1 E280A mutation from familial Alzheimer’s disease reveal intraneuronal Aβ42 peptide accumulation, hyperphosphorylation of TAU, oxidative stress, apoptosis and Ca2+ flux dysregulation: Therapeutic Implications

Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by progressive memory loss and cognitive disturbance as a consequence of the loss of cholinergic neurons in the brain, neuritic plaques and hyperphosphorylation of TAU protein. Although the underlying mechanisms leading to these events are unclear, mutations in presenilin 1 (PSEN1), e.g., E280A (PSEN1 E280A), are causative factors for autosomal dominant early-onset familial AD (FAD). Despite advances in the understanding of the physiopathology of AD, there are no efficient therapies to date. Limitations in culturing brain-derived live neurons might explain the limited effectiveness of AD research. Here, we show that mesenchymal stromal (stem) cells (MSCs) can be used to model FAD, providing novel opportunities to study cellular mechanisms and to establish therapeutic strategies. Indeed, we cultured MSCs with the FAD mutation PSEN1 E280A and wild-type (WT) PSEN1 from umbilical cords and characterized the transdifferentiation of these cells into cholinergic-like neurons (ChLNs). PSEN1 E280A ChLNs but not WT PSEN1 ChLNs exhibited increased intra- and extracellular Aβ42 peptide and TAU phosphorylation (at residues Ser202/Thr205), recapitulating the molecular pathogenesis of FAD caused by mutant PSEN1. Furthermore, PSEN1 E280A ChLNs presented oxidative stress (OS) as evidenced by the oxidation of DJ-1Cys106-SH into DJ-1Cys106-SO3 and the detection of DCF-positive cells and apoptosis markers such as activated pro-apoptosis proteins p53, c-JUN, PUMA and CASPASE-3 and the concomitant loss of the mitochondrial membrane potential and DNA fragmentation. Additionally, mutant ChLNs displayed Ca2+ flux dysregulation and deficient acetylcholinesterase (AChE) activity compared to control ChLNs. Interestingly, the inhibitor JNK SP600125 almost completely blocked TAU phosphorylation. Our findings demonstrate that FAD MSC-derived cholinergic neurons with the PSEN1 E280A mutation are a valid model of AD and provide important clues for the identification of targetable pathological molecules.

Alzheimer's disease (AD) is a chronic neurodegenerative condition characterized by loss of memory, reasoning and decision-making functions [1] due to the severe loss of cholinergic neurons from the nucleus basalis magnocellularis of Meynert and cholinergic projections to the cortex and hippocampus [2]. The neuropathological profile of AD is associated with the extracellular accumulation of insoluble forms of amyloid-β (Aβ) in plaques and intracellular aggregation of the microtubule protein TAU in neurofibrillary tangles [3]. Aβ is derived by the proteolytic cleavage of amyloid  precursor protein (APP). APP is first cleaved by β-secretase, which then undergoes additional cleavages by γ-secretase to generate a series of peptides prone to aggregation. Most mutations in the presenilin 1 (PSEN 1) gene, which codes for the catalytic component of γ-secretase [4], result in the overproduction of Aβ, specifically, the 42-amino acid Aβ isoform (Aβ 1-42 , hereafter Aβ 42 ) [5], and occur most frequently in familial AD (FAD; http://www.molgen.ua.ac.be/ADMutations/). Glu280Ala (p. E280A, c.839A > C, exon 8) in PSEN1 is a well-characterized FAD mutation found in a large kindred localized in Antioquia, Colombia [6][7][8][9] that shows typical phenotypes of AD with complete penetrance [10]. Similar to the majority of dominant-negative PSEN1 mutations [11,12], PSEN1 E280A produces increased Aβ 42 deposition [13], hippocampal neuron loss [14], and A TAU accumulation in young adults [15,16].
Despite advances in the understanding of the physiopathology of AD [17], there are no efficient therapies to date. Although limitations in culturing brain-derived live neurons might slow AD research, the rapid advances in cellular genetic reprogramming, in particular the induction of somatic cells (e.g., fibroblast) into stem cells (e.g., human induced pluripotent stem cells, hiPSCs), has led to the modeling of FAD PSEN1 mutations in vitro [18][19][20][21]. Obtaining iPSCs from patients bearing PSEN1 mutations is appealing; however, the isolation and purification procedures are technically challenging, expensive, time consuming and labor intensive. Alternatively, the human mesenchymal stromal (stem) cells derived from Wharton's jelly tissue (WJ-MSCs) are multipotent cells that can differentiate and/or transdifferentiate into mesodermal and ectodermal lineage cells [22][23][24][25].
Because MSCs might be equivalent to human embryonic stem cells (hESCs) and hiPSCs [26,27], these cells have become an interesting and promising tool for modeling FAD PSEN1 E280A in vitro.
The aim of the present study was to establish an in vitro cellular model that reveals the major pathologic features of the FAD PSEN1 E280A mutation, thereby enabling investigation of the pathomechanisms of early onset FAD. Therefore, Aβ 42 production, TAU phosphorylation, oxidative stress (OS), cell death, and neuronal dysfunction were investigated in cholinergic-like neurons (ChLNs) derived from wild-type (control) and PSEN1 E280A MSCs. We demonstrate for the first time that FAD PSEN1 E280A pathology can be recapitulated in MSC-derived ChLNs. These findings in ChLNs show great promise for modeling human FAD in vitro and identifying therapeutic targets for AD treatment.

Materials and Methods
The collection and use of umbilical cords from newborns were approved by the Ethics

Isolation and Expansion of hWJ-MSCs
The human umbilical cords were obtained from ten healthy, natural childbirths (Tissue

Identification of the PSEN1 E280A mutation
The PSEN1 E280A mutation was detected by PCR using mismatch primers and digestion of the products with Bsm1 [28]. Digested products were separated on a 3% agarose gel.
According to the electrophoretic patterns, the samples were classified as wild-type (WT) or mutant PSEN1 E280A. Based on this classification, the TBC# WJMSC-11 WT PSEN1 and TBC# WJMSC-12 PSEN1 E280A cell lines were selected for further experiments.

APOE genotyping analysis
Genotyping of the APOE polymorphism was performed using polymerase chain reaction amplification of a 244-bp fragment followed by digestion with HhaI as described by [29]. After a new centrifugation, cells were fixed with freshly prepared Carnoy's solution.

Karyotyping
Metaphase spreads were analyzed after staining with quinacrine (Sigma) for karyotyping.
Analysis was performed on three different primary cultures counting 20 metaphases for each sample.

Colony-forming units assay
The colony formation assay is an in vitro cell survival assay based on the ability of a single cell to grow into a colony [30]. WJMSC-11 WT and WJMSC-12 PSEN1 E280A cells were seeded at a density of 200 cells/ well on 6-well plates followed by the addition of 3 mL of regular culture medium. The cultures were left to grow in a humidified atmosphere with 5% CO 2 at 37 °C for 15 days. The culture medium was changed twice a week. After 15 days of cultivation, both WT and mutant PSEN-1 cells were stained with 0.5% crystal violet and counted using the cell counter plugin from ImageJ program. The experiment was conducted three times.

Adipogenic differentiation
Adipogenic differentiation was performed according to [31] with minor modifications.

Osteogenic differentiation
Osteogenic differentiation was performed according to [31] with minor modifications.
Briefly, WT and mutant cells at passages 4-7 were plated at a density of 10,000 cells/ cm 2 in 12-well plates in regular culture medium. After 72 h, the culture medium was replaced by osteogenic differentiation medium containing high-glucose DMEM (Sigma), 10% FBS, 1 µM dexamethasone (Alfa Aesar, cat # A17590), 250 µM sodium ascorbate (Sigma, cat # A4034), and 10 mM β-glycerophosphate (Alfa Aesar, cat # L03425). The medium was changed every 3-4 days. Control cells were kept in regular culture medium. After 28 days of induction, cells were fixed in 4% FA and stained with standard Von Kossa Staining.

Chondrogenic differentiation
Chondrogenic differentiation was performed according to [32] with minor modifications.
After 28 days of induction, cells were fixed in 4% formaldehyde, stained with toluidine blue for 2 min at room temperature and viewed by light microscopy.

Cholinergic-like neuron (ChLN) differentiation
ChLN differentiation was performed according to [25]. WT and mutant MSCs were seeded at 1-1.5 x 10 4 cells/ cm 2 in laminin-treated culture plates for 24 h in regular culture medium.
Then, the medium was removed, and cells were incubated in minimal culture medium

Immunofluorescence analysis
For the analysis of neural-, Alzheimer's disease-, oxidative stress-and cell death-related  [34]. Briefly, after solubilization of the peptide (Sigma Cat #A9810) in DMSO, the "unaggregated" peptide was obtained by dissolving the DMSO-solubilized peptide in water and used immediately (0 days). To obtain the "large oligomers", 10mM Tris was added to DMSO-solubilized peptide solution and incubated it for 15 days at 4 °C.
The determination of the aggregation state of Aβ42 was performed by Western analysis of SDS-PAGE as described above. The assessment was repeated three times in independent experiments.

Evaluation of intracellular hydrogen peroxide (H 2 O 2 ) by fluorescence microscopy
To determine the levels of intracellular H 2 O 2 , we used 2′,7′-dichlorofluorescein diacetate ( Quantitative data and figures from the sub-G 0 /G 1 population were obtained using FlowJo 7.6.2 Data Analysis Software. The assessment was repeated three times in independent experiments blind to experimenter and flow cytometer analyst.

Acetylcholinesterase activity measurement
We determined the acetylcholinesterase (AChE) activity using the AChE Assay Kit (Abcam, Cat# ab138871) according to the manufacturer's protocol. Briefly, ChLNs at days 0, 2 and 4 post differentiation were detached with 0.25% trypsin and mechanically lysed by freezing/sonication. Lysates were centrifuged at 13,000 rpm for 15 min, and supernatants were used for protein quantification by the BCA method (see above) and the detection of AChE activity. AChE degrades the neurotransmitter acetylcholine (ACh) into choline and acetic acid. We used the DTNB (5,5′-dithiobis(2-nitrobenzoic acid)) reagent to quantify the thiocholine produced from the hydrolysis of acetylthiocholine by AChE. The absorption intensity of the DTNB adduct was used to measure the amount of thiocholine formed, which was proportional to AChE activity. We read the absorbance in a microplate reader at ~410 nm. The data obtained were compared to the standard curve values, and the AChE amounts (mU) were normalized to protein values (mU/ mg protein). The assessment was repeated three times in independent experiments blind to experimenter.

Measurement of A 1-42 peptide in culture medium
The level of A 1-42 peptide was measured according to a previous report [35]  The peaks of the fluorescence transients were found by calculating the average of four consecutive points and identifying those points that gave the highest average value (F max ).
The amplitudes of the Ca 2+ -related fluorescence transients were expressed relative to the resting fluorescence (ΔF/F) and were calculated by the following formula: ΔF/F=(F max -F rest )/(F rest -F bg ). For the calculation of the fluorescence intensities, ImageJ was used. The terms fluorescence intensity was used as an indirect indicator of intracellular Ca 2+ concentration. The assessment was repeated three times in independent experiments blind to experimenter.

JNK inhibition experiment
The ChLNs were left in regular medium for 0, 2 and 4 days alone or co-incubated with the anthrapyrazolone JNK inhibitor SP600125 (1 µM final concentration). This compound competes with ATP to inhibit the phosphorylation of c-JUN. After this time, cells were evaluated for p-TAU and t-TAU protein expression by immunofluorescence, as described above. The assessment was repeated three times in independent experiments blind to experimenter.

Photomicrography and image analysis
Light microscopy photographs were taken using a Zeiss Axiostart 50 Fluorescence

Data analysis
In this experimental design, a vial of MSCs was thawed, cultured and the cell suspension was pipetted at a standardized cellular density of 2.6 x 10 4 cells/cm 2 into different wells of a 24-well plate. Cells (i.e., the biological and observational unit [38]) were randomized to wells by simple randomization (sampling without replacement method), and then wells (i.e., the experimental units) were randomized to treatments by similar method.
Experiments were conducted in triplicate wells. The data from individual replicate wells were averaged to yield a value of n=1 for that experiment and this was repeated on three occasions blind to experimenter and/ or flow cytometer analyst for a final value of n=3 [38]. Based on the assumptions that the experimental unit (i.e. the well) data comply with independence of observations, the dependent variable is normally distributed in each treatment group (Shapiro-Wilk test), and there is homogeneity of variances (Levene's test), the statistical significance was determined by a One-way analysis of variance (ANOVA) followed by Tukey's post hoc comparison calculated with GraphPad Prism 5.0 software.
Additionally, MSCs were transdifferentiated into cholinergic-like neurons (ChLNs) from WJ-MSCs using a new method [25]. As shown in Figure 2, WT PSEN1 and PSEN1 E280A WJ-MSCs cultured in minimal culture medium (MCm) for 7 days expressed basal levels of protein MAP2 ( Fig. 2A and 2B) and  Tub III ( Fig. 2A and 2C) and undetectable levels of GFAP ( Fig. 2A and 2D) and ChAT ( Fig. 2A and 2E). As expected, when the cells were exposed to cholinergic-N-Run medium (Ch-N-Rm) for 7 days [25], the levels of protein MAP2 ( Fig. 2A and B),  Tub III ( Fig. 2A and 2C) and ChAT ( Fig. 2A and 2E) were significantly higher than those in cells exposed to MCm. Noticeably, ChLNs remained negative for the specific glial cell lineage marker GFAP ( Fig. 2A and 2D). These observations were confirmed by immunofluorescence ( Fig. 2G-J). Because the enzyme AChE catalyzes the breakdown of the neurotransmitter acetylcholine (ACh, [40], we evaluated whether ChLNs expressed a catalytically functional AChE enzyme. As shown in  Fig. 1A-J). Moreover, both WT and PSEN1 E280A MSCs displayed normal ∆Ψ m (Suppl. Fig. 2A), undetectable levels of ROS (Suppl. The above observations prompted us to evaluate the same cell parameters in ChLNs. Therefore, WT PSEN1 and PSEN1 E280A ChLNs were left in regular culture medium until 0, 2 and 4 days post transdifferentiation. We initially verify that the A 42 antibody (e.g., cat# AB5078P, Sigma-Aldrich) recognize monomers as well as oligomers of A 42 . Effectively, the A 42 antibody was capable to recognize the "unaggregated" (monomers) and large oligomeric forms of synthetic A 42 peptide (Fig. 3A). Western blot revealed that WT PSEN1 ChLNs displayed low or undetectable levels of intracellular A 42 oligomers ( Fig. 3A and 3B) and oxidized DJ-1 (Fig. 3A and 3C), whereas flow cytometry showed no loss of  m (Fig. 4A-C) and no ROS generation (Fig. 4D-F) at any time tested. However, PSEN1 E280A ChLNs exhibited significantly higher levels of intracellular A 42 oligomers ( Fig. 3A and 3B) and oxidized DJ-1 (Fig. 3A and 3C) and lower  m than WT PSEN1 ChLNs at days 2 and 4 ( Fig. 4B and 4C). These observations were confirmed by immunofluorescence microscopy (Fig. 3D-K, and Fig. 4J'-O'). Interestingly, PSEN1 E280A ChLNs showed high levels of ROS production as early as day 0 (Fig. 4D, M'
The above findings compelled us to evaluate whether ChLNs expressed a catalytically functional acetylcholinesterase (AChE) enzyme at similar times post differentiation. As shown in Figure 6K, the AChE enzyme showed similar catalytic activity at days 2 and 4 in both WT PSEN1 and PSEN1 E280A ChLNs. However, the AChE enzyme activity in PSEN1 E280A ChLNs was significantly lower (at least 5-fold) than that in WT PSEN1 ChLNs at day 4.

Discussion
Currently, the neuropathology of AD includes extracellular deposits of A in plaques, intracellular neurofibrillary tangles comprising hyperphosphorylated TAU, synaptic dysfunction, and neuronal death. In an effort to explain such observations, several theories have been proposed [46]; however, the amyloid cascade hypothesis has prevailed for more than 25 years [47]. The A hypothesis postulates that an imbalance in the production of extracellular Aβ 42 plaques by mutations in at least three genes (e.g., APP, PSEN1; PSEN2) is an early initiating factor in AD. However, experimental therapies targeting A have thus far been unsuccessful [48]. Several factors have probably contributed to the failures in AD drug development, including unsuitable preclinical research models that do not fully recapitulate the human disease; consequently, druggable targets remain missing [49]. the catalyzation of the formation and production of ROS (e.g., H 2 O 2 ) by A metal (e.g., Cu 2+ , Cu 1+ , Fe 2+ ) complexes [54]; and targeting of mitochondria [55]. Whatever the mechanism, we demonstrated for the first time that Aβ 42 endogenously produces H 2 O 2 in PSEN1 E280A ChLNs. Because H 2 O 2 can function as a second messenger [56], it might also activate other redox proteins, such as apoptosis-signal regulating kinase 1 (ASK-1, [57], which in turn directly or indirectly activate other signaling pathways, e.g., JNK/c-JUN [58]. We found that Aβ 42 induced phosphorylation of c-JUN [59], p53 and PUMA in PSEN1 E280A ChLNs. Notably, c-JUN-and p53-dependent apoptosis is triggered by transactivation of the pro-apoptotic gene PUMA [60][61][62]. Remarkably, JNK can also stabilize and activate p53 [63]. Taken together, these results suggest that Aβ 42  H 2 O 2 activates a cascade of events leading to the JNK> c-JUN, p53 > PUMA pathway. Although PUMA has been shown to cooperate with direct activator proteins (e.g., BAX) to promote mitochondrial outer membrane permeabilization (MOMP) and apoptosis [64], the exact mechanism by which MOMP occurs is not fully understood [65]. Interestingly, PSEN1 Neuronal calcium (Ca 2+ ) dyshomeostasis has been proposed to play a crucial role in AD disease progression [68]. However, the mechanisms of Ca 2+ dysregulation are not clear. In contrast to Demuro and Parker [69], who found that intracellular  [70].
We found that the PSEN1 E280A ChLN response to ACh was significantly reduced by day 4 post transdifferentiation. Notably, A has been shown to directly affect 7 nicotinic ACh receptor (7 nAChR) function by acting as an agonist (~100 nM) and a negative modulator (at high concentrations) [71]. Consistent with this view, we confirmed that PSEN1 E280A ChLNs secreted aberrant amounts of Aβ 42 e.g., ~2500-f.i.) compared to WT PSEN1 ChLNs These observations confirm that overproduction of extracellular Aβ 42 is a paramount feature of the majority of PSEN1 mutations in vitro and in vivo [5], including the E280A mutation [21]. Despite these observations further investigation is required to determine whether 7 nAChRs specifically are affected by A  in PSEN1 E280A ChLNs.
In contrast to others (e.g., [72]), our observations do not support the common view that extracellular A is capable of increasing neuronal Ca 2+ flux through Aβ 42 -forming pores.
However, we do not discard the possibility that given a longer incubation time,  [75,76]. Moreover, these findings comply with the notion that Aβ 42 accumulation affects TAU pathology [77]. However, the molecular link between A and TAU is still not yet completely defined. In agreement with others [78], our data suggest that JNK is a strong candidate TAU kinase involved in the hyperphosphorylation of TAU in PSEN1 E280A ChLNs. This assumption is supported by two observations. First, JNK phosphorylates TAU at Ser 202 / Thr 205 [45], two phosphorylation epitopes identified in the present study. Second, PSEN1 E280A ChLNs exposed to the JNK inhibitor SP600125 significantly reduced TAU phosphorylation. Given that JNK plays a pivotal role in both OS-induced apoptosis and TAU phosphorylation, these findings identify JNK as a potential therapeutic target [79]. Although we do not discard the possibility that other kinases might be implicated in TAU pathology (e.g., LRRK2, GSK-3, Cdk5) , our findings suggest that JNK plays a key role in TAU hyperphosphorylation in PSEN1 E280A ChLNs.
Outstandingly, MSC-derived ChLNs not only replicate the pathophysiology of AD, i.e., intracellular accumulation of A42 and TAU phosphorylation as similarly reported in the PSEN1 iPSC model [19,20], but also replicate the intracellular aggregation of A42 and OS phenotypes in AD model iPSCs [18,80]. Therefore, MSC-derived PSEN1 neurons should be considered equivalent to iPSC-derived PSEN1 neurons. Moreover, our results suggest that PSEN1 E280A-induced neural alterations may precede Aβ 42 deposition and that those alterations represent longstanding effects of intracellular Aβ 42 oligomeric toxicity and possibly even developmental changes. The molecular alterations might start when neurons develop into neuron-specific cholinergic-type cells or may even exist at birth.
These findings may explain why functional and structural brain changes manifest in children (9-17 years old) and young individuals (18-26 years) who are carriers of the PSEN1 E280A mutation [81,82]. Furthermore, these observations suggest that intracellular Aβ 42 oligomeric toxicity is an early and slowly progressive process that might damage neuronal cells in a TAU-dependent and independent fashion (OS,  m shutdown, apoptosis and intraneuronal Ca 2+ dysregulation) more than two decades before the stage of dementia [6,83].