The Alzheimer Disease-Causing Presenilin-1 L435F Mutation Causes Increased Production of Soluble Aβ43 Species in Patient-Derived iPSC-Neurons, Closely Mimicking Matched Patient Brain Tissue

Abstract

Familial Alzheimer disease-causing mutations in Presenilin 1 (PSEN1) are generally thought to shift the processing of APP toward longer, more amyloidogenic Aβ fragments. However, certain PSEN1 mutations cause severe reduction in gamma secretase function when expressed in the homozygous state, thus challenging the amyloid hypothesis. We sought to evaluate the effects of one such mutation, PSEN1 L435F, in more physiologic conditions and genetic contexts by using human induced pluripotent stem cell (iPSC)-derived neurons from an individual with familial AD (fAD) linked to the PSEN1 L435F mutation, and compared the biochemical phenotype of the iPS-derived neurons with brain tissue obtained at autopsy from the same patient. Our results demonstrate that in the endogenous heterozygous state, the PSEN1 L435F mutation causes a large increase in soluble Aβ43 but does not change the overall levels of soluble Aβ40 or Aβ42 when compared with control iPSC-neurons. Increased pathologically phosphorylated tau species were also observed in PSEN1-mutant iPSC-neurons. Concordant changes in Aβ species were present in autopsy brain tissue from the same patient. Finally, the feasibility of using Aβ43 immunohistochemistry of brain tissue to identify fAD cases was evaluated in a limited autopsy case series with the finding that strong Aβ43 staining occurred only in fAD cases.

INTRODUCTION

The amyloid hypothesis, strongly supported by both genetic and neuropathologic evidence, has been central to much Alzheimer disease (AD) research. For example, the hypothesis that accumulation of Aβ represents the critical initiating event in the pathogenesis of AD lies at the heart of many ongoing clinical trials for early intervention (1). The observation that presenilin is the catalytic unit of gamma secretase, which generates Aβ, and that mutations in presenilin lead to autosomal dominant early onset AD, is frequently cited as strong support for the amyloid hypothesis. However, some familial AD (fAD)-causing mutations in Presenilin 1 (PSEN1) result in severe reduction of gamma secretase activity (2). This observation raises the alternative possibility that reduced gamma secretase enzyme activity, instead of Aβ, may cause downstream changes resulting in AD pathology. The fAD-causing PSEN1 L435F mutation is within this subcategory of mutations and shows greatly reduced activity in vitro in the homozygous state in multiple model systems (2, 3).

However, the effects of the PSEN1 L435F mutation are uncertain. Some studies suggest a near-complete loss of function, while others favor reduced function and a bias toward the production of 43 amino acid Aβ species (Aβ43) (3–7). Overexpression of the PSEN1 L435F mutation in PSEN2 knockout mouse embryonic fibroblasts or HEK cells results in an absolute increase in Aβ43 species compared with wildtype PSEN1 (4, 5). Discordantly, heterozygous expression of PSEN1 L435F in PSEN2 knockout mice causes an overall decrease in the amount of Aβ43 species present in brain tissue when compared with wildtype PSEN1 PSEN2–/– mice (7). Aβ43, although low in abundance when compared with the more-studied Aβ40 and Aβ42 species, is potently amyloidogenic, akin to Aβ42, and is present at increased levels in fAD and sporadic AD (sAD) brain lysates (8, 9). Thus, understanding the effects of the PSEN1 L435F mutation on Aβ43 production in human cells carrying the allele in heterozygous form would improve knowledge of how this mutation might contribute to the development of AD.

Induced pluripotent stem cells (iPSCs) are well-suited to addressing this question. iPSCs share the complete genetics of donor patients and have been extensively used to model AD in vitro in human neurons. In general, human iPSC-neuron models of fAD show a relative increase in the production of Aβ42 compared with Aβ40 and often develop increased levels of pathologically phosphorylated tau protein downstream of these changes (10, 11). More recently, Aβ43 species have been a focus of investigation. iPSC-neurons harboring multiple AD-causing PSEN1 mutations produce increased Aβ43 compared with wildtype cells (12). With regard to the PSEN1 L435F mutation, there is no reported human iPSC model.

We took advantage of an ongoing brain donation program to derive an iPSC model from an individual with fAD and the PSEN1 L435F mutation and a sex and ApoE genotype matched control. There was a full neuropathologic examination of the brain from both subjects, confirming the diagnosis according to current criteria (13, 14). We were then able to clarify the overall effects of the PSEN1 L435F mutation on the levels of Aβ secreted by human iPSC-neurons and on the levels of phosphorylated tau (pTau) present within these cells. By pairing the PSEN1 mutant and control cell lines with matched donated brains, we compared our findings in iPS-derived neurons with that observed directly in matched brain tissue (Fig. 1). In addition, the observation that Aβ43 species are increased in both fAD iPSCs and brain tissue led us to evaluate the feasibility of using Aβ43 immunohistochemistry to screen for fAD cases within archival patient brain tissues.

FIGURE 1.

Experimental overview.

MATERIALS AND METHODS

Plasmids and Lentiviral Production

pLVX-TetOne-Puro-hNGN2 (Clontech, Mountain View, CA) lentiviral plasmid for constitutive expression of puromycin resistance and doxycycline-inducible expression of human neurogenin 2 (NGN2) was a gift from K.A. Worringer (Novartis Institutes for BioMedical Research, Boston, MA). VSVG-pseudotyped lentivirus was prepared and concentrated via ultracentrifugation as previously described (15).

Brain Procurement, Lysate Preparation, and Immunostaining

All human brain tissues were obtained from the Massachusetts Alzheimer’s Disease Research Center (MADRC, P30AG062421). PBS-soluble brain extracts from frontal cortex were prepared by dounce-homogenization in ice-cold PBS–/– with protease inhibitor (Cell Signaling, Cat #5871S) followed by centrifugation at 10 000g for 10 minutes. Five-micron formalin-fixed paraffin-embedded brain sections were obtained from the MADRC. Immunostaining of brain sections for Aβ (DAKO, Santa Clara, CA. 6F/3D, 1:600) and Aβ43 (ThermoFisher, Waltham, MA, Cat# 44-340, 1:500) was performed using a Bond Rx autostainer with the following settings: Bake and Dewax, IHC Protocol F 60 minutes, HIER 20 minutes with ER2 (Leica Biosystems, Buffalo Grove Il). Stained slides were imaged on an Olympus BX40 microscope and photographed using a QColor3 camera and QImaging software (Olympus, Tokyo, Japan).

Neuropathologic Evaluation

Autopsy evaluation of each donated brain was carried out by a board-certified neuropathologist at the MADRC and was consistent with current guidelines for evaluation of neurodegenerative disease (13, 14), including immunohistochemistry for Aβ, tau, α-synuclein, and phospho-TDP-43, as well as assessments for vascular disease.

Primary Fibroblast Line Derivation

At the time of autopsy, consent was also obtained for iPSC generation and genetic studies according to a protocol approved by the Partners Institutional Review Board. Primary fibroblast lines were derived via explant culture at the time of autopsy from 2 patients (Table 1; lines C1 and fAD1). A sterile 1–2 cm² piece of skin was removed from the inner thigh. Skin samples were then transported in sterile saline and held at 4°C until dissection. Subcutaneous fatty tissues were dissected away and the tissue was subdivided into 1–2 mm² fragments. Tissue fragments were allowed to attach to treated T75 flasks (Corning CellBind, Corning, NY) at RT for ∼5 minutes, then fed with low volume fibroblast media (5–7 mL). Media was changed once/week until cells reached 80%–90% confluence at ∼4 weeks. Tissue fragments were then removed and fibroblasts were trypsinized and cryopreserved at p0–p2 using Synth-a-Freeze freezing media (ThermoFisher) in isopropanol freezing chambers. Fibroblast culture medium consisted of DMEM, 10% fetal bovine serum, l-glutamine (2 mM), and Antibiotic-Antimycotic (1×) (Gibco, ThermoFisher, Waltham, MA).

TABLE 1.

iPSC Cell Line Patient Demographics

Case #	Age	Sex	Diagnosis	Mutation	Braak Stage	Thal Phase	CERAD Density	Diffuse Aβ	CAA (0–3)	HTN (0–3)	TDP-43	ApoE	Brain Weight	Race
2012, C1	>90	F	Control	N/A	II	0	0	None	0	2	Neg.	3/3	1120	W
2048, fAD1	53	F	Familial AD	PSEN1 L435F (c.1303C>T)	VI	3	3	Frequent	3	0	Neg.	3/3	1180	W

Braak, Thal, and CERAD staging as described (14); CAA, cerebral amyloid angiopathy; HTN, hypertensive cerebrovascular disease; AopE, ApoE genotype.

iPSC Generation and Culture

iPSCs were commercially generated from primary fibroblast cultures using mRNA reprogramming (16) (Cellular Reprogramming Inc., Pasadena, CA). iPSC cultures were maintained in mTeSR media (StemCell Technologies, Vancouver, CA) on Matrigel coated dishes (Corning). Routine passaging was performed using Accutase (Gibco, ThermoFisher). Routine karyotyping, mycoplasma testing, and STR analysis were performed at WiCell (Madison, WI).

iPSC Neuronal Differentiation

iPS-neurons were produced by doxycycline-driven expression of hNGN2 combined with SMAD and WNT inhibition as previously described, with minor modifications (17). For stable integration of the inducible NGN2 lentivirus, 300K iPSCs were plated as single cells into one well of 6-well plate in the presence of thiazovivin and infected with pLVX-TetOne-Puro-hNGN2 lentivirus O/N at 24 hours after plating. Twenty four hours after lentiviral transduction, cells were passaged with accutase treatment into a 10-cm plate and selected with 5 µg/mL puromycin 24 hours later (Gibco, ThermoFisher, Cat# A11138-03). Resulting puromycin-resistant iPSCs were expanded and cultured under continuous puromycin selection. At Day (D0), iPSCs stably expressing the inducible NGN2 construct were passaged as single cells with accutase and plated at 400–500 k cells per well on Matrigel coated 6-well plates in mTeSR medium (Stemcell Technologies) with rock inhibitor (thiazovivin, 1 µM, EMD Millipore, Burlington, MA). Doxycycline (2 µg/mL, Sigma, St. Louis, MO) was added on plating at D0 to induce NGN2 expression and maintained in culture medium thereafter. On D1, medium was switched to N2 media (DMEM: F12 [Gibco], Glutamax [1%, Gibco], dextrose [0.3%, EM Science, Hatfield, PA], N2 [1%, Gibco]) supplemented with SB431542 (10 µM, Tocris, Bristol, UK), LDN-193189 (100 nM, Stemgent, Cambridge, MA), XAV939 (2 µM, Stemgent), and doxycycline. On D2, cells were fed with N2 medium with SB/XAV/LDN at one-half the concentration of D1 and doxycycline at full concentration. On D3, cells were fed with N2 medium supplemented with NT3 (10 ng/mL, PreproTech, Rocky Hill, NJ), BDNF (10 ng/mL R&D Systems, Minneapolis, MN), GDNF (10 ng/mL, R&D Systems), and doxycycline. On D4, cells were switched to NBM media (neurobasal medium, minus phenol red [Gibco], glutamax [1%], dextrose [0.3%], NEAA [0.5%, Gibco], and B27 [2%, Gibco]) supplemented with NT3, BDNF, GDNF, doxycycline, and the antimitotic FUdR (10 µM, Sigma). Cells were fed with NBM media plus D4 supplements every 2–3 days from D4 until D14, at which point FUdR was removed from the media and cells were fed weekly thereafter. Two percent Horse serum (Gibco 26-050-088) was added to media beginning at D8 to improve neuronal survival. For immunohistochemistry, differentiating neurons were passaged at D4 using accutase treatment and plated onto 96 well plates (30–40k cells/well) with 2% horse serum added at this time to improve survival. Ninety six well-plates were precoated with poly-d-lysine (Corning 356640) and further coated with laminin (10 µg/mL, Sigma), fibronectin (2 µg/mL, Sigma), and Matrigel (2.5×, Corning) in DMEM:F12 for 3 hours at RT. Where indicated, the gamma-secretase inhibitor DAPT (10 µM, Sigma) was applied with media changes.

Western Blot

Cells were harvested at D28 on ice into ice-cold RIPA buffer with protease inhibitor (Cell Signaling, Danvers, MA, Cat #5871S) using a cell scraper. After scraping, cells were lysed with 10 passes using insulin syringes and pelleted for 10 minutes × 10 kg at 4°C. Resulting supernatant and pellets were saved at –80°C. Western blots were performed on 10 µg samples of soluble supernatants using the Invitrogen NuPage Novex Gel System according to the manufacturer’s instructions (ThermoFisher). Fluorescent secondary antibodies were used at a concentration of 1:5000. Blots were imaged on a LI-COR Odyssey system and analyzed using LI-COR Image Studio (LI-COR Biosciences, Lincoln, NE).

iPSC Genotyping

Genomic DNA was extracted from iPSC cultures of each line and isolated using the Qiagen DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany). The PSEN1 L435F locus was PRC-amplified, gel-purified, and subsequently genotyped using Sanger sequencing at the Massachusetts General Hospital sequencing facility, using PCR primers as previously described (5). PSEN1 forward primer, 5′-3′: TTGCCTGAAAATGCTTTCATAATTAT. PSEN1 reverse primer, 5′-3′: GGAATGCTAATTGGTCCATAAAAG. ApoE genotyping was performed on genomic DNA isolated from iPSC lines using a TaqMan SNP Genotyping Assay according to the manufacturer’s instructions (ThermoFisher) with allele-specific primer sets (APOE4-112 and APOE4-158).

Enzyme-Linked Immunosorbent Assay

Enzyme-linked immunosorbent assay (ELISA) measurements of Aβ species were run according to the manufacturer’s instructions. Cell media was aspirated, centrifuged to remove debris, and frozen at –80°C prior to measurements, which were performed in single runs for each analyte. Supernatant was diluted 1:5 prior to measurement of Aβ40 and Aβ42, and run undiluted for Aβ43. Detection reagents were: Human Aβ 1–40 (298-64601, FUJIFILM Wako Chemicals U.S.A. Corporation, Richmond, VA), human Aβ 1–42 (WAKO 296-64401), and human Aβ 1–43 (27710, Immuno-Biological Laboratories, Inc., Minneapolis, MN). Assay plates were read on a Wallac Victor2 at 450 nM (Perkin Elmer, Waltham, MA).

Lactate Dehydrogenase Cytotoxicity Assay

Lactate dehydrogenase (LDH) assay was performed with the LDH-Glo cytotoxicity assay (76326-472, Promega, Madison, WI) according to the manufacturer’s protocols, utilizing the manufacturer’s LDH storage buffer to preserve cell culture supernatant at –80°C prior to analysis (200 mM Tris-HCl [pH 7.3], 10% glycerol, 1% BSA, sterile filtered, media diluted 10× into storage buffer). Assay plates were read on a Wallac Victor2 (Perkin Elmer).

iPSC/iPSC-Neuron Immunohistochemistry, Microscopy, and Image Analysis

iPSCs and iPSC-neurons were fixed with 4% paraformaldehyde in PBS for 30 minutes at RT. Cells were washed with PBS and permeabilized for 15 minutes at RT with 0.1% Triton X-100. Cells were then washed again with PBS and blocked for 1 hour at RT with PBS–/–, 5% NGS. Primary antibodies were applied O/N at 4°C in blocking media. Secondary antibodies were applied for 1 hour at RT at 1:500 in PBS–/–. DAPI was applied for 15 minutes at RT in PBS–/–. Stained cells were imaged on a Zeiss Imager Z2 microscope with Axiocam 512 (iPSCs) or a Biotek Cytation 5 microscope (iPSC neurons) (Carl Zeiss AG, Oberkochen, Germany, Biotek Instruments Inc., Winooski, VT). Neuronal images were analyzed with Cell profiler (18) to determine the overall percentage of neurons (% of nuclei with overlapping Map2/Tuj1 staining). Custom ImageJ macros were used to evaluate overall neurite density and the phospho-tau staining intensity within TUJ1-positive neurites.

Statistical Analyses

Calculation of standard deviation, standard error, one-tailed and two-tailed homoscedastic T-tests, and ANOVAs (with replication) were performed using Microsoft Excel Analysis ToolPak and R.

RESULTS

Generation and Characterization of Control and PSEN1 L435F iPSCs From Brain Donors

Two cases were selected for iPSC reprogramming: One >90-year-old cognitively normal female (MADRC_2012, C1) and one 53-year-old female fAD patient carrying the PSEN1 L435F mutation (MADRC_2048, fAD1) (Table 1; Fig. 2). Both cases were ApoE 3/3 genotype. The mutation carrier was a member of a previously described early onset fAD kindred with known PSEN1 L435F mutation (3). Full neuropathologic evaluation was performed on each patient to verify diagnosis and assess for concurrent neurodegenerative changes with immunohistochemistry for Aβ, tau, α-synuclein, and phospho-TDP-43 (see Table 1 for demographics and neuropathologic evaluation). The control patient’s brain showed moderate cerebrovascular disease and Braak stage II of VI neurofibrillary tangle (NFT) distribution, consistent with normal aging (14). There was no Aβ deposition in the control brain. The PSEN1 L435F mutation carrier demonstrated severe AD neuropathologic changes assessed as A2B3C3, with Thal stage 3 of 5 for Aβ distribution, Braak stage VI of VI for NFTs, and a CERAD score 3 of 3 for Aβ-positive neuritic plaque density (14). Primary dermal fibroblast lines from each donor were derived at the time of autopsy via explant culture (see Materials and Methods). Subsequently, iPS cells were derived from both patients using mRNA reprogramming and subjected to basic characterization assays. Both iPS lines demonstrate normal 46XX karyotype and uniformly express the pluripotency markers Oct-4, Nanog, Sox2, and Tra-1-60 by immunohistochemistry (Fig. 3A–D). Sanger sequencing of iPSC DNA confirmed the presence of the PSEN1 L435F mutation in the fAD patient, consistent with the documented clinical history (Fig. 3E).

FIGURE 2.

Gross and microscopic findings. (A) Gross appearance of PSEN1 L435F donor brain. (B) Luxol hematoxylin and eosin image of fAD entorhinal cortex showing abundant “cotton wool” plaques. (C) Tau immunohistochemistry of fAD hippocampus (CA2) demonstrating abundant NFTs around plaques. (D) Gross appearance of control donor brain. (E) Luxol hematoxylin and eosin image of control entorhinal cortex (F) Bielchowsky silver stain of control hippocampus (CA2).

FIGURE 3.

Pluripotency marker expression, karyotyping, and Sanger sequencing of control and PSEN1 L435F iPSCs. Immunohistochemistry for Oct4, Nanog, Sox2 and Tra-1-60 expression in control (A) and PSEN1 L435F (B) iPSCs. (C, D) Normal 46 XX karyotype in control (C) and PSEN1 L435F (D) iPSCs. (E) Sanger sequencing confirming heterozygous PSEN1 L435F (c.1303 C>T) mutation in the fAD line.

Neuronal Differentiation of iPSCs via NGN2 Induction Combined With SMAD and WNT Inhibition

A NGN2-based direct-conversion approach was used to derive high-purity cultures of neurons from control and PSEN1 mutant iPSCs (Table 1; lines C1 and fAD1). Lentivirus mediated introduction of doxycycline-inducible human NGN2 in combination with SMAD and WNT inhibition was used as previously described (17), but with the addition of a ubiquitously expressed puromycin resistance cassette to allow selection of iPSCs carrying the inducible transgene. With this approach, over 94% of cells were positive for the neuronal marker Beta-III tubulin at 28 days after the start of differentiation and there was no statistically significant difference in the percentage of neurons produced by either of the 2 iPSC lines (Control 94.6 ± 4.1% [SD] and PSEN1 L435F 94.6 ± 4.1%, p = 0.997, Fig. 4B, C, E–G). Map2 staining results were similar (Control 80.0 ± 11.3% [SD] and PSEN1 L435F 86.6 ± 8.5%, p = 0.467, Fig. 4A, D, G). Treatment of cells with the gamma secretase inhibitor DAPT (10 µM) did not significantly influence the percentage of Beta-III tubulin/Tuj1 or Map2-positive neurons (TUJ1: Control 98.4 ± 2.0% [SD] and PSEN1 L435F 96.9 ± 3.3%, p = 0.527, MAP2: Control 85.8 ± 5.0% [SD] and PSEN1 L435F 81.3 ± 15.9%, p = 0.665, Fig. 4G). Two factor ANOVA did not reveal significant contributions from either iPSC line or DAPT treatment for TUJ1 percentage (p = 0.977 for line and p = 0.203 for DAPT treatment) or MAP2 percentage (p = 0.873 for line and p = 0.963 for DAPT treatment). Neuronal cultures were grown without a glial feeder layer allowing isolation of neuronal phenotypes, although possibly at some expense to long-term culture stability. Lentivirus transduced iPSCs were passaged >10 times under continuous puromycin selection with no appreciable loss of neuronal differentiation efficiency.

FIGURE 4.

Neuronal differentiation of control and PSEN1 L435F mutant iPSCs. Neuronal marker expression in control (A–C) and PSEN1 mutant (D–F) iPSC-neurons stained for MAP2 (A, D) or TUJ1 (B, C, E, F) in green with or without DAPT treatment as indicated. Blue = DAPI, scale bar = 200 µM. (G) Quantification of percent neuronal nuclei by TUJ1 and MAP2 staining with and without DAPT treatment. (H) Neurotoxicity minus culture media baseline (LDH, mU/mL). ± standard error.

PSEN1 L345F iPSC-Neurons Secrete Increased Levels of Aβ43 and Have Increased Aβ42/40 and Aβ43/40 Ratios

In order to evaluate the effect of heterozygous PSEN1 L435F mutation on Aβ secretion in human neurons, control and mutant neurons were produced in parallel and multiple Aβ species were measured in the culture medium using length-specific ELISA assays. iPS-neurons derived from control and PSEN1 L435F iPSCs were cultured until day 28 of differentiation, at which point media and cells were harvested. A subset of cultures was exposed to the gamma-secretase inhibitor DAPT for 14 days prior to media harvest. Compared with control, PSEN1 L435F iPSC-neurons produce dramatically increased levels of Aβ43 species with preserved overall levels of Aβ40 and Aβ42 (n = 5, Table 2; Fig. 5A). In addition, Aβ43/40 and Aβ42/40 ratios are increased in PSEN1 L435F iPSC-neurons (n = 5, Table 2; Fig. 5B). In the presence of the gamma secretase inhibitor DAPT, there was near-complete abrogation of Aβ40, Aβ42, and Aβ43 production with no significant differences in between-line comparisons (n = 5, Table 2; Fig. 5A). 43/40 and 42/40 ratios could not be measured in the presence of DAPT due to absence of detectable Aβ40 in the culture medium for both cell lines. Measurement of LDH in media to assess overall levels of cell death showed no significant difference over baseline between either iPSC line (p = 0.261) or in DAPT+ versus DAPT– conditions (p = 0.823) by two-way ANOVA (Fig. 4H).

TABLE 2.

Aβ Species ELISA Results

	Analyte (pmol/L)	Control iPSC-Neurons	PSEN1 L435F iPSC-Neurons	p Value
–DAPT	Aβ 1–43	0.03 ± 0.03	5.08 ± 1.45	0.008**
	Aβ 1–42	14.25 ± 3.32	15.73 ± 3.95	0.781
	Aβ 1–40	108.21 ± 30.31	97.95 ± 26.40	0.805
+DAPT	Aβ 1–43	0.00 ± 0.00	0.37 ± 0.37	0.347
	Aβ 1–42	0.00 ± 0.00	0.07 ± 0.07	0.347
	Aβ 1–40	0.00 ± 0.00	0.00 ± 0.00	n/a
Ratios	Aβ 43/40	0.0002 ± 0.0002	0.052 ± 0.004	1.79E–09***
	Aβ 42/40	0.14 ± 0.01	0.164 ± 0.004	0.025*

ELISA results are presented in pmol/L ± standard error. p Values are calculated for each control versus PSEN1 L435F comparison. n = 5.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

FIGURE 5.

Aβ species secreted by control and PSEN1 L435F iPSC-neurons. (A) Aβ43, Aβ42, and Aβ40 species measured in culture media at 28 days post NGN2 induction, both with and without gamma secretase inhibitor ± standard error. (B) Aβ43/40 and Aβ42/40 ratios ± standard error. *p < 0.05; **p < 0.01; and ***p < 0.001.

PSEN1 L435F iPSC-Neurons Show Increased pTau

Increased pathologic phosphorylation of tau is believed to occur downstream of Aβ production in AD and is variably seen in iPSC-neuron models (10). To assess for increased pTau species, D28 iPSC-neurons with and without DAPT treatment were immunostained for pTau (PHF1) and a pan-neuronal marker (TUJ1/βIII-tubulin). PHF1 staining was present in all neurites as expected for embryonic neuronal populations. However, average PHF1 stain intensity in neurites was increased in PSEN1 L435F neurons compared with control neurons both with and without DAPT treatment (p = 0.019 without DAPT and p = 0.030 with DAPT, Fig. 6A, B). Although within-cell-line pairwise comparisons for effect of DAPT treatment on PHF1 intensity were not significant (p = 0.234 for Control and p = 0.100 for PSEN1), two-way ANOVA revealed a significant contribution of both cell line (p = 0.001), and DAPT treatment on PHF1 staining intensity (p = 0.03). There was no significant difference in TUJ1-positive neurite density between the two lines or with/without DAPT treatment by two-way ANOVA (p = 0.177 for line and p = 0.981 for ±DAPT, Fig. 6C).

FIGURE 6.

PHF1 pTau expression. (A) Representative immunohistochemistry for TUJ1 and PHF1 for control and PSEN1 mutant iPSC-neurons with (+D) and without (–D) DAPT treatment (green = TUJ1, red = PHF1, blue = DAPI). Images are exposed and processed identically for each antibody. (B) PHF1 average neuronal intensity for both cell lines ± DAPT in a.u. ± standard deviation, *p < 0.05. (C) TUJ1 average neuronal density (pixels/image) for both cell lines ± DAPT ± standard deviation.

Additionally, Western blots were performed on D28 cultures of iPSC neurons to evaluate pTau levels. On Western blot, pTau (AT8) to total tau (DAKO) ratios are elevated in PSEN1 mutant neurons compared with wildtype neurons under standard culture conditions (p = 0.0314) and in the presence of the gamma secretase inhibitor (DAPT) (p = 0.0029) (Fig. 7A, B). DAPT treatment caused a reduction in AT8/Total Tau ratio in control neurons (statistically significant, p = 0.0070), but this did not reach statistical significance in PSEN1 mutant neurons (p = 0.1038). Two-way ANOVA revealed significant contributions to AT8/Total Tau ratio from both cell line (p = 0.0030) and DAPT treatment (p = 0.0289). In sum, the PSEN1 mutant line had more pTau at baseline than the control line; both control lines and PSEN1 mutant line had a decrement in pTau after DAPT treatment, although the PSEN1 line continued to have more. Given the relatively long half-life of pTau, it is difficult to interpret to what extent the elevated pTau in PSEN1 mutant line is due to Aβ production (or even to other gamma secretase functions), but prior studies are consistent with the elevation in pTau being due, at least in part, to the elevation of Aβ seen in the PSEN1 line.

FIGURE 7.

pTau/Total tau measurement by Western blot. (A) AT8/Total tau ratio measured by Western blot ± standard error for Control and PSEN1 mutant neurons at D28 with (+D) and without (–D) DAPT treatment. n = 4 for PSEN1 +D, n = 5 for other conditions. (B) Western blot for A: L, ladder; C, control; P, PSEN1; PBr, matched PSEN1 mutant frontal cortex; Rec, recombinant 2N4R HIS-tagged tau. *p < 0.05; **p < 0.01; and ***p < 0.001.

Because inhibition of Notch signaling is one effect of DAPT treatment, it is possible that DAPT-treated neurons would exhibit accelerated maturation, which might influence pTau levels (19). During early embryonic development, 3R isoforms of tau predominate with increased 4R production as time progresses (20). To assess for potential differences in neuronal maturation, the 4R tau protein expression was evaluated by Western blot at D28. Western blot with total tau (DAKO) and 4R-tau-specific antibodies did not detect 4R tau protein expression in either control or PSEN1 mutant neurons either with or without DAPT treatment (Supplementary DataFig. S1A, B).

Matched Patient Brain Tissues Show Concordant Levels of Soluble Aβ Species

In order to assess the biological relevance of increased Aβ43 production in PSEN L435F iPSC-neurons, Western blot was performed on soluble extracts of frontal cortex (BA9) procured at autopsy from the same donor used to derive the iPSC line (MADRC_2048, fAD1). In comparison to the control iPSC/brain donor (MADRC_2012, C1), frontal cortex from the PSEN1 L435F patient shows greatly increased soluble Aβ43, Aβ42, and Aβ40 species (p = 0.003, 0.0003, 0.0007, respectively, Fig. 8). This finding matches the above results with iPSC-neurons derived from the same patients. The level of Aβ43 in the PSEN1 L435F brain was also increased in comparison to another fAD case caused by the PSEN1 H163R mutation as well as 3 sAD cases and 2 additional control brains by one-way ANOVA (p < 0.006, Fig. 8; Supplementary Data).

FIGURE 8.

Aβ species in human brain measured by ELISA. ELISA measurement of Aβ43, Aβ42, and Aβ40 species in PBS soluble extracts of frontal cortex from AD patients and controls ± standard error between 2 technical replicates. iPSC Donor = brain sample matched to iPSC line.

Aβ43 Immunohistochemistry Differs Between fAD and sAD

Given the observation that Aβ43 species are greatly increased in PSEN1 L435F brain tissue, along with previous work demonstrating that increased levels of Aβ43 occur downstream of multiple PSEN1 and APP mutations (5, 12), we next set out to determine whether Aβ43 immunohistochemistry could be used to distinguish between fAD and sAD brain tissues obtained at autopsy (Fig. 9). Sections of frontal cortex were stained for Aβ43 on a panel of 10 patients with advanced sAD, 9 patients with fAD, 3 Down syndrome patients, 1 patient with familial cerebral amyloid angiopathy ([CAA], Iowa mutation), and one control (Table 3). Aβ43 staining intensity in Aβ plaques was scored from 1 to 3, blinded to diagnosis. For each section, a total Aβ stain was used to verify the presence of plaques and to evaluate for CAA.

FIGURE 9.

Aβ43 immunohistochemistry in fAD, sAD, Down syndrome, and familial CAA (fCCA) brain tissue. Representative Aβ43 immunohistochemistry shows stronger staining in PSEN1 L435F mutants and a subset of other PSEN1 mutants compared with fAD with APP mutation, sAD patients, Down syndrome, fCAA, and control.

TABLE 3.

Aβ43 Immunohistochemistry Results, Frontal Cortex

	ADRC#	Age	Sex	ApoE Genotype	Diagnosis	Aβ 43 Plaques	Aβ 43 NFTs	Aβ 43 CAA	Braak Stage	Thal Phase	CERAD Density
Sporadic AD	sAD1	72	Female	3/3	sAD	+	+	n.a.	VI	5	3
	sAD2	78	Female	3/4	sAD	+	0	0	V	4	3
	sAD3	88	Female	4/4	sAD	+	+	+	VI	5	3
	sAD4	>90	Male	3/4	sAD	++	+	n.a.	V	4	2
	sAD5E	73	Female	3/4	sAD	+	0	n.a.	VI	5	3
	sAD6	70	Female	3/3	sAD	0	+	0	VI	4	3
	sAD7	76	Male	3/4	sAD	0	+	n.a.	V	4	3
	sAD8E	>90	Male	3/4	sAD	0	0	0	VI	3	2
	sAD9	85	Female	3/4	sAD	0	0	0	VI	3	2
	sAD10E	85	Female	3/4	sAD	0	0	n.a.	V	3	2
CAA	fCAA1	45	Male	3/4	fCAA-Iowa	0	0	+	–	–	0
Familial AD	fAD1iPSC, E	53	Female	3/3	fAD-PSEN1 L435F	+++	0	++	VI	3	3
	fAD2	55	Male	3/3	fAD-PSEN1 L435F	+++	0	++	V	–	–
	fAD3	59	Female	–	fAD-PSEN1 L435F	+++	0	++	IV	–	–
	fAD4	58	Male	2/3	fAD-PSEN1	+++	+	+	VI	5	3
	fAD5 E	59	Female	3/3	fAD-PSEN1 H163R	+	+	++	VI	5	3
	fAD6	51	Male	3/3	fAD-PSEN1	0	+	n.a.	VI	4	3
	fAD7	39	Male	3/3	fAD-PSEN1	0	+	0	VI	4	3
	fAD8	39	Male	3/3	fAD-PSEN1	+	+	+	VI	3	3
	fAD9	52	Male	3/4	fAD-APP V717I	+	+	0	VI	5	3
Down syndrome	DS1	70	Female	3/4	DS	0	+	+	VI	5	3
	DS2	57	Male	3/3	DS	0	+	0	VI	5	3
	DS3	59	Male	3/3	DS	0	+	0	–	–	–
Control	C1iPSC, E	>90	Female	3/3	Control	0	0	0	II	0	0

Five-µM frontal cortex sections stained for Aβ43. Aβ43 features scored as none (0), mild (+), moderate (++), or strong (+++). CERAD density refers to neuritic plaques. iPSC, matched to iPSC line; E, sample also present in ELISA (Fig. 8); –, data unavailable.

Although some Aβ43-positive plaques were seen in 7/9 fAD cases and in 5/10 sAD cases, strong positive staining of Aβ43 in plaques was seen only in fAD cases (4/9), including all 3 cases with the PSEN1 L435F mutation (Fig. 9). Only weak to moderate staining was seen in sAD cases (4/10 sAD with weak staining, 1/10 sAD with moderate staining). Aβ43 plaque staining was not observed in Down syndrome cases, fCAA, or the control case. Weak Aβ43 staining of CAA was present in the fCAA case, in 1/6 sAD cases, and in 6/8 cases of fAD with CAA on total Aβ stain in the examined area.

Surprisingly, many cases displayed strong Aβ43 staining on a subset of tau aggregates when stained with a rabbit polyclonal Aβ43 antibody. Aβ43 labeling of NFTs, dystrophic neurites, and ghost tangles was seen in 14 of 24 cases tested, across all diagnostic categories (Table 2; Supplementary DataFig. S2). Aβ43 labeling of NFTs was present in cases both with and without concurrent Aβ43 labeling of plaques. In some instances, fibrillar tau species negative for Aβ43 staining were visible within plaques immediately adjacent to NFTs with Aβ43 labeling (Supplementary DataFig. S2). This finding demonstrates that the observed staining pattern is not likely due to general cross-reactivity with aggregated tau species. However, 3 alternative Aβ43-specific monoclonal antibodies did stain Aβ plaques but did not stain NFTs, making the significance of this finding uncertain. Previous work with a different polyclonal anti-Aβ antibody also showed ghost tangles (21), although this is infrequent with anti-Aβ monoclonal reagents, supporting the possibility that some polyclonal antibodies against Aβ might be selectively deposited on extracellular β-pleated sheet structures, such as ghost tangles, or that some Aβ species may be codeposited on ghost tangles.

DISCUSSION

Overall, this work first demonstrates directly that biochemical alterations measured in iPS-neurons can closely reflect equivalent changes measured in brain. This suggests that cell autonomous features, such as those which impact neuronal protein processing, can be meaningfully modeled by such model systems. In the context of AD, these results clarify the effects of PSEN1 L435F using the first patient-derived iPSC model with this genotype and demonstrate the influence of this mutation on APP processing. We utilize a human model system with endogenous genetics in order to complement existing mouse and human cell culture studies requiring experimental genetic manipulation. We find that iPSC-neurons carrying the PSEN1 L435F mutation produce elevated levels of Aβ43 with preserved levels of Aβ42 and Aβ40 compared with iPSC-neurons derived from an autopsy-confirmed control patient. In addition, we find that PSEN1 L435F mutant human iPSC-neurons exhibit elevated Aβ43/40 and Aβ42/40 ratios, consistent with other iPSC-neuron models of AD (10, 11). These findings support the notion that heterozygous PSEN1 L435F mutation leads to altered PSEN1 function which promotes AD pathogenesis and does not simply lead to a reduction in overall PSEN1 activity. Our results are in contrast to previous work showing decreased secreted Aβ43 in some PSEN1 L435F model systems and support studies demonstrating increased Aβ43 downstream of this mutation (3–7).

Furthermore, PSEN1 L435F mutant neurons exhibit increased pTau compared with control iPSC-neurons. This finding is consistent with the hypothesis that increased pTau in AD is downstream of increasing Aβ species of longer length (i.e. Aβ42 and Aβ43) (22). DAPT treatment reduced pTau levels in iPSC-neurons, confirming a link between gamma secretase function and potentially pathogenic phosphorylation of the tau protein. However, increased levels of neurotoxicity were not observed in cultures with higher levels of pTau. The 3R/4R tau ratio is not altered following DAPT treatment of iPSC-neurons, indicating that accelerated neuronal maturation is unlikely to play a role in decreasing pTau levels downstream of gamma secretase inhibition. In fact, Western blots failed to detect 4R tau isoforms at D28. These results are consistent with prior studies of human iPSC-derived cortical neurons of similar age (20). In summary, heterozygous PSEN1 L435F mutation seen in fAD does not appear to cause an overall reduction of Aβ production, but an increase specifically in pathogenic longer Aβ species and subsequent increase in pTau, thus supporting the amyloid hypothesis of AD.

The NGN2-based neuronal differentiation protocol presented here represents a stepwise improvement over prior techniques (17). We utilize a ubiquitously expressed selection cassette to limit silencing of virally inserted genes, allowing for consistent production of high-purity neuronal cultures for >10 passages after a single viral transduction. This approach reduces the overall amount of lentivirus required and simplifies the cell culture workflow. We find this technique particularly useful when working with multiple iPSC lines in parallel.

The derivation of iPSCs from patients with autopsy-confirmed diagnoses offers an advantage over approaches using clinical syndrome to identify potential iPSC donors. This is critically important for creation of “Control” iPSC lines due to the near-universal presence of AD neuropathologic changes at autopsy in the aged population (23). By deriving iPSCs from a >90-year-old patient with minimal AD changes at autopsy, this study removes asymptomatic and early sAD as a potential confounding variable in control iPSCs in contrast to other work using younger clinical or biomarker-negative controls who may later go on to develop AD (24). A potential reason that sAD iPSC lines have produced variable results to date is that autopsy confirmation of AD pathology has not been available to eliminate the estimated 50% of cases with a clinical diagnosis of probable AD, but final pathologic diagnosis of another age-related neurodegenerative disease, either alone or in conjunction with AD (25). Stated another way, only ∼30% of clinically probable sAD cases actually have sAD as the only neurodegenerative disease at autopsy and these cases may contaminate subsequent iPSC models (25). Furthermore, full neuropathologic evaluation prior to iPSC generation makes it possible to control for concurrent neuropathologic changes, such as cerebrovascular disease, Lewy body disease, and TDP-43 proteinopathies—all of which are known to be present in moderate frequencies even in confirmed sAD cases (25).

Additionally, utilizing an autopsy-confirmed cohort of neurodegenerative disease patients for iPSC models provides the ability to validate in vitro findings directly in matched tissue. Indeed, we found concordant increases in Aβ43 and Aβ42 in PSEN1 L435F iPSC-neurons and matched brain tissue. Brain lysate from the PSEN1 L435F iPSC donor displayed higher levels of Aβ43 than any of the other cases evaluated. Relatively high levels of soluble Aβ40 were observed in case sAD8, but not in the other 2 sAD cases subjected to Aβ ELISA. The predominance of Aβ40 in case sAD8 likely reflects the presence of CAA, which is composed primarily of Aβ40 species (26). On IHC, sAD8 demonstrated moderate CAA in frontal cortex, whereas cases sAD5 and sAD10 had none. Both fAD cases with high levels of Aβ40 species also demonstrated CAA in the frontal cortex.

The observation of increased Aβ43 species in PSEN1 L435F mutant brain tissue led us to evaluate whether or not an Aβ43-specific immunostain could be used to distinguish PSEN1 mutant fAD brain tissues from sAD. We found strong Aβ43 labeling of plaques in a subset fAD cases, including all 3 cases tested with the PSEN1 L435F mutation. Only weak to moderate plaque staining was observed in sAD cases, consistent with this hypothesis. However, Aβ43 staining was somewhat variable between cases, with some PSEN1 fAD cases showing only weak plaque labeling. Therefore, strong Aβ43 staining may be considered suggestive of fAD, whereas weak or moderate staining is consistent with a diagnosis of either sAD or fAD. Aβ43 labeling of plaques was not observed in Down syndrome or familial CAA cases.

Unexpectedly, a single Aβ43 antibody strongly labeled a subset of NFTs in fAD, sAD, and Down syndrome cases, both with and without concurrent Aβ43 labeling of plaques. NFTs labeled with Aβ43 were predominantly within living neurons and ghost tangles, with less labeling of dystrophic neurites at the cores of Aβ plaques. Three alternative commercially available Aβ43-specific monoclonal antibodies did not show NFT staining, although they bound to the plaques in much the same pattern. The significance of the finding that the polyclonal reagent detected tangles is uncertain and possibly related to off-target binding of the antibody. However, intracellular Aβ40 and Aβ42 species have been implicated in the pathogenesis of AD (27, 28).

Future studies utilizing the iPSC lines presented here could focus on clarifying the downstream factors that lead to increased pTau in PSEN1 L435F mutant neurons, particularly the role of intracellular Aβ species, β-CTFs, and endosomal function (12). Our group has derived additional iPSC lines from patients with autopsy confirmation of AD diagnosis and matched brain tissue, including fAD with PSEN1 and APP mutations as well as sAD. Future work will include using these iPSC/brain pairs to provide a broader understanding of the relationship between iPSC-neurons and brain tissues from matched patients. Unfortunately, it is not possible to expand this analysis to all the AD cases used in this work because many predate the discovery of iPSC technology and subsequent routine collection of dermal fibroblasts for reprogramming. Finally, although this study focused on a genetic form of AD, expanding the approach to include autopsy-confirmed cases of sAD with and without potentially confounding neuropathologic processes could help to address some of the conflicting results found using iPSC models derived from clinically diagnosed individuals.

D.H.O. is a recipient of an Alzheimer’s Association Clinician Scientist Fellowship (2018-AASCF-592307) and a Jack Satter Foundation Award. DHO is partially supported by the Dr. and Mrs. E. P. Richardson, Jr Fund for Neuropathology at MGH. NIH support was provided by the National Institute on Aging (Grant R01AG058002). Additional funding was provided by the the Cure Alzheimer's Fund (CIRCUITS Consortium). The Massachusetts Alzheimer’s Disease Research Center is supported by the National Institute on Aging (Grant P30AG062421).

The authors have no duality or conflicts of interest to declare.