Ava i l ab l e on l i ne a t www.sc i enced i r ec t . com

www.e l sev i e r . com/ loca te / sc r

Stem Cell Research (2013) 10, 213–227

Prevention of β-amyloid induced toxicity inhuman iPS cell-derived neurons by inhibitionof Cyclin-dependent kinases and associatedcell cycle events

Xiaohong Xu a, b, Ying Lei b, Jie Luo b, Jamie Wang b, Shu Zhang b,Xiu-Juan Yang c, Mu Sun c, Emile Nuwaysir d, Guohuang Fan b, Jing Zhao b,Lei Lei a,⁎, Zhong Zhong b,⁎

a Department of Histology and Embryology, Harbin Medical University, Harbin 150081, Chinab Regenerative Medicine DPU, GlaxoSmithKline (China) R&D Co., Ltd., Shanghai 201203, Chinac Neurodegeneration DPU, GlaxoSmithKline (China) R&D Co., Ltd., Shanghai 201203, Chinad Cellular Dynamics International, Inc. 525 Science Drive, Madison, WI 53711, USA

Received 6 August 2012; received in revised form 22 November 2012; accepted 22 November 2012Available online 7 December 2012

Abstract Alzheimer's disease (AD) is a neurodegenerative disorder that causes progressive memory and cognitive decline dueto the selective neuronal loss in the cortex and hippocampus of the brains. Generation of human induced pluoripotent stem(hiPS) cells holds great promise for disease modeling and drug discovery in AD. In this study, we used neurons with forebrainmarker expression from two unrelated hiPS cell lines. As both populations of neurons were vulnerable to β-amyloid 1–42(Aβ1–42) aggregates, a hallmark of AD pathology, we used them to investigate cellular mediators of Aβ1–42 toxicity. Weobserved in neurons differentiated from both hiPS cell lines that Aβ induced toxicity correlated with cell cycle re-entry andwas inhibited by pharmacological inhibitors or shRNAs against Cyclin-dependent kinase 2 (Cdk2). As one of the hiPS cell lineshas been developed commercially to supply large quantities of differentiated neurons (iCell® Neurons), we screened achemical library containing several hundred compounds and discovered several small molecules as effective blockers againstAβ1–42 toxicity, including a Cdk2 inhibitor. To our knowledge, this is the first demonstration of an Aβ toxicity screen usinghiPS cell-derived neurons. This study provided an excellent example of how hiPS cells can be used for disease modeling andhigh-throughput compound screening for neurodegenerative diseases.

© 2012 Elsevier B.V. All rights reserved.

Abbreviations: iPS cells, induced pluripotent stem cells; hiPS-C4, hiPS cell line UC C0406 iPS-C4; ES cells, embryonic stem cells; NPCs,

neural progenitor cells; CCEs, cell cycle events; Aβ, β-amyloid; NFTs, neurofibrillary tangles; APP, amyloid precursor protein; AD, Alzheimer'sdisease; PD, Parkinson disease; Cdk, Cyclin-dependent kinase; shRNA, small hairpin RNA; Rb, retinoblastoma protein; TUNEL, terminaldeoxynucleotidyl transferase-mediated dUTP-biotin nick end labeling; FACS, fluorescence-activated cell sorting; CBPs, calcium-bindingproteins.⁎ Correspondence to: L. Lei, No. 194 Xuefu Road, Nangang District, Harbin 150081, China. Fax: +86 451 87503326 or Z. Zhong, Building 3, 898

Halei Road, Zhangjiang Hi-tech Park, Pudong, Shanghai 201203, China. Fax: +86 21 61590844.E-mail addresses: leil086@yahoo.com.cn (L. Lei), zhong.z.zhong@gsk.com (Z. Zhong).

1873-5061/$ - see front matter © 2012 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.scr.2012.11.005

214 X. Xu et al.

Introduction

Induced pluripotent stem (iPS) cells are a type of pluripotentstem cells by forced expression of reprogramming transcrip-tion factors (Takahashi and Yamanaka, 2006; Yu et al., 2007).Similar to human embryonic stem (ES) cells, human iPS (hiPS)cells can be differentiated into derivatives of the threeprimary germ layers, such as cardiomyocytes of mesoderm(Wei et al., 2012), pancreatic cells of endoderm (Zhang et al.,2009) and neurons of ectoderm (Bardy et al., 2012; Petit etal., 2011). This breakthrough holds great promise for itsapplications in cell transplantation, human disease modelingand drug discovery (Laustriat et al., 2010). Despite manyattempts to use hiPS cells to model human neurodegenerativediseases for the last five years since they have been available(Brennand et al., 2010; Dimos et al., 2008; Ebert et al., 2009;Israel et al., 2012; Lee et al., 2009; Shi et al., 2012; Songet al., 2011), there are still many challenges. For someneurodegenerative diseases with long latency periods such asAlzheimer's disease (AD) and Parkinson disease (PD), it isunclear whether the late-onset disease phenotypes could berecapitulated in iPS cellular models.

AD is one of themost common neurodegenerative disordersin aged individuals. It is characterized by neuropathologicalhallmarks including amyloid plaques, neurofibrillary tangles(NFTs) and eventually neuronal loss in the cerebral cortex andhippocampus. β-amyloid (Aβ) is a peptide of 36–43 aminoacids which is processed by successive action of the β and γsecretases from the amyloid precursor protein (APP) and it ismost commonly known to be associated with AD (Vassar,2005). Since β-amyloid 1–42 (Aβ1–42) is one of the maincomponents of the amyloid plaques, it is thought that theaccumulation of Aβ1–42 in the brain correlates with ADseverity significantly (Cummings and Cotman, 1995). However,the recent failure of clinical studies with γ-secretaseinhibitor (semagacestat) (Schor, 2011) has raised questionsabout whether the clearance of Aβ plaques would treat ADeffectively. Therefore, in our study we focused on theinvestigation of key intracellular mediators of Aβ toxicitythrough screens of chemical inhibitors against Aβ1–42 toxiceffects on hiPS cell-derived neurons.

Among the many cellular pathways contributing to celldeath, cell cycle dysregulation is thought to be particu-larly relevant to neurons. Attempts to re-enter cell cycleinduced by insults would conflict with their terminallydifferentiated, non-dividing state (Copani et al., 2001;Greene et al., 2007; Yang et al., 2003). In proliferatingcells, cell cycle is precisely regulated by the interactionsbetween Cyclin-dependent kinases (Cdks) and their obli-gate Cyclin partners. The Cdks are activated via bindingwith their corresponding Cyclins, and then these activatedcomplexes phosphorylate downstream substrates to initi-ate a series of cellular events such as DNA replication,chromosome separation and cell division. Tumor suppres-sor retinoblastoma protein (Rb) is one substrate of theCdk4/6 and Cdk2. Unphosphorylated Rb plays a role insequestering transcription factors required for S phaseentry. Once the Rb is phosphorylated by the Cdks, thephosphorylated Rb (phos-Rb) reduces its affinity for thetranscription factor E2F1. E2F1 is then released from theinhibitory complex E2F-DP and directs the transcription ofthe S phase specific genes (Nevins, 1992). In post-mitotic

cells such as neurons, the cells are held in G0 phase andthey withdraw from the cell cycle by lack of active Cdk–Cyclin complexes for cell cycle progression.

In some neurodegenerative diseases such as AD, PD andamyotrophic lateral sclerosis, ectopic cell cycle events (CCEs)have been observed in post-mitotic neurons (Bonda et al.,2009; Lim and Qi, 2003; Ranganathan and Bowser, 2003; Smithet al., 2004). In the 1990s, several groups reported that cellcycle markers were expressed in the brains of AD patients(Baumann et al., 1993; Vincent et al., 1996). Moreover, viafluorescence in situ hybridization (FISH) technique, Yang andhis colleagues demonstrated the DNA synthesis process in theneurons of the AD patients (Yang et al., 2001). In addition tothe above findings in AD patients, several in vitro studiesdemonstrated that blockade of the G1–S transition using aCyclin D1 antisense or a dominant-negative mutant of Cdk4/6could prevent the Aβ induced apoptosis (Copani et al., 2001;Park et al., 1997). Thus, a substantial body of evidence hasassociated the aberrant cell cycle re-entry in differentiatedneurons with neuronal apoptosis in AD.

In this study, we modeled neuronal loss in human ADbrains by exposing human forebrain neurons derived from iPScells to the insult of Aβ1–42 aggregates. Those neurons weresensitive to Aβ toxicity and showed occurrence of CCEsafter Aβ1–42 treatment. Moreover, the aberrant cell cyclere-entry and neuronal apoptosis were prevented by Cdkinhibitors. Importantly, this finding was confirmed in acommercially available forebrain neuronal population(iCell® Neurons) derived from a different hiPS cell line.The availability of large batches of iCell Neurons with consistentquality enabled a small-molecule screen for compoundsthat blocked the toxic effects of Aβ1–42 and demonstratedthe successful application of hiPS cells in drug screens forneurodegenerative diseases.

Materials and methods

Chemicals and peptides

Roscovitine and Cdk4 inhibitor II were purchased fromCalbiochem (San Diego. CA, USA). 5-Bromo-2′-deoxyuridine(BrdU), Olomoucine and GW8510 were from Sigma Aldrich (St.Louis, MO, USA). Cdk2 inhibitor II was from Santa Cruz (SantaCruz, CA, USA). PD0332991 was from Axon Medchem (9713 GZGroningen, Netherlands). The recombinant Aβ1–42 peptideswere purchased from rPeptide (Athens, Georgia, USA). Thepreparation of Aβ1–42 oligomers followed the previousestablished protocol (Dahlgren et al., 2002) and western blotanalysis was performed to evaluate the formation of Aβ1–42aggregates with Aβ antibody (6E10, Covance, Princeton, NJ,USA) (Supplemental Fig. 1).

hiPS cell culture, forebrain neuron differentiationand iCell Neurons culture

hiPS cell line UC C0406 iPS-C4 (hiPS-C4) was generated fromthe urine cells of a healthy woman and maintained aspreviously described (Zhou et al., 2011). Neural differenti-ation of forebrain neurons from hiPS-C4 cells was performedusing a previously established protocol with some modifica-tions (Chambers et al., 2009). Briefly, hiPS-C4 cells were

215Cdk inhibition prevents Aβ induced toxicity in hiPS cell-derived neurons.

disaggregated using collagenase/dispase (2 mg/mL), thenwashed by hiPS medium (80% DMEM/F12, 20% knockoutserum replacement, 1% non-essential amino acids, 2 mML-glutamine and 0.1 mM β-mercaptoethanol), and pre-plated on gelatin for 1 h (hour) at 37 °C to remove MEFs.The non-adherent hiPS cell clumps were cultured on petridishes in N2B27 medium (DMEM-F12/Neural Basal medium1:1 with 1% N2, 2% B27, 1% non-essential amino acids, and2 mM L-glutamine) supplemented with 10 μM Rock inhibitorY-27632 (Sigma Aldrich) for 8 h. Then the unattachedaggregates were collected and plated on dishes pre-coatedwith 10 μg/mL poly-L-ornithine (Sigma Aldrich) and 10 μg/mLlaminin in N2B27 medium supplemented with 500 ng/mLNoggin (PeproTech, Rocky Hill, NJ, USA) and 10 μM SB-431542(Sigma Aldrich). After 12–14 days of culture, cells werepassaged with a cell scraper at a split ratio of 1:1. Those cells(neural progenitor cells, NPCs) were cultured in N2B27mediumsupplemented with bFGF (20 ng/mL), EGF (20 ng/mL) andBDNF (PeproTech, 20 ng/mL), and passaged once every4–5 days. For neural differentiation, NPCs were cultured inN2B27 medium supplemented with BDNF (PeproTech,20 ng/mL), GDNF (PeproTech, 20 ng/mL) and NT3 (PeproTech,10 ng/mL) in 6 cm dishes at a density of 200,000 cells/cm2.After 7 days of culture, cells were digested with accutase andfiltered with 40 μm cell strainer (BD Bioscience, San Jose,California, USA) to collect single cells. In the final step ofdifferentiation, those differentiated cells were seeded onplates pre-coated with 10 μg/mL poly-L-ornithine (SigmaAldrich) and 10 μg/mL laminin at a density of 50,000–100,000 cells/cm2 for another 2–3 weeks of culture in N2B27medium supplemented with BDNF (PeproTech, 20 ng/mL),GDNF (PeproTech, 20 ng/mL) and NT3 (PeproTech, 10 ng/mL),cAMP (N6,2′-O-Dibutyryladenosine 3′,5′-cyclic monophos-phate, Sigma Aldrich, 10 μM), and 0.2 mM ascorbic acid(STEMCELL Technologies, Vancouver, BC, Canada,). Differen-tiated neurons were cultured in fresh N2B27 medium withoutgrowth factors for 1–2 day(s) before Aβ toxicity assay. Allgrowth factors and reagents were from Life Technologies(Carlsbad, CA, USA) unless noted otherwise.

iCell Neurons were purchased from Cellular DynamicsInternational (Madison, WI, USA) and cultured according tothe manufacturer's instructions. Cells were used for Aβtoxicity assay after 5–7 days post-thaw.

The human biological samples were sourced ethically andall experiments involved in application of human sampleswere approved by GlaxoSmithKline Human Biological SampleUser Committee.

Cell viability assay

Cell viability was measured by CellTiter-Glo LuminescentCell Viability Assay (Promega, Madison, WI, USA), accordingto the manufacturer's instructions, and luminescence wasmeasured by the EnVision microplate reader (Perkin Elmer,Waltham, MA, USA) or Glomax 96 microplate luminometer(Promega).

Immunofluoresence and quantification

Cells were fixed with 4% paraformaldehyde (PFA) for 15 minat room temperature. They were later permeabilized in 0.3%

Triton-X100 for 20 min at room temperature, blocked with3% normal donkey serum containing 0.1% Triton-X100 inphosphate buffered saline (PBS) for 1 h at room tempera-ture and stained with the primary antibodies at 4 °Covernight. The following primary antibodies were used:anti-vGAT, anti-vGLUT2 (Synaptic Systems, Goettingen,Germany); anti-GABA, anti-MAP2 (Sigma Aldrich); anti-SOX2, anti-Cdk2, anti-Cdk4 (Santa Cruz); anti-phos-Rb(Ser795), anti-Cyclin D1 (Cell Signaling Technology, Essex,Massachusetts, USA); anti-Nestin, anti-Tubulin, beta IIIisoform, C-terminus, clone TU-20 (TuJ1, Millipore, Danvers,MA, USA); anti-FOXG1 (Abcam, Cambridge, UK); anti-GFAP(Dako, Carpinteria, CA, USA). For BrdU staining, cells wereincubated with 2 M HCl at room temperature for 30 min todenature DNA, and incubated with primary antibodies at4 °C overnight. After washing 3 times with PBS containing0.1% Triton-X100, the cells were incubated with corre-sponding Alexa Fluor 488 and Alexa Fluor 546 secondaryantibodies (all from Life Technologies) for 1 h at roomtemperature, washed 3 times with PBS containing 0.1%Triton-X100, incubated with 1 μg/mL 4′, 6-diamidino-2-phenylindole (DAPI, Sigma Aldrich) for 10 min.

For TUNEL (terminal deoxynucleotidyl transferase-mediateddUTP-biotin nick end labeling) staining, In Situ Cell DeathDetection Kit, Fluorescein (Roche, Basel, Switzerland) wasused. Briefly, cells were fixed with 4% PFA at room tem-perature for 15 min and washed 3 times with PBS beforebeing stained with TUNEL reagent for 1 h in the absence oflight. For double staining of BrdU and TUNEL, cells wereincubated with 2 M HCl at room temperature for 30 min todenature DNA, washed with PBS, incubated with anti-BrdU(Santa Cruz) antibodies at 4 °C overnight, washed with PBSand incubated with Alexa Fluor 594 goat anti-rat IgG (LifeTechnologies) and TUNEL reagent for 1 h at room temper-ature, and then washed cells 3 times with PBS.

For quantification, images were acquired from 6 to8 non-overlapping fields on each coverslips using an LSM710confocal microscope (Carl Zeiss, Oberkochen, Germany).Cell numbers were counted in a double-blinded fashion.Data used for analysis were collected from 3 independentexperiments with 2 to 3 replicates (corresponding to 2 to 3coverslips) for every group in each experiment.

Immunoblotting

The cells were lysed with RIPA buffer (Sigma Aldrich)containing Halt™ Protease and Phosphatase Inhibitor Cock-tail (Thermo Fisher, Philadelphia, PA, USA) for 20 min onice. The samples were incubated at 70 °C for 15 min inNupage sample buffer (Life Technologies) and proteins wereseparated on Nupage 4%–12% Bris-Tris gels followed bytransfer to nitrocellulose membrane (Life Technologies) forimmunoblotting. The following primary antibodies wereused for detection: anti-Cdk2, anti-Cdk4 (Santa Cruz);anti-Cyclin D1, anti-E2F1, anti-P21, anti-phos-Rb (Ser795)(Cell Signaling Technology); anti-Cyclin A (Abcam); anti-Aβ(6E10, Convance); anti-β-Actin-peroxidase (Sigma Aldrich).All the secondary antibodies used for visualization wereeither goat anti-mouse or goat anti-rabbit purchased fromSigma Aldrich. Blots were developed with the SuperSignalWest Pico Chemiluminescent Substrate or SuperSignal West

216 X. Xu et al.

Femto Maximum Sensitivity Substrate Kit (Thermo Fisher)and visualized by the ImageQuant™ LAS 4000 biomolecularimager (GE Healthcare Life Sciences, Pittsburgh, PA, USA).Densitometry measurement of protein bands intensity wascarried out using ImageQuantTL software (GE HealthcareLife Sciences).

RNA isolation, cDNA synthesis and real-time PCR

Cells were lysed using RLT Buffer and RNA was purified usingthe RNeasy Mini Kit and the RNase-Free DNase Set (Qiagen,Valencia, CA). For all samples, cDNA was generated usingPrimeScript® RT reagent Kit (Takara, Shiga, Japan). Quan-titative real-time PCR was run on the LightCycler® 480 II(Roche, Basel, Switzerland) using primers listed in Supple-mental Table 1.

Electrophysiology

Patch clamp recordings were performed on hiPS-C4 cell-derived neurons on Day 21. Neurons were perfused contin-uously with buffer containing the following (in mM, all fromSigma Aldrich): 140 NaCl, 5 KCl, 2 CaCl2, 1 MgCl2, 10 HEPES,10 glucose (added before use) with pH 7.4. The thin-walledborosilicate capillary glass (WPI, TW-150F-4) was pulled toa resistance of 3–6 Mω using a horizontal pipette puller(P-97, Sutter) and filled with intracellular solutioncontaining the following: (in mM, all from Sigma Aldrich):140 KCl, 1 MgCl2, 2 CaCl2, 5 EDTA, 10 HEPES with pH 7.3.Action potentials were recorded in the whole-cell currentclamp configuration. A current was injected to hold themembrane potential at around −60 mV, and 10 ms of 200pA depolarized current was injected until action potentialswere observed. Recordings were performed using an AxonCNS Multiclamp 700B patch clamp amplifier and Digidata1440A (Molecular Devices). Data were digitized at 10 kHzand analyzed with Clampex 10.2, clampfit 10.2 Software(Molecular Devices).

Neurite outgrowth assay

After treatment with Aβ1–42 aggregates and compounds,the neurons were fixed with 4% PFA and then stained withanti-TuJ1 antibody (Millipore) which specifically labels betaIII-tubulin in axons and dendrites. The Alexa Fluor 488donkey anti-mouse IgG secondary antibodies (Life Technol-ogies) and DAPI (Sigma Aldrich) were used before finalwashing. The cells were scanned with Cellomics ArrayScanVTI HCS Reader (Thermo Fisher) and analyzed with itsNeuronal profiling V4 algorithm. The average length perneurite was used for assay readout.

shRNA construction, lentivirus generation andinfection

Human Cdk2 and Cdk4 small hairpins RNA (shRNA) wereprepared in pLVX-shRNA2 vector (Clontech, Mountain View,CA, USA), based on the sequence of 5′-GGCCAGGAGTTACTTCTATGC-3′ and 5′-GCTGGAAATGCTGACCTTTAA-3′, re-spectively. Lentivirus production and titer test were carried

out following the previously published protocol (Crittendenet al., 2007). iCell Neurons were thawed and cultured for3 days when virus at the multiplicity of infection (MOI) of 5was added and incubated for 8 h before changing to freshmedium. Cells were treated with 5 μM Aβ1–42 for 48 h after5 days of infection. Then the number of GFP (+) cells andtotal cells were counted with Acumen Explorer X3 (TTPLabTech, Royston, UK) within 2 h post-fixation with 4% PFA.All the virus related experiments were approved by theGenetically Modified Organism (GMO) Management Commit-tee of GlaxoSmithKline R&D China.

Flow cytometry analysis

iCell Neurons in culture were treated with TrypLE™ (LifeTechnologies) to form a single cell suspension. Cells werestained with LIVE/DEAD® Fixable Red Dead Cell Stain (LifeTechnologies) prior to fixation by formaldehyde and perme-abilization in fluorescence-activated cell sorting (FACS)buffer containing 0.1% saponin. Cells were then stainedwith a combination of mouse anti-human TuJ1-Alexa Fluor488 and mouse anti-human Nestin Alexa Fluor 647 antibodiesor with the appropriately conjugated isotype controls (allantibodies from BD Biosciences). Stained cells were analyzedon an Accuri C6 flow cytometer. Flow cytometry dataanalysis was performed using FCS Express (De Novo Soft-ware) quantifying the percent TuJ1 (+)/Nestin (−) eventswithin the live cell population determined from the LIVE/DEAD staining.

Aβ toxicity assay and compound screens

The iCell Neurons were thawed and cultured for 5 days at adensity of 8000 cells/well in 384-well plates beforecompound addition and Aβ1–42 incubation. In the primaryscreen, 50 nL of each compound in 10 mM stock from a GSKproprietary tool compound set was transferred to 50 μLmedia by the Echo Liquid Handling System (Labcyte,Sunnyvale, CA, USA). After 2 h of incubation with com-pounds containing media, 5 μM of Aβ1–42 or 0.2% DMSOdissolved in fresh culture medium was added to the cells for48 h treatment. Cell viability was measured by CellTiter-GloLuminescent Cell Viability Assay (Promega). The hits wereselected if the compounds increased the cell viability by atleast 3 times the standard deviation of the CellTiter-GloLuminescent measure over the mean cell viability of the cellgroup treated by Aβ1–42 only.

Statistics

Dose–response curves were generated from XLfit (IDBS,Burlington, MA, USA) using the dose–response onsite formulaas follows: fit=(A+(B/(1+((x/C)^D)))); inv=((((B/(y–A))–1)^(1/D))*C); res=(y-fit). Other results were analyzed using theGraphPad Prism 5 software (GraphPad Software, San Diego, CA,USA) and were presented in the form of mean±SD. Unpairedstudent's t-test or two-way ANOVA with Bonferroni post-testswere used to evaluate the significance of differences betweenmeans. In all cases, significance was noted at *Pb0.05,**Pb0.01, ***Pb0.001.

Figure 1 Characterization of forebrain neurons derived from hiPS-C4 cells. A) Experimental scheme of neural differentiation fromhiPS-C4 cells. B, BDNF; G, GDNF; N, NT3; AA, ascorbic acid. Cells on Day 21 were ready for Aβ toxicity assay. B) Cells were labeledwith antibodies for Nestin (red), FOXG1 (red) and SOX2 (green) on Day 0. Scale bar, 20 μm. C) Cells were labeled with antibodies forGFAP (red), TuJ1 (green), SOX2 (red) and MAP2 (green) on Day 21. DAPI (blue) was used for nuclei staining. Scale bar, 20 μm.D) Analysis of cell percentages during different time points of differentiation. E) Real-time PCR results for cells on Day 21.F) Spontaneous action potentials. G) Evoked action potentials. The values used for each column were mean±SD from 3 biologicalreplicates.

217Cdk inhibition prevents Aβ induced toxicity in hiPS cell-derived neurons.

Figure 2 Aβ1–42 induced neuronal apoptosis and aberrant cell cycle re-entry in hiPS-C4 cell-derived neurons. A) Cell viability wasmeasured at 24 h and 48 h. B) BrdU, red; TuJ1, green. C) The percentages of BrdU (+)/TuJ1 (+) cells were used for analysis. D) BrdU,red; TUNEL, green. E) The percentages of BrdU (+)/TUNEL (−) cells and BrdU (+)/TUNEL (+) cells were used for analysis. F) Cyclin D1,green; TuJ1, red. G) Cdk4, green; TuJ1, red. H) Cdk2, green; TuJ1, red. I) Phos-Rb, green; TuJ1, red. DAPI (blue) was used for nucleistaining and scale bar=20 μm for all images. The values used for each column were mean±SD from 3 biological replicates and***Pb0.001 was determined by unpaired t-test.

218 X. Xu et al.

219Cdk inhibition prevents Aβ induced toxicity in hiPS cell-derived neurons.

Results

Generation of forebrain neurons from hiPS-C4 cells

In neurodegenerative diseases, it is common that theselective neuronal loss occurs in or initiates from restrictedregions of the brain. For example, progressive neuronal losshas been observed in the cerebral cortex of AD patients.Recently, human regional specific neurons were successfullydifferentiated from ES and iPS cells (Bissonnette et al.,2011; Hester et al., 2009; Zeng et al., 2010). Neuronal

Figure 3 Cdk inhibitors prevented neuronal apoptosis in hiPS-C4 cecompounds (Comp.) were generated with XLfit curve fitting softwaD) The percentages of TUNEL (+) cells were used for analysis. E) Theused for each column were mean±SD from 3 biological replicates an

differentiation without exogenous morphogens will generatethe neuroepithelia cells that differentiate into corticalprogenitors and subsequently glutamatergic neurons andGABAergic interneurons by default (Liu and Zhang, 2011). Inthis study, we followed the previously established Dual-SMAD inhibition protocol (Chambers et al., 2009) to generateforebrain neurons in the absence of factors that confer theregional specificity. After about two weeks of neuralinduction with Noggin and SB431542 (Fig. 1A), we success-fully obtained Nestin (+)/SOX2 (+) neural progenitor cells(NPCs) from hiPS-C4 cells (Fig. 1B). Those NPCs have

ll-derived neurons. A–B) Full dose–response curves for indicatedre. C) TuJ1, red; TUNEL, green; DAPI, blue. Scale bar, 20 μm.percentages of TuJ1 (+) cells were used for analysis. The valuesd ***Pb0.001 was determined by unpaired t-test.

Figure 4 Prevention of Aβ1–42 induced cellular toxicity via inhibition of Cdks in iCell Neurons. A) iCell Neurons labeled withantibodies for TuJ1 (green) on Day 1; MAP2 (green), GABA (red), vGAT (green) and vGLUT2 (red) on Day 14. Hoechst (blue) was usedfor nuclei staining. Scale bar, 50 μm. B) Post-thaw purity of iCell Neurons as measured on Day 1 by flow cytometry for TuJ1 (neuronalmarker) and Nestin (neural stem/progenitor marker). C) Cell viability was analyzed with XLFit curve fitting software at different dosesand time points of Aβ1–42 treatment. D) Neurite was stained with TuJ1 antibodies (green) after 24 h Aβ1–42 treatment. E) Full dose–response curves for indicated compounds (Comp.) were generated with XLfit curve fitting software. F) For neurite outgrowth assay,cells were stained by anti-TuJ1 antibodies (green) after 36 h Aβ1–42 treatment. GW, GW8510. G) Average length per neurite was usedfor analysis. H–I) Aβ1–42 induced neuronal death in cells after 5 days of infection by lentiviruses of shRNA2, shCdk2 and shCdk4. Thepercentages of survived GFP (+) cells in different groups were obtained by the ratios of GFP (+) cells before and after Aβ1–42treatment. The values used for each column were mean±SD from 3 biological replicates; *Pb0.05, **Pb0.01, ***Pb0.001 weredetermined by unpaired t-test. Scale bar=20 μm for images in D), F) and H).

220 X. Xu et al.

221Cdk inhibition prevents Aβ induced toxicity in hiPS cell-derived neurons.

predominantly a forebrain identity as they highly expressedforebrain marker FOXG1 (Fig. 1B). We monitored theexpression of NPCs, glial cell and differentiated neuronmarkers SOX2, GFAP, TuJ1 and MAP2 from Day 0 to Day 28.TuJ1 and MAP2 expression increased over the course ofdifferentiation and reached the highest level on Day 21,when the majority of NPCs became differentiated and lostSOX2 expression. In contrast, the GFAP (+) glial cells startedto appear on Day 21 and further increased on Day 28(Figs. 1C,D). In order to determine the expression file ofthose differentiated neurons, real-time PCR analysis wascarried out on Day 21. Those cells expressed relative highlevels of markers for glutamatergic neurons (vGLUT1andvGLUT2), GABAergic neurons (GAD65 and GAD67) andcholinergic neurons (ChAT and VAChT), and with a lowpercentage of cells expressing dopaminergic neuron markerTH (Fig. 1E). Moreover, they also expressed high levels offorebrain marker genes (FOXG1, OTX2) and with relativelylow levels of midbrain genes (PAX2) and hindbrain genes(OLIG2 and HOXB6) on Day 21 (Fig. 1E). Electrophysiologyanalysis demonstrated that spontaneous action potentials(Fig. 1F) and evoked action potentials (Fig. 1G) were presentin those neurons on Day 21. We wanted to ask if thosefunctional forebrain neurons derived from hiPS-C4 cellswould be appropriate for modeling human AD in vitro, sincethey include the susceptible neurons in human AD brainssuch as glutamatergic neurons, GABAergic neurons andcholinergic neurons.

Induction of neuronal death and aberrant cell cyclere-entry in hiPS-C4 cell-derived neurons by Aβ1–42aggregates

The toxicity of Aβ1–42 to neurons has been demonstrated inprimary cortical and hippocampal neurons from rodentbrains, and differentiated neurons from neuronal cell lines(Copani et al., 1999; Frasca et al., 2004; Troy et al., 2000;Varvel et al., 2008). However, studies with Aβ1–42 toxicityon hiPS cell-derived neurons are lacking. In this study,different doses of Aβ1–42 were added to fully differentiatedhiPS-C4 cell-derived neurons, and cell viability was mea-sured at 24 h and 48 h after treatment, respectively. Aβ1–42 reduced neuronal survival in a dose- and time-dependentmanner, and with about 25% of neurons dying within 24 hand 60% of neurons dying within 48 h after treatment at5 μM Aβ1–42 (Fig. 2A).

A series of evidence for cell cycle re-entry has been foundin human AD brains, such as altered or inappropriateexpression of Cyclin D1, Cdk2/4, and Rb phospharylation inAD neurons compared to age-matched controls (Busser etal., 1998; Mosch et al., 2007; Ranganathan et al., 2001;Smith et al., 1999). In mammalian cells, complexes of CyclinD1-Cdk4/6 and Cyclin E-Cdk2 regulate the G1–S phasetransition via phosphorylating Rb, which acts as a brake toS phase entry. The activity of Cdks could be positivelyregulated by increased expression of Cyclins and proteinshuttling of Cdks/Cyclins from cytoplasm to nucleus. Toanalyze cell cycle re-entry in hiPS cell-derived neurons,hiPS-C4 cell-derived neurons were treated with 5 μM Aβ1–42 aggregates and expression pattern of G1 phase relatedproteins were analyzed at different time points post-

treatment. We found that Cyclin D1 was mainly detected inthe cytoplasm and gradually accumulated in the nucleararea after 3 h to 8 h exposure to the Aβ1–42 (Fig. 2F).Similar to Cyclin D1, Cdk4 and Cdk2 also showed trans-locations from the cytoplasm to the nucleus after Aβ1–42treatment (Figs. 2G,H), which was consistent with theprevious report in homocysteine treated rat cortical neurons(Ye and Blain, 2010). Thus, in response to Aβ1–42 toxicity,Cyclin D1, Cdk2 and Cdk4 appeared to translocate to thenucleus, where they could form active complexes of CyclinD1–Cdk4 and Cyclin E–Cdk2 and phosphorylate their nuclearsubstrates. Rb phosphorylation was highly detected in thenuclear areas of the treated neurons, while it was barelypresent in the controlled neurons (Fig. 2I). These observa-tions indicated that those differentiated neurons hadre-entered cell cycle and reached the G1–S phase boundaryafter Aβ1–42 treatment.

To further determine whether Aβ1–42 induced cell cyclere-activation in hiPS-C4 cell-derived neurons, BrdU was usedto label newly synthesized DNA during S phase of the cellcycle. We found a significant increase of BrdU (+)/TuJ1 (+)neurons after Aβ1–42 treatment when compared with theuntreated groups (4.5%±3.2% versus 16.6%±4.6%) (Figs. 2B,C). The results suggested that Aβ1–42 triggered the neuronsto re-entered cell cycle, and most of these cells (92.8%±5.4%) traversing S phase were also undergoing apoptosis asindicated by TUNEL signal (Figs. 2D,E).

Prevention of neuronal apoptosis in hiPS-C4cell-derived neurons by Cdk inhibitors

Having demonstrated that the hiPS-C4 cell-derived neuronswith aberrant occurrence of CCEs would trigger neuronalapoptosis, we were interested in testing whether blocking Cdkactivity would protect these neurons. Two validated Cdkinhibitors GW8510 and Cdk2 inhibitor II (Davis et al., 2001;Sielecki et al., 2000) were found to protect hiPS-C4 cell-derived neurons against Aβ1–42 toxicity in a dose-dependentmanner (Figs. 3A,B). TUNEL staining was performed to furtherdemonstrate reduction of programmed cell death induced byAβ1–42 for 24 h from 47.94%±4.26% to 33.41%±4.94% and33.39%±4.02% by 1 μM GW8510 and 1 μM Cdk2 inhibitor II,respectively, close to what was observed in the untreatedgroups (22.95%±4.63%) (Figs. 3C,D). Meanwhile, 1 μMGW8510and 1 μM Cdk2 inhibitor II prevented neuronal loss with62.91%±5.47% and 59.56%±4.98% TuJ1 (+) cells, comparedwith the Aβ1–42 treated groups (48.39%±4.37%) anduntreated control group (71.62%±4.84%) (Figs. 3C,E).

Confirmation of cellular toxicity and cell cyclere-entry induced by Aβ1–42 in iCell Neurons

It is well known that hiPS cells have different differentiationpropensities among clones. There are also concerns that thegenetic variability of iPS cell clones contribute to phenotypicdifferences of differentiated cells. Therefore, it is importantthat we employed an independent hiPS cell clone to confirmour findings in hiPS-C4 cell-derived neurons. iCell Neuronsare commercially available neurons derived from hiPS cells.They were similar to our hiPS-C4 cell-derived neurons in thatthey also represented forebrain identity andwere predominantly

222 X. Xu et al.

223Cdk inhibition prevents Aβ induced toxicity in hiPS cell-derived neurons.

GABAergic neurons and glutamatergic neurons (Fig. 4A,Supplemental Fig. 3). Importantly, iCell Neurons have onedistinct advantage for our study as they were highly pureneuronswith 98.41% TuJ1 (+)/Nestin (−) cells characterized byFACS (Fig. 4B) and few undifferentiated NPCs that mightinterfere with cell imaging and biochemical characterizationof CCEs, were left in the population.

iCell Neurons were found to be similarly susceptible to Aβtoxicity as hiPS-C4 cell-derived neurons. Aβ1–42 reducedneuronal survival in a dose- and time-dependent manner, andabout 50% of iCell Neurons died within 24 h and 64% diedwithin 48 h after treatment by 5 μM Aβ1–42 (Fig. 4C). With apopulation of pure neurons, neurite outgrowth could bereadily imaged and quantified. When treated with Aβ1–42 at5 μM and 10 μM for 24 h, iCell Neurons reduced their averageneurite length to 71.48±1.29% and 38.77±3.65%, respective-ly, normalized to the control group (100±3.8%) (Fig. 4D), afinding consistent with the recent report in rat hippocampalneurons (Nguyen et al., 2012). Moreover, GW8510 rescued theneurite outgrowth reduction in a dose-dependent manner.After 36 h incubation, Aβ1–42 reduced the average neuritelength of iCell Neurons to 25.40%±4.57% compared with thecontrol group (100%±5.54%) (Figs. 4F,G). GW8510 significantlyrescued the neurite length to 56.38±5.24%, 49.56±6.20%,46.13±10.07% and 38.08%±7.36% at concentrations of 1 μM,0.3 μM, 0.1 μM and 0.03 μM, respectively (Figs. 4F,G).

As GW8510 a pan Cdk inhibitor (inhibiting Cdk2/1/4) wasalso effective in blocking Aβ induced toxicity in hiPS-C4cell-derived neurons. We were curious about the role ofCdk4, another key player of the G1–S transition in Aβinduced cell cycle re-entry. Previous attempts in thebiochemical characterization of Cdk2 and Cdk4 in responseto Aβ toxicity in hiPS-C4 cell-derived neurons were compli-cated due to the existence of identical proteins fromproliferating cells in that population. With a population ofpure neurons, we were able to detect expression changes ofcell cycle proteins after Aβ1–42 treatments. In our study,we observed that the Cyclin D1 expression in iCell Neuronsincreased from 3 h after Aβ1–42 treatment and theincreased level could be sustained for 20 h (Figs. 5A,C).However, no significant increase in the Cdk4 expression wasseen (Figs. 5A,D). The phos-Rb expression was significantlyincreased from 3 h to 8 h after Aβ1–42 treatment andgradually decreased to the level of untreated group within20 h (Figs. 5A,B). We found that Cdk2 expression increasedgreatly by 3 h and 8 h treatment and the Cyclin A expressionincreased from 8 h to 20 h after exposure to Aβ1–42(Figs. 5A,E,F). These changes were expected to facilitatethe insulted neurons to transit from G1 phase to S phase,which were consistent with our observation that E2F1expression gradually increased from 3 h to 20 h after Aβ1–42 treatment (Figs. 5A,H). Moreover, we discovered that theCdk inhibitor protein P21 increased by 3 h after treatment,and then decreased significantly (Figs. 5A,G). In addition,

Figure 5 Immunoblotting analysis of CCEs caused by Aβ1–42 in iCepoints (0 h, 3 h, 8 h and 20 h) after 5 μM Aβ1–42 treatment. anti-phanti-P21, anti-E2F1 and anti-Actin antibodies were used for immunobwere shown. The values used for each column were mean±SD from 3group were determined by unpaired t-test. I–K) 1 μM GW8510 bloctreatment. The values used for each column were mean±SD from 3 bDMSO group were determined by two-way ANOVA with Bonferroni p

we observed that 1 μM GW8510 significantly reduced theexpression of phos-Rb (Figs. 5I,J) and E2F1 (Figs. 5I,K), whichwere markers for cells actively progressing through cellcycle G1 phase. This finding was consistent with theobservation reported in rat stroke model (Osuga et al.,2000), and suggested that Cdk inhibitor GW8510 couldprevent the aberrant cell cycle re-entry caused by Aβ1–42in iCell Neurons.

We selected 6 Cdk inhibitors with different specificitiesto Cdk2 and Cdk4, as shown in Table 1 to test theirprotective effects. Interestingly, two Cdk4 specific inhibi-tors Cdk4 inhibitor II and PD0332991 failed to showsignificant protective effects, while all other inhibitorsthat inhibit Cdk2 protected neurons against Aβ1–42 toxicityin a dose-dependent manner (Fig. 4E). This may indicatethat inhibition of Cdk2 rather than inhibition of Cdk4 couldrescue the neuronal death in iCell Neurons caused by Aβ1–42 aggregates. To further confirm that it is inhibition of Cdk2rather than Cdk4 was protective against Aβ1–42 toxicity, weemployed shRNAs against Cdk2 and Cdk4 in this toxic cellularmodel. shCdk2 and shCdk4 could efficiently knockdown theendogenous expression of Cdk2 (Supplemental Fig. 3A) andCdk4 in iCell Neurons (Supplemental Fig. 3B). We observedthat shCdk2 significantly increased the survival of GFP (+)cells to 89.96±5.17% compared with the shRNA2 control group(58.6±3.76%) (Figs. 4H,I). Intriguingly, cells transfected withshCdk4 also increased the survival of GFP (+) cells, althoughthe specificity of shCdk4 was not ideal as some reduction ofCdk2 was also observed. Moreover, it appears that shCdk2protected neuronal deathmore efficiently than shCdk4. Theseresults were consistent with previous reports in Aβ treated ratcortical neurons (Copani et al., 1999, 2001).

Screen for inhibitors of Aβ1–42 toxicity in iCellNeurons

For drug screening, it is critical to have large quantities ofcells with consistent quality. iCell Neurons fulfilled thesecriteria and allowed us to establish a high-throughput assayusing Aβ1–42 as the insult. Several hundred compounds froma GSK proprietary compounds library were screened at 10 μMfor each compound. 19 compounds were selected as hitsbased on the selection criteria outlined in the Materials andmethods (Fig. 6). Among the 19 hits, one of them was a Cdk2inhibitor confirming the reliability and sensitivity of ourscreening platform based on hiPS cell-derived neurons.

Discussion

AD is becoming more prevalent given that there is no cure andour population is aging. A cellular model that could accuratelyrecapitulate AD pathology is critical for investigating thepathogenic mechanisms and searching for treatments. Over

ll Neurons. A) Cell lystates were collected at a few different timeos-Rb (S795), anti-Cdk4, anti-Cyclin D1, anti-Cdk2, anti-Cyclin A,lotting. B–H) Relative protein expression levels versus 0 h groupbiological replicates; *Pb0.05, **Pb0.01, ***Pb0.001 versus 0 hked the increased phos-Rb and E2F1 after 3 h and 8 h Aβ1–42iological replicates and **Pb0.01, ***Pb0.001 versus the value ofost-tests for multiple comparisons.

Table 1 Commercial inhibitors of Cyclin-dependentkinases

Name Supplier Catalogno.

Inhibitingtarget(s)

Roscovitine Calbiochem 557367 Cdk a1, Cdk2, Cdk5Olomoucine Sigma Aldrich O0866 Cdk1, Cdk2, Cdk5GW8510 Sigma Aldrich G7791 Cdk2, Cdk1, Cdk4Cdk2 inhibitor II Santa Cruz sc-221409 Cdk2Cdk4 inhibitor II Calbiochem 219477 Cdk4PD0332991 Axon Medchem 1505 Cdk4, Cdk6a Cdk: Cyclin-dependent kinase.

224 X. Xu et al.

the past decades, scientists have used primary neurons fromrodent brains or immortalized neuronal cell lines to study ADmechanisms and for drug screening. However, there are issuesand limitations for drug discovery studies with these models.For example, there are important genetic and physiologicaldifferences between human and animal brains, and primaryneurons from rodent do not reflect neuronal conditions inhuman beings accurately. On the other hand, the immortal-ized cell lines lose many features of human cells and theirintracellular regulations are significantly different from truehuman neurons. The development of hiPS cell technologyholds great potential for drug discoveries for AD, as hiPS cellswith identical genetic makeup of patients' specific geneticbackgrounds can be readily established (Israel et al., 2012; Shiet al., 2012; Yagi et al., 2011). Importantly, cortical neuronsderived from these hiPS cell lines carry important features ofAD pathology, such as increased generation of Aβ peptide.However, it is disappointing that no neuronal loss from theaccumulation of Aβ peptide is observed and these iPS cell linescould not be used for screening compounds against Aβtoxicity. In our attempt to model AD, instead of using iPScells carrying genetic mutations implicated in the disease, weused the neurons derived from healthy-subject iPS cells andestablished a toxicity assay using Aβ1–42 aggregates. To our

Figure 6 Screen for inhibitors of Aβ1–42 toxicity in iCell Neuroagainst Aβ1–42 toxicity showed that 19 compounds (blue dots) prot

knowledge, this is the first demonstration of the Aβ toxicityassay using hiPS cell-derived neurons, which substantiates thepotential of the hiPS cells in high-throughput compoundscreens.

One issue with using iPS cells is that random integrationof exogenous reprogramming genes might lead to clonalheterogeneity and possible functional diversity. Therefore,it is necessary to validate findings from one hiPS cell clonewith other independently derived hiPS cell clones. Anotherpotential issue is that it is usually challenging to generate alarge amount of highly purified cells for high-throughputdrug screening. Recently, iPS cell-derived derivatives havebecome commercially available as cryopreserved assay-ready cells (Prescott, 2011), relieving scientists from thelengthy, laborious and complicated differentiation process.General acceptance of these products in research and drugdiscovery, however, requires detailed characterization andvalidation studies (Chai et al., 2012; Whitemarsh et al.,2012). In this study, we carried out side-by-side comparisonsof the iCell Neurons with our own iPS cell differentiatedneurons and confirmed that abnormal CCEs after Aβ1–42treatment and rescuing the neuronal apoptosis by Cdkinhibition could be observed in both cell populations,consistent with the findings in H9 ES cell-derived neurons(Supplemental Fig. 4). The availability of iCell Neurons inlarge quantities should enable high-throughput screeningwith a more physiologically relevant cellular model.

A series of studies have demonstrated that GABAergicneurons were found to be affected in AD patients (Inagumaet al., 1992; Mikkonen et al., 1999; Takahashi et al., 2010),as well as cholinergic neurons and glutamatergic neurons(Danysz et al., 2000; Muir et al., 1994; Selkoe, 2002).However, some researchers believe that GABAergic neuronswere relatively resistant in AD due to their high expression ofcalcium-binding proteins (CBPs) (Mattson and Magnus,2006). In our cellular ADmodel, weemployed amixed neuronalculture expressing GABAergic, glutamatergic, and cholinergicneuron markers and we did not find that specific types ofneurons demonstrated significantly selective sensitivity to

ns. Scatter plot of all compounds with their protective effectsected iCell Neurons against Aβ1–42 toxicity.

225Cdk inhibition prevents Aβ induced toxicity in hiPS cell-derived neurons.

Aβ1–42 (Supplemental Fig. 5). It is possible that the neuronsused in these experiments are not mature enough to expresshigh levels of CBPs to afford the protection, compared with theones in AD patients. Moreover, different neuron cultureprotocols, Aβ species, and treatment time may also lead tothe different neuronal susceptibility to toxic exposure (Kranticet al., 2011; Pakaski et al., 1998). Further investigations maybe necessary to clarify the association between neuronalmature extent and their susceptibility to Aβ toxic effects.

Cdk inhibitors have been widely tested in animal models ofcentral nervous system diseases including AD, PD, stroke andtraumatic brain injury (Hilton et al., 2008; Jorda et al., 2003;Osuga et al., 2000). Though they have been demonstrated toimprove behavioral outcomes and increase neuronal survivalin those animal models, the lack of specificity of those Cdkinhibitors may cause side effects and other issues. In previousreports, Park et al. demonstrated that dominant negativeforms of Cdk4 and Cdk6, but not Cdk2 and Cdk3, prevented Aβinduced neuronal death (Giovanni et al., 1999; Park et al.,1997). While Copani et al. found that a dominant negativemutant of Cdk2 also protected neurons against Aβ toxicity(Copani et al., 1999, 2001). In our study, we found that bothshCdk2 and shCdk4 could decrease neuronal death caused byAβ, but shCdk2 were more effective than shCdk4. Thisdiscrepancy might be caused by the different experimentalparadigms including different forms of Aβ aggregates,different gene silencing methods and different neuronal celltypes used in those studies. It is still not clear from the currentstudy which factors from the aberrant CCEs trigger theapoptosis pathways. The E2F1 related signaling pathways aregenerally accepted to be involved in neuronal apoptosis. Forinstance, E2F1 could induce caspase activation by directlyincreasing intracellular Apaf-1 levels (Furukawa et al., 2002).It also inhibits the NF-KB related survival signals and favors theaccumulation of ROS, which induce the occurrence ofapoptosis (Phillips et al., 1999). Furthermore, E2F1 alsomight induce the expression of other apoptosis relatedgenes, such as Bcl2 (Eischen et al., 2001), Cdc2 (Konishi andBonni, 2003) and Bim (Biswas et al., 2005). In this study, wefound that Aβ1–42 treatment triggered DNA synthesis andinduced the increased expression of Cyclin D1, Cdk2 and CyclinA, though it should be noted that the cell cycle re-entry is onlyone of the potential risk factors from the multitude ofactivated signaling pathways in AD. The evidence suggeststhat other activated mitogenic pathways also exist in a similarmanner. For example, MAPK/ERK1/2 signaling and PLC/IP3/PKC/JNK signaling have been reported to be involved in thecell cycle-reentry in AD researches (Frasca et al., 2004; Grantet al., 2001; Lopez-Bergami and Ronai, 2008). The complexityin AD mechanisms may account for our results demonstratingthat targeting the singlemolecule (Cdk2 inhibition) only partlyrescues and delays the neuronal apoptosis rather thancompletely blocks neuronal death. Therefore, it will likelybe necessary to find additional signaling components inaddition of targeting Cdk2 for therapeutic interventions forAD.

In summary, to establish a cellular model for AD, we usedAβ1–42 aggregates as insults to induce the neuronal apoptosisin the hiPS cell-derived neurons. We observed that abnormalcell cycle re-entry occurred in hiPS cell-derived neurons afterAβ1–42 treatment, and that a series of small-molecule Cdkinhibitors and shRNAs against Cdk2 and Cdk4 could efficiently

block the cellular toxicity elicited by Aβ1–42. Furthermore,we performed a high-throughput drug screen and found somepotent small molecules which could efficiently preventneurons against Aβ toxic effects. Therefore, our studyprovided an excellent example of how hiPS cells can be usedin disease modeling and high-throughput compound screeningfor neurodegenerative diseases.

Disclosure statement

The authors declare that Emile Nuwaysir is working forCellular Dynamics International, Inc. and holds for stockoptions in the company. There are no other potentialconflicts to disclose.

Acknowledgments

We would like to express our sincere gratitude to our collaboratorsfrom Cellular Dynamics International Inc., especially to SusanDeLaura, Lucas Chase, Jeff Grinager, David Majewski and MonicaStrathman for the support on the preparation of manuscript. We alsodeeply appreciated Dwight Morrow and John McNeish (both fromRegenerative Medicine DPU of GlaxoSmithKline, USA) for theircomments and editing of the manuscript. This work also receivedsupport from the State Key Development Program of Basic Researchof China (Project No. 2012CBA01303).

Appendix A. Supplementary data

Supplementary data to this article can be found online athttp://dx.doi.org/10.1016/j.scr.2012.11.005.

References

Bardy, J., Chen, A., Lim, Y.M., Wu, S.M., Wei, S., Han, W., Chan,K.K., Reuveny, S., Oh, S., 2012. Microcarrier suspension culturesfor high density expansion and differentiation of humanpluripotent stem cells to neural progenitor cells. Tissue Eng.Part C Methods http://dx.doi.org/10.1089/ten.TEC.2012.0146.

Baumann, K., Mandelkow, E.M., Biernat, J., Piwnica-Worms, H.,Mandelkow, E., 1993. Abnormal Alzheimer-like phosphorylationof tau-protein by cyclin-dependent kinases cdk2 and cdk5. FEBSLett. 336, 417–424.

Bissonnette, C.J., Lyass, L., Bhattacharyya, B.J., Belmadani, A.,Miller, R.J., Kessler, J.A., 2011. The controlled generation offunctional basal forebrain cholinergic neurons from humanembryonic stem cells. Stem Cells 29, 802–811.

Biswas, S.C., Liu, D.X., Greene, L.A., 2005. Bim is a direct target ofa neuronal E2F-dependent apoptotic pathway. J. Neurosci. 25,8349–8358.

Bonda, D.J., Evans, T.A., Santocanale, C., Llosa, J.C., Vina, J.,Bajic, V.P., Castellani, R.J., Siedlak, S.L., Perry, G., Smith,M.A., Lee, H.G., 2009. Evidence for the progression through S-phase in the ectopic cell cycle re-entry of neurons in Alzheimerdisease. Aging 1, 382–388.

Brennand, K.J., Simone, A., Jou, J., Gelboin-Burkhart, C., Tran, N.,Sangar, S., Li, Y., Mu, Y., Chen, G., Yu, D., McCarthy, S., Sebat,J., Gage, F.H., 2010. Modelling schizophrenia using humaninduced pluripotent stem cells. Nature 473, 221–225.

Busser, J., Geldmacher, D.S., Herrup, K., 1998. Ectopic cell cycleproteins predict the sites of neuronal cell death in Alzheimer'sdisease brain. J. Neurosci. 18, 2801–2807.

226 X. Xu et al.

Chai, X., Dage, J.L., Citron, M., 2012. Constitutive secretion of tauprotein by an unconventional mechanism. Neurobiol. Dis. 48,356–366.

Chambers, S.M., Fasano, C.A., Papapetrou, E.P., Tomishima, M.,Sadelain, M., Studer, L., 2009. Highly efficient neural conversionof human ES and iPS cells by dual inhibition of SMAD signaling.Nat. Biotechnol. 27, 275–280.

Copani, A., Condorelli, F., Caruso, A., Vancheri, C., Sala, A.,Giuffrida Stella, A.M., Canonico, P.L., Nicoletti, F., Sortino,M.A., 1999. Mitotic signaling by beta-amyloid causes neuronaldeath. FASEB J. 13, 2225–2234.

Copani, A., Uberti, D., Sortino, M.A., Bruno, V., Nicoletti, F., Memo,M., 2001. Activation of cell-cycle-associated proteins in neuronaldeath: a mandatory or dispensable path? Trends Neurosci. 24,25–31.

Crittenden, J.R., Heidersbach, A., McManus, M.T., 2007. Lentiviralstrategies for RNAi knockdown of neuronal genes. Curr. Protoc.Neurosci. 5.26.1–5.26.21 (Chapter 5, Unit 5.26).

Cummings, B.J., Cotman, C.W., 1995. Image analysis of beta-amyloid load in Alzheimer's disease and relation to dementiaseverity. Lancet 346, 1524–1528.

Dahlgren, K.N., Manelli, A.M., Stine Jr., W.B., Baker, L.K., Krafft,G.A., LaDu, M.J., 2002. Oligomeric and fibrillar species ofamyloid-beta peptides differentially affect neuronal viability. J.Biol. Chem. 277, 32046–32053.

Danysz, W., Parsons, C.G., Mobius, H.J., Stoffler, A., Quack, G., 2000.Neuroprotective and symptomatological action of memantinerelevant for Alzheimer's disease—a unified glutamatergic hypoth-esis on the mechanism of action. Neurotox. Res. 2, 85–97.

Davis, S.T., Benson, B.G., Bramson, H.N., Chapman, D.E., Dickerson,S.H., Dold, K.M., Eberwein, D.J., Edelstein, M., Frye, S.V.,Gampe Jr., R.T., Griffin, R.J., Harris, P.A., Hassell, A.M.,Holmes, W.D., Hunter, R.N., Knick, V.B., Lackey, K., Lovejoy,B., Luzzio, M.J., Murray, D., Parker, P., Rocque, W.J.,Shewchuk, L., Veal, J.M., Walker, D.H., Kuyper, L.F., 2001.Prevention of chemotherapy-induced alopecia in rats by CDKinhibitors. Science 291, 134–137.

Dimos, J.T., Rodolfa, K.T., Niakan, K.K., Weisenthal, L.M.,Mitsumoto, H., Chung, W., Croft, G.F., Saphier, G., Leibel, R.,Goland, R., Wichterle, H., Henderson, C.E., Eggan, K., 2008.Induced pluripotent stem cells generated from patients with ALScan be differentiated into motor neurons. Science 321,1218–1221.

Ebert, A.D., Yu, J., Rose Jr., F.F., Mattis, V.B., Lorson, C.L.,Thomson, J.A., Svendsen, C.N., 2009. Induced pluripotent stemcells from a spinal muscular atrophy patient. Nature 457,277–280.

Eischen, C.M., Packham, G., Nip, J., Fee, B.E., Hiebert, S.W.,Zambetti, G.P., Cleveland, J.L., 2001. Bcl-2 is an apoptotictarget suppressed by both c-Myc and E2F-1. Oncogene 20,6983–6993.

Frasca, G., Chiechio, S., Vancheri, C., Nicoletti, F., Copani, A.,Angela Sortino, M., 2004. Beta-amyloid-activated cell cycle inSH-SY5Y neuroblastoma cells: correlation with the MAP kinasepathway. J. Mol. Neurosci. 22, 231–236.

Furukawa, Y., Nishimura, N., Furukawa, Y., Satoh, M., Endo, H.,Iwase, S., Yamada, H., Matsuda, M., Kano, Y., Nakamura, M.,2002. Apaf-1 is a mediator of E2F-1-induced apoptosis. J. Biol.Chem. 277, 39760–39768.

Giovanni, A., Wirtz-Brugger, F., Keramaris, E., Slack, R., Park, D.S.,1999. Involvement of cell cycle elements, cyclin-dependentkinases, pRb, and E2F x DP, in B-amyloid-induced neuronaldeath. J. Biol. Chem. 274, 19011–19016.

Grant, E.R., Errico, M.A., Emanuel, S.L., Benjamin, D., McMillian,M.K., Wadsworth, S.A., Zivin, R.A., Zhong, Z., 2001. Protectionagainst glutamate toxicity through inhibition of the p44/42mitogen-activated protein kinase pathway in neuronally differ-entiated P19 cells. Biochem. Pharmacol. 62, 283–296.

Greene, L.A., Liu, D.X., Troy, C.M., Biswas, S.C., 2007. Cell cyclemolecules define a pathway required for neuron death indevelopment and disease. Biochim. Biophys. Acta 1772, 392–401.

Hester, M.E., Murtha, M.J., Song, S., Rao, M., Miranda, C.J., Meyer,K., Tian, J., Boulting, G., Schaffer, D.V., Zhu, M.X., Pfaff, S.L.,Gage, F.H., Kaspar, B.K., 2009. Rapid and efficient generation offunctional motor neurons from human pluripotent stem cellsusing gene delivered transcription factor codes. Mol. Ther. 19,1905–1912.

Hilton, G.D., Stoica, B.A., Byrnes, K.R., Faden, A.I., 2008.Roscovitine reduces neuronal loss, glial activation, and neuro-logic deficits after brain trauma. J. Cereb. Blood Flow Metab. 28,1845–1859.

Inaguma, Y., Shinohara, H., Inagaki, T., Kato, K., 1992. Immunore-active parvalbumin concentrations in parahippocampal gyrusdecrease in patients with Alzheimer's disease. J. Neurol. Sci.110, 57–61.

Israel, M.A., Yuan, S.H., Bardy, C., Reyna, S.M., Mu, Y., Herrera, C.,Hefferan, M.P., Van Gorp, S., Nazor, K.L., Boscolo, F.S., Carson,C.T., Laurent, L.C., Marsala, M., Gage, F.H., Remes, A.M., Koo,E.H., Goldstein, L.S., 2012. Probing sporadic and familialAlzheimer's disease using induced pluripotent stem cells. Nature482, 216–220.

Jorda, E.G., Verdaguer, E., Canudas, A.M., Jimenez, A., Bruna, A.,Caelles, C., Bravo, R., Escubedo, E., Pubill, D., Camarasa, J.,Pallas, M., Camins, A., 2003. Neuroprotective action offlavopiridol, a cyclin-dependent kinase inhibitor, in colchicine-induced apoptosis. Neuropharmacology 45, 672–683.

Konishi, Y., Bonni, A., 2003. The E2F–Cdc2 cell-cycle pathwayspecifically mediates activity deprivation-induced apoptosis ofpostmitotic neurons. J. Neurosci. 23, 1649–1658.

Krantic, S., Isorce, N., Mechawar, N., Davoli, M.A., Vignault, E.,Albuquerque, M., Chabot, J.G., Moyse, E., Chauvin, J.P.,Aubert, I., McLaurin, J., Quirion, R., 2011. HippocampalGABAergic neurons are susceptible to amyloid-beta toxicity invitro and are decreased in number in the Alzheimer's diseaseTgCRND8 mouse model. J. Alzheimer's Dis. 29, 293–308.

Laustriat, D., Gide, J., Peschanski, M., 2010. Human pluripotentstem cells in drug discovery and predictive toxicology. Biochem.Soc. Trans. 38, 1051–1057.

Lee, G., Papapetrou, E.P., Kim, H., Chambers, S.M., Tomishima,M.J., Fasano, C.A., Ganat, Y.M., Menon, J., Shimizu, F., Viale,A., Tabar, V., Sadelain, M., Studer, L., 2009. Modellingpathogenesis and treatment of familial dysautonomia usingpatient-specific iPSCs. Nature 461, 402–406.

Lim, A.C., Qi, R.Z., 2003. Cyclin-dependent kinases in neuraldevelopment and degeneration. J. Alzheimer's Dis. 5, 329–335.

Liu, H., Zhang, S.C., 2011. Specification of neuronal and glialsubtypes from human pluripotent stem cells. Cell. Mol. Life Sci.68, 3995–4008.

Lopez-Bergami, P., Ronai, Z., 2008. Requirements for PKC-augmented JNK activation by MKK4/7. Int. J. Biochem. CellBiol. 40, 1055–1064.

Mattson, M.P., Magnus, T., 2006. Ageing and neuronal vulnerability.Nat. Rev. Neurosci. 7, 278–294.

Mikkonen, M., Alafuzoff, I., Tapiola, T., Soininen, H., Miettinen, R.,1999. Subfield-and layer-specific changes in parvalbumin,calretinin and calbindin-D28K immunoreactivity in the entorhi-nal cortex in Alzheimer's disease. Neuroscience 92, 515–532.

Mosch, B., Morawski, M., Mittag, A., Lenz, D., Tarnok, A., Arendt,T., 2007. Aneuploidy and DNA replication in the normal humanbrain and Alzheimer's disease. J. Neurosci. 27, 6859–6867.

Muir, J.L., Everitt, B.J., Robbins, T.W., 1994. AMPA-inducedexcitotoxic lesions of the basal forebrain: a significant role forthe cortical cholinergic system in attentional function. J.Neurosci. 14, 2313–2326.

Nevins, J.R., 1992. E2F: a link between the Rb tumor suppressorprotein and viral oncoproteins. Science 258, 424–429.

227Cdk inhibition prevents Aβ induced toxicity in hiPS cell-derived neurons.

Nguyen, L., Wright, S., Lee, M., Ren, Z., Sauer, J.M., Hoffman, W.,Zago, W., Kinney, G.G., Bova, M.P., 2012. Quantifying amyloidbeta (abeta)-mediated changes in neuronal morphology inprimary cultures: implications for phenotypic screening. J.Biomol. Screen. 17, 835–842.

Osuga, H., Osuga, S., Wang, F., Fetni, R., Hogan, M.J., Slack, R.S.,Hakim, A.M., Ikeda, J.E., Park, D.S., 2000. Cyclin-dependentkinases as a therapeutic target for stroke. Proc. Natl. Acad. Sci.U. S. A. 97, 10254–10259.

Pakaski, M., Farkas, Z., Kasa Jr., P., Forgon, M., Papp, H., Zarandi,M., Penke, B., Kasa Sr., P., 1998. Vulnerability of smallGABAergic neurons to human beta-amyloid pentapeptide. BrainRes. 796, 239–246.

Park, D.S., Levine, B., Ferrari, G., Greene, L.A., 1997. Cyclindependent kinase inhibitors and dominant negative cyclindependent kinase 4 and 6 promote survival of NGF-deprivedsympathetic neurons. J. Neurosci. 17, 8975–8983.

Petit, I., Kesner, N.S., Karry, R., Robicsek, O., Aberdam, E., Muller,F.J., Aberdam, D., Ben-Shachar, D., 2011. Induced pluripotentstem cells from hair follicles as a cellular model for neurodevelop-mental disorders. Stem Cell Res. 8, 134–140.

Phillips, A.C., Ernst, M.K., Bates, S., Rice, N.R., Vousden, K.H.,1999. E2F-1 potentiates cell death by blocking antiapoptoticsignaling pathways. Mol. Cell 4, 771–781.

Prescott, C., 2011. The business of exploiting induced pluripotent stemcells. Philos. Trans. R. Soc. Lond. B Biol. Sci. 366, 2323–2328.

Ranganathan, S., Bowser, R., 2003. Alterations in G 1 to S phasecell-cycle regulators during amyotrophic lateral sclerosis. Am. J.Pathol. 162, 823–835.

Ranganathan, S., Scudiere, S., Bowser, R., 2001. Hyperphosphorylationof the retinoblastoma gene product and altered subcellulardistribution of E2F-1 during Alzheimer's disease and amyotrophiclateral sclerosis. J. Alzheimer's Dis. 3, 377–385.

Schor, N.F., 2011. What the halted phase III gamma-secretase inhibitortrial may (or may not) be telling us. Ann. Neurol. 69, 237–239.

Selkoe, D.J., 2002. Alzheimer's disease is a synaptic failure. Science298, 789–791.

Shi, Y., Kirwan, P., Smith, J., MacLean, G., Orkin, S.H., Livesey,F.J., 2012. A human stem cell model of early Alzheimer's diseasepathology in Down syndrome. Sci. Transl. Med. 4, 124–129.

Sielecki, T.M., Boylan, J.F., Benfield, P.A., Trainor, G.L., 2000.Cyclin-dependent kinase inhibitors: useful targets in cell cycleregulation. J. Med. Chem. 43, 1–18.

Smith, M., Nagy, Z., Esiri, M., 1999. Cell cycle-related proteinexpression in vascular dementia and Alzheimer's disease.Neurosci. Lett. 271, 45–48.

Smith, P.D., O'Hare, M.J., Park, D.S., 2004. CDKs: taking on a role asmediators of dopaminergic loss in Parkinson's disease. TrendsMol. Med. 10, 445–451.

Song, B., Sun, G., Herszfeld, D., Sylvain, A., Campanale, N.V., Hirst,C.E., Caine, S., Parkington, H.C., Tonta, M.A., Coleman, H.A.,Short, M., Ricardo, S.D., Reubinoff, B., Bernard, C.C., 2011.Neural differentiation of patient specific iPS cells as a novelapproach to study the pathophysiology of multiple sclerosis.Stem Cell Res. 8, 259–273.

Takahashi, K., Yamanaka, S., 2006. Induction of pluripotent stemcells from mouse embryonic and adult fibroblast cultures bydefined factors. Cell 126, 663–676.

Takahashi, H., Brasnjevic, I., Rutten, B.P., Van Der Kolk, N., Perl,D.P., Bouras, C., Steinbusch, H.W., Schmitz, C., Hof, P.R.,Dickstein, D.L., 2010. Hippocampal interneuron loss in an APP/PS1 double mutant mouse and in Alzheimer's disease. BrainStruct. Funct. 214, 145–160.

Troy, C.M., Rabacchi, S.A., Friedman, W.J., Frappier, T.F., Brown,K., Shelanski, M.L., 2000. Caspase-2 mediates neuronal celldeath induced by beta-amyloid. J. Neurosci. 20, 1386–1392.

Varvel, N.H., Bhaskar, K., Patil, A.R., Pimplikar, S.W., Herrup, K.,Lamb, B.T., 2008. Abeta oligomers induce neuronal cell cycleevents in Alzheimer's disease. J. Neurosci. 28, 10786–10793.

Vassar, R., 2005. beta-Secretase, APP and Abeta in Alzheimer'sdisease. Subcell. Biochem. 38, 79–103.

Vincent, I., Rosado, M., Davies, P., 1996. Mitotic mechanisms inAlzheimer's disease? J. Cell Biol. 132, 413–425.

Wei, H., Tan, G., Manasi, Qiu, S., Kong, G., Yong, P., Koh, C., Ooi,T.H., Lim, S.Y., Wong, P., Gan, S.U., Shim, W., 2012. One-stepderivation of cardiomyocytes and mesenchymal stem cells fromhuman pluripotent stem cells. Stem Cell Res. 9, 87–100.

Whitemarsh, R.C., Strathman, M.J., Chase, L.G., Stankewicz, C.,Tepp, W.H., Johnson, E.A., Pellett, S., 2012. Novel applicationof human neurons derived from induced pluripotent stem cellsfor highly sensitive botulinum neurotoxin detection. Toxicol. Sci.126, 426–435.

Yagi, T., Ito, D., Okada, Y., Akamatsu, W., Nihei, Y., Yoshizaki, T.,Yamanaka, S., Okano, H., Suzuki, N., 2011. Modeling familialAlzheimer's disease with induced pluripotent stem cells. Hum.Mol. Genet. 20, 4530–4539.

Yang, Y., Geldmacher, D.S., Herrup, K., 2001. DNA replicationprecedes neuronal cell death in Alzheimer's disease. J. Neurosci.21, 2661–2668.

Yang, Y., Mufson, E.J., Herrup, K., 2003. Neuronal cell death ispreceded by cell cycle events at all stages of Alzheimer's disease.J. Neurosci. 23, 2557–2563.

Ye, W., Blain, S.W., 2010. S phase entry causes homocysteine-induced death while ataxia telangiectasia and Rad3 relatedprotein functions anti-apoptotically to protect neurons. Brain133, 2295–2312.

Yu, J., Vodyanik, M.A., Smuga-Otto, K., Antosiewicz-Bourget, J.,Frane, J.L., Tian, S., Nie, J., Jonsdottir, G.A., Ruotti, V., Stewart,R., Slukvin, I.I., Thomson, J.A., 2007. Induced pluripotent stemcell lines derived from human somatic cells. Science 318,1917–1920.

Zeng, H., Guo, M., Martins-Taylor, K., Wang, X., Zhang, Z., Park,J.W., Zhan, S., Kronenberg, M.S., Lichtler, A., Liu, H.X., Chen,F.P., Yue, L., Li, X.J., Xu, R.H., 2010. Specification of region-specific neurons including forebrain glutamatergic neurons fromhuman induced pluripotent stem cells. PLoS One 5, e11853.

Zhang, D., Jiang, W., Liu, M., Sui, X., Yin, X., Chen, S., Shi, Y.,Deng, H., 2009. Highly efficient differentiation of human EScells and iPS cells into mature pancreatic insulin-producingcells. Cell Res. 19, 429–438.

Zhou, T., Benda, C., Duzinger, S., Huang, Y., Li, X., Li, Y., Guo, X.,Cao, G., Chen, S., Hao, L., Chan, Y.C., Ng, K.M., Ho, J.C.,Wieser, M., Wu, J., Redl, H., Tse, H.F., Grillari, J., Grillari-Voglauer, R., Pei, D., Esteban, M.A., 2011. Generation ofinduced pluripotent stem cells from urine. J. Am. Soc. Nephrol.22, 1221–1228.