Alzheimer's disease-related amyloid-β induces synaptotoxicity in human iPS cell-derived neurons

Human induced pluripotent stem cell (iPSC)-derived neurons have been proposed to be a highly valuable cellular model for studying the pathomechanisms of Alzheimer's disease (AD). Studies employing patient-specific human iPSCs as models of familial and sporadic forms of AD described elevated levels of AD-related amyloid-β (Aβ). However, none of the present AD iPSC studies could recapitulate the synaptotoxic actions of Aβ, which are crucial early events in a cascade that eventually leads to vast brain degeneration. Here we established highly reproducible, human iPSC-derived cortical cultures as a cellular model to study the synaptotoxic effects of Aβ. We developed a highly efficient immunopurification procedure yielding immature neurons that express markers of deep layer cortical pyramidal neurons and GABAergic interneurons. Upon long-term cultivation, purified cells differentiated into mature neurons exhibiting the generation of action potentials and excitatory glutamatergic and inhibitory GABAergic synapses. Most interestingly, these iPSC-derived human neurons were strongly susceptible to the synaptotoxic actions of Aβ. Application of Aβ for 8 days led to a reduction in the overall FM4–64 and vGlut1 staining of vesicles in neurites, indicating a loss of vesicle clusters. A selective analysis of presynaptic vesicle clusters on dendrites did not reveal a significant change, thus suggesting that Aβ impaired axonal vesicle clusters. In addition, electrophysiological patch-clamp recordings of AMPA receptor-mediated miniature EPSCs revealed an Aβ-induced reduction in amplitudes, indicating an impairment of postsynaptic AMPA receptors. A loss of postsynaptic AMPA receptor clusters was confirmed by immunocytochemical stainings for GluA1. Incubation with Aβ for 8 days did not result in a significant loss of neurites or cell death. In summary, we describe a highly reproducible cellular AD model based on human iPSC-derived cortical neurons that enables the mechanistic analysis of Aβ-induced synaptic pathomechanisms and the development of novel therapeutic approaches.

In Alzheimer's disease (AD), synapse damage and synapse loss are thought to underlie cognitive deficits. 1 Oligomers of the amyloid-β (Aβ) peptide appear to induce synaptic failure as an early event in the etiology of AD. [2][3][4] However, despite its wellestablished synapse-impairing effects in rodent models, 5-7 the synaptotoxic actions of Aβ most relevant for the human disease have not been identified in a human model system. Several studies have investigated the synaptotoxic effects of Aβ in cultured rodent neurons and in transgenic mouse models revealing a multitude of potential mechanisms affecting synapses. Postsynaptic Aβ actions result in the loss of functional (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type) glutamate receptors, [8][9][10] involve long-term depression-like mechanisms, 9,11,12 and lead to the degradation of the entire postsynapse (dendritic spines). 9,11,13 In addition, several distinct presynaptic Aβ actions on the synaptic vesicle cycle have been described. 10,14 Furthermore, Aβ-induced impairments of axonal transport regulation and Aβ-induced axon degeneration have been found in rodent neurons. [15][16][17] This puzzling diversity of Aβ-induced synapse-related defects raises the question whether all of them are involved in the early pathomechanisms of human AD.
In addition to well-established animal systems, the modelling of human neurological disease pathologies by human induced pluripotent stem cell (hiPSC) technology 18 has been proposed as an innovative approach. [19][20][21] The in vitro differentiation of hiPSCs to excitable neurons has been reported using a variety of protocols. [22][23][24] However, quantitative analysis of both functional glutamatergic and GABAergic synapses has been difficult to achieve. 19,25,26 In addition to studying the functional properties of iPSC-derived human neurons from healthy individuals, the in vitro differentiation of 1 patient-derived iPSCs has been used to model complex neurodevelopmental and neurodegenerative diseases. 19,27,28 Recently, iPSCs derived from AD patients have been reported to exhibit increased secretion of Aβ upon in vitro neuronal differentiation; however, neither a loss of synapses nor an impairment of synapse function was detected. 21,[29][30][31][32][33] Here we describe a hiPSC-based, carefully optimized in vitro differentiation protocol, including a novel immunopanning step, which enabled us to study the deleterious effects of application of Aβ on human cortical neurons and on human synapses.

Results
Neural differentiation of hiPSCs and immunopurification of hiPSC-derived immature neurons. hiPSCs were cultured (Supplementary Figure S1) and in vitro differentiated using an embryoid body (EB) system similar to published protocols. 22 After initial differentiation, EBs were plated on a matrigel substrate leading to the formation of paired box protein 6 (Pax6)-expressing neuroepithelial rosettes (Supplementary Figure S2) that further differentiated to heterogeneous cultures also containing non-neuronal cells (Figures 1a and b). After 6-8 weeks of in vitro differentiation, heterogeneous cultures were dissociated to single cells, which were subjected to immunopurification. Classical immunopanning 34 with specific modifications was performed using the neural cell adhesion molecule (NCAM) antibody VIN-IS-53 to isolate immature neurons expressing NCAM at a high level. To quantify immunopanning efficiency, dissociated cells without immunopanning (control), dissociated cells isolated by NCAM immunopanning, and dissociated cells non-adherent to the panning plates, respectively (Figure 1c), were immunocytochemically stained for NCAM and the neuronal marker microtubule-associated protein 2 (MAP2) 1 day after immunopurification (Figures 1d and e). The fraction of MAP2-positive cells was strongly increased in cells isolated by NCAM immunopanning (91.2 ± 4.3%) as compared with control cells (28.1 ± 20.6%) and to cells nonadherent to the panning plates (12.2 ± 7.4%) (Figure 1g). The fraction of NCAM-positive cells was also increased by immunopanning ( Figure 1f); however, as expected from the low level NCAM expression in neural precursor cells, the increase was less pronounced as compared with MAP2. We next characterized the immunopurified immature neurons using immunocytochemistry. Staining for cortical marker proteins revealed that the vast majority of MAP2-positive cells expressed markers of deep layer cortical neurons (Ctip2 (chicken ovalbumin upstream promoter transcription factor-interacting protein 2), Tbr1 (T-box, brain, 1)), while only 5.0 ± 1.4% of the MAP2-positive neurons expressed the upper layer marker special AT-rich sequence-binding protein 2 (Satb2; Figures 1h and i). Similar to the composition of neuronal cell types in the in vivo cortex, 15.7 ± 1.7% of the MAP2-positive neurons were GABAergic (glutamic acid decarboxylase 67 (GAD67) positive) (Figures 1h and i). Survival of immature neurons was not affected by the immunopanning procedure (Figure 1c). In summary, NCAM immunopanning of hiPSC-derived heterogeneous cultures resulted in highly purified MAP2-positive immature deep-layer cortical neurons.
Morphological maturation and action potential generation in hiPSC-derived neurons purified by immunopanning. Further cultivation of the MAP2-positive immature human neurons led to the formation of extended neurites after 1 week and to the formation of a dense neuritic network of mature neurons exhibiting enlarged somata at 8 weeks after immunopurification ( Figure 2a). These cultures contained only very few glial cells due to the inhibition of proliferation of non-neuronal cells (see Materials and Methods). To study whether morphologically maturated iPSC-derived human neurons exhibit essential functional properties typical of cultured cortical neurons, we performed a basic electrophysiological characterization. Whole-cell patch-clamp recordings (at 8 weeks after immunopurification) of the membrane potential (57.8 ± 0.7 mV resting potential) in current-clamp mode revealed the generation of action potentials upon injection of depolarizing current in all neurons tested ( Figure 2b). To demonstrate the expression of voltagedependent Na + currents, we did whole-cell recordings in voltage-clamp mode at a holding potential of − 60 mV.
Step depolarizations of the membrane potential elicited typical inward Na + currents that were blocked by addition of tetrodotoxin (TTX; 1 μM; Figure 2c). Thus, electrophysiological analysis revealed essential functional properties such as electrical excitability and TTX-sensitive voltage-dependent Na + currents in hiPSC-derived neurons at 8 weeks after immunopurification.
Characterization of functional synapses in mature hiPSC-derived neurons purified by immunopanning. At this stage of maturation, the formation of synaptic structures was indicated by vesicle clusters immunocytochemically stained for the synaptic vesicle-associated proteins VAMP2 (vesicle-associated membrane protein 2)/synaptobrevin and  (Figures 3a and b). The presence of both glutamate and γ-aminobutyric acid (GABA) containing presynaptic vesicle clusters was confirmed by a punctate immunostaining on dendrites for the vesicular glutamate transporter 1 (vGlut1, colocalized with postsynaptic postsynaptic density protein 95 (PSD95) puncta) and for the vesicular GABA transporter (vGAT), respectively (Figures 3c  and d). Moreover, the formation of functional synapses was indicated by spontaneous miniature postsynaptic currents (mPSCs) that were observed by whole-cell patch-clamp recording at − 60 mV holding potential (Figures 3e and f). We further characterized spontaneous mPSCs pharmacologically by using specific antagonists. AMPA receptor-mediated miniature excitatory postsynaptic currents (mEPSCs) were isolated by addition of gabazine (10 μM) and TTX (1 μM), and were blocked by the AMPA receptor antagonist 6,7-dinitroquinoxaline-2,3-dione (DNQX; 10 μM; Figure 3e). GABA A receptor-mediated mPSCs were isolated by addition of DNQX and TTX and were blocked by addition of gabazine ( Figure 3f). This demonstrates the presence of both functional glutamatergic and GABAergic synapses. Taken together, our findings demonstrate that at 8 weeks after immunopurification our iPSC-derived human neurons exhibit essential functional properties, such as excitability, synaptic activity, and functional glutamatergic and GABAergic synapses.  37 To test for potential unspecific effects of the 7PA2 supernatant, we immunodepleted Aβ from the 7PA2 conditioned medium by using the anti-Aβ monoclonal antibody IC16. 37,38 Aβ-induced defects were analysed 8 days after Aβ application. First, we studied the effects of Aβ on cycling synaptic vesicle clusters in the processes of iPSC-derived human neurons. Cycling vesicle clusters were stained by the fluorescent dye FM4-64 (N-(3-triethylammoniumpropyl)-4-(6-(4-(diethylamino) phenyl) hexatrienyl) pyridinium dibromide). 39 Uptake of extracellularly added FM4-64 (10 μM) in recycling vesicles was stimulated by triggering the exocytosis and endocytosis of synaptic vesicles with extracellular electrical stimulation. After removal of extracellular FM4-64 dye, FM4-64-stained vesicle clusters were imaged by fluorescence microscopy (Figure 4a). For each FM4-64 staining experiment, the number of fluorescent puncta (i.e., vesicle clusters) was determined after thresholding the FM4-64 fluorescence image (see Figure 4a) and was multiplied with the mean total fluorescence intensity per punctum, resulting in an overall FM4-64 fluorescence signal per area of neuropil. Strikingly, 8 days after addition of Aβ, we observed a strong, significant reduction in the overall FM4-64 fluorescence signal (Figures 4a and b) indicating a toxic effect of Aβ on recycling vesicle clusters. Addition of immunodepleted 7PA2 supernatant did not affect the overall FM4-64 fluorescence signal demonstrating a specific effect of Aβ contained in the 7PA2-conditioned medium. The Aβ-induced reduction of the overall FM4-64 fluorescence signal was largely due to a reduction in the number of FM4-64-stained puncta (vesicle clusters) per area of neuropil as indicated by plotting FM4-64 puncta density versus the mean total fluorescence intensity per punctum (Figure 4b). We also studied the destaining of FM4-64-stained vesicle clusters by triggering vesicle re-exocytosis with extracellular electrical stimulation (1200 stimuli at 20 Hz (for 60 s)); Figure 4c). The mean destaining kinetics of FM4-64 puncta were not significantly affected by addition of Aβ, thus further supporting that a reduction in the number of vesicle clusters was the major Aβ-induced alteration.
To further confirm an Aβ-induced impairment of vesicle clusters, we immunocytochemically stained iPSC-derived human neurons for the synaptic vesicle marker vGlut1 after incubation with Aβ. Again, we observed a significant reduction in the density of vGlut1 immunopositive puncta (vesicle clusters) in neurites of Aβ-treated neurons as compared with vehicle-treated control cultures (Figures 4d and e). Addition of immunodepleted 7PA2 supernatant did not affect the number of vGlut1 puncta, again demonstrating a specific effect of Aβ contained in the 7PA2-conditioned medium. To check for Aβ-induced changes in the density of processes and in cell density, we fluorescently stained iPSC-derived human neurons by addition of the membrane permeable dye calcein-AM directly following the FM4-64 staining/destaining experiments ( Figure 4f). Eight days after addition of Aβ, the number of neuronal cell bodies and the density of processes (determined by counting processes crossing a randomly chosen line) were not significantly affected (Figures 4g and h). In summary, our results indicate that addition of Aβ results in a strong reduction in the density of vesicle clusters in the neurites of iPSC-derived human neurons.
Aβ induced functional impairment of glutamatergic synapses in iPSC-derived human neurons. We next studied whether the addition of Aβ induced deleterious changes at bona fide synaptic sites contacting postsynaptic dendrites. To focus on presynaptic vesicle clusters, we reanalysed the FM4-64 puncta (from the FM4-64 staining/ destaining experiments described above) that contacted the proximal dendrites of iPSC-derived human neurons. Proximal dendrites arising from neuron somata were identified by calcein-AM staining of whole neurons after FM4-64 staining/ destaining experiments, and FM4-64 puncta were analysed after thresholding the FM4-64 fluorescence image (Figure 5a). Eight days after addition of Aβ, we did not observe a significant change in the dendritic density of FM4-64 puncta or a significant change in total FM4-64 fluorescence intensity per punctum or a significant change in the overall presynaptic FM4-64 fluorescence signal per dendrite ( Figure 5b). Furthermore, the quantitative analysis of the stimulation-induced destaining of FM4-64 puncta did not reveal any significant changes in exocytosis (Figure 5c). We also reanalysed the presynaptic vGlut1 immunopositive puncta (described above) located on dendrites, which were identified by MAP2 co-immunostaining of dendrites ( Figure 5d). Again, 8 days after addition of Aβ, we observed  (Figure 5e). Taken together, the analysis of FM4-64 fluorescent and of vGlut1 immunopositive puncta on dendrites thus revealed no significant presynaptic alterations at a relatively early phase of incubation with Aβ, suggesting that bona fide presynaptic vesicle clusters and functional release properties were not affected by short-term application of Aβ.
We next addressed whether the deleterious effects of Aβ on functional properties of glutamatergic synapses, which are well described in rodent neurons, are also observed in iPSCderived human neurons. We recorded AMPA receptormediated mEPSCs (AMPA mEPSCs) using the whole-cell patch-clamp technique (1 μM TTX, 10 μM gabazine added, holding potential: − 60 mV). We again applied Aβ containing supernatant of the 7PA2 CHO cell line at 8 weeks after immunopurification and studied its effects on AMPA mEPSCs 8 days later. Addition of Aβ resulted in a strong, significant reduction of the mean amplitude of AMPA mEPSCs  (Figures 6a and b), indicating a loss of postsynaptic AMPA receptor function. This amplitude reduction was not observed upon addition of 7PA2 supernatant immunodepleted of Aβ, demonstrating a specific Aβ action. Addition of Aβ did not lead to a significant reduction of the mean frequency of AMPA mEPSCs (Figures 6a and b). The observed trend to a reduced AMPA mEPSC frequency is likely to be caused by limitations in the detection of small amplitude minis. To further confirm an Aβ induced impairment of postsynaptic AMPA receptors, we performed immunocytochemical stainings for GluA1(GluR1) (AMPA receptor subunit 1), PSD95, and MAP2 (to visualize dendrites) in iPSC-derived human neurons. Eight days addition of Aβ resulted in a significant reduction of the density of GluA1 immunopositive puncta on MAP2-immunostained dendrites (Figures 6c and d). Furthermore, the density of synaptic GluA1 immunopositive puncta (colocalized with PSD95 puncta) was also significantly reduced 8 days after Aβ application (Figure 6d). The dendritic density of PSD95 immunopositive puncta was not significantly altered ( Figures  6c and d). These Aβ-induced changes were not detectable upon addition of 7PA2 supernatant immunodepleted of Aβ, again demonstrating a specific Aβ action. Taken together, the observed Aβ-induced reduction of AMPA mEPSC amplitudes and of dendritic GluA1 expression indicates an early action of Aβ on postsynaptic AMPA receptors in iPSC-derived human neurons. Interestingly, as indicated by the lack of Aβ effects on vesicle clusters on dendrites, an early Aβ-induced presynaptic impairment was not detectable.
Aβ-induced effects on cell survival, cellular stress and tau protein in hiPSC-derived neurons. To study whether our short-term application of Aβ compromises cell survival, we stained healthy neurons and nuclei of disrupted cells, respectively by performing a live/dead assay (see Materials and Methods). Healthy neurons were fluorescently stained by addition of membrane permeable calcein-AM, which is intracellularly converted to fluorescent calcein. Nuclei without an intact plasma membrane were stained by Ethidium Homodimer 1 (Figure 7a). At 8 days of incubation, we did not observe significant effects of the addition of Aβ on cell survival (Figure 7b). Because Aβ-induced cellular stress might be detectable prior to cell death, we studied the expression of marker proteins for endoplasmic reticulum (ER) stress and autophagy by western blotting analysis. Interestingly, we found a slight Aβ-induced (8 days incubation) increase in the autophagy marker microtubule-associated protein 1A/1B-light chain II (LC3-II) and in the ER stress markers binding immunoglobulin protein (BiP) and CCAATenhancer-binding protein homologous protein (CHOP) (Figures 7c and d). This is in line with the recent description of increased ER stress in AD patient-derived iPSC lines upon neuronal differentiation. 40 We further analysed Aβ-induced changes in tau protein expression and phosphorylation state by western blotting. Although the expression level of tau protein appeared not to be affected by short-term application of Aβ (8 days incubation), an increased phosphorylation of tau was clearly detectable (Figures 7e and f). This finding in human neurons is in line with the Aβ-induced phosphorylation of tau protein, which has been described in mouse models previously. 4,41 In summary, our results indicate that at an early phase of Aβ actions defects in synapse function are clearly revealed in iPSC-derived human neurons by changes in postsynaptic AMPA receptors. In addition, Aβ-induced alterations in neuritic vesicle clusters, in ER stress/autophagy marker proteins, and in tau protein phosphorylation were found. Our work thus establishes that iPSC-derived human neurons represent an innovative model system to study the molecular mechanisms of Aβ-induced synapse damage in AD.

Discussion
In this study, we used hiPSCs to establish an innovative cell culture model of cortical neurons for specific aspects of AD. We analysed the effects of Aβ on human neurons, human synapses, and synaptic vesicle clusters. We observed a loss of axonal vesicle clusters, a loss of postsynaptic AMPA receptors, and increased tau protein phosphorylation, thus revealing complex deleterious actions of Aβ.
In vitro differentiation of human iPSCs and immunopurification of iPSC-derived immature human neurons. In vitro differentiation of human pluripotent stem cells to neurons has been established by several groups. 22,23,42 Without addition of specific morphogens/growth factors, 43 iPSCs differentiate to dorsal, cortical-like neuronal progenitor cells characterized by Pax6 expression; 20,22,44 however, other contaminating cell types are also generated. The invention of immunopanning protocols 34 enables the highly efficient purification of specific types of neuronal precursors, neurons, or astrocytes. [45][46][47] Similarly, the presence of NCAM has been used for selecting neuronally differentiated cells. 48,49 Here we have developed an immunopanning procedure based on NCAM antibodies, which yielded 490% MAP2-positive immature neurons. Immunocytochemical characterization of the purified neurons revealed typical cortical-like cultures consisting of mainly glutamaterigc and about 20% GABAergic neurons. Functional maturation of human iPSC-derived immature neurons, including synapse formation. In line with our results, it has been shown that neurons derived from human iPSCs generate action potentials and exhibit voltageactivated Na + and K + currents. 19,24,44,50 In addition, our iPSC-derived human neurons fired repetitive action potentials reminiscent of the firing pattern of regular spiking cortical pyramidal neurons. 24 Similar to primary cultured rodent neurons, 51 the formation of functional synapses has also been observed in human iPSC-derived neurons; however, reliable synapse formation and function appears to be difficult to achieve. 19,25,26 Importantly, our iPSC in vitro differentiation protocol resulted reliably in the formation of both functional glutamatergic and GABAergic synapses, thus leading to synaptically active networks.
Synaptotoxic effects of Aβ in human iPSC-derived neurons. We used our innovative model system for studying synaptic aspects of AD. Oligomeric Aβ peptides are thought to be crucial molecular entities in AD 1-3,5,52 and exhibit-in rodent neurons-toxic effects at glutamatergic synapses leading ultimately to synapse loss. 8,11,37,[52][53][54][55][56] The initial deleterious actions of Aβ on the function of glutamatergic synapses might be specifically localized either presynaptically or postsynaptically. In rodent neurons, an initial Aβinduced removal of postsynaptic AMPA receptors by facilitating AMPA receptor endocytosis has been described. 9,10,12 Furthermore, Aβ disrupts membrane trafficking and synaptic recruitment of AMPA receptors by reducing surface expression. 57 In line with an initially postsynaptic mechanism, we found a pronounced Aβ-induced reduction in both the amplitudes of AMPAR-mediated mEPSCs and in postsynaptic AMPA receptor clusters in hiPSC-derived neurons. In addition, also presynaptic mechanisms of action of Aβ affecting both exocytosis and endocytosis have been proposed in rodent neurons. 10,14 However, our analysis of both FM-stained and immunocytochemically stained vesicle clusters apposed to dendrites in hiPSC-derived neurons did not reveal any short-term presynaptic effects of Aβ, thus indicating that Aβ-induced alterations in postsynaptic AMPA receptors might be of crucial importance in the human disease.
In addition to Aβ-induced synaptotoxicity, AD is characterized by the formation of neurofibrillary tangles and tau hyperphosphorylation. 58 Aβ-induced tau phosphorylation 4,13,41 has been proposed to be an early event leading to the formation of neurofibrillary tangles typical of tauopathy. 41 In line with familial AD patient-derived iPSC-derived neurons, 33 our data demonstrate that Aβ-induced tau phosphorylation occurs in hiPSC-derived neurons after short-term application of Aβ. Furthermore, we observed a slight Aβ-induced increase in proteins related to the ER stress-induced unfolded protein response (UPR). UPR includes enhanced expression of chaperones such as BiP, increased degradation of misfolded proteins, and activates autophagy. 59 Failure of this system to restore ER homeostasis initiates apoptotic mechanisms. In line with our results, Aβ oligomers have been described to induce the UPR. 60 Effects of Aβ on vesicle clusters in hiPSC-derived neurons. Intriguingly, upon Aβ application we observed a reduced density of non-synaptic axonal vesicle clusters, which are well known to undergo exocytosis and endocytosis. 61,62 Because we did not observe any Aβ-induced changes in the FM destaining kinetics (reflecting exocytosis) and in the FM fluorescence intensity (reflecting endocytosis), an Aβ-induced impairment of the axonal transport of vesicles without axon degeneration appeared to underlie the reduction in the density of axonal vesicle clusters. In line with this idea, in rodent neurons an impairment of the axonal transport by Aβ has been observed. [15][16][17] Aβ-induced defects in the transport of organelles, for example, mitochondria, 63 and in the transport of dense core and synaptic vesicles 63,64 without axon degeneration have been described previously. Thus our results in iPSCderived human neurons strongly support an important role of an Aβ-induced impairment in axonal transport of synaptic vesicles at early stages of the human disease.
Conclusion: human iPSC derived neurons as an innovative model system to study toxic effects of Aβ. Here we have established hiPSC-derived neurons as an innovative AD model system to study the effects of Aβ. Toxic effects of Aβ were complex, including both non-synaptic effects on axonal vesicles and bona fide postsynaptic impairment of AMPA receptors. In addition, an Aβ induced increase in tau protein phosphorylation was found. Most importantly, our iPSC-derived human model system will aid the development of novel AD therapeutics based on the investigation of the molecular mechanisms of deleterious Aβ actions in human neurons. were cultured according to standard procedures in mTeSr medium (STEMCELL Technologies, Köln, Germany) or in E8 medium 67 on hESC qualified matrigel (Corning, Corning, NY, USA) coated cell culture six-well plates (Sarstedt, Nürnbrecht, Germany). Neural differentiation was performed according to a modified protocol of Li et al. 22 EBs were allowed to form in T25 cell culture flasks (Sarstedt) in N2B27 medium containing 50% DMEM/ F12 (Biochrom, Berlin, Germany), 50% Neurobasal, 2 mM Glutamax, 100 units/ml penicillin and 100 μg/ml streptomycin, 0.1 mM β-mercaptoethanol, 1% B27 without vitamin A (all Gibco, Darmstadt, Germany), 3 μg/ml heparin (Sigma Aldrich, St. Louis, MO, USA), and a modified N2 supplement with final concentrations of 7.5 μg/ml human insulin, 50 μg/ml BSA, 8 μg/ml putrescine, 3.1 ng/ml progesterone, 2.6 ng/ml sodium selenite (all Sigma Aldrich), and 25 μg/ml human holo-transferrin (Calbiochem, Darmstadt, Germany). After 14 days of suspension culture, EBs were plated on 100-mm cell culture dishes coated with growth factor reduced matrigel (Corning) in N2B27 medium. Six-to-8 weeks after starting differentiation, heterogeneous cultures were harvested for immunopanning by scraping in DPBS without calcium and magnesium (PanBiotech, Aidenbach, Germany).
Immunopanning of hiPSC-derived immature neurons and neuronal maturation. Panning plates (100 mm, bacterial petri dish quality; Falcon/Corning, Darmstadt, Germany) were prepared by incubating goat anti mouse IgG+IgM antibody (Jackson Immunoresearch, Suffolk, UK) at 10 μg/ml in 9 ml of 50 mM Tris HCl/pH 9.5 overnight at 4°C. After three washing steps with DPBS, the panning plates were incubated overnight at 4°C with the primary mouse anti-NCAM antibody (VIN-IS-53, kindly provided by Dr. Peter Andrews, University of Sheffield, UK) (0.55 μg/cm 2 in 0.2% BSA Fraction V (Gibco)/DPBS, 7 ml per plate). Plates were washed twice with DPBS and once with 0.02% BSA. Cells harvested from one 100-mm differentiation plate were enzymatically digested in accutase (Sigma Aldrich) for 30 min. After enzyme removal, cells were dissociated in 0.2% BSA in DPBS supplemented with 1 mg/ml glucose (Sigma Aldrich) and 0.33 mM sodium pyruvate (Gibco), pelleted at 200 g, resuspended in 0.02% BSA plus glucose and sodium pyruvate, and distributed to four panning plates (7 ml per plate). Cells were incubated on the panning plates for 10 min at RT. Non-adherent cells were removed by washing 10 times with DPBS. Adherent cells were detached by repeated pipetting (P1000, Eppendorf, Wesseling-Berzdorf, Germany) and rinsing of the plates with 0.02% BSA. Cells were pelleted at 130 × g for 10 min. Twenty-five thousand cells were plated on PO (1 mg/ml in borate buffer, coating overnight at RT) and laminin (10 mg/ml, coating 3 h at 37°C) coated coverslips (12 mm, Assistant, Sondheim/Rhön, Germany) in a 10-μl drop of neuronal culture medium, that is, Neurobasal, 2 mM Glutamax, 1 mM Sodium pyruvate, 100 units/ml penicillin, and 100 μg/ml streptomycin (all Gibco), 2% NS21 supplement, 68 25 μg/ml human insulin, 100 μg/ml BSA, 16 μg/ml putrescine, 62 ng/ml progesterone, 40 ng/ml sodium selenite (all Sigma Aldrich), 50 μg/ml human holo-transferrin (Calbiochem), and 30 ng/ml BDNF (Peprotech, Hamburg, Germany). Two millilitres of medium was added after neurons had attached (30 min at 37°C). To inhibit proliferation of neural stem cells, cultures were treated with 10 μM cytosine β-D-arabinofuranoside hydrochloride (Sigma Aldrich) for the first 5 days. After removal of the mitotic inhibitor, half of the medium was changed once per week over a period of 2-3 months. Evaporation was compensated for by the addition of aqua bidest (Sigma Aldrich).
All procedures were carried out as described. 37 In brief, conditioned neuronal culture medium containing cell-secreted Aβ was obtained from a CHO cell line (7PA2) that expresses mutant human APP751 (V717F). 35,36 As determined by ELISA (Aβ 40 ), 7PA2conditioned medium contained Aβ at a concentration of 25-50 ng/ml. Aβ containing 7PA2-conditioned medium was immunodepleted with anti-Aβ antibody (IC16) 38 coupled to NHS-Sepharose (GE Healthcare, Chalfont St Giles, UK) by overnight incubation at 4°C followed by centrifugation. Depletion of Aβ was confirmed by immunoblotting.
Immunocytochemistry and immunoblotting. Immunocytochemistry was carried out according to standard protocols, and quantification of images was carried out as described previously. 37 Antibodies used are listed in Supplementary Table S1. DAPI (10 μg/ml) was used to stain nuclei. Immunoblotting was performed according to standard protocols as described previously. 69 Briefly, cells were scraped in 2 × loading buffer (200 mM Tris/HCl pH 6.8, 5% SDS, 20% glycerol, 0.04% bromphenol blue, 200 mM DTT), boiled for 5 min and centrifuged at 16.000 × g for 3 min. Lysates of different samples were adjusted for GAPDH content after analyzing triplicates by SDS-PAGE and immunoblotting. Adjusted samples were separated by SDS-PAGE (12%, Tris-Glycine) and blotted on nitrocellulose membranes. Immunoblotting and protein detection was performed according to standard procedures. Membranes were blocked with 10% Roti-Block (Carl Roth, Karlsruhe, Germany) in Tris-buffered saline with 0.1% Tween 20. Proteins of interest were visualized using Western Bright Quantum ECL (Biozym, Hessisch Oldendorf, Germany) and the Chemi Doc XRS System (Biorad, Munich, Germany).
Calcein staining, fluorescence imaging, and analysis of cell death. For staining of healthy cells, we incubated cells with the calcein-green component (1/2000) of the LIVE/DEAD Kit (Molecular Probes, Darmstadt, Germany) for 10 min at 37°C followed by immunocytochemistry (or direct imaging), which was performed according to standard procedures. 37 Fluorescence imaging was carried out on a Zeiss Axiovert 200 M inverted fluorescence microscope (Zeiss, Oberkochen, Germany) equipped with a 12-bit monochrome CoolSNAP ES2 CCD camera (Photometrics, Tucson, AZ, USA) using the MetaVue and Metamorph (analysis) software (Visitron Systems, Puchheim, Germany). Aβ-induced cell death was studied by using the LIVE/ DEAD Kit (Molecular Probes) according to the manufacturer's instructions.
For studying exocytosis of FM4-64 stained vesicles, destaining of FM4-64 puncta was elicited by extracellular electrical stimulation (1200 stimuli at 20 Hz (for 60 s)). After 100 s of destaining, a second stronger stimulation (2000 stimuli at 20 Hz without imaging (for 100 s)) was carried out to reach background fluorescence levels. Quantitative image analysis was carried out by defining regions of interest (ROI) around FM4-64 puncta. Total fluorescence intensity was determined for each FM4-64 punctum using the Metamorph software (Molecular Devices). Total background intensity was determined after the second stimulation within the same ROI and was subtracted from the total fluorescence intensity within the ROI to obtain the total FM4-64 fluorescence signal per punctum. Decay time constants of FM4-64 destaining were determined by monoexponentially fitting (SigmaPlot11 software, Systat Software Inc, San Jose, CA, USA) the mean fluorescence decay after averaging the normalized (intensity at the start of stimulation was set to 100%) destaining curves of FM4-64 puncta obtained from a field of view ( Figure 4) or a proximal dendrite ( Figure 5).
Statistics. All data are given as means ± S.E.M. Results were statistically analysed using one-way ANOVA with Holm-Sidak posthoc test (SigmaPlot11 software).