Alzheimer’s in a dish – induced pluripotent stem cell-based disease modeling
Since the discovery of the induced pluripotent stem cell (iPSC) technique more than a decade ago, extensive progress has been made to develop clinically relevant cell culture systems. Alzheimer’s disease (AD) is the most common neurodegenerative disease, accounting for approximately two thirds of all cases of dementia. The massively increasing number of affected individuals explains the major interest of research in this disease as well as the strong need for better understanding of disease mechanisms.
IPSC-derived neural cells have been widely used to recapitulating key aspects of AD. In this Review we highlight the progress made in studying AD pathophysiology and address the currently available techniques, such as specific differentiation techniques for AD-relevant cell types as well as 2D and 3D cultures. Finally, we critically discuss the key challenges and future directions of this field and how some of the major limitations of the iPSC technique may be overcome.
Stem cell-based disease models have the potential to induce a paradigm shift in biomedical research. In particular, the combination of the iPSC technology with recent advances in gene editing or 3D cell cultures represents a breakthrough for in vitro disease modeling and provides a platform for a better understanding of disease mechanisms in human cells and the discovery of novel therapeutics.
KeywordsAlzheimer’s disease Induced pluripotent stem cells iPSC-derived neurons iPSC-derived astrocytes iPSC-derived microglia Disease modeling
Alzheimer’ s disease
Amyloid precursor protein
Clustered regularly interspaced short palindromic repeats
Hematopoietic progenitor cell
- iN cell
Induced neuronal cell
Induced pluripotent stem cell
Neural cell adhesion molecule 1
Neural progenitor cells
Presenilin 1 protein
Presenilin 1 gene
RAN binding protein 17
AD is the most common age-related neurodegenerative disease affecting memory and executive functions up to behavioral and neuropsychiatric changes in late stages. An estimated 40 million people, mostly older than 60 years, have dementia worldwide, and this number is expected to double every 20 years, until at least 2050 . The World Health Organization predicts that by 2040, neurodegenerative disease will overtake cancer to become the second leading cause of death after cardiovascular diseases . Thus, understanding cellular and molecular pathomechanisms of these disorders is of major importance. Current approved AD therapies are only symptomatic, and no treatment exists to fully stop progression or prevent disease onset.
For decades, primary rodent cell cultures represented the gold standard of mammalian in vitro systems. Despite a significant gain of knowledge on basic neural cell biology and disease pathophysiology obtained from these cells, translation to clinical research remains challenging. Cellular functions, morphology or expression of several disease-relevant proteins such as ApoE strongly differs between rodent and human cells (see following chapters). The establishment of the iPSC culture technique represents one of the most important innovations in biomedical research in this century. By reprogramming somatic human cells into an embryonic stem cell-like state, pluripotent cells are generated which can be further differentiated into virtually any somatic cell. The combination with gene editing techniques such as CRISPR/Cas9 further increases the value of this technique. The newest achievements are 3D models or self-organizing organoids which resemble patient-in-a-dish models of organs and can even display key features of organ-specific cytoarchitecture [3, 4].
Here we review the current research on iPSC-based modeling of AD, highlight current advances and discuss future challenges of the iPSC technology.
Alzheimer’s disease pathophysiology
The microtubule-associated protein tau has six isoforms which vary in length due to alternative splicing of exons 2, 3 and 10. Three isoforms contain 3 microtubule-binding repeats (3RTau) the other three isoforms contain 4 repeats (4RTau). In embryonic cells only one of the 3RTau isoforms is expressed, whereas mature neurons show presence of all six isoforms. Tau belongs to the major phosphoproteins in the brain. In the AD brain, tau is abnormally phosphorylated. In addition, tau is subjected to other posttranslational modifications which may further contribute to its pathologic role in AD, such as acetylation or N- and C-terminal truncation (reviewed in ). These modifications lead to altered tau folding thereby increasing its aggregation propensity. Aberrantly modified tau mediates toxicity downstream of Aβ [13, 14, 15] and knockout of endogenous tau abolished axonal transport deficits in APP transgenic mice  confirming its essential role in AD.
Genetically, AD is divided into two forms, familial (fAD) and sporadic AD (sAD). Familial cases have a predominantly early-onset (< 60 years, also called early-onset AD (EOAD)) while sAD cases with no or less familial evidence have later age of onset (> 60 years, late-onset AD (LOAD)) . FAD mutations either 1) increase the total production of Aβ, 2) shift the production to more toxic and aggregation-prone Aβ42 or 3) increase the aggregation propensity of the peptide, depending whether the mutation is localized 1) around the β-secretase cleavage, 2) at the γ–secretase cleavage site or 3) within the Aβ sequence . This explains the increased Aβ deposition in the brains of fAD patients. However, fAD accounts for less than 1% of all AD cases . The strongest nongenetic risk factors for sAD are age, medical factors such as diabetes mellitus, high blood pressure or hypercholesterolemia as well as socioeconomic factors such as low education or lack of social contacts . Although sAD is not associated with mutations that ultimately cause disease onset, genetic risk genes play a role in at least 80% of all AD cases . The major genetic risk factor is the presence of the apolipoprotein Ɛ4 (APOE4) allele. ApoE is a glycoprotein expressed mainly in the liver and in the brain. It is involved in transport of cholesterol across different tissues and cells and has been implicated in neuronal growth and repair, immune response, and activation of lipolytic enzymes . ApoE has three different isoforms, ApoE2, ApoE3 and ApoE4, which differ only in two amino acid residues. While APOE3/3 is the most common genotype and does not affect the AD risk, the presence of one or two copies of APOE4 increases the risk by tree or twelve fold, respectively . In vitro, ApoE4 has been shown to enhance aggregation of Aβ compared to ApoE3 . Analyses of ApoE in mouse models show that ApoE4 is associated with increased levels of Aβ, as well as amyloid plaque deposition . Furthermore, the different ApoE isoforms may differentially affect energy metabolism, neuroinflammation and synaptic function . More risk genes have been identified with different population frequencies and odds ratios including TREM2, BIN1, CLU, SORL1, ABCA7, CD33 and others (reviewed in [27, 28]).
Many of the AD risk genes show highest expression in glia, i.e. astrocytes (e.g. APOE, CLU, SORL1) or microglia (e.g. TREM2, CD33)  highlighting the importance of glia, especially microglia and astrocytes, for AD pathophysiology. In our opinion, in the last decade research focused too much on neurons, disregarding the role of glia. However, this recently changed and consequently, the number of studies in iPSC-derived glia increases and new differentiation protocols to obtain human astrocytes and microglia are being developed (see chapter Differentiation into disease-relevant cell types). Human glia structurally and functionally differs from murine glia [29, 30, 31, 32]. Hence, the use of human iPSC-derived glia is of high importance to i) verify already existing data from animal studies and ii) acquire new insights into the mechanisms by which human glia contribute to the pathogenesis of sAD.
iPSCs and disease modeling
Since the development of the iPSC technology, a major concern included the use of inserting retro viruses into the genome, which may cause inadvertent effects such as tumorigenesis. To circumvent these limitations, non-integrative techniques were established, including episomal vectors [35, 42], Sendai-virus , mRNA transfection  or piggyback-mediated OSKM transposition .
Differentiation into disease-relevant cell types
Pluripotent cells have the capacity to differentiate into three germ layers: endoderm, ectoderm and mesoderm. In turn, these multipotent cells can give rise to terminally differentiated cell types. Ectoderm gives rise to neuroectoderm and NPCs, which can finally be differentiated into neuronal or glial cells (i.e. astrocytes and oligodendrocytes). During neurodevelopment strict temporal and spatial expression of patterning factors determine cell fate. The most pivotal transcription factors are SMAD transcription factors, which play a fundamental role in transforming growth factor β (TGFβ) and bone morphogenic protein (BMP) signaling . In these pathways two different classes of receptor-regulated SMAD transcription factors, SMAD1/5/8 and SMAD2/3, are responsible for the respective signaling pathway. Inhibition of both classes of SMAD transcription factors is in turn a defining part of neuralation as their inhibition leads to the formation of the neural tube . During indirect neuronal differentiation, antagonizing TGFβ and BMP receptors causes dual SMAD inhibition, and has been shown effective in iPSC differentiation to neural cells , yielding over 80% of PAX6 positive neural progenitor cells (NPCs). Currently, NPC yields have been improved by the use of more stable, specific, and efficient small molecules . NPCs in turn can be differentiated to a mixed population of neuronal cells. Specific morphogens can be applied directing a specific cell fate. Sonic hedgehog induces a ventral fate such as GABAergic and motor neurons or sonic hedgehog in combination with retinoic acid induces midbrain dopaminergic neurons [55, 57, 58, 59].
To overcome laborious methods and impure yields, direct conversion of iPSCs to neurons was developed [60, 61] (Fig. 3). Viral overexpression of Ngn2 yielded a pure population of excitatory glutamatergic neurons within 2 weeks, called iN cells . Besides the quick and efficient aspects of directed differentiation, it also allows for a pure single cell type culture. This can be an advantage in order to study cell type-specific phenotypes, which can be diluted out in mixed neuronal cultures, and therefore not detected. In the scope of AD research, this enables the definition of pathogenic contributions of distinct cell types.
Until recently, astrocyte differentiation protocols were laborious and inefficient [63, 64]. By applying astrocyte-specific morphogens to NPCs, such as fibroblast growth factor 2, Heregulin-1β, insulin-like growth factor 1 and Activin A, functional astrocytes can now be derived within 30 days from the NPC stage . Alternatively, astrocytes can be differentiated from glial progenitor cells (GPCs). GPCs are generated similar to NPCs, however, platelet-derived growth factor was added during progenitor development. Addition of N2, B27, serum and Leukemia inhibitory factor to the GPCs resulted in formation of astrocytes after a total culture time of 6 to 8 weeks . Similar to induced neurons, it is also possible to directly convert fibroblasts to astrocytes by the forced overexpression of NFIA, NFIB and SOX9, yielding functional astrocytes in 2 weeks . However, this was only shown for murine fibroblasts. Human fibroblasts could be directly converted to astrocyte-like precursors by applying a chemical cocktail (histone deacetylase inhibitor valproic acid, the GSK3β inhibitor CHIR99021, the TGFβ inhibitor 616,452, the lysine specific histone demethylase LSD1 inhibitor tranylcypromine, the cyclic AMP inducer forskolin, and a histone methylation inhibitor DZNep). Astrocyte-like precursors had to be further differentiated into astrocytes but the efficiency was rather low with 15% GFAP positive and 40% S100β positive cells .
Simplified protocols repurposed commercial primary astrocyte maintenance medium for astrocyte induction when applied to NPCs . Both types of methods yielded functionally active astrocytes, which were able to take up glutamate and showed an immune response to several stimuli [65, 66, 69], including Aβ peptides, making these cells especially suitable to elucidate pathomechanisms in AD. This was evident by the switch from supportive glycolytic astrocytes, to oxidative, reactive astrocytes in cells expressing a PS1 fAD mutation , as well as a lack of neuronal support shown in sAD iPSC-astrocyte models [71, 72]. This supports the general understanding that astrocytes in AD lose their homeostatic support function and gain a toxic effect, thereby actively contribute to neuronal dysfunction in AD.
The development of microglia does not have an ectodermal origin, in fact, this cell type is derived from mesodermal hematopoietic stem cells, specifically yolk sack-derived c-Myb-independent primitive macrophages . It was not until recently that feasible microglia differentiation protocols were developed (see  for a complete overview). In short, recent studies pushed the development of erythromyeloid progenitors (EMPs), which are then directed to myelopoiesis by application of morphogens such as interleukin 3, and macrophage colony stimulating factor (M-CSF), additional factors are optional [75, 76], with one study using an initial selection step . Further, maturation and microglial enrichment requires supplementation with CSFR1 ligand interleukin 34 instead of M-CSF . IPSC-microglia were able to phagocytose fibrilar Aβ , express AD relevant genes such as APOE and TREM2 [75, 78], and were reactive in co-culture with neurons expressing fAD mutations . Furthermore, these genes were upregulated when microglia were treated with fibrilar Aβ . In addition, studies focusing on other neurodegenerative disorders have found specific phagocytosis deficits in iPSC-microglia, a mechanism also known to be affected in AD  indicating that iPSC-derived microglia could be an appropriate cell model to study AD pathogenesis.
Neural cells are routinely differentiated in a conventional two-dimensional (2D) culture setting. However, differentiating iPSCs in 3D cultures has been shown to increase cellular maturation, evident by elevated expression of synaptic marker synaptotagmin and neural cell adhesion molecule 1 in NPC-derived neurons . It should be noted that these NPCs, named ReN cells, did not originate from iPSCs. ReN cells are immortalized human neural progenitor cells, originally derived from the ventral mesencephalon region of the human fetal brain, which have the ability to differentiate into neurons and glial cells. 3D cultures can be established through culturing cells in scaffolds, such as Matrigel or synthetic hydrogels, or scaffold-free by self-organizing or transcription factor patterned organoids [3, 60, 82, 83, 84]. Using single type cell cultures , the effect of 3D culture conditions can be easily analysed for a certain cell type while the microenvironment, especially in hydrogel-based 3D cultures, is controllable. However, single cells do not recapitulate the complexity of neurons interacting with each other and with glial cells. On the other hand, the microenvironment and access to nutrients in self-organizing organoids cannot be controlled well, in addition to lack of cytoarchitectural structures. Patterned 3D corticospheres allow for the development of structured cortical lamination comprising of both superficial and deep cortical layer neurons, including non-reactive astrocytes . Allowing the spheres to grow in suspension rather than on semi-solid Matrigel or hydrogel enables the access of transcription factors and thus closer control of the cell microenvironment. However, to achieve matured 3D organoids the center will be deprived of nutrients and oxygen, creating a necrotic core, which may confound meaningful results . To overcome this, vasculature will need to be integrated into the organoids, which has been tried, but is still in a preliminary stage when using iPSCs (see  for more comprehensive review). While all 3D model systems have their advantages and disadvantages, evidently, these models showed valuable improvements to understanding AD pathology as mentioned below.
Even though iPSC-derived neural cells can mimic AD pathomechanisms to a certain extent in 2D cultures, the presence of robust Aβ plaque and neurofibrillary tangle pathology is overtly absent . By using 3D cultures, Aβ peptides are trapped in the matrix and not removed during medium changes and can therefore form aggregates in the 3D environment. Evidently, fAD models showed increase Aβ aggregations in hydrogel-based neuro-organoids  and single cell models . Another major finding was that Matrigel-based 3D cell cultures expressed 4RTau isoforms, which were not detected in 2D cultures  at the same time point. In long-term cultures (~ 6 months) of dopaminergic neurons, 4RTau expression was observed also in 2D . In addition to Aβ and Tau pathology, other AD-related pathways such as P21-activated kinase (PAK) were also found to be dysregulated in fAD cells in 3D but not in 2D . PAK signalling is sensitive to mechanotransduction, which may explain why this pathway is affected by fAD mutations in 3D cultures, but is due to its absence not affected in 2D cultures. When exposed to Aβ aggregates a significant PAK-mediated response was observed, which was absent in 2D cultures . Using 3D iPSC-neuronal models could allow for the study of AD relevant pathomechanisms in a setting that more closely reflects the in vivo context.
AD pathophysiology in iPSC-derived cells
Aβ and tau pathology
A prominent pathological feature of AD, the increased production of Aβ42, Aβ40 and increased Aβ42/40 ratio, was replicated in iPSC-derived mixed neuronal cultures generated from fAD fibroblasts [63, 87, 88, 89] and also in fAD NPCs . However, Aβ pathology was shown to be cell type specific, as fAD and sAD fibroblasts did not secrete increased levels of Aβ40 . Not only neurons but also fAD astrocytes were shown to secrete increased levels of Aβ42 and displayed reduced capacity to take up Aβ42 [70, 91], evidently showing that astrocytes are not mere bystanders in AD pathology. Besides aberrant Aβ metabolism, AD is also defined by the presence of hyperphosphorylated Tau aggregates. As described above, iPSC-derived neurons are known to be resemble immature neurons and do not express 4RTau in the timeframe of in vitro neuronal differentiation . 4RTau isoforms are more prone to aggregate due to the additional microtubule-binding domain. When iPSC-neurons were cultured for an extended period of time, up to 356 days in vitro, a shift from solely 3RTau expression to co-expression with 4RTau was observed .
Even in the absence of 4RTau expression, iPSC-neurons from both fAD and sAD patients showed an increase in phosphorylated tau (pTau) [89, 94]. In fAD neurons carrying an APP duplication (APPdp), major Tau kinase GSK-3β showed increased activity which was rescued by β-secretase inhibition . In addition, extracellular application of Aβ to HCS-derived iPSC-neurons induced Tau phosphorylation without affecting total Tau protein levels .
Even though most hallmarks were prominent in mixed neuronal cells derived from fAD patients, sAD patient-derived neuronal cells did not show a robust Aβ42/40 ratio or pTau increase [63, 89, 96, 97]. An explanation for these irregular observations in sAD neurons is the extensive variation of genetic background as described above. In addition, the heterogeneity of these cultures could mask distinct pathomechanisms in specific cell types. Using pure cultures of distinct neuronal subtypes could indicate which cell type contributes to specific disease mechanisms in sAD [89, 98].
Elevated Aβ production in neuronal cells is caused by aberrant APP processing. In iPSC-neurons the APPV717F mutation caused increased generation of Aβ and sAPPβ , induced by mislocalization and enrichment of APP in early endosomes. As BACE-1 shows highest activity in early endosomes, the accumulation of APP in its close proximity enables BACE-1 to readily cleave an abundance of APP. Enlarged endosomes have been observed in APPdp neurons but not in fAD neurons carrying the PSEN1ΔE9 mutation [89, 99]. Nonetheless, in the absence of enlarged or increased numbers of endosomes, a decrease in APP endo- and transcytosis was found . In addition, elevated levels of Rab11 were observed - a molecule playing a prominent role in receptor recycling and transport. Endocytosis of low-density lipoprotein (LDL) was also shown to be defective, with the decrease of LRP1 surface expression suggested as causative mechanism. These endocytosis deficits were rescued with β-secretase inhibition and were not dependent on the specific fAD mutation (e.g. APP or presenilin 1 (PS1)) . Consistently, sAD neuronal models also displayed deficits in endocytosis, for example by decrease of the endocytosis-mediating protein clathrin . The presence of endosome pathology independent of AD type or presence of APP or PS1 mutations, indicate that endocytosis and vesicle recycling may be a fundamental factor in AD pathogenesis.
As the presence of the APOE4 allele is the major genetic risk factor in sAD, this field of research has recently refocussed on neural models expressing different APOE variants, using either isogenic or patient derived cells distinguished based on APOE status [71, 72, 101]. ApoE is mainly expressed by astrocytes in the brain, and only expressed by neurons under stress conditions or trauma . Patient-derived iPSC-astrocyte models carrying homozygous APOE3 or APOE4 secreted ApoE with a different lipidation status, apoE4 lipoprotein particles were less lipidated compared to ApoE3 particles . Furthermore, isogenic homozygous APOE4 astrocytes both produced and secreted less ApoE than APOE3/3 cells . Gene regulation in APOE4/4 astrocytes was different compared to APOE3/3 astrocytes, with genes associated with lipid metabolism being prominently affected. Additionally, APOE4/4 astrocytes were less efficient in clearing Aβ42 which was shown to be a lysosome dependent effect . APOE4/4 astrocytes had a decreased ability to support neuronal and synaptic homeostasis, as neurons in co-culture expressed decreased amounts of neuronal and synaptic markers, in addition to decreased survival rates shown for isogenic , and patient-derived iPSC-astrocytes .
Neurons carrying the APOE4/4 allele showed increased Aβ40, Aβ42 and sAPPβ secretion as well as elevated Tau phosphorylation [71, 101]. Interestingly, inhibition of Aβ production did not decrease pTau in these cells. In a mixed neuronal culture expressing ApoE4, especially GABAergic neurons showed prominent degeneration due to increased apoptosis . Defective GABAergic neurons have previously been described in (pre-clinical) AD, and non-demented subjects carrying APOE4, which can cause network hyperexcitability . During differentiation, APOE4/4 cells take on a neuronal fate earlier than APOE3/3 cells and lose their ability to proliferate . Subsequently, APOE4 neurons showed an increase in synaptic density and neurotransmitter release, eventhough synaptic genes were downregulated, indicating the presence of an overcompensatory mechanism. These structural alterations could lead to network hyperexcitability, a known phenomenon in AD described above. Corresponding to endosomal pathology found in sAD and fAD neurons described earlier, APOE4/4 neurons also displayed an increased number of early endosomes. Genetic correction of APOE4/4 to APOE3/3 decreased these phenotypes [71, 101]. The APOE4/4 effects were also ameliorated by small molecule structure corrector PH002, which changes the conformation of ApoE4, resembling the conformation of ApoE3, as physiological Aβ40/Aβ42 ratio was restored and pTau levels were decreased . In addition, neuronal cells from a patient lacking the APOE gene (APOE −/−), showed a phenotype similar to APOE3/3 neurons, indicating that ApoE4 could have a gain of toxic function.
Next to astrocytes, microglia are the second major contributor of ApoE, especially plaque-associated ApoE, in the AD brain . IPSC-microglia carrying the APOE4/4 genotype showed a positive regulation of immune system-related genes, whereas genes involved in cell movement and development were downregulated . These cells had shorter and fewer processes compared to APOE3/3 microglia and showed slower phagocytosis of Aβ42. However, when seeded in neuronal organoids, APOE4/4 microglia display longer processes. Microglia with shorter processes have been described to be more efficient in Aβ uptake in the brain, indicating APOE4/4 microglia in the 3D environment have a reduced ability to sense Aβ42, which resulted in an increase in extracellular Aβ levels in the neuro-organoid.
Aβ has been described to interfere with synaptic processes and mediate synaptic degeneration and toxicity (reviewed in ). Neuronal and astrocytic iPSC-derived cells are electrophysiologically functional evident by spontaneous and evoked activity [89, 95, 106]. In addition, pre- and post-synaptic markers have been detected within close proximity, indicating the presence of functional synapses. Hence, iPSC-derived models enable the study of synaptic function and dysfunction. Synaptotoxic processes previously established in other AD models have likewise been observed in iPSC-derived neurons. Application of Aβ to healthy iPSC-neurons reduced the number of synaptic recycling pool (RP) vesicles but did not affect synaptic vesicle release or the abundance of pre-synaptic puncta . The RP vesicle deficit was caused by a defective synaptic vesicle reuptake mechanism, whereas the exocytosis kinetics were unchanged (e.g. no change in synaptic vesicle release). Both, exogenous Aβ application and endogenous Aβ in fAD neurons failed to show pre-synaptic mechanistical effects [89, 95]. However, in the post-synapse deficits were detected in the form of reduced post-synaptic puncta, and reduced AMPA receptor-mediated mEPSCs, indicating that Aβ had a pronounced post-synaptic effect in iPSC neurons . Interestingly, pre- and postsynaptic protein levels were also unaffected in cultures of sAD patient-derived glutamatergic neurons in which Aβ and pTau levels were not dysregulated . This could imply that either Aβ and/or pTau pathology is needed for a synaptotoxic phenotype, or the mere presence of one neuronal cell type (glutametergic neurons) cannot recapitulate the intricacy of a complex in vivo synaptic system.
Endoplasmic reticulum and mitochondrial stress
Mitochondrial metabolism, ER homeostasis and lysosomal clearance are essential to maintain neuronal and glial functions. In AD, however, these pivotal organelles have been implicated in contributing to neuronal degeneration. Early pathological events include the accumulation of misfolded protein in the ER, triggering a stress mechanism called the unfolded protein response (UPR) . Unfolded protein can be sequestered and cleared by autophagy, which appears to be dysfunctional in AD. The UPR was upregulated in fAD and sAD iPSC-neurons , as well as in healthy iPSC-neurons after application of Aβ, where ER stress proteins LC3-II, BiP and CHOP were increased . The aberrant accumulation of these proteins recapitulated the aforementioned deficits in autophagy. Accordingly, autophagy markers LC3-II and p62 levels where increased in fAD fibroblasts and neurons [108, 109]. In addition, decreased acidification of lysosomes, and increased levels of early endosomes/lysosomes were detected, indicating that the autophagic degradation phase was obstructed and lysosomal functioning was compromised . Further, presenilin-deficient iPSC-neurons showed altered lysosomal function, reduced autophagy and decreased levels of cytosolic and nuclear calcium . Dysfunctional calcium homeostasis was also observed in iPSC-astrocytes derived from a patient with the PSEN1ΔE9 mutation, contributing to loss of homeostatic neuronal support by astrocytes .
Oxidative stress, caused by increased mitochondrial ROS production, is a central process in normal aging. However, aberrant ROS production is observed in AD brains and has been linked to disease progression. Alterations in mitochondrial membrane composition, structure and subcellular localization of mitochondria were reported in AD . Accordingly, increased ROS levels were observed in iPSC-derived neurons and astrocytes from fAD and sAD patients [63, 98]. Interestingly, protein levels of mitochondrial oxidative phosphorylation chain complexes were upregulated in sAD neurons in the absence of alterations in Aβ or pTau, indicating that mitochondrial dysfunction may be an early event upstream of Aβ or Tau pathology .
Stem cell-based disease models have the potential to induce a paradigm shift in biomedical research. Although several concerns still limit the value and validity of iPSC-based studies, most of these limitations can be overcome by following some general guidelines. Studies on iPSC-based modeling of AD are often not reproducible or show opposite results. Better standardization of differentiation protocols – currently most labs use their own protocols, better characterization of the obtained cell cultures and use of more defined media and coatings will reduce the variability among labs and increase reproducibility. As described above, another concern, which is related to poor reproducibility, is the heterogeneity among different individuals and the heterogenous nature of sAD. To overcome this limitation, higher donor numbers, patients as well as controls, will be necessary. Alternatively, isogenic cells should be used instead of AD patient-derived cells, wherever possible. This further allows the analysis of specific disease risk genes, such as APOE or TREM2. To increase cellular maturity of iPSC-derived cells, which is lost during reprogramming of adult somatic cells, 3D cell cultures and organoid technologies represent a major step forward.
Taken together, iPSCs – either patient-derived or isogenic – constitute a breakthrough in modeling human diseases. Although several hurdles still have to be overcome, this technique allows understanding the earliest events in AD and connecting them to their underlying molecular mechanisms, providing the basis for identification of disease pathomechanisms and new therapeutic targets.
We thank Stephanie Davaz, Debora Wanner, Julian Birnbaum for fluorescence images and Manuel Gersbacher for creating the APP processing figure.
CT drafted the manuscript, SDL and CT wrote it. Both authors read and approved the final version.
CT and SDL was supported by a grant from the Neuroscience Center Zurich, University of Zurich, Switzerland and from the Betty and David Koetser Foundation for Brain Research, Zurich, Switzerland.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
- 9.Tackenberg C, Nitsch RM. The secreted APP ectodomain sAPPα, but not sAPPβ, protects neurons against Aβ oligomer-induced dendritic spine loss and increased tau phosphorylation. Molecular Brain. 2019;12(1):29.Google Scholar
- 14.Tackenberg C. Human tau-dependent toxicity in APP transgenic cultures depends on calcium influx through N-methyl-D-aspartate receptors. Matters. 2016. https://doi.org/10.19185/matters.201602000028.
- 18.Selkoe DJ. Alzheimer's disease. Cold Spring Harbor perspectives in biology. 2011;9(2):600–14.Google Scholar
- 20.F C. Nongenetic Risk Factors for Alzheimer’s Disease. In: D G, E S, editors. Neurodegenerative Diseases. London: Springer; 2014.Google Scholar
- 24.Sanan DA, Weisgraber KH, Russell SJ, Mahley RW, Huang D, Saunders A, et al. Apolipoprotein E associates with beta amyloid peptide of Alzheimer's disease to form novel monofibrils. Isoform apoE4 associates more efficiently than apoE3. J Clin Invest. 1994;94(2):860–9.PubMedPubMedCentralCrossRefGoogle Scholar
- 43.Fusaki N, Ban H, Nishiyama A, Saeki K, Hasegawa M. Efficient induction of transgene-free human pluripotent stem cells using a vector based on Sendai virus, an RNA virus that does not integrate into the host genome. Proc. Jpn. Acad. Ser. B Phys. Biol. Sci. 2009;85(8):348–62.PubMedPubMedCentralCrossRefGoogle Scholar
- 56.Pauly MG, Krajka V, Stengel F, Seibler P, Klein C, Capetian P. Adherent vs. Free-Floating Neural Induction by Dual SMAD Inhibition for Neurosphere Cultures Derived from Human Induced Pluripotent Stem Cells. Front Cell Dev Biol. 2018;6:3. https://doi.org/10.3389/fcell.2018.00003.
- 58.Cooper O, Hargus G, Deleidi M, Blak A, Osborn T, Marlow E, et al. Differentiation of human ES and Parkinson's disease iPS cells into ventral midbrain dopaminergic neurons requires a high activity form of SHH, FGF8a and specific regionalization by retinoic acid. Mol Cell Neurosci. 2010;45(3):258–66.PubMedPubMedCentralCrossRefGoogle Scholar
- 72.Zhao J, Davis MD, Atagi Y, Shinohara M, Graff-Radford NR, Younkin SG, et al. APOE epsilon4/epsilon4 diminishes neurotrophic function of human iPSC-derived astrocytes. Hum Mol Genet. 2017;26(14):2690–700.Google Scholar
- 78.Haenseler W, Sansom SN, Buchrieser J, Newey SE, Moore CS, Nicholls FJ, et al. A highly efficient human pluripotent stem cell microglia model displays a neuronal-co-culture-specific expression profile and inflammatory response. Stem Cell Reports. 2017;8(6):1727–42.PubMedPubMedCentralCrossRefGoogle Scholar
- 80.Garcia-Reitboeck P, Phillips A, Piers TM, Villegas-Llerena C, Butler M, Mallach A, et al. Human induced pluripotent stem cell-derived microglia-like cells harboring TREM2 missense mutations show specific deficits in phagocytosis. Cell Rep. 2018;24(9):2300–11.PubMedPubMedCentralCrossRefGoogle Scholar
- 93.Sposito T, Preza E, Mahoney CJ, Seto-Salvia N, Ryan NS, Morris HR, et al. Developmental regulation of tau splicing is disrupted in stem cell-derived neurons from frontotemporal dementia patients with the 10 + 16 splice-site mutation in MAPT. Hum Mol Genet. 2015;24(18):5260–9.PubMedPubMedCentralCrossRefGoogle Scholar
- 98.Birnbaum JH, Wanner D, Gietl AF, Saake A, Hock C, Nitsch RM, et al. Oxidative stress and altered mitochondrial protein expression in the absence of amyloid-beta and tau pathology in iPSC-derived neurons from sporadic Alzheimer’s disease patients. Stem Cell Res. 2018;27:121–30.PubMedCrossRefPubMedCentralGoogle Scholar
- 105.Tackenberg C, Ghori A, Brandt R. Thin, Stubby or Mushroom: Spine Pathology in Alzheimer's. In: Lahiri DK, editor. Advances in Alzheimer's Research. 2. Sharjah: Bentham Science Publishers; 2014. p. 375–94.Google Scholar
- 106.Iovino M, Agathou S, Gonzalez-Rueda A, Del Castillo Velasco-Herrera M, Borroni B, Alberici A, et al. Early maturation and distinct tau pathology in induced pluripotent stem cell-derived neurons from patients with MAPT mutations. Brain. 2015;138(Pt 11):3345–59.PubMedPubMedCentralCrossRefGoogle Scholar
- 109.Martin-Maestro P, Gargini R, A Sproul, A, Garcia E, Anton LC, Noggle S, et al. Mitophagy Failure in Fibroblasts and iPSC-Derived Neurons of Alzheimer's Disease-Associated Presenilin 1 Mutation. Front Mol Neurosci. 2017;10:291.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.