The Use of Pluripotent Stem Cell-Derived Organoids to Study Extracellular Matrix Development during Neural Degeneration

The mechanism that causes the Alzheimer’s disease (AD) pathologies, including amyloid plaque, neurofibrillary tangles, and neuron death, is not well understood due to the lack of robust study models for human brain. Three-dimensional organoid systems based on human pluripotent stem cells (hPSCs) have shown a promising potential to model neurodegenerative diseases, including AD. These systems, in combination with engineering tools, allow in vitro generation of brain-like tissues that recapitulate complex cell-cell and cell-extracellular matrix (ECM) interactions. Brain ECMs play important roles in neural differentiation, proliferation, neuronal network, and AD progression. In this contribution related to brain ECMs, recent advances in modeling AD pathology and progression based on hPSC-derived neural cells, tissues, and brain organoids were reviewed and summarized. In addition, the roles of ECMs in neural differentiation of hPSCs and the influences of heparan sulfate proteoglycans, chondroitin sulfate proteoglycans, and hyaluronic acid on the progression of neurodegeneration were discussed. The advantages that use stem cell-based organoids to study neural degeneration and to investigate the effects of ECM development on the disease progression were highlighted. The contents of this article are significant for understanding cell-matrix interactions in stem cell microenvironment for treating neural degeneration.


Alzheimer's Disease Pathology and Progression
Alzheimer's disease (AD) is a very common, incurable age-associated and economically costly disease characterized by progressive neurodegeneration, which causes deterioration and damage of neurons within the cerebral cortex, loss of memory, and cognitive decline [1]. It is the most common type of dementia and over 30 million people are suffering from this disease all over the world [2]. About 500,000 people die of AD every year and this number increases year by year. Currently, no effective treatments or available drugs can cure AD and the total costs in 2010 were around $172 billion in the United States [2], which makes it a heavy economic burden not only on the dementia patients but also on our global society. Most of clinical cases in AD develop late-onset symptoms (after the age of 65), also called sporadic AD (SAD). And about 2-5% of disease burden is an early-onset type, called familial AD (FAD). FAD is related to genetic mutations, such as mutations in amyloid precursor protein (APP), presenilin-1 (PS1), and presenilin-2 (PS2) [3]. The gene of PS1 encodes the encodes the catalytic subunit of γ-secretase that mediates APP cleavage for the generation of amyloid β42 (Aβ42) peptides [4].
AD is identified by three cardinal features in human brain: Aβ plaques, neurofibrillary tangles (NFTs), and neuron loss. Aβ plaques are usually formed because of the increased production or deficient clearance of neuronal Aβ caused by unknown reasons. APP is proteolyzed to the soluble and nontoxic Aβ monomers by β-and γ-secretases. The monomeric Aβ peptides then aggregate to form toxic Aβ oligomers and further generate the insoluble fibrils, which ultimately form the plaques (Figure 1) [5]. Some extracellular matrix (ECM) components, e.g., heparan sulfate proteoglycans (HSPG), are considered to promote extracellular formation of amyloid plaques [6]. In addition, a healthy body has the mechanism to remove Aβ peptides and balance their generation and clearance. For instance, matrix metalloproteinase 3 (MMP3), a collagen IV proteinase, could degrade Aβ components [7,8]. Also, a defective clearance of Aβ and an unbalance between Aβ generation and clearance have been demonstrated in SAD [9]. Thus, the peculiar generation and flawed removal of Aβ peptides cause Aβ accumulation. The more accumulations of Aβ peptides occur, the more senile plaques are formed in the human brain. . It is postulated that AD may be caused by the deposition of Aβ and tau hyperphosphorylation-derived neurofibrillary tangles (NFTs), both of which could activate the caspase-associated apoptosis. In AD brain the monomeric Aβ peptides aggregate to form toxic Aβ oligomers and further generate the insoluble fibrils, which ultimately form the plaques. The toxic Aβ species could trigger an inflammatory response and increase the level of ROS, which may cause neuron death. On the other hand, toxic Aβ species may be transferred into cells and trigger the apoptosis of neurons. NFTs could either cause neuronal cell death or trigger apoptosis of neurons. Aβ may interplay with tau-derived NFTs formation, while critical questions about Aβ-induced tau pathology in AD are still unanswered. ECM: extracellular matrix; APP: amyloid precursor protein; ROS: reactive oxygen species; SAT3: signal transducer and activator of transcription 3; JNK: c-Jun N-terminal kinase; GSK-3β: glycogen synthase kinase-3β.
The NFTs in AD are composed of paired helical filaments consisting of hyperphosphorylated tau (p-tau), a protein associated with the microtubules [10]. The normal form of tau is soluble, whereas it depicts strong ability to assemble and become insoluble under physiological conditions. The over . It is postulated that AD may be caused by the deposition of Aβ and tau hyperphosphorylation-derived neurofibrillary tangles (NFTs), both of which could activate the caspase-associated apoptosis. In AD brain the monomeric Aβ peptides aggregate to form toxic Aβ oligomers and further generate the insoluble fibrils, which ultimately form the plaques. The toxic Aβ species could trigger an inflammatory response and increase the level of ROS, which may cause neuron death. On the other hand, toxic Aβ species may be transferred into cells and trigger the apoptosis of neurons. NFTs could either cause neuronal cell death or trigger apoptosis of neurons. Aβ may interplay with tau-derived NFTs formation, while critical questions about Aβ-induced tau pathology in AD are still unanswered. ECM: extracellular matrix; APP: amyloid precursor protein; ROS: reactive oxygen species; SAT3: signal transducer and activator of transcription 3; JNK: c-Jun N-terminal kinase; GSK-3β: glycogen synthase kinase-3β. The NFTs in AD are composed of paired helical filaments consisting of hyperphosphorylated tau (p-tau), a protein associated with the microtubules [10]. The normal form of tau is soluble, whereas it depicts strong ability to assemble and become insoluble under physiological conditions. The over phosphorylation of tau contributes to its disassembly from microtubules and its aggregation to form insoluble fibers, that is, the NFTs. Protein kinases, such as mitogen-activated protein kinase (MAPK), glycogen synthase kinase-3 (GSK-3), and Jun N-terminal kinase/stress-activated protein kinases (JNK/SAPK), could upregulate phosphorylation of tau. Many studies support the notion that Aβ may interplay with tau-derived NFT formation, while critical questions about Aβ-induced tau pathology in AD are still unanswered. Overall, the perturbation of tau kinases and tau phosphatases leads to the abnormal hyperphosphorylation of tau, which results in NFT generation and toxicity in neurons.
The Aβ plaque and NFT formation show toxicity to neuronal cells. However, the causes of synaptic dysfunctions and selective loss of vulnerable neurons, particularly the subtypes, such as pyramidal, cholinergic, noradrenergic, and serotonergic neurons, remain elusive [11]. Studies show that Aβ plaques could trigger an inflammatory process by increasing the secretion of pro-inflammatory factors, e.g., IL-6 and TNF-α, which impair microglia and the surrounding neural cells, and finally result in neurodegeneration and neuron loss [12,13]. Based on the Aβ plaque deposition and cortical NFT formation, AD initiates when the connection between entorhinal cortex and hippocampus begins to disappear at the transentorhinal stage [5,14]. Other neuronal cells, such as GABAergic interneurons, show their function loss during the advanced stage of AD, i.e., the isocortical stage.

Current Challenges and the Demand for a Good AD Model
Although the pathophysiological changes of AD are characterized, it is unclear what factors induce the disease and how selective neuronal loss in AD is caused. Many hypotheses have been proposed for the causes of pathological characteristics [5,9]. These hypotheses include: (1) amyloid hypothesis, which posits that the gene mutations of APP and PS1/2 induce the overexpression or deficient removal of Aβ toxic species, and thus lead to the other two cardinal changes associated with AD ( Figure 1); (2) tau hypothesis, which believes that the hyperphosphorylated tau protein disassembles from microtubules and subsequently induces neuronal death ( Figure 1); (3) unknown triggering hypothesis, which postulates that some uncertain factors lead to the neurodegeneration, both directly and indirectly, through the formation of plaques and tangles. The last case seems to incorporate both the amyloid and tau hypotheses to interpret the etiology of AD. Although both amyloid and tau hypotheses could explain some of the pathological hallmarks and clinical observations about AD, there is still a controversy over which one occurs first. Thus, it is necessary to establish the disease models of AD, given that it is difficult to obtain patient brain samples.
To date, the modeling of AD for mechanistic and pathological studies mainly relies on non-human animal models and non-neuronal human cells. Non-neuronal cells usually lack the specific cellular structures of neurons and fail to recapitulate the signaling pathway and other physiology of neurons. Therefore, they cannot capture the three pathological changes in AD [15]. The transgenic animal models, usually the mice, have been developed for discovering genetic mutations in FAD [16][17][18]. To some extent, these models could interpret the pathogenicity of Aβ accumulation, plaque formation, and tauopathy. For instance, one study has developed a transgenic AD mouse model with a mutation in APP [17]. The model showed senile plaque formation and disconnection of synapses, but the presence of NFTs and neuronal loss was never found. Although animal models displayed the benefits for understanding AD neuropathology, they fail to reflect human brain anatomy, genetics, and the AD-associated neuron loss. More importantly, pharmacological testing and candidate drug target screening for AD using these animal models have shown no successful development for AD therapeutics till now [19]. Therefore, a more robust and suitable model that could capture all the three cardinal characteristics of AD is demanded. Recently, human adult neural stem cell-derived models through 3D culture in vitro, although limited by the cell source, were found to increase Aβ plaque depositions and tau phosphorylation [20,21]. In another study, Choi et al. reported a 3D thin-layer culture system by embedding human neural stem cells into Matrigel to model AD [22]. The overexpression of FAD mutations in APP and PS1 in embedded cells could induce both extracellular Aβ plaque and phosphorylated tau aggregation in the soma and neurites. These studies accurately paved the way for using human induced pluripotent stem cells (hiPSCs), a more reproducible and scalable cell type, to model AD in vitro.

Human Pluripotent Stem Cells for Modeling AD
Human PSCs (hPSCs) include human embryonic stem cells (ESCs) derived from blastocysts and hiPSCs reprogrammed from somatic cells. HPSCs show the properties of unlimited self-renewal and potent differentiation capacity. Theoretically, hPSCs could give rise to the three-germ layers, including ectoderm, endoderm, and mesoderm, and generate every cell type in the body (e.g., neurons, heart, pancreatic, and liver cells). HiPSCs were pioneered by Shinya Yamanaka's lab in 2006 that the introduction of four reprogramming factors including Oct4, Sox2, c-Myc, and Klf4 (also dubbed Yamanaka factors) could convert adult somatic cells into PSCs [23]. Using hiPSCs to model AD shows great advantages compared to other cell lines or model systems. First, patient-specific cell lines can be easily established just through collecting biopsy from patients (e.g., drops of blood or a piece of skin) and reprogramming the tissue to hiPSCs ( Figure 2). Therefore, the ethical controversy using human embryos for human ESCs can be avoided. For AD, the patient-specific hiPSCs loaded with gene mutations can easily recapitulate the mechanism of disease progression, especially for the early-onset type. Thus, the transgenic technology for specific genetic manipulation used in animal models becomes unnecessary. HiPSCs are also clinically advantageous since the use of autologous tissue ideally surpasses the patient's immune rejection. Moreover, hiPSCs with unlimited self-renewal capacity could help to provide large numbers of patient-specific neuronal cells for in vitro research and clinical applications. paved the way for using human induced pluripotent stem cells (hiPSCs), a more reproducible and scalable cell type, to model AD in vitro.

Human Pluripotent Stem Cells for Modeling AD
Human PSCs (hPSCs) include human embryonic stem cells (ESCs) derived from blastocysts and hiPSCs reprogrammed from somatic cells. HPSCs show the properties of unlimited self-renewal and potent differentiation capacity. Theoretically, hPSCs could give rise to the three-germ layers, including ectoderm, endoderm, and mesoderm, and generate every cell type in the body (e.g., neurons, heart, pancreatic, and liver cells). HiPSCs were pioneered by Shinya Yamanaka's lab in 2006 that the introduction of four reprogramming factors including Oct4, Sox2, c-Myc, and Klf4 (also dubbed Yamanaka factors) could convert adult somatic cells into PSCs [23]. Using hiPSCs to model AD shows great advantages compared to other cell lines or model systems. First, patient-specific cell lines can be easily established just through collecting biopsy from patients (e.g., drops of blood or a piece of skin) and reprogramming the tissue to hiPSCs ( Figure 2). Therefore, the ethical controversy using human embryos for human ESCs can be avoided. For AD, the patient-specific hiPSCs loaded with gene mutations can easily recapitulate the mechanism of disease progression, especially for the early-onset type. Thus, the transgenic technology for specific genetic manipulation used in animal models becomes unnecessary. HiPSCs are also clinically advantageous since the use of autologous tissue ideally surpasses the patient's immune rejection. Moreover, hiPSCs with unlimited selfrenewal capacity could help to provide large numbers of patient-specific neuronal cells for in vitro research and clinical applications.

AD Models Using hPSCs
Generally, three methods are utilized to establish AD pathological phenotype using hPSCs (summarized in Table 1). The first method is chemical induction, in which extrinsic chemicals, such

AD Models Using hPSCs
Generally, three methods are utilized to establish AD pathological phenotype using hPSCs (summarized in Table 1). The first method is chemical induction, in which extrinsic chemicals, such as Aβ42 oligomers and aftin5 (an Aβ42 inducer), are used to treat healthy hPSC-derived neural cells and induce AD-related phenotypes [24][25][26]. The addition of extrinsic chemicals to the normal hPSC-derived neurons could recapitulate some AD events, such as cytotoxicity of neuronal cells. However, the progression and some important AD pathologies, such as extracellular Aβ plaque formation, are difficult to recapitulate. The second method is to use genetic tools, such as lentivirus vectors or CRISPR-Cas9 gene-editing system, to overexpress AD-related genes, such as APP, PS1/2, and apolipoprotein E3/E4 (APOE3/E4), in hPSC-derived cells [27][28][29]. Using human ESC-derived neurons with recombinant human APOE2, APOE3, or APOE4, Huang et al. showed that all three APOE isoforms could upregulate APP and Aβ production but with different efficacy (APOE4 > APOE3 > APOE2) [29]. Park et al. co-cultured neurons and astrocytes derived from hPSCs overexpressing FAD-APP mutations with immortalized SV40 microglia in a microfluidics-based system to study neuro-inflammatory activity in AD [30]. For this approach, overexpression of mutant proteins is needed to induce AD phenotype.     The third method is to use AD patient-specific iPSCs carrying AD phenotype and mutations [42,59]. HiPSCs derived from familial and sporadic AD patients have been differentiated into various types of neuronal cells for studying specific AD pathologies ( Table 1). The patients generally had genetic mutations in APP, PS1, or PS2, in the case of FAD, and mutations in APOE3/E4 for SAD. Most of the hPSC-derived AD models used either 2D or embryoid body/neurosphere differentiation protocols to generate forebrain neurons, such as cortical glutamatergic neurons, GABAergic interneurons, or cholinergic neurons. The neuronal cells were characterized through gene and/or neuronal marker expression and tested for action potentials and calcium-handling ability for functional assessments. The AD-related pathology including elevated Aβ42 production and hyperphosphorylated tau was indicated in these models [71,72].
There are two main classes of mutations for familial AD: (1) mutations in APP; and (2) the mutations in PS1 or PS2 [73]. Aβ is a cleavage product of APP, so the dysfunction of APP processing can cause AD ( Figure 1). PS1/PS2 mediate the regulated proteolytic events of several proteins including γ-secretase [39], which plays an important role in Aβ generation. Mutations in PS1/PS2 also lead to Aβ plaque accumulation. Human iPSC lines were derived from familial (APP mutation) and sporadic AD patients and differentiated into cortical neurons (expressing TBR1 and SATB2) [37]. In FAD-derived cells, extracellular Aβ42 increased along with the decrease in intracellular Aβ2 compared to control neural cells. For SAD-derived cells, extracellular Aβ42 did not change while intracellular Aβ42 decreased compared to control of neural cells. Astrocytes also accumulated intracellular Aβ, which increased endoplasmic reticulum (ER) stress and reactive oxygen species (ROS). To improve ER stress and inhibit ROS generation, drugs including BSI, DHA, NSC23766, and dibenzoylmethane were evaluated. DHA decreased BiP (an HSP70 molecular chaperone located in the lumen of the ER) protein level, cleaved caspase 4 and peroxiredoxin-4, and decreased ROS generation. Differential response to DHA was observed for different patient-specific hiPSCs, indicating that DHA may be used for a subset of AD patients.
To evaluate if hiPSCs can recapitulate AD phenotype in the patients and understand the relationship between APP processing and tau phosphorylation, hiPSCs were derived from two familial (mutations in APP) and two sporadic AD patients [36]. Neurons from AD patients (two familial and one sporadic) had elevated levels of Aβ40, p-tau, and active glycogen synthase kinase-3β (aGSK-3β). β-secretase inhibitors, but not γ-secretase inhibitors, reduced p-tau and aGSK-3β levels. In AD, synaptic loss is one pathological observation, and the decreased synapsin I is usually observed in the affected human brain. In this study, no synaptic loss was observed and an extended culture period may be required to observe synaptic loss.
Neural progenitor cells (NPCs) derived from hiPSCs from PS1 FAD patients were also characterized [39]. The differentiation from hiPSCs used monolayer-based protocol with dual SMAD inhibition for 14 days. The derived NPCs displayed normal electrophysiology. However, an increased ratio of Aβ42/Aβ40 was observed in the conditioned medium compared to the control cell lines. Genetic profiling identified the genes that were differentially regulated due to PS1 mutations in order to find the molecules that might have a developmental role in FAD pathology. Three targets were found, including NLRP2, ASB2, and NDP. In particular, NDP expression decreased in the hippocampus of the late-onset AD brains. Moreover, synaptic dysfunction was observed in hiPSC-derived cortical neurons with autosomal dominant AD mutations or trisomy of chromosome 21 [58]. The released Aβ peptides from the FAD neurons with APP or PS1 mutations blocked hippocampal long-term potentiation (LTP), while neurons with trisomy 21 inhibited LTP through extracellular tau [58]. In another study, APOE4-expressing neurons showed elevated levels of tau phosphorylation, which was not related to the increased Aβ production, and the cells displayed GABAergic neuron degeneration [66].

Neural Tissue Patterning of hPSCs
The advancements using hPSCs to model AD started with the generation of neuronal cells or tissues that are directly affected by the disease through neural patterning (Figure 2). The key to success of this utilization is to efficiently generate specific neuronal subtypes or brain-like tissues from hPSCs. Neural tissue patterning is a complex process governed by intrinsic and extrinsic factors including morphogens and cell-ECM interactions. During the development of mammalian brain, early neural progenitors of the neural tube are derived through the inhibition of bone morphogenetic protein (BMP) and activin signaling. The cells are then specified into neuronal subtypes of each brain region along the rostral-caudal axis by tuning Wnt and retinoic acid (RA) signaling pathways, and along the dorsal-ventral axis mainly by the gradient of sonic hedgehog (SHH) proteins. Based on this knowledge, protocols utilizing the morphogens to mimic brain development have been established to produce brain region-specific neuronal subtypes derived from hPSCs ( Figure 3). For example, cortical glutamatergic neurons and GABAergic neurons from hPSCs were acquired by the modulation of SHH and fibroblast growth factor (FGF)-2 signals [25,74], while hindbrain/spinal motor neurons were efficiently derived through the activation of SHH, RA, and Wnt signaling pathways [75]. HPSC-derived spherical brain-like tissues, such as cerebral organoids with definable forebrain, midbrain, and hindbrain/spinal cord layers [76], and cortical spheroids with laminated regions [77] were generated by the default neural development without extrinsic factors. Interestingly, recent studies have obtained the whole spectrum of brain region-specific neural progenitors and neuronal subtypes from hiPSCs [78,79], which indicates the remarkable value of hiPSC-based models for the study and treatment of patient-specific neurological diseases. For AD, the Aβ deposition, tau tangles, and neuronal loss ultimately occur in the cerebral cortex, leading to the cortical symptoms related to language, attention, and visuospatial orientation. Therefore, studies modeling AD with hiPSCs often begin with the generation of forebrain cortical neurons. Differentiation to forebrain neuronal fates is based on the "default" induction strategies without exogenous patterning factors after dual-SMAD inhibition in a monolayer culture or embryoid body (EB)-based culture [77,80]. exogenous patterning factors after dual-SMAD inhibition in a monolayer culture or embryoid body (EB)-based culture [77,80]. Regional specification in neural differentiation of pluripotent stem cells (PSCs) mimicking in vivo regional patterning, and the use of the derived forebrain organoids for AD modeling. The regional patterning of brain organoids was achieved by cell signaling modulators and the corresponding organoids generated were used for disease modeling. DA: dopaminergic; FGF: fibroblast growth factor; RA: retinoic acid; SHH: Sonic Hedgehog; BDNF: Brain-derived neurotropic growth factor; SDF: Stromal cell-derived factor. Adapted and revised from [81].

A Novel Neural Patterning Method: Organoid Technology
The organoid systems derived from EB-based cultures emerge as a mixed cell population in a 3D platform. Relying on the intrinsic ability of self-organization of hiPSCs, and sometimes with the help of suitable exogenous factors (e.g., Matrigel), organoids recapitulate a large number of biological events in vivo [82,83]. Organoids are 3D spatial organization including heterogeneous tissue-specific cells, cell-cell interactions, and cell-ECM interactions, and exhibit certain physiological functions similar to tissues or organs in human body [84]. During organoid formation, ECM is one of the most important patterning factors that regulate and assist the self-organization and differentiation of stem cells within the organoids (Figure 4) [85]. ECMs, either secreted by stem cells or derived from artificial scaffolds, provide physical supports for cell attachment and organization and additional supplementation of signaling cues for cell growth and differentiation [86,87]. This organoid technology bridges a gap in the existing in vitro 3D culture systems by providing a robust approach for tissue patterning of stem cells and is more representative of in vivo situations [88][89][90]. 3. Regional specification in neural differentiation of pluripotent stem cells (PSCs) mimicking in vivo regional patterning, and the use of the derived forebrain organoids for AD modeling. The regional patterning of brain organoids was achieved by cell signaling modulators and the corresponding organoids generated were used for disease modeling. DA: dopaminergic; FGF: fibroblast growth factor; RA: retinoic acid; SHH: Sonic Hedgehog; BDNF: Brain-derived neurotropic growth factor; SDF: Stromal cell-derived factor. Adapted and revised from [81].

A Novel Neural Patterning Method: Organoid Technology
The organoid systems derived from EB-based cultures emerge as a mixed cell population in a 3D platform. Relying on the intrinsic ability of self-organization of hiPSCs, and sometimes with the help of suitable exogenous factors (e.g., Matrigel), organoids recapitulate a large number of biological events in vivo [82,83]. Organoids are 3D spatial organization including heterogeneous tissue-specific cells, cell-cell interactions, and cell-ECM interactions, and exhibit certain physiological functions similar to tissues or organs in human body [84]. During organoid formation, ECM is one of the most important patterning factors that regulate and assist the self-organization and differentiation of stem cells within the organoids (Figure 4) [85]. ECMs, either secreted by stem cells or derived from artificial scaffolds, provide physical supports for cell attachment and organization and additional supplementation of signaling cues for cell growth and differentiation [86,87]. This organoid technology bridges a gap in the existing in vitro 3D culture systems by providing a robust approach for tissue patterning of stem cells and is more representative of in vivo situations [88][89][90]. The first generation of neural rosettes from human ESCs was derived in 2001 by forming spontaneously differentiated EBs and then plating EBs on coated dishes for neuroepithelial cluster formation [91]. Combining the EB-based rosette patterning and serum-free culture, the so-called SFEBq approach, i.e., serum-free floating culture of EB-like aggregates with quick re-aggregation, was developed and could generate remarkably large rosettes with lumens and apicobasal structures from hPSCs in 2008 [92]. In 2011, the self-organizing optic cups with retinal specifications from hPSCs were derived using the entirely 3D EB-based neural culture [93]. This study indicates that neural tissues with histologically accurate architecture could be patterned only in 3D floating culture of hPSCs. All these studies paved the road for the advent of a new neural patterning technology, that is, organoid. In 2013, Lancaster et al. derived cerebral organoids with a broad brain regionalization to model microcephaly using EB-based culture with Matrigel embedding and agitation [76]. The organoids could reach up to 4 mm in diameter with fluid-filled cavities that resembled ventricles rather than the small neural-tube-like lumens found in rosettes. After that, the organoid-based neural patterning methods of hPSCs have been utilized for mimicking human brain development, modeling neurological disease in vitro, and testing potential drug candidates [81].
Using the SFEBq method, multilayer structures were derived from hiPSCs with lower and upper cortical layer fates in dorsal forebrain [94]. The 3D aggregates can generate inhibitory neurons and display neuronal connectivity (expression of synapsin I and PSD95). Genomics study showed the correlation to the brain development at postconceptional weeks 4-10. Another 3D culture was used to generate laminated cerebral cortex-like structure from hPSCs [77]. The cells were induced for forebrain fate using dual SMAD inhibition for six days, followed by FGF-2/epidermal growth factor treatment for 19 days, and the maturation with brain-derived neurotrophic factor (BDNF) and NT3. At day 18, 85% PAX6+ cells and >80% of FOXG1+ cells were observed. The organoid size increased from 300 µm The first generation of neural rosettes from human ESCs was derived in 2001 by forming spontaneously differentiated EBs and then plating EBs on coated dishes for neuroepithelial cluster formation [91]. Combining the EB-based rosette patterning and serum-free culture, the so-called SFEBq approach, i.e., serum-free floating culture of EB-like aggregates with quick re-aggregation, was developed and could generate remarkably large rosettes with lumens and apicobasal structures from hPSCs in 2008 [92]. In 2011, the self-organizing optic cups with retinal specifications from hPSCs were derived using the entirely 3D EB-based neural culture [93]. This study indicates that neural tissues with histologically accurate architecture could be patterned only in 3D floating culture of hPSCs. All these studies paved the road for the advent of a new neural patterning technology, that is, organoid. In 2013, Lancaster et al. derived cerebral organoids with a broad brain regionalization to model microcephaly using EB-based culture with Matrigel embedding and agitation [76]. The organoids could reach up to 4 mm in diameter with fluid-filled cavities that resembled ventricles rather than the small neural-tube-like lumens found in rosettes. After that, the organoid-based neural patterning methods of hPSCs have been utilized for mimicking human brain development, modeling neurological disease in vitro, and testing potential drug candidates [81].
Using the SFEBq method, multilayer structures were derived from hiPSCs with lower and upper cortical layer fates in dorsal forebrain [94]. The 3D aggregates can generate inhibitory neurons and display neuronal connectivity (expression of synapsin I and PSD95). Genomics study showed the correlation to the brain development at postconceptional weeks 4-10. Another 3D culture was used to generate laminated cerebral cortex-like structure from hPSCs [77]. The cells were induced for forebrain fate using dual SMAD inhibition for six days, followed by FGF-2/epidermal growth factor treatment for 19 days, and the maturation with brain-derived neurotrophic factor (BDNF) and NT3.
At day 18, 85% PAX6+ cells and >80% of FOXG1+ cells were observed. The organoid size increased from 300 µm at two weeks to 4 mm at 2.5 months. The formed human cortical spheroids corresponded to the late mid-fetal periods. The spheroids had a distinct ventricular zone and sub-ventricular zone. The cells also formed superficial cortical layers (expressing BRN2 and SATB2 for layer II-IV, emerged at day 76) and deep cortical layers (expressing CTIP2-layer V and TBR1-layer VI, showed up at day 52). The presence of GFAP+ cells showed the astrogenesis (3-8% cells). Functional characterizations include: (1) calcium imaging for spontaneous calcium spikes; (2) Na + /K + currents; and (3) the firing of action potentials. Slices of spheroids showed spontaneous synaptic activity, which can be reduced by a glutamate receptor blocker. This study demonstrated the patterning and specification of different neuronal and glial cell types which can be used for large-scale drug screening.
The organoid technology is a promising method to mimic human brain development and could generate brain-like tissues for modeling neurological diseases, including autosomal recessive primary microcephaly [95], Zika virus infection [96], autism spectrum disorder [97], Timothy syndrome [98], brain tumors [99,100], and others (see the review by Amin et al., 2018 [101]). Lots of studies have shown the feasibility of the organoid methods to model AD pathogenesis [24,52,60,61]. Our previous studies found that dynamic culture of cerebral organoids promoted cortical layer structure compared to static culture, and the 3D brain organoids from hiPSCs with PS1 M146V mutation could recapitulate some AD-related phenotype, such as Aβ secretion and neuron death [59,102]. By using hiPSCs with APOE4/E3 mutations, Lin et al. showed that APOE4 forebrain organoids displayed the increased Aβ aggregation and hyperphosphorylated tau [61]. Taken together, 3D hPSC-derived organoids provide structured neural tissues and specific microenvironments that can be used to model AD-associated pathology.

Effects of ECMs on Neural Patterning of hPSCs
The central nervous system (CNS) is characterized by a functional network of neurons, glia, and their secreted ECMs. ECMs constitute the essential physical structures for neural cells and provide diverse biochemical signals for neurogenesis, neural cell migration and differentiation, and synaptic plasticity. ECMs also play an important role in neural tissue patterning of hPSCs. Studies performed with decellularized ECMs from different PSC derivatives indicated that ECMs guided PSC differentiation into the cell types residing in the tissue from which the ECMs were derived [103,104]. The ECM-stem cell interactions are mainly mediated by integrins, a large family of heterodimeric transmembrane receptors connecting with an intracellular cytoskeleton. Thus, cell migration, organization, and differentiation could be regulated through ECM-integrin-cytoskeleton connections. For instance, the binding between α6β1 integrin and laminin is the key player for neural stem cell adhesion to the vascular structures [105]. ECMs may also control stem cell fate decisions through the modulation of intracellular signaling. For example, our previous study demonstrated that undifferentiated PSC-derived ECMs could upregulate Wnt/β-catenin signaling, while NPC-derived ECMs promoted neural patterning of PSCs through the inhibition of Wnt signaling pathway [106].
The biophysical properties, such as elastic modulus, geometry, and Poisson's ratio of matrix, also impact growth, proliferation, and neural differentiation of PSCs [107,108]. Stem cells cultured on hydrogels with varied stiffness indicated that substrate elastic modulus can alter critical cellular events, such as ECM assembly, cell motility, and cell spreading [109][110][111]. For neural patterning, a large number of studies showed the important effects of matrix stiffness or elastic modulus (Table 2) [111][112][113][114][115][116][117][118]. Saha et al. first showed that soft substrates supported neuron differentiation from NPCs, while a higher elastic modulus (E~1 to 10 kPa) promoted glial cell generation [111]. Leipzig et al. further demonstrated that substrates with Young's modulus (E) below 1 kPa favored neuronal differentiation, those in the range of 1 and 3.5 kPa supported astrocytes, and above 7 kPa could enhance oligodendrocyte derivation [112]. By culturing hPSCs on Matrigel-coated polyacrylamide substrates, a very soft matrix (E~0.1 kPa) was found to support early neural differentiation of hPSCs [119]. Normally, cells sense elasticity during the attachment on the substrate through focal adhesions and formation of stress fibers. Their responses to the matrix properties rely on myosin-directed contraction and cell-ECM adhesions, which involve integrins, cadherins, and other adhesion molecules [120]. The Poisson's ratio is another important biophysical cue that influences stem cell behaviors, as the nuclei of ESCs exhibit a negative Poisson's ratio in the pluripotent-state [121]. Our previous work found that Poisson's ratio of matrix could confound the effects of elastic modulus on PSC neural differentiation [108]. In conclusion, ECMs serve as a reservoir of biochemical and biophysical factors that impact stem cell growth, organization, and differentiation.   Note: PA, polyacrylamide; vmIPNs, variable moduli interpenetrating polymer networks; MAC, methacrylamide chitosan; PDMS, polydimethylsiloxane; HA, hyaluronic acid; PLLA, poly-L-lactide acid; PLGA: poly(lactic co-glycolic acid); PCL, polycaprolactone; PEGDA, polyethylene glycol diacrylate; PET, polyrethylene terephthalate; PEKK, poly(etherketoneketone); PCU, polycarbonate-urethane; GAG, glycosaminoglycan; PEG, polyethylene glycol; PU, polyurethane; PDGFR, platelet-derived growth factor receptor beta (PDGFRβ) signaling; EMT, epithelial-mesenchymal-transition.

Proteoglycans in the ECMs of AD Brain
The brain ECMs express low amounts of fibrous proteins, such as collagen and laminin, but high amounts of glycosaminoglycans (GAG) compared to other tissue types. The ECMs in the brain include glycoproteins and proteoglycans originated from neurons or glia. Proteoglycans (PGs) are a group of glycoproteins that carry covalently bound sulfated GAG chains which play important roles in cell differentiation, tissue morphogenesis, and phenotypic stabilization via cell-matrix adhesiveness and/or binding to growth regulators. GAGs are composed of repeating disaccharide units attached to a core protein through serine residue and carbohydrate linkage regions. Depending on the disaccharide structures, PGs can be divided into chondroitin/dermatan sulfate, heparin/heparan sulfate (HS), and keratan sulfate side chains. In the CNS, the majority of proteoglycans have either heparin sulfate or chondroitin sulfate side chains. PGs are expressed at different stages of human brain development and are produced non-homogenously by different neural cell types [144].

Chondroitin Sulfate Proteoglycans (CSPGs) in Brain Development
CSPG family consists of four members: (1) lecticans, a family including aggrecan, versican, neurocan, and brevican; (2) phosphacan; (3) small leucine-rich proteoglycans, e.g., decorin and biglycan; (4) other CSPGs including neuroglycan-C and NG2 [145]. Distribution and expression of these CSPGs vary during human brain development. Aggrecan and versican are distributed in ECMs of various non-neuronal tissues as well as neural tissues, whereas neurocan and brevican appear to show a neural tissue-specific distribution. Versicans are found to express in the post-natal development of brains. Brevican is expressed predominantly by astrocytes, not neurons. NG2, a large transmembrane CSPG, is expressed by oligodendrocyte progenitor cells in CNS. Neurocan, a secretory CSPG, is reported to be synthesized mainly by neuronal cells.
The CSPGs are either secreted as ECMs or inserted to the plasma membrane of the cells. The attachment of CSPGs to the cell membrane is either through a collagenous ligand or by core proteins and cell surface-associated CSPGs such as aggrecan. The concentrations of aggrecan and brevican increase during brain development. The concentration of neurocan also increases with development to reach a peak in the early postnatal period and declines thereafter. The full-length neurocan is the major variant in pre-and early postnatal brains, while it is hardly detectable in the mature brain. Phosphacan and receptor-type protein tyrosine phosphatase are produced by both glial cells and neuronal cells. Variations in sulfate content and localization on the same GAG chains or the same core protein are tissue-specific or cell-specific. The expression of specific GAG biosynthetic and modifying enzymes by individual neural cell types dictates the type of carbohydrate modification of the core proteins expressed in each cell type [146].

Heparin/HSPG in Brain Development
Heparan sulfate proteoglycans are heavily glycosylated proteins, in which some HS and GAG chains are covalently attached to a core protein, which can be either surface proteins (e.g., syndecans/glypicans) or secreted proteins (e.g., agrin/collagen type XVIII/perlecan) [147]. N-Syndecan (syndecan-3), a transmembrane HSPG, has the capacity to bind to heparin-binding growth factors with a neurite-promoting activity, such as FGF-2, pleiotrophin/HBGAM, and midkine. This matrix-growth factor interaction is found to be the basis for the axonal development of neural cells. Also, N-syndecan works as a receptor molecule for pleiotrophin, which then enhances the phosphorylation of intracellular cytoskeleton-regulating molecules, such as cortactin and fyn-kinase. It was found that the association of cortactin and fyn-kinase with N-syndecan was increased after induction of long-lasting synaptic sensitivity called long-term potentiation.
N-syndecan regulates the neuronal activity-dependent connectivity through the fyn signaling pathway. N-syndecan, together with syndecan-2, also appears to be involved in the maturation of synapses. Syndecan-2 interacts with calcium/calmodulin-dependent serine protein kinase, a PDZ family protein that induces the formation of mature dendritic spines, and supports the development of postsynaptic specialization [148]. Heparin is one of the glycosaminoglycans that bind to different proteins and thereby promote various neural functions, such as preventing blood clotting, neuronal communication, and brain development.

Impacts of CSPG on AD Pathology
The motor and sensory areas of the cortical brain, which show cytoskeletal changes during AD development, were found to express prominent levels of CSPGs. These results suggested the importance of CSPGs on AD development, e.g., synaptic loss. Numerous molecular events occur in the post-synaptic density of an excitatory synapse in response to a train of pre-synaptic action potentials. One of these factors is the presence of peri-synaptic ECMs, a lattice of chondroitin sulfate-bearing proteoglycans, termed lecticans (e.g., aggrecan, brevican, neurocan, and versican), bound to hyaluronic acid near their N-terminus and tenascin-R near their C-terminus. Alterations in brain ECMs occur early in AD progression. The synaptic loss is observed prior to neuronal cell death, and the loss of synapses in the outer molecular layer of the hippocampal dentate gyrus is more highly correlated with cognitive impairment than other classical AD pathology including NFTs and senile plaques. ECMs can respond to network activity by incorporating secreted molecules, by shedding extracellular domains of transmembrane molecules, or by freeing products of activity-dependent proteolytic cleavage as signaling messengers.
Lecticans, also known as hyalectans, have been studied for their influence on the plasticity of synapses because of the elevated expression in AD patients. Lecticans are found to contribute to diminished synaptic plasticity. Involvement of CSPGs in hippocampal synaptic plasticity was confirmed using the enzyme that specifically digests chondroitin sulfates, chondroitinase ABC (ChABC) [149]. Injection of ChABC in young AD mice elevated lectican levels and resulted in a reversal of contextual fear memory deficits and restoration to normal long-term potentiation. ChABC treatment results in the removal of CS chains from lectican core proteins (and other CS bearing PGs). These studies demonstrated that removal of CS chains stimulated neural plasticity. Increasing evidence has demonstrated that CS-bearing ECM molecules increase with age and are associated with AD. CS-bearing PGs bind to Aβ, but it remains unknown whether Aβ binds to brain-derived lecticans. Injection of ChABC increased synaptic density surrounding plaques and reduced Aβ burden in the s/m of the hippocampus [150].

Impacts of Heparin/HSPGs on AD Pathology
HSPGs are found to grow concomitantly with amyloid filament formation. HSPGs have a negative charge and thereby concentrate on the ligands and protect them from proteolysis while making insoluble complexes [151]. Heparin and HSPGs regulate the activity of a number of proteases involved in AD. HSPGs promote Aβ aggregation, stabilize Aβ fibrils, and inhibit Aβ degradation. Perlecan and agrin are found to be the important HSPGs that are involved in the pathogenesis and localized with Aβ deposits in AD brain. Perlecan accelerates Aβ fibril formation and maintains the Aβ fibril stability by using Perlecan's HS-GAG chains [152]. Agrin also binds to Aβ and accelerates fibril formation and protects fibrillar Aβ from proteolysis ( Figure 5) [153]. HSPGs bind Aβ with high affinity and promote intracellular Aβ uptake in multiple cell types and thereby increase the cytotoxicity [154].
correlated with cognitive impairment than other classical AD pathology including NFTs and senile plaques. ECMs can respond to network activity by incorporating secreted molecules, by shedding extracellular domains of transmembrane molecules, or by freeing products of activity-dependent proteolytic cleavage as signaling messengers.
Lecticans, also known as hyalectans, have been studied for their influence on the plasticity of synapses because of the elevated expression in AD patients. Lecticans are found to contribute to diminished synaptic plasticity. Involvement of CSPGs in hippocampal synaptic plasticity was confirmed using the enzyme that specifically digests chondroitin sulfates, chondroitinase ABC (ChABC) [149]. Injection of ChABC in young AD mice elevated lectican levels and resulted in a reversal of contextual fear memory deficits and restoration to normal long-term potentiation. ChABC treatment results in the removal of CS chains from lectican core proteins (and other CS bearing PGs). These studies demonstrated that removal of CS chains stimulated neural plasticity. Increasing evidence has demonstrated that CS-bearing ECM molecules increase with age and are associated with AD. CS-bearing PGs bind to Aβ, but it remains unknown whether Aβ binds to brain-derived lecticans. Injection of ChABC increased synaptic density surrounding plaques and reduced Aβ burden in the s/m of the hippocampus [150].

Impacts of Heparin/HSPGs on AD Pathology
HSPGs are found to grow concomitantly with amyloid filament formation. HSPGs have a negative charge and thereby concentrate on the ligands and protect them from proteolysis while making insoluble complexes [151]. Heparin and HSPGs regulate the activity of a number of proteases involved in AD. HSPGs promote Aβ aggregation, stabilize Aβ fibrils, and inhibit Aβ degradation. Perlecan and agrin are found to be the important HSPGs that are involved in the pathogenesis and localized with Aβ deposits in AD brain. Perlecan accelerates Aβ fibril formation and maintains the Aβ fibril stability by using Perlecan's HS-GAG chains [152]. Agrin also binds to Aβ and accelerates fibril formation and protects fibrillar Aβ from proteolysis ( Figure 5) [153]. HSPGs bind Aβ with high affinity and promote intracellular Aβ uptake in multiple cell types and thereby increase the cytotoxicity [154].  HSPGs are also the target of Aβ-induced oxidative stress production. Studies propose that minimal chain length is a prerequisite for efficient fibril polymerization and deposition. Fragmentation of HS using heparinase III was found to inhibit the polymerization and deposition of amyloid peptides [155]. Heparinase is a glucuronidase that specifically cleaves heparan sulfate, thereby preventing Aβ-HSPG interactions. The formed HS oligosaccharides then bind amyloid monomers to prevent them from polymerizing and assembling into larger aggregates. Treatment of cells with heparinase III also reduced the HSPG-based Aβ-induced ROS production.
HSPGs modulate the Aβ-associated neuroinflammation and Aβ clearance from the human brain. Several in vitro studies indicated the interaction of Aβ with GAGs including HS and heparin, a HS analogue with a higher sulfation degree [156]. All those studies describe that the N-terminus of Aβ is a HS-binding motif and this sequence supports the interaction between Aβ and HSPGs [157]. Studies also suggest that the Aβ peptides compete with FGF-2 for HS binding site. The interaction depends on several factors, such as the sulfation pattern, chain length of the GAGs, and the pH. The higher sulfation in heparin boosted the affinity and the degree of Aβ-HS complex formation was higher compared to Aβ-HSPG complex. Lower binding of heparin fragments shorter than 6-sugar units to Aβ also suggested the requirement of GAG chain length for the interaction [158]. It is also proposed that the Aβ-HS interaction is mutually protective because HS is protected from heparanase degradation and Aβ is protected from protease degradation. Cellular degradation of HSPGs using heparanase (an endo-β-glucuronidase that specifically cleaves HS side chains of HSPGs) decreased the cytotoxicity of Aβ peptides on the cells [154]. Aβ clearance from the cells is performed in different ways, including degradation by proteolytic enzymes or receptor-mediated Aβ transport across the blood-brain barrier, in which the main receptor is low-density lipoprotein receptor-related protein-1 (LRP-1) [159,160]. The complex interactions of Aβ precursor protein ApoE, LRP-1, and HSPG facilitate Aβ internalization. Recent studies found that HSPGs and LRP-1 mediate Aβ internalization in a seemingly cooperative manner, in which HSPGs regulate Aβ binding to cell surface through LRP-1 [159]. These findings suggest that cell surface HS mediates Aβ internalization and toxicity. Tau binding and internalization also depend on HSPGs, which are critical mediators of tau fibril internalization. Tau binding to HSPGs is required for transcellular propagation [161]. Further understanding of the interactions between ECM molecules and Aβ peptides and their effects on neuronal differentiation should help to develop new treatment methods for AD.

Heparin-Based Therapy for Neural Degeneration
Aβ-related activation and the followed neuroinflammation are dependent on the region 1-11 of the peptide [162]. Related studies of the molecules that have the ability to pharmacologically target this region indicate the possible therapeutic effects of heparin [163]. In vitro treatment of neurons with heparin reduced Aβ-associated cytotoxicity, revealing the capability of heparin on AD treatment. Heparin, which has been widely used as an anticoagulant in clinical use, has the capability to bind to the region 12-17 of Aβ (HHQK) [157], near to the contact and complement activation residues. The heparin-binding regions of these proteins are characterized by clusters of arginines and lysines. These clusters form centers of highly positive charge density that electrostatically interact with the acidic groups of heparin.
The binding of heparin to the regions of Aβ occurs as electrostatic interactions between cationic sites of Aβ and anionic sulfate residues of heparin. This causes steric changes in the Aβ and influences the biological function, which in turn reduces the inflammation of neural cells by inhibiting Aβ complement and contact systems and thereby reduces the neurodegeneration. Complement and contact systems are related inflammatory cascades and are dependent on the reciprocal activation. These systems share a major inhibitor factor (C1 INH) with each other (Figure 6), which is downregulated in AD brain. Inhibition of complement activation by heparin might also be due to the potentiation of C1-1NH. C1-1NH attaches to the enzymes and works as an inhibitor of spontaneous activation of the complementary system. The addition of low molecular weight heparin reduced plaques and Aβ accumulation in a mouse AD model [164]. Association between heparin and the Aβ peptides was found to be pH-dependent. Heparin-binding to the peptides reduced the concentration of the peptides and the expression of kinnogen.
Biologically active beta-site amyloid precursor protein cleaving enzyme 1 (BACE1) interacts directly with HS/heparin and has a high relative affinity for these polysaccharides. It has been reported that heparin inhibits BACE1 activity in vitro and thereby regulates the production of Aβ peptides [163,165]. It is also suggested that Aβ is a metalloprotein that expresses both high and low metal affinity sites. The AD patients are characterized by elevated levels of copper and zinc in the hippocampal region. These metals have been found to accelerate the accumulation of Aβ and promote the ROS release to the cells. Cu 2+ acts as a bridge between anionic groups on HSPGs and cationic groups of the NFTs. It was also found that free carboxyl groups are required for HSPG binding to NFTs. Heparin can bind to the extracellular superoxide dismutase, a homotetrameric Cu 2+ and Zn 2+ containing glycoprotein, and regulate the activity of these metals which reduce accumulation of Aβ peptides ( Figure 6). peptides was found to be pH-dependent. Heparin-binding to the peptides reduced the concentration of the peptides and the expression of kinnogen. Biologically active beta-site amyloid precursor protein cleaving enzyme 1 (BACE1) interacts directly with HS/heparin and has a high relative affinity for these polysaccharides. It has been reported that heparin inhibits BACE1 activity in vitro and thereby regulates the production of Aβ peptides [163,165]. It is also suggested that Aβ is a metalloprotein that expresses both high and low metal affinity sites. The AD patients are characterized by elevated levels of copper and zinc in the hippocampal region. These metals have been found to accelerate the accumulation of Aβ and promote the ROS release to the cells. Cu 2+ acts as a bridge between anionic groups on HSPGs and cationic groups of the NFTs. It was also found that free carboxyl groups are required for HSPG binding to NFTs. Heparin can bind to the extracellular superoxide dismutase, a homotetrameric Cu 2+ and Zn 2+ containing glycoprotein, and regulate the activity of these metals which reduce accumulation of Aβ peptides ( Figure 6).

Studying ECM Effects in hiPSC-Derived Forebrain Organoids
Combining the knowledge of forebrain organoids and the ECM development in AD pathology, our studies used heparin (competes for Aβ affinity with HSPG), heparinase III (digests HSPGs), chondroitinase (digests CSPGs), and hyaluronic acid (HA) to treat the cortical and hippocampal forebrain spheroids/organoids exposed to Aβ42 oligomers [166]. Briefly, cortical spheroids were generated using dual SMAD inhibition followed with FGF-2 and cyclopamine (an Shh inhibitor). In the case of hippocampal differentiation, the hiPSCs were treated with dual SMAD inhibitors, cyclopamine, and IWP4 (a Wnt inhibitor), followed by treatment with CHIR99021 (a Wnt activator) and BDNF. Exogenous addition of Aβ42 oligomers was found to reduce the expression of β-tubulin III, suggesting the associated neuronal cell death. The treatments of ECM-related molecules with Aβ42 oligomers

Studying ECM Effects in hiPSC-Derived Forebrain Organoids
Combining the knowledge of forebrain organoids and the ECM development in AD pathology, our studies used heparin (competes for Aβ affinity with HSPG), heparinase III (digests HSPGs), chondroitinase (digests CSPGs), and hyaluronic acid (HA) to treat the cortical and hippocampal forebrain spheroids/organoids exposed to Aβ42 oligomers [166]. Briefly, cortical spheroids were generated using dual SMAD inhibition followed with FGF-2 and cyclopamine (an Shh inhibitor). In the case of hippocampal differentiation, the hiPSCs were treated with dual SMAD inhibitors, cyclopamine, and IWP4 (a Wnt inhibitor), followed by treatment with CHIR99021 (a Wnt activator) and BDNF. Exogenous addition of Aβ42 oligomers was found to reduce the expression of β-tubulin III, suggesting the associated neuronal cell death. The treatments of ECM-related molecules with Aβ42 oligomers promoted β-tubulin III expression, indicating the protective effect of these ECM related molecules. Further analysis on the forebrain and hippocampal markers revealed that ECM enzymes were capable of rescuing TBR1+ cells in cortical forebrain spheroids and PROX1+ neurons in the hippocampal spheroids in the presence of Aβ42 oligomers. These results suggest the necessity of further evaluation of various brain ECMs on forebrain spheroid patterning and cell survival and the related neurological disorders to derive better therapeutic interventions. In addition, heparin-conjugated HA hydrogels were synthesized and characterized [107]. A comparative study with heparin-HA hydrogels with varying modulus reveals that the lower modulus (300-400 Pa) hydrogel supports the forebrain fate whereas the higher modulus (1000-1300 Pa) hydrogel supports hindbrain fate of hiPSCs. This patterning effects may be regulated through modulation of the Hippo/YAP signaling.

Conclusions and Perspectives
Recently, the findings of adult neurogenesis in humans have sharply declined, which raises questions about animal models to study human central nervous system [167]. The emerging hPSC technology and organoid methods represent promising opportunities to investigate the development of human brain and neurological diseases. These stem cell-based organoid systems have recapitulated some biological events, such as temporal and spatial organization of brain tissues, neuronal-glial cell interactions, and cell-matrix interactions. However, these models need to show promise to reveal new features of normal and pathological phenomena of neurogenesis. One aspect of the interests is that the patient-derived hPSCs with mutations, such as APP or PS1/PS2, could be used to study the roles of specific genes in FAD development and progression. Another aspect is that organoids provide a more advanced in vitro tool to investigate complex niche interactions, such as how HSPGs are involved in Aβ pathogenesis. While the organoid models have great advantages, they are also challenged by the low homogeneity and reproducibility. The brain organoids may be different from each other not only in size but also structure, which makes it difficult to control the quality and predict the readouts. Future improvements, such as novel biomaterials combined with new culture system design (e.g., microfluidics), may be utilized to better control neural patterning in a more accurate and predictable way.