Brain organoids: building higher-order complexity and neural circuitry models

Abstract

Brain organoids are 3D tissue models of the human brain that are derived from pluripotent stem cells (PSCs). They have enabled studies that were previously stymied by the inaccessibility of human brain tissue or the limitations of mouse models of some brain diseases. Despite their enormous potential, brain organoids have had significant limitations that prevented them from recapitulating the full complexity of the human brain and reduced their utility in disease studies. We describe recent progress in addressing these limitations, especially building complex organoids that recapitulate the interactions between multiple brain regions, and reconstructing in vitro the neural circuitry present in vivo. These major advances in the human brain organoid technology will remarkably facilitate brain disease modeling and neuroscience research.

Brain organoids – simplicity drives complexity

Understanding the molecular mechanisms of brain disorders requires extensive studies in animal and human models. The significant differences between brain development in humans and rodents, and the drawbacks of some mouse models of brain disease, necessitate the use of human brain models. The difficulty of accessing human brain tissue, and the fact that little research can be done on post-mortem tissue, have urged the field to develop in vitro models of the human brain. PSCs provide an unprecedented opportunity for bioengineering human brain cells in 2D and 3D cultures. Brain organoids are self-organized 3D cellular systems which include different brain cell types and can generate structures with functional features that resemble those of different brain regions [1]. While 2D cultures have enabled the study of particular cell types of the human brain in isolation, organoids can recapitulate more of the structural complexity of the human brain, such as the cellular diversity and representation of multiple cell types and layers that work together to govern brain development and function [2,3], and can recapitulate the gene expression profiles [4], epigenetic landscapes [5], and some synaptic functions [6] of the human brain.

After the initial success of building simple ‘generic’ organoids that recapitulate the features of several mixed brain regions, important questions (and aspirations) arose: whether organoids that recapitulate specific brain regions can be developed, and whether the inter-region circuitry observed in vivo can be reconstructed in this in vitro model. Moreover, significant efforts have been invested over recent years to eliminate other limitations of the organoid models, such as by developing organoid culture conditions and organoids that can be used to study aging and aging-related disorders. In this review we therefore focus on these important aspects which have not been covered in other recent reviews: namely the aspects of organoid models and culture technologies that allow recapitulation of neural circuitry and electrical activity recordings, and how the major limitations of organoids can be overcome to maximize their potential and applications in brain disease studies, including those relevant to aging and neurodegenerative disorders.

Building higher-order complexity in brain organoids

Brain regions

Multiple simple forms of brain organoids have been generated over the past >10 years; however, these have suffered from two cardinal limitations. First, brain/cerebral organoids spontaneously differentiated from PSCs recapitulate a rudimentary brain-like tissue encompassing several mixed and unorganized brain regions that are not well demarcated [2]. Efforts then ensued to develop brain region-specific organoids that recapitulate specific and well-characterized brain regions. These brain region-specific organoids were developed via recapitulating in vitro the patterning cues and morphogenetic signals that drive the neurodevelopment of each brain region in vivo (Figure 1) [7,8]. For example, neural induction cues in organoids encompassed inducing the neuroectodermal fate by dual SMAD inhibition (BMP inhibition + TGF-β inhibition) during the first few days of induction [8–10], followed by directing the organoids towards the cortical fate by using WNT and TGF-β inhibition [3], or a striatal fate with lateral ganglionic eminence (LGE) (see Glossary) cells using WNT inhibition, TGF-β, RXR, and SHH stimulation [11,12]. Subpallium organoids with medial ganglionic eminence (MGE) characteristics were developed using WNT inhibition plus FGF-β, EGF, and SHH stimulation [13]. Moreover, thalamic organoids were generated by using dual SMAD inhibition followed by treatment with BMP7, insulin, and MEK inhibitor [14]. While activation of the WNT pathway, plus either BMP pathway activation or SHH pathway inhibition led to hippocampal organoid generation [10,15], modulation of WNT pathway combined with the BMP inhibition, SHH activation and FGF-β led to hypothalamic specification [7,10]. With regards to midbrain organoid specification, multiple protocols have been devised which utilized BMP and TGF-β inhibitors along with WNT, SHH, and FGF-8 pathway activators [7,9,16,17]. Furthermore, organoids recapitulating hindbrain and spinal cord have been developed via caudal fate patterning using retinoic acid (RA) and/or FGF pathway stimulators [18–24]. In addition to these patterning cues and signaling pathway modulators, neurotrophic factors such as brain-derived neurotrophic factor (BDNF) and glial cell-derived neurotrophic factor (GDNF) were often added to the media to enhance the maturation of organoids [7,17,20]. In addition to directly adding morphogens to the organoid culture, a recently developed approach involved exposing the PSCs or embryoid bodies simultaneously to multiple orthogonal morphogen gradients in a microfluidic device. This enabled more controllable chemical gradients and generated neural tube or different brain region organoids at the same time [25,26]. Collectively, controlling fate determination cues and signal transduction pathways at the right time during the process of organoid development has enabled the generation of a large array of region-specific brain organoids.

Figure 1. Organoids recapitulating specific brain regions.

(A) Signaling pathways during neural tube development. Along the dorsal–ventral axis, SHH signaling drives ventral fate whereas WNT and BMP pathway activation favors dorsalization. In the anterior–posterior axis, low WNT concentration promotes forebrain specification whereas a higher WNT concentration leads to caudal fate. FGFs and retinoic acid (RA) also participate in the caudal fate induction. (B) Generation of region-specific brain organoids and the necessary patterning cues. Forebrain, midbrain, hindbrain, and spinal cord organoids can be induced by modifying the signaling pathways, as indicated in the corresponding frames. In the frames, ‘,’ indicates ‘and’, whereas ‘/’ indicates ‘or’. Abbreviations: BMPi, bone morphogenetic protein inhibitor; FGF-β, fibroblast growth factor β; LGE, lateral ganglionic eminence; MEKi, mitogen-activated protein kinase inhibitor; RXR, retinoid X receptor; SDF1, stromal cell-derived factor 1; SHHa, sonic hedgehog pathway activator; TGF-βi, transforming growth factor β inhibitor; WNTa, wingless/integrated pathway activator; WNTi, wingless/integrated pathway inhibitor. Figure created with BioRender.

Neural circuitry

The second limitation of previous brain organoid models was the inability to recapitulate the interactions that occur in vivo between multiple brain regions, including signal transduction cascades and cell–cell crosstalk, cellular migration, and importantly the neural circuitry that develops between two or more brain regions. A pioneering solution to this problem was to fuse individually developed brain region-specific organoids to generate ‘multi-region assembloids’ (Figure 2) [12,13,27]. The assembloids were fused by embedding individual organoids or spheroids together in a Matrigel droplet or by coculturing them in a microcentrifuge tube or plate for several days. For example, forebrain assembloids were developed by fusing individually induced dorsal forebrain/pallium organoids containing glutamatergic neurons with ventral forebrain/subpallium organoids containing γ-aminobutyric acid (GABA)-ergic neurons [12,13,27]. These forebrain assembloids recapitulated the dorsal–ventral axis and GABAergic interneuron migration, and showed extension of axonal projections and formation of synaptic assemblies [13]. This provided a model to study interneuron migration which is critical for cortical circuit formation [12] and for understanding the molecular mechanisms of migrational deficits of cortical interneurons in Timothy syndrome [27]. Another assembloid model for studying interneuron migration was developed by fusing MGE and cortical organoids, where interneurons migrated from MGE organoids and integrated into the cortical organoids and formed excitatory postsynaptic densities [28].

Figure 2. Methods to generate assembloids that recapitulate inter-brain-region interactions and neural circuitry.

Brain organoid assembloids include multi-region assembloids and multi-lineage assembloids. Multi-region assembloids can be used to recapitulate in vitro interactions between different brain regions and multi-lineage assembloids can integrate vascular structures and microglia cells into brain organoids. Multi-region assembloids are generated by fusing individually formed brain region-specific organoids using different methods such as (1) coculturing in tubes, (2) coembedding in Matrigel, (3) coculturing in multiwell plates, (4) acoustic devices, (5) 3D bioprinting, (6,9) 3D printed embedding molds, (7) transwell plates, and (8) biohybrid robot-on-a-chip. Multi-lineage assembloids refer to brain organoids with endothelial cells, pericytes, and microglial cells. These nonneuronal lineage-originated cells were incorporated into brain organoids by (2) coembedding in Matrigel, (3) coculturing in multiwell plates, (5) 3D bioprinting, or (10) microfluidic devices. Abbreviations: MGE, medial ganglionic eminence; MSN, medium spiny neuron; SVZ, subventricular zone; VZ, ventricular zone. Figure created with BioRender.

Importantly, assembloids enable long-range neuronal activity and circuit formation which can be analyzed using calcium imaging or electrophysiological recordings. For example, the normal activity of migrated interneurons was observed by the calcium transient and active calcium surge in the MGE and cortical assembloids [28]. Corticothalamic assembloids which can recapitulate corticothalamocortical circuits showed reciprocal thalamocortical projections similar to the reciprocal connections between the thalamic nuclei and the cortex that occur in vivo [14]. Interestingly, thalamic neurons in the corticothalamic assembloids exhibited higher firing frequency than those in non-fused thalamic organoids, suggesting functional interactions within the assembloid [14]. Furthermore, striatal and midbrain organoids that were fused to model a part of the basal ganglia exhibited electrophysiologically active nigrostriatal and striatonigral pathways. Overexpression of α-synuclein (α-syn) in these striatal–midbrain assembloids showed α-syn transportation from striatal to midbrain organoids along dopaminergic axons, and triggered nigrostriatal system damage reminiscent of Parkinson’s disease (PD) [29]. Network analysis showed that the network burst frequency and the number of spikes per burst are higher in striatal–midbrain assembloids than in individual striatal and midbrain organoids, which indicated higher network activity in assembloids compared with the individual organoids [29]. Optogenetics can be coupled with calcium imaging or electrophysiology to facilitate analysis of the spatiotemporal pattern of neuronal activity in specific neuronal cell types [30]. For example, corticostriatal assembloids were developed by fusing cortical and striatal organoids and these assembloids encompassed corticostriatal projections and the cortical neurons sent axonal projections into the striatal organoids where they formed synapses with medium spiny neurons (MSNs). Optogenetic stimulation of the cortical neurons in these assembloids led to evoked calcium responses in the striatal MSNs. Combined with electrophysiology, optogenetic stimulation of the cortical neurons was shown to induce firing and evoke excitatory postsynaptic currents and potentials in the MSNs, which further demonstrates that cortical neurons form functional synaptic connections with MSNs in the striatal organoids [11]. Assembling striatal organoids with functional MSNs with midbrain substantia nigra organoids, MSNs extended projections from the striatum to the substantia nigra and formed synaptic connections with GABAergic neurons. Calcium imaging showed increased neuronal activity in assembloids compared with individual striatal organoids. In addition, optogenetic stimulation of striatal neurons induced evoked inhibitory postsynaptic currents in substantia nigra neurons, indicating that synaptic connections were formed between the neurons in striatal and substantia nigra organoids. These results indicate that these assembloids can be used to model striatonigral circuits relevant to studying motor disorders [31]. Moreover, thalamocortical assembloids have been used to study the circuit-level effects of calcium voltagegated channel subunit α1G (CACNA1G) gene variants [32]. Optogenetic stimulation of diencephalic organoids, the developmental precursor to thalamic glutamatergic neurons, increased firing rates and calcium responses of the neurons in the cortical organoids [32].

A more complex milestone was the fusion of multiple (i.e., more than two) brain region-specific organoids or neural structures to form three- or four-way assembloids (Figure 2). For example, corticomotor assembloids encompassing cortical organoids, spinal cord organoids and skeletal muscle spheroids were developed, where neurons in the cortical organoids connected with the spinal cord organoids and activated muscle spheroids to generate contraction. Rabies virus tracing, calcium imaging, and patch-clamp recordings showed that the corticofugal neurons project to and connect with the motor neurons of the spinal organoids that in turn connect to the muscle spheroids [33]. Photostimulation of the cortical organoids with glutamate uncaging resulted in muscle contraction, which suggested the presence of a corticospinal muscle functional unit. Similarly, cerebral or midbrain organoids were assembled with multiple motor neuron spheroids and connected to a muscle bundle on a solid substrate. In this system, stimulation of the organoids induced muscle bundle movement, and glutamate or levodopa treatment of the cerebral or midbrain organoids increased the corresponding electrophysiological signals including voltage, number of spikes, and mean firing rate. These electrophysiological signals transferred to the muscle bundle and increased muscle movement, indicating that a functional brain–motor system can be developed [34]. Furthermore, cortical and thalamic organoids were assembled with retinal organoids to reconstruct visual circuits. In these assembloids, the retinal ganglion cells extended axons deep into assembloids which recapitulate the projections of the visual system, and the astrocytes in the thalamic organoids also migrated into the retinal organoids, akin to the in vivo interactions between these regions [35]. Four-brain region assembloids have been developed by assembling cortical, diencephalic, dorsal spinal cord, and somatosensory organoids, aiming to recapitulate the human ascending spinothalamic pathway. In these assembloids, optogenetic stimulation of sensory neurons was transmitted through spinal and thalamic neurons to cortical neurons, and calcium levels were increased in all the other regions [36]. These advanced organoid models recapitulate the in vivo neural circuits and brain region interactions and have great potential for deconstructing the circuitry components of complex animal behaviors, ranging from motor and sensory to cognitive and executive functions. For example, memory formation, which involves cortical [37], corticohippocampal [38], corticothalamocortical [39], mesocortical [40], basal ganglia [41], and Papez [42] circuits, may be studied using the assembloid models. Furthermore, the corticothalamic and corticothalamocortical assembloids might be useful for studying sensory processing, consciousness, and learning [39]. Reward-related behaviors which involve mesocortical pathways [40] and basal ganglia circuits [43] can be investigated using forebrain–midbrain and midbrain–striatal assembloids. Moreover, brain–tumor assembloid models have pushed the boundaries of tumor–brain interaction studies (Box 1). Despite the aforementioned enormous potential of assembloids, challenges remain, such as (i) the lack of a standard method for fusing the organoids into assembloids, which has led to variability between assembloids; and (ii) the inability to recapitulate the white matter between brain regions by using traditional fusion methods, which has made assembloid models unsuitable for studying axon bundles, an important feature of inter-regional functional connectivity [44]. However, recent technologies have helped to solve these challenges, as summarized in Box 2.

Box 1. Assembloid models for studying tumor–brain interactions.

Brain–tumor interactions have been studied by assembling the brain organoids with tumor organoids. For example, the invasive nature of the glioblastoma cells was studied by using glioblastoma–cerebral assembloids. Time-lapse imaging over 90 h showed that glioblastoma cells can invade the cerebral organoid as single cells or as cell clusters. Tumor microtube networks which resemble the direct intercellular communication routes and tumor–normal cell connections (indicated by the synapse-like junctions between the tumor and normal cells) were observed in the assembloids [120]. In glioblastoma organoids assembled with dorsal and ventral forebrain organoids, it was observed that glioblastoma cells specifically invade and selectively inhibit axonal growth in the dorsal forebrain organoids. Neuronal activity in dorsal forebrain organoids was significantly increased, which may have been caused by a decrease in vesicular GABA transporter expression [121]. The capacity of glioblastoma integration into neuronal circuitry can be studied by transplanting glioblastoma organoids into adult mouse brains. A monosynaptic viral tracing system can help to establish brain-wide connectivity of glioblastoma cells into diverse neurotransmitter networks [122]. Assembloids between cortical organoids and small-cell lung cancer aggregates have indicated that astrocyte recruitment to the tumor microenvironment is driven by the small-cell lung cancer cell-secreted brain development factor, Reelin; in turn, astrocytes promote cancer cell growth in the brain, highlighting the importance of astrocyte and cancer cell interactions in metastatic brain conditions [123]. In summary, tumor–brain assembloids hold great potential for investigating tumor–brain interactions including tumor invasion mechanisms, tumor–brain connections, and interactive changes in the microenvironment.

Box 2. Approaches for addressing the limitations of assembloid models.

The use of embedding molds and microfluidic technologies has greatly helped to standardize the development of assembloids and reduce variability. A fusion method via designed embedding molds has helped to position ventral midbrain, striatal, and cortical organoids in a linear array that resembles their anterior–posterior positioning in vivo [124]. For more spatially controlled fusion and arrangement of organoids, an acoustofluidic method was developed in which the spatial location of organoids was controlled in a contact-free, label-free, and minimal impact manner by regulating dynamic acoustic fields within a hexagonal acoustofluidic device. The spatial control enabled precise regulation of neuron projection parameters, projection maturation, and neural progenitor cell (NPC) division in the assembloids. Moreover, by taking advantage of automatically controlled microfluidic technology, two-way assembloids were generated in a controlled way [125]. Alternatively, electrospray can generate cell microcapsules, and various cell microcapsules can be flexibly arranged to generate region-specific brain organoids and assembloids [126].

The axon bundles in white matter between brain regions have recently been recapitulated by growing cerebral organoids on two chambers located at the ends of a microdevice. The reciprocally connected axons in a narrow channel between two organoids have enabled the recapitulation of axon bundles and the mimicking of inter-regional cortical connections [127]. When the microdevice was equipped with a multielectrode array (MEA) layer and polydimethylsiloxane (PDMS) microfluidic layer, MEA recordings showed that these organoids with inter-regional axonal tracts exhibited more robust neuronal activity compared with fused cerebral organoids. Thus, recapitulating axonal connections within regions can contribute to enhanced network activity in assembloids [44]. Altogether, these improved technologies have supported the standardization and optimization of assembloid models to make them more applicable to future studies.

Synaptic function and electrical activity in organoids

The improved organoid models described earlier showed evidence of mature synapses and electrically active neuronal networks which are essential features for robust neuroscience studies. For example, synaptogenesis in midbrain organoids was observed between presynaptic and postsynaptic terminals stained with respective markers [45]. Sparse neuronal labeling also revealed the presence of dendritic spine structures in cerebral organoids [46]. Furthermore, electron microscopy demonstrated the presence of vesicle-filled synaptic structures in cerebral organoids [47].

Spontaneous electrical activity emerges in the developing brain independently of sensory input and plays a significant role in neural circuit formation, neuronal migration, and maturation [48]. Neuronal electrical activity has been demonstrated in different organoid models via calcium imaging [24,28,45,49,50] and electrophysiology [7,45,46,51]. Calcium imaging and genetically encoded calcium indicators (GCaMPs) have been used to label neurons in organoids and confirmed that organoid neurons display calcium activities that can be blocked by the sodium channel blocker, tetrodotoxin [28]. Intracellular calcium dynamics resembling in vivo cerebellar circuit patterns indicated functionally mature network activity in cerebellar organoids, and Purkinje cells in the organoids actively contributed to the formation of functional neural networks [24]. To precisely measure synaptic functions, methods for electrophysiological recording from organoids have evolved rapidly (Figure 3). Whole-cell patch-clamping has been used to record the electrophysiological properties of the cellular membrane and intracellular action potentials of individual neurons in brain organoids [46,51]. This provides individual neuron activity recordings with high temporal resolution, but the lower spatial range limits access to complex network activity [51]. Furthermore, multielectrode arrays (MEAs) can be used to monitor the activity of neuronal networks within organoids, as well as the evolution of network activity over time [6,45,46,52,53]. For example, MEAs recordings showed an increased mean firing rate in spinal organoids treated with pain-evoking chemicals, and demonstrated the feasibility of modeling nociception circuitry [53]. Moreover, the electrical activity of neurons in cerebral organoids from 2 months to 10 months was recorded with MEAs, and network events or synchronized ring across multiple electrodes were observed, in addition to nested oscillatory events that initiated at ~4 months and transitioned to stronger and more variable oscillations at ~10 months, patterns that resemble some preterm electroencephalography (EEG) recordings [6]. Neuronal firing frequency can be modulated in midbrain organoids by using a dopamine receptor agonist, quinpirole [45]. Methods for recording functional activity in brain organoids using calcium imaging and MEAs have provided valuable standardized methods for recording neuronal networks at the synaptic and network levels [54].

Figure 3. Methods for measuring electrophysiological activity in brain organoids.

The techniques for obtaining electrophysiological recordings from brain organoids have been improved to enable long-term recording without interfering with organoid structures. These approaches include patch-clamp, multielectrode arrays (MEAs), 3D MEAs, 3D multifunctional mesoscale framework (MMF), shell MEAs, ‘pocket’-like MEAs, cyborg organoids, and kirigami electronics. Figure based on data from [6,45,46,51–53,55–59,62]. Figure created with BioRender.

However, a limitation of neural activity recording using MEAs is that the MEAs are implanted at, and record mostly from, the bottom of the organoids, and thus do not record all organoid neurons. As a solution, the traditional 2D MEAs have been modified into 3D MEAs which can robustly record functional neural networks throughout the organoids [55]. Inserting the 3D MEAs into the organoids, however, can inevitably damage the organoids. Therefore, a new microfabricated 3D framework was developed consisting of 3D microelectrodes – the 3D multifunctional mesoscale framework (3D MMF) – that can envelop the organoids, and the electrical interface enabled recording from organoids without damaging them. 2D flexible electronics have been folded into a designed 3D structure to precisely match the morphology of the organoids which will be integrated into the 3D MMF, and thus can monitor the electrophysiological activities of the organoid through intimate contact across its surface [56]. The 3D MMF has been used to record wave spreading, firing, and bursting events with less damage to the organoids [56]. Similar 3D-shaped shell MEAs generated with tunable polymer leaflets and conductive polymer-coated metal electrodes, that are adjustable for organoids of different sizes, have facilitated long-term recordings of organoids without disrupting the organoid inner structure [57]. Another highly stretchable 3D dual MEAs device enabled continuous medium exchange when recording by using an electrode ‘pocket’-like design in which a pair of fabricated stretchable MEAs are aligned vertically around the ‘pocket’ [58]. Furthermore, to achieve long-term 3D recordings of single-cell electrophysiology in organoids, elastic tissue-like nanoelectronics using serpentine structures in the mesh were designed [59,60]. Stem cells were cocultured with the elastic nanoelectronics and were used to reconstruct brain organoids by folding the nanoelectronics; these were referred to as ‘cyborg organoids’ [59]. These elastic nanoelectronics can stretch to accommodate tissue growth and can be linked to an external recording set-up for long-term measurement. Thus, the stem cell-derived neuronal cells and nanoelectronics are in close contact, and cellular activity can be recorded as single-cell electrophysiology during the organoid development process [60]. Further advanced cyborg organoids using a liquid metal polymer conductor-based mesh neuro-interface showed neural spike, synchronization, and oscillation activities in hippocampal organoids [61]. However, while being a promising, long-term stable technology for electrophysiological recordings from organoids, the cells in the cyborg organoids must be in contact with the nanoelectronics from the initiation of organoid induction, and this inevitably interferes with organoid cytoarchitecture and development. Therefore, a new 3D folding electrical recording platform, kirigami, has been developed that consists of ultrathin electrodes that can geometrically accommodate organoids in suspension, thereby enabling their long-term integration and continuous recording of neuronal activity in brain organoids [62].

Enhancing the capabilities of brain organoid models

Previous brain organoid models suffered significant problems that limited their capabilities and potential applications. Recent endeavors involving innovative strategies such as long-term culture devices and multi-lineage assembloids which include vascular and microglial cells inside organoids have significantly alleviated these limitations (Figure 2). We later summarize some major limitations of previous organoid models, and the approaches devised to resolve them and thus significantly enhance the capabilities of the organoid models.

Cell death in brain organoids

Culturing organoids is a lengthy process that presents challenges, including cell death in the organoids during culture that is probably due to inadequate penetration of oxygen and nutrients to the core. Thus, shaking approaches during organoid culturing were introduced, as detailed in Box 3 and Figure 4. Another strategy to enhance nutrient access to the inside of organoids involved slicing the organoids and culturing them at the air–liquid interface, and this enhanced neuronal survival and maturation (Figure 4) [46,63]. Alternatively, to avoid slicing, whole brain organoids were cultured on a microfluidic platform by incorporating perfusable culture chambers, air–liquid interfaces, and a ‘one-stop’ protocol (Figure 4). This microfluidic platform minimized hypoxic core formation and enhanced organoid uniformity [64].

Box 3. Agitation methods for long-term culture of organoids.

Spinning bioreactors and agitation devices (see Figure 4 in the main text) have enabled organoids to be cultured for up to 1 year [2]. Given that the bulky nature of these spinning bioreactors makes them difficult to install in standard tissue-culture incubators, miniaturized (mini-) bioreactors were developed (see Figure 4 in the main text) [7]. These mini-bioreactors use a smaller volume of medium and thus reduce the cost of the procedure. Another type of spinning bioreactor, the rotating wall vessel (RWV) bioreactor, has been used for organoid long-term culture. The RWV bioreactor gently drags the fluid in the cylinder on its axis, and the spheroids/organoids in it rotate in a circular path [128,129]. It provides an environment with lower shear and turbulence which can reduce potential damage to the organoids and facilitate 3D organoid formation by providing the cells with the spatial freedom to colocalize and self-assemble [130]. Recently, to investigate the mechanical forces that can affect the formation of the brain organoids, researchers cultured brain organoids in a vertical mixing bioreactor [131]. Vertical mixing enabled the high turbulent energy which is related to the stirring force around organoids and maintained inter-organoid distances. Interestingly, the vertical mixed brain organoids showed an inverted structure compared with orbital mixed organoids. Neural progenitor cells (NPCs) were located at the peripheral edge of the organoids and the neurons were generated in the center of the organoids. Single-cell RNA sequencing (scRNA-seq) analysis indicated that the neurons in the inverted brain organoids resembled GABAergic neurons of the ventral forebrain [131]. Thus, the fluid dynamics in culturing medium not only correlated with the organoids survival, differentiation, and maturation but also correlated with organoid structure and the cell types that were generated. To avoid the need for specialized equipment or bioreactors in the aforementioned methods, and to make organoid generation more simple and affordable by most laboratories, orbital shakers (see Figure 4 in the main text) were tested and compared with the bioreactors [2,132]. Interestingly, orbital shakers were found to produce lower shear stress compared with the mini-bioreactors; these latter devices exhibited lower speeds and higher shear stress that may negatively affect organoid development [132]. It thus seemed that orbital shakers, in lieu of spinning bioreactors, enable easier, more affordable, and more robust generation of brain organoids.

Figure 4. Methods for long-term culture of brain organoids.

Various approaches have been devised, including orbital shaker, spinning bioreactor, spinΩ mini bioreactor, vertical bioreactor, rotating-wall vessel bioreactor, and air–liquid interphase culture for sliced and whole organoid and organoid-on-a-chip models to improve the long-term culturing of brain organoids. Abbreviation: RWV, rotating wall vessel. Figure created with BioRender.

Moreover, vascularization of organoids was attempted to enhance nutrient access to the organoid core, reduce cell death, and recapitulate the blood–brain barrier (BBB). Organoids were coembedded in Matrigel along with endothelial cells derived from the same induced PSC (iPSC) line, and vasculature-like tubular structures formed around the organoids [65]. Alternatively, PSCs or PSC-derived neural progenitor cells (NPCs) and endothelial cells cocultured from the beginning of induction [66,67], or vascular spheroids from PSCs or umbilical vein endothelial cells, were assembled with the brain organoids and resulted in angiogenesis [68,69]. The assembly of blood vessel organoids with cerebral organoids facilitated the modeling of BBB characteristics in vitro and helped to model neurovascular interactions, especially in cerebral cavernous malformation [70]. Furthermore, organoid-on-chip approaches have been used to recapitulate the BBB. Organoid-on-chips consist of microfluidic channels and chambers containing the organoids, and thus resemble a computer chip. This allows the flow of the medium through a vascularized organoid in a microfluidic system, and enabled temporal synchronization and spatial orientation between the vascular cells and organoids [71], enhanced neuronal differentiation, and reduced cell death (Figure 4). Further advances using organ-on-chip platform modeling of the human BBB may enable precision medicine testing of BBB properties of individual patients with iPSC-derived brain organoids and endothelial cells in microfluidic organ-on-chip devices.

Lengthy gliogenesis process

In addition to neurons, the spontaneous induction of astrocytes [51] and oligodendrocyte progenitor cells [13] has been demonstrated in organoids, although they tend to appear later – consistent with the late switch between neurogenesis to gliogenesis in embryogenesis [72]. The addition of oligodendrocyte lineage growth factors such as platelet-derived growth factor AA, insulin-like growth factor 1, and hepatocyte growth factor, as well as hormones including thyroid hormone and insulin, has facilitated the maturation and myelination of oligodendrocytes to a great extent [73,74]. However, these protocols require >3 months of differentiation to achieve myelination. To overcome this limitation, in addition to external growth factors and hormones, the use of human OLIG2 or SOX10 PSC reporter lines which express the essential transcription factors for oligodendrocyte development has shortened the time needed to develop mature oligodendrocytes with a myelin sheath layer to a minimum of 42 days [75,76].

Lack of microglia in organoids

A lack of immune cell types such as microglia, that are necessary to study immune responses in brain diseases including neuroimmune and neurodegenerative disorders, affected earlier organoid models. To address this, integration of microglia was recently achieved by coculturing microglia with the organoids [77]. Microglia-like cells (MLCs) derived from iPSCs were cocultured with organoids, and the MLCs showed differential migration ability, intracellular Ca²⁺ signaling, and responses to proinflammatory stimuli [78]. In addition to MLCs, microglial precursors (erythromyeloid progenitors derived from iPSCs) have been integrated with brain organoids and transplanting them into mouse brain has allowed homeostatic human microglia development in vivo [79]. Similarly, macrophage precursor cells were integrated with midbrain organoids, and this led the organoids to release cytokines and chemokines, and influenced oxidative stress, immune response, and synaptic remodeling-related gene expression [80]. Myeloid-specific transcription factor PU.1 overexpression also induced MLCs that showed an intact complement and chemokine system when incorporated into brain organoids [81]. These methods for incorporating microglia into brain organoids have enhanced the ability of organoids to model brain diseases that involve a significant immune component.

Low efficiency and reproducibility of brain organoid generation protocols

Brain organoid generation involves lengthy protocols that produce organoids with significant variations between them. Thus, the use of 3D bioprinting to generate organoids has been adopted to reduce the limitations of traditional protocols. 3D bioprinting of neural tissues requires specialized bioinks that can support cell survival and differentiation into mature neural cell types. Fibrin-based bioinks have been used to support neural tissue growth from iPSCs and iPSC-derived NPCs in a microfluidic device [82,83]. Alternatively, extrusion-based bioprinting with an alginate/Matrigel bioink containing iPSC-derived NPCs was used to generate mature neural tissue [84]. Furthermore, programmable and patterned vascularized organoids composed of neural stem cells, neurons, and endothelial cells were achieved within days via orthogonally inducing iPSCs into neuronal and endothelial lineage through the forced overexpression of transcription factors [85]. In a recent study, the bioprinted 3D neural tissues formed neuron–astrocyte networks and cortical striatal projections resembling functional neural circuits [86]. An advanced microfabricated embryoid body disk device can be used for scaling up the organoid generation process and decreasing organoid variability [87]. Altogether, these advances in the use of bioprinting to generate organoids have facilitated accelerated nutrient perfusion into organoids and enhanced the speed and reproducibility of organoid generation.

Suitability for modeling aging-dependent and neurodegenerative disorders

The earlier described improvements in the organoid models, including enabling long-term culture of organoids and integrating microglia into organoids, may enhance the suitability of organoids for modeling aging-associated brain disorders. However, another challenge in that regard is that the organoids represent early developmental stages and do not express aging-associated molecular and cellular signatures of neurodegenerative disorders such as Alzheimer’s disease (AD) and PD. In an effort to simulate aging in organoids, midbrain organoids were maintained in BDNF and GDNF without antioxidants for up to 60 days. These organoids exhibited aging-related features of human midbrain, such as upregulated senescence-related genes and DNA damage markers [88]. A senescence-inducing drug, hydroxyurea, was also used in spheroids to trigger aging-like phenotypes [89]. However, this drug, which acts via inhibiting ribonucleotide reductase and DNA synthesis/replication, could induce nonspecific effects on cells, and this remains a possible caveat that will need to be addressed. A small-molecule amyloid β 42 (Aβ42) inducer, Aftin-5, has also been used in organoids to accelerate the generation of the Aβ42 and enhance the Aβ42/Aβ40 ratio [90]. Sporadic AD pathogenesis can be modeled by adding human serum to organoids since serum exposure caused by BBB disruption in AD patients is a key risk factor in AD [91]. A hallmark of PD pathology, α-syn aggregation, usually occurs later in disease. In organoids, it can be accelerated by an optogenetics-assisted induction system [92] or by seeding of preformed α-syn fibrils [93]. The dopaminergic neurotoxin 1-methyl-4-phenyl-1,-2,3,6-tetrahydropyridine (MPTP), which is used to trigger PD neuropathology, can induce cell death in dopaminergic neurons [17]. Overall, the organoid models have become better suited for modeling aging-associated neurodegenerative conditions.

Brain organoids as disease models

The advanced organoid models have enabled studies of multiple types of brain disorders and neurological conditions. As examples, we review recent applications in two classes of neurodegenerative disorders (AD/dementia and PD) and an infectious disease [coronavirus disease 2019 (COVID-19)-associated neurological conditions].

Alzheimer’s disease and dementia

Brain organoids with AD-associated genetic mutations or variants have shown extracellular deposition of Aβ, including Aβ plaques, high levels of phosphorylated tau protein in the somata and neurites, and detergent-insoluble filamentous tau [94]. Cerebral organoids carrying the apolipoprotein E ε4 (APOE ε4, or simply APOE4) variant, the strongest genetic risk factor for sporadic AD, showed increased Aβ aggregates and tau phosphorylation compared to APOE3 organoids. Neurons with APOE4 variant showed an increased number of synapses and astrocytes showed impaired Aβ uptake and increased cholesterol accumulation [95]. Elevated Aβ40 and Aβ42 levels and some cellular apoptosis at the edge of the organoids were associated with reductions of synaptic proteins in AD patient organoids [96]. Furthermore, neuronal network activity testing in cerebral organoids showed disrupted calcium signaling in AD organoids, characterized by the lack of synchronization of calcium transients and the increased amplitudes of spontaneous calcium transients in the organoids [50]. In mutant Tau cerebral organoids modeling frontotemporal demantia (FTD), impairment in the autophagy-lysosomal pathway was observed, which led to loss of glutamatergic neurons in 6-month-aged organoids [97]. Moreover, the generation of multi-lineage assembloids (Figure 2) has facilitated the study of neuroinflammation in AD. When MLCs were assembled with organoids, intact complement and chemokine systems protected the parenchyma from cellular and molecular damage caused by Aβ, and reduced Aβ-induced expression of genes associated with apoptosis, ferroptosis, and AD stage III [81]. MLCs from APOE4 homozygous iPSCs were cocultured with APP duplication organoids, and these assembloids showed increased Aβ deposits and phospho-tau levels compared with those cocultured with healthy APOE3 MLCs in an in vitro model of AD [95].

Parkinson’s disease

Midbrain organoids are the most commonly used organoid model to study PD-relevant dopaminergic neuron pathologies. Midbrain organoids encompass dopaminergic neurons that produce and secrete dopamine [98]. Midbrain organoids carrying the LRRK2 G2019S mutation, a common genetic cause of PD, recapitulate disease-relevant phenotypes [88,98] such as a decrease in the number and complexity of dopaminergic neurons [98]. Moreover, midbrain organoids carrying triplications of SNCA, the gene encoding α-syn, promote the formation of Lewy body-like inclusions with a spherically symmetric morphology and an eosinophilic core containing α-syn with ubiquitin [99]. When SNCA-overexpressing midbrain organoids were assembled with striatal organoids, they exhibited shorter, fewer, and random axonal projections compared with wild-type organoids, as well as decreased electrophysiological and network activity. α-Syn propagation via retrograde movement through dopaminergic neuron axonal projections was observed and α-syn accumulated in midbrain organoids. These assembloids provide a model to study the ‘prion-like’ propagation hypothesis which suggests that cell-to-cell transmission of pathological α-syn contributes to synucleinopathies [29].

COVID-19-associated neurological conditions

The COVID-19 pandemic has introduced a range of neurological conditions associated with severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection [100]. Studies on post-mortem brains of COVID-19 patients showed the presence of viral RNA and protein, including subgenomic RNA, indicative of active virus replication in the brain, and infectious virus could be isolated from human brain [101–103]. Moreover, studies on patients who recovered from acute COVID-19 suggested long-term structural brain changes or cognitive/neuropsychiatric deficits [104–106]. Organoids have been widely used to explore the potential tropism of SARS-CoV-2 in brain cells, especially given the lack of robust animal models to study SARS-CoV-2 central nervous system (CNS) tropism. For example, brain region-specific organoids, including cortical, hippocampal, hypothalamic, midbrain, and choroid plexus organoids, have demonstrated possible SARS-CoV-2 infection to different extents, and the virus was reported to target neurons, astrocytes, and neurovascular cells [10,107–109]. However, the differential susceptibilities to infection of different cell types or organoid models might be attributed to different expression levels of the SARS-CoV-2 entry factors such as angiotensin-converting enzyme 2 (ACE2), which showed higher expression in mature neurons than in undifferentiated NPCs [110,111].

Brain organoid models have also been used to study the effect of SARS-CoV-2 infection on neuropathologies associated with neurodegenerative conditions [112]. For example, cortical organoids showed tau mislocalization from axons to somata in infected neurons and tau threonine-231 phosphorylation, as occurs in tauopathies and dementias [107]. The assembly of blood vessel organoids with cortical organoids has facilitated BBB-like features with tight and adherens junctions, as well as pericyte, astrocyte, and microglia/macrophage cells. After SARS-CoV-2 infection, the cortical–blood vessel organoids exhibited Aβ accumulations, tau mislocalization and upregulation, and microglia and astrocyte activation induced by proinflammatory cytokines [113]. Moreover, astrocytes can increase SARS-CoV-2 infection in brain organoids and when co-cultured with iPSC-derived neurons with APOE4 allele, neurons showed more severe infection compared with those with the APOE3 allele [114]. Infection reduced neurite number and length and triggered synaptic loss, especially in APOE4 neurons [114]. Together, studies using brain organoid models suggested a possible association between COVID-19 and the risk of developing neurological or neurodegenerative conditions; however, more support for that possibility is needed from animal models and clinical studies.

Concluding remarks

Brain organoids have paved the way to investigate the molecular and cellular mechanisms of brain development and disease in human models. Advances in organoid modeling of the human brain have significantly improved the capabilities of the model, such as the ability to reconstruct neural circuitry and to generate reproducible organoid batches that can be cultured for extended periods of time or that encompass essential elements of the brain such as microglia or BBB. More work, however, will be necessary to recapitulate further features of the human brain in organoids and to realize the full potential of organoids for understanding the molecular and circuit bases of behavior and disease (see Outstanding questions). Moreover, the organoids will help to reveal disease phenotypes and test new therapeutic approaches. For example, the combined use of assembloids and organoid transplantation has promoted investigation of new therapeutic strategies such as antisense oligonucleotides for Timothy syndrome [115]. Recent efforts to integrate human brain organoids into mouse or rat frontoparietal, somatosensory, and visual cortex have suggested a possible future therapeutic strategy for brain injury [116–118]. However, ethical challenges will first need to be overcome [119].

Outstanding questions.

Can organoids enable the deconstruction and ‘teasing out’ of the cellular and circuit bases of animal behavior in vitro?

Can organoids further facilitate modeling of CNS–peripheral nervous system (PNS) interactions and brain–peripheral organ axes?

Can organoids be used to identify regenerative therapies that can precisely rescue defective circuits that underlie many brain diseases?

Can organoids exhibit some level of ‘sentience’ and learning behaviors, as previously suggested for some iPSC-derived neuron models?

Can the organoid technology, wedded with artificial intelligence, facilitate studying the neuroscience of human intelligence?

Can organoids better recapitulate the gradual and aging-dependent neurodegenerative disease progression process (beyond recapitulating the pathologies) that occurs in human AD and PD?

Can mixed-species (human and non-human primate) brain region organoid assembloids help to delineate human-specific circuits that may underlie some features of brain development and advanced cognitive functions that are specific to humans?

Highlights.

Organoid models can specifically recapitulate a large array of brain regions to enable circuitry reconstruction in vitro.

Organoids enable modeling of the interactions between multiple brain regions.

Brain organoid and assembloid models can recapitulate some of the neural circuits observed in vivo and display electrophysiologically active neurons and synapses.

Advances in organoid protocols have eliminated some major limitations and have improved the applicability of organoid models for studying aging and neurodegenerative conditions.

Glossary

Acoustofluidic method

a technology that uses acoustic waves to control fluids and particles within fluids.

Alzheimer’s disease (AD)

a neurodegenerative disease characterized by pathological hallmarks such as protein aggregation and amyloid plaques in the brain caused by the accumulation of Aβ and hyperphosphorylated tau protein in the brain.

Amyloid β (Aβ)

the main peptide that forms plaques in AD. It is produced by selective processing of the amyloid precursor protein (APP) by the γ-secretase enzyme complex. Impairment in Aβ clearance has been highly associated with AD.

Calcium imaging

a microscopy technique to optically record neuronal cellular activity by fluorescent molecules that bind to calcium – increased intracellular calcium is a measure of neuronal activity.

Glutamate uncaging

a technique used to study synaptic function and neural circuitry. It involves the use of caged glutamate – an inactive form of the glutamate bound to a light-sensitive protecting group. When exposed to a specific wavelength of light, the protecting group is removed, ‘uncaging’ the glutamate and allowing it to activate receptors locally.

Lateral ganglionic eminence (LGE)

a progenitor domain that generates striatal projection neurons and interneurons.

Lewy body

round, reddish inclusions in nerve cells that are composed of an abnormally folded and aggregated α-synuclein (α-syn) protein.

Medial ganglionic eminence (MGE)

a vital source of GABAergic interneuron progenitors and cortical interneurons.

Medium spiny neurons (MSNs)

a type of projection neuron in the striatum that sends inhibitory projections to basal ganglia nuclei.

Microglia

the resident innate immune cells of the brain, which react to pathogens, toxins, and cell debris by phagocytosis as well as by the production of cytokines and chemokines. In neurodegenerative diseases, microglial activation is induced by neuronal stress and degeneration.

Multielectrode arrays (MEAs)

grids consisting of tightly spaced microscopic metal electrodes. They are used to measure the spontaneous firing activity of neurons.

Neural progenitor cells (NPCs)

CNS progenitor cells that generate most of the different types of glial and neuronal cells.

Optogenetics

a technique that can activate or inhibit specific types, subtypes, or subpopulations of neurons/cells in a high temporal resolution by expressing light-sensitive proteins called opsins.

Parkinson’s disease (PD)

a neurodegenerative disease characterized by the degeneration mainly of dopaminergic neurons in the substantia nigra pars compacta and increased α-syn accumulation and aggregation in the form of Lewy bodies and Lewy neurites.

Patch-clamp recordings

an electrophysiological technique used to measure the cell membrane potential and the amount of current passing across the cell membrane.

Rabies virus tracing

a technique that uses the movement of engineered rabies virus between cells as a label to determine the connectivity of a neuron.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)

a highly transmissible and pathogenic coronavirus that caused the COVID-19 pandemic. Infection leads to a highly contagious severe respiratory disease that can spread in aerosols and may affect different organs of the body, including the CNS.

Spheroids

a ‘primitive’ form of 3D cultures that are made of tissue-related single or multicellular cell types and represent partial tissue organization. This term has recently been limited to cellular systems generated by putting several already patterned cells that have less self-organization based on the recent nomenclature consensus.

Synucleinopathies

a diverse group of clinically and pathologically heterogeneous neurodegenerative disorders characterized by pathologic aggregates of insoluble α-syn in neurons and glia.

Tau protein

a microtubule-associated protein in neurons that tends to self-aggregate to form β-sheet structures which further accumulate as neurofibrillary tangles in AD and other dementias.

Tauopathies

a class of neurodegenerative disorders pathologically defined by the deposition of abnormal tau protein in the brain.

Timothy syndrome

a rare disorder affecting the heart and nervous system. It is caused by mutations in the gene CACNA1C which encodes a calcium channel in the neuron cell membrane.

Tropism

the ability of a pathogen (e.g., a virus) to preferentially infect particular cell types rather than others because of the availability of infection-facilitating host cell machinery and/or virus infection-regulating immune responses in those cell types.