The Brain Organoid Technology: Diversity of Protocols and Challenges

Abstract

The brain organoid technology emerged a little over a decade ago. During this short time span, the handling approach has seen tremendous advancements in order to solve current obstacles and enable the development of new applications. Using these methodologies, the fundamental characteristics of the majority of the brain regions may be mimicked in organoids; however, the existing brain organoids cannot be regarded an exact replica of the human brain or its anatomical regions. This chapter will present some of the biological phenomena on which the brain organoid technology relies. Following this, a summary of the gross common structure and timeline of the brain organoid protocols along with their main components and strategies for their improvement is included. A special selection of protocols for each major brain region will be presented with their origin, rationale, and key specifics. Finally, some of the daunting challenges to brain organoid technology will be highlighted.

1. Introduction

More than a decade has passed since the first brain organoid emerged from the lab of one of the fathers of organoid technology – Yoshiki Sasai. During that time, an increasing number of researchers worldwide got attracted by the potential of this technology and the hope that we may learn how to build our brains in vitro and thus further adopted and developed it [1]. A few years before the first brain organoid protocol was published, Zhang et al. developed a protocol for the generation of free-floating embryoid bodies (EB) from dissociated human embryonic stem cells (hESC), which afterward could be directed to differentiate into neural precursors in rosette structures. This technique actually was an offspring of the methods for aggregated suspension culture from dissociated fetal brain cells like the one by Bjerkvig et al. from 1986 [2, 3]. The team used cells from post-neurulation stages, when loss of potency is present and which can affect their self-organization and differentiation capacities, as some recent studies reveal, while Zhang et al. used embryonic cells from a much earlier stage of the embryogenesis and therefore could direct them to form neural rosettes (resembling early neural tube-like structures). Noteworthy to mention is the contemporary study by Doetschman et al., in which they could observe nerve cells generated spontaneously and stochastically from EBs formed from blastocyst cells using undefined medium. However, they lacked the knowledge of the molecular mechanisms and ways to direct them to a desired cellular fate [3, 4, 5, 6]. Zhang et al. succeeded in generating EBs, which were directed to structures recapitulating very early stages of neural tube formation and comprised mostly of neural precursor cells, thanks to the newer discoveries on the neural induction mechanisms [2]. However, the goal of their study was to get transplantable neural precursors and not to recapitulate later-stage brain regions. Therefore, their protocol was much shorter and more straightforward than the ones for organoids.

By the turn of the twentieth century, the recently acquired knowledge of the molecular control of the embryonic brain patterning got consolidated in updated models of neural induction and morphogenesis. It became evident that the brain morphogenesis is spatiotemporally organized and orchestrated by a dynamic network of transient patterning centers (organizers) that secrete a bunch of molecular controllers (morphogens) that can trigger identity change of nearby cells in a concentration−/distance-dependent fashion. These morphogens are activators or inhibitors of a handful of signaling pathways. In addition, more evidence was accumulated that the neural progenitors without extrinsic signals develop anterior neural specification [7]. So, during the first decade of the new millennium, protocols for stem cell differentiation towards identities from the major brain regions were developed based on this framework. Besides, some of the growth and differentiation conditions were also significantly improved; for example, with the popular dual SMAD inhibition strategy for improved neural induction or with the usage of the survival enhancer of dissociated cells -Y-27632 (ROCK inhibitor that diminishes dissociation-induced apoptosis) [8, 9]. This review will present some of the key biological phenomena engaged in brain organoid generation. The structure and the main components of the brain organoid protocols will be analyzed and summarized. The protocols specific to different brain regions will be presented and analyzed to introduce and guide the readers in this growing field. Furthermore, some of the major obstacles related to brain organoid technology and recapitulating the brain and its regions will be discussed.

2. Brain organoid protocols: general timeline, stages, strategies to improve outcome and classification

Most of the current brain organoid protocols use the dissociation-re-aggregation method to form 3D aggregates from the stem cells. After some period of recovery and establishment of cell–cell contacts and maturation, these aggregates become EBs, which are used as a starting point for organoid formation and development. The EBs are not formed from every stem cell aggregate. So far, the studies have revealed a set of important variables [10, 11]. Firstly, the critical cell mass, i.e., a sufficiently large number of cells, is needed to form EB. Secondly, size and uniformity have a huge effect on the homogeneity and structure of the EBs and may impact the fate choice. Studies found that the addition of Y-27632 is beneficial for these results. To overcome these issues, geometric confinement approaches are often used. Thirdly, the time for aggregation also plays a critical role. These and other variables secure the establishment of an appropriate intra-EB microenvironment to trigger the intrinsic embryo developmental programs. The EBs are multicellular aggregates of stem cells, which resemble the inner cell mass of the pre-gastrulation embryo. These cells have the potential to develop into cells from all three germ layers [12]. Afterward, the cells from the EBs were then committed to neural development through medium change and addition in the medium of various signaling molecules. Initially, during this process of differentiation EBs are directed to neuroectoderm formation. The neuroectoderm then gets transformed to neuroepithelium, which depending on the culturing conditions (as dynamic or static culture) attains the form of large neuroepithelium bulges or small neural rosettes, thus recapitulating the neural tube. These structures further differentiate into more specific brain regions according to the applied patterning scheme [13, 14, 15, 16].

The summarized workflow of the procedures and timelines can be divided schematically into the following stages (Figure 1), (note that the steps and timeline are exemplary and can vary depending on the organism, used patterning strategies, goals, and other factors) [13, 14, 15, 16]:

1. Dissociation of the stem cells and re-aggregation– day −1.

2. Formation of embryoid bodies – day 0–1. To the embryoid body medium usually is added Y-27632 to increase cell survival. Microscopically the edges of EB have a thin bright ring.

3. Neural induction – the cells are directed to neuroectodermal fate. Commonly is around 3−6 (10) days for human organoids. The neural induction medium is often supplied with dual SMAD inhibitors (as SB431542 or A83–01 and LDN193189 or dorsomorphin). Microscopically, the edges of EB have a thicker bright ring.

4. Differentiation – the neuroectoderm is further transformed to neuroepithelium and further expanded. Sometimes this stage is subdivided into two:

5. Region-specific patterning − The differentiation/patterning medium is commonly supplemented with patterning molecules, sometimes also with dual SMAD inhibitors.

6. Tissue induction and growth − The medium, sometimes, has neural patterning molecules. Often the organoids are embedded or encapsulated in gel in order to prevent the diffusion of the secreted morphogens.

7. Organoid growth − Maturation medium is commonly used without patterning molecules but instead with growth factors (such as BDNF, NT-3), antioxidants (such as ascorbic acid), and other additives, which enhance the organoid maturation and long-term survival.

Figure 1.

A schematic timeline of generalized brain organoid protocol: A. stages of the treatments; B. timeline of the 3D culture development and growth; C. timeline of the neural differentiation. After [14, 15, 16, 17].

Patterning agents: Currently, the available protocols for brain organoids try to mimic the molecular control exercised by the embryonal network of brain patterning centers. The major signaling pathways elaborated or triggered by them are the pathways of fibroblast growth factor (FGF), sonic hedgehog (Shh), Wnt, bone morphogenic protein (BMP), and nodal/activin. This patterning network with its morphogen gradients can be presented in a 3D coordinate system where at each spatial coordinate of the brain primordium corresponds a combination of morphogens (which usually are the first messengers of these pathways) with specific concentrations (Figure 2) [19, 20]. Based on this model, the researchers approximate the morphogen combination patterning the region of interest and determine the appropriate concentrations afterward. Although in the brain organoid protocols usually are used as morphogens the first messengers of these pathways, in some scenarios are used secondary messengers or their analogs instead – for example, when it is cost-efficient; or it is difficult to work with them; or is needed partial activation of the pathway or for another reason [21]. In general, the used patterning agents can be divided by their ability to caudalize or ventralize the cell populations regarding the most anterior and dorsal end of the central nervous system primordium. As caudalizing agents, the first messengers from the Wnt and BMP families are often employed, while Shh and Wnt are used as ventralizing agents.

Figure 2.

The morphogen gradients in the developing embryonal CNS. T – Telencephalon; D – Diencephalon; MB – Midbrain; R1-R8 – Rhombomeres of the rhombencephalon; SC – Spinal cord. After [18].

Oxygenation and metabolite exchange: To cope with the lack of vascularization and the hindered oxygen and metabolite exchange, several strategies are developed:

1. Static-to-spinning strategy – during the neural induction and early regional specification phases, the organoids are grown under static conditions and, during the longer stages of growth and maturation, are transferred to a shaker or spinning reactor [22].

2. Sliced strategy – the growing organoids are regularly sliced during the growth and maturation phase [23].

3. Air-liquid interface strategy – the maturing organoids, after 2 months are embedded in agarose and cut into slices, one side of which is exposed to air during the consequent culturing [24].

4. Hyperoxygenation – the air above the brain organoid culture is supplied with extra oxygen (up to 40%) [25].

5. Transplantation in animal brain – the maturing organoids are transplanted into the brain of the host animal after one and half months of in vitro culture. After a few weeks, the organoids get vascularized and continue to grow in the host [26].

Mediums and Supplements: They are the fundamental components of the organoid microenvironment; however, there is no strict rule on which formulation to use. It requires experimenting with a few recipe/brand choices and their combinations to achieve optimum output. Some commonly used mediums are DMEM/F12; Neurobasal, GlasgowMEM, BrainPhys, mTESR1, Essential medium, etc. [13, 14, 15, 16, 27, 28].

As supplements are commonly used: KSR (knockout serum replacement), N2, B27, Gem21 Neuroplex, etc. [13, 14, 15, 16, 27, 28].

Stage-dependent medium composition specifics: The biological transformation of the EBs and consequently organoids require adjustment of the microenvironment according to the dynamic transient necessities of the culture. During the initial stages are used DMEM/F12, essential medium supplied with KSR, etc. The medium is often changed during the differentiation stages to Neurobasal, DMEM/F12 with N2, b27(−A). For the maturation stage, often the choice is for Neurobasal, DMEM/F12, or their combination with added supplements [13, 14, 15, 16, 17, 27, 28].

Culture dishes and substrates: Dishes with U or V-shaped wells with ultra-low attachment characteristics or hanging drop culture are suggested for the initial stages. At later stages, there is more freedom of dish choice – petri dish, micro-wells, and microfluidics [13, 14, 15, 16, 17, 27, 28]. So far, if a substrate is used at all in the protocols, the most commonly used is Matrigel (basement membrane matrix produced from mouse sarcoma cells); it is used either in differentiation or maturation stages or both [13, 28].

Classification: The current brain organoid protocols can be classified in several ways. One of the commonest is based on the approaches for differentiation and can be divided into two groups: 1. Guided (use external agents to direct the patterning) and 2. Unguided (accentuate on the intrinsic developmental programs) (Figure 3) [13, 14, 15, 16, 17, 28]. Another way is based on the anatomical region in which the organoids recapitulate – telencephalic, diencephalic, mesencephalic, and rhombencephalic [14, 15, 16, 17, 27, 28]. Here will be used in a nested fashion.

Figure 3.

Approaches for generation of brain organoids. (a). Unguided - without use of patterning agents; (B). And (C). Guided – Utilizing patterning agents to direct the patterning process. (B). Generating region-specific organoids. (C). Generation of fussed organoids from separately grown guided region-specific organoids.

3. Brain organoid protocols – Regional specification

3.1 Guided brain organoids - telencephalon

The telencephalon is the largest part of the brain in humans. Regardless of its astonishing evolutionary expansion in primates, the telencephalon retains its major subdivisions among the vertebrates: 1. The dorsal telencephalon or pallium, which via the thalamus receives most sensory afferents, and 2. Ventral telencephalon subpallium is mainly involved in motor functions (Figure 4) [29]. Major organizers are the roof and floor plate and ANR, cortical hem, PSB (pallial-subpallial boundary), or antihem. It may seem that the telencephalon organoids should be the easiest to grow and develop since the default fate of the early embryonic ectodermal cells is one of the rostral telencephalic neurons. However, in reality, this presented a challenge. The first experimental protocols for neural progenitor’s induction could derive only neural cells with caudal identity or at best, the midbrain’s one [31, 32]. By the early 2000s, an established and empirically discovered neural inducer was the retinoic acid, which induces only neural cells with posterior identity [33].

Figure 4.

Organizing centers in the brain primordia with their main morphogens. ANR - anterior neural ridge; ZLI - zona limitans intrathalamica; PSB – Pallial-subpallial boundary; LT - Lamina terminalis; MHB - midbrain-hindbrain boundary. After [29, 30].

In the 1990s, few endogenously synthesized molecules were discovered with neural inducing activity, namely Chordin and Follistatin, but they worked for amphibian organisms, while such knowledge was still lacking for the mammalians. One of the leading reasons for this situation was the usage of the animal cap assay as the main workhorse for such discoveries, which is from Xenopus sp., while for mammals, such an assay was missing [32]. Kawasaki et al. in 2000 did an extensive test with a series of perspective molecules for neural induction as FGF2, FGF8, Shh, HGF (hepatocyte growth factor), EGF, PDGF (platelet-derived growth factor), LIF, LiCl (activator of Wnt signaling) on mESC [32, 34]. However, the authors failed to induce neurons with any of them. Instead, they discovered that co-culturing with stromal PA6 cells promotes neural differentiation of mESC, which they named stromal cell-derived inducing activity (SDIA), which seemed to be suppressed by BMP4 treatment (around that time, it was known that BMP4 inhibits neural differentiation [35]. A few years later, Watanabe et al. used a similarly defined medium, but they used SFEB (serum-free floating culture of embryoid body-like aggregates) type culturing instead of attached one [36]. They applied at the beginning during the induction phase Nodal antagonist (LeftyA) since it was found in 2004 that Nodal inhibits ES differentiation to neural fate [37]. In addition, they applied Wnt antagonist (Dkk1), which was in concordance with the neural anteriorizating role of Wnt-inhibitors and caudalizing one of Wnts in the early developmental stages [22]. Later, the authors applied various signaling molecules known to be involved in the brain patterning as Shh and Wnt3a [38, 39].

Eiraku et al. successfully modified the protocol of Watanabe et al. for brain organoids [1, 36]. One of the breakthroughs was shortening the initial aggregation step, which allowed the earlier restoration of the cell–cell interactions. The other facilitating condition was the usage of small bottom micro-well spheroid plates instead of large bottom petri dishes, which improved the aggregation by reducing the number and increasing the size of the formed aggregates. During the late neural induction phase, in the aggregates was formed a cavity through apoptosis similar to those observed by Coucouvanis & Martin in EBs [40]. Thus, forming polarized neuroepithelial structures with an apical surface inside and a basal side outside, which are later divided into smaller neuroepithelial rosettes. Like the mESC culture, the hESC aggregates also formed an internal cavity covered with neuroepithelium; however, they did not reform into smaller rosettes but into mushroom-like shapes. The authors initially applied Dkk1 and LeftyA and later different combinations of FGF8, FGFR3-fc, and BMP4. In this way, they generated polarized tissues recapitulating early corticogenesis of different regions of the forebrain primordium. Few years later, the authors improved this protocol by replacing the partly defined KSR (Knockout Serum Replacement) medium with chemically-defined components [41]. In addition, they added ECM components (as Matrigel, fibronectin, laminin, and laminin/entactin complex) from day 1, which significantly improved the morphological stability of the organoids, especially the later complex. Initially, they applied IWP2 (Wnt inhibitor) and later optionally Shh and FGF8. Thus, they generated cortical neuroepithelium resembling cortical hem-like or lateral ganglionic eminence–like tissues (with Shh addition).

Paşca et al. proposed another approach for generating cortical organoids [42]. Instead of using some of the classical morphogens, they relied on the default neural fate of the ectoderm and the use of some mitogens and growth factors that were found to be able to induce differentiation. Initially, they applied the dual SMAD-inhibition strategy. Afterward, a combination of FGF2 and EGF was applied. Earlier studies have shown the mitogen action of FGF2 and EGF, plus for the EGF, it was found that at later stages, it can also induce differentiation towards neurons and astrocytes [43, 44]. In the next step, the organoids were transferred in a medium with neurotrophic factors BDNF and NT3. As a result, they generated cortical organoids with neurons and astrocytes analogous to the late mid-fetal stage (19–24 weeks post-conception). Next year Qian et al. developed a series of protocols for region-specific organoids [45]. The first type was the forebrain one. They started with a dual SMAD-inhibition strategy (dorsomorphine and A83–01) without morphogens. For the patterning stage, they used CHIR99021, SB431542, and Wnt3a and embedded the organoids in Matrigel. In the next stage, the organoids were detached from the Matrigel and placed in a custom spinning bioreactor. For the maturation stage, growth factors such as BDNF, GDNF, etc., were added. As a result, the authors acquired forebrain organoids with all six cortex layers, which were identified by their respective markers as REELIN, CUX1, BRN2, SATB2, CTIP2, or TBR1.

3.1.1 Guided brain organoids - dorsal telencephalon

The dorsal telencephalon or pallium, for most of its parts, has laminar organization. It can be subdivided into dorsal pallium (isocortex), medial pallium (hippocampus), and lateral pallium (olfactory cortex). The mostly non-neuronal formation of choroid plexus is also included in this region [29, 46]. Kadoshima et al. improved the protocol of Nasu et al. for human dorsal telencephalon organoids [41, 47]. They used the same inhibitors but with tripled cell density. They added FBS and Matrigel in the final step, which improved the long-term growth. In this way, they got organoids of cortical neuroepithelium recapitulating the fetal cortex in the second trimester. Often these organoids formed spontaneously intracortical dorsocaudal-ventrorostral polarity. The neuroepithelium flanking tissue often had cortical hem markers. In a dose-dependent manner with SAG (Shh agonist), they could imitate dorsoventral gradient and generate LGE (lateral ganglionic eminence) and MGE (medial ganglionic eminence) type cells.

Further Mariani et al. optimized the protocol for human-induced pluripotent stem cells (hiPSC) [48]. They used 3-fold higher cell density and Dkk1, SB431542, and BMPRIA-Fc as inhibitors of Wnt, TGF-β/activin/nodal, and BMP pathways. They acquired tissues with a dorsal pallial telencephalic identity corresponding to embryonic human cerebral cortex 8−10 weeks post-conception with stratifying cytoarchitecture, including radial glial cells expressing neural progenitor proteins, intermediate progenitors, and maturing neurons. Later, they used this protocol for patient-specific organoids for autism research [49]. Choroid plexus develops on the dorsal side of the brain ventricles, and its chief functions are the secretion of the cerebrospinal fluid (which has important roles in the homeostasis and neurogenesis) and barrier for the molecular exchange with the vascular system [50]. Therefore, it is the desired target for researchers and medicians alike.

Sakaguchi et al. used the procedure of Kadoshima et al. as a foundation for the production of additional region-specific organoids [51]. At that time, it was known that Wnt and BMP signaling had a leading role in the telencephalon patterning of that region during embryogenesis [52, 53]. First, they explored the ways to generate choroid plexus organoids by using a series of BMPs and a Wnt agonist (CHIR 99021) with different concentrations. As a result, it was found that BMP4 plus CHIR 99021 is the most potent combination, followed by BMP2 and BMP7, and treatment only with CHIR 99021 produces hem-like tissues. The authors also found that the choroid plexus or cortical hem organoids start to produce the patterning ligands characteristic for these organizers in the embryo. Further, they attempted to generate hippocampal primordium tissues. This was achieved by shortening the exposure time to BMP4 plus CHIR 99021, with medium change and additional oxygen supply.

Pellegrini et al. used the protocol for unguided brain organoids of Lancaster et al. to develop a scheme for choroid plexus organoids [13, 54]. The forebrain organoids often have a small percentage of cells with choroid plexus identity. To increase this ratio, an uncommon scheme with pulsed treatment was developed. The authors used the protocols of Watanabe et al. and Sakaguchi et al. to select BMP4 and CHIR 99021 as patterning factors [51, 55]. The generated organoids formed cysts filled with a liquid whose content recapitulates one of the cerebrospinal fluids. Also, the authors found tight barrier formation, the medium and intra-organoid fluid with junction types and transporters characteristic of the native choroid plexus. The marvel of this protocol is that for the first time, such regional organoids could secrete fluid that emulates cerebrospinal fluid and form structures with brain-barrier-like functions. This provides the researchers with a unique platform to study in vitro the mechanisms of the secretion of cerebrospinal fluid or to develop better drugs capable of crossing the blood–brain barrier.

3.1.2 Guided brain organoids - ventral telencephalon

The ventral telencephalon or subpallium has complex and rather dynamic organization and structure during embryogenesis. At E12.5 in mouse, its main subdivisions are: lateral ganglionic eminence (LGE) (produces the striatal components), medial ganglionic eminence (MGE) (pallidum proper and produces globus pallidus), AEP (peduncle/internal capsule, and produces many sublenticular components of the extended amygdala), POC (commissural preoptic area) [56]. So far, specific protocols recapitulating LGE and MGE are developed as described below. Xiang et al. explored the ways to generate brain organoids recapitulating the medial ganglionic eminence [57]. They upgraded the protocol of Maroof et al. for hESC-derived GABAergic interneurons and the one from Nicholas et al. for human pluripotent stem cells (hPSCs)-derived forebrain interneurons [58, 59]. At the neural induction stage, they applied dual SMAD and Wnt inhibitors. For the ventral patterning, they used Shh and purmorphamine. After over 70 days in culture, the MGE organoids comprised diverse cell populations mostly with MGE identity, the largest portion being intermediates or not yet committed, and a quarter were interneurons.

Cederquist et al. applied an unusual approach to generate MGE organoids [60]. They genetically modified a small batch of hPSCs to express sonic hedgehog (Shh) in the presence of doxycycline. Afterward, these cells were dissociated and re-aggregated in small aggregates. On top of them were aggregated 10-fold more hPSCs in this way was created big chimeric spheroid from small Shh producing cell cluster surrounded by a larger mass of non-genetically modified hPSCs. During the neural induction period, dual SMAD and Wnt inhibition and doxycycline were used. In the regional patterning phase, only doxycycline was used. With immunohistochemistry, the authors showed the formation of Shh gradient spreading from the artificial organizer and also that there is regionalization of the organoids recapitulating the dorsoventral patterning of the telencephalon with at least five domains i.e., close to the organizer, the cells had an identity of ventroposterior hypothalamus, next was anterodorsal hypothalamus, then medial ganglionic eminence (MGE) and lateral ganglionic eminence (LGE), and neocortex.

Birey et al. utilized the protocol of Pasca et al. for cortical organoids as the base [42, 61]. In order to get more ventralized organoids, they added IWP-2 (Wnt inhibitor) from the late third of the neural induction stage onwards. And a week later, they added Shh activator (SAG). After 105 days of growth and maturation, the subpallium organoids contained cells with markers for ventral neural progenitors and GABAergic cells. Miura et al. used, as a starting point, the activin-based protocol of Arber et al. for striatal neuron induction in 2D culture to develop a protocol for striatal brain organoids [62, 63]. So, for regional differentiation, initially, they added activin A and IPW2 (Wnt inhibitor). In addition, through transcriptomic analysis, they found that the retinoid X receptor gamma (RXRG) is highly expressed in this region. They added RA agonist (SR11237) and optimized the concentration and time window and found that a week post neural induction phase to be the best time for the agonist application. Afterward, they did a single-cell transcriptomic analysis of the organoids and found that most of the cells were with LGE identity, with over half being GABAergic neurons.

3.2 Guided brain organoids - diencephalon

Diencephalon has still debatable embryonic organization and subdivisions. In mouse at E13.5, it is divided into the hypothalamus, prethalamus, epithalamus, thalamus, and pretectum (Figure 3) [64]. Major organizers dorsoventrally are the roof and floor plates and transversely – ZLI (zona limitans intrathalamica). Currently, there are protocols for the hypothalamus, prethalamus, and thalamus as follows.

Shiraishi et al. developed a protocol for stem cell generation of tissues resembling major parts of the diencephalon, and although they did not label them as organoids, it is valuable to be included here [21]. A key component in their rationale was the introduction of the FGF signaling pathway inhibition since FGFs were local caudalizing factors for that region. Their tests showed that the partial intracellular inhibition of MAPK/ERK kinase (MEK) with PD0325901 is better than the total inhibition through the cell membrane FGFR receptor with SU5402. However, these inhibitors brought mediocre gain of cells with thalamic markers. A recent finding by Suzuki-Hirano et al. for strong transient expression of BMP7 in this region helped them to increase the gain significantly by adding BMP7 [65]. Looking further on how to get more differentiated local regionalization, they experimented with activators of Shh (SAG) and Wnt (CHIR 99021) (which are expressed by ZLI). By adding PD0325901 and BMP7 on day 4 and replaced with SAG on day 7, they acquired cell populations from the prethalamus, thalamus (from ventricular and mantle zones), and pretectum in a single sample. When instead SAG was used CHIR99021, it led to an increase of cells with prethalamic markers.

Xiang et al. adapted the protocol of Shiraishi et al. for human forebrain organoids [66]. They changed the Shiraishi et al. timeline and protocol by using dual SMAD inhibition and insulin during the neural induction phase. While for the thalamic patterning phase, they removed them, and it was necessary to inhibit the MEK–ERK signaling (with PD0325901) to prevent excessive caudalization and BMP7 as it promotes thalamic differentiation. They obtained thalamic organoids with strong expression of the thalamic markers TCF7L2 and GBX2. Medina-Cano et al. in order to shorten the time for organoid generation, used epiblast-like cells (EpiLCs) instead of the common blastocyst stem cells [67]. During the first 2 days of neural induction and anterior-posterior patterning phase, they applied dual SMAD along with Wnt and FGF inhibition. For another 2 days during the neuroepithelium expansion, the inhibitors were replaced with FGF8b. From day one till five, the EB were embedded in Matrigel and afterward transferred to an orbital shaker. They found that addition of high levels of FGF8b and no BMP7 (as in [66]) generates organoids with prethalamus identity with higher efficiency.

Ozone et al. improved an adapted for 3D culture previous protocol for 2D hypothalamic cells differentiation [68]. They used twice less KSR supplement for the neural induction medium. While for the patterning phase were, added Shh agonist (SAG) and BMP4. Thus, they gained tissues with the expression of ventral hypothalamic markers. Next protocol developed by Qian et al. was for hypothalamus organoids [45]. They employed dual SMAD inhibition for neural induction. For the specific patterning, they applied Wnt3a, Shh, and purmorphamine. At the differentiation stage, the organoids were transferred to a spinning bioreactor supplied with FGF2 and CTNF (ciliary neurotrophic factor). As a result, the generated organoids after the patterning phase expressed markers for the early hypothalamus development, which later matured to specific hypothalamus cell populations.

3.3 Guided brain organoids - midbrain or mesencephalon

The mesencephalon mainly can be divided (in mouse at E11) into dorsal part - tectum and ventral one - tegmentum. From the ventral midbrain primordium arises substantia nigra, which plays important roles in controlling movement and sensory processing [69, 70]. Its pathology is the key factor in some widespread diseases with a great socio-economic burden as the Parkinson’s one [71]. Therefore, significant interest was paid to the midbrain organoids; however, the majority of them are focused on the ventral region (Figure 4). Major organizers dorsoventrally are the roof and floor plates and transversely – MHB (midbrain-hindbrain boundary).

Tieng et al. modified 2D culture protocol for midbrain dopaminergic neurons [72, 73]. In the induction phase is used dual SMAD-inhibition plus a cocktail of Shh, purmorphamine (Shh activator), and FGF8 to promote floor plate identity. Shortly, Wnt activator (CHIR99021) is added afterward. The authors also found a positive effect on the maturation by inhibiting Notch pathway, which is known to arrest the proliferation and promote differentiation.

Jo et al. used a slightly different induction and patterning scheme than Tieng et al. by changing the treatment time windows [14]. At the start of the dual SMAD-inhibition, they also added Wnt activator (CHIR99021). On day four were added Shh and FGF8. At the tissue growth phase was added Matrigel and organoids were placed on orbital shaker till the end of the cultivation. In this way, the authors obtained midbrain organoids with neuromelanin-containing and dopamine-secreting neurons. The last protocol developed by Qian et al. was for midbrain organoids and was also based on the 2D protocol for dopaminergic neurons by Kriks et al. [45, 73]. Here, they again used dual SMAD inhibition, but supplied from the beginning with FGF8, Shh and Purmorphamine. Later CHIR99021 was added followed by removal of FGF8, Shh and Purmorphamine and SB431542. Afterwards the organoids were transferred into spinning bioreactors and supplied with growth factors. The obtained midbrain organoids had approx. 50% cells with dopaminergic markers and even more with floor plate identity.

Monzel et al. modified the 2D protocol by Reinhardt et al. for the differentiation of neural precursor cells [74, 75]. Initially, hESC was treated with SB431542, dorsomorphin, CHIR99021, and purmorphamine, and after day four, the first two molecules were removed, and ascorbic acid was added. Thus, after several days, human neuroepithelial stem cells were generated and plated on Matrigel for maintenance. Next, for differentiation into the midbrain, dopaminergic fate CHIR99021 was substituted with FGF8 for 8 days. For the organoid culture, these differentiated cells were dissociated and re-aggregated in the same maintenance medium. On day eight, the aggregates were transferred to Matrigel droplets, and on day 10 were placed long-term in a differentiation medium (with purmorphamine till day 16) on an orbital shaker. After 2 months in culture, the midbrain organoids had over 60% positive midbrain neurons with small percentages of other subtypes and astrocytes.

In an attempt to model Parkinson’s disease with midbrain organoids, Smits et al. adapted this protocol by doubling the length of the dual SMAD-inhibition period and afterward reduced 4-fold the concentrations and used SAG instead of purmorphamine in the initial steps [76]. For the differentiation step, FGF8 was used CHIR99021 in static culture. Consequently, Nickels et al. combined these two protocols in order to improve the reproducibility and viability of the organoids [77]. They achieved this by adjusting the initial cell count and changing the timeline of the treatments with respective signaling molecules.

3.4 Guided brain organoids - rhombencephalon or hindbrain

Rhombencephalon is a major part of the 3-vesicle brain primordium in mammals (E9 stage in mouse). In the transition to 5-vesicle brain primordium (E11 in mouse), it divides into two parts - the rostral part metencephalon, which gives rise dorsally to the cerebellum and ventrally to the pons; and the caudal part myelencephalon which gives rise to the medulla oblongata (Figure 4) [70]. Major organizers dorsoventrally are the roof and floor plates and transversely - MHB and some rhombomere borders. All main regions were recapitulated by the current hindbrain organoid protocols as described below.

Muguruma et al. upgraded their protocol for 2D culture of cerebellar cells [78, 79]. It would be logical if the authors simply applied some of the morphogens synthesized by the dorsal hindbrain primordia and/or the MHB to direct the cells towards cerebellar identity. However, this approach tried by them, or other researchers led to rather a low efficiency of cerebellar cell generation, especially Purkinje ones [80, 81]. So, they tried somewhat to recapitulate induction of the isthmic organizer, which in turn secretes the needed signaling molecules for cerebellar differentiation. They adapted the strategy of Wataya et al. and used a high dose of the weak caudalizing agent insulin followed by FGF2 (which moderately increased the expression of Wnt1 and FGF8) [78, 82]. Along with that, they applied SB431542, the concentration of which was reduced after a week together with FGF2 [78]. In addition, they found that exogenous FGF19 and SDF1 (stromal-derived factor-1 synthesized by the adjacent meninges) can facilitate cerebellar plate formation.

Brain stem organoids protocol was developed by Eura et al. based on the one from Paşca et al. [42, 83]. Initially, the aggregates were treated with FGF2, EGF, and insulin along with dual-SMAD inhibitors. In addition, they added progesterone and transferrin (engaged in iron metabolism) known to promote dopaminergic differentiation and protection from recent studies [84, 85, 86]. On day 22, the used patterning and growth factors were replaced with a cocktail of neurotrophic factors and other small molecules, promoting differentiation and maturation for another week. Afterward, no growth factors were added. In this way, they generated human brainstem organoids (hBSOs), containing midbrain/hindbrain progenitors, noradrenergic and cholinergic neurons, dopaminergic neurons, and neural crest lineage cells. Molchanova et al. developed a protocol for rostral brainstem organoids in a two-step approach [87]. First, they differentiated the stem cells into caudalized human neuroepithelial ones using dual SMAD inhibition along with Wnt and Shh activation. Afterward, the neurospheres were dissociated and re-aggregated to initiate the formation of hindbrain organoids. The caudalized aggregates were treated for a week with Wnt and Shh activators for further differentiation before transfer to the shaker for maturation. After two and 4 months, the organoids had large cell populations with pons and medulla oblongata identity and smaller populations of astrocytes.

Valiulahi et al. used as a starting point the protocol of Kirks et al. for generation of dopamine neurons [73, 88]. Instead of using caudalizing Wnt agonist, they decided to try with the strong caudalizing agent RA. To determine the optimal time window and concentration series of tests were performed and was found that max concentration RA from day one to 13 gives the highest percentage of 5-HT neurons. For the 3D brain organoid protocol, they experimented to find the optimum between RA plus purmorphamine treatment start and the generated structure and set it to day 5 till 13. So, in the end, the dual SMAD inhibitors were omitted in the neural induction phase. Before the end of the patterning phase, the organoids were embedded in Matrigel and later placed on a shaker. As result, they acquired organoids with diverse cell identity which overall resembles the ones from the caudal rhombomeres R5−R8 (which later and the largest population (30–40%) being 5-HT-expressing neurons.

3.5 Unguided brain organoids

Lancaster et al. were the first to develop unguided brain organoids [13]. As a base, they used the protocol of Xia et al. for neuroepithelial differentiation [89]. They added a low concentration of FGF2 and ROCK inhibitor at the initial stage. After 6 days, EB was transferred to a neural induction medium and they began forming neuroepithelial tissues. At the differentiation step, differentiation medium and Matrigel embedding were used. In the final stage, the organoids were transferred to a spinning bioreactor and RA was added. The generated brain organoids were rather large (up to 4 mm) with complex heterogeneous tissues resembling various brain regions such as forebrain, midbrain, and hindbrain.

As you can see the majority of the organoid protocols utilizes only a handful of morphogens and signaling molecules related to the main signaling pathways engaged in the embryonal brain patterning. Among the key points for the success of the protocol is to find the right combination of signaling molecules and their right concentration, the right time and length to apply them. Although most researchers do not publish their error-and-trials reports until they get to the right combination, few do, as Xiang et al. [57, 88], which can be of great help for the newcomer in the field of brain organoids. The quest for better organoids, the demands for cheaper and more reproducible organoids along with other factors in the last few years inspired the researchers to experiment with and introduce some bioengineering approaches in the generation of brain organoids. Howbeit, due to volume limitations here, the reader is encouraged to honor the excellent review by Yi et al. dedicated to the bioengineering approaches in organoids in general [90].

4. Challenges on the path of brain organoids technology

Brain organoid technology is a little over than decade old, and it is still in its infancy. Therefore, it is no surprise that the generated brain organoids still suffer from significant discrepancies compared to the native brain. Here will be discussed some of the important issues.

Vascularization: The brain is one of the most metabolically active organs in the body requiring a high oxygen supply; therefore, it has one of the densest vascularization networks [91, 92]. In most of the current protocols, the brain organoids lack vascularization. Lack of organoid vascularization of such presumably active tissues worsens the viability of the cells, causes necrosis, limits the organoid size, and disturbs the tissue structure [93, 94]. The trouble with the vascularization is mostly due to the fact that the vascular epithelia originate from different germ layers than the neurons and the macroglia [95]. In a recent report, Pham et al. used a dual culture approach by separate differentiation of endothelial cells, which afterward were co-cultured with the formed brain organoids [95]. However, further research is needed to determine how functional the brain organoids are and how well they recapitulate the native brain vascularization. Another proposed approach by Mansour et al. was the implantation of the organoid in the living brain in order to be provoked vascularization from the surrounding tissues, thus securing its long-term survival [96].

Artificial maturation/aging strategies: Currently, most of the organoids develop and differentiate with comparable speed to the one observed during the natural neurogenesis for the particular species [47]. This can be problematic in some applications. For example, if we want to produce an appropriate neural patch for an injured patient using his/her own hiPSC, then this patch generation should happen sufficiently fast in order for the patch to serve its purpose. Alternatively, if we want to study aging-related diseases such as Parkinson’s. Borghese et al. used the Notch-inhibition strategy to accelerate the neuronal differentiation of ESC in vitro and in vivo [97]. However, Notch signaling plays an important role in brain patterning and neuronal specification so that such inhibition may interfere with the desired differentiation results. Miller et al. used progerin (truncated form of lamin, which is associated with premature aging) to induce aging human iPSC [98]. Later Vera et al. proposed another approach by downregulation of telomerase which induces telomere shortening [99]. Recently, Li et al. developed organoids with accelerated growth by using mutant cells; however, such genetically modified organoids cannot be used in translation [100].

Morphological discrepancies with native brain: One of the commonest problems with current brain organoids is that within a single organoid are formed several neuroepithelial rosettes, each acting as a separate center of morphogenesis, while in the normal embryo is formed only one. Recently Knight et al. proposed a protocol for single rosette generation [101]. They achieved a high percentage of single rosette organoids by imposing geometrical confinement on the growth in custom micro-patterned plates and/or with ROCK-inhibition.

Glial cells: Most of the brain organoids have little or no glial cells. This is not a big surprise if we consider the gliogenesis timeline during the normal ontogenesis. Significant amounts of astrocytes start to be generated only at the late stages of the embryogenesis, while for the oligodendrocytes, this time starts with the postnatal period [102, 103]. Howbeit, the situation is very different for the microglia, which start to invade the neural tube relatively early in the embryogenesis, and by its end, all the microglia are present [103]. They are absent from the brain organoids because they are produced by different germ layers than the neuroectoderm. So, the lack of glia in the brain organoids is logical, and if we want brain organoids to be better copies of the real brains, then we either need to grow them for a comparable time length of the in vivo brain development or to find shorter ways (for the macroglia) or to introduce somehow the missing germ layer in the system (for the microglia).

There is some recent progress in the resolution of these issues. Ormel et al. modified the unguided protocol of Lancaster et al. by tweaking its timeline and reducing the concentrations of some of the additives, and as a result, they got whole-brain organoids with innate microglial cells [104]. However, this protocol may not be a feasible solution for the researchers who need more specific organoids. Later Bejoy et al. developed a different strategy by dual culturing the brain organoids and the microglia mesodermal progenitors and afterward co-culturing them so that the microglia can invade the organoids [105]. As for the macroglia, Paşca et al. and later Yakoub generated brain organoids that developed together neurons and astrocytes, howbeit they were forebrain specific [42, 106]. Recently Shaker et al. published a protocol for cortical brain organoids that could be used to develop myelinating oligodendrocytes along with astrocytes [107]. All these are encouraging results, so hopefully, we expect to see organoids with full-spectrum glial cells in the future.

5. Concluding remarks

For its short existence, the brain organoid technology made respectable progress and drew lots of interest. It became a rather efficient tool and platform to improve the scientific knowledge of the brain evolution and development and to address clinical problems in a general or personalized approach for brain-related pathologies [13, 45, 48, 49, 108].

Howbeit, there is still large room for improvement and development of this technology since many of its issues wait for practical and efficient solutions and applications.

The functional vascularization of the organoids is probably the most penalizing challenge. When it gets solved, this will open the door for bigger and better replicas of the brain with wider applications as the vascularization network is not only a nutrition/metabolite carrier but also an active part of brain development [109].

Most of the current brain organoids are relatively simple mimics of the much more complex brain. For example, at the moment we cannot recapitulate the overwhelming complexity of such a minute brain region as hypothalamus, which is of comparable size to the biggest brain organoids. So far, it is known that tens of signaling molecules dynamically drive the patterning of hypothalamus primordium during early embryogenesis, many of them distributed in a gradient manner with respectable precision [110]. However, the existing hypothalamic brain organoid protocols utilize just two of them in extrinsically uniform concentration. The generated hypothalamic organoids are coarsely differentiated with limited heterogeneity and cell diversity. At later stages, it seems that the control of hypothalamus patterning is also rather complex and poorly known and hard to be recapitulated. Probably better knowledge and experimenting are needed to get an improved recapitulation of this brain region. A promising move in this direction is recent single-cell analysis studies that hold the promise to improve our understanding of the hypothalamus patterning mechanisms [111]. The situation is not very different for the other brain regions than the hypothalamus.

Wider adoption of engineering techniques can help for better and more natural, dynamic control of the microenvironment [90]. They can also improve reproducibility, automate and reduce the costs per organoid.

Another approach to increase the brain organoid complexity is through the fusion of separately grown region-specific ones, thus forming complex structures named assembloids (Figure 3C) [112]. Howbeit, this approach is rather young and so far had found limited success and application.

Although the present of the brain organoid technology still seems challenging, looking at the astonishing advances in the embryology and microtechnologies, there is hope that its future could be encouraging with improved products and applications.

Conflict of interest

“The authors declare no conflict of interest.”