Production of Highly Uniform Midbrain Organoids from Human Pluripotent Stem Cells

Brain organoids have been considered as an advanced platform for in vitro disease modeling and drug screening, but numerous roadblocks exist, such as lack of large-scale production technology and lengthy protocols with multiple manipulation steps, impeding the industrial translation of brain organoid technology. Here, we describe the high-speed and large-scale production of midbrain organoids using a high-throughput screening-compatible platform within 30 days. Micro midbrain organoids (µMOs) exhibit a highly uniform morphology and gene expression pattern with minimal variability. Notably, µMOs show dramatically accelerated maturation, resulting in the generation of functional µMOs within only 30 days of differentiation. Furthermore, individual µMOs display highly consistent responsiveness to neurotoxin, suggesting their usefulness as an in vitro high-throughput drug toxicity screening platform. Collectively, our data indicate that µMO technology could represent an advanced and robust platform for in vitro disease modeling and drug screening for human neuronal diseases.


Introduction
Drug discovery is a high-risk but high-return endeavor, with the vast majority of drug candidates failing to demonstrate efficacy or being associated with unexpected toxicity during preclinical and clinical trial testing [1].Additionally, many drug candidates exhibit a lack of efficacy during clinical trials despite demonstrating therapeutic effects in animal models [2,3], marking interspecies differences as another hurdle to overcome for efficient drug discovery.Induced pluripotent stem cell (iPSC) technology [4] has recently been considered as a useful tool for not only reproducing patient-specific pathophysiology [5,6] but also minimizing the use of preclinical animal models, potentially enabling highly efficient and cost-effective patient-specific drug development [7].However, 2D cell culture-based iPSC technology is associated with another critical problem: the incomplete recapitulation of disease pathophysiology due to its limited capacity for reproducing both the in vivo niche and cellular interactions [8,9].
Organoid technology, a recent breakthrough in the stem cell field, has been highlighted as the most innovative stem cell technology for understanding pathological developmental processes and the mechanisms underlying various diseases [10].Indeed, recent studies using organoid technology have unveiled the mechanism of several diseases including the infamous coronavirus 2019 (COVID- 19) disease and have discovered novel therapeutic candidates by closely recapitulating disease-specific pathophysiology at the organ level [11].Notwithstanding, industrial translation of organoid technology, including brain organoids, has been hampered by multiple technical roadblocks such as: (1) lack of proper culture technology for producing adult organ-like mature organoids [12,13]; (2) biased cellular composition [14,15]; (3) low efficiency and reproducibility [16,17]; (4) variability among individual organoids both within and across batches [16]; and (5) frequently observed uncontrolled differentiation, leading to outgrowths [18,19], and necrotic dead core caused by insufficient delivery of oxygen and nutrients to the organoid core due to lack of a vascular system [20,21].Moreover, the process of generating relatively mature brain organoids takes at least 3-6 months and involves multiple manipulation steps such as seeding the starting cells, embryoid body formation, embedding into the extracellular matrix, and using an orbital shaker to facilitate organoid maturation [14,15,22,23].Such a lengthy process with multiple manipulation steps, together with the aforementioned technical roadblocks in brain organoid production, results inevitably in an extremely high cost for producing brain organoids.Therefore, the development of key technologies that could overcome all the hurdles and limitations associated with organoid generation is required for the industrialization of organoid technology toward drug screening for human neuronal diseases.
To facilitate the industrialization of brain organoid technology, the current study aims to develop the high-speed and large-scale standardized production protocol for the generation of midbrain organoids using a high-throughput screening-(HTS-) compatible platform.Under this platform, we were able to generate micro midbrain organoids (µMOs) within 4 weeks of differentiation.Compared with typical midbrain organoids, µMOs exhibit a uniform and consistent morphology, size, and gene expression pattern.Strikingly, µMOs exhibit dramatically accelerated maturation within only 30 days of differentiation, such as the homogeneous distribution of mature dopaminergic neurons, presence of other in vivo-like cellular compartments including glial cells, production of neuromelanin-like granules, and mature electrophysiological features.The highly uniform µMOs produced by a high-speed and industry-scale methodology also facilitate in vitro drug toxicity screening.Our novel and robust protocol for generating µMOs may represent a key technology for the industrial application of brain organoids such as high-throughput drug screening for Parkinson's disease and other neurodegenerative diseases.

Results
2.1.One-Step Generation of Micro Midbrain Organoids Using a High-Throughput Platform.To overcome the technical roadblocks associated with brain organoid technology, such as a lengthy procedure with multiple manipulation steps, the lack of a standardized protocol for producing uniform brain organoids with consistent quality, variability in organoid morphology, and other features within and across batches, and the need for a highly scalable production platform, we first tried to establish a one-step simplified protocol for producing midbrain organoids (MOs) using a high-throughput screening-(HTS-) compatible platform such as 96-and 384well plates.As we used a high-throughput platform for all the MOs generation steps without transferring MOs into other cell culture formats, we sought to optimize the starting cell number (5 × 10 2 , 1× 10 3 , 3× 10 3 , 5× 10 3 , 7× 10 3 , and 1 × 10 4 ) by making slight modifications to our previous MO protocol [23] (Figure 1(a)).We did not make use of the orbital shaker in this study even during the organoid maturation step because a proper flow could not be created in a high-throughput platform using a typical orbital shaker.
We first evaluated the morphology and gene expression patterns of early MOs on DIV (days in vitro) 7. MOs generated from a different number of starting cells displayed a uniform morphology but distinct size based on the starting cell number (Figure S1).MOs on DIV 7 showed inactivation of pluripotency markers without activation of mesendodermal genes (Figure S1).In contrast, MOs generated from a different number of starting cells showed activation of ectodermal lineage markers whereas MOs generated from a low number of starting cells (5 × 10 2 and 1 × 10 3 ) showed higher expression of early neuronal markers such as N-CAD, PLZF, SOX2, and FGF8 compared with MOs derived from a highcell number (7 × 10 3 and 1 × 10 4 ) (Figure S1).This result indicated that MOs generated from 5 × 10 2 to 1 × 10 3 starting cells, which have relatively higher surface-to-volume ratios, are more sensitive to neuroectodermal differentiation cues than MOs from higher cell numbers.However, MOs on DIV 14 displayed quite diverse morphological patterns based on their starting cell numbers (Figure 1(b)).MOs generated from low-cell numbers (5 × 10 2 , 1 × 10 3 , and 3 × 10 3 ) showed uniform and consistent morphology containing developing neural rosettes (Figure 1(b)).In contrast, MOs formed from high-cell numbers (5 × 10 3 , 7 × 10 3 , and 1 × 10 4 ) displayed relatively poor quality and uncontrolled outgrowth formation (Figure 1(b)).The expression levels of apoptosis markers increased in a starting cell number-dependent manner, lending support to our finding (Figure S1).Both forebrain (FOXG1 and LHX2) and hindbrain (HA1 and HB4) markers were suppressed in all conditions (Figure 1

Stem Cells International
in MOs generated from all numbers of starting cells (Figure 1(d)).
Similarly, midbrain-specific dopaminergic (mDA) neurons expressing tyrosine hydroxylase (TH) were observed in MOs from all conditions (Figure S1).However, the expression levels of TH were relatively lower in MOs generated from high-cell numbers compared with MOs from low-cell numbers (Figure 1(d)).
To determine whether our high-throughput platform without the use of a typical orbital shaker is sufficient for producing MOs containing mature mDA neurons, we induced the maturation of MOs.Similar to MOs on DIV 14, MOs on DIV 21 that had formed from a low number of starting cells (5 × 10 2 and 1 × 10 3 ) showed a uniform morphology (Figure 1(b)).However, MOs that had formed from a high number of starting cells (3 × 10 3 , 5 × 10 3 , 7 × 10 3 , and 1 × 10 4 ) were of relatively poor quality, resulting in severe outgrowths in more than 60% of MOs (Figure 1(b)) consistent with the increased expression levels of apoptotic markers and increased number of caspase 3+ cells in MOs obtained from a high number of starting cells (Figure S2).Although the size of MOs obtained from a low number of cells was significantly smaller on DIV 7, MOs from DIV 14 to 21 showed similar size regardless of the starting cell number (Figure 1(c)), indicating that the number of starting cells does not affect the growth of the resultant MOs after the maturation step.Indeed, MOs from all conditions exhibited a similar gene expression profile in which midbrain-specific markers were specifically activated at both the RNA and protein levels (Figures 1(e) and 1(f ) and Figure S2), however, forebrain and hindbrain markers were not expressed in all group of MO (Figure S2).Although the numbers of TH + mDA neurons are similar in MOs from all conditions on DIV 21 (Figures 1(e) and 1(f )), we observed dramatic differences in MOs on DIV 30 (Figures 1(g) and 1(h)).We found a highly enriched and homogeneous distribution of mDA neurons in MOs derived from 5 × 10 2 cells but a low number of mDA neurons in MOs from other conditions.This data indicate that 5 × 10 2 cells is an optimal number of cells to start with for producing homogeneous MOs using a highthroughput platform (Figure 1(i) and Video Clip 1).For the rest of the analysis we focused on MOs generated from a starting population of 5 × 10 2 cells and we hereafter refer to MOs from 5 × 10 2 cells as micro MOs (µMOs).Taken together, our data indicate that the generation of MOs from a low number of starting cells 5 × 10 2 would be beneficial for both the robust specification into the midbrain and prevention of undesired differentiation into other brain regions, efficiently inhibiting the formation of both outgrowths and apoptosis level in our experimental setting using a high-throughput platform.
The presence of neuromelanin-like granules accumulating in the substantia nigra pars compacta in the human midbrain is a major criterion for assessing the functional maturation level of MOs [15,24].Although we previously described the homogeneous distribution of neuromelaninlike granules in our MOs [23], in general, the accumulation of neuromelanin-like granules was only partially observed in a limited area in MOs even after a long-term maturation phase, i.e., 70 DIV [23] and 80 DIV [15] of differentiation.Thus, we next assessed whether µMOs after only 30 DIV of differentiation could already show the accumulation of neuromelanin-like granules in the presence or absence of dopamine (Figure 3(a)).For this, we treated µMOs on one half of a 96-well plate with dopamine for 10 days and omitted dopamine for the other half.To our surprise, we readily observed the accumulation of neuromelanin-like granules on the outside of µMOs, indicating that mDA neurons might secrete neuromelanin-like granules (Figure 3(b)).All the dopamine-treated µMOs showed the production of neuromelanin-like granules and individual µMOs displayed uniform accumulation patterns in which virtually 100% of µMOs successfully produced neuromelanin-like granules in a similar manner, suggesting the functional homogeneity of individual µMOs (Figure 3(b)).Using Fontana-Masson staining, we also confirmed the presence of neuromelaninlike granules in µMOs but not in cerebral organoids (COs) that had been matured during the same period (Figure 3(c)).Notably, µMOs that were not treated with dopamine also rarely exhibited Fontana-Masson-positive granules (Figure 3 (c)).As neuromelanin-like granules are rarely observed in typical MOs even after a long-term maturation step (usually 70-80 DIV after differentiation) [15,23], we concluded that the functional maturation of µMOs is much faster than that of typical MOs.Furthermore, we also confirmed the production of dopamine in µMOs after only 30 days of differentiation, which is typically observed at 96-140 days of differentiation using previously described protocols [23] (Figure 3(d)), indicating that µMOs could functionally mature within 30 DIV of differentiation.Similarly, dopamine release could be readily observed in µMOs at DIV 30, although its level was further increased in µMOs at DIV 60 (Figure 3

Stem Cells International
firing rate compared with µMOs from DIV 20 (Figures 3(g) and 3(h), Figure S5).This data indicate that the number of synapses in µMOs is substantially increased within 30 DIV of differentiation.Furthermore, µMOs from DIV 30 also showed increased burst activity and burst duration compared with µMOs from DIV 20 (Figure S5), also indicating that the neural network of µMOs could be further strengthened by the increased number of synapses within 30 DIV of differentiation.By analyzing signal synchronization and neural network of µMOs from DIV 20 to 30 (Figures 3(i) and 3(j)), we also found that µMOs from DIV 30 show significantly strong neural network formation compared with µMOs from DIV 20 (Figure S5).Collectively, µMOs are functionally mature, and their functional maturation could be robustly achieved within only 30 DIV of differentiation.12 Stem Cells International 2.3.Large-Scale Production of Uniform µMOs with Minimal Variability.Variability in organoids within and across batches and obtained from different cell lines is a major roadblock for translating brain organoid technology into the clinic [16].As the instability of brain organoids has the potential to impede the application of brain organoids for in vitro disease modeling and library-scale drug screening, a standardized and reproducible approach is required for producing uniform brain organoids of consistent quality on an industrial scale [16,17].In the current study, we described the high-speed and large-scale production of functionally mature µMOs containing mature mDA neurons using a high-throughput platform.
We then sought to address the potential variability among individual µMOs from distinct cell lines and batches.Specifically, we explored the consistency of individual µMOs produced in a 96-well plate.We first produced µMOs from H9 hESCs labeled with GFP and then assessed their morphological consistency by measuring the shape and size of individual µMOs using a high-content imaging system (Figure 4(a)).We observed a uniform morphology in virtually 100% of 96 welland 384 well-derived individual µMOs (Figure 4(a)).Moreover, µMOs exhibited a relatively uniform size, ranging from 0.9 to 1.3 mm (Figure 4(c)).In contrast, typical MOs showed relatively inconsistent quality, with heterogeneous morphology and size (Figures 4(b) and 4(c), Figure S6).To measure the qualitative consistency of µMOs at the molecular level, we first isolated mRNA from 96 individual µMOs from a 96-well plate and compared their gene expression pattern.We used typical MOs of three different classes of quality (high, medium, and low) based on their morphological differences as controls (Figure S6).As expected, typical MOs of different qualities displayed distinct transcriptional levels of two representative midbrain markers, TH and ASCL1 (Figure S6).To our surprise, the vast majority of µMOs exhibited higher expression levels of TH and ASCL1 compared with typical MOs with high quality (92.7% for TH and 94.7% for ASCL1) (Figure S6).We next performed high-content imaging of 96 individual µMOs from a 96-well plate to address the homogeneous expression of TH and MAP2 at the protein level (Figure 4  (d)).Using whole-mount imaging of the 96 µMOs in which the abundant neurites expressing both TH and MAP2 are clearly visible (Figures 4(d) and 4(e)), we compared the mean brightness of TH and MAP2 signals from each µMO (Figures 4(f ) and 4(g)).When we did this, we achieved consistent signals of TH and MAP2 in individual µMOs, with 87% and 81% of µMOs exhibiting a similar range of TH and MAP2 brightness, respectively, or even higher than average signals (Figures 4(f ) and 4(g)).Taken together, our data indicate that there is minimal inter-organoid variability with our µMO strategy.
Next, we sought to address the cell line variability in the production of µMOs.To this end, we produced both µMOs and typical MOs from another hESCs (H9) and two lines of our homemade hiPSCs (lines 1 and 2).Compared with typical MOs, µMOs generated from different cell lines showed consistent morphology and size (Figure S6).Moreover, µMOs generated from different cell lines also exhibited higher expression levels of midbrain markers (Figure S6).We also assessed the batch-to-batch variability in µMO generation.For this, we produced µMOs from three independent experiments (H1 hESCs) and compared their morphology and size.µMOs from all three independent batches exhibited uniform morphology and size on DIV 1, 7, 14, 21, and 30 (Figure S6).Furthermore, our RNAseq analysis using µMOs from three independent batches display nearly indistinguishable gene expression profiles of multiple markers examined (Figures S3 and S4).Collectively, our data indicate that our novel strategy might serve as a universal platform for producing highly uniform MOs with minimal variability among organoids within and across batches and obtained from different cell lines.

Highly Consistent µMOs Facilitate µMO-Based
In Vitro Toxicity Assay.Finally, to assess whether our µMOs could serve as a drug-screening platform, we evaluated the potential of µMOs as an in vitro high-throughput toxicity assay platform.To this end, we treated µMOs with 1-methyl-4phenyl-1,2,3,6-tetrahydropyridine (MPTP), a well-known neurotoxin and determined the its effect on µMO morphology, size, and gene expression profile.We treated DIV 30 µMOs for 10 days with different concentrations of MPTP (150, 300, and 500 µM).The MPTP-treated µMOs exhibited clear size reduction in a dose-dependent manner (Figure S7), which could be explained by the increased expression levels of apoptosis-related markers upon MPTP treatment (Figure S7).In contrast, the expression levels of midbrain markers TH and ASCL1 were suppressed upon MPTP treatment, also in a dose-dependent manner (Figure S7).MPTP-mediated cell death could also be confirmed in a dose-dependent manner by the gradually increased number of TUNEL + cells (Figure S7).
In order to explore neurotoxin-mediated toxicity in detail, we next treated rotenone (100 µM) and MPTP (150 µM) on DIV30 µMOs for 2 days and determined their toxicity by assessing cleaved-caspase 3 and TUNEL-positive cells (Figure 5(a)).As a positive control, we treated H 2 O 2 on µMOs for 30 min (Figure 5(a)).In contrast to nontreated µMOs, both H 2 O 2 and rotenone treated µMOs exhibited massive cell death, as evidenced by the increased numbers of both cleaved-caspase 3 and TUNEL-positive cells (Figures 5(b) and 5(c)).As expected, MPTP-treated µMOs exhibited the relatively mild toxicity compared with both H 2 O 2 and rotenone treated µMOs (Figures 5(b) and 5(c)) because MPTP exerts its role as a dopaminergic neurotoxin via astrocytes which exist relatively rarely in µMOs.Notably, the increase in apoptosis also destroys the morphology of the nerve, resulting in the fragmentation of the nerve structure (Figure 5(b)).
To use our µMOs for in vitro high-throughput drug screening, including toxicity screening, individual µMOs should display similar responses to drug treatment.Thus, we next investigated the response of individual µMOs to MPTP treatment.For this, we treated µMOs with MPTP (500 µM) and assessed the drug response of randomly selected µMOs (10 µMOs) by counting the number of TUNEL + cells.In contrast to control µMOs, MPTP-treated µMOs exhibited significant increases in the number of TUNEL + cells (Figure S7).Notably, the number of TUNEL + cells in randomly selected individual µMOs was in a similar range (Figure S7) (10.9% TUNEL + cells in MPTP-treated µMOs), indicating that individual µMOs exhibit a highly consistent response against MPTP.Taken together, our data indicate that µMOs exhibit a highly consistent and dosedependent response to neurotoxin in terms of morphology, size, gene expression pattern, and cell death, rendering µMOs highly suitable for in vitro high-throughput toxicity screening.

Discussion
Brain organoid technology has opened up a new avenue for more precise and reliable in vitro disease modeling and drug screening for various neuronal diseases by closely recapitulating in vivo microenvironments than 2D differentiation protocols that produce mostly a singular differentiated cell type without forming the in vivo-like tissue architecture composed of multiple cellular components [9,26].However, roadblocks remain and impede the translation of brain organoid technology into the clinic or industry.Recently, we and others have successfully described advanced brain organoid technologies for overcoming some of the technical roadblocks associated with brain organoids such as: (1) lowefficiency and reproducibility [16]; (2) biased cellular composition [14,15]; (3) necrotic dead core [20,21]; and (4) heterogeneous structure including undesired differentiation into other tissues [18,19] or other brain regions [27].However, fundamental issues persist and need to be overcome before initiating industrial translation of brain organoid technology such as: (1) a lengthy protocol for producing brain organoids resulting in extremely high-production costs; (2) variability in individual organoids both within and across batches and from distinct starting cell sources, and (3) the lack of a standardized large-scale production platform.
In the current study, we describe a scalable and highspeed method for robustly generating highly uniform MOs.Using our novel strategy, we were able to generate µMOs with uniform and consistent morphology, size (Figure 4(a)-4(c)), and gene expression profile (Figure 4(d)-4(g)).Furthermore, our µMOs also exhibited dramatically reduced Stem Cells International formation of outgrowths (Figure 1(b)), which is a common technical issue in brain organoid production [20,21].Notably, the frequently observed variability among individual MOs, distinct starting cell lines, and different batches were efficiently minimized using our µMO generation method (Figure 4, Figure S6).More importantly, we could generate functionally mature µMOs within only 30 DIVs of differentiation, as evidenced by the formation of neuromelanin-like granules and mature electrophysiological activity (Figure 3).Taken together, our data indicate that the high-speed and large-scale production of MOs with dramatically accelerated functional maturation could be achieved using a highthroughput platform.
A recent study described the production of uniform MOs by employing an automated workflow system, producing automated MOs (AMOs) that are highly uniform in terms of morphology and gene expression pattern [28].However, some challenging issues remained with this previous technology, such as: (1) the requirement of a costly robotic system; (2) the absence of oligodendrocytes and the presence of an extremely rare population of GABAergic and glutamatergic neurons, and (3) the requirement for an additional predifferentiation step to produce neural precursor cells [30].In contrast, µMOs could be robustly and reproducibly generated under a typical lab setting without the need for a costly automation system.Furthermore, using our novel and robust 16 Stem Cells International method, we could reduce the price required for producing MOs by up to 90% compared with previously described protocols [14,15] including ours [23].Our µMOs also displayed an in vivo-like cell type composition, with the abundant presence of distinct subtype neurons, such as GABAergic and glutamatergic neurons, and glial cells, such as astrocytes and oligodendrocytes.Moreover, our strategy does not require any predifferentiation step of hPSCs.
In the previous study [30], the production of uniform AMOs could be achieved using predifferentiated small molecule neural precursor cells (smNPCs) but not using undifferentiated hPSCs.Indeed, although smNPC-derived AMOs were highly homogeneous, AMOs from hPSCs were relatively heterogeneous with significant variability both within and among batches [30].As hPSCs are a major cell source routinely used for in vitro disease modeling and drug screening, a user-friendly method is needed that does not necessitate a costly automation system as well as an additional predifferentiation step.In the line with this issue, µMO technology could be a universal and user-friendly strategy for producing highly uniform and functionally mature MOs with minimal variability among individual MOs both within and across batches and obtained from distinct starting cell sources.
We performed in vitro high-throughput toxicity screening using µMOs to validate the compatibility of µMOs in high-throughput screening.We subjected µMOs to treatment with MPTP, a nigral toxin and then assessed the morphological change of µMOs as a primary readout and their transcriptional status (apoptosis and midbrain markers) as a secondary readout.Following treatment of µMOs with MPTP, we detected a clear dose-dependent size reduction and transcriptional alteration, indicating the potential of using µMO technology to achieve rapid and reliable toxicity screening.Moreover, individual µMOs exhibited a similar response to MPTP treatment, suggesting that they have high-throughput screening compatibility for evaluating the in vitro toxicity of drugs.As functional astrocytes produce MPP+, a metabolized toxic form of MPTP, the MPTPmediated cell death observed in µMOs indicates that our µMOs with functional astrocytes are highly compatible for neurotoxin-based in vitro disease modeling of Parkinson's disease.Collectively, our novel approach for the high-speed and large-scale production of µMOs might represent an advanced platform for high-throughput in vitro disease modeling and drug screening for Parkinson's and other neuronal diseases.
For generating typical MOs, our previously described protocol was used [23].In brief, 1 × 10 4 single cell-dissociated hPSCs were plated per well in ultralow-attachment U-bottom 96-well plates using EBM for generating EBs.At 24 hr after plating, the medium was changed to BGM.The conditions for inducing neuroectodermal commitment, mesencephalon specification, and mesencephalic floor plate specification during the procedure for generating µMOs were used for producing typical MOs.Typical MOs were embedded into growth factor-reduced matrigel (Corning), and matrigel-embedded typical MOs were transferred onto 6-cm petri dishes containing BGM supplemented with 100 ng/ml of FGF8, 2 μM SAG, BMM was replaced every other DIV.A detailed description of stem cell culture, gene expression analysis, cryosection, immunohistochemistry, Fontana-Masson staining, electrophysiology, whole mount staining and high-content imaging, in vitro high-throughput screening of drug toxicity, TUNEL assay, morphological image analysis, is provided in supplemental experimental procedures.4.3.Generation of Cerebral Organoids.To generate COs, 1 × 10 4 single cell-dissociated hESCs were plated on each well of ultralow-attachment U-bottom 96-well plates (Corning).At 24 hr after plating, BGM supplemented with 2 μM dorsomorphin and 2 μM A83-01 were added to EBs.The medium was replaced every other day.After 7 days of differentiation, COs were embedded into growth factor-reduced matrigel (Corning) droplets.Matrigel-embedded COs were transferred to 6-cm petri dishes containing BGM supplemented with 200 ng/ml laminin and 2.5 μg/ml insulin for inducing basal-apical lamination.After 2 days, COs were transferred into ultralow-attachment 6-well plates (Corning) containing BMM and cultured on an orbital shaker (Stuart).The culture medium was replaced every other day.
4.8.Fontana-Masson Staining.The neuromelanin pigment was stained by the FontanaMasson Stain Kit (Abcam) according to the manufacturer's instructions using cryosectioned organoids.4.9.Electrophysiology.To evaluate the functional maturation level of the MOs, we used a 32-channel neural probe.The width and thickness of the probe were 220 and 20 μm.The microelectrodes (14 × 14 μm) were located in the probe's tip, with 30 μm spacing between the electrodes.The microelectrodes were coated with black platinum through electroplating to improve the recording capacity.The average impedance of black Pt-coated microelectrodes was measured 19 AE 1 kΩ at 1 kHz in 0.1 M PBS.To measure neural signals in the incubator, we used a miniaturized recording structure consisting of a custom microdrive to adjust the vertical position of the probe, a PDMS chamber to load the MOs, and an acrylic enclosure to prevent rapid media evaporation [29].After loading the MO in the center of the PDMS chamber, we slowly inserted the neural probe into the organoid using the custom microdrive.After filling the chamber with fresh medium, we put the MO inserted with the neural probe into the acrylic enclosure.We measured neural signals from the MO in the incubator.Neural signals recorded from the neural probe were processed and digitized through an RHD2132 amplifier board connected to an RHD2000 evaluation system (20 KS/s per channel, 0.3-Hz high-pass filter, 6-Hz low-pass filter, and 16-bit ADC).Neural signals were recorded for at least 3 min and from at least 3 MOs per condition.
4.10.Whole Mount Staining and High-Content Imaging.For immunostaining of whole µMOs without microsectioning, whole µMOs were fixed with 4% paraformaldehyde (Sigma) for 15-20 min at room temperature and washed three times with PBST containing DPBS supplemented with 0.1% Triton X-100.Fixed µMOs were transferred into blocking solution containing DPBS supplemented with 6% BSA (Sigma), 0.2% Triton X-100, and 0.01% sodium azide (Biosolution) and then incubated overnight on rocking platform mixer at room temperature.The samples were incubated with primary antibodies for 48 hr on a rocking platform mixer at room temperature.After washing three times with PBST, µMOs were incubated with fluorescence-labeled secondary antibody and TOPRO-3 on a rocking platform mixer for 48 hr at room temperature.After rinsing three times with PBST, 96-well plates containing µMOs were imaged using a THUNDER Imager Live Cell & 3D Cell Culture & 3D Assay (Leica Microsystems).Cell imaging and visualization of optical sections in 3D were carried out on a Leica DMi8 microscope using the Thunder Imaging System (Leica Microsystems).Stacks were taken at 2 μm intervals in the z-axis.To quantify the mean brightness of TH and MAP2 signals from each µMO, the images of individual µMO were analyzed using with Image J software (https://imagej.nih.gov/ij/download.html).

4.11.
In Vitro High-Throughput Screening of Drug Toxicity.
For the in vitro toxicity assay using µMOs, 96 well-derived DIV 20 µMOs were treated with different concentrations of MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) (150, 300, and 500 µM).After 10 days of drug treatment, µMOs were analyzed to measure changes in their morphology and size.To assess the transcriptional alterations of drug-treated µMOs, randomly selected 10 µMOs were used for qPCR.The drug response of individual µMOs to MPTP was determined by performing TUNEL staining.
To compare neurotoxin-mediated toxicity upon treatment of different toxins, we also treated rotenone (100 µM) and MPTP (150 µM) on DIV 30 µMOs for 2 days and determined their toxicity by assessing cleaved-caspase 3 and TUNELpositive cells.As a positive control, we treated H 2 O 2 on µMOs for 30 min.Nontreated µMOs were used as negative control.
4.12.TUNEL Assay.To observe apoptotic cells in µMOs, the TUNEL assay was performed using the In Situ Cell Death Detection Kit fluorescein (Roche) according to the manufacturer's protocol.In briefriefly, microsectioned µMOs on glass slides were prepared for the TUNEL assay.Samples were permeabilized with permeabilization solution for 2 min on ice and washed three times with PBS.They were then placed in citrate buffer, subjected to 750-W microwave irradiation for 1 min, and then translocated to cool distilled water.Samples were immersed for 30 min in tris-HCl (0.1 M and pH 7.5), containing 3% BSA and 20% normal bovine serum at room temperature.The TUNEL reaction mixture was then added to the µMO section.Organoid sections were blocked and costained with antibody raised against TUJ1.Quantification of TUNEL signals was performed by counting TUNEL-positive cells out of DAPI-positive cells using Image J software (https://imagej.nih.gov/ij/download.html).
4.13.RNAseq Analysis.RNA-seq data paired-end files were mapped to the human genome (UCSC and hg38) build using the RNA-seq aligner STAR (v2.5.2b).After mapping, the gene expression level was calculated as Fragments Per Kilobase Million (FPKM) by Cufflinks, the transcriptome assembly and differential expression analysis tool (v2.2.1), based on the default options.To determine clustering in gene expression, statistical analyses were performed with the heatmap.2function (gplots package) in the statistical software R (v3.3.2). 4.14.Morphological Image Analysis.For morphological analysis, the images of individual organoids were taken under an inverted microscope (Olympus) and the images for quantifying the diameter and area (including dead-core areas) of organoids were analyzed using Image J (https://imagej.nih.gov/ij/download.html)software.4.15.Statistical Analysis.For statistical analysis, the unpaired t-test was used for calculating p values.All the values were calculated from at least triplicate experiments, and the p values were presented as * p<0:05.

FIGURE 3 :
FIGURE 3: Accelerated functional maturation patterns in µMOs.(a) Schematic illustration depicting the procedure for evaluating the formation of neuromelanin-like granules in µMOs.(b) Accumulation of neuromelanin-like granules in µMOs (DIV 30).The neuromelanin-like granules were uniformly formed in all dopamine-treated µMOs (left) but not in untreated µMOs (right).The formation of neuromelanin-like granules was observed on the outside of µMOs.Scale bar, 100 μm.(c) Fontana-Masson staining of µMOs (DIV 30).Cerebral organoid was used as a negative control.Scale bar, 100 μm.(d) Confocal images showing the expression pattern of dopamine in µMOs (DIV 30).Scale bar, 20 μm.(e) Quantification of dopamine release in µMOs (DIV 30 and 60) mnt, minutes.(f, g) Color-mapped raster plots showing neural signals recorded from a 32-channel neural probe in typical MOs (f ) on DIV 20, 30, 70, and 130 and µMOs (g) on DIV 20 and 30.(h) Bar graph displaying mean firing rate of neural signals in typical MOs and µMOs on DIV 20 and 30 (n = 5 independent samples for typical MOs, n = 4 independent samples for µMOs on DIV 20, and n = 3 independent samples for µMOs on DIV 30).Data are presented as mean AE SEM.(i) Bar graph displaying synchronization between neural signals recorded from each electrode in µMOs on DIV 20 and 30 (n = 4 independent samples for µMOs on DIV 20 and n = 3 independent samples for µMOs on DIV 30).Data are presented as mean AE SEM.(j) Bar graph displaying the total number of connected electrodes in µMOs on DIV 20 and 30 (n = 4 independent samples for µMOs on DIV 20 and n = 3 independent samples for µMOs on DIV 30).Data are presented as mean AE SEM.

FIGURE 4 :
FIGURE 4: Minimal variability in µMOs within and across batches and obtained from different cell lines.(a) Images of individual µMOs produced in 96-(left) and 384-well plates (right) using GFP-labelled hESCs.(b) Representative morphological images of typical MOs (i) and µMOs derived from 96-(ii) to 384-well (iii) plates.Scale bar, 1 mm.(c) Average diameter of typical MOs and µMOs from H1 hESCs, H9 hESCs, and hiPSCs (DIV 30), 20 typical MOs or µMOs from each cell line were used for analysis.Data are presented as mean AE SD from three independent experiments.(d) Overview of high-content imaging showing the expression patterns of TH and MAP2 in 96 individual µMOs from 96-well plate.Scale bar, 100 μm.(e) Representative examples of high-content imaging.Scale bar, 100 μm.(f and g) Brightness of TH (f ) and MAP2 (g) signals in 96 individual µMOs.The continuous line represents the mean value of 96 individual µMOs from a 96-well plate and the dotted lines show the range of SD.

FIGURE 5 :
FIGURE 5: In vitro high-throughput toxicity assay using µMOs.(a) Schematic illustration depicting the in vitro toxicity assay using µMOs.(b and c) Confocal images showing the presence of TUNEL and cleaved-caspase 3 signals in µMOs after treatment with H 2 O 2 , MPTP, and rotenone.Scale bar, 100 μm.