Reliability of high-quantity human brain organoids for modeling microcephaly, glioma invasion and drug screening

Three-dimensional brain organoids derived from hiPSCs can model aspects of human brain development, neural network function, and genetic disease. Despite their potential for translational applications and drug discovery, current organoid methods suffer from major limitations: heterogeneity in morphology and regional identity, variable sizes, activation of stress pathways that impair cell-type specification, limited batch sizes preventing robust statistics, and poor reproducibility across lines and batches. Many protocols rely on embryoid body (EB) formation and extracellular matrix embedding, introducing variability in EB size and quality that affects downstream differentiation outcomes. The research question addressed is whether a simplified, scalable, EB-free approach can generate large numbers of human brain organoids with controlled size, reduced stress signaling, consistent cytoarchitecture and cell diversity, and functionality, thereby enabling reliable disease modeling and medium-throughput drug screening. The study proposes a Hi-Q platform using custom microwells and spinner bioreactors to standardize early neurosphere formation and downstream differentiation, and evaluates its reliability across multiple hiPSC lines and disease contexts.

Prior work established brain organoids as models of embryonic cortical development and disease, including microcephaly, but highlighted challenges with reproducibility and stress-induced transcriptional artifacts that alter neuronal diversity. EB-based methods and various microwell approaches have attempted to reduce variability, yet EB size heterogeneity affects cell lineage composition and maturation. Comprehensive analyses (e.g., Bhaduri et al.) reported activation of cellular stress pathways in organoids that impair molecular subtype specification. Single-cell atlases (e.g., Kanton et al., Nowakowski et al.) provide benchmarks for comparing organoid cell-type composition and maturation against primary human brain. Protocols using extrinsic factors (retinoic acid, BDNF) can push maturation but may trade off homogeneity. Recent organoid applications include modeling rare neurogenetic conditions and glioma invasion, but preclinical drug screening in 3D tissue has been limited by organoid quantity and quality constraints. This study situates its Hi-Q method within these efforts, aiming to overcome EB-related variability, stress pathway activation, and scale limitations.

Platform design and differentiation: hiPSCs (80% confluency) were dissociated into single cells (Accutase, 5 min, 37°C), counted and seeded into custom-designed coating-free spherical microwell plates fabricated from medical-grade Cyclo-Olefin-Copolymer (COC). Each of 24 wells contained 185 inverted pyramid-shaped microwells (1×1 mm opening, ~180 µm round base), promoting uniform sphere formation via cell adhesion and identical diffusion conditions. Approximately 10,000 hiPSCs were seeded per microwell in neural induction medium (NIM) with ROCK inhibitor for the first 24 h; ROCK inhibitor was omitted thereafter to avoid stress and mesendodermal bias. Over 5 days, uniform neurospheres formed, displaying neural rosettes and apical cilia. Neurospheres were transferred to spinner-flask bioreactors (75 ml neurosphere medium; 25 rpm), cultured 4 days, then switched to brain organoid differentiation medium with SB431542 (5 µM) and Dorsomorphin (0.5 µM) for undirected neural differentiation. After 21 days, cultures were switched to a maturation medium without SMAD inhibitors and maintained up to day 150 at 25 rpm. Scale and lines: Organoids were generated from six hiPSC lines (four healthy, two patient-derived for microcephaly), yielding ~15,373 organoids across 39 batches. Size uniformity and growth kinetics were assessed across lines and batches. Single-cell RNA sequencing: At days 60, 90, and 150, nuclei from three organoids per time point were sequenced (Chromium Single Cell 3′ v3; NovaSeq 6000; ~25k reads/nucleus). Data processing used CellRanger (GRCh38), Scanpy v1.8.2, scrublet (doublet removal), normalization/log-transform, highly variable gene selection, regression of total counts and mitochondrial percentage, PCA, Harmony for batch smoothing, BBKNN for kNN, ForceAtlas2 for graph layout, Leiden clustering, and Lasso logistic regression trained on primary embryonic brain labels (Nowakowski et al.) to annotate cell types. Diffusion pseudotime (DPT) analysis started from proliferative radial glia. Cross-study integration compared Hi-Q to EB-based datasets (Kanton et al., 90/180 days; others at 60 days) using Seurat v5 and Harmony. Histology and immunostaining: Whole-mount clearing (ethanol series, ethyl cinnamate) and confocal imaging (Zeiss LSM 880) quantified markers: progenitors (SOX2, Nestin, PAX6), early neurons (DCX), neuronal/cilia (acetylated α-tubulin, Arl13B), pan-neuronal (TUJ-1), cortical neurons (MAP2, Tau, CTIP2), Purkinje marker (PCP4), inhibitory neuron marker (GAD67), synaptic proteins (Synapsin-1, PSD95), stress markers (pH2AX for DNA damage), apoptosis (TUNEL), and oSVZ markers (PTPRZ1, phospho-vimentin). Functional assays (calcium imaging): Organoids at days 30, 40, 50, and 150 were loaded via bolus injection with OGB-1-AM and imaged (wide-field; 14–20 Hz) to assess spontaneous activity and responses to bath-applied glutamate (1 mM) and GABA (1 mM). Pharmacological blockers included TTX (1 µM; Na+ channels), APV (100 µM; NMDA), NBQX (50 µM; AMPA), and NiCl2 (voltage-gated Ca2+). Cryopreservation: Day 18 organoids were frozen in CS10 medium (10 organoids per 1 ml vial; controlled freezing: 24–48 h at −80°C then liquid nitrogen). Thawing at 37°C followed by 48 h static recovery, then spinner-flask culture. Viability, size, cytoarchitecture, and recovery percentage were assessed; older organoids (day 35–40) did not recover. Disease modeling: Patient-derived hiPSCs with CDK5RAP2 homozygous mutation (primary microcephaly) and ERCC6/CSB mutation (Cockayne syndrome, secondary microcephaly) were differentiated via Hi-Q. Cytoarchitecture (VZ presence/diameter, progenitor compaction, cortical plate formation) and stress markers (pH2AX, TUNEL) were quantified; mitotic spindle orientation assessed via p-vimentin to infer symmetric vs asymmetric divisions. Glioma invasion and drug screening: mCherry-labeled patient-derived GSC line #450 was applied as spheres or single cells to day-50 Hi-Q organoids. A 96-well assay used automated imaging (Operetta High-Content; Brightfield and DsRed channels) at 24 h and 72 h. Image analysis (PerkinElmer Columbus) detected organoid boundaries and GSC invasion spots; hits were defined as compounds with 0–1 invasion spot at both timepoints. A library of 180 validated selective inhibitors (Target Selective Screening Library; 5 µM initial screen, 1 µM secondary) was tested. Quantitative 3D imaging and computational volume occupancy metrics validated invasion inhibition. In vivo validation: GFP-tagged GSC lines (#1 and #472) were grafted into NOD-SCID mouse striatum. After one week, animals received saline or combination therapy (Selumetinib 20 mg/kg oral; Fulvestrant 1 mg s.c.) daily for three weeks. At eight weeks post-engraftment, brains were fixed and serially sectioned; tumor volumes and invasion into corpus callosum, optic tract, and ventricular walls were quantified.

Scalability and reproducibility: The Hi-Q platform generated ~15,373 organoids across 39 batches. Day 5 neurospheres were uniform across batches (average ~0.5 mm diameter). Measuring 300 randomly selected organoids across four hiPSC lines showed high intra-batch size consistency and proportional growth from day 20 to day 60 without aberrant variation. Integrity was high, with only 0–3 disintegrated organoids per 300 sampled (Fig. 1J). Cell diversity and stress: Time-resolved scRNA-seq (16,228 cells across days 60, 90, 150) identified apical progenitors, radial glia (RG), proliferating RG, IPCs, developing neurons, early neurons, excitatory (EN) and inhibitory (IN) neurons, astrocytes/oligodendrocytes. Pseudotime revealed two trajectories: dorsal telencephalon lineage to EN (GLI3, EOMES, GRIA2) and ventral lineage to IN (CCND2, DLX5, GAD2). Hi-Q organoids exhibited lower expression of stress markers PGK1, ARCN1, GORASP2 than published organoid datasets (Bhaduri; Kanton) and slightly higher than adult brain (Allen Brain Atlas), indicating reduced ectopic stress. Compared to EB-based organoids, Hi-Q organoids had fewer astrocytes and a higher proportion of proliferative populations, suggesting slightly less maturation at matched ages. Maturation and cytoarchitecture: Day 20 organoids primarily expressed progenitor and early neuronal markers, with limited axonal segregation of MAP2/TUJ-1. By day 60, cortical plates were thicker and more distinct, with increased Tau, PCP4, PSD95, Synapsin-1, and CTIP2 expression; VZ/SVZ and oSVZ markers were present by scRNA and immunostaining. Functional activity: Across ages 30, 40, 50, 150, 42–54% of cells were spontaneously active. TTX (1 µM) strongly dampened activity at days 40, 50, 150 (p=1.39E-14; 7.96E-34; 3.88E-48). All tested cells responded to glutamate (1 mM) and GABA (1 mM) with large, long-lasting Ca2+ transients. Glutamate-induced signals at day 50 were reduced by APV (100 µM; N=3, n=133; p=2.84E-12) and further by APV+NBQX (50 µM; N=3, n=108; p=3.50E-29). GABA responses were reduced by TTX (N=3, n=204; p=3.85E-14) and nearly abolished by TTX+NiCl2, indicating functional expression of TTX-sensitive Na+ channels, voltage-gated Ca2+ channels, AMPA/NMDA receptors, and GABAA-mediated depolarization-dependent Ca2+ influx. Cryopreservation: Day 18 organoids achieved 75–90% recovery after thawing, with no significant size changes versus controls, similar VZ and primitive cortical plate cytoarchitecture, and by day 60, synapsin-1-positive mature neuronal architectures indistinguishable from controls. Day 35–40 organoids failed to recover, showing elevated TUNEL and damaged cytoarchitecture. Disease modeling: CDK5RAP2 organoids (primary microcephaly) were significantly smaller at day 30, with smaller, disorganized VZs and dispersed TUJ-1+ neurons; spindle orientation analysis showed predominant vertical divisions (premature differentiation) versus horizontal divisions in controls (progenitor expansion). CSB organoids (Cockayne syndrome) were initially larger but collapsed later; they lacked recognizable VZs, showed weak SOX2+ progenitor density and diffuse neurons, and exhibited significantly increased pH2AX+ and TUNEL+ cells, indicating extensive DNA damage and apoptosis. Glioma invasion and drug screen: Hi-Q organoids supported robust invasion by patient-derived GSCs (spheres or single cells), with protrusions and microtubes mimicking in vivo behavior. A medium-throughput screen of 180 compounds (5 µM) identified 16 hits inhibiting invasion; secondary screening (1 µM) highlighted Selumetinib (MEK inhibitor) and Fulvestrant (SERD) as effective. Quantitative 3D imaging showed significant reductions in invasion depth, protrusion number/length, GSC foci number/diameter versus DMSO controls. In NOD-SCID xenografts (GSC#1, GSC#472), combined Selumetinib+Fulvestrant treatment reduced striatal tumor volumes and invasion into corpus callosum, optic tract, and ventricular walls compared to saline.

The Hi-Q platform directly addresses key barriers in brain organoid research by standardizing early neurosphere formation in coating-free microwells, omitting EB and extracellular matrix steps that contribute to heterogeneity. The result is large-scale production of uniform organoids with consistent cell diversity and reduced ectopic stress signaling, improving reliability for disease modeling and screening. Time-resolved scRNA-seq confirms reproducible maturation trajectories toward excitatory and inhibitory neuronal lineages and alignment of cell-type composition across organoids of the same age, with lower stress marker expression than prior organoid datasets. Functional assays demonstrate development of active neuronal networks responsive to major neurotransmitters and ion channel blockers, validating physiological relevance. Cryopreservation at early stages enables biobanking and flexible workflows, an important advance for patient-derived models. Modeling two distinct neurogenetic disorders showcases versatility: CDK5RAP2 microcephaly reflects progenitor division defects leading to premature differentiation, while CSB progeria-associated pathology reveals DNA damage–driven disorganization and cell death. Finally, the glioma invasion assay adapts Hi-Q organoids to medium-throughput screening, identifying clinically relevant inhibitors (Selumetinib, Fulvestrant) that also reduce invasion in vivo, supporting translational potential. Together, these findings show that a scalable, low-stress, reproducible organoid platform can underpin personalized modeling and targeted drug discovery in human brain tissue contexts.

This study introduces a robust, EB-free Hi-Q organoid platform that produces thousands of uniform human brain organoids across multiple hiPSC lines with reproducible cytoarchitecture, cell diversity, functional neuronal networks, and reduced culture-induced stress. The method supports cryopreservation and re-culturing, enabling biobanking, and reliably models divergent neurogenetic diseases (primary microcephaly due to CDK5RAP2 mutation and progeria-associated Cockayne syndrome). Hi-Q organoids also provide a physiologically relevant substrate for glioma invasion assays and medium-throughput drug screening, yielding candidate inhibitors (Selumetinib and Fulvestrant) validated in mouse xenografts. Future directions include enhancing maturation where desired via optional factors (e.g., retinoic acid, BDNF, CNTF), extending cryopreservation protocols to later developmental stages, broadening disease panels for precision medicine, and integrating high-content analytics for larger-scale screening across diverse patient-derived lines.

Hi-Q organoids, while showing reduced stress marker expression compared to other in vitro datasets, still do not fully match in vivo brain conditions and exhibit slightly higher stress than adult brain. Relative to EB-based protocols, Hi-Q organoids appear less mature at matched ages, with fewer astrocytes and more proliferative populations, reflecting the unguided differentiation and omission of maturation factors. Cryopreservation was successful at day 18 but failed for day 35–40 organoids, indicating stage-dependent viability and the need for optimized protocols for mature tissues. Skipping EB formation may limit contributions from mesoderm/endoderm-derived signals that could influence maturation, and the protocol’s undirected differentiation may require optional supplementation to achieve specific regional identities or advanced maturation states. Drug screening was medium-throughput and limited to 180 selective inhibitors; broader libraries and multiple GSC lines would strengthen generalizability.