Three-dimensional midbrain organoids: a next-generation tool for Parkinson’s disease modelling and drug discovery

Abstract

Parkinson’s disease (PD), a progressive neurodegenerative disorder marked by dopaminergic (DA) neuron loss and Lewy body formation, lacks therapies to halt neurodegeneration. Current models, including 2D cultures and animal studies, fail to fully recapitulate human midbrain complexity, underscoring the need for advanced human-relevant disease modelling systems. Midbrain organoids (MOs), three-dimensional (3D) stem cell-derived neuronal structures mimicking midbrain architecture, have emerged as transformative tools for modelling PD. These organoids replicate key pathological hallmarks and enable disease mechanistic studies and drug screening for PD. Recent advances of research in MOs include genetic modelling of PD-linked mutations (e.g., LRRK2, GBA1, DNAJC6), optogenetics-assisted α-synuclein (α-syn) protein aggregation systems, and high-throughput drug testing platforms. MOs also show promise for cell replacement therapy, with successful integration and functional recovery in animal PD models. However, challenges such as batch variability, limited vascularization, incomplete neuronal maturation, and high costs hinder reproducibility and scalability. Future directions focus on integrating vascular networks, microglia co-cultures, automated workflows, and assembloid technologies to enhance pathophysiological relevance and translational potential in PD. By addressing these limitations, research in MOs could revolutionize PD research, offering critical insights into disease mechanisms and accelerating therapeutic discovery for PD patients.

Introduction

Parkinson’s disease (PD) affected an estimated 11.9 million individuals worldwide in 2021, with projections suggesting a rise to 15.6 million in 2030, 20.4 million in 2040, and 25.2 million by 2050, driven primarily by global population aging and growth [1]. PD imposes a substantial dual burden in aging societies, with annual U.S. costs exceeding $50 billion from both direct healthcare expenses and indirect losses like reduced productivity and caregiver strain. Beyond economics, patients face social isolation and stigma, while families and healthcare systems grapple with caregiver burnout, rising hospitalizations, and disparities in specialized care access [2, 3]. Its pathological hallmark includes Lewy bodies formation and the gradual loss of dopamine (DA) -producing neurons in the substantia nigra pars compacta (SNpc) of the midbrain [4, 5]. As DA levels fall, patients develop characteristic PD motor symptoms, including resting tremor, rigidity, bradykinesia (slowness of movement) and postural instability, as well as a variety of non-motor features such as sleep disturbances, mood changes, and autonomic dysfunction [5].

Despite intensive research, there are currently no therapies that can halt or reverse the underlying neurodegeneration; available treatments (for example, levodopa and dopamine agonists) primarily aim to replenish DA or mimic its action and thus relieve patient symptoms [6, 7]. Deep brain stimulation (DBS), a surgical therapy used to manage motor symptoms and medication-induced complications, has also been applied in selected patients with early-stage PD [8]. However, while DBS can significantly improve motor function and quality of life, there is no conclusive evidence that it can alter the course of neurodegeneration or slow disease progression. Moreover, individual responses to these medications vary widely, and long-term use can lead to complications such as motor fluctuations and dyskinesias [9]. This clinical variability, together with PD’s complex and multifactorial nature, continues to challenge the development of truly disease-modifying treatments. To address these challenges, researchers are leveraging advanced model systems, including cell-based assays, animal models, and patient-derived neuronal cultures, to recapitulate α-synuclein (α-syn) pathology, DA neuron degeneration, and genetic risk factors, thereby enabling mechanistic studies and preclinical testing of novel therapeutic strategies [10, 11].

Critically, PD occurs naturally only in humans with no spontaneous counterpart in other species due to unique vulnerabilities in human nigral neurons and distinct aging processes. This species exclusivity complicates disease modelling, as animal models fail to fully replicate human pathophysiology and drug responses, underscoring the need for human-relevant systems. In addition, recently the regulatory agencies, such as the U.S. Food and Drug Administration (FDA), have advocated for a reduction in animal model use in drug testing and disease modelling [12]. This shift underscores the growing need for advanced human-relevant cell-based systems to study neurodegenerative disorders, particularly PD. More recently, disease modelling using patient-derived induced pluripotent stem cells (iPSCs) has gained traction [13]. These iPSCs retain the patient’s unique genetic background and mutations, offering a personalized platform for studying PD [14]. In vitro, iPSCs can be differentiated into DA neurons, but this process is time-consuming, and the resulting two-dimensional (2D) DA neuron cultures lack the structural and functional complexity of human midbrain [15].

To overcome these limitations, researchers have developed 3D midbrain organoids (MOs) that more accurately mimic the architecture, cellular diversity, and microenvironment of the midbrain, a region central to PD pathology [16]. The MOs approaches have demonstrated the ability to recapitulate key PD phenotypes and are increasingly used to investigate disease mechanisms and evaluate potential therapeutics in preclinical studies [17]. In this review, we explore the recent advancements in MOs technologies, evaluate their application in PD research, and identify current challenges and future directions to enhance their translational impacts.

Midbrain organoids (MOs) models

MOs recapitulate the structural and functional features of human midbrain, a region critically involved in producing DA that regulate motor activity. While MOs offer advantages in modelling 3D human tissue architecture, they present unique limitations compared to 2D systems. Critically, MOs develop hypoxic cores absent in 2D cultures and lack functional long-range neural connections. As summarized in Table 1, both platforms have complementary strengths: 2D models excel in high-throughput screening, whereas MOs better recapitulate human disease phenotypes like spontaneous α-synuclein aggregation. The development of these organoids builds upon advances in 2D culture systems—particularly the influential work by Kriks et al. (2011), who established a floor plate-based patterning protocol to direct human pluripotent stem cells (hPSCs) toward a midbrain dopaminergic fate [18]. Although their method was originally developed in 2D, it laid the groundwork for subsequent adaptations in 3D cultures, marking significant progress in the use of hPSC-derived systems for PD modelling.

Table 1.

Comparison of 2D vs. 3D models for parkinson’s research

Aspect	2D Models	3D models	PD Research Implication
Physiological Relevance	Low: Lack 3D architecture	High: Recapitulates tissue organization	MOs better model cell-matrix interaction in SNPc
Disease Phenotypes	Artificial α-syn induction	Spontaneous α-syn/Lewy pathology	MOs capture natural protein aggregation dynamics
Hypoxia Artifacts	Absent	Present in cores (> 200 μm)	MOs may overrepresent hypoxic stress vs. human brain
Neural Circuitry	None	Local synapses onlym no long-range connections	Both to model nigrostriatal pathwayfail
Throughput & Cost	High throughput; Low cost	Low throughput; high cost	2D better for initial drug screening
Reproducibility	High (standardized protocols)	Variable (batch to batch heterogeneity)	2D more reliable for toxicity assays
Key Utility	Target validation, high-throughput screening	Pathogenesis studies; host-graft interaction modelling	Models serve complementary roles

The floor plate patterning strategy mimics the ventral midline region of the developing neural tube and guides neuronal differentiation via the secretion of morphogens like Sonic Hedgehog (SHH). By exposing stem cells to a combination of SHH, WNT pathway activators, and fibroblast growth factors, researchers induce a floor-plate-like identity, enabling robust generation of midbrain DA (mDA) neurons. These principles have since been translated into 3D MOs models, where the spatial organization allows for self-assembly into tissue-like structures more closely resembling the in vivo midbrain. In these 3D cultures, further patterning refinement with neurotrophic factors such as brain-derived neurotrophic factor (BDNF) and glial cell line-derived neurotrophic factor (GDNF) enhances DA neuron survival and maturation (Fig. 1) [19–21]. Human-derived MOs typically mature within 40–50 days, acquiring electrophysiological properties and expressing Tyrosine Hydroxylase (TH), a key marker for mDA neurons. Studies by Jo et al. demonstrated that these organoids can also produce neuromelanin, a characteristic feature of adult human midbrain tissue [22]. Functional maturation of MOs was evaluated through a combination of electrophysiological assessments (e.g., spontaneous action potentials and synaptic activity), neurochemical profiling (including immunostaining for dopaminergic markers and dopamine quantification), and observation of disease-relevant phenotypes such as α-synuclein aggregation and neuromelanin production. Together, these assays provide functional validation of neuronal identity and maturity beyond molecular markers, as summarized in Table 2.

Fig. 1.

Generation of Midbrain Organoids. Human MOs are generated from hPSCs through a multistep protocol that recapitulates early midbrain development in vitro based on Jo et al. (2016) protocol [22]. In the initial phase, hPSCs are aggregated to form embryoid bodies (EBs). This aggregation establishes a 3D foundation for subsequent lineage specification. The neural induction and floor plate patterning phase directs the EB toward a midbrain neuroectoderm fate. As differentiation progresses, the EBs are embedded in a Matrigel matrix to promote organoid assembly by providing structural support and establishing a microenvironment conducive to 3D tissue organization. Finally, the mature phase is reached as the organoids continue to differentiate; functional properties characteristic of midbrain tissue become evident, including the emergence of TH-positive DA neurons and the formation of synaptic networks

Table 2.

Comparative summary of human MOs models and their key features

Authors (Year)	Cell Origin	Primary Application	Organoid Type	Maturation Assessment Method	Culture Duration (Days)	DA Survival Rate (%)	Uniformity (CV)	% TH⁺ DA Neurons	DA Neuron Subtype	Key Readouts/Findings
Jo et al., 2016	hESCs	Technology development	Floor-plate–patterned 3D MOs	Electrophysiological recordings, immunostaining, and RNA-seq for gene expression analyses	45–70	~ 50%	High (> 30%)	~ 54% (Day 45)	A9 (SNpc-type)	Neuromelanin present; functional electrophysiology; first MO model producing mature DA neurons
Fiorenzano et al., 2021	Multiple hPSC lines	Technology development	Silk-scaffolded MOs	Single-cell RNA sequencing, electrophysiology, neuromelanin staining, and immunohistochemistry	~ 60	~ 60%	Moderate (20%)	~ 20% (Day 60)	A9 & A10, with an A9-like cluster	Neuromelanin present; single-cell profiling showed diverse DA subtypes; improved tissue architecture
Kim et al., 2019	LRRK2 G2019S iPSCs	Genetic modeling	LRRK2 G2019S mutant MOs	qRT-PCR, immunofluorescence, FACS for synapsin+ cells, dopamine quantification, and gene set enrichment analysis	60	LRRK2 mutation cells increased DA death up to 20%	Not reported	~ 60% (Day 60)	A9 (SNpc)	DA neuron loss; TXNIP identified as mediator of G2019S pathology
Jo et al., 2021	GBA1⁻/⁻ ± α-syn iPSCs	Genetic modeling	GBA1⁻/⁻ + α-syn triplication MOs	Immunostaining for midbrain DA marker and electrophysiology assay	60–90	GBA1⁻/⁻ and α-syn triplication greatly increased dopaminergic apoptosis up to 5% and 10%	Uniform ventral midbrain patterning across different genetic background	~ 30% (Day 90)	A9 (SNpc)	Lewy body–like α-syn aggregates; synergistic PD phenotypes from dual genetic hits
Wulansari et al., 2021	DNAJC6 KO iPSCs	Genetic modeling	DNAJC6 KO MOs	Immunostaining for midbrain DA marker	~ 50–70	Mutant organoids showed ~ 50% of WT TH⁺ neuron yield	Very uniform (> 95% FOXA2⁺/LMX1A⁺)	~ 80% TH⁺/LMX1A⁺ (Day 60)	A9 (SNpc) and A10-like	Impaired DA development (WNT–LMX1A disruption) + α-syn accumulation/degeneration
Kim et al., 2023	OASIS iPSCs	Genetic modeling, Drug screening	Optogenetic α-syn MOs	Immunostaining for midbrain DA marker, α-syn aggregation staining	~ 30–60	OASIS caused rapid DA neurotoxicity (percentage are not specified)	Not specified	Not specified	A9 (SNpc)	Accelerated DA loss upon light activation; used to screen BAG956 for α-syn clearance
Renner et al., 2021	hPSC-derived MOs	Drug screening	Automated HTS (“AMOs”)	Immunostaining for TH and MAP2 (DA and neuron marker)	45–60	~ 70%	Very low (< 5%)	~ 50%	A9/A10	High-throughput, single-cell toxicity screening; reduced batch variability
An et al., 2024	PD iPSC-derived MOs	Drug screening	Au nanodot–patterned 3D MOs	Immunofluorescence and qPCR for dopamine neuron marker, and multi-electrode array recordings	Not specified	Not assessed (focus on electrochemical detection).	Not applicable	Not specified	A9/A10	Nanostructured electrode for real-time monitoring of drug responses in 3D MOs
Zheng et al., 2023	hPSC-derived MOs	Transplantation	Standard MOs → grafted into PD mouse	Immunofluorescence, neuromelanin staining, and HPLC-based dopamine quantification	~ 25 days in vitro (pre-graft), 6- and 12-weeks post-transplant.	2.57% TH⁺ at 6 weeks vs. 14.11% TH⁺ at 12 weeks post-transplant	Not explicitly measured	~ 0.80% (D15 pre-graft)	A9 (SNpc)	Graft survival, axon projection, synapse formation; restored motor function in PD mouse
Cai et al., 2025	hPSC-derived MOs	Technology development	VID scaffold–embedded MOs	single-cell RNA sequencing, immunostaining, dopamine quantification, neuromelanin staining, and electrophysiology assay	70	~ 85%	Moderate (15%)	> 65%	A9/A10	Scaffold improved DA neuron survival & zonal midbrain organization (addresses vascularization)
Sabaté-Soler et al., 2022	hPSC MOs + microglia	Technology development	Microglia-integrated MOs	Immunostaining of DA marker and RNA-Seq after microglia co-culture	~ 60	Not reported	Not addressed.	Not quantified	A9/A10	Increased synaptic development & electrophysiological maturity (immune-integrated model)
Mohamed et al., 2022	hPSC MOs	Technology development	Microfabricated disk MOs	Scalable hMOS viability, Immunostaining of TH and MAP1	60	> 80%	Low (< 10%)	> 60%	A9/A10	High-throughput organoid generation (addresses production bottleneck)
Gerasimova et al., 2025	MOs + T cells	Technology development	T cell–MO co-culture model	Immunostaining of FOXA2 and TH, qPCR of DA marker, and viability assay	30–60; 7 days with T-cells	Co-culture caused significant neuronal death (~ 5%)	Not applicable	Not quantified	A9/A10	Platform to study adaptive immune effects on midbrain neurons
Reumann et al., 2023	Assembloid MOs	Technology development	Ventral midbrain–striatum–cortex assembloid	Immunostaining of FOXA2 and TH	Ventral midbrain, striatal, and cortical organoids (∼20–25 days old) Assembloids were analyzed up to ~ day 90–109 (post-fusion)	Midbrain region maintained TH⁺ cells through > 90 days	High reproducibility – ~96.0% (± 2.4% s.d.) of assemblies stayed fused after 3 days in culture	~ 60% of midbrain-derived axons in forebrain were TH⁺	Mixed A10/A9 identities	Recreated DA circuit connectivity in vitro (multi-region assembloid for PD circuit modeling)

Recent advances using single-cell transcriptomics have mapped the developmental trajectory of MOs, revealing a diversity of DA neuron subtypes. Notably, one cluster showed a transcriptional signature closely resembling A9 SNpc DA neurons, the primary population affected in PD. These A9-like neurons express appropriate molecular markers and display spontaneous firing activity, reflecting intrinsic electrophysiological properties characteristic of dopaminergic neurons. While these findings support the maturation of DA neuron-like populations within MOs, it is important to note that organoids lack long-range afferent and efferent connections found in vivo. Nonetheless, MOs offer a valuable human-derived platform to study midbrain patterning and dopaminergic differentiation, with growing utility in PD pathogenesis research and preclinical screening [23].

Application of human MOs

MOs are engineered to recreate the intricate cellular architecture and microenvironment of the human midbrain, making them indispensable for PD modelling studies. Their 3D structure allows researchers to investigate dopaminergic neuron development and model PD-related genetic mutations in a human-relevant context [17]. In addition, MOs better reflect certain features of human tissue physiology compared to 2D systems, although they still lack vasculature and full neural circuitry. Several studies have begun to explore their potential for examining cellular responses under specific environmental or genetic conditions [24]. Table 1 provides a comprehensive overview of human MOs models, summarizing their methodologies, dopaminergic features, and applications in PD research. These models support mechanistic investigations of disease progression, such as neuronal degeneration and α-synuclein pathology, and offer platforms for preclinical drug screening. In the following section, we explore these applications in greater detail (Fig. 2).

Fig. 2.

Applications of Human Midbrain Organoids in PD Research. This figure illustrates the use of human MOs in PD research, highlighting three key applications: (1) Human MOs containing PD-associated mutations serve as models to study disease mechanisms and their role in PD progression. Additionally, gene correction strategies allow researchers to assess the potential of mutation reversal as a regenerative therapy for PD patients (2) Drug Screening – Affected human MOs serve as platforms for testing potential neuroprotective compounds. Drugs are administered, and their effects are assessed using molecular and cellular analyses to identify promising candidates for PD treatment (3) Transplantation – Healthy human MOs are transplanted into PD animal models to evaluate their regenerative potential in restoring dopaminergic function. Successful integration and functional recovery are indicated by improved motor behavior and neurochemical balance. These applications demonstrate the versatility of MOs in advancing PD modelling, drug discovery, and regenerative therapies

MOs genetic models

MOs are derived from hPSCs, including iPSCs that are reprogrammed from adult somatic cells. iPSCs generated from patients with specific PD mutations capture the relevant genetic profiles, allowing researchers to closely mimic the disease’s inherent complexities within an in vitro setting [25]. Additionally, iPSCs derived from healthy individuals can be genetically edited to introduce PD-related mutations, which minimizes the confounding effects of genetic background variability and provides a controlled environment to delineate mutation-specific phenotypic outcomes [26]. This model enables the direct observation of phenotypic changes that occur as a consequence of PD-related mutations, such as the progressive degeneration of DA neurons [26]. A study by Kim et al. (2019) developed a robust 3D in vitro model using isogenic MOs derived from iPSCs, comparing those with the LRRK2 G2019S mutation to matched controls. Their model recapitulated hallmark PD features. Notably, the thiol-oxidoreductase TXNIP emerged as a key mediator in the protein-protein interaction network within mutant organoids, potentially contributing to neurodegeneration. This study underscores the value of MOs for identifying novel pathogenic mechanisms and therapeutic targets in PD [27]. Another study by Jo et al. (2021) examined the combined impact of glucocerebrosidase (GBA1) deficiency and α-syn overexpression using a dual perturbation model. This led to the accumulation of detergent-resistant, β-sheet–rich α-syn aggregates, forming Lewy body–like inclusions. Similar inclusions were also observed in patient-derived organoids carrying SNCA triplication when treated with a glucocerebrosidase inhibitor. These findings highlight the utility of MOs in modelling Lewy body-related neurodegeneration and in screening therapeutic interventions targeting α-syn pathology [28].

A study by Wulansari et al. (2021) investigated early-onset PD linked to DNAJC6 mutations using CRISPR-Cas9-modified brain organoids. They observed that loss-of-function mutations in DNAJC6 impaired endocytosis and disrupted the WNT–LMX1A signalling axis, which is crucial for mDA neurons development. These defects led to both developmental vulnerabilities and degenerative phenotypes, including α-syn aggregation, increased neuronal firing, and mitochondrial/lysosomal dysfunctions. This dual impact highlights how genetic mutations can disrupt both early neurodevelopment and later neurodegeneration in PD [29]. More recently, Kim et al. (2023) introduced an optogenetics-assisted α-syn aggregation induction system (OASIS) in iPSC-derived MOs, rapidly inducing pathological features akin to human PD. OASIS provides a dynamic platform to study PD pathogenesis and screen aggregation-targeted therapies [30]. In conclusion, these studies underscore the immense promise of MOs as a platform for genetic modelling in PD. Derived from hPSCs—either as patient-specific iPSCs carrying defined PD mutations or as healthy cells genetically modified to introduce PD risk factors—MOs recapitulate both the developmental and degenerative pathologies observed in PD. Furthermore, these models can be used to explore regenerative strategies, such as testing whether genetic correction of PD mutations can restore normal midbrain phenotypes and function. MOs provide a complementary and human-relevant platform for studying PD mechanisms and identifying therapeutic targets, especially in contexts where traditional 2D cultures may not capture complex tissue-level interactions.

Drug testing with MOs

MOs have emerged as a powerful platform for drug screening, offering a physiologically relevant model that closely mimics human midbrain conditions. These 3D organoids contain functional DA neurons, making them particularly valuable for studying PD and other neurodegenerative disorders. MOs have the potential to better reflect certain aspects of human midbrain physiology than traditional 2D cultures, offering a valuable in vitro system for assessing drug responses in a more tissue-like context. In this perspective, Kim et al. (2023) demonstrated the utility of OASIS in MOs for accelerating PD pathology modelling and drug screening. Their study identified BAG956, a PI3K inhibitor, as a promising compound capable of reversing PD-related phenotypes by promoting autophagic clearance of pathological α-syn aggregates. These results underscore the potential of MOs in therapeutic development; however, certain challenges remain. Batch-to-batch variability in organoid production and heterogeneous drug distribution within the tissue can introduce inconsistencies that may limit the reproducibility of experimental outcomes [30]. In addition, these results are based on only two iPSC lines and have not yet been replicated across additional genetic backgrounds, so it remains unclear how robust BAG956’s effects will be in broader patient populations.

To address these limitations, Renner et al. (2021) developed a high-throughput toxicity screening (HTS) platform using human MOs, demonstrating a scalable and reproducible method for evaluating drug toxicity at the single-cell level. Their automated workflow significantly reduces variability in organoid production and drug administration, ensuring more consistent experimental outcomes. Additionally, their study leveraged single-cell resolution analysis, overcoming the issue of heterogeneous drug distribution by precisely assessing neurotoxic effects within organoids [31]. Yet the reliance on viability and aggregate-load endpoints may overlook subtler functional impairments, and no head-to-head comparison with animal or clinical data has validated its predictive power.

Further advancing drug screening capabilities, An et al. (2024) introduced a mesoporous gold (Au) nanodot-patterned 3D concave electrode system to enhance drug evaluation in PD patient-derived MOs. This innovative approach improves electrophysiological monitoring, allowing for real-time assessment of neuronal activity and drug responses within organoids. By integrating nanotechnology with organoid-based models, the study provides a more precise and scalable method for evaluating therapeutic compounds, further strengthening the translational potential of MOs in PD research [32]. These advancements highlight the evolving potential of MOs as sophisticated drug screening platforms. Importantly, MOs may be particularly suitable for capturing complex tissue-level drug responses, while 2D systems remain indispensable for high-throughput, cost-effective early-stage screening. While MOs offer structural and cellular advantages over conventional systems, further validation and head-to-head comparisons are needed to establish their translational reliability and predictive accuracy.

Transplantation of MOs in PD animal models

MOs hold significant promise as a potential cell replacement therapy for PD, addressing one of the primary pathological features of the disease—the progressive loss of DA neurons in the SNpc. Some ongoing trials, such as BlueRock Therapeutics and the Japanese iPSC program, have collectively established the safety and early efficacy of hESC and iPSCderived dopaminergic progenitor grafts, which showed motor improvements and increased dopamine PET uptake without tumorigenesis or graftrelated dyskinesias [33, 34]. While these single-cell suspensions can replenish lost DA neurons, they lack the native 3D microenvironment of the midbrain. MOs, in contrast, offer a preorganized tissue construct complete with supporting glia and extracellular matrix, preformed synaptic networks, and enhanced cell–cell support that boosts graft survival, maturation, and rapid functional integration. Moreover, patientspecific organoids enable modeling of host immune responses and rigorous preclinical safety testing—features that can derisk and refine nextgeneration PD cell therapies [35]. Given that the depletion of these neurons leads to impaired dopamine signalling, contributing to the characteristic motor dysfunctions observed in PD patients, transplanting MOs represents an innovative strategy to replenish lost DA neurons and restore functional connectivity in affected brain regions. However, direct comparisons between MO transplantation and conventional cell suspension approaches remain limited, and more rigorous head-to-head studies are needed to determine the specific therapeutic contexts where organoids may offer clear advantages.

Although there are currently no ongoing clinical trials of MO transplantation in PD patients, several preclinical studies have demonstrated its feasibility and therapeutic potential in animal models. A key study conducted by Zheng et al. (2023) demonstrated the therapeutic potential of human iPSC-derived MOs by evaluating their ability to integrate into host neural circuits and recover motor function in a PD mouse model. The researchers generated 3D MOs from human iPSCs and transplanted them into the striatum of immunodeficient mice to assess their survival, differentiation, and axonal innervation over extended periods, observing robust graft survival and continued maturation for up to 6 months post-transplantation. Their findings revealed that the transplanted MOs successfully matured into DA neurons, extended axonal projections into host brain regions, and established functional synaptic connections within the striatum. Critically, motor recovery was rigorously quantified using standardized behavioural tests (such as the cylinder test and rotarod), showing that mice receiving human MOs transplants exhibited significant and sustained improvements in motor function, suggesting that the transplanted organoids contributed to restoring dopaminergic signalling. Furthermore, comprehensive histological analysis at these long-term endpoints revealed no evidence of tumor formation or other overt adverse pathological events associated with the grafts, indicating a favourable safety profile within the observation period [36].

Similarly, Fu et al. generated human iPSC-derived MOs enriched for dopaminergic progenitors and transplanted dissociated organoid cells into the striatum of 6-OHDA-lesioned PD mice. The grafts survived long-term, with over 80% of engrafted cells differentiating into TH⁺ dopaminergic neurons that released dopamine, integrated into host circuits as shown by a genetically encoded dopamine sensor, and significantly improved motor deficits as early as four weeks post-transplantation [37].

To ensure these promising preclinical results can be replicated and safely translated, Pașca et al. (2025) propose a fourstage consensus pipeline to standardize neural organoid, assembloid, and transplantation research. First, human pluripotent stem cells undergo rigorous quality control—including pluripotency assays, genome integrity validation, and pathogen screening—to ensure a reliable starting material. Second, in vitro differentiation and characterization are standardized via singlecell transcriptomics, immunocytochemistry, morphological analysis, calcium imaging, and electrophysiology to uniformly assess organoid identity and function. Third, researchers assemble assembloids or guide organoid formation in a hypothesisdriven manner, tailoring culture conditions and regionspecific patterning to address specific biological questions. Finally, transplantation procedures are systematized to include donor consent, batchlevel differentiation QC, immunecompatibility testing, imagingbased engraftment monitoring, and postgraft functional assays. By emphasizing transparent reporting, data sharing, and alignment of experimental design with precise scientific objectives, this framework aims to enhance reproducibility, facilitate crosslaboratory comparisons, and accelerate the translational impact of organoid and assembloid technologies [38].

While this study underscores the potential of MOs-based cell therapy, further research is required to optimize transplantation protocols, improve graft survival, and ensure long-term integration without adverse immune responses. Challenges such as efficient vascularization, immune compatibility, and functional recovery beyond motor symptoms must be addressed before MOs can be translated into clinical applications for PD treatment. Nevertheless, studies like Zheng et al. (2023) and Fu et al. (2024) pave the way for future regenerative medicine approaches, offering a more physiologically relevant alternative to conventional stem cell-derived DA neuron therapies.

Limitations and challenges

MOs offer unparalleled opportunities for modelling human nigral development and PD, but their broad adoption is hampered by several interrelated technical challenges.

Reproducibility

Reproducibility remains a key bottleneck: organoid size, cellular composition, and DA subtype ratios (A9 vs. A10) vary greatly not only between laboratories but even between batches in the same lab. Factors such as hPSC line origin, passage number, precise morphogen timing and dose, and lot‑to‑lot variability in Matrigel all contribute to heterogeneity. Despite efforts to employ defined, xeno‑free media, synthetic hydrogels, and automated bioreactors, achieving the consistency required for drug screening or clinical applications has yet to be fully realized [23, 39–41].

Vascularization and immune components

Most MOs protocols like 2D monolayer cultures, lack intrinsic vasculature and resident immune cells. However, their 3D architecture both exacerbates and partially mitigates these shortcomings. As organoids grow beyond 200–500 μm in thickness, diffusion limitations become critical, often resulting in necrotic cores that are rare in thin 2D cultures, yet the same 3D spatial organization allows for more realistic cell–cell interactions which is a prerequisite for modeling complex immune crosstalk even in the absence of microglia. Although bioengineered scaffolds mimic blood vessel networks and improve oxygen and nutrient delivery [42], fully perfusable systems are still under development. Similarly, the absence of microglia and other glial cells hampers the modelling of neuroinflammatory pathways critical to PD. While integrating iPSC-derived microglia has been shown to increase synaptic density and enhance electrophysiological maturity, synchronizing their development with neurons remains challenging. Additionally, incorporating astrocytes, oligodendrocytes, and endothelial cells continues to be a persistent hurdle [43]. In contrast, 2D systems face identical hurdles for glial and endothelial integration but completely lack the structural context needed to recapitulate tissue-level pathophysiology.

Maturation issues

Even when organoids survive and differentiate, they rarely achieve adult‑like maturity. Additionally, MOs lack long-range afferent and efferent connections, limiting their capacity to model full nigrostriatal circuits. After 60–100 days in culture, they often resemble fetal or early postnatal tissue. For example, neuromelanin granules are underdeveloped, action potentials are immature, and key terminal differentiation markers: aldehyde dehydrogenase 1 family member A1 (ALDH1A1), G protein-activated inward rectifier potassium channel 2 (GIRK2), dopamine transporter (DAT), and vesicular monoamine transporter 2 (VMAT2) are expressed only weakly. Without afferent connections, systemic cues, or mechanical stimuli, organoid neurons cannot fully recapitulate adult electrophysiological and metabolic phenotypes. Additionally, MOs frequently exhibit necrosis in their central regions due to limited oxygen and nutrient diffusion, which restricts long-term viability and maturation. The lack of vascularization leads to hypoxic and apoptotic zones, especially in larger organoids, thereby compromising structural integrity and functional outcomes. Strategies such as extended culture beyond 200 days, in vivo transplantation, and co‑culture with striatal spheroids show promise, but are technically demanding and not yet standardized [44, 45].

Scalability and cost

Scalability and cost issues limit high‑throughput and translational uses. Generating uniform organoids requires weeks to months, expensive growth factors (e.g., SHH, FGF8, BDNF), specialized matrices, and manual quality control. Although microfabricated microwell arrays, suspension bioreactors, and robotic liquid handlers can improve throughput and reproducibility, the per-organoid cost remains prohibitive for large-scale screening. Moreover, downstream assays—such as high-content imaging, single-cell transcriptomics, and electrophysiology—add further expense [46]. By contrast, 2D cultures afford higher throughput and lower cost but cannot reproduce 3D cell–cell interactions.

Translational gaps

Many challenges in organoid research, such as batch variability and the absence of clinical candidates, which are also limitations of animal models and 2D systems, underscoring the need for complementary approaches. Despite proof-of-concept successes in identifying compounds that modulate PD phenotypes—such as BAG956 in OASIS-based screens or toxicity hits in automated pipelines [30] —questions remain around robustness and real-world translation. Most drug-screen studies use a limited number of iPSC lines and lack replication across diverse genetic backgrounds, making it unclear how broadly effective these compounds will be. Furthermore, no organoid-derived candidates have yet advanced to clinical trials, underscoring that current platforms remain predominantly in the discovery stage. Bridging this gap will require head‑to‑head validation against animal models, expanded sample sizes, and early‑phase pharmacokinetic and safety assessments in vivo.

Ethical considerations

While both 2D neuronal cultures and 3D organoid systems raise fundamental ethical issues around donor consent and biosafety, MOs introduce additional layers of complexity, most notably concerns about emergent network activity, moral status, and transplantation as quasi-tissue grafts.

First, as organoid complexity increases, questions about moral status arise. While most neuro-ethicists consider actual consciousness in organoids highly unlikely—especially in the absence of environmental stimulation—it is still wise to proceed cautiously. If the electrophysiological patterns of organoids begin to closely resemble those observed in living brains, adopting a precautionary approach becomes essential. This cautious stance emphasizes the need for robust, multidisciplinary guidelines to ensure responsible research and to effectively navigate the ethical challenges as these systems progress toward clinical applications [47].

Second, informed consent for donor cells must reflect both current uses and unforeseeable future applications. Broad, one-time consent may fail to capture donor preferences around complex assembloids or chimeric models; dynamic, project-specific consent with the option to withdraw is recommended to uphold autonomy and trust [48]. Consent documents should transparently discuss potential commercial uses, data privacy risks (e.g., genomic sequencing), and ethical oversight plans.

Third, transplantation risks demand rigorous preclinical evaluation. Beyond immunogenic rejection and tumorigenicity, grafted MOs may integrate in unpredictable ways, raising safety and long-term monitoring challenges. Regulatory agencies such as the FDA require Good Manufacturing Practice (GMP)–compliant protocols for any cell-based therapy; scaling MOs to GMP standards—defined media, xeno-free matrices, and batch release criteria—remains a major translational hurdle [49].

Finally, existing oversight frameworks offer mixed guidance. The International Society for Stem Cell Research (ISSCR) currently does not impose organoid-specific rules, and the National Academies recently declined to implement new regulations for human brain models [47]. However, as MOs approach clinical application, establishing clear ethical guidelines, informed-consent best practices, and GMP pathways will be essential to ensure responsible innovation and patient safety.

In summary, while midbrain organoids (MOs) hold tremendous potential for modelling human nigral development and PD, their application introduces unique challenges compared to traditional 2D cultures. MOs provide three-dimensional tissue architecture and multicellular interactions that better recapitulate in vivo midbrain biology, but also introduce barriers such as batch heterogeneity, hypoxic necrosis, and limited long-term maturation. On the technical side, MOs full application is limited by several unresolved technical and ethical challenges (Fig. 3). On the technical side, issues such as reproducibility, lack of vascularization and immune components, limited maturation, and scalability constraints impede broader adoption. Equally important, ethical considerations—including concerns over emerging moral status, informed consent for donor cells, and the safety of transplantation—further complicate their translational pathway. Overcoming these challenges will require multidisciplinary innovations encompassing bioengineering, standardized protocols, integrated multicellular systems, and robust ethical and regulatory frameworks. Continued progress in these areas is essential to transform MOs from promising experimental models into reliable tools for drug discovery and clinically responsible applications.

Fig. 3.

Current Challenges and Emerging Solutions in Midbrain Organoid Technology. Schematic representation of the major technical limitations in human MOs systems and the bioengineering innovations devised to address them. Batch-to-batch variability and inter-organoid heterogeneity can be overcome by automated high-throughput screening (HTS) platforms and robotic liquid-handling workflows, which enforce standardized timing, reagent dispensing, and quality control. The absence of vascularization and immune components—which leads to necrotic cores and incomplete maturation—is countered by vascular network–inspired scaffolds (e.g., VID), microfluidic perfusion chips, and co-culture with iPSC-derived endothelial cells, microglia, or T cells. To overcome persistent maturation issues and better model neural circuitry, researchers are developing multi-region assembloids that fuse different regions of the brain. Finally, the high cost and limited scalability of current protocols are being addressed through microfabricated disk technology, suspension bioreactors, and AI-driven process analytics, enabling hands-off generation of hundreds of uniform organoids for drug screening and translational applications

Future directions

In previous section, major roadblocks in MOs technology have been highlighted. In this section, we tackle each challenge in turn—summarizing both what has been demonstrated in the literature and what remains to be developed.

Vascularization and perfusion

To date, MOs vascularization has already been advanced by Cai et al.’s vascular network–inspired diffusible (VID) scaffolds, which significantly improved oxygen and nutrient diffusion, reduced core necrosis and hypoxia, and enhanced dopaminergic neuron survival and zonal architecture [42]. Building on this success, future work could embed iPSC-derived endothelial progenitor cells with neuroepithelial precursors in microfluidic devices. Researchers might use 3D bioprinting to create branched, vessel-like channels in hydrogels. In addition, integrating on-chip flow controls such as peristaltic pumps or micro-actuators can mimic natural blood flow. These steps help maintain steady oxygen and nutrient levels, which are crucial for the long-term maturation of organoids.

Immune integration and neuroinflammation

Thus far, integrating immune cells into human MOs markedly enhances their physiological relevance. For example, incorporating microglia into brain organoids not only provides an essential innate immune component but also improves neuronal maturation and functionality through mechanisms such as synaptic remodelling and reduced oxidative stress [43, 50]. Similarly, the development of co-culture models with peripheral T cells revealed how adaptive immune cells interact with neurons to influence neurodegeneration, particularly in PD [51]. These results suggest that immune-integrated organoids can recapitulate the complex interplay between neural and immune cells seen in the human brain.

Looking to the future based on previous findings, combining both resident microglia and infiltrating T cells within MOs represents an exciting approach to modelling human neuroinflammation and neurodegenerative conditions. Such multifaceted platforms would allow researchers to dissect how innate and adaptive immune responses together drive selective neuronal vulnerability and synaptic alterations, thereby offering valuable insights into disease mechanisms. Ultimately, these advanced co-culture systems could pave the way for precision medicine applications, enabling better screening of therapeutics tailored to modulate neuroimmune interactions in conditions like PD.

Automated high-throughput platforms

Currently, automated workflows for human MOs are paving the way for more reproducible and scalable systems for 3D-based chemical screening. In one study, a fully automated high-throughput workflow was established for generating, maintaining, and analysing MOs in standard 96-well plates, ensuring consistent morphology, gene expression, and functional neural activity across batches [31, 52]. This automated process not only minimizes inter-organoid variability but also enables robust, single-cell level assessments of drug effects, accelerating the discovery of novel compounds relevant for PD research.

Building on this progress, combining automated organoid systems with artificial intelligence (AI) offers a powerful new way to understand complex biological data. AI algorithms can handle the huge amounts of information generated by detailed organoid images and genetic activity measurements. This helps researchers spot subtle patterns and connections that might be overlooked otherwise. As demonstrated by recent work, these combined high-speed platforms allow scientists to identify important changes in cells without needing to know the exact cause first. This ultimately helps in developing more personalized treatments by linking how cells respond to potential patient outcomes [53]. Together, these innovations promise to enhance our ability to model midbrain pathology in vitro and to streamline the drug discovery process for neurodegenerative diseases.

Circuit-level assembloids

So far, spatially arranged assembloids that combine ventral midbrain, striatum, and cortex have shown active dopaminergic projections and synaptic connections. These structures mimic key features of the human nigrostriatal pathway and enables detailed studies of neuronal migration, projection patterns, and drug responses [54]. Similarly, midbrain–hindbrain assembloids engineered by Gomez-Giro et al. (2025) have faithfully modelled neuron-to-neuron spread of pathological α-syn aggregates in a caudo-rostral manner, providing a platform to dissect proteopathic seeding and test anti-spreading interventions [55].

Building on these proof-of-concept studies, future efforts should standardize multi-region fusion protocols to ensure reproducible circuit assembly, integrate optogenetic or chemogenetic actuators for controlled interrogation of pathway function, and combine live-cell calcium imaging and multielectrode array recordings to quantify dynamic network activity. Such advanced assembloid platforms will not only deepen our mechanistic understanding of PD-related network dysfunction but also serve as high-content screening tools for circuit-targeted therapeutics [56].

Microfabricated disks and scalability

Up to now, microfabricated disks technology, as described by Mohamed et al. (2022), offers a powerful solution to the longstanding issues of throughput and cost in MOs generation. Organoids were cultured in defined microwells etched into a rotating disk, which have tiny, identical wells where hundreds of organoids can grow at once while the disk spins. This makes the organoids much more uniform, reduces hands-on work, and cuts down on wasted materials. Because multiple disks can run together on simple equipment like a shaker, researchers can make many more organoids without a big jump in cost or effort [46]. This is key for large drug screens or studies before human trials.

Looking ahead, connecting these disks to other automated tools will make them even more powerful and reliable. Adding robot arms could handle feeding and dosing the organoids automatically. Built-in sensors could check organoid health without disturbing them. Tweaking the well shapes or disk materials might also improve nutrient flow, preventing dead spots inside larger organoids. Finally, pairing these automated disks with AI for quality checks could create a fully hands-off system for mass-producing organoids. This paves the way for cheaper drug discovery and even manufacturing for future cell therapies.

In summary, these emerging innovations—from perfusable vascular networks and immune-integrated co-cultures to fully automated, high-throughput platforms, sophisticated assembloid systems, and scalable microfabricated disks—collectively tackle the key limitations of MOs in PD research (Fig. 3). By enhancing oxygen and nutrient delivery, incorporating neuroimmune interactions, standardizing production, and faithfully recapitulating circuit complexity, these advances promise to transform organoids into reproducible, translational models. Ultimately, harnessing this multidisciplinary toolkit will accelerate drug discovery and pave the way toward clinical applications of MOs technologies.

Conclusion

Human MOs represent a significant advance in modelling the cellular complexity and selective neuronal vulnerability characteristic of PD. These 3D models more faithfully replicate tissue architecture than 2D monolayers, generating DA subtypes with functional properties closer to in vivo midbrain neurons and recapitulate key genetic and pathological hallmarks of PD. However, compared to 2D systems, MOs face greater challenges with protocol variability, vascular/immune component integration, neuronal maturation, and production costs, constraining their reproducibility and translational potential.

Moving forward, establishing community-wide standards for protocol reporting and inter-laboratory ring trials will be critical for benchmarking organoid fidelity. The incorporation of perfusable microvasculature, resident microglia, and other glial lineages will enhance physiological relevance and enable modelling of neuroinflammatory processes. Quantitative metrics—spanning transcriptomic, proteomic, electrophysiological, and metabolic domains—are needed to define and verify mature midbrain phenotypes. Finally, advances in microfabrication, automation, and AI-based analytics will drive down costs and enable high-throughput screening, while assembloid technologies will allow reconstruction of nigrostriatal and multisystem circuits implicated in both motor and non-motor PD features. By addressing these challenges, MOs are poised to become indispensable tools for elucidating PD mechanisms and accelerating the discovery of disease-modifying therapies.

Simultaneously, developing clear ethical and regulatory frameworks, covering organoid sentience assessments, dynamic donor consent, and GMP-compliant manufacturing, will be critical to translate these models safely into the clinic. By combining technical innovation with ethical rigor, human MOs are poised to revolutionize our understanding of PD pathogenesis and spearhead the discovery of truly disease-modifying therapies.

Author contributions

Z.D.Z, R.E reviewed the literature and drafted the manuscript. E.K.T. provide critical comments, revised and touched up the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by Singapore National Medical Research Council (NMRC) grants, including OF-IRG, CS-IRG, HLCA2024, ASTaR, OF LCG 000207, and a clinical translational research programme in Parkinson’s disease, as well as Singhealth-Duke-NUS AM Position grant.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.