Insights into neurodevelopmental features of Huntington's disease from stem cell-derived models including organoids

Introduction

Huntington's disease (HD) is primarily recognized as an adult-onset neurodegenerative disorder,^1,2 yet it also involves significant neurodevelopmental alterations well before clinical symptom onset. ³ Wild-type huntingtin (wtHTT) protein regulates neuronal survival, axonal and vesicular trafficking, and brain development, with wtHTT deletion causing embryonic lethality.^4–7 Given the diverse functions of wtHTT, it is not surprising that studies increasingly find that mutant huntingtin (mHTT) has adverse effects on neurodevelopment.

Comparisons between normal and HD fetal brains suggest that mHTT disrupts brain development during early stages, including the mislocalization of junctional complex proteins, which may contribute to impaired neural progenitor cell (NPC) polarity and differentiation. Specifically, fetal HD brains appear to have: (1) altered distributions of tight and adherens junction proteins; (2) reduced numbers of mitotic progenitor cells; (3) changes in the quantity and morphology of cilia at the apical surface; and (4) an increased ratio of intermediate progenitor cells to apical radial glia—two types of NPCs—in the ventricular zone of the developing cortex. ⁷ Research in mice, human fetal tissue, and organoids further suggests that mHTT subverts early developmental processes. The mutant protein may impair neuronal survival by disrupting the delivery of brain-derived neurotrophic factor (BDNF), decreasing cellular ATP/ADP ratios, impairing autophagy, altering ciliogenesis in striatal cells, and/or changing mitochondrial function.^5–9

Neurodevelopmental perturbations in HD leave lasting footprints in the adolescent and adult brain, manifesting as structural abnormalities including reductions in intracranial volume, ¹⁰ striatal enlargement, ¹¹ and neuronal migration defects such as periventricular nodular heterotopias—clusters of mispositioned neurons that disrupt cortical layering and architecture. ¹² HD is also associated with widespread dysregulation of neurodevelopmental gene expression, particularly in cortical regions, potentially predisposing these areas to later degeneration. ³ Beyond structural changes, HD mutation carriers exhibit functional alterations such as early hyperconnectivity between the striatum and cerebellum, which diminishes over time. ¹¹ Investigating these various neurodevelopmental features in human stem cell-derived models has emerged as a promising approach to understand the earliest molecular features of HD.

Multiple lines of evidence support the view that HD has a neurodevelopmental component, it remains unclear whether stem cell-based models consistently recapitulate these alterations and have the capacity to unveil novel disease features. This review highlights critical insights gained from studies using stem cells, stem cell-derived neural progeny, brain organoids, and assembloids to investigate the mechanisms underlying neurodevelopmental changes in HD (Figure 1). Specifically, we discuss the ability of stem cell-based model systems to reproduce disease-associated characteristics, the contributions of organoids to our understanding of HD pathogenesis, the advantages and limitations of each model system, and recent advances in computational and transcriptomic methods. Collectively, these models have yielded findings that support the view that HD is not purely a late-onset neurodegenerative disorder but also involves early neurodevelopmental disruptions that predispose the brain to later cognitive and motor impairments.

OPEN IN VIEWER

Figure 1.

Progression of stem cell models used in HD research, demonstrating advances from two- to three-dimensional tissue culture systems and major findings across the HD field. (A) Induced pluripotent stem cell (iPSC) lines show disease-specific gene expression patterns. Lines with longer CAG repeats are more vulnerable to stress. ¹³ (B) Mechanism by which FAN1 stabilizes mutant HTT (mHTT) CAG repeat in medium-sized spiny neurons (MSNs): sequestering MLH1 and PMS2, preventing the binding of MSH2 and MSH3 on hairpin loops formed by larger CAG repeats, thereby stopping mismatch repair (MMR) activation and CAG expansion. ¹⁴ (C) HD neuruloids display reduced lumen size, an expanded central PAX6 domain, impaired radial symmetry, and a weakened or absent collagen layer in the central neural structure. ¹⁵ (D) Cerebral organoids with the HD-causing mutation demonstrate a downregulation of CHCHD2 (coiled-coil-helix-coiled-coil-helix domain containing 2) particularly in NPCs leading to neurometabolic failures and mitochondrial dysfunction. HD organoids also display aberrant cytoarchitecture and premature differentiation. ¹⁶ (E) HD striatum (Str) and midbrain substantia nigra (SN) assembloids display projection defects of MSNs and reduced calcium activity. ¹⁷ HD = Huntington's Disease. hESCs = Human Embryonic Stem Cells. PGD = Preimplantation Genetic Diagnosis. ASOs = Antisense Oligonucleotides. Created with BioRender.com.

Recapitulating disease-associated characteristics in stem cell models

Functional and developmental capabilities of 2D and 3D HD models—utility, strengths, and limitations

Recent findings using brain organoids

This section highlights the use of organoids and assembloids to model HD, focusing on neurodevelopmental abnormalities and transcriptional changes observed in these model systems. It covers defects in cortical layering and changes in organoid size linked to impaired neuroepithelial organization. We also emphasize the value of assembloids for studying interregional brain interactions and address ongoing efforts to refine organoid models to both understand HD pathogenesis and identify potential therapeutic targets.

In addition to mirroring neurodevelopmental patterns of the human fetus, organoids have complex organ-like spatial cell arrangements that foster both paracrine and direct contact-mediated interactions.^32,52 Compared to stem cells and their neural progeny, brain organoids possess two features that may offer further insight into pathogenic mechanisms. First, because they consist of differentiated cells, the enhanced quality control mechanisms reported in stem cells are likely absent.^22,53 Second, their adaptability to extended culture durations (e.g., four months ⁹ ) enables them to mimic neurodevelopmental stages and provides sufficient time for early disease-associated characteristics to emerge. Although organoids cannot replicate the age-dependent features of HD or other neurodegenerative disorders, their multicellular composition and long-term viability in culture enable investigations of early disease-associated characteristics and their variation across cell types.

Recent research suggests that altered brain development sets the stage for neurodegeneration during adulthood.^1,7 Given that brain organoids recapitulate aspects of neurodevelopment, researchers have explored whether this model system mirrors the aberrant developmental patterns reported in HD fetal brains.^{9,15–17,33,34,43,50} Specifically, investigators have sought to determine if organoids similarly exhibit: (1) altered distributions of tight and adherens junction proteins^9,15,16,33; (2) reduced numbers of mitotic progenitor cells^{9,16,17,33,43}; (3) changes in the quantity and morphology of cilia at the apical surface^9,33,34; and (4) an increased ratio of intermediate progenitor cells to apical radial glia in the ventricular zone of the developing cortex. ³³ In general, organoid and assembloid studies mimic the altered neurodevelopmental patterns observed in HD fetal brains.^{7,9,15–17,33,34,43,50} With respect to organoid size, the literature is inconsistent on whether HD organoids are larger or smaller than their wild-type counterparts.^{9,16,33,34,43} Of the five studies available for review, three reported that HD organoids are typically smaller.^9,16,33 This size reduction likely stems from the presence of smaller neural rosettes, indicating a reduced apical radial glia population and impaired neuroepithelial organization.^{15,16,33,43,50}

Close comparison of the transcriptional profiles of HD versus wild-type organoids has provided insights into the mechanistic underpinnings of early neurodevelopmental changes. In HD organoids, upregulated genes are primarily associated with the following: neurodevelopment, epigenetic regulation, cytoskeletal organization, extracellular matrix organization, cell cycle regulation, stress response, and protein translation/degradation.^{9,16,33,34,43,50} In contrast, downregulated genes are largely linked to the following: later stages of neurodevelopment (e.g., neuronal and glial maturation and migration), ventral telencephalon development, interneuron specification, cytoskeletal organization and motility, cell cycle regulation, and neurotransmission and metabolic signaling—particularly within GABAergic and glutamatergic pathways.^{9,15,16,33,34,43,50} Many differentially expressed genes overlap with those identified in postmortem brain tissue from HD patients, suggesting that their dysregulation is an early event and persists throughout the patient's life.^9,15 This sustained dysregulation may contribute to ongoing neurodegeneration and symptom progression, highlighting potential targets for therapeutic intervention across different stages of the disease.

Collectively, these transcriptional changes likely contribute to a developmental shift in HD organoids involving premature neurogenesis and delayed maturation of the resulting postmitotic neurons.^16,33,34,43 In the developing cerebral cortex, neurons migrate along apical radial glial scaffolds to the cortical plate, where they are sequentially deposited in an inside-out pattern to form the six-layered cortical structure. Early born neurons populate the deeper layers (layers 6 and 5), while later-born neurons contribute to the upper layers (layers 4-2). As they migrate toward their final destinations, neurons transiently express markers specific to the cortical layers they traverse.^54–56 Expanding on this in the context of HD pathogenesis, Liu et al. observed an accumulation of cortical layer 5 neurons (CTIP2⁺) in HD cortical organoids, accompanied by a reduction in upper-layer neurons (SATB2⁺, layers 4-2). ³³ The accumulation of layer 5 neurons in HD organoids suggests that upper-layer neurons fail to complete their migration. ³³ To assess whether this developmental shift renders HD organoids ‘older’ or ‘younger’ relative to their wild-type counterparts, many studies have analyzed organoid transcriptional profiles alongside those of human fetal brains. To date, these findings are mixed: some studies suggest HD organoids are ‘older’, while others suggest they are ‘younger’.^33,43,50 Given the role of mitochondria in cellular aging, these shifts in developmental timing may relate partially to mHTT-induced mitochondrial dysfunction. Organoid studies show that mHTT disrupts mitochondrial dynamics, reduces oxidative phosphorylation, and contributes to neurodegeneration.^16,57 These findings align with broader mitochondrial impairments in HD, including deficits in energy metabolism and neuronal signaling that are critical for cognitive function and synaptic plasticity.

In summary, the HD organoid field remains in its early stages, with much still to be uncovered. Scientists are actively refining organoid generation protocols and beginning to elucidate the cellular and molecular mechanisms underlying HD pathogenesis. Below is a summary of key organoid studies from the past year (2024–2025).

Traditionally, organoid research has relied on models of single brain regions. Liu et al. employed HD cortical organoids as a spatiotemporal model to characterize neurodevelopmental defects. ³³ Organoids carrying CAG repeat expansions in the adult-onset range (55 or 59 repeats) had disrupted junctional complexes in neural tubes, delayed maturation of postmitotic neurons, aberrant specification of cortical neuron subtypes, and abnormal subcortical projections during corticogenesis. Regardless of disease status, the organoids contained polyQ assemblies associated with Golgi stacks and clathrin-positive vesicles; however, the structural and functional characteristics of these assemblies differed between genotypes. In HD organoids, the polyQ assemblies were shorter and scaffolded fewer Golgi stacks and clathrin-positive vesicles, which coincided with reduced ADP-ribosylation factor 1 (ARF1) recruitment to these assemblies. Pharmacological inhibition of ARF1 in wild-type organoids reproduced the junctional complex defects present in HD organoids, implicating ARF1-mediated Golgi dysfunction as a downstream consequence of mHTT. Together, these findings link polyQ pathology to disrupted cellular organization during cortical development and provide mechanistic insight into how mHTT impairs corticogenesis. ³³ Lisowski et al. added to the growing body of HD research by generating isogenic iPSC lines with 70 CAG/CAA repeats in one or both HTT alleles, as well as a line in which the repeat expansion was eliminated from both alleles. ¹⁶ They conducted a comparative analysis of iPSCs, NPCs, and cerebral, cortical, and midbrain organoids derived from these lines. This approach led them to identify coiled-coil-helix-coiled-coil-helix domain containing 2 (CHCHD2)-mediated neurometabolic failure as an early feature of HD pathogenesis. ¹⁶

As another example of brain region-specific modeling, Galimberti et al. generated telencephalic organoids that include both dorsal and ventral cell fates. ⁹ During human neurodevelopment, the telencephalon gives rise to the cerebral cortex and basal ganglia, including the striatum. Because the cerebral cortex and striatum are the two most vulnerable brain regions in HD, this model holds promise for uncovering pathogenic mechanisms. The telencephalic organoids were derived from a panel of isogenic hESC lines with varying CAG repeat expansion lengths (20, 48, 56, or 72 repeats). HD organoids displayed abnormalities in ventral fate acquisition, resulting in fewer GABAergic neurons. In contrast, dorsal cell populations were less affected and primarily showed defects in maturation. ⁹

In HD patients, selective regional vulnerability and SRI lead to a mosaic of repeat expansion lengths across brain cells. To investigate the role of this mosaicism in HD pathogenesis, Galimberti et al. generated organoids composed of both mutant and wild-type cells. Compared to monoculture HD organoids, many abnormalities were rescued in the mosaic organoids. These findings suggest wild-type cells exert a beneficial non-cell-autonomous effect, which could be leveraged for therapeutic development. ⁹

To capture interregional interactions and cell diversity, the organoid field has increasingly prioritized the production and examination of assembloids. Notably, Wu et al. were the first to develop a protocol for generating HD organoids that exhibit mHTT aggregates rather than assemblies. ¹⁷ This protocol involves fusing striatum and midbrain substantia nigra (SN) organoids to form Str-SN assembloids. As this model mimics other disease-associated characteristics, including impairments in neuronal projections and decreased calcium signaling, it holds promise as a platform to study disease progression and test potential therapeutic strategies. ¹⁷ Świtońska-Kurkowska et al. were the first investigators to generate brain organoids modeling juvenile-onset HD that are larger than their wild-type counterparts. ³⁴ This may be due to their use of fused dorsal-ventral assembloids, which mimic interactions between the developing cortex and striatum. These HD assembloids had enlarged neural rosettes and impaired neurogenesis, marked by delayed neuronal maturation and elevated expression of proliferation-related genes. While the impaired neurogenesis aligns with previous findings from cerebral and cortical organoids,^16,33,43 the presence of enlarged neural rosettes contrasts with earlier reports.^{15,16,33,43,50} In addition to having altered proportions of inhibitory neurons and glia, HD assembloids also exhibited increased abundance of choroid plexus cells. These cells secrete transthyretin (TTR) into the cerebrospinal fluid, from which it may enter the bloodstream—highlighting TTR as a potential blood-based biomarker for HD. ³⁴

Comparison of the ease and utility of various stem cell-derived model systems

Table 1 summarizes and compares features of stem cell models. iPSCs and hESCs provide genetically defined, indefinitely self-renewing platforms for modeling HD. While it can be challenging to create isogenic iPSC and hESC lines, doing so is critically important as isogenic controls eliminate potentially spurious differences elicited by genetic changes outside of the disease mutation. Stem cells require careful maintenance to prevent unwanted differentiation, yet their capacity to differentiate into any cell type makes them a powerful tool for studying disease mechanisms. While they can model transcriptional dysregulation and mimic early HD neurodevelopmental stages, they are limited in their ability to replicate cellular circuitry. Moreover, they only fully exhibit aberrant HTT behavior or SRI when proteasomal stressors or hyper-expanded HTT are present—a limitation likely due to the enhanced quality control mechanisms present in stem cells.^22,53

OPEN IN VIEWER

Table 1.

Comparison of 2D and 3D stem cell-based models for HD research.

This table summarizes the ability of human stem cell-derived systems to recapitulate key disease-associated characteristics of HD. Created with BioRender.com.

NPCs, which can be generated from pluripotent stem cells in approximately 3–4 weeks, ⁵⁸ require less maintenance than their undifferentiated counterparts, making them attractive for functional assays and drug screening. Differentiated neurons can model aberrant HTT behavior, SRI, transcriptional dysregulation, cellular circuitry, and early neurodevelopmental stages, making them a compelling option for modeling disease pathology. The challenge of maintaining neurons for extended periods, however, may prevent certain disease-associated characteristics from being captured. Moreover, some features may only manifest in the context of complex, multicellular environments.

Brain organoids provide greater structural and cellular complexity but are technically demanding due to variability in their maturation. They often require 1–2 months of culture to develop meaningful architecture and HD-relevant phenotypes. Organoids permit broader modeling of aberrant HTT behavior, cellular circuitry, transcriptional dysregulation, and neurodevelopmental changes across different areas of the brain. Unguided differentiation relies on the intrinsic capacity of pluripotent stem cells to self-pattern and produce diverse brain-like regions, offering heterogeneity and the potential to model interregional interactions. ⁵⁹ These benefits also make the organoids highly variable, raising concerns about reproducibility. In contrast, direct differentiation relies on applying defined differentiation factors to guide organoid development towards specific regional identities, such as dorsal forebrain or midbrain. This ensures greater reproducibility and regional precision while at the same time sacrificing diversity and spontaneous organization that may reveal emergent developmental phenotypes.

Assembloids augment the ability of organoids to model circuitry, but are the most labor intensive, requiring the coordinated fusion and maintenance of two or more region-specific organoids. While offering invaluable information, their reproducibility and scalability remain limited.

Two-dimensional iPSC, NPC, and neuronal cultures primarily capture cell-autonomous processes, such as alterations in gene regulation and proteostasis that are intrinsic to affected cells. In contrast, organoids and assembloids facilitate investigation of non-cell-autonomous mechanisms, including interactions between neuronal subtypes, between neurons and glia, ⁶⁰ or between different brain regions. Overall, NPCs offer the balance of speed and accessibility, while organoids and assembloids sacrifice ease for greater neurodevelopmental and in vivo disease relevance.

Advanced computational and transcriptomic methods are opening new doors in organoid research:

We conclude by highlighting the still unrealized potential of organoids. Incorporation of artificial intelligence (AI) and deep learning strategies into brain organoid research is likely to transform the analysis of neurodevelopmental and neurodegenerative processes. For example, AI-based analysis resulted in the development of a high-throughput, micropatterned, brain organoid-based platform for drug screening. ⁶¹ This platform employs deep convolutional neural networks (CNNs), which extract features that distinguish HD from non-HD organoid images, such as the expansion of neuroepithelial cores. The AI-based system detected subtle rescue effects evoked by bromodomain inhibitors on neurodevelopmental abnormalities in HD organoids. ⁶¹ The potential of AI to quantify both the toxic and beneficial effects of pharmacological treatments could be invaluable to the drug discovery process.

In addition to evaluating phenotypic rescue, AI can be used to analyze organoid structure and development by enhancing high-resolution imaging modalities and facilitating the interpretation of complex datasets. An interesting example is an AI-optimized segmentation and analysis pipeline that combines 3D volumetric magnetic resonance imaging (MRI) data with deep learning algorithms to produce 3D masks of organoids and segment aberrant features, such as cysts. ⁶² This pipeline enables longitudinal monitoring of cerebral organoid growth and, as a result, supports automated assessment of organoid quality, making it easier to track structural changes over time. ⁶² The integration of AI with advanced imaging analysis has the potential to uncover disease-associated characteristics that are often overlooked in 2D and 3D model systems.

Bulk and single-cell RNA sequencing (RNA-seq) have revolutionized biomedical research. Both approaches, however, require tissue dissociation, hindering their ability to assess spatial changes and cell-cell interactions. To address this limitation, researchers used MRI data from pre-manifestation HD patients alongside postmortem single-nucleus RNA-seq data to correlate cell loss with dysregulated genes, including those involved in development. ³ More recently, spatial transcriptomics have provided high-resolution gene expression maps of intact brain tissue and organoid sections, preserving their spatial organization. In HD mice, spatial transcriptomics revealed regional and age-dependent changes. These changes were detected at birth, with the gene expression profile at postnatal day 0 (P0) displaying a compensatory activated state among many cell types that shifted towards downregulation by 4 weeks of age. Using gene expression maps, the authors assessed regional vulnerability within the brain. They found heightened glycolysis in the caudate putamen as early as P0, alterations to MSN identity and maturation, and enrichment of astrocytes in the ventricular zone. ⁶³ Single-cell transcriptomics has also revealed changes in glia, astrocytes, and striatal cell states as well as compensatory mechanisms,^64–66 matching some of the results gleaned using spatial transcriptomic data.

Likewise, organoid morphology inferred by spatial transcriptomics can be used to determine when and how cellular and tissue architecture deviates from normal.^67,68 For example, in HD midbrain organoids, single-cell gene expression maps have revealed upregulation of genes related to synaptic signaling, axon guidance, and neurodevelopmental defects that can be aided by spatial mapping to delineate where changes occur in relation to organoid structure and infer cell-cell interactions. ¹⁶ Rather than simply examining the gene expression map of an organoid section, spatial transcriptomics can help construct 3D molecular maps of organoids using serial sections and probe-based arrays. ⁶⁹ Overall, spatial transcriptomics is already being used in HD mouse models and to visualize brain organoid transcription profiles, so utilizing it in HD organoids may further our understanding of developmental alterations resulting from the disease.

Beyond advances in AI and transcriptomics, organoid intelligence-based biocomputing represents an exciting frontier. ⁷⁰ Recent research demonstrates that brain organoids harbor the essential molecular machinery and functional properties required for learning and memory, including synaptic plasticity, expression of immediate early genes, and highly interconnected, maturing neural networks bearing signatures of criticality. Furthermore, high-density microelectrode array recordings and stimulation of highly patterned electrical activity demonstrate that organoids are capable of both spontaneous and evoked, input-specific forms of plasticity. These advances speak to their potential as dynamic in vitro model systems to characterize disease-related network dysfunction. ⁷¹ Researchers have expanded organoid computing reservoirs by creating paired systems in which two organoids establish functional network connections. Experiments using virtual environments with reward and penalty stimuli demonstrate that these paired organoid systems can learn to “navigate” a digital maze. ⁷² Paired organoid reservoirs could be integrated into drug-discovery platforms to assess drug-induced changes in cognition, which cannot be assessed using 2D systems, and bypass blood-brain barrier permeability issues encountered in in vivo systems. In the study of HD, for example, high-density microelectrode array recordings from disease and non-disease organoids could serve as functional read-outs for changes in organoid network formation.

Together, these developments highlight the convergence of biological and computational sciences, revealing novel opportunities to understand and treat HD. The dual functionality of organoids as both disease models and biocomputing units underscores their potential as powerful tools with which to study disease mechanisms and advance therapeutic discovery. By bringing together AI, spatial transcriptomics, and biocomputing, researchers may soon achieve fundamental advances in our understanding and treatment of neurodevelopmental disturbances in HD and other neurological disorders.