Our mission
Brain disorders remain among the most difficult diseases to understand because both the nervous system and the disorders that affect it are extraordinarily complex. The human brain is assembled over a prolonged period through coordinated programs of cell specification, migration, maturation and circuit formation, and many disease-associated genetic changes exert their effects during these difficult-to-access stages (Cell, 2022).
Our premise is that by reconstructing key steps of human neural development and gaining experimental access to the molecular, cellular and circuit processes that build the nervous system, we can uncover biology that has remained hidden and understand how it is altered in disease.
Over the years, our laboratory has taken a systematic approach to this problem. Starting with patient-derived neurons, we developed regionalized neural organoids to capture defined stages and regions of human development, assembloids to reconstruct interactions among cells and brain regions, and xenotransplantation approaches to study human neural circuits as they mature and function in vivo (reviewed in Nature, 2018; Science, 2019; Cell, 2022). Together, these complementary systems provide experimental access across levels of organization, from molecular and cellular programs to migration, connectivity and circuit function.
We use these approaches both to uncover fundamental principles of nervous system development and to determine how disease-associated genetic variation disrupts them. Our work is rooted in basic science, with an eye toward revealing biology that is particularly relevant to neurological and psychiatric disorders. In selected cases, mechanistic discoveries point to ways of intervening in disease, allowing us to carry insights from human models toward therapeutic development and the clinic (Annual Review of Neuroscience, 2025).
This combination of stem cell-derived systems, circuit reconstruction and in vivo experimentation is helping define an emerging field of stem cell neuroscience: using experimentally accessible human neural systems to study aspects of nervous system development, function and disease that have otherwise been difficult to reach.
How human neural cells develop
Our work with human pluripotent stem cells began soon after cellular reprogramming made it possible to derive neurons from individual patients. Using hiPS cell-derived neurons in 2D culture, we showed that cells from individuals with Timothy syndrome recapitulate molecular and cellular features of the disorder (Nature Medicine, 2011).
To capture later stages of corticogenesis that were difficult to reach in conventional 2D cultures, we moved into 3D. Building on the capacity of pluripotent stem cells to self-organize, we introduced the use of instructive developmental signals to derive regionalized neural organoids with defined identities. Our first cortical spheroids (later called cortical organoids) progressively generated major cortical cell types, including excitatory neurons and astrocytes, and allowed us to characterize their physiological maturation (Nature Methods, 2015). We subsequently extended this strategy to generate subpallial, striatal, thalamic, spinal cord and other neural identities, as well as oligodendrocytes and additional neuronal and glial populations (Neuron, 2017; Nature Neuroscience, 2019; Nature Biotechnology, 2020; Cell, 2020; Nature, 2025; Science, 2025).
A major focus has been making these preparations reproducible and scalable. We benchmarked them against primary human tissue, including through single-cell atlases of the developing cerebral cortex and spinal cord (Cell, 2021; Nature Neuroscience, 2023), quantified reproducibility across lines and experiments (Nature Methods, 2019), and developed approaches for producing organoids at scale (Nature Biomedical Engineering, 2025). This has enabled studies across increasingly large numbers of human stem cell lines, including nearly 100 patient and control lines in a recent autism study (Nature, 2026).
More recently, we moved beyond developing one regional preparation at a time to systematically interrogating how developmental signals generate neural diversity. By screening combinations of instructive signals, we identified conditions that produce a much broader repertoire of neural identities and provide a framework for studying how regional and cellular fates are specified across the nervous system (Cell Stem Cell, 2024).
Long-term culture has also given us access to another defining feature of human development: its tempo. We maintained some cultures for years, allowing us to follow astrocytes as they progressed from fetal-like toward more mature states (Neuron, 2017), map gene regulation and chromatin across prolonged stages of corticogenesis (Science, 2020), and show that developmental timing is remarkably preserved in vitro. Cortical organoids undergo a prenatal-to-postnatal-like transition at approximately 9-10 months, including the canonical NMDA receptor subunit switch (Nature Neuroscience, 2021). These systems allow us to ask what sets the unusually prolonged tempo of human neural development and how it may be altered in disease.
How human circuits assemble
Neural regions do not develop in isolation. To study interactions between independently specified tissues, we introduced assembloids, bringing together regionalized neural organoids and cell populations in a modular system.
In our first forebrain assembloids, we combined cortical and subpallial tissues and directly followed human interneurons as they migrated into the cortex and formed functional connections with excitatory neurons (Nature, 2017). These experiments revealed a remarkable capacity for human neural tissues to self-assemble ex vivo: when the appropriate developmental components are brought together, cells carry intrinsic programs that help guide migration, recognition and circuit formation.
We then extended this approach to other circuits. Cortico-striatal assembloids allowed us to study long-range projections and functional connectivity between cortex and striatum (Nature Biotechnology, 2020), while cortico-spinal-muscle assembloids reconstructed a descending motor pathway in which cortical activity propagates through spinal motor neurons to drive muscle contraction (Cell, 2020). More recently, thalamocortical assembloids provided access to reciprocal thalamus-cortex connectivity and its disruption by disease-associated variants (Neuron, 2024); a four-part ascending somatosensory assembloid linked sensory neurons, spinal cord, thalamus and cortex to examine information flow across a human sensory pathway (Nature, 2025); and midline assembloids recreated organizer activity and commissural axon guidance in the developing spinal cord (Science, 2025).
These systems allow us to dissect how cells migrate, axons find their targets, regions become functionally coupled and activity propagates through developing circuits. They also make it possible to perturb defined components systematically. Using large-scale CRISPR screens in assembloids, we have mapped disease-associated genes onto specific developmental processes and cell-cell interactions (Nature, 2023).
Some questions require the environment of a living brain. We therefore developed approaches to transplant human cortical organoids into the developing rodent brain, where they become vascularized, mature further, receive sensory input and participate in functional circuits (Nature, 2022). More recently, in developmental xenocortication, we transplanted human cortical organoids into mice with extensive pallial depletion, creating space for human cortical tissue to develop, diversify and integrate in vivo (Nature, 2026). These approaches extend the study of human circuit assembly from defined ex vivo systems to maturation, integration and function in the living brain.
Disease mechanisms
Many neurological and psychiatric disorders are genetic, but knowing the gene is not the same as knowing how disease develops. We use patient-derived and genetically engineered models to ask when during human development a perturbation first becomes consequential, which cell types and processes are affected, and whether a phenotype arises within a cell or emerges through interactions with other cells and circuits.
This approach has allowed us to study disease across very different levels of neural organization. Work on CACNA1C in Timothy syndrome progressed from abnormalities in patient-derived neurons to defects in interneuron migration and circuit activity (Nature Medicine, 2011; Cell Stem Cell, 2022). In 22q11.2 deletion syndrome, cortical models from multiple individuals revealed abnormalities in neuronal excitability and calcium signaling (Nature Medicine, 2020). Cortico-striatal assembloids derived from individuals with 22q13.3 deletion involving SHANK3 uncovered abnormalities in human cortico-striatal circuit function (Nature Biotechnology, 2020).
As we reconstruct additional circuits, we can ask how pathogenic variants act in the developmental context in which the affected gene normally functions. Thalamocortical assembloids revealed circuit consequences of CACNA1G variants associated with neurological disease (Neuron, 2024), while our four-part ascending sensory assembloid showed how loss- and gain-of-function variants in SCN9A produce opposing effects on activity across the human sensory pathway, connecting a sodium-channel defect to circuit-level phenotypes associated with pain insensitivity and extreme pain (Nature, 2025).
We also ask whether genetically distinct disorders converge on common developmental biology. In a large study spanning diverse genetic forms of autism, we found both convergence and divergence across molecular programs, cell types and developmental trajectories (Nature, 2026). At a different scale, CRISPR screens in forebrain assembloids allowed us to test hundreds of neurodevelopmental disorder genes in parallel and map them onto defined processes such as interneuron generation and migration (Nature, 2023).
Not all insults are genetic. We have also used these models to study hypoxic injury of the developing human brain. Cortical organoids revealed selective vulnerability of intermediate progenitors to oxygen deprivation and implicated the unfolded protein response, findings that we validated in primary human tissue (Nature Medicine, 2019). More recently, xenocortication allowed us to revisit hypoxic injury in vivo and measure cellular, imaging and behavioral consequences in mice carrying large human cortical grafts (Nature, 2026).
Together, these studies aim to move from association to mechanism: identifying where, when and how genetic or environmental perturbations alter the development and function of human neural circuits.
Therapeutic development
One of our major goals is to carry mechanistic discoveries from human stem cell models toward therapeutic development and, when possible, the clinic. Timothy syndrome provides a long-term example of this approach, in which successive models of the same disorder revealed increasingly precise levels of mechanism and eventually pointed to a therapeutic strategy.
We began with hiPS cell-derived neurons from individuals with Timothy syndrome, identifying abnormalities in calcium signaling and excitability (Nature Medicine, 2011). As we moved into cortical organoids, we found that the disease-causing CACNA1C variant disrupts a developmentally regulated switch in exon usage. Forebrain assembloids then revealed abnormalities in interneuron migration and allowed us to dissect how altered calcium-channel signaling disrupts this developmental process in a multicellular context (Cell Stem Cell, 2022).
These findings suggested a way to intervene at the level of CACNA1C RNA processing. We developed antisense oligonucleotides (ASOs) that shift splicing away from the disease-associated exon 8A and toward the alternative exon 8. We showed that this strategy corrects molecular and physiological abnormalities in patient-derived neurons and cortical organoids, rescues interneuron migration in assembloids, and reverses disease-associated phenotypes after transplantation of human cells in vivo (Nature, 2024).
The Timothy syndrome program therefore traces an arc from patient-derived neurons to developmental mechanism, circuit-level phenotypes, in vivo validation and therapeutic development. The program is now advancing toward a clinical trial.
Ethics, community and training
As human neural models become more complex, ethical questions increasingly become part of the science itself. We have worked with bioethicists, neuroscientists, social scientists, patient advocates and other scholars to examine the implications of organoids, assembloids, transplantation and related approaches, from early discussions of organoid research to broader international efforts (Nature, 2018; Nature Neuroscience, 2022). We also convened interdisciplinary meetings at Asilomar to consider the scientific and ethical challenges raised by increasingly complex human neural models and to develop principles for their responsible study. These discussions contributed to an international Asilomar report on the responsible development of stem cell-based neural models (Science, 2025).
We have also worked to build a common framework for the field. Together with colleagues, we helped establish nomenclature and practical standards for generating, characterizing and reporting neural organoids and assembloids (Nature, 2022; Nature, 2024). These efforts are intended to make experiments more reproducible, results easier to compare and the limitations of each model more explicit.
Finally, we have placed strong emphasis on sharing methods and training others to use them. Through open protocols, hands-on courses, international workshops and scientific meetings, hundreds of investigators have been trained in these approaches, and methods developed in the lab have now been implemented by hundreds of groups worldwide. Our goal is not only to develop new experimental systems, but to help build a rigorous, open and responsible community around stem cell neuroscience.