Brain-on-a-chip devices are thumbnail-sized microfluidic platforms that grow living human neurons and supporting brain cells inside engineered compartments, recreating aspects of brain structure and function that flat cell cultures and animal models cannot. Built using techniques borrowed from the semiconductor industry, these chips allow researchers to watch neural circuits form, study diseases like Alzheimer’s in real time, and test drugs on human tissue without opening a skull. The technology sits at the intersection of neuroscience, materials engineering, and stem cell biology, and it has matured rapidly enough that U.S. law now formally recognizes organ-on-a-chip platforms as legitimate alternatives to animal testing in preclinical drug development.
How the Chips Are Made
Most brain-on-a-chip devices are fabricated using some combination of soft lithography, photolithography, and 3D printing. The dominant structural material is a silicone-based polymer called polydimethylsiloxane, or PDMS, which is optically transparent, flexible, nontoxic, and permeable to gases. That combination of properties makes it possible to culture living cells inside the chip while imaging them under a microscope in real time. Other materials used for specific applications include polycarbonate, glass, silicon wafers, and biological scaffolds made from collagen, matrigel, or hyaluronic acid.1PubMed Central. Microfluidic Brain-on-a-Chip: Perspectives for Mimicking Neural System Disorders
The basic architecture involves tiny channels, typically measured in micrometers, etched into these materials and then sealed together in layers. Fluid flows through the channels at precisely controlled rates, delivering nutrients, removing waste, and exposing cells to mechanical forces that mimic conditions inside the body. What makes this more than a fancy petri dish is the geometry: the channels can be shaped to separate different cell populations into distinct compartments while still allowing them to communicate through narrow connecting passages. That spatial control is the foundation for everything that follows.
Recreating the Blood-Brain Barrier
One of the most important structures in the brain, and one of the hardest to study, is the blood-brain barrier. This tightly sealed lining of brain capillaries controls what gets from the bloodstream into the brain. Most drug candidates for neurological diseases fail precisely because they cannot cross this barrier, and traditional lab models using flat membrane inserts do a poor job of replicating the conditions that keep the barrier tight.
Microfluidic chips have a real advantage here because they can expose endothelial cells to fluid flow. In actual brain capillaries, blood pushes against the vessel lining with a wall shear stress of roughly 0.3 to 2 pascals. That mechanical force is not just incidental; it actively promotes the formation of tight junctions between cells, making the barrier less leaky. Conventional well-plate setups cannot replicate this shear, but chip-based systems can.2PubMed Central. Microfluidic organ-on-chip technology for blood-brain barrier research One early demonstration showed that applying shear stress to endothelial cells grown on a chip tripled the transendothelial electrical resistance, a direct measure of how tight the barrier is.3PubMed. BBB on chip: microfluidic platform to mechanically and biochemically modulate blood-brain barrier function
Beyond shear stress, these chips also incorporate multiple cell types. The real blood-brain barrier is not made of endothelial cells alone; it depends on interactions with astrocytes and pericytes, the supporting cells that wrap around blood vessels in the brain. Research using induced pluripotent stem cells (iPSCs), which are adult cells reprogrammed to behave like embryonic stem cells, showed that iPSC-derived neural cultures could support endothelial maturation just as well as primary astrocytes and pericytes, achieving comparably low permeability.4Cell Stem Cell. A Human Blood-Brain Barrier Organ-Chip Recapitulates In Vivo-like Barrier Functions That finding matters because it means researchers can build an entire barrier from a single patient’s cells, opening a path toward personalized models.
Wiring Neural Circuits With Directional Control
Neurons in the brain do not connect randomly. Information flows in specific directions along defined pathways, and recreating that directional wiring has been a central engineering challenge. Microfluidic chips address this by using asymmetric microchannels, passages shaped so that axons can grow forward through them easily but have difficulty growing backward. Researchers found that by tuning the geometry of these channels, they could guide axons from hippocampal neurons to form predefined connections between two compartments within the first six days of culture. When those chips were coupled with microelectrode arrays, the recordings confirmed that electrical signals propagated in the intended direction, and this organized connectivity lasted for up to 25 days.5PubMed Central. Design of Cultured Neuron Networks in vitro with Predefined Connectivity Using Asymmetric Microfluidic Channels
The ability to impose directionality distinguishes these systems from traditional neural cultures where cells connect in whatever pattern chance dictates. If you want to model a circuit where region A sends signals to region B but not the reverse, asymmetric channels give you that lever. Combined with microelectrode arrays that can both stimulate and record from individual electrodes, researchers now have a way to build simple neural circuits with known architecture and then probe how those circuits respond to drugs, toxins, or disease-related mutations.6PubMed Central. Microelectrode Arrays Technology for Brain-on-a-Chip Applications
Multi-Region Models and Brain-Wide Connectivity
The human brain is not one uniform tissue. Cortex, hippocampus, substantia nigra, and striatum each contain different mixes of neuron types and neurotransmitter systems. When neurons from different brain regions are cultured together in a single undivided dish, they merge into an unstructured mass. A multiregional brain-on-a-chip solves this by housing neurons from different areas in separate but connected compartments. One study seeded neurons derived from different brain regions into distinct chambers and found that the interconnected co-culture reduced overall firing activity compared to neurons cultured in isolation, and also shifted the proportions of astrocytes and specific neuronal subtypes, more closely resembling what happens in a living brain. The researchers then used the platform to study phencyclidine (PCP), a drug known to induce schizophrenia-like symptoms, and tracked how its effects on one compartment rippled through to connected regions.7PubMed Central. Neurons derived from different brain regions are inherently different in vitro: a novel multiregional brain-on-a-chip
Other designs have pushed this further by differentiating human neural stem cells into dopaminergic and GABAergic neurons simultaneously in different compartments of the same device, generating circuits that use multiple neurotransmitter systems. The neurospheres formed robust connections across all compartments, demonstrating that a single chip can host multineurotransmission circuits modeling the functional connectivity between distinct human brain regions.8ACS Biomaterials Science & Engineering. Brain-on-a-Chip Device for Modeling Multiregional Networks
Building the Neurovascular Unit
The brain’s functional unit is not just neurons. It includes endothelial cells lining blood vessels, pericytes stabilizing those vessels, astrocytes bridging between vessels and neurons, and oligodendrocytes insulating axons. Reconstructing this whole ensemble, the neurovascular unit, on a single chip is a significant step up from modeling any one component. A microfluidic platform that combined human neural stem cells with a functional microvascular barrier achieved over 90% cell viability for both the endothelial and neural-glial populations, and confirmed stable expression of key markers: the endothelial cells expressed the adhesion molecule CD31 at roughly 99%, while the neural stem cells expressed the protein nestin at about 97%.9PubMed Central. A neurovascular unit-on-a-chip: culture and differentiation of human neural stem cells in a three-dimensional microfluidic environment
Vascularization itself presents a chicken-and-egg problem: neurons need blood supply to survive long term, but growing functional blood vessels in a chip is not trivial. One approach used iPSC-derived endothelial cells co-cultured with motor neuron spheroids in a microfluidic device. The vascular networks self-assembled into tubes with an average lumen diameter of about 60 micrometers, and their presence boosted neuronal activity. Calcium oscillations, a proxy for neural firing, increased in both frequency and amplitude when vascular networks were present compared to neuron-only controls.10Nature (Scientific Reports). Engineered 3D vascular and neuronal networks in a microfluidic platform
Glial Cells and Neuroinflammation
For years, neuroscience research focused overwhelmingly on neurons, but glial cells, astrocytes, oligodendrocytes, and microglia, now account for a large share of what researchers want to model on chips. Microglia in particular are the brain’s resident immune cells and play central roles in neuroinflammation, a process linked to Alzheimer’s, Parkinson’s, and multiple sclerosis. A humanized neural tissue-on-chip model differentiating neurons, astrocytes, oligodendrocytes, and microglia all from the same patient’s iPSCs has been developed for studying neuroinflammation and testing therapeutic approaches, including extracellular vesicles derived from mesenchymal stem cells.11PubMed. Differentiation of Neurons, Astrocytes, Oligodendrocytes and Microglia From Human Induced Pluripotent Stem Cells to Form Neural Tissue-On-Chip Because all four cell types come from the same genetic background, the system avoids the immunological mismatches that plague experiments using cells sourced from different donors.
Modeling Neurodegenerative Disease
Disease-specific brain chips are one of the field’s most active frontiers. For Alzheimer’s disease, researchers have built isogenic brain-chip models using cells that carry familial Alzheimer’s mutations alongside genetically corrected control cells from the same donor. In these chips, the brain-channel compartment showed decreased levels of amyloid-beta 42, total tau, and phosphorylated tau 181 in protein lysates, while the vascular channel showed increased levels of total tau and phosphorylated tau, recapitulating the kind of protein trafficking across the blood-brain barrier that happens in actual Alzheimer’s patients.12PubMed Central. Modeling neurovascular dysfunction in Alzheimer’s disease using an isogenic brain-chip model
Parkinson’s disease chips focus on a different set of molecular culprits. Alpha-synuclein aggregates, the hallmark of Parkinson’s pathology, damage mitochondria, causing fragmentation and deficiency in the mitochondrial complex I pathway. Dopaminergic neurons generated from Parkinson’s patients carrying the A53T mutation or triplication of the SNCA gene accumulate alpha-synuclein inclusions and show impaired cellular morphology and membrane potential on chip.13Nature Communications. Neuropathogenesis-on-chips for neurodegenerative diseases These patient-derived models offer something animal models generally cannot: human-specific disease biology in a controlled, observable environment.
Infectious Disease and Cross-Organ Modeling
Brain chips are not limited to neurodegeneration. During the COVID-19 pandemic, researchers connected a lung alveolus chip to a blood-brain barrier chip, creating a linked system that could test how SARS-CoV-2 infection in lung tissue translates into neurological damage. Direct exposure of the BBB chip to the virus caused mild disruption, but perfusing it with medium from the infected lung chip caused more severe injuries: endothelial dysfunction, pericyte detachment, and neuroinflammation. Gene expression analysis showed that the brain endothelium downregulated actin cytoskeleton genes while glial cells upregulated inflammatory pathways.14PubMed Central. Blood–brain barrier injury and neuroinflammation induced by SARS-CoV-2 in a lung–brain microphysiological system This linked-organ approach captures something crucial: the brain damage from COVID-19 appears to be driven largely by circulating inflammatory signals from the infected lungs, not just by the virus directly reaching the brain.
Drug Screening and Toxicity Testing
Pharmaceutical companies spend enormous sums testing drugs that ultimately fail in human trials, often because animal models poorly predict human neurotoxicity. Brain-on-a-chip platforms offer a way to screen compounds against human tissue earlier in the pipeline. A tetra-culture brain chip containing endothelial cells, neuroblastoma cells, microglia, and astrocytes was used to measure the effects of organophosphate compounds, chemicals found in pesticides and nerve agents. The system tested barrier penetration, acetylcholinesterase inhibition, and cell viability for four different organophosphates, and the in vitro toxicity results correlated with available animal data, suggesting the chip could serve as a cost-effective alternative to animal testing for this class of compounds.15Scientific Reports. Three-dimensional (3D) tetra-culture brain on chip platform for organophosphate toxicity screening
An especially promising direction is integrating biosensors directly into the chip for real-time monitoring. Rather than removing samples for offline analysis, embedded detection systems can continuously track biomarkers like neurotransmitter levels, barrier permeability, and inflammatory mediators as a drug flows through the system.16PubMed Central. Biosensors Integration in Blood-Brain Barrier-on-a-Chip: Emerging Platform for Monitoring Neurodegenerative Diseases That kind of continuous readout gives pharmacologists a dynamic picture of how a drug affects the brain over hours or days, not just a single-timepoint snapshot.
Personalized Models From Patient Cells
The convergence of iPSC technology with organ-on-a-chip platforms has created a plausible pathway to personalized neurology. When blood-brain barrier chips were built using iPSCs from patients with neurological diseases, the chips predicted disease-specific deficiencies in transporter proteins and disruptions to barrier integrity that matched known clinical features of those diseases.17PubMed. Human iPSC-Derived Blood-Brain Barrier Chips Enable Disease Modeling and Personalized Medicine Applications In principle, a clinician could take a blood sample from a patient, reprogram the cells into iPSCs, differentiate them into the relevant brain cell types, seed them on a chip, and test which drugs work best for that specific individual’s biology.
This is not yet routine clinical practice. Generating iPSC lines and differentiating them takes weeks, and the per-patient cost remains high. But the concept provides a framework for studying how genetic variation affects drug response in a human-relevant system, an approach sometimes called pharmacogenomics. Because the chips allow environmental factors like toxin exposure to be layered on top of a patient’s genetic background, they offer a platform for studying gene-environment interactions that are nearly impossible to isolate in living patients.18PubMed. iPSC-derived organ-on-a-chip models for personalized human genetics and pharmacogenomics studies
The PDMS Absorption Problem
PDMS is popular for good reasons, but it has a serious flaw: it absorbs small hydrophobic molecules. Many neuropsychiatric drugs are hydrophobic, which means a chip made of PDMS can silently soak up a meaningful fraction of the drug you are trying to test, producing false low-potency readings. Research specifically measuring this absorption effect across 13 neuropsychopharmaceuticals found that sorption increased with compound hydrophobicity for PDMS-based devices, though the effect was less pronounced in alternative materials like polycarbonate.19PubMed Central. Sorption of Neuropsychopharmaca in Microfluidic Materials for In Vitro Studies
This is not a minor technical nuisance. If the platform is supposed to tell you how much of a drug crosses the blood-brain barrier and what it does to neurons on the other side, losing an unknown fraction of that drug into the chip walls undermines the entire measurement. The field is gradually transitioning away from PDMS toward nonabsorptive materials, but this creates trade-offs: some alternative polymers are harder to mold, less transparent, or more expensive. Thermoplastics like cyclic olefin copolymers and hybrid resins are emerging as replacements that preserve the optical clarity and moldability researchers need while sharply reducing molecular absorption.20ACS Biomaterials Science & Engineering. Beyond Polydimethylsiloxane: Alternative Materials for Fabrication of Organ-on-a-Chip Devices and Microphysiological Systems
Mechanical Injury on a Chip
Not all brain damage is molecular. Traumatic brain injury involves physical deformation of tissue, and modeling that deformation reproducibly in the lab is tricky. A microfluidic device designed specifically for TBI research used a three-layer design to compress cortical spheroids at a 25% strain and a strain rate of 3.125 per second, comparable to mild TBI. After 24 hours, the compressed samples showed up to 8% cell death on average, providing a controlled, repeatable injury that can be paired with drug screens or genetic analyses.21MINDS@UW Madison. Traumatic brain injury on-a-chip: a microfluidic device for the compression of cortical spheroids The ability to deliver a precisely calibrated mechanical insult to human neural tissue and then watch the cellular response unfold is something animal models offer only crudely, since you cannot easily control the exact forces applied to a mouse brain.
Regulatory Recognition and Reduced Animal Testing
For decades, U.S. federal law effectively required animal testing before a drug could enter human clinical trials. The FDA Modernization Act 2.0 changed that by amending the 1938 Federal Food, Drug, and Cosmetic Act to formally allow alternatives including cell-based assays, organoids, and organs-on-chips in preclinical development.22PubMed Central. FDA Modernization Act 2.0: transitioning beyond animal models with human cells, organoids, and AI/ML-based approaches The law does not ban animal testing; it removes the legal mandate that made it the only accepted path. For brain-on-a-chip developers, this regulatory shift is a strong tailwind. Pharmaceutical companies now have legal cover to substitute chip-based assays where they can demonstrate equivalence or superiority to animal data, and the economics favor it: chips are cheaper per test than maintaining animal colonies, produce results faster, and generate data from human rather than rodent tissue.
Still, validation remains the bottleneck. Regulators want standardized protocols, reproducibility data, and head-to-head comparisons showing that chip-based results predict human outcomes at least as well as animal experiments. The field is working through this, but the regulatory door is now open in a way it was not five years ago.
3D Bioprinting and the Next Generation of Chips
As fabrication technology advances, 3D bioprinting is beginning to supplement traditional lithography-based methods. Bioprinting allows researchers to deposit cells and supporting hydrogels in precise three-dimensional patterns, creating structures that are harder to achieve with flat channel-based designs. One application used melt-electrowriting to print a nervous system chip with biomimetic microchannels and compartmented chambers that enabled aligned axonal networks and organized spatial arrangement of cell types. When used to study viral spread through neural tissue, the printed chip revealed that Schwann cells participated in axon-to-cell viral transmission but appeared resistant to infection themselves.23Elsevier. Recent advances in 3D-printing-based organ-on-a-chip
Bioprinting also opens the door to architectures that more closely resemble the layered, curved, and folded geometry of actual brain tissue. Cortical layers, for instance, are not flat sheets sitting in channels; they are stacked, curved structures with precise cell-type organization at each depth. Printing these layers with spatial fidelity is still more aspiration than routine, but early demonstrations suggest the approach can produce tissue constructs with a degree of anatomical realism that channel-based chips cannot match. The combination of printed scaffolds with microfluidic perfusion, electrodes for electrical recording, and biosensors for chemical monitoring is converging toward platforms that watch a small piece of human brain doing something genuinely brain-like, one chip at a time.