How 3D neural models, calcium imaging, and electrophysiology reveal functional changes alongside conventional cell-health endpoints.
Beyond survival: Why functional assessment matters
A compound can leave neurons fully viable while altering how they communicate. This distinction makes central nervous system (CNS) safety assessment more demanding than running a simple cytotoxicity assay. Human neural cultures can add mechanistic and functional evidence during preclinical development, provided researchers carefully consider and match each model and readout to the question at hand.
What is CNS safety assessment?
Preclinical CNS safety assessment investigates unintended effects that a candidate compound may have on the brain, spinal cord, or associated neural functions. It evaluates changes in neuronal survival and structure, altered synaptic signaling, abnormal network activity, and other effects that might impact development decisions. Early identification of liabilities helps to frame follow-up experiments to guide candidate prioritization.
Neurotoxicity refers broadly to harmful effects on nervous tissue or its function, while safety pharmacology examines undesirable pharmacodynamic effects on physiological functions. The terms overlap, but a viability readout cannot serve as a surrogate for functional safety.
Structural liabilities encompass cell loss, neurite retraction, or altered synaptic architecture, whereas functional impairments manifest as altered neuronal excitability, firing rates, or network synchrony in fully viable cells. This distinction guides assay choice: morphology and viability answer different questions from electrical or calcium-based recordings.
Why is CNS toxicity difficult to predict preclinically?
Neural activity arises from interactions among excitatory and inhibitory neurons, astrocytes and, in many contexts, other glial and vascular cells. A compound acting on one population can change the behavior of its larger network. For example, reducing inhibitory signaling can alter bursts and synchrony without immediately reducing the number of viable neurons.
Exposure timing dictates phenotypic outcomes. An ion-channel effect may appear quickly, whereas injury involving mitochondrial function, synapses, or cellular stress may evolve over repeated exposure. Species-dependent biology and the developmental state of stem-cell-derived neurons can also complicate interpretation. Human-cell origin addresses one source of uncertainty, but does not reproduce the adult brain, drug distribution, or the full systemic response.
No single model captures every potential liability. A useful assessment combines orthogonal endpoints and verifies whether the biological system engages the mechanism of interest at a measurable level.
Why is cell viability alone insufficient?
Standard ATP quantification, membrane-integrity measurements, and live/dead staining indicate aspects of cellular health. They remain useful for detecting overt cytotoxicity, establishing tolerable concentration ranges. However, a network that stops firing because most cells died poses a different question from one that becomes silent while remaining viable.
Functional disruption can occur without overt cell death. A candidate compound may change spontaneous firing frequency, burst duration, network synchrony, or the balance between excitation and inhibition at a concentration that does not cause a detectable viability loss. In such a case, viability and network assays give complementary evidence rather than competing verdicts.
Neither a single functional signal nor an unchanged viability result establishes clinical risk. Exposure, concentration, reference controls, and concordance across readouts still matter. A defined neuronal viability and apoptosis assay illustrates the value of a cell-health endpoint, whereas functional assays address a separate dimension of the question.
Comparing neural model formats: 2D monolayers vs. 3D systems
Human-derived 3D neural models organize neural cells in a three-dimensional volume rather than on a flat two-dimensional substrate. The format facilitates multicellular contact, synaptic maturation, and long-term functional monitoring.
Nevertheless, traditional two-dimensional cultures maintain an important role in drug discovery, because accessible cells facilitate cell-body and neurite measurements, imaging, and straightforward electrophysiological testing. However, a monolayer cannot replicate tissue-level spatial organization or multicellular architecture. Additionally, exposure to artificial surfaces can alter cellular morphology, function, and network properties.
Neural spheroids and microtissues are three-dimensional aggregates of neural cells. These terms typically denote purpose-built models constructed from defined ratios of neuronal and glial cells with controlled tissue dimensions. Standardized production protocols, precise cell-ration control, and scalable culture platforms render these 3D systems compatible with higher-throughput applications, enabling their integration into early-stage drug discovery.
In contrast, neural organoids generated through self-organized stem-cell differentiation capture developmental patterning and tissue heterogeneity. They are useful for disease research or developmental questions but can demand longer maturation and require careful control of variability.
Defined composition and self-organization are not mutually exclusive, nor does one always outperform the other. The choice turns on the required biological features, reproducibility, throughput, and the strength of the evidence linking a measured endpoint to the intended use.
Which cell populations matter in a cortical neural model?
Glutamatergic excitatory neurons and GABAergic inhibitory neurons shape neural network firing and synchrony, while astrocytes support metabolic homeostasis and influence synaptic environments, including extracellular transmitter handling.
Cell ratios can influence baseline network dynamics and modulate responses to test compounds. For instance, a neuron-rich culture may facilitate electrophysiological or calcium recording, while a culture with more astrocytes may be better suited for studying metabolic support. Researchers should document cell identity, proportions, maturation markers, and functional responses rather than assuming that an aggregate recapitulates human cortical composition.
Microglia, oligodendrocytes, and vascular or blood-brain barrier components become especially relevant to questions involving inflammation, myelination, or compound access. If a mechanism requires a missing population, a negative result has limited meaning.
What is functional neurotoxicity?
Functional neurotoxicity is a harmful, compound-associated change in neural signaling or network behavior, which may arise without immediate cell death. Experimental indicators can include altered excitability, firing rate, burst pattern, synchrony, or excitation–inhibition balance. These observations require interpretation, as a desired pharmacological action can also change network activity. Concentration-response relationships, time course, washout or recovery, and comparison with reference compounds help distinguish mechanisms. Measurements at baseline and after dosing can reveal an acute change or a delayed divergence. Pairing them with viability and morphology helps determine whether the functional result accompanies cellular injury.
How can neuronal network activity be measured in vitro?
Functional neurotoxicity manifests as altered excitability, firing frequency, burst duration, or network synchrony. Two primary technologies capture these electrical events in vitro: Calcium flux imaging and microelectrode array (MEA) electrophysiology.
Calcium Imaging:
Calcium imaging follows changes in intracellular calcium with fluorescent indicators or genetically encoded sensors. In an active culture, it can show the timing and distribution of calcium transients and synchronous oscillations. Plate-compatible systems can support comparisons across doses and experimental conditions. Calcium signals are an indirect proxy for electrical activity and can also reflect non-spiking calcium biology. Various parameters such as dye loading, optical depth, and acquisition rate affect interpretation.
Electrophysiology:
MEA records extracellular electrical signals from cells close enough to couple with its electrodes. Analysis can quantify spikes, bursts, network bursts, and synchrony over repeated sessions. The recording is label-free, but electrode contact, tissue positioning, and the fraction of cells sampled matter, particularly in a 3D aggregate. A lack of detected spikes may reflect poor coupling rather than an inactive tissue.
High-density electrode arrays can sample electrical activity at multiple positions around a neural aggregate. The quality of the signal still depends on tissue placement and coupling, so spatial detail must be interpreted alongside recording geometry.
Calcium imaging offers optical sampling of activity-related intracellular signals. MEA records extracellular electrical events with finer temporal resolution. Their spatial sampling, throughput, and confounders differ. Neither is intrinsically superior for every study. Combined with viability, morphology, and molecular assays, they can help separate an early network perturbation from later cellular damage.
Why does long-term functionality matter?
Acute exposure tests immediate responses. Repeated or prolonged dosing addresses different questions, including whether an early perturbation persists, accumulates, or resolves. Delayed changes may be missed if a culture is observed only once soon after treatment.
Accurate longitudinal interpretation requires a stable baseline, consistent dosing protocols, and proper vehicle controls. Repeated measurements of the same experimental units can track the trajectory, but researchers should control for handling and reporter effects. End-of-study viability and structural measurements help contextualize any functional decline.
How should you choose a 3D neural model for CNS safety assessment?
Choosing a suitable 3D model depends on biological, technical, practical and financial aspects. Ultimately, the specific research question should dictate the platform of choice.
The evaluation typically begins with the cellular composition of the system. A suitable model must contain all the essential cell types at appropriate ratios to model the desired effects, and it needs to remain functionally stable over the target incubation window.
Once the biological baseline is established, one must weigh the physiological complexity of the model against scalability and batch-to-batch reproducibility. Furthermore, the practical utility of a neural in vitro system heavily depends on compatibility with relevant endpoints, such as functional calcium flux analysis and microelectrode array assays alongside biochemical or immunofluorescence-based imaging assays.
Finally, decisions in drug development are inevitably commercial. While the immediate cost per data point is often a primary consideration, making informed choices requires looking beyond the initial purchase price to understand the total cost of ownership of a technology, including the hands-on operator time, validation efforts, required instrumentation, and quality control.
Model development and standardization: When does in-house development make sense?
In-house model development allows end-to-end control over biological parameters, enabling teams to tailor cell sources, ratios, genetic modifications, and maturation timelines to specific mechanistic questions. However, this flexibility requires developing robust differentiation protocols, aggregation workflows, and culture conditions. Establishing this robustly in-house demands substantial hands-on expertise and effort.
Commercially pre-prepared cultures can circumvent much of this upfront work. Providers absorb the burden of batch-to-batch standard operating procedures, raw material qualification, and baseline quality control. They often supply detailed characterization packages covering cell identity, viability and marker expression. Nevertheless, in-study qualification remains important, as production documentation cannot substitute for endpoint-specific controls under the conditions of the actual experiment.
In-house development makes sense when proprietary biological customization is paramount. Commercially sourced models excel when throughput, batch-to-batch reproducibility, shortened assay development timelines, and total cost of ownership take priority.
How InSphero supports CNS safety – 3D InSight™ Human Neural Microtissues
Evaluating CNS safety pharmacology in drug discovery requires robust, human-relevant in vitro systems capable of recapitulating complex network electrophysiology. 3D InSight™ Human Neural Microtissues address this need by offering assay-ready, quality-controlled, and reproducible 3D microtissues engineered for scalable testing.
Assembled with defined, tightly controlled proportions of human iPSC-derived excitatory and inhibitory neurons alongside supporting astrocytes, these microtissues establish functional synaptic communication and spontaneous network firing. The intrinsic electrical activity provides a stable, physiologically relevant baseline for pharmacological studies, allowing researchers to evaluate compound-induced changes in firing frequency, burst duration, and network synchrony. Tracking these network responses across concentration-response curves, acute versus chronic exposures, and recovery phases informs mechanism-of-action studies.
Delivering mature assay-ready microtissues from validated cell lots eliminates in-house differentiation variability and ensures high batch-to-batch consistency. Furthermore, the microtissues are provided in ANSI/SLAS-compatible Akura™ 96-well and 384-well formats designed for robust handling and throughput-compatible automated workflows.
Where can 3D neural models contribute, and what remains uncertain?
With suitable validation, these systems can contribute to exploratory CNS safety screening, functional pharmacology, neurotoxicity investigations, compound prioritization, and mechanism-of-action studies. Disease modeling and target-engagement questions require a model that expresses the relevant disease phenotype or target. Safety pharmacology applications require a suitable functional baseline and pharmacological controls.
However, as with any system, limitations remain. Many systems omit vasculature or some glial populations. IPSC-derived cells may reflect immature developmental states and culture formats may not appropriately model systemic exposure. Researchers can strengthen inference by specifying context of use, reporting negative as well as positive controls, and integrating orthogonal evidence.
Future progress depends less on making every model more elaborate than on defining reproducible endpoints, documenting the exposure conditions under which they work, and testing how well those endpoints inform real development decisions.
References and further reading
- Toxicity assessment using neural organoids: innovative approaches and challenges. Toxicological Research (2025). DOI: 10.1007/s43188-025-00279-y.
- Application of a high-density microelectrode array assay using a 3D human iPSC-derived brain microphysiological system model for in vitro neurotoxicity screening of environmental compounds. Archives of Toxicology (2025). DOI: 10.1007/s00204-025-04043-x.
