Research Letter
Ontogeny of schooling behavior and stage-specific vulnerability to valproic acid in Xenopus laevis larvae
2026 Volume 51 Issue 9 Pages 463-469
Details
2026 Volume 51 Issue 9 Pages 463-469
Perinatal exposure to chemicals such as valproic acid (VPA) increases the risk of neurodevelopmental disorders, including autism spectrum disorder (ASD). Because maternal drug metabolism complicates the assessment of fetal impacts in mammalian models, Xenopus laevis tadpoles have emerged as an effective alternative for neurotoxicology. Tadpoles exhibit “schooling,” a polarized social aggregation behavior. Previous studies, which exposed tadpoles to VPA continuously from stage NF42 to NF49, suggested that VPA-induced social behavior abnormalities are triggered during this period. In this study, we re-evaluated the ontogeny of schooling behavior and its vulnerability to VPA by administering short-term exposures at specific developmental stages. Using computational modeling and statistical analyses, we compared the spatial distribution of live tadpoles with computer-simulated random models. Untreated tadpoles exhibited significant, non-random social aggregation as early as stage NF42, indicating that the sensorimotor circuitry for schooling is established earlier than previously assumed. Furthermore, tadpoles exposed to VPA for 4 days starting at stages NF42 and NF48 maintained schooling behavior, whereas VPA exposure starting at stage NF52 significantly disrupted spatial cohesion. These findings indicate that while social behavior is acquired prior to NF42, NF52 represents a critical window of heightened pharmacological vulnerability to teratogens, providing new insights into the etiology of environmentally induced neurodevelopmental disorders.
Perinatal and early developmental stages are characterized by rapid and highly organized neurodevelopmental events, including neurogenesis, synaptic pruning, and the dynamic establishment of functional neural circuits (Ta et al., 2022). Exposure to environmental factors and drugs during these developmental windows can alter brain wiring and induce long-term behavioral abnormalities, elevating the risk of neurodevelopmental disorders such as ASD (Roullet et al., 2013). Epidemiological and translational studies demonstrate that prenatal exposure to VPA, a broad-spectrum antiepileptic drug functioning as a histone deacetylase (HDAC) inhibitor and g-aminobutyric acid (GABA) transaminase inhibitor, significantly increases the incidence of these disorders (Bromley et al., 2013; James et al., 2015).
In mammalian models, assessing direct fetal neurotoxicity is limited by confounding variables such as maternal drug metabolism and the blood-placental barrier (Kotta-Loizou et al., 2024). Consequently, Xenopus laevis has become a valuable model organism in developmental neurotoxicology (Carotenuto et al., 2023; Exner and Willsey, 2021). Because Xenopus larvae develop externally in an aqueous environment, precise, stage-specific pharmacological interventions are possible (Buchholz, 2015). Xenopus tadpoles exhibit schooling, a consistent and polarized social aggregation behavior (Katz et al., 1981; Lopez et al., 2021). Schooling requires complex sensorimotor integration, relying on the maturation of the optic tectum for visual collision avoidance and the lateral line system for hydrodynamic mechanosensory feedback (James et al., 2015).
Previous research indicated that tadpoles exposed to VPA continuously from NF42 to NF49, compared to unexposed controls, exhibited impaired social behavior, suggesting that VPA-induced behavioral abnormalities are triggered between NF42 and NF49 (James et al., 2015). However, such prolonged exposure paradigms, spanning approximately 10 days from stage NF42 to NF49, may confound the timing of neural circuit emergence with the exact timing of the window of pharmacological vulnerability. Inhibition of a behavior by a drug does not necessarily imply that the underlying neural circuit is formed during that specific exposure period. Instead, it may indicate that an already existing circuit has entered a highly sensitive state due to epigenetic modifications, structural reorganization, or specific neurotransmitter receptor transitions (Ta et al., 2022). The developmental progression of Xenopus from NF52 through metamorphosis depends on thyroid hormone signaling and large-scale synaptic reorganization, closely resembling perinatal brain development in humans (Buchholz, 2015; Sachs and Buchholz, 2017). Thus, identifying stage-specific vulnerabilities in Xenopus offers relevant translational insights into human neurodevelopment. Furthermore, establishing a simple screening system that yields results in a shorter timeframe is of significant importance, as it enables the rapid assessment of the effects of various chemical substances on the behavior of organisms.
In this study, we re-evaluate the acquisition of schooling behavior in Xenopus laevis and its inhibition by VPA. By utilizing computational modeling for spatial analysis of nearest neighbors via Delaunay triangulation and a robust statistical framework including multiple-group comparisons, we separate the ontogeny of social behavior from its pharmacological vulnerability (Lopez et al., 2021). By comparing live tadpole distributions with computer-simulated random models, we define the precise developmental onset of schooling and demonstrate that VPA sensitivity represents a later, stage-specific critical window of vulnerability rather than the origin of sociality.
Wild-type Xenopus laevis embryos were obtained via standard in vitro fertilization following human chorionic gonadotropin (hCG) injection into adult frogs (Ishihara et al., 2011). To eliminate chemical variables and environmental confounders, embryos and larvae were reared in a strictly controlled aqueous environment consisting of aerated, dechlorinated tap water (standardized rearing water). Animals were maintained at 24°C under a 12-hr light/dark cycle. Larvae were fed a commercial tropical fish diet (Sera Micron) every other day. Developmental staging followed the Nieuwkoop and Faber (NF) normal table (Nieuwkoop and Faber, 1994). Prior to pharmacological exposure and behavioral assays, tadpoles were acclimated to the experimental conditions in fresh dechlorinated tap water for 24 hr.
Valproic acid exposureTadpoles at developmental stages NF42, 48, and 52 were exposed to VPA at a final concentration of 1 mM. Three 1-L glass beakers were prepared for each group, containing 500 mL of rearing water either without VPA (control group) or with VPA (treatment group); ten tadpoles were placed in each beaker for a four-day exposure period (n=30 per group). No food was provided during the treatment period. Water was not changed to avoid stressing the tadpoles. Twenty tadpoles were randomly selected from the 30 in each group for use in behavioral experiments. Before conducting behavioral assays, we confirmed that 0.1-1 mM VPA did not affect mortality during the rearing period from NF42 to NF49.
Behavioral assay apparatusBehavioral assays were conducted with minor modifications to the methods described previously (James et al., 2015). Schooling assays were performed in a 17-cm diameter circular glass arena (Fig. 1A). To prevent artifacts caused by experimenter presence, visual disturbances, or environmental noise, the arena was placed inside an opaque, soundproof enclosure. Uniform illumination was provided from below using a flat LED light pad. The behavioral stimulus was a brief mechanical vibration generated by a standard dental vibrator positioned under the arena. This vibration induced temporary scattering of the tadpoles, followed by the re-establishment of their spatial distribution.
Automated experimental setup and timeline for behavioral assays. (A) Apparatus for evaluating schooling behavior. A GoPro camera for image acquisition and a vortexer (vibrator) to stimulate movement were enclosed in a soundproof box. Operations were coordinated by a Python script running on a Raspberry Pi accessed via SSH. (B) Experimental schedule. The protocol consisted of 2.5-min cycles over 60 min. A vibration stimulus induced swimming, and an image was captured once the tadpoles ceased movement to analyze their spatial configuration.
The entire apparatus was controlled by an Internet of Things (IoT) framework using a Raspberry Pi accessible remotely via Secure Shell (SSH). A custom Python script controlled both the IoT relay powering the vortexer and a mounted GoPro camera (Fig. 1A). The protocol consisted of delivering the vibration stimulus, waiting 2.5 min for swimming to cease and spatial structures to form, and then capturing a high-resolution image. This cycle was repeated over a 60-min trial, obtaining 13 independent spatial configuration images per cohort (Fig. 1B).
Computational tracking and spatial analysisRaw images were processed using ImageJ to extract the 2D Cartesian coordinates (X, Y) of the head and gut of each tadpole in every frame. The analysis of spatial relationships among tadpoles utilized a custom Python algorithm performing Delaunay triangulation (Lopez et al., 2021). Delaunay triangulation provides a mathematically robust method for objectively defining a network of nearest neighbors on a 2D plane without requiring arbitrary distance thresholds, thereby effectively excluding distant individuals that do not influence local schooling dynamics. From this triangulated mesh, the script calculated the “Nearest Neighbor Distance (cm),” representing the absolute spatial distance between adjacent individuals connected by the Delaunay graph, and the “Inter-individual Angle (degrees),” indicating the relative alignment of head-to-tail swimming vectors between nearest neighbors, ranging from 0° (perfectly parallel) to 180° (anti-parallel) (Katz et al., 1981).
Simulated random spatial modelsTo conclusively demonstrate the presence of non-random schooling behavior, empirical biological data must be tested against a null hypothesis of a purely random distribution. A Python-based simulation was developed to generate 8 independent datasets (R1 to R8). Each dataset simulated the random placement of coordinate pairs within a virtual 17-cm circular boundary, matching the sample sizes and physical constraints of the biological assays. The generation of random tadpole data in this simulation combined random sampling from a Continuous Uniform Distribution with relative position calculations using Polar Coordinates.
StatisticsTreating all inter-tadpole distances calculated from multiple frames as independent variables causes pseudoreplication. To avoid this, data from three independent biological replicates were pooled and aggregated before analysis. For statistical testing, a Student’s t-test was used for comparisons between two groups (e.g., control vs. VPA-treated groups at each stage). For comparisons among multiple groups (e.g., across different developmental stages or random simulation models), an analysis of variance (ANOVA) was performed. When a significant main effect was observed, Welch’s t-test followed by the Benjamini-Hochberg procedure was used for post-hoc multiple comparison corrections. Statistical significance was defined as p < 0.05.
Before evaluating live Xenopus behavior, the simulated null models (R1–R8) were analyzed to establish a robust baseline. As shown in Fig. 2A, B, and C, the probability density functions of nearest-neighbor distances for all 8 random simulations exhibited a broad, flat distribution, with the majority of distances ranging from 2.5 cm to 7.5 cm. The density peaks were low (maximum probability density ~0.16), indicating a sparse, uncoordinated scattering of points across the arena typical of a lack of social cohesion. ANOVA among the R1–R8 groups revealed no statistically significant differences, confirming that the simulations provide a stable and reproducible baseline representing random spatial noise.
Spatial analysis of simulated random models and the ontogeny of schooling behavior. (A-C) Simulated random spatial models (R1–R8). (A) Comparison of inter-tadpole distances across the 8 random simulations. (B, C) Probability density distributions of nearest-neighbor distances for the random models, showing broad, flat distributions characteristic of random scatter. (D-E) Ontogeny of schooling behavior in untreated tadpoles (NF42-NF50) compared to the R1 baseline. Differences between groups were evaluated using ANOVA followed by Welch's t-test and the Benjamini-Hochberg procedure for post-hoc corrections. (D) Comparison of inter-tadpole distances across developmental stages. (E) Probability density distributions of nearest-neighbor distances. A significant leftward shift indicating spatial cohesion is evident from stage NF42. (F) Probability density of inter-individual swimming angles, showing progressive alignment at later stages. Different characters indicate a statistically significant difference.
To determine the developmental onset of schooling, the spatial distributions of untreated tadpoles at consecutive developmental stages (NF42, NF45, NF47, NF48, NF49, and NF50) were plotted against a representative random model (R1) (Fig. 2D, E).
At the early stage of NF42, the probability density distribution of nearest-neighbor distances demonstrated a significant leftward shift compared to the R1 baseline. The density peak at NF42 sharpened considerably, localizing around 2.5 cm to 5.0 cm, with a probability density reaching 0.175. This indicates that NF42 tadpoles actively form aggregations after agitation rather than settling randomly. ANOVA and post-hoc tests corrected for multiple comparisons confirmed that the NF42 distribution differed significantly from R1 (p < 0.001), providing evidence that the neural circuits governing basic social aggregation are functional by this early developmental stage.
As development progressed, the degree of spatial cohesion strengthened, with inter-individual distances decreasing progressively up to NF48. This suggests ongoing refinement and maturation of the sensorimotor integration required for complex schooling behavior. Alignment angles also showed progressive structuring (Fig. 2F); compared to the flat, uniform angle distribution inherent to the R1 random model, later stages exhibited a higher probability of parallel and near-parallel orientations.
Stage-specific inhibition by valproic acidHaving established that the basic circuitry for schooling is acquired by NF42 and refined up to NF50, the effect of a 4-day VPA exposure on this behavior was evaluated starting at three distinct developmental checkpoints: NF42, NF48, and NF52 (Fig. 3).
Stage-specific effects of valproic acid (VPA) on schooling behavior. Tadpoles were exposed to VPA for 4 days starting at developmental stages NF42, NF48, and NF52. Comparisons between the control and treated groups at each developmental stage were evaluated using t-tests. (A) Tadpoles exposed starting at NF42 showed no significant difference in spatial distribution between the control and VPA-treated groups (n.s.). (B) Tadpoles exposed starting at NF48 similarly exhibited no significant disruption of schooling behavior (n.s.). (C) Only when exposed starting at NF52 did VPA induce a loss of schooling behavior, resulting in significantly increased inter-tadpole distances compared to the cohesive control group (***, p<0.001).
First, regarding the impact of a 4-day VPA exposure starting at NF42 (Fig. 3A), tadpoles exhibited active swimming and basic group formation similar to the control group. The mean nearest-neighbor distance was approximately 4.0 cm for the control group and approximately 4.0 cm for the VPA-treated group. A t-test showed no significant difference in the distribution between the two groups (n.s.), indicating that the initial formation of the sensorimotor schooling circuit is unaffected by VPA.
Next, for the 4-day VPA exposure starting at NF48 (Fig. 3B), both cohorts maintained cohesive schooling. The mean distance was approximately 4.0 cm for the control group and approximately 4.0 cm for the VPA-treated group. Similar to the earlier stage, no significant disruption of spatial configuration was detected (n.s.).
However, a distinct divergence was observed when the 4-day VPA exposure started at NF52 (Fig. 3C). Control tadpoles maintained cohesive, clustered schools (mean distance ~3.5 cm). In contrast, tadpoles exposed to VPA from this stage failed to aggregate, showing a strong tendency for independent, isolated swimming. Their spatial distribution widened considerably, resulting in a mean distance of approximately 6.0 cm, which closely resembled the uniform scattering seen in the simulated random models. A t-test confirmed that the difference between the control and VPA distributions at NF52 was highly significant (p < 0.001), demonstrating a loss of social behavior induced specifically by exposure at this developmental stage.
The computational spatial analysis and random simulation modeling presented in this study prompt a reevaluation of the ontogeny of social behavior in developing amphibians. Previous studies, which examined social behavior by comparing unexposed groups with tadpoles exposed to VPA continuously from NF42 to NF49, suggested that VPA-induced social behavior abnormalities are triggered between NF42 and NF49 (James et al., 2015). However, our data, based on defined developmental stages and statistical comparisons against computer-generated random models, update this interpretation. The current data clearly show that Xenopus laevis tadpoles possess the capacity for non-random social aggregation as early as NF42. This implies that the basic neural architecture necessary for this behavior (likely networks integrating lateral line mechanoreception with early visual processing in the optic tectum) is already established and functionally active (Katz et al., 1981).
According to James et al. (2015), VPA exposure during the period spanning stages NF42 to NF49 has widespread effects on neural circuit formation. However, in the present study, no behavioral effects were observed following VPA exposure at stage NF42. This suggests that the critical period of susceptibility to VPA’s effects on neurodevelopment may lie in a treatment window beginning more than four days after stage NF42. Given that changes in gene expression are thought to precede phenotypic changes, future research should investigate the impact of VPA treatment on the expression of genes involved in neurodevelopment. The finding that tadpoles undergoing a 4-day acute exposure starting at NF42 and NF48 display consistent schooling behavior and remain resistant to the neurotoxic effects of VPA separates the acquisition of the behavior from the onset of its pharmacological vulnerability. VPA acts as a HDAC inhibitor and enhances GABAergic neurotransmission (Ghodke-Puranik et al., 2013; Kim et al., 2019). In the developing Xenopus brain, populations of GABAergic interneurons in the optic tectum—a major integrative hub for multisensory processing and collision avoidance—undergo spatial and functional reorganization between NF46 and NF48 (Wullimann et al., 2005). Concurrently, during this specific developmental window, global histone acetylation patterns (e.g., H3K9) shift dynamically, driving changes in gene expression essential for synaptic maturation (Ta et al., 2022).
Therefore, the emergence of VPA sensitivity due to exposure starting at NF52 likely reflects a critical developmental transition within the underlying circuitry, rather than its de novo formation. While early schooling circuits (at stages NF42 and NF48) may rely on simpler, deterministic sensorimotor loops, the circuitry at NF52 appears to transition into a more complex network dependent on mature GABAergic inhibition and a precisely regulated epigenetic environment (Kim et al., 2019). Exposure to VPA at the NF52 stage disrupts this epigenetically driven maturation phase. As demonstrated in previous studies, such inhibition leads to hyperconnected, hyperexcitable neural networks lacking the appropriate signal-to-noise ratio required to compute the spatial algorithms necessary for social cohesion, ultimately resulting in behavioral collapse (James et al., 2015).
Furthermore, stage NF52 in Xenopus laevis represents a period of extensive, thyroid hormone-dependent brain remodeling that shares cellular and molecular homologies with human perinatal brain development (spanning late gestation to early postnatal life) (Buchholz, 2015). Because Xenopus develops externally, targeting this specific window (NF48–NF52) allows researchers to bypass the confounding variables of the mammalian placental barrier and directly observe how chemical exposure alters the final wiring of social networks during a phase analogous to the human third trimester. This provides the model with high translational value for exploring the etiology of neurodevelopmental disorders such as ASD.
In conclusion, by implementing an analytical pipeline featuring ANOVA and post-hoc tests against simulated random arrays, this study mapped the ontogeny of social behavior in Xenopus. It establishes that while the behavioral circuitry is acquired prior to NF42, a critical window of vulnerability to epigenetic teratogens emerges at stage NF52. Future investigations utilizing this vulnerability window at NF52, combined with targeted transcriptomics and electrophysiological analyses, are expected to identify the specific HDAC targets and synaptic alterations uncoupled by VPA exposure, thereby elucidating the mechanistic origins of chemically induced neurodevelopmental disorders.
We thank Editage (https://www.editage.jp/) for their thorough and critical reading and revision of the manuscript.
FundingThis work was supported in part by the JSPS KAKENHI (grant number: 22K06312).
Conflict of interestThe authors have no competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Data availabilityThe data in this study are included in the article/supplementary materials. Contact the corresponding author(s) directly to request the underlying data.
Author contributionsMomoka Yamashita: Investigation. Hayase Horie: Investigation. Akinori Ishihara: Conceptualization, Validation, Writing – original draft, Visualization, Project administration.
Ethical approval and consent to participateAll breeding and experimental procedures were approved by the Shizuoka University Animal Experiment Committee (permits #2019F-10 and #2020F-11) under the International Guidelines on the Welfare and Management of Animals (Ministry of the Environment).
Patient consent for publicationNot applicable.
