Perfluorooctanesulfonic acid (PFOS) antagonizes gamma-aminobutyric acid (GABA) receptors in larval zebrafish and mammalian models

Abstract

Per- and polyfluoroalkyl substances are a class of synthetic chemicals detected ubiquitously in the environment, humans, and wildlife. Perfluorooctanesulfonic acid (PFOS) is one prevalent chemical previously shown to cause adverse effects on nervous system function across in vivo and in vitro models, including dark-phase hyperactivity in larval zebrafish. The objective of this study was to evaluate the role of gamma-aminobutyric acid receptors (GABARs), GABAAR and GABABR, as mediators of dark-phase hyperactivity in PFOS-exposed larval zebrafish. Zebrafish were acutely exposed to 7.87 to 120 μM PFOS, 0.68 to 12.4 μM picrotoxin (GABAAR antagonist), 0.77 to 14.05 μM propofol (GABAAR-positive allosteric modulator), 4.4 to 80 μM saclofen (GABABR antagonist), 0.43 to 7.87 μM CGP13501 (GABABR-positive allosteric modulator), or the solvent control 0.4% dimethyl sulfoxide 60 min before behavior assessment at 5 days post fertilization. Co-exposures to positive allosteric modulators and PFOS were performed. Acute exposure to PFOS caused transient dark-phase hyperactivity. Concentration-dependent dark-phase hypoactivity was observed following acute propofol or CGP13501 exposure, in contrast to the concentration-dependent hyperactivity caused by acute picrotoxin exposure. Saclofen exposure provoked a modest reduction in dark-phase motor activity at the highest concentration tested. PFOS-induced hyperactivity was rescued to baseline activity by co-exposure to propofol or CGP13501. To assess relevance across species, electrophysiological measurements were performed in cultured mouse cortical neurons and BrainSpheres derived from human-induced pluripotent stem cells. PFOS exposure reduced GABAAR-mediated currents in mouse neurons. GABAAR- and GABABR-dependent units in BrainSphere-derived neural networks exhibited increased spiking activity following PFOS exposure. This study demonstrates that PFOS antagonizes GABARs in zebrafish, mouse, and human experimental systems. Taken together, this study supports the concept that early life-stage zebrafish can be used to rapidly identify causative mechanisms, conserved across taxa, by which xenobiotic agents alter neuroactivity.

Keywords: GABA; PFOS; behavior; neurotoxicology; zebrafish.

Fig. 1.

Dark-phase hyperactivity is observed in 5 dpf zebrafish larvae acutely exposed to PFOS. (A) Schematic representation of the exposure paradigm in which 5 dpf larvae were exposed to 7.87 to 120 µM PFOS or 0.4% DMSO as a vehicle control 60 min before behavioral assessment. (B) Locomotor activity over time following exposure to 120 µM PFOS (blue; n = 23) or 0.4% DMSO (grey; n = 45) in the light phase (yellow; 1200 to 2400 s) at 13,238 lux and dark-phase (white; 2400 to 3720 s) at 0 lux of the light–dark transition test. Data are mean distance moved per second (cm/s) ± standard error. (C) Box- and violin-plots of motor activity in each 10 min assay phase (L1, L2, D1, D2) following 7.87 to 120 µM PFOS (n = 21 to 24) exposure or 0.4% DMSO (n = 45). Data are summed distance moved in 2 min intervals (cm/2 min) for each larva. (D) Box- and violin plots of motor activity (cm) in the 3 s following the dark–light (VSR1) or light–dark (VSR2) transition following 7.87 to 120 µM PFOS (n = 21 to 24) exposure or 0.4% DMSO (n = 45). Boxes indicate the median and interquartile range (IQR), whiskers indicate the calculated minimum (25th percentile −1.5 × IQR) and the calculated maximum (75th percentile +1.5 × IQR), and dots indicate the outliers beyond the calculated minima and maxima. Violins describe the kernel probability density of the underlying data. Significance (P < 0.05) is displayed as different letters and was determined by Tukey-adjusted estimated marginal means following either a (C) generalized additive mixed effects model (Fig. S1) or a (D) linear mixed effects model. Summary data are located in Excel Tables S1 to S3 (Owen 2025a). dpf, days post fertilization; L, light; D, dark; DMSO, dimethyl sulfoxide; PFOS, perfluorooctanesulfonic acid; VSR, visual startle response; IQR, interquartile range.

Fig. 2.

The optimal PFOS exposure window to trigger dark-phase hyperactivity in 5 dpf zebrafish is between 30 and 60 min. Box- and violin-plots signifying the distance moved (cm) in 2 min periods across the first 10 min in the dark (D1) and the next 10 min in the dark (D2) at 0 lux for each larva exposed to 120 µM PFOS (blue) or 0.4% DMSO (grey) for (A) 30 min, (B) 60 min, (C), 120 min, or (D) 240 min before locomotor assessment. Replicate numbers range from 45 to 47 larvae per test group. Light phase and visual startle response activity can be found in the supplement (Fig. S2). Boxes indicate the median and interquartile range (IQR), whiskers indicate the calculated minimum (25th percentile −1.5 × IQR) and the calculated maximum (75th percentile +1.5 × IQR), and dots indicate the outliers beyond the calculated minima and maxima. Violins describe the kernel probability density of the underlying data. Significance (P < 0.05) is displayed as different letters and was determined by Tukey-adjusted estimated marginal means following a generalized additive mixed effects model (Fig. S3). Summary data are located in Excel Tables S4 to S13. D, dark; DMSO, dimethyl sulfoxide; PFOS, perfluorooctanesulfonic acid; IQR, interquartile range.

Fig. 3.

Exposure to GABAR modulators picrotoxin, propofol, and CGP13501, but not saclofen, causes changes in swimming behavior at 5 dpf. Representative time series of locomotor activity of larvae exposed to (A) 6.96 µM GABA_AR antagonist picrotoxin (purple; n = 24), (C) 2.46 µM GABA_AR PAM propofol (orange; n = 23), (E) 25.09 µM GABA_BR antagonist saclofen (blue; n = 20), or (G) 2.46 µM GABA_BR PAM CGP13501 (pink; n = 14) compared with 0.4% DMSO control (grey; n = 35-48). Data are mean distance moved per second (cm/s) ± standard error across the light phase (yellow; 1200 to 2400 s) at 13,238 lux and dark phase (white; 2400 to 3720 s) at 0 lux of the light–dark transition test. Box- and violin-plots of motor activity in the D1 and D2 phases following exposure to (B) 0.88 to 12.4 µM picrotoxin, (D) 0.77 to 14.05 µM propofol, (F) 4.4 to 80 µM saclofen, or (H) 0.43 to 7.87 µM CGP13501. Light-phase data can be found in the supplement (Fig. S5). Data are summed distance moved in 2 min intervals (cm/2 min) for each larva. Boxes indicate the median and interquartile range (IQR), whiskers indicate the calculated minimum (25th percentile −1.5 × IQR) and the calculated maximum (75th percentile +1.5 × IQR), and dots indicate the outliers beyond the calculated minima and maxima. Violins describe the kernel probability density of the underlying data. Significance (P < 0.05) is displayed as different letters and was determined by Tukey-adjusted estimated marginal means following a generalized additive mixed effects model (Fig. S6). Summary data are located in Excel Tables S14 to S18. GABAR, GABA receptor; dpf, days post fertilization; D, dark; DMSO, dimethyl sulfoxide; IQR, interquartile range.

Fig. 4.

Co-exposure to the GABA_AR PAM propofol and PFOS blunts dark-phase hyperactivity to control levels. Locomotor activity of 5 dpf larvae exposed to (A) 0.43 µM GABA_AR PAM propofol (orange; n = 68), (B) 120 µM PFOS (blue; n = 72), or (C) propofol and PFOS (lavender-blue; n = 69) compared with 0.4% DMSO control (gray; n = 67). Data are mean distance moved per second (cm/s) ± standard error across the light phase (yellow; 1200 to 2400 s) at 13,238 lux and dark phase (white; 2400 to 3720 s) at 0 lux of the light–dark transition test. Box- and violin-plots of summed distance moved in 2 min intervals (cm/2 min) for each larva in the (D) D1 phase and D2 phase. Light-phase data can be found in the supplement (Fig. S7). Boxes indicate the median and interquartile range (IQR), whiskers indicate the calculated minimum (25th percentile −1.5 × IQR) and the calculated maximum (75th percentile +1.5 × IQR), and dots indicate the outliers beyond the calculated minima and maxima. Violins describe the kernel probability density of the underlying data. Significance (P < 0.05) is displayed as different letters and was determined by Tukey-adjusted estimated marginal means following a generalized additive mixed effects model (Fig. S8). Summary data are located in Excel Tables S19 and S20. GABA_AR, GABA_A receptor; PFOS, perfluorooctanesulfonic acid; PAM, positive allosteric modulator; dpf, days post fertilization; D, dark; DMSO, dimethyl sulfoxide; IQR, interquartile range.

Fig. 5.

Dark-phase hyperactivity is rescued by GABA_BR PAM CGP13501 and PFOS co-exposure. Locomotor activity of 5 dpf larvae exposed to (A) 1.38 µM GABA_BR PAM CGP13501 (pink; n = 65), (B) 120 µM PFOS (blue; n = 67), or (C) the co-exposure of CGP13501 and PFOS (lilac; n = 67) compared with 0.4% DMSO control (grey; n = 70). Data are mean distance moved per second (cm/s) ± standard error across the light phase (yellow; 1200 to 2400 s) at 13,238 lux and dark-phase (white; 2400 to 3720 s) at 0 lux of the light–dark transition test. Box- and violin-plots of summed distance moved in 2 min intervals (cm/2 min) for each larva in the (D) D1 phase and D2 phase. Light-phase data can be found in the supplement (Fig. S7). Boxes indicate the median and interquartile range (IQR), whiskers indicate the calculated minimum (25th percentile −1.5 × IQR) and the calculated maximum (75th percentile +1.5 × IQR), and dots indicate the outliers beyond the calculated minima and maxima. Violins describe the kernel probability density of the underlying data. Significance (P < 0.05) is displayed as different letters and was determined by Tukey-adjusted estimated marginal means following a generalized additive mixed effects model (Fig. S8). Summary data are located in Excel Tables S21 and S22. GABA_BR, GABA_B receptor; PFOS, perfluorooctanesulfonic acid; PAM, positive allosteric modulator; dpf, days post fertilization; D, dark; DMSO, dimethyl sulfoxide; IQR, interquartile range.

Fig. 6.

GABA_AR-mediated currents are reduced in mouse cortical neurons exposed to PFOS. (A) Infrared differential interference contrast image of a patch-clamped cultured mouse cortical neuron. (B) Example experiment showing a representative pharmacologically isolated GABA_AR-mediated response to 50 µM GABA that was inhibited by subsequent co-exposure of 120 µM PFOS. Extracellular solution was supplemented by 10 µM NBQX, 50 µM APV, and 3 µM CGP55845 to block AMPAR, NMDAR, and GABA_BR, respectively. (C) Representative GABA currents in a control cell measured in control recording solution and a cell pre-treated and continuously superfused with PFOS containing solution. (D) Individual and mean GABA current (IGABA) in control cells (control, mean ± SE, n = 7) and cells pre-treated and during PFOS application (+PFOS, mean ± SE, n = 6). ***P = 0.001, Mann–Whitney U-test. Summary data are located in Excel Table S23. PFOS, perfluorooctanesulfonic acid; IGABA, GABA_A receptor-mediated current.

Fig. 7.

PFOS exposure increases spiking of GABA_AR- and GABA_BR-dependent units in human BrainSphere-derived neural networks. (A) Schematic representation of the exposure paradigm for GABA_A and GABA_B unit identification and PFOS treatment using microelectrode array (MEA) recordings. After baseline recording 1, the neurotransmitter GABA followed by the respective antagonist bicuculline or saclofen were applied for unit identification. Normalized spike count (y-axis) of BrainSpheres exposed to increasing concentrations of PFOS (x-axis) for (B) GABA_A and (C) GABA_B units. The dotted horizontal line represents the control median. Boxes indicate the median and interquartile range (IQR), X indicates the mean, whiskers indicate the calculated minimum (25th percentile −1.5 × IQR) and the calculated maximum (75th percentile +1.5 × IQR), and dots indicate the outliers beyond the calculated minima and maxima. Violins describe the kernel probability density of the underlying data. Values above violins represent adjusted P-values obtained from a Friedman test followed by a Conover post hoc test. P-values are Benjamini–Hochberg adjusted. Data inclusion criteria are visualized in the supplement (Fig. S9) (Owen 2025c). Summary data are located in Excel Table S24. PFOS, perfluorooctanesulfonic acid.

Fig. 8.

Possible binding of PFOS within the human α₁β₂γ₁ GABA_AR pore, as predicted by HADDOCK. The structure of GABA_AR is based on PDB ID: 8SG0, modified to meet HADDOCK input requirements. (A) Cartoon representation of the highest ranked structure within the best-ranked cluster with α-subunits in yellow, β-subunits in blue, and the γ-subunit in green. The inset displays PFOS as stick representation in its predicted binding position. (B) Energy analysis of the 91 structures within the best ranked cluster. The binding interface of PFOS within the GABA_AR pore has an extensive buried surface area (BSA) and is mainly stabilized by van der Waals (vdW) forces, whereas electrostatics (ES) have a minor contribution. Boxes indicate the median and interquartile range (IQR), whiskers indicate the calculated minimum (25th percentile −1.5 × IQR) and the calculated maximum (75th percentile +1.5 × IQR), and dots indicate the outliers beyond the calculated minima and maxima. PFOS, perfluorooctanesulfonic acid; vdW, van der Waals; ES, electrostatics; BSA, buried surface area.

Fig. 9.

Discovery of mechanisms underlying sulfonic acid PFAS-induced behavioral phenotypes. This study aimed to build on previous work revealing a shared hyperactivity phenotype for sulfonic acid PFAS^A (Gaballah et al. 2020). Visual startle response (VSR) hyperactivity was found to persist to 8 dpf and require ppard function^B (Gutsfeld et al. 2024). Transient dark-phase hyperactivity was hypothesized to be mediated by a chemical-receptor interaction (Gutsfeld et al. 2024), leading to the examination of PFOS antagonism of the GABARs in larval zebrafish and mammalian models in the current study^C (Owen 2025b). PFAS, per- and polyfluoroalkyl substances; PFOS, perfluorooctanesulfonic acid; PFHxS, perfluorohexanesulfonic acid; VSR, visual startle response; dpf, days post fertilization.