Quadrimodal in vivo assay for detection of adverse neuro-behavioral effects: validation using dextromethorphan

Drug discovery notoriously suffers from high attrition rates during clinical trials. A major contributing factor to this issue is the frequent failure to identify adverse neural functional effects of compounds in the CNS at earlier stages of drug development than currently possible. To address this unmet need, the (side-)effects of the widely used antitussive dextromethorphan were assessed employing a neuro-behavioral assay originally developed for evaluation of anesthetics. This high-content in vivo assay provides quadrimodal readouts, reflecting behavioral changes of a weakly electric fish model induced by the compound to be tested. Two of these readouts are proxies of the neural activity of the pacemaker nucleus, an endogenous oscillator that drives the periodic electric organ discharges. Administration of dextromethorphan induced a decrease in the frequency of these discharges, an increase in the generation of amplitude-frequency modulations of this electric behavior, and the production of an abnormal motor behavior, consisting of full-transverse-plane rotations of the fish. The rate of occurrence of each of these behaviors was dose dependent. The changes in discharge frequency and rotation of the fish’s body persisted for several hours and in some individuals even for at least 2 days. We hypothesize that the drug-induced changes in the two electric behaviors are mediated by inhibition of N -methyl-D-aspartate (NMDA) receptors and by activation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-receptors in the pacemaker nucleus. The results of this study serve as proof of concept for the applicability of the neuro-behavioral assay for the evaluation of (side-) effects of drug candidates on neural function in the CNS.

Drug discovery is a complex, lengthy, and costly process. On average, it takes more than 10 years to develop an innovative drug. Most of this time—about 8 years—is spent on conducting clinical trials (Brown et al., 2022; Eastern Research Group, 2024). As a conservative estimate, the out-of-pocket cost per drug is approximately $173 million, but approximately $879 million after accounting for the costs of failures and capital (Eastern Research Group, 2024). Around two-thirds of the overall drug development expenditures incur during the three clinical phases.

The primary goal of drug development is the identification and optimization of a compound that interferes with a ‘druggable’ target causally linked to the disease or condition for which treatment is sought with high selectivity and efficacy. Nevertheless, comparable efforts are undertaken to minimize adverse effects of the drug candidate. Although Target Safety Assessment as an integral part of the research program during drug discovery allows investigators to make Go/No-Go decisions before expensive preclinical and clinical testing, most unwanted side effects are discovered only at later stages of drug development. Between 2014 and 2023, the likelihood of approval by the Food and Drug Administration (FDA) of a drug candidate that entered Phase I of the clinical stage was just 6.7% (Citeline, 2024). The biggest hurdle is Phase II, with just 28% of the drug candidates tested in this phase advancing to Phase III.

A major contributing factor to the late (and thus extremely costly) failures of drug candidates is neural toxicity in the central nervous system (CNS). A comprehensive longitudinal review of AstraZeneca’s small-molecule drug projects from 2005 to 2010 has shown that this issue accounts for nearly one-quarter of failures across the entire drug-discovery process (Cook et al., 2014), yet only a minor fraction of these failures becomes evident during Good Laboratory Practice toxicity testing in the preclinical phase (Roberts et al., 2014).

Most of the adverse neural effects cause disturbances in CNS function (such as headache, nausea, dizziness, anxiety, depression, seizure/convulsion, or suicidal ideation). These effects are far more difficult to predict by the currently used animal models than the less frequent structural adverse effects, which are identifiable as CNS tissue damage in histological material (Walker et al., 2018).

To address the unmet need for animal models that can predict potential adverse effects of drug candidates on CNS function earlier than is currently possible, our laboratory has developed a novel high-content in vivo assay (Eske et al., 2023; Lehotzky et al., 2023; Ahmed and Zupanc, 2025; Zupanc and Ahmed, 2025; Zupanc and Hong, 2026). This so-called Neuro-Behavioral Assay (NBA) is grounded on the weakly electric brown ghost knifefish as an animal model. This teleost generates a continuous train of extremely regular discharges of low amplitude (<1 V) and high frequency (up to 1 kHz) with its electric organ (Zupanc and Bullock, 2005). The timing of each electric organ discharge (EOD) is controlled in a one-to-one fashion by command spikes originating from an endogenous oscillator in the medulla oblongata: the pacemaker nucleus. Thus, the frequency of the EOD is identical to the frequency of the neural oscillations of the pacemaker nucleus (Meyer, 1984), enabling investigators to use the behavioral output (the EOD) as a proxy of the neural activity of this brainstem structure. This unique feature provides an enormous advantage of the NBA because the electric behavior can be readily recorded non-invasively by external electrodes in the tank in which the fish is kept (Figure 1).

Initially, we showed that anesthetic drugs are powerful modulators of the EOD frequency (Eske et al., 2023; Lehotzky et al., 2023; Ahmed and Zupanc, 2025; Zupanc and Ahmed, 2025), indicating interference with vital body functions controlled by brainstem nuclei. Since anesthetics can also modulate thalamic input to the pacemaker nucleus, causing transient amplitude-frequency modulations of the EOD (‘chirps’), extraction of these signals from the recorded EOD enables investigators in addition to studying the effect of anesthetic compounds on higher brain function. Moreover, since movements of the fish relative to the stationary recording electrodes are reflected by alterations in the amplitude of the recorded dipole-like electric field of the fish, it is possible to extract information about relative changes in locomotor activity from EOD recordings. This third readout of the NBA has made it possible to quantify the reduction in locomotor activity induced by anesthetics and the recovery of locomotor activity after anesthesia, and to correlate these changes with those observed in the same assay run in EOD frequency and chirp rate (Ahmed and Zupanc, 2025).

While the NBA is a powerful tool for studying the neural effects of anesthetics in the CNS, its applicability is not restricted to these agents. Recent work from our lab has shown that the assay can detect a robust, reversible, dose-dependent neural effect of the antihistamine diphenhydramine (best known under its brand name Benadryl), reflected by a significant drop in the EOD-frequency readout (Zupanc and Hong, 2026). Whereas this finding was an important proof of concept, it has remained elusive whether non-anesthetic drugs induce changes only in the EOD frequency readout of the NBA, and whether different drugs produce different ‘signatures’ of the assay readouts. These differences may be evident not only by the type of behavioral output affected, but also by the amount of drug-induced change relative to baseline levels.

To address these questions, we tested in the present study the effect of dextromethorphan (DXM) on neural functions in the CNS, assayed by the NBA. This drug is the most frequently used over-the-counter antitussive, found in cold and flu medications both as the only active ingredient (such as Delsym 12-h Cough, Robitussin Cough Long Acting®, Triaminic Long-Acting Cough®, and Vick DayQuil Cough®) and as one of several active ingredients in numerous combination products. DXM is also frequently abused as an illicit drug, often in combination with other drugs. At therapeutic doses, it may cause several neural side effects, such as dizziness, lightheadedness, drowsiness, nervousness, restlessness, and nausea. Symptoms of overdose include changes in vision, difficulty breathing, hallucinating, seizures, and coma (HFS Patient Medication Information, 2026). These well-documented adverse effects made DXM well suited for testing our hypothesis that these side effects are reflected by modulation of the behavioral outputs of the NBA, and that the type of output affected and the amount of change induced produce a distinct readout signature. The results of our investigation not only confirmed this notion but also established a novel drug-induced neuro-behavioral effect, in the following referred to as Full Transverse Plane Rotation (FTPR). Evaluation of changes in this behavioral readout elevates the previous trimodal version of the NBA to an even more powerful quadrimodal version.

Animals

A total of 8 adult brown ghost electric knifefish (Apteronotus macrostomus) of the Apteronotus leptorhynchus species group were used in this study. They originated from their natural habitat in Colombia and were imported by a tropical fish dealer (AliKhan Tropical Fish, Richmond Hill, NY, United States).

The quantitative data presented in this study were obtained from 6 of these fish. The body weight of these individuals ranged from 4.0 g to 7.7 g (mean: 6.1 g), and the total length ranged from 119 mm to 160 mm (mean: 144 mm). Mean baseline EOD frequencies, obtained by averaging the 2 mean baseline frequencies measured in Section 3.1.1 (see below), extended from 706 Hz to 854 Hz. This range indicates that the sample included both females and males, as females tend to discharge at lower frequencies than males, with little overlap between the sexes (Zupanc et al., 2014). One fish (MK-14), which died of causes not related to the assay runs near the end of the project duration, was sexed through gonadal inspection. It was a female with white eggs that constituted 1.3% of the body weight. Both features indicate a low degree of sexual maturity of this individual (brown ghost electric knifefish are seasonal breeders that undergo cyclic changes in sexual maturity). Consistent with its identification as a female, its mean baseline EOD frequency was very low: 706 Hz.

Several weeks before running an assay, the fish were transferred from a community tank to an isolation tank (50 cm × 30 cm × 30 cm), in the following referred to as ‘home tank.’ The latter was equipped with an aquarium thermostat heater and an air-driven corner filter. The fish was kept under a 12:12-h light-dark photoperiod at a conductivity of approximately 150–200 μS/cm, pH values of 7.7–7.9, and water temperatures of 26 °C–28 °C.

Generation of readouts of the neuro-behavioral assay

The neuro-behavioral effects of DXM were assessed quantitatively through the 4 readouts of the NBA—EOD frequency, chirp rate, LAI, and FTPR. The first 3 readouts were extracted from the recorded EOD, whereas the fourth readout was obtained from the video-recorded motor behavior of the fish.

For the recording of the EOD, the home tank of each fish was equipped with an opaque cylindrical PVC tube (length: 190 mm; inner diameter: 38 mm; outer diameter: 42 mm), which the fish, as nocturnal animals, utilize as daytime shelter (Figure 1). While running the assay, the open ends of the tube were closed with a coarse plastic mesh netting to prevent the fish from leaving the tube, thereby allowing for uninterrupted high-quality recordings. Differential recording of the fish’s EOD was accomplished through a pair of stainless-steel electrodes built into the tube. A strip of stainless steel placed near the tube served as a ground electrode. The recorded EOD signal was AC amplified (gain: 30×; low-pass filter: none; high-pass filter: 200 Hz) by a CED 1902 amplifier (Cambridge Electronic Design, Cambridge, England) and then digitized at a sampling rate of 50 kHz using a CED Micro 1401 mkII analog-to-digital converter (Cambridge Electronic Design), a Lenovo ThinkCentre V50s SFF desktop computer (equipped with an Intel Octa Core i7-10,700, 32 GB RAM, 512 GB NVMe +1 TB HDD), and the software program Spike 2 versions 5.21 and 11.05 (Cambridge Electronic Design).

The EOD frequency was calculated using the automated procedure by Lehotzky et al. (2023). Since the EOD frequency is temperature dependent, all frequency measurements were adjusted to a reference temperature of 26.0 °C, employing a Q 10 of 1.56 (Zupanc et al., 2003). Water temperature was measured every minute with a calibrated digital thermometer (Fisher Brand, Model 15–077-8, 11.705,843; Thermo Fisher Scientific, Waltham, Massachusetts, United States; accuracy ±0.05 °C). Chirps were detected using the automated approach of Eske et al. (2023). LAI was determined by automated analysis of amplitude modulations of the recorded EOD (Ahmed and Zupanc, 2025).

For video recording of the fish’s motor activity (cf. Supplementary Material), a Sony HDR-CX405 HD Video Recording Handycam Camcorder (Sony Corp.) mounted on a tripod was used. The rate of FTPRs was determined by manual analysis of playbacks of the recorded videos.

Experimental design

To assess the effect of DXM, each run of the NBA consisted of three parts (Figure 2). To establish baseline activities, the EOD and the motor behavior of the fish were recorded in its home tank for 30 min (‘pre-treatment period’). Then, the shelter tube with the fish was transferred to a separate, smaller aquarium (‘experimental tank;’ 30 cm × 20 cm × 20 cm), where the fish was exposed for 15 min to a specific concentration of DXM (DXM Pharmaceutical) dissolved in water from the fish’s home tank (‘test experiments’) or to home-tank water only (‘control experiments’). The handling of the shelter tube with the fish was identical under both conditions, and the temperatures of the water of the home tank and the experimental tank were very similar (±0.5 °C). After 15 min in the experimental tank, the shelter tube with the fish was returned to the home tank, where the EOD and video recordings continued for 285 min (‘post-treatment period’).

Data analysis

The variability of sampled EOD frequency measurements was characterized by computing the coefficient of variation (CV) (Equation 1). The CV is defined as

Data analysis was carried out using Excel for Microsoft 365 MSO (version 2508) (Microsoft). Statistical analysis of data was performed employing IBM SPSS Statistics (version 29.0.1.1 (244)) (IBM Corporation).

Effect of 0.002% DXM

In the first set of experiments, we examined the effects of a 15-min immersion of the fish into a 0.002% w/v solution of DXM in water from its home tank. The effects were assessed by comparing the quantitative measures of the 4 readouts of the NBA (EOD frequency, chirp rate, LAI, and FTPR) during the 285-min post-treatment period with the baseline measures established during the 30-min pre-treatment period. As a control, the fish was immersed in water from its home tank only, while keeping all other details of the experimental procedure identical to test runs of the NBA.

EOD frequency

The EOD frequency is distinguished by its low variability over time at a constant temperature. To quantify the degree of variability in the 6 fish examined, we analyzed the 360 baseline frequency measurements (collected in 5-s segments) in each of these fish during the 30-min pre-treatment period of both the control experiment and the test experiment. In this sample, the CV ranged from 0.074% to 0.303% (mean: 0.152%; median: 0.129%). This result underscores the extremely low background (baseline EOD frequency), enabling detection of even small signals (frequency changes) in the frequency readout of the NBA.

Exposure of the fish to 0.002% DXM induced a decrease in EOD frequency during the post-treatment period compared to the baseline in each of the 6 fish (Figures 3A,C). These decreases ranged from −1.3 Hz to −7.1 Hz (mean: −3.9 Hz; median: −4.2 Hz). By contrast, a 15-min exposure of the same fish to water from the respective home tank only, without addition of the drug, resulted in both EOD frequency decreases (3 fish) and EOD frequency increases (3 fish), ranging from −1.3 Hz to +3.3 Hz (mean: +0.4 Hz; median: +0.2 Hz) (Figures 3B,C). Compared to the frequency changes observed under control conditions, the frequency decreases induced in the 6 fish under test conditions were significantly different at p < 0.05 (Wilcoxon signed rank test, 2-tailed).

Chirp rate

Immersion of the fish into the 0.002% solution of DXM in home-tank water for 15 min revealed in each of the 6 individuals an increase in the rate of chirping during the post-treatment period compared to baseline. However, this increase was marginal in 3 fish (MK-07, MK-13, MK-14) and within the range of the spontaneous chirp rates observed in the present investigation during the baseline recordings and in previous studies (Engler et al., 2000; Zupanc et al., 2001; Eske et al., 2023). We therefore excluded these fish from further analysis.

In the other 3 fish (MK-11, MK-12, MK-15), the chirp rate was markedly (3–18 times) higher during the post-treatment period compared to baseline (Figure 4, left column). In control experiments, decreases in chirp rate were observed in 2 of these fish (MK-11, MK-15). In the other fish (MK-12), very minor increases in chirp rate occurred during the post-treatment period relative to baseline, under control conditions (Figure 4, right column).

LAI

Comparison of the LAIs extracted from the EOD recordings during the post-treatment period with the LAIs computed from the EOD recordings during the pre-treatment period did not indicate any significant difference between test experiments and control experiments in the 6 fish (p > 0.05, Wilcoxon signed rank test, 2-tailed). Congruent with this finding, the means and medians of the relative changes (mean LAI during the post-treatment period divided by mean LAI during the pre-treatment period) under these 2 conditions were similar (means 0.873 versus 0.869; medians: 0.853 versus 0.808).

FTPR and other motor behavioral patterns

Whereas no statistically significant differences in LAI between test and control conditions could be found, changes in several other motor patterns were induced by DXM. They are shown in the video available as Supplementary Material S1. The most common one was quivering of the anal and pectoral fins, which occurred as early as 2–6 min after the start of the fish’s immersion into the 0.002% w/v DXM solution. Additional alterations in motor behaviors, typically commencing a few minutes later, comprised of jerking motions and adoption of a belly-up posture. The most conspicuous motor activity evoked by administration of DXM was rotation in the transverse plane along the fish’s longitudinal axis. Such behavior occurred at various degrees and including 360° rotations, which we refer to as FTPRs. The rate of the latter was determined manually by counting the number of FTPRs over 5-min segments. The investigator who performed the FTPR scoring was blinded to the treatment condition.

During the 30-min baseline recording in the fish’s home tank, very few or no FTPRs were observed (Figure 5, left and right columns). However, immersion of the fish into 0.002% w/v DXM in the experimental tank induced manifold increases in the number of FTPRs displayed, with rates as high as 165 per 5 min (Figure 5, left column). Most of this increased motor activity began toward the end of the 15-min exposure to DXM or shortly after the fish’s return to its home tank. Notably, in each of the 6 fish examined, the FTPRs occurred in bouts during the post-treatment period.

Comparison of the FTPR rates sampled during the 57 5-min time segments of the 285-min post-treatment period with the rates sampled during the 6 5-min time segments of the baseline period showed a highly significant difference in 5 of the 6 fish (p < 0.01, Mann-Whitney U Test, 2- tailed). In the sixth fish (MK-13), the FTPR rate also increased, but this increase was concentrated in 2 bouts. The first bout, with FTPR rates as high as 118/5 min, commenced within minutes after the fish was returned from the experimental tank to its home tank. The other bout, with a peak FTPR rate of 11/5min, was evident at about 3.5 h after the fish’s return to its home tank. By contrast, the FTPR rates during the long interval between these two bouts and during the remaining recording after the second bout were rather similar to the baseline rates, thus having a ‘dilutive’ effect on the FTPR rate during the 285-min post-treatment period. As a consequence, the difference in the FTPR rates between the pre-treatment period and the post-treatment period was not statistically significant (p > 0.05, Mann-Whitney U Test, 2-tailed).

By contrast, in control experiments (in which the fish were immersed in the experimental tank to water from the home tank only), the median FTPR rates of the baseline period and the post-treatment period were 0 per 5 min in each of the 6 fish examined (Figure 5, right column). No significant difference of the FTPR rates between the 2 time periods was evident in any of these 6 fish (p > 0.05, Mann-Whitney U Test, 2-tailed).

Dose-related effects of DXM

To examine whether the above-described effects of DXM on EOD frequency, chirp rate and FTPR rate were dose related, we ran the NBA at 2 additional concentrations (0.00067% and 0.00133%) in each of the 6 fish. We then combined the results of these assay runs with those obtained through treatment of the fish with 0% (=control; reflecting baseline activity) and 0.002% of the drug (see Section 3.1). Whereas the EOD frequency and FTPR readouts were analyzed in each of the 6 fish, chirp rate was analyzed only in the 3 fish (MK-11, MK-12, MK-15) that had responded to the DXM treatment at 0.002% with a significant increase in this behavioral output. Like in the experiments in Section 3.1, each assay run consisted of a 30-min pre-treatment period, a 15-min exposure to the drug (or water from the fish’s home tank only), and a 285-min post-treatment period.

EOD frequency

To examine the dose-response relationship between DXM and the EOD frequency, we first determined the mean EOD frequencies during the 30-min pre-treatment period and the 285-min post-treatment period at each concentration. Then, we subtracted the latter from the former. At 0%, 0.00067%, and 0.00133% of DXM, the responses of the fish were variable, including both frequency increases and frequency decreases. As reported in Section 3.1.1, a 15-min immersion of the fish into a 0.002% DXM solution induced in each individuum a decrease in EOD frequency. Spearman correlation indicated that there was a significant moderate negative relationship between the concentration of DXM and the EOD frequency (r S = 0.506, p < 0.05, 2-tailed; n = 1 NBA run at each of the 4 concentrations in each of the 6 fish examined) (Figure 6).

Chirp rate

To examine the dose-response relationship between DXM and the generation of chirps, we assessed the ‘net’ change in this behavior induced by the drug at each concentration. The net chirp rate was computed by subtracting the mean chirp rate during the pre-treatment period from the mean chirp rate during the post-treatment period. At concentrations of 0%, 0.00067%, and 0.00133% of DXM, a 15-min exposure of the fish to the drug resulted in both decreases and increases in chirp rate. At the highest concentration tested — 0.002% — the drug induced in each of the 3 fish a rise in chirp rate. This increase was 3–18 times higher during the post-treatment period, compared to the pre-treatment period. Spearman correlation indicated that there was a significant strong positive relationship between the concentration of DXM and the net chirp rate (r S = 0.734, p < 0.01, 2-tailed; n = 1 NBA run at each of the 4 concentrations in each of the 3 fish examined) (Figure 7).

FTPR rate

To examine the dose-response relationship between DXM and the generation of FTPRs, we assessed the ‘net’ change in this behavior induced by the drug at each concentration, obtained by subtracting the mean FTPR rate during the pre-treatment period from the mean FTPR rate during the post-treatment period. As shown in Figure 8, the net FTPR rate increased with increasing concentration of DXM. This dose-dependency is mirrored by a strong positive correlation of the DXM concentration and the net mean FTPR rate (r S = 0.786, p < 0.001, 2-tailed; n = 1 NBA run at each of the 4 concentrations in each of the 6 fish examined).

Long-term effects of DXM

For assessing possible long-term effects of DXM, the following experiment was run over 3 consecutive days. On Day 1, DXM was administered through a 15-min immersion of the fish in a 0.002% solution, applying the standard protocol described in Section 2.3. On Day 2 and Day 3, the EOD and the motor behavior of the fish were recorded in its home tank at the same time as the experiment on Day 1 had been carried out (10:00–15:30 local time) to avoid possible effects of circadian rhythms on the rate of the behavioral patterns used as readouts of the NBA. In contrast to Day 1, on Days 2 and 3 the fish remained in its home tank throughout the 330 min recording session.

After analysis of the chirp data, we decided to exclude this behavioral output from further consideration. Two of the fish (MK-11 and MK-15) exhibited extremely high spontaneous rates of chirping during the pre-treatment periods on Day 1 (28/h and 48/h), which matched (in MK-15) or exceeded (in MK-11) the mean chirp rates during the corresponding post-treatment sessions. While the rate of chirping is typically very low, with average rates of 1–5 per hour, in very rare instances it can be extremely high, as also noted previously (Engler et al., 2000). In the latter study, one fish produced 268 chirps within a 1-h recording period. The reasons for these episodes of extremely high rates of spontaneous chirping are unknown.

EOD frequency

On Day 1, in each of the 4 fish examined, the mean EOD frequency was lower during the 285-min post-treatment period than the mean EOD frequency during the 30-min pre-treatment period (Figure 9). This result replicates the findings obtained in earlier experiments in the same 4 fish (see Section 3.1.1). On both Day 2 and Day 3, the mean EOD frequency determined over 330 min remained lower than the mean pre-treatment frequency on Day 1 in 2 fish, whereas it was higher in the other 2 fish.

FTPR rate

On Day 1, in each of the 4 fish examined, the mean FTPR rate (per 5 min) was higher during the 285-min post-treatment period than the mean FTPR rate (per 5 min) during the 30-min pre-treatment period (Figure 10), thus replicating the findings obtained in earlier experiments in the same 4 fish (see Section 3.1.4). On Day 2, in each of the 4 fish, the mean FTPR rate (per 5 min) during the 330 min recording had decreased, compared to the mean FTPR rate (per 5 min) during the respective 30-min pre-treatment period of Day 1 and was approximating this baseline activity in 3 fish (MK-12, MK-13, MK-15). However, in 1 fish (MK-11), it was still markedly high. On Day 3, the mean FTPR rate (per 5 min) during the 330 min recording was virtually undistinguishable from the respective baseline activity in Fish MK-12, MK-13, and MK-15 but was still clearly elevated in Fish MK-11, suggesting that the latter fish still had not fully recovered from the modulatory effect of DXM on the generation of FTPRs.

Distinct signature of NBA readouts produced by DXM

In this study, we addressed two major questions. First, can neuro-behavioral (side-)effects of DXM be detected in one or several readouts of the NBA by changes of the respective behavior, relative to baseline activity? Second, does the combination of these readouts create a distinct signature of the neuro-behavioral effects of DXM, which differentiates it from the signatures of other drugs?

Our assay runs produced an affirmative answer to each of these two questions. They showed that DXM induces dose-dependent changes in 3 readouts of the NBA: (i) consistent decrease in EOD frequency across all fish examined, comprising a few hertz, (ii) pronounced increase in chirp rate in a select number of individual fish, (iii) generation of FTPRs, which were rare in non-treated fish, but occurred at very high rates, often in form of bouts, and for prolonged periods of time after the 15-min immersion of the fish in a solution of the drug. No statistically significant changes were found in the LAI between fish subjected to test conditions versus control conditions.

The increase in chirp rate after administration of DXM only in about half but not all the fish examined, is reminiscent of findings in previous investigations testing the effect of specific compounds on NBA readouts. Anesthesia induced by tricaine methanesulfonate or eugenol resulted in a similar bimodal distribution among fish in regard of chirp rates—they were markedly elevated in about half of the fish but hardly so in the other half (Eske et al., 2023; Lehotzky et al., 2023; Ahmed and Zupanc, 2025). Remarkably, anesthesia induced by urethane triggered increases in chirp rate in every individual fish, and these increases were much larger than those induced (in select individuals) by tricaine methanesulfonate or eugenol (Eske et al., 2023; Zupanc and Ahmed, 2025). The reasons for the difference in the magnitude of the effect of tricaine methanesulfonate, eugenol, and DXM on the one side and urethane on the other are unknown. The inter-individual differences in the effect of the former 3 compounds are perhaps related to genomic/phenotypic diversity of the fish, as we use wildtype animals in the NBA. The importance of this aspect for a better understanding of inter-individual differences in the effect of drugs is increasingly recognized in drug discovery (Harrill et al., 2009; French et al., 2015; Gatti et al., 2018; Dumont et al., 2024).

FTPRs were absent when we had tested the anesthetics tricaine methanesulfonate, urethane, or eugenol, and the anti-histamine diphenhydramine (Eske et al., 2023; Lehotzky et al., 2023; Ahmed and Zupanc, 2025; Zupanc and Ahmed, 2025; Zupanc and Hong, 2026). However, the occurrence of this behavior in the present study after treatment with DXM establishes a new (fourth) readout of the NBA. Although the FTPR information is the only of the 4 readouts that is not extracted from the fish’s EOD, it is recorded on video during the same assay run when EOD frequency, chirp rate, and LAI information are recorded via the fish’s EOD. The time course of the FTPRs can, thus, be compared directly with the latter 3 readouts. Since FTPRs can be readily quantified, this readout adds an important measure of an ‘abnormal’ behavior to the portfolio of the other neuro-behavioral patterns measured in the NBA.

The second major question addressed in the present investigation was whether the changes in the NBA readouts induced by DXM are distinct from the changes observed in response to treatment with different drugs. Although only 4 other drugs have been studied thus far, and therefore caution must be exercised regarding this limited sample, the previously tested compounds produced a signature notably different from the signature obtained after DXM administration in the present investigation. Each of the anesthetic drugs triggered a dramatic reduction in relative locomotor activity as well as pronounced decreases in EOD frequency exceeding 100 Hz and in some cases reaching nearly 300 Hz at the most effective concentration of the respective anesthetic agent (Eske et al., 2023; Lehotzky et al., 2023; Ahmed and Zupanc, 2025; Zupanc and Ahmed, 2025). Diphenhydramine induced a rather modest reduction in EOD frequency ranging from around 40–100 Hz in individual fish, yet it did not cause changes in chirp rate or LAI (Zupanc and Hong, 2026). Neither the 3 anesthetic drugs nor diphenhydramine resulted in the generation of FTPRs. Moreover, a long-term persistence of a drug effect, particularly on the production of FTPRs, was observed only after administration of DXM. After treatment with each of the other 4 drugs, recovery of the altered neuro-behavioral output occurred within a few hours, whereas recovery from DXM may take as long as several days.

Based on the current investigation and our previous studies employing the NBA, we hypothesize that other neuroactive drugs are likely to induce significant changes in one or several of the 4 readouts of the NBA and produce distinct signatures of individual drugs in terms of type, magnitude, and time course of the changes in the neuro-behavioral outputs of this assay.

Comparison of the DXM concentrations used in the assay runs with the concentrations found in human patients

DXM was approved under the brand name Romilar as an antitussive by the FDA in 1958. It is contained as the active ingredient for the treatment of cough in numerous OTC medicines (Dicpinigaitis et al., 2014). Due to the long history and the wide use of DXM, its adverse effects are well documented. The severity of these effects depends not only on the dosage of the drug but also on an individual’s capability to metabolize the drug. After administration, DXM is rapidly metabolized into the pharmacologically active dextrorphan by CYP2D6, a polymorphic enzyme of the cytochrome P450 superfamily. In a study examining healthy male volunteers in the United States, it was found that 84% of the subjects were fast metabolizers of DXM, whereas 9% of the sample were slow metabolizers (Woodworth et al., 1987). The half-life of DXM was 2 h in fast metabolizers, but up to 19 h in the slow metabolizers (Capon et al., 1996). This puts the latter individuals at increased risk of developing adverse reactions to DXM. It is possible that the inter-individual differences in the neuro-behavioral effects revealed by the NBA are caused by genetic variations of the fish homologue of CYP2D6.

In humans, a blood-plasma concentration of 0.0001–0.04 mg/L is considered to be within the therapeutic (‘normal’) range, while concentrations of 0.1–2.8 mg/L and 1.1–3 (−20) mg/L are regarded as toxic and comatose/fatal, respectively (Schulz et al., 2020). Notably, the concentration of DXM is approximately 68 times higher in the brain than it is in the serum (Steinberg et al., 1996), likely due to the lipophilicity of the drug.

How do these concentrations compare with those used for testing in the present study? As a standard concentration of DXM in most of the experiments, we used 0.002%, i.e., 20 mg/L. In the dose-response experiments, we tested two additional concentrations, 0.0013% (=13 mg/L) and 0.00067% (=6.7 mg/L). All doses were applied by a 15-min immersion of the fish into a solution of the drug at the respective concentration. Since we observed the first changes in EOD frequency, chirp rate, and rate of FTPRs within a few minutes, it is evident that DXM had entered the body, including the brain, rapidly. However, neither the blood-serum concentration nor the concentration of the drug in the brain are known. Measurements of these concentrations are challenging due to the small amount of blood and brain tissue available, even if the fish would be killed for maximizing the amount of the samples. For example, the brain weight of a fish of the mean size used in the present study (total body mass 6.1 g; total length: 144 mm) is just around 70 mg (cf. Figure 13 in Zupanc and Horschke, 1995). To be able to determine the actual concentration of DXM and its metabolites, it will be desirable to develop protocols for quantitative analysis of the small amounts of these molecules in the fish body.

Towards target deconvolution

As a phenotypic in-vivo assay, the NBA enables investigators to evaluate the integrated neuro-behavioral response of a model organism to a compound. Since the neural mechanisms controlling two of the behavioral outputs assessed in the assay—EOD frequency and chirp generation—are fairly well understood, opportunities exist for downstream target deconvolution and, thus, drug optimization.

The final motor output of these two electric behaviors is controlled by the pacemaker nucleus. In fish of the size used in the present study, this midline structure in the medulla oblongata consists of roughly 200,000 cells (Sîrbulescu et al., 2014). Yet, only about 110 cells form its oscillatory neural network—approximately 90 pacemaker cells and 20 relay cells (Dye and Heiligenberg, 1987; Sîrbulescu et al., 2014; Zupanc et al., 2014; Ilieş and Zupanc, 2023). The pacemaker cells are confined to the pacemaker nucleus and connected via chemical and electrotonic synapses to each other and to the relay cells (Elekes and Szabo, 1985; Dye and Heiligenberg, 1987). The relay cells convey the output of the pacemaker nucleus to the electromotor neurons—modified spinal motoneurons—in the spinal cord. Their hypertrophied axonal terminals called electrocytes form the fish’s electric organ (de Oliveira-Castro, 1955; Bennett, 1971; Waxman et al., 1972). The synchronized depolarizations of the individual electrocytes produce the discharges of the electric organ. Each EOD is driven in a one-to-one fashion by a volley of command spikes from the pacemaker nucleus (Meyer, 1984) — a feature that enables investigators to use the EOD frequency as a proxy of the frequency of the neural oscillations of the pacemaker nucleus.

The frequency of the oscillations of the pacemaker nucleus (and thus of the EOD) is determined by several factors, including intrinsic properties of the pacemaker and relay cells (such as size and ion channel density) and the potassium concentration in the extracellular space (Zupanc et al., 2014; Hartman et al., 2021; Ilieş and Zupanc, 2023; Kawasaki and Zupanc, 2025). Transient modulations of the EOD are controlled by input received from two upstream nuclei (Kawasaki et al., 1988; Keller et al., 1991; Zupanc and Heiligenberg, 1992; Heiligenberg et al., 1996): frequency changes of up to approximately ±20 Hz by the sublemniscal prepacemaker nucleus (SPPn) in the mesencephalon, and chirp-like frequency-amplitude modulations by a cluster of neurons within the central posterior/prepacemaker nucleus in the diencephalon (CP/PPn), hence referred to as CP/PPn-C. These two types of modulations are mediated by different glutamate receptor types on the relay cells (Dye et al., 1989; Heiligenberg et al., 1996): transient frequency changes by input from the SPPn via N -methyl-D-aspartate (NMDA)-receptors, and chirp-like modulations by activation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-receptors through input from the CP/PPn-C.

We speculate that the decrease in EOD frequency and the increase in chirp rate induced by DXM involve blockade of NMDA receptors and activation of AMPA receptors. DXM has been shown to interact with a variety of neurotransmitter receptors and ion channels, of which blocking of NMDA glutamate receptors is believed to be the chief action (Ferkany et al., 1988; Netzer et al., 1993). For guiding the design of future experiments, we propose the following hypothesis: Binding of DXM to the NMDA receptors at the postsynaptic sites where the relay cells receive input from the SPPn reduces sodium current, leading to a lowering of the excitatory drive of the relay cells, which in turn results in a decrease in the oscillation frequency of the pacemaker nucleus. As a secondary effect, the ability of DXM to inhibit voltage-gated sodium channels may also contribute to the lowering of the oscillation frequency.

At the same time, AMPA-receptor-dependent signaling at the relay cells may be promoted by the glutamate surge produced by the blocking of the NMDA receptors. According to our working hypothesis, the stimulation of the AMPA receptors causes depolarization and thereby generation of frequency-amplitude modulations, evident as chirp-like modulations of the EOD. Similar stimulation of AMPA receptors by a surge of glutamate induced by blocking of NMDA receptors is thought to mediate the antidepressant action of DXM—a discovery that led to the repurposing of the drug for treatment of depression (for reviews, see Duman et al., 2016; Kowalczyk et al., 2026).

In addition to the changes in EOD frequency and chirp rate, DXM induced a marked increase in FTPR rate in each of the fish examined. However, unlike the above discussed potential mechanisms underlying the observed changes in the two electric behaviors, the neural substrate mediating FTPRs is unknown. Despite its nature as a motor behavior, it is unclear whether DXM affects primarily neural pathways controlling the motor output of this behavioral pattern, or whether the primary target of the drug are sensory systems, thereby leading to impairment of sensory guided behaviors. In humans, abuse of DXM by taking doses that exceed severalfold those recommended for treatment of acute cough produce both perceptual changes (such as visual hallucination) and motor impairment (such as ataxia). Given that DXM can induce abnormal behavior in both humans and fish, it would be interesting to further examine how the production of FTPRs relates to the symptoms of abuse of this drug in humans.

The NBA as a high-content assay for evaluation of neuro-behavioral effects of drugs

Presence of statistically significant differences between control and test conditions in EOD frequency and chirp rate demonstrates that a compound tested in the NBA is able to pass the blood-brain barrier. This interpretation is founded on the fact that each of the two behavioral outputs is exclusively controlled by the dynamics of the pacemaker nucleus and the brain sites that provide input to this nucleus, but not by any peripheral system.

It is likely that in most cases an increase in FTPR rate or a reduction in the LAI are also controlled by central neural pathways, but since the mechanisms underlying the changes in these behaviors are not known, one needs to consider peripheral actions as possible primary cause. For example, a compound may exert a paralytic effect through interaction with the cholinergic system at the neuromuscular junction, thereby resulting in a decrease in LAI. Still, even in such a case one would need to examine possible interference with cholinergic systems within the central systems.

Will a compound that passes the blood-brain barrier in brown ghost knifefish also pass it in mammalian species, including humans? Like in mammals, the blood-brain barrier of modern ray-finned bony fish (a taxon that includes knifefish and zebrafish), is composed of brain microvascular endothelial cells forming intercellular tight junctions (Cserr and Bundgaard, 1984; Bundgaard and Abbott, 2008). This structural similarity of the blood-brain barrier extends to the molecular level. For example, Claudin-5a is a zebrafish ortholog of the mammalian Claudin-5. Both proteins are highly expressed in endothelial cells of brain capillaries, and both play a critical role in the functional formation of the blood-brain barrier (Li et al., 2022). Thus, we hypothesize that any compound found in the NBA to induce changes in EOD frequency and/or chirp rate can be predicted to pass the blood-brain barrier of humans with reasonable likelihood. Such predictive power of the NBA, if validated, would be highly valuable. First, it would allow investigators to identify potential neural adverse effects of a drug candidate designed to target peripheral organs. Second, it could be used to assess whether a drug candidate targeted at sites within the CNS is capable of passing the blood-brain barrier. If necessary, the assay could also be used to optimize the permeability of this compound by molecular modification. Currently available technologies address these tasks with rather limited success—despite substantial efforts undertaken to develop better in silico, in vitro, and in vivo methods to predict the permeability of molecules in the human brain (for review, see Gupta et al., 2024).

One of the strengths of the NBA is its ability to assess differential effects on different neural systems. In the current investigation, the changes in the readouts induced by DXM indicated that at least 3 different systems are affected by the drug. We hypothesize that such diversity of changes in NBA readouts is indicative of a broad spectrum of (side-)effects of a drug in humans. DXM exhibits such a broad range of adverse effects (HFS Patient Medication Information: American Society of Health-System Pharmacists, 2026). They include dizziness, lightheadedness, drowsiness, nervousness, restlessness, nausea, and vomiting when the drug is taken at therapeutic doses. Symptoms of overdose range from milder effects, such as nausea, vomiting, drowsiness, dizziness, and unsteadiness to more serious effects, such as changes in vision, difficulty breathing, fast heartbeat, visual and auditory hallucinations, seizures, and coma.

Another strength of the NBA is that it evaluates the neuro-behavioral effects of a compound on a continuous scale, whereas many in vivo assays employed in drug discovery, particularly those used for high-throughput screening, assess binary behavioral endpoints, such as presence versus absence of a specific reflex in response to drug administration. This feature of the NBA allows investigators, among others, to quantitatively analyze recovery of the animal from the effect of a drug. In the present study, we showed that a 15-min exposure of the fish to a 0.002% concentration of DXM induces changes in 3 of the 4 readouts of the NBA. Each of these effects lasted for at least several hours and in some individuals even for more than 2 days. At first sight, such long-term neuro-behavioral effects of DXM are surprising, as information available on numerous websites universally states that intoxication effects of overdoses of DXM in humans last for 3–6 h. However, in (what is to our knowledge) the only study that examined quantitatively and under controlled conditions the time course of several behavioral, physiological, and cognitive effects at supratherapeutic doses of DXM in human volunteers, several measures indicated persistent impairment of psychomotor performance (Reissig et al., 2012). For example, 360 min after administration of DXM, the human probands performed almost as poorly in the Balance task (involving balancing on one foot with eyes closed) at high doses as they did after 120 min and 240 min, and significantly different from how they performed after receiving a placebo.

By contrast, diphenhydramine, an antihistamine with adverse effects commonly assumed to a few hours, exhibits relatively fast recovery when tested in the NBA (Zupanc and Hong, 2026). Although the magnitude of the behavioral change (decrease in EOD frequency) induced by this OTC drug is, at its most effective concentration, about one order of magnitude greater than the frequency reduction occurring in response to administration of DXM at its most effective concentration, recovery of 90% of the difference between the minimum EOD frequency and the mean baseline EOD frequency is achieved within 2–4 h. Similar differences in recovery time were found among anesthetic drugs, and these differences reflected the differences in recovery time when these anesthetic agents are used for inducing general anesthesia in human and veterinary medicine (Eske et al., 2023; Lehotzky et al., 2023; Ahmed and Zupanc, 2025; Zupanc and Ahmed, 2025).

Based on the findings of the present study and our previous investigations, we hypothesize that the NBA is well suited to make predictions about the time course of recovery from neuro-behavioral (side-)effects of a drug. Such an application of the assay might be particularly powerful when evaluating potential differences in recovery time of structural or compositional variants of a molecule.