Photomotor response data analysis approach to assess chemical neurotoxicity with the zebrafish embryo

Text S1: Chemical stability Chemical stability of azinphos-methyl, propoxur, tricaine and 3,4-dichloroaniline was determined in three independent experiments by measuring UV/VIS – absorption at 231 nm, 270 nm, 312 nm and 297 nm, respectively, using a nanodrop nd-1000 photometer (peQlab Biotechnology GmbH; Erlangen; Germany). For each compound standard curves were prepared and the concentrations of the test compounds were measured after exposures at 26°C for 24 h and 48 h in 24 well plates with one embryo in 2 mL per well and 10 wells per concentration. Within 48 h of exposure, declines of the nominal concentrations were found to be below 1% (azinphos-methyl, tricaine), 5% (propoxur) and 15% (3,4-dichloroaniline). Due to the little changes in concentrations over time biological effects were related to nominal exposure concentrations for these test chemicals. For the lipophilic pyrethroids esfenvalerate and flucythrinate, the exposure concentrations were below the detection limit for UV/VIS analysis. For esfenvalerate, a strong deviation from nominal concentrations and a rapid concentration decline within 24 h exposure has previously been observed in a semistatic exposure setup (Klüver et al., 2015). Due to the similarities in structure and physico-chemical properties to esfenvalerate we assume that the actual concentration of flucythrinate in the wells also strongly deviated from nominal concentrations. However, as we lack analytical data we relate observed biological effects to nominal concentrations of these compounds and need to note that actual concentrations were possibly lower so that therefore calculated effect concentrations may be overestimated for these compounds.


Retrieval of AFT data (LC50)
Lethal effect concentrations for adult or juvenile zebrafish were not available for all selected test compounds. Therefore, given the overall high correlation of fathead minnow with zebrafish AFT data (Belanger et al., 2013) the respective AFT LC50 values for fathead minnow (Pimephales promelas) were retrieved from an US-EPA database (Russom et al., 1997). For azinphos-methyl, propoxur, flucythrinate, tricaine and 3,4-dichloroaniline those values are 0.20 µM, 42.0 µM, 1.0 nM, 302.0 µM a nd 47 µM, respectively (Table 1). For esfenvalerate, the LC50 of the structurally similar fenvalerate was used, which differs from esfenvalerate by stereoisometric composition (3.6 nM). is the effect at concentration ci : is the concentration of substance i : is the effect in controls : is the maximum effect at curve saturation : is the slope of the curve 50 : is the constant describing the concentration of substance causing 50 % of the maximum effect The model was fitted to the concentration depending effect data and the model parameters were estimated using the program JMP (SAS, Marlow, UK) for fish toxicity test data or BDMS (available at https://www.epa.gov/bmds) for PMR data. A 4-parameter logistic regression model was used for analysis of FET test results (min and max were set to 0 % and 100 %), a 4-parameter regression model was used for PMR effects based on the 2D-density approach (min was set to 19 %). Effect concentrations and model parameters are summarized in Table S1.

Residual analysis
In order to evaluate how well the model fits the PMR data, we calculated for each data point (representing a PMR effect at a certain substance concentration) the standardized residuals (eq. 2) and plotted them against the corresponding prediction estimate (eq. 1, see above). The scatter of the residuals is concentration-independent and showed no heteroscedasticity (Fig. S1). Therefore, we concluded that the logistic model explains the PMR effect adequately for all five neuroactive substances (azinphosmethyl, propoxur, esfenvalerate, flucythrinate and tricaine). The PMR effects measured in three independent experiments showed no difference in the standardized residuals, which was a strong indication that the measurements of the PMR effects were reproducible. The residual standard error [RSE; (eq. 4)] for the regression ranged from 6 % for propoxur to 10 % for tricaine, which means that the average response deviated from the regression line by a maximum of 10 % of PMR effect. The standardized residuals were found to be in a range of +/-2 for all compounds and the controls (mean of 0 +/-0.2 and standard deviation of maximal 1.6) (Fig. S1).
is the residual of an effect at a concentration ci is the standardized residual of an effect at a concentration ci : is the residual standard error : is the amount of data points used for model fitting

Text S3: Control variability
The PMR is characterized by an intrinsic variability, which can partially be minimized by controlling certain parameters, such as the time of the day when the analysis is conducted and the temperature. These parameters were controlled in the present study. While the typical response pattern was observed in all replicates, some variability in the magnitude of the motion index could still be observed. Figure S1A shows the raw data time course of the motion index that was observed in the present study. As indicated by the ratio of the motion indices for the excitation versus the pre-pulse phases, the motion index ( Fig. S1B and Tab. S1) can be normalized by the motion index of the refractory phase. The use of surface densitiy areas as performed for the assessment of PMR data in this study is another way to normalize the motion indices, since for each replicate the surface density areas are compared to the respective replicate controls (for details refer to Material and Methods in the main article).

Text S4: PMR effect quantification based on the OA-approach
The PMR effect quantification based on the OA-approach was performed as described in the Material and Methods section in the main article. The distribution of the Motion Index values, detected during 5 -25 s of a PMR measurement were described by fitting a density function to the data. The OA of density curves for control and respective treatment groups is a measure for the similarity of the movement activity. One minus the OA is used as PMR effect parameter based on this so called OA-approach. Concentration-dependent PMR effects were modeled using a 4-parameter logistic regression model (eq. 1, see above). The regression curves (Fig. S2) and residual plots ( Fig. S3) are shown for the five neuroactive compounds azinphosmethyl, propoxur, flucythrinate, esfenvalerate and tricaine and the narcotic reference compound 3,4-dichloroaniline. The resulting parameter estimates and the benchmark concentrations are shown in Table S2.

Text S5: Morphological effects
Morphological effects were analysed by automated image assessment (see main article for details) for the exposure of embryos to two selected concentrations, i.e. the effect concentration of the PMR (either EC50 or benchmark concentration) and the LC50 (or maximum water solubility if no lethality was observed). The compound-specific results of the quantitative assessment of the various morphological features can be found in the supplemental excel file. In order to summarize the data the means and standard deviations between controls and exposed embryos were compared (Table S6). With the exception of small difference in the contour-yolk distance of propoxur and several endpoints for tricaine no morphological effects for embryos exposed to PMR effect concentrations were observed. Slightly more changes in morphology were observed for higher effect concentrations (at the LC50 or maximum water solubility). Hence, except for tricaine potential secondary effects on the PMR caused by morphological changes can be excluded.

A B
Fig. S3: Concentration response curves for PMR effects (•) in zebrafish embryos (30 -35 hpf) exposed to different concentrations of the five neuroactive compounds, acetylcholine esterase inhibitors azinphos-methyl (a) and propoxur (b); voltage gated sodium channel agonists flucythrinate (c) and esfenvalerate (d), and the sodium channel antagonist tricaine. 3,4-Dichloroaniline (f) served as a non-neuroactive reference compound with a narcotic mode of action. Data points (•) represent differences in the overlapping area (OA) of the activity parameter density between control and the respective chemically treated embryos (OA approach). The control variability of the PMR is indicated by open circles (○). A logistic model was fitted to the data and used to calculate EC50 values and benchmark concentrations (BMC). The LC10 for fish embryos at 48 hpf is indicated by a vertical line (─ • ─). In the case that no mortality was observed, the maximum water solubility level is indicated by a vertical line (▬ ▬).

Fig. S4: Residual plots for PMR effects estimated by the overlapping area (OA) approach
Standardized residuals depending on predicted response is shown for the five neuroactive chemicals azinphosmethyl, propoxur, flucythrinate, esfenvalerate and tricaine. Mean, standard deviation of standardized residuals and residual square error for each compound were calculated.

Fig. S5: Lethal effects of the test compounds in the FET test
Effect -% mortality, concentration -µM