Rapid determination of chemical losses in a microplate bioassay using fluorescence spectroscopy

Abstract

The increase in production and innovation of chemicals that humans interface with has enhanced the need for rapid toxicity testing of new and existing chemicals. This need, along with efforts to reduce animal testing, has led to the development of high-throughput bioassays typically conducted in microplates. These bioassays offer time and resource advantages over traditional animal models; however, significant chemical losses can occur in microplates. Current methods for measuring chemical losses in microplates require extensive sample preparation and highly sensitive instruments. We propose the use of fluorescence spectroscopy to measure chemical losses in high-throughput bioassays as a low resource alternative to the existing methods. A fluorescent plate reader was used to develop methods for quantifying the aqueous concentrations of two chemicals, 2-hydroxynaphthalene and acridine, in microwells of a 96-well microplate. A high-throughput, 5 day embryonic zebrafish bioassay was used as the model bioassay for method development. Chemical losses were attributed to a combination of photodegradation, sorption, and uptake by the zebrafish embryos, kinetics of which were derived from a pseudo-first order model. Chemical uptake amount was calculated to be approximately 50% and 21% of the total chemical amount added for 2-hydroxynaphthalene and acridine, respectively. Unexpected cranial deformities were observed in the embryonic zebrafish, suggesting further investigation of potential additive toxicity of the ultraviolet radiation exposure from fluorescence measurements and chemical exposure is needed. Nonetheless, this novel method provides a rapid, low resource approach to measuring chemical losses in microplates that can be extended to a variety of autofluorescent chemicals and microplate-based bioassays.

Graphical Abstract

A novel fluorescencespectroscopy method that can be used as a low resource, rapid, and green approach for measuring chemical losses in polystyrene microplates commonly used in bioassays.

1. Introduction

Adverse ecological and human health effects associated with new and existing chemicals used in pharmaceuticals, personal care products, agriculture, and industry have become increasingly prominent in recent decades.^1,2 This has prompted a push for high-throughput, inexpensive methods of chemical risk assessment to inform prioritization and regulatory needs. In accordance with the Tox21 initiative to reduce animal testing, high-throughput in vitro bioassays have gained popularity as a screening step for a variety of relevant toxicological endpoints, such as cytotoxicity, endocrine disruption, and aryl hydrocarbon receptor agonism or antagonism.^3–5 In addition, high-throughput platforms for in vivo toxicity assessments using aquatic organisms such as zebrafish (Danio rerio), Daphnia magna, and Caenorhabditis elegans have been established.^6–8 These high-throughput in vitro and in vivo bioassays are typically conducted in plastic microplates that can have well numbers ranging from 6 to 1536 wells, with 96-well plates used most often.⁹

Despite the cost and time benefits associated with high-throughput bioassays, a prominent challenge for bioassays conducted in microplates is chemical losses. Chemical losses can occur due to volatilization and sorption (adherence and diffusion of the chemical to the well walls).¹⁰ Chemical losses in microplates are of concern for bioassays because bioassays often try to establish a dose–response relationship between chemical dose and measured effect. Nominal concentration is often used to describe dose; however, chemical losses can create a large discrepancy between nominal concentration and the actual aqueous concentration.^10,11 Thus, the measured effect associated with a nominal concentration may be the response to a much lower actual concentration, and resulting dose–response curves may underestimate bioactivity.

Microplates are typically made of polystyrene, polypropylene, cyclic olefin copolymer, or cyclic olefin polymer, which can act as a sink for compounds with low aqueous solubility.⁹ Sorption losses are a function of the chemicals hydrophobicity, which can be numerically described by its octanol–water partition coefficient (K_ow). Fischer et al.¹² generated a model of chemical losses due to sorption to polystyrene microplates and aqueous media based on chemical K_ow and reported losses of 20% to 50% for chemicals with K_ow > 3, losses of 50% to 80% for chemicals with K_ow > 4, and losses over 80% for chemicals with K_ow > 6. Similarly, Huchthausen et al.¹³ reported a 4-fold reduction in aqueous concentrations compared to nominal concentrations for a variety of organic acid pharmaceuticals in polystyrene microplates. The aforementioned studies that have quantified chemical losses in microplates have relied on extensive extraction and analytical techniques, such as solid phase microextraction paired with liquid or gas chromatography and high-resolution mass spectrometry.^10–13 These methods are time and resource consuming, making it difficult to match the high-throughput nature of the bioassays conducted in the microplates. Additionally, these methods are typically not aligned with green analytical chemistry principles, which aim to improve practicality and decrease negative environmental and health impacts associated with analytical methods.^14,15 Although the models generated from these data sets provide an alternative to measuring chemical concentrations directly by predicting chemical losses without requiring any additional resources, a gap still exists for low resource analytical methods to explicitly measure chemical concentrations in microwells.

We propose the use of fluorescence spectroscopy to measure chemical losses in high-throughput bioassays using a fluorescent plate reader as a low resource alternative to the existing analytical methods in literature. Plate readers are commonly used in high-throughput bioassay procedures, as the dose-dependent variable in many in vitro bioassays is luminescence, UV absorption, or fluorescence measured by a plate reader.^16,17 Therefore, this approach allows for the quantification of chemical losses in a bioassay procedure without the need for additional instrumentation or sample extraction and preparation steps. This approach requires the chemicals of interest to exhibit autofluorescent properties. Autofluorescence is fluorescence that occurs naturally for a given chemical compound based on its structural properties. Properties such as rigidity, like that observed in aromatic structures, conjugated double bonds, and the absence of halogen atoms can all contribute to a compounds autofluorescence.^18,19 Classes of autofluorescent chemicals of interest include polycyclic aromatic hydrocarbons (PAHs) and their derivatives,^20,21 pesticides,^22,23 and pharmaceuticals.^23,24

The development of a fluorescent plate reader method for quantifying chemical losses can be summarized in the following objectives: (1) develop a quantitative method for target chemicals in media with a fluorescent plate reader, (2) establish background fluorescence and potential interference from biological components, (3) quantify and describe kinetics of abiotic and biotic losses in an exposure experiment, and (4) compare abiotic losses to existing predictive models. This method was developed using a high-throughput embryonic zebrafish assay as the model bioassay, which is a well-established high-throughput in vivo assay used to assess chemical toxicity.^25,26 The model chemicals used in method development were two environmentally relevant polycyclic aromatic hydrocarbons previously shown to be toxic in embryonic zebrafish, 2-hydroxynaphthalene and acridine.²⁷

2. Materials and methods

2.1. Chemicals

Acridine (99.6%) (CAS: 260-94-6) and 2-hydroxynaphthalene (99.7%) (CAS: 135-19-3) and dimethyl sulfoxide (DMSO) (CAS: 67-68-5) (99.5%) were purchased from Sigma Aldrich (St. Louis, MO). Acridine and 2-hydroxynaphthalene stocks were prepared in DMSO at a concentration of 10 mM for use in zebrafish exposure studies.

2.2. Zebrafish husbandry

Tropical 5D wild-type zebrafish were maintained in 14 hours light/10 hours dark using protocols in compliance with the Institutional Animal Care and Use Committee (ACUP 2021–0166) at the Sinnhuber Aquatic Research Laboratory at Oregon State University. Fertilized zebrafish embryos were collected at ~4 hours post fertilization (hpf) and enzymatically dechorionated.²⁸ Dechorionated embryos were hand-pipetted into microwells containing 100 μL of embryo media (EM) for all experiments. The preparation of the plates used for exposure tests is described in Section 2.8. All plates were kept in a dark 28 °C room for the duration of the experiments, removed only for fluorescent measurements and morphology screening. Plates containing zebrafish embryos were placed on a 235 rpm shaker table for the first day of development, and then removed from the shaker table and stored on a stagnant shelf for the remainder of the experiments. This is common practice for zebrafish embryo chemical exposure, as it allows for mixing in the first 24 hours of development.²⁹ Zebrafish embryos were allowed to develop for 5 days. After 5 days, zebrafish embryos were assessed for mortality and 8 developmental endpoints pertaining to: cranium (eye/snout/jaw), axis, edema (heart/yolk sac), swim bladder/somites, lower trunk/caudal fin, brain, skin, and notochord development described in Truong et al.⁸

2.3. Fluorescent plate reader operation

All fluorescence spectroscopy methods were developed with the BioTek Synergy Mx microplate reader. The microplates used in all experiments were polystyrene black-walled 96-well plates purchased from Life Technologies Corporation (Carlsbad, CA), which were used to eliminate interfering signals from neighboring wells produced during fluorescent measurements. All measurements were taken from the top of the plate with sensitivity and filter width settings of 75 and 20, respectively. The plate reader was validated for fluorescent measurements with the corners, sensitivity, and linearity tests per the operator manual. Measurements were corrected for Raman scattering, or inelastic scattering, by subtracting media blank measurements from the intensities measured in PAH containing wells.

2.4. Quantitative method development

To create quantitative methods for measuring the test chemicals (2-hydroxynaphthalene and acridine), the optimal excitation and emission wavelength pairing for each chemical was first determined using 3-dimensional fluorescent scans. 3-Dimensional scans collect fluorescent emissions over a range of excitation and emission wavelengths to create an excitation/emission matrix (EEM). The maximum of this matrix corresponds to the optimal excitation and emission wavelengths for the chemical. The difference between the optimal excitation and emission wavelengths is the chemical’s Stokes shift, which is often used to identify chemicals.³⁰

After the appropriate excitation and emission wavelengths were determined, standards of the individual chemicals were prepared in EM. Standard concentrations ranged from 0 to ~35 μM, with three wells prepared per concentration. Plates were measured immediately after standard solutions were added to minimize the influence of abiotic losses. Note that fluorescent intensities are dependent on both concentration and path length, or the depth of the solution in each well. Thus, the volume of standard solutions added to each well was equivalent to the volume of media used in all experiments, to keep the path length variable constant. Method detection limits (MDLs) were determined according to EPA procedure 40 CFR 136 Appendix B (US EPA, 2016), with n = 8 replicates for standards and blanks.

2.5. Establishing background fluorescence and impact of UV on zebrafish embryos

The excitation wavelengths often used for measuring autofluorescent compounds fall in the ultraviolet (UV) range, which are potentially damaging to zebrafish embryos.³¹ However, the exposure to UV radiation is very brief, approximately 100 μs, in a fluorescent plate reader. Therefore, it is important to determine if the UV radiation doses the embryos received during plate reader measurements were lethal or disrupted development. To do so, microplates were prepared with zebrafish embryos in 100 μL EM per well. Embryos were divided into 7 groups with 10 embryos per group. The control group (group 7) were not exposed to UV, while groups 1 to 6 received one UV dose per day for up to 5 days. The group number corresponding to number of days the embryos were exposed to UV (i.e. group 1 received one UV dose on day 1, group 2 received one UV dose on day 1 and one UV dose on day 2, etc.). Mortality and morphological defects were assessed at ~120 hpf.

2.6. Determining photodegradation rate

Although fluorescence spectroscopy is commonly considered a non-destructive technique,^32,33 photodegradation of the chemicals catalyzed by the UV light remains a possibility. PAHs are known to undergo photodegradation with exposure to UV light.^34,35 Photodegradation experiments were carried out with the plate reader to determine the contribution of photodegradation to the overall abiotic losses. Plates were prepared with experimental wells containing 100 μL EM with ~35 μM of 2-hydroxynaphthalene or acridine, and control wells containing 100 μL EM. Fluorescent measurements were taken in rapid succession, approximately one minute apart, to minimize the impacts of sorption of the chemicals to the well walls. Thus, the PAH losses observed with each subsequent measurement were assumed to be due to photodegradation.

2.7. Quantifying sorption losses

PAHs are also known to adsorb to the well walls of 96-well plates due to their hydrophobicity. Chlebowski et al.³⁶ investigated the sorption losses of a variety of PAHs and nitrogen containing PAHs with polystyrene 96-well plates and reported losses up to 43% due to sorption. Acridine and 2-hydroxynaphthalene have estimated log K_ow of 3.4 and 2.7, respectively.^37,38 Therefore, losses due to sorption of the PAHs to the microplate well walls were expected, particularly for acridine. To quantify chemical losses due to sorption of the chemicals to the well walls, sorption control wells (n = 16) containing ~35 μM of 2-hydroxynaphthalene or acridine in 100 μL EM were prepared on the same plates used in the bioassay described in the following section. Aqueous concentrations of 2-hydroxynaphthalene and acridine were measured daily. Changes in aqueous chemical concentrations were assumed to be due to sorption and photodegradation. Volatilization of the compounds was assumed to be negligible based on Henry’s coefficients as reported in the Hazard Substances Data Bank (HSDB): 4 × 10⁻⁷ atm m³ mol⁻¹ and 2.7 × 10⁻⁸ atm m³ mol⁻¹ for acridine and 2-hydroxynaphthalene, respectively.

Because the sorption control wells were on the same plates the zebrafish bioassays were conducted on, the plates were under two different conditions over the course of the 5 day experiment. Plates were on a 235 rpm shaker table for the first day of exposure, then kept on a stagnant shelf from days 1 to 5. To determine the impact of agitation on sorption of the chemicals, a separate set of microplates were prepared with experimental wells containing 100 μL of ~35 μM of 2-hydroxynaphthalene or acridine and control wells containing 100 μL EM. These plates were measured for initial PAH concentrations with the fluorescent plate reader before being sealed with parafilm and allowed to equilibrate for 24 hours either on a 235 rpm shaker table or on a stagnant shelf in the dark. After 24 hours, plates were measured again with the fluorescent plate reader and PAH losses in plates under both conditions were quantified.

2.8. PAH exposure test

Duplicate 96-well microplates were prepared for PAH exposure tests, with all wells containing 100 μL of liquid volume. The target initial concentration of 2-hydroxynaphthalene and acridine was 35 μM, the lowest observed adverse effect level (LOAEL).²⁷ DMSO content in the wells was 0.28% and 0.33% for acridine and 2-hydroxynaphthalene, respectively. EM solutions were hand-pipetted into microwells, and initial PAH measurements were taken with the plate reader before embryos were added. Embryos were then added to wells via hand-pipetting and checked under a microscope for any anomalies before plates were measured again with the plate reader. Each plate contained 16 replicate wells for zebrafish embryos exposed to 2-hydroxynaphthalene or acridine, providing in total 32 replicates per chemical exposure. The initial fluorescent plate reader measurements on day one was followed by one measurement per day for the remaining 4 days incubation period. Plates were sealed with parafilm in between fluorescent measurements; however, these covers were temporarily removed for each fluorescent measurement.

In addition to the sorption control wells described in Section 2.7, these plates contained control wells with zebrafish embryos and embryo media. These wells served as UV exposure controls to confirm UV exposure from fluorescent measurements was not impacting the zebrafish embryos. Parathion was selected as the chemical for the positive toxicity control group as standard for this bioassay.³⁹ Embryos exposed to parathion were not measured with the plate reader. At ~120 hpf, zebrafish embryos were assessed for mortality and morphological abnormalities.

2.9. Rate derivation and comparison to predictive models

A pseudo-first order model for the aqueous PAH concentrations over the course of the PAH exposure tests was fit to the experimental data. The goal of this modeling was to help determine the contribution of photodegradation, sorption, and uptake to the decreases in aqueous concentrations of 2-hydroxynaphthalene and acridine over the course of the experiment. The net uptake by zebrafish embryos, sorption to the microwells, and photodegradation were considered in the development of the model. Thus, the pseudo-first order model can be described with the equations listed in Table 1, where $C$ is the concentration of the PAH in the aqueous phase (μM), $C_{\mathrm{i}}$ is the initial aqueous PAH concentration (μM), and $k$ is the rate constant of the reaction described in the rate constant column. All rate constants have units of inverse days and were derived in the manner described below.

Table 1.

Process for deriving rate constants in kinetic model. Bolded constants are the constants derived in the corresponding process step

Step	Data set used	Equation	Rate constant
1	Section 2.6	$C(t)=C_{\mathrm{i}}{e}^{-k_{1}t}$	$k_1={\mathit{k}}_{\mathbf{photo}}$
2	Section 2.7	$C(t)=C_{\mathrm{i}}{e}^{-k_{2}t}$	$k_2=k_{\text{photo}}+{\mathit{k}}_{\mathbf{sorption}}$
3	Section 2.8	$C(t)=C_{\mathrm{i}}{e}^{-k_{3}t}$	$k_3=k_{\text{photo}}+k_{\text{sorption}}+{\mathit{k}}_{\mathbf{uptake}}$

In all cases, quantified aqueous PAH concentrations were linearized to the equations shown in Table 1 to perform regression analysis:

$$\mathrm{l}\mathrm{n}\left(\frac{C\left(t\right)}{C_0}\right)=-kt$$ (1)

where $k$ is the slope of the linear regression line fit to the linearized experimental data. Photodegradation rates were derived from photodegradation experiments. Because the timeline of this experiment did not match the timeline of the PAH exposure test, the photodegradation rate was derived first as a function of the number of measurements taken. This rate was then converted to inverse days with the ratio of measurements per days; one measurement per day to match the daily measurement strategy in the PAH exposure test. Sorption and uptake loss rates were assumed to be negligible due to the short timescale of the experiment and absence of zebrafish embryos in the wells. Sorption rate was derived from the losses in the abiotic wells of the PAH exposure test. Uptake losses were assumed to be zero because of the absence of zebrafish embryos in these wells. Finally, uptake rate was derived from the data collected from zebrafish embryo-containing wells in the PAH exposure test.

The aqueous concentrations measured in control wells (wells without embryos) over time were compared to the predicted concentration over time using the Fischer et al. model for sorption to polystyrene microwells.¹² This model considers input parameters such as the geometry of the wells, protein and lipid concentrations in the aqueous phase, and properties of the chemical of interest. Furthermore, the mass balance model created by Armitage et al.,¹¹ IV-MBM EQP v2.0, was used to predict the equilibrium concentration of our chemicals and compare with our final measured concentrations to determine if equilibrium was reached in our system. The full description of the model inputs is provided in Tables S4–S6.^†

3. Results

3.1. Quantitative method development

The excitation/emission wavelengths for acridine and 2-hydroxynaphthalene were 350/429 nm and 270/355 nm, respectively. The EEMs of these chemicals that were used to determine these measurement wavelengths are provided in Fig. S1.^† Linear calibration curves were established between concentration and fluorescent intensity for both compounds (R² ≥ 0.99) and are shown in Fig. S2.^† Because these measurements rely on the autofluorescence of the compounds, sensitivity is highly variable. Acridine was found to have a greater quantum yield than 2-hydroxynaphthalene, thus the MDL for acridine (0.65 μM) is approximately six times lower than that of 2-hydroxynaphthalene (3.97 μM). The MDL for 2-hydroxynaphthalene is still an order of magnitude lower than the initial concentrations used in exposure tests, therefore its low sensitivity does not impede measurements in future tests.

AGREE and BAGI metric approaches were used to assess the greenness of this quantitative method. This method scored a 0.62 on the 0 to 1 AGREE scale, and 72.5 on the 25 to 100 BAGI scale, indicating that this method is reasonably green and very practical.^14,15 The primary aspects of the method that reduced its greenness on the AGREE scale were the use of an off-line analysis and the use of non-biobased reagents. The primary aspects of the method that reduced its practicality on the BAGI scale were the fact that this method only measures 1 analyte at a time, and that 100 μL of material is needed. A full description of the AGREE and BAGI inputs and outputs is provided in Tables S2 and S3.^†

3.2. Impact of zebrafish embryo development on background fluorescence

The UV exposure test produced several important findings. Primarily, embryo mortality and development were not significantly impacted by the acridine and 2-hydroxynaphthalene measurement methods. The incidence of adverse effects for UV exposed embryos was not statistically significantly different than the control embryos that were not exposed to UV, as shown in Fig. S3.^† Therefore, the UV radiation dose the embryos received during their 5 days incubation did not lead to adverse health effects for the wavelengths used. Background fluorescence for normal embryos in all exposure groups were therefore averaged together and displayed in Fig. 1. While there has been documentation of adverse outcomes for UV exposed zebrafish embryos, including mortality, developmental abnormalities, decreases in hatch rate, and increased inflammation, the UV doses in these studies are orders of magnitude higher than the doses received by embryos in the present study, which was estimated to be 6.1 mJ cm⁻².^31,40,41 Banerjee and Leptin⁴² reported no significant difference in outcome for zebrafish embryos exposed to 6 mJ cm⁻², however these outcomes were reported for a single UV dose, whereas the present study entails up to six UV doses (one dose per day).

Fig. 1.

Change in normalized background fluorescence for 2-hydroxynaphthalene (A and B) and acridine (C and D) wavelengths for wells containing zebrafish embryos (ZE) with normal development (A and C) or abnormal development (B and D).

The second finding of importance is that the background fluorescence was not impacted by normal zebrafish embryo development over the duration of the experiment. This is illustrated by the overlap in background fluorescence of media and embryo containing wells in Fig. 1A and C. This becomes apparent only after normalizing the background intensity of wells containing zebrafish embryos to the initial measured background intensity after embryo placement. In comparing the background fluorescence directly before and after zebrafish embryo placement, it was observed that the placement of zebrafish embryos resulted in a 13 ± 3% and 13 ± 7% decrease in background fluorescent intensity for 2-hydroxynaphthalene and acridine wavelengths, respectively. This decrease is potentially due to the absorption of the excitation wavelength radiation by the embryo. Therefore, while embryo placement in the wells was observed to decrease the background fluorescence, the background fluorescence did not change after the initial reduction for the remainder of the 5 day incubation.

The third finding of importance from the UV exposure study is that the death of zebrafish embryos, as well as cranial deformities, had a significant impact on the background fluorescence. The incidence of these endpoints was no greater than 20% of the total number of zebrafish embryos. For zebrafish embryos that developed cranial deformities or mortality, a significant increase in background fluorescent intensity was measured, as shown in the red data series in Fig. 1B and D. This increase was more prominent for the acridine wavelength, where background fluorescent intensity in impacted wells increased by a factor of 2.8 to 22.8 compared to 1.2 to 1.7 for the 2-hydroxynaphthalene wavelength. The increase in fluorescence in the wells that contained dead or abnormal embryos may be due to cell/yolk sac lysis and release of fluorescent material into the aqueous phase. Similar reports of increased autofluorescence following cell death have been reported for a range of biological systems, including plant cells,⁴³ human eosinophils,⁴⁴ and zebrafish embryos, where increased fluorescent emissions were attributed to lipofuscin-like autofluorescence.⁴⁵

These results indicate that measurements for the target compounds may be obscured by embryo death or cranial malformation. Therefore, in the following PAH exposure experiment, sharp increases in fluorescence, defined as an inflection point in the first derivative of the measured fluorescent intensity, were considered to indicate embryo abnormality. Subsequent measurements from those wells were removed from further data analysis. Fluorescent inflections were in complete agreement with screening results for this exposure test. In other words, fluorescent inflections were measured in all wells that also had observed embryo mortality and/or cranial deformation in the screening results, and no fluorescent inflections were measured in wells that contained normal zebrafish embryos.

3.3. Derivation of first-order rate constants from measured PAH concentrations

The use of a regression based pseudo-first order model to describe photodegradation, sorption, and zebrafish embryo uptake was generally successful. The linear regression models for the linearized data are shown in Fig. 2.

Fig. 2.

Linear regression fit for derivation of rate constants $k_1$ (A), $k_2$ (C and D), and $k_3$ (E and F) as defined in Table 1. The legend in B applies to all panels in this figure. CI = confidence interval, LR = linear regression. The number of replicates included in each measurement is reported in Table S1.^†

Kinetic rate constants for $k_2$ and $k_3$ were derived in a piece-wise manner for t ≤ 1 day and t ≥ 1 day to accommodate the change in conditions (shaken for t < 1 day, stationary for t ≥ 1 day), as it was found that the shaker table had a significant impact on sorption (Fig. S4^†).

The linear fit for 2-hydroxynaphthalene data was generally poorer than that of acridine (seen in the R² values reported in Fig. 2C–F), possibly due in part to the lower sensitivity for 2-hydroxynaphthalene measurements. The exception to this was photodegradation, where the fit for 2-hydroxynaphthalene was slightly better than acridine (R² = 0.72 vs. 0.68 for 2-hydroxynaphthalene and acridine, respectively). However, the linear regression fits for the $k_1$ were poorer than that of $k_2$ and $k_3$ (seen in the R² values reported in Fig. 2A compared to Fig. 2C–F), due in part to the shallowness of the slope.

Comparisons of rate constants for photodegradation, sorption, and uptake reveal differences in chemical transport and behavior based on their physicochemical properties. Photodegradation was a minor contributor to losses in the system for both compounds compared to sorption and embryo uptake, as shown in Table 2. Additionally, the rate of photodegradation was independent of mixing via the shaker table, whereas the rate of sorption and uptake changed significantly between t ≤ 1 day and t ≥ 1 day.

Table 2.

Summary of kinetic constants derived from the linear regression of experimental data and used to model PAH concentrations in controls well and embryo wells.

Rate constant (day−1)	2-Hydroxynaphthalene			Acridine
	t ≤ 1 day		t ≥ 1 day	t ≤ 1 day		t ≥ 1 day
$k_1$		0.007 ± 0.0006			0.007 ± 0.0006
$k_2$	0.25 ± 0.006		N/A	0.38 ± 0.008		0.19 ± 0.002
$k_3$	0.62 ± 0.04		0.22 ± 0.01	0.7 ± 0.018		0.16 ± 0.004
$k_{\text{photo}}$		0.011 ± 0.0008			0.007 ± 0.0006
$k_{\text{sorption}}$	0.24 ± 0.006		N/A	0.37 ± 0.008		0.18 ± 0.002
$k_{\text{uptake}}$	0.39 ± 0.04		0.22 ± 0.01	0.32 ± 0.02		N/A

Error represents standard error. N/A is used to indicate where rate constant was ≈ 0

Sorption rates were significantly higher when plates were on the shaker table (t ≤ 1 day) for both PAHs. This may be due to a larger surface area of the well walls in contact with the embryo media while plates were on the shaker table, leaving more sites available for PAH sorption. Additionally, the concentration gradient serving as the driving force for diffusion of the PAHs into the polystyrene was also greatest at the start of the experiment. While factors such as serum content or dissolved organic matter in media has been shown to impact sorption,^12,46 the impact of mixing on sorption has not been widely reported on in the literature. However, losses due to sorption observed in this study are consistent with losses reported in other zebrafish exposure studies, where control concentration decline has been attributed to sorption to well or tank walls.^47,48 There was a factor of 2 decrease in sorption rate when plates were removed from the shaker table for acridine, while no rate of sorption was derived for 2-hydroxynaphthalene after removal from the shaker table. The low sorption rate of 2-hydroxynaphthalene after removal from the shaker table suggests that equilibrium may have been reached between the media concentrations and well walls. Acridine is expected to adsorb to the well walls to a greater extent than 2-hydroxynaphthalene due to its higher log K_ow value, supporting why more acridine abiotic losses attributed to sorption occurred even after plates were removed from the shaker table. Sorption losses are discussed further in Section 3.4.

The rate of PAH uptake by zebrafish embryos decreased after plates were removed from the shaker table for a factor of ~2 for 2-hydroxynaphthalene, while uptake of acridine appeared to cease completely. No uptake rate was reported for acridine at t ≥ 1 in Table 2 because the derived rate and error range encompassed 0. While this may suggest that acridine uptake was only occurring in the first day of exposure, these results may also indicate that there were interfering signals being measured at the acridine wavelength for the zebrafish embryo containing wells. Interfering signals could come from abnormal embryo development, as discussed previously, but may also be due to PAH metabolism. Certain PAH metabolites can exhibit autofluorescent properties as well, thus PAH metabolites excreted by the zebrafish embryos could also contribute to background fluorescence.²⁰

Although zebrafish embryo uptake rate constants for 2-hydroxynaphthalene and acridine do not exist in the literature, other studies have investigated uptake kinetics of PAHs and polar compounds. Brox et al.⁴⁹ derived uptake rates for a variety of ionic and non-ionic polar compounds in embryonic zebrafish and reported uptake rates within the range of those derived in this study. The same order of magnitude for 2-hydroxynaphthalene uptake rate was reported for valproic acid, a comparable compound based on molecular weight and K_ow.⁴⁹ A limitation of this study is the inability to distinguish between PAH accumulation and PAH metabolism. PAH metabolism is predominantly carried out by cytochrome P450 (CYP) enzymes, which are both expressed and active in embryonic zebrafish.⁵⁰ Therefore, uptake rate cannot be equated to bioaccumulation rate, as this would overestimate bioaccumulation by neglecting metabolism. Further work on the analysis of PAH content in the embryos, as well as PAH metabolites, would help differentiate between uptake, metabolism, and bioaccumulation.

The rates derived for $k_{\text{photo}}$, $k_{\text{sorption}}$, and $k_{\text{uptake}}$ reported in Table 2 fit well to the experimental data for 2-hydroxynaphthalene and acridine, as shown in Fig. 3.

Fig. 3.

PAH concentrations for 2-hydroxynaphthalene (A) and acridine (B) and first-order model fit with kinetic parameters described in Table 2. The vertical line indicates when plates were removed from the shaker table at t = 1 day.

The initial decrease in measured PAH concentrations was not sufficiently captured by the t ≤ 1 day models, as seen by the large error particularly in Fig. 3A. This indicates that initial transport of the chemicals while plates are on the shaker table may be more complex and require a more elaborate model. Designing future experiments to include more data points on the first day of incubation would improve model fitness in this area and help determine if a different modeling strategy is more appropriate.

Media concentrations of 2-hydroxynaphthalene and acridine measured with fluorescence in the PAH exposure test allowed for the estimation of not only rates of abiotic losses and PAH uptake, but also PAH dose. Total uptake mass per zebrafish embryo, corrected for abiotic losses, was estimated to be 2.09 ± 0.89 nmol and 0.67 ± 0.26 nmol for 2-hydroxynaphthalene and acridine, respectively. Because uptake rate was estimated to be ~0 for acridine after 24 hours of exposure, the uptake amount reported for acridine accounts only for the first day of exposure. Estimates for uptake amount for both PAHs are significantly lower than the respective exposure masses of 4.16 ± 0.11 nmol and 3.15 ± 0.15 nmol, indicating that the previously reported toxicity metrics may understate the toxicity of these PAHs.²⁷ A strength of this method is the ability to measure concentrations in individual wells, capturing the heterogeneity of the concentrations across the plate with zebrafish embryos that are assumed to respond in a reproducible manner within a population. The close clustering of all control measurements demonstrates that abiotic losses are consistent across the plate, therefore variability in embryo containing wells is likely due to the variability in the embryos themselves.

3.4. Comparison of experimental data to existing predictive models

The concentrations of 2-hydroxynaphthalene or acridine in control wells (wells without zebrafish embryos) can be compared to existing predictive models for sorption of chemicals to microwells developed by Fischer et al. and Armitage et al.^11,12 While the Fischer et al. model generated predicted media concentrations over the course of the exposure experiment, the Armitage et al. model predicted the final equilibrium concentration of the system. The results of the model output compared to the measured concentrations and first-order fit are shown in Fig. 4.

Fig. 4.

Comparison of predicted media concentrations of 2-hydroxynaphthalene (A) and acridine (B) to average media concentrations measured in control wells with the fluorescence spectroscopy method. The regression based first-order model fit to the experimental data is also shown for comparison.

Although the predicted 2-hydroxynaphthalene concentrations from the Fischer et al. model was near the measured concentration after 5 days, the model deviated significantly from the measured concentrations in the first 24 hours. This may be due to the reduced temporal and spatial resolution of the Excel model used, which was noted by Fischer et al. to impact the accuracy of the model in the initial stage.¹² The opposite trend is observed for acridine, where the fit of the Fischer et al. model in the first 24 hours of the experiment was in good agreement with the measured concentrations, but the model deviated increasingly from the measured concentrations as time progressed. A potential source of error in parameterizing the Fischer et al. model is the partitioning coefficient for polystyrene and water and the diffusion coefficients. The partitioning coefficients for polystyrene and water (K_PS/w) for 2-hydroxynaphthalene and acridine were estimated using the linear correlation between K_ow and K_PS/w presented in Fischer et al., which had a reported R² of 0.82. The estimated log K_PS/w for 2-hydroxynaphthalene and acridine were 1.46 and 1.85, respectively. Because no relationship between other chemical properties and the diffusion coefficient was established in Fischer et al., the average diffusion coefficient from the dataset used to develop the Fischer et al. model was used for both 2-hydroxynaphthalene and acridine, as suggested by the authors. Future work to derive K_PS/w and diffusion coefficients for 2-hydroxynaphthalene and acridine experimentally would likely improve model fitness, however there is still reasonably good agreement between measured and predicted concentrations shown in Fig. 4.

The predicted equilibrium concentration using the IV-MBM EQP v2.0 model was lower than the 2-hydroxynaphthalene concentrations measured after 5 days, suggesting that equilibrium was not yet reached between the polystyrene phase and aqueous phase for this chemical. However, the measured concentrations of 2-hydroxynaphthalene did not change significantly after the first day of the experiment. This may indicate the system reached equilibrium after 1 day, with the actual equilibrium media concentration being approximately 31.5 μM instead of 25 μM. While the Armitage et al. model may have overestimated the partitioning of 2-hydroxynaphthalene into the polystyrene phase, this model appears to have underestimated partitioning into polystyrene for acridine. This is supported by the observation that acridine concentrations measured after 5 days were below the predicted equilibrium concentration. The duration of experiments should be extended into the future to try to capture the system in equilibrium, as 5 days did not appear to be sufficient in the present study.

3.5. Zebrafish bioassay results

Endpoints that were observed in zebrafish embryos exposed to 2-hydroxynaphthalene or acridine in the present study are in agreement with a previous exposure study.²⁷ Many endpoints were shared amongst embryos exposed to acridine or 2-hydroxynaphthalene, namely mortality, yolk sac and pericardial edema, and pectoral fin malformation. Axis and cranial deformities were also observed in embryos exposed to 35 μM acridine. The incidence of individual endpoints is reported in Fig. S5,^† with any effect observed in 100% and 97% of embryonic zebrafish exposed to acridine and 2-hydroxynaphthalene, respectively. All measured endpoints for acridine matched the expected endpoints, however we report an additional endpoint for 2-hydroxynaphthalene, cranial deformation, that was not observed previously for 2-hydroxynaphthalene exposure.²⁷ These additional deformities may reflect additive toxicity effects from the 2-hydroxynaphthalene and UV co-exposure that could not be accounted for by individual exposures. Almeida et al.⁵¹ reported similar phenomena for zebrafish embryos exposed to UV and triclosan, denoting that increases in toxicity may also be due to photodegradation of the compound. However, further work is needed to elucidate potential products of photodegradation, and their toxicity to zebrafish embryos.

In comparing wells with a detected fluorescent inflection and wells with observed mortality or morphological endpoints, fluorescent inflections were no longer in good agreement with screening results as they were in the UV only exposure study. This is predominantly due to the PAH signal obscuring potential increases in fluorescence from embryo mortality or deformation. For 2-hydroxynaphthalene, which had an order of magnitude weaker fluorescent intensity than acridine, 26% of embryos that had a screening hit for mortality or cranial deformity did not have an observed fluorescent inflection. For acridine, 63% of embryos with a screening hit did not have an observed fluorescent inflection. Thus, the stronger the chemical’s autofluorescence, the less likely fluorescent inflections will be detected. This may explain why acridine uptake appeared to cease after 24 hours, as further reductions in acridine concentrations may have been obscured by the increased background fluorescence associated with embryo mortality. However, only one well out of all screened wells (1.6%) had a normal embryo with a fluorescent inflection. It should be noted that it is possible that embryos died in between the time of the final fluorescent measurement and the time they were screened, which would also contribute to the low incidence of fluorescent inflections compared to screening results. Daily screening steps should be included in future experiments to ensure all abnormal embryos are removed from the fluorescent data set.

4. Conclusions

In conclusion, the novel fluorescent plate reader method presented in this study provided insights into the uptake and abiotic losses of 2-hydroxynaphthalene and acridine in a high-throughput embryonic zebrafish bioassay. This plate reader method required no sample pooling or extraction steps and allowed for the quantification of μM level concentrations of PAHs in individual microwells containing 100 μL of aqueous solution. The use of this method with zebrafish embryos did not negatively impact zebrafish embryo development, nor did normal embryo development impact background fluorescence. However, abnormal embryo development resulted in a measurable change in background fluorescence. Pseudo-first order kinetic models considering embryo uptake, photodegradation, and sorption fit well with experimental data, with uptake and sorption being the major contributors to overall PAH losses. Existing predictive models fit also the experimental data moderately well, however future work to derive key model parameters, K_PS/w and diffusion coefficients, is needed to improve model accuracy. Phenotypic outcomes and embryo mortality in PAH exposed embryos were consistent with previous reports, except for additional cranial deformities observed in the present study with 2-hydroxynaphthalene. This cautions potential additive toxicity of the UV exposure and chemical exposure that this plate reader method entails, necessitating further investigation with a larger suite of chemicals. Nevertheless, this novel, low resource method provides rapid information on chemical losses in a microplate-based bioassay that can be applied to a variety of microplate-based bioassays and autofluorescent chemicals.

Data availability

The data supporting this article have been included as part of the ESI.^†