Naphthenic acid fraction compounds, produced by the extraction of bitumen from oil sands, alter survival and behaviour of juvenile yellow perch (Perca flavescens)

Our study followed a before–after-control-impact design (Underwood 1993) (Fig. 1a). After a period of acclimatization to captivity (4 days), groups of five wild-caught fish were held in replicate mesocosms filled with clean lake water for 1 week (hereafter, “baseline phase”), then exposed to one of three NAFC treatments for a second week (hereafter, “exposure phase”). The day NAFCs were added (i.e., 22 September 2020) is referred to as “day 0” (Fig. 1a). NAFC treatments consisted of three nominal concentrations (clean lake water as a negative control, 2 mg/L NAFC, and 15 mg/L NAFC), with four mesocosms per treatment distributed over the study area in a random block design (Fig. 1b). The experiment was conducted at the QE3 Living Lab, an outdoor experimental ecotoxicology facility on the property of the Queen's University's Biological Station near Elgin, Ontario, Canada. Animal husbandry and procedures were approved by the Queen's University Animal Care Committee (Orihel 2020–1829).

In early September 2020, juvenile yellow perch were collected from Warner Lake, Ontario (9.2 ha, mean depth 2.9 m, maximum depth 6.4 m, 44°33′′55′′N 76°19′′35′′W) using baited minnow traps (Gee; 42 × 19 cm, 6.4 mm square galvanized mesh). Captured fish were placed into 16 L holding containers filled with lake water and transported to shore, where they were transferred to aerated holding tanks (300 L). Fish of similar size (6.3 g ± 1.2 g mass, 6.9 cm ± 0.7 cm fork length, mean ± SD) were selected to reduce any associated individual variation in other measures (OECD 2013). Retained fish (n = 60) were transferred to the 12 outdoor mesocosms (5 fish per mesocosm) and acclimatized for 4 days to reduce the effects of handling stress on fish. The mesocosms (75 cm diameter × 91 cm height; Fig. 1c) were lined with food-grade polyethylene plastic bags and filled with 60 L of filtered water from Warner Lake. Fish were fed ad libitum daily with frozen brine shrimp (Artemia spp.) delivered through a plastic tube to reduce disturbance and habituate fish to the food delivery process prior to food stimulus tests. Water temperature, pH, dissolved oxygen, conductivity, salinity, and ammonia levels in the mesocosms were measured daily using Hana (H198129) and Hach meters (HQ40d) and ammonia test strips (API). Light intensity and temperature were monitored hourly using HOBO data loggers (Onert, Bourne, Massachusetts, USA). All water parameters remained within the recommended ranges for freshwater fish throughout the experiment (Tables S1 and S2) (OECD 2013). We monitored the mesocosms daily for any mortalities, and all fish were weighed (Thermo Fisher Scientific Denver Instrument P-403, ±0.001 g) and photographed (Canon EOS 100 D) at the beginning (day −6) and end of the experiment (day 6) or on the day they were recorded as a mortality. Photographs were later analyzed using ImageJ2 to determine fork length.

This experiment used an NAFC extract prepared from a large volume sample from an active tailings pond in Canada's oil sands region. The preparation and analysis of this NAFC extract have been described in detail elsewhere (Marentette et al. 2015a, b; Bartlett et al. 2017; Reynolds et al. 2022). In summary, a large volume (<1000 L) of OSPW was acidified to pH < 2 and allowed to settle for 24 h prior to centrifugation. Following centrifugation, the insoluble pellet was resuspended in 0.1 N NaOH (pH 12) to solubilize organic acids and repeatedly centrifuged to remove suspended solids and humins. The alkaline mixture was filtered through a column of diethlyaminoethyl-cellulose (DEAE-cellulose) to remove high-molecular-weight humic-like materials. The alkaline DEAE-celluose column eluent solution was then subjected to repeated liquid-liquid extractions using DCM to remove neutral organics such as PACs. The total NAFC concentration of the extract was 2.64 × 10³ mg/L, as measured via negative-ion electrospray ionization high-resolution mass spectrometry (ESI-HRMS, Orbitrap) using a method similar to Headley et al. (2013). The extract was stored in darkness at pH 10 and 4 °C until diluted with lake water to nominal concentrations of 2 and 15 mg/L for this experiment. The initial pH of the NAFC treatments ranged from 8.7 to 10.0 and was titrated to 8.0 ± 0.1 with 0.1 mol/L hydrochloric acid. Throughout the experiment, pH values ranged from 7.0 to 7.8 for controls, 7.2 to 7.7 for the 2 mg/L NAFC treatments, and 7.0 to 7.7 for the 15 mg/L NAFC treatments (Table S1). Water samples from the treatment and control mesocosms were collected in amber glass bottles on day 0 (2 h after NAFC addition), day 3, and day 6 of the exposure phase, and stored at 4 °C until analysis to confirm target NAFC concentrations were achieved.

NAFCs were extracted from mesocosm water samples following a previously reported procedure (Headley et al. 2002). Briefly, Biotage Isolute ENV + cartridges (Biotage, Uppsala, Sweden) were mounted in a solid-phase extraction (SPE) manifold, rinsed with 6 mL of MilliQ water, 6 mL of LCMS-grade methanol (Fisher Scientific, Hampton, NH, USA), and conditioned with 6 mL of MilliQ water. Samples were then acidified to pH < 2, and then extracted at approximately 5 mL/min under gentle vacuum (approximately 50–60 torr). Following sample loading, SPE cartridges were rinsed with 5–6 mL of MilliQ water under gravity to de-salt. Samples were eluted from cartridges with 6 mL of LCMS-grade methanol under gravity, then evaporated under gentle N₂ flow in a water bath at approximately 40 °C. Samples were reconstituted into 50:50 acetonitrile:water with 0.1% NH₄OH, then transferred to LCMS vials for analysis.

Concentrations of NAFCs were quantified using an external standard calibration method (Bauer et al. 2015) using NAFC extracts derived from OSPW (Rogers et al. 2002). This OSPW-derived NAFC mixture was originally quantified against a Kodak NA mixture via FTIR, as described by Rogers et al. (2002). Sample extracts were bracketed by calibration standards and blanks for every 10 samples. All samples were run in a single sequence batch, except where sample dilutions were required to reach the method's instrument linear calibration range between 1.0 and 100 mg/L, which were analyzed in a second batch with independent blanks and calibration brackets. The method detection limit (using 100 mL sample aliquots) was 0.10 mg/L, with an estimated extraction efficiency of 105% (rsd ≤ 4%). NAFC concentrations are reported as semiquantitative results as authentic reference standards are not available. Sample data from these non-targeted analyses were pre-processed in XCalibur version 2.2. (Thermo Fisher Scientific, Waltham, MA, USA), then formulae were assigned to accurate mass data at a level-4 identification confidence level (Schymanski et al. 2014) using Composer 1.5.6 software (Sierra Analytics, Modesto CA, USA). Formula assignment was limited to compounds between 100 and 500 m/z with a maximum tolerable mass error of ≤3 parts per million.

Quantification provided by HRMS and Orbitrap-MS should be considered semiquantitative. A lack of standardized NA or NAFCs, as well as a lack of agreement on analytical methodology, similarly limits all available methods from definitively quantifying NAFCs (Kovalchik et al. 2017). Further, there are many factors affecting ion abundances and ionization efficiency in a mass spectrometer, which will selectively promote or suppress the response factor of particular compounds and compound classes (Palacio Lozano et al. 2020). Comparisons based on spectral features, such as proportions of heteroatomic classes or particular formulae, must be interpreted only for relative trends and do not directly report absolute amounts of each.

We assessed three behaviours in assays at different time points: swim activity, foraging activity, and antipredator responses. All behavioural assays involved filming the groups of fish using overhead-mounted digital cameras (Canon HF R800; Fig. S1). Swim activity was assessed by filming the fish in their 60 L exposure mesocosms for 4 min in the absence of an introduced stimulus (“no-stimulus activity test”). A 4 min film time was chosen based on modified methods from Mathis and Smith (1993) and Chivers and Smith (1995). Videos of regular swim behaviour were recorded in the morning (9–11 am) during both the baseline (day −1) and exposure (day 5) phases. Foraging activity was assessed by filming the fish for 4 min periods before (pre-food stimulus) and after (post-food stimulus) a food stimulus was introduced on day 5 (“food stimulus test”). The food stimulus consisted of an injection of frozen brine shrimp (2 g) delivered through a plastic tube in the same manner that food was delivered throughout the experiment. To assess anti-predator behaviour, fish from each mesocosm were transferred together from the exposure mesocosms to 60 L cylindrical test tanks at the end of the exposure phase (day 6), allowed to acclimatize for 10 min, and then filmed for 4 min periods before (pre-predator stimulus) and after (post-predator stimulus) the introduction of a predator stimulus (“predator stimulus test”). The predator stimulus was a model bird head resembling the great blue heron (Ardea herodias; Fig. S1), a common local predator of juvenile yellow perch. The model was mounted on a rod, introduced at the centre of the test tank in a downward motion to strike the bottom of the tank, and then immediately retracted.

Behavioural videos were reviewed with the display overlaid with an 11 × 11 grid, where the grid width was approximately one fish body length (Fig. S2). During each behavioural test, we recorded the number of equilibrium losses (underbelly of a fish was visible); activity, indicated by the number of grid lines crossed; shoaling, indicated by a shoaling index (see below); number of burst swim events (sudden increase in velocity); total time spent frozen (seconds); and space use, indicated by a space use score (see below). Equilibrium losses, activity, burst events, and time spent frozen were all tracked continuously for each fish within a replicate trial (Table S3).

Shoaling activity and space use were scored every 15 s during behavioural tests as per previous experiments (Mathis and Smith 1993; Chivers and Smith 1995). Shoaling was scored using an index of shoal cohesiveness, where 1 = all fish were at least one body length away from their nearest neighbour and 5 = all five fish were within one body length of each other. Space use was assessed with the grid overlay divided into two zones of equal area encompassing the centre and outer perimeter of the mesocosm, with individual fish located in the outer zone scoring 1 and fish scoring 2 if they were in the middle zone (Fig. S2a). During food stimulus tests, a second space use score was assessed, where the grid overlay was divided into three equal areas and scores ranged from 1 for each fish in the farthest zone from the food injection site to 3 for each fish in the food injection zone (Fig. S2b). The scores for each measure at 15 s intervals were summed over the pre- and post-food stimulus periods for each test and then averaged by the number of fish present in the trial to give each trial period a final score. Final shoaling scores ranged from 0 to 1, where 0 indicates no time spent shoaling and 1 indicates the majority of the trial was spent shoaling. Space use (location) scores ranged from 0 to 2, where 0 indicates no time spent in the centre of the arena and 2 indicates the majority of the trial in the centre of the arena. Space use (location) scores in relation to a food stimulus ranged from 0 to 3, where 0 indicates no time spent near the food stimulus and 3 indicates the majority of the trial spent near the food stimulus. Behavioural data were extracted from the videos by an observer tracking an individual fish during each assessment, with corresponding exposure treatments unknown to minimize observer bias. To ensure behavioural observations were credited to the correct individual fish during assessments, each video was assessed multiple times, each time tracking and observing a different individual fish.

Concentrations of total NAFCs measured in water samples from NAFC-treated mesocosms were similar to nominal treatment levels (i.e., 2.1 ± 0.2 mg/L for the 2 mg/L treatment and 13.3 ± 5.1 mg/L for the 15 mg/L treatment; mean ± SD, n = 10–11/treatment) and were stable over time during the exposure phase (Fig. 2). These treatment concentrations are below those reported for fresh oil sands tailings (19.7–85 mg/L; Mahaffey and Dubé 2017) and are at the low end of the range of concentrations reported for reclamation ponds (5–40 mg/L; Anderson et al. 2012; Kavanagh et al. 2013) and bitumen-affected wetlands in Canada's oil sands region (3.3–72 mg/L; Vander Meulen et al. 2021a). Background concentrations of NAFCs in natural water in control mesocosms were 1.2 ± 0.3 mg/L (mean ± SD; n = 11), which is comparable to values we reported previously for natural lake water (∼0.5 mg/L) (Reynolds et al. 2022; Robinson et al. 2023) and reference wetlands in Canada's oil sands region unaffected by bitumen-derived inputs (Vander Meulen et al. 2021b). The background acid-extractable compounds that were detected in natural water included substantial amounts of z = 0 compounds (i.e., C_9, C₁₀, C₁₆, and C₁₈) that were likely fatty acids, as well as oxygen-rich formulae (e.g., O₃-, O₄-, O₅-, and O₆-containing formulae) reflective of ubiquitous humic and/or fulvic materials, similar to previous observations from Athabasca region surface waters observed with this method (Sun et al. 2017).

The NAFC extract used for our study was composed primarily of classical NAs (O₂ species; Fig. 3) with 15–16 carbon atoms and 2–4 double bond equivalents. Although classical NAs have been implicated as a primary organic contaminant in OSPW (Morandi et al. 2015; Hughes et al. 2017), recent work has demonstrated that classical NAs are not the only cause for concern. Formulae that are more abundant in aged bitumen-derived inputs, which include compounds with oxygen-rich and/or aromatic bonding configurations (Bauer et al. 2019b), such as hydroxylated aldehydes and other oxidative structural moieties, can also cause deleterious effects (Sun et al. 2017; Bauer et al. 2019a). The molecular composition of OSPW may also include substantial amounts of O₃- and O₄-NAFCs (Vander Meulen et al. 2021a). The NAFC extract in this case appears to be composed primarily of classical NAs (Fig. 3), so associated toxic outcomes should be attributed to this specific compound class.

Exposure to 15 mg/L NAFCs substantially reduced fish survival. Fish exposed to the 15 mg/L NAFC treatment had average survival rates of 80% during the baseline phase that decreased to 25% after 6 days of NAFC exposure (Fig. 4; Fig. S4). In contrast, during the baseline and exposure phases, control fish had average survival rates of 80% and 75%, respectively, and fish exposed to the 2 mg/L NAFC treatment had average survival rates of 85% and 70%, respectively. Over the entire experiment, survival varied with NAFC treatment (GLM, F_2,18 = 4.02, P = 0.036) and the treatment × exposure phase interaction (F_2,18 = 3.88, P = 0.039), which was driven by decreased survival in the 15 mg/L treatment compared to the other two groups but not between exposure phases (F_1,18 = 0.97, P = 0.34). Survival of yellow perch did not differ between treatments during the baseline phase (GLM, F_2,9 = 0.072, P = 0.93), but did differ significantly between NAFC concentrations during the exposure phase (GLM, F_2,9 = 7.25, P = 0.013), with the 15 mg/L treatment resulting in decreased survival (Fig. 4). However, K-M survival curves representing the entire experimental period identified significant differences between treatments (K-M; X²₍₂₎ = 14.3, P < 0.001; Fig. S4).

Our findings of substantially reduced yellow perch survival following exposure to 15 mg/L NAFCs provide important new context for the impacts of NAFCs on fish survival. A recent study on embryonic fathead minnows using the same NAFC extract reported an LC₅₀ value of 28.5 mg/L and a 10% benchmark dose value of 22.3 mg/L (Reynolds et al. 2022). Laboratory-based studies using NAFCs derived from the same settling basin (Industry A, but sampled in different years) and using the same extraction method as this study reported an EC₅₀ value for hatch success of 11 mg/L for walleye embryos and 23.2 mg/L for fathead minnow embryos (Marentette et al. 2015b). Another study using the same NAFC extract resulted in an LC₅₀ value of 32.8 mg/L for fathead minnow embryos and an LC₅₀ value of 51.8 mg/L for fathead minnow larvae (Kavanagh et al. 2012). Fathead minnow is a robust fish species with a higher tolerance to chemical exposure, whereas yellow perch tend to be more sensitive (Teather and Parrott 2006). A study involving young-of-the-year yellow perch reported 100% mortality in <96 h of exposure to 6.8 mg/L NAFCs isolated from surface waters of the West InPit (WIP) settling basin (Syncrude Canada Ltd.) (Nero et al. 2006). However, inconsistencies in source materials and chemical analysis methods may preclude direct comparisons of results between studies. Thus, our findings of reduced survival in yellow perch exposed to 15 mg/L NAFCs, a similar concentration to walleye LC₅₀ values and a much lower concentration than LC₅₀ values reported for fathead minnows using the same extract, highlight the differences in sensitivities between species. Comparing differences in sensitivities between fish species exposed to NAFCs allows for a more thorough understanding of how OSPW may affect aquatic communities. In addition, 15 mg/L NAFCs is at the low end of the range of concentrations reported for bitumen-affected wetlands in Canada's oil sands region (3.3–72 mg/L, Vander Meulen et al. 2021a). Our findings provide important information for establishing regulations for the safe release of NAFCs into aquatic ecosystems.

Total activity measured by the number of grid lines crossed during the observations tended to be greater in the post-stimulus periods than the pre-stimulus periods, with generally lower activity levels in the 15 mg/L treatment (Figs. 5d–5f). During the no-stimulus activity tests, there was no significant effect of treatment during the baseline phase on day −1 (GLM, F_2.9 = 0.53, P = 0.61) or during the exposure phase on day 5 (GLM, F_2,8 = 0.21, P = 0.82), although activity tended to be greater during the exposure phase (Fig. 5d). In the predator stimulus test, the number of grid lines crossed pre-predator stimulus did not vary significantly between NAFC concentrations (GLM, F_2,8 = 1.9, P = 0.21), although activity levels were lower in the 15 mg/L treatment compared to the others (Fig. 5e). During the post-predator stimulus period, there was also no overall significant effect of treatment (GLM, F_2,8 = 3.57, P = 0.078), but there was a clear stepwise pattern of lower activity level with increasing NAFC concentration (Fig. 5e). In the post-predator stimulus period of the predator stimulus test, fish crossed an average of 28 lines/min in the control treatment, 19 lines/min in the 2 mg/L NAFC treatment, and 5 lines/min in the 15 mg/L NAFC treatment. In the food stimulus test, the number of grid lines crossed did not vary significantly between treatments, either pre-food stimulus (GLM, F_2,8 = 0.21, P = 0.82) or post-food stimulus (GLM, F_2,8 = 0.71, P = 0.52), while the control treatment was associated with the highest observed activity levels (Fig. 5f). Our findings are consistent with Philibert et al. (2019), who reported decreased activity following an alarm cue stimulus consistent with an anti-predator response in fish exposed to OSPW during development. Similar decreases in activity and constrained anti-predator behaviours have been observed in fish exposed to crude oil (Barron et al. 2005; Kochhann et al. 2015; Lari et al. 2016) and polycyclic aromatic hydrocarbons (PACs) (Vignet et al. 2014; Vignet et al. 2015; Johansen et al. 2017).

The occurrence of equilibrium losses in yellow perch was significantly influenced by NAFC exposure in the three behavioural tests (Figs. 5a–5c), with fish in the 15 mg/L NAFC treatment being the only ones to lose equilibrium. In the no-stimulus activity test during the baseline phase, no equilibrium losses were observed, whereas fish in two out of three replicate mesocosms in the 15 mg/L treatment did lose equilibrium during the exposure phase (one twice and one seven times; Fig 5a), driving a significant effect of treatment in the exposure phase (GLM, F_2,8 = 10.79, P = 0.0053). In the predator stimulus test, equilibrium losses significantly increased with NAFC exposure during both the pre-predator stimulus and post-predator stimulus periods (GLM, both F_2,8 = 5.19, P = 0.036), although these equilibrium losses occurred in only one of the 15 mg/L NAFC replicate mesocosms (twice pre-predator stimulus and thrice post-predator stimulus; Fig. 5b). In the food stimulus test, equilibrium losses again occurred exclusively in the 15 mg/L treatment in both the pre-food stimulus (GLM, F_2,8 = 11.4, P = 0.0046) and post-food stimulus (GLM, F_2,8 = 5.19, P = 0.036) periods (Fig. 5c). Equilibrium losses have also been observed in juvenile pink salmon exposed to crude oil, which is notable as NAFCs are found in oil residues (Yang et al. 2021). Reynolds et al. (2022) observed similar equilibrium losses in recently hatched fathead minnows exposed to 21 and 29.5 mg/L of NAFCs, concentrations that were also associated with higher rates of mortality. One study suggests that loss of equilibrium in fish can serve as an alternate endpoint that can identify a larger portion of the population that is under stress and potential forthcoming mortality (Gosselin and Anderson 2020). If OSPW and NAFCs are released, fish in the Canadian oil sands region would not likely be statically exposed to contaminants as they were in this study, and thus it is relevant to understand both the immediate effects on survival and the changes in behaviour that decrease survival over time.

Our study experimentally manipulated the exposure of wild juvenile yellow perch to different concentrations of NAFCs in outdoor aquatic mesocosms to identify changes in fish survival and behaviour caused by NAFCs. Such survival and behavioural changes in response to NAFCs would have been impossible to isolate in a field study of wild fish in Canada's oil sands region. However, the changes in survival and behaviour we observed in yellow perch under experimental conditions could occur in wild fish under similar exposure scenarios in the oil sands region (e.g., in reclamation ponds constructed on former open-mine pits, in the case of an accidental breach of a tailings pond into a natural aquatic ecosystem, or possibly downstream of future outfalls releasing treated industrial effluent). If so, this could have implications for the health of fish populations. First, we observed a substantial reduction in survival in yellow perch exposed to 15 mg/L NAFCs. Previous studies spanning decades found exposure to OSPW in the range of 2–4 mg/L NAFCs has immunological and reproductive effects on yellow perch (van den Heuvel et al. 2012; Hogan et al. 2018). However, few studies have investigated the effects of NAFCs isolated from OSPW on yellow perch survival (Nero et al. 2006), and thus our findings provide important new information on the effects of NAFCs on yellow perch survival that can be compared to previous studies using NAFCs from the same settling basin and extraction methods (Kavanagh et al. 2012; Marentette et al. 2015b; Reynolds et al. 2022). Second, we observed that yellow perch exposed to 2 or 15 mg/L NAFCs changed fish activity in response to a predator stimulus. Changes in short-term activity can alter how fish encounter and respond to both resources and risks in the environment (Saaristo et al. 2018). Individuals that can quickly and accurately detect and assess risk are favoured by natural selection, and thus, disruption of a sensory or motor system due to contaminant exposure can have important implications for individual survival and fitness. Even small changes in fish survival can have a major impact on fish recruitment (Scott and Sloman 2004), and thus, increased susceptibility to predation, as a consequence of NAFC exposure could potentially lead to population declines. Third, we observed that yellow perch exposed to 15 mg/L NAFC experienced losses in equilibrium. If fish are unable to maintain equilibrium, this would affect their ability to forage for food, avoid predators, and reproduce with mates, among other normal activities critical for survival.