Geographic variation in behavioral avoidance of a widespread contaminant by Gulf killifish Fundulus grandis

Understanding the behavioral response of estuarine fishes to environmental contaminants is critical for assessing the ecological consequences of disturbances. Here, we examined geographic variation in avoidance behavior to petroleum-contaminated sediments in the Gulf killifish (Fundulus grandis), a sentinel species in northern Gulf of Mexico marshes. Adult fish were collected from four geographically distinct estuaries: Terrebonne Bay, Louisiana; Mississippi Sound, Alabama; Pensacola Bay, Florida; and Cedar Key, Florida. Choice experiments tested fish preference to either uncontaminated sediments or sediments contaminated with fresh or weathered oil. Results found significant differences among source locations with individuals from western Gulf sites (Louisiana and Alabama) exhibiting strong avoidance of fresh oil, while those from eastern sites (Florida) showed limited or no avoidance. No source location avoided weathered oil. These findings suggest that sensitivity to oil contamination varies regionally, potentially reflecting local exposure histories, environmental adaptation, or other sources of geographic variation. The diminished response to weathered oil across all locations highlights a reduced ability to detect residual contamination and indicates that these fish may use more volatile compounds that precipitate or degrade faster as a cue. This study underscores the importance of incorporating behavioral and geographic-level variability into impact assessments and emphasizes the need for regionally-tailored conservation strategies.

Estuarine ecosystems provide numerous ecosystem services, ranging from storm protection (Barbier, 2020) to nutrient reduction (Seitzinger, 1988) and enhanced production of commercially and recreationally important species (Beck et al., 2001). Salt marshes are widespread and vital estuarine habitats that facilitate energy transfer to adjacent coastal waters (Ziegler et al., 2021; Keppeler et al., 2024). Due to the highly dynamic conditions of estuaries, these systems often depend on a limited number of resident taxa (Day et al., 1989; Whitfield, 1994). The estuarine resident Gulf killifish (Fundulus grandis) is a mid-trophic level species essential for maintaining trophic connectivity and ecosystem resilience (Subrahmanyam and Drake, 1975; Rozas and LaSalle, 1990; Costanza et al., 1993; Tett et al., 2013). Fundulus grandis is a common and ecologically important resident of northern Gulf of Mexico salt marshes and has become a sentinel species for assessing environmental disturbance due to its abundance (Able et al., 2015), high site fidelity (Nelson et al., 2014; Vastano et al., 2017; Jensen et al., 2019), and central role in food webs (McCann et al., 2017; Oken et al., 2023). Its hardiness, ease of capture, and ability to thrive in captivity also make it an ideal species for laboratory studies of environmental stressors (Whitehead et al., 2011; Dubansky et al., 2013, 2017).

Populations of estuarine fishes often experience localized environmental conditions and stressors that can shape physiological tolerance and behavioral traits, including responses to contaminants. Previous studies have shown that population-level differences in sensitivity to environmental toxins can arise from both genetic differentiation and phenotypic plasticity (Whitehead et al., 2011; Schlenker et al., 2019). For F. grandis, such spatial variability could result in differential susceptibility to hydrocarbon exposure, depending on historical oiling, contaminant loads, or other co-occurring stressors such as hypoxia and salinity fluctuations. As a result, individuals from different areas may vary in their capacity to detect and avoid contaminated sediments, with potential consequences for habitat use, exposure risk, and resilience to oiling events (Fodrie et al., 2014). Understanding this variability is crucial for evaluating how widespread disturbances like oil spills affect estuarine species at scales relevant to conservation and management.

The 2010 Deepwater Horizon (DwH) oil spill released approximately 4.9 million barrels of crude oil into the northern Gulf, impacting over 1,700 km of shoreline, predominantly saltmarshes (McNutt et al., 2012; Michel et al., 2013; Nixon et al., 2016). Initial predictions anticipated severe ecological disruption, particularly to nearshore fish communities. However, subsequent field surveys often found minimal to no long-term population-level impacts on coastal fishes (Fodrie and Heck, 2011; Moody et al., 2013; Fodrie et al., 2014; Able et al., 2015; Schaefer et al., 2016; Lewis et al., 2022). This apparent resilience has been attributed to several factors, including behavioral avoidance of contaminated habitats, patchy oil distribution, and rapid recolonization from unaffected areas (Fodrie et al., 2014; Martin, 2017; Martin et al., 2020a). Laboratory experiments, however, have consistently documented sublethal physiological and behavioral effects of oil exposure on estuarine fishes, including reduced foraging efficiency, impaired locomotion, and altered olfactory responses (Claireaux et al., 2004; Whitehead et al., 2011; Dubansky et al., 2013; McDonald et al., 2022). Specifically, F. grandis demonstrates strong avoidance of fresh oil-contaminated sediments, but this response is diminished with weathered oil or after prior exposure to oil, suggesting potential impairment of sensory detection mechanisms (Martin, 2017; Martin et al., 2020b). These behavioral alterations may impact habitat selection, foraging success, and ultimately, fitness.

Despite the growing body of evidence for behavioral responses to oil exposure, little is known about intraspecific variation in these responses across geographically or environmentally distinct areas. Given the extensive spatial heterogeneity in oiling during the DwH spill, during which Louisiana, Mississippi, and Alabama marshes received substantially greater shoreline oiling than many portions of the eastern Gulf (Michel et al., 2013), as well as ongoing regional differences in oil and gas extraction, transport, and refining activities, it is critical to examine whether geographic variability exists in their ability to detect and avoid oil-contaminated habitats. Such variation could have significant implications for population resilience and the persistence of ecological functions in impacted areas. In this study, we experimentally test the behavioral responses of F. grandis from various areas throughout the Gulf to fresh and weathered oil-contaminated sediments at a concentration previously shown to induce behavioral avoidance. We hypothesize that (1) fish will exhibit avoidance of fresh oil-contaminated sediments relative to uncontaminated controls or weathered oil-contaminated sediments, and (2) the strength of avoidance behavior will vary among fish origins. Such variation could reflect differences in historical exposure to hydrocarbons, local environmental conditions, developmental effects, or pre-existing geographic variation. Because these mechanisms are not directly tested here, our goal is to identify whether behavioral differences exist among fishes from different areas rather than determining an underlying cause.

Fish collection

Adult (>65mm total length) Gulf killifish (F. grandis) were collected from four geographically distinct estuarine locations along the northern Gulf of Mexico: Terrebonne Bay, Louisiana (TB); Mississippi Sound, Alabama (MS); Pensacola Bay, Florida (PB); and Cedar Key, Florida (CK) (Figure 1A). Source locations were selected to span a broad longitudinal gradient across the northern Gulf, while staying within the species’ range. These locations also differ in their histories of oil exposure associated with both the DwH spill (Michel et al., 2013) and broader regional oil and gas activities. The objective was to determine whether behavioral responses to oil differed among fish collected from geographically-separated estuaries representing different environmental histories. All collection locations contained similar salinity regimes typical of northern Gulf of Mexico estuaries, ranging from 5–10 psu.

At each site, fish were captured using baited minnow traps placed along the marsh edge in shallow intertidal zones (Able et al., 2015; Vastano et al., 2017). After capture, fish were transported to the laboratory and housed in aerated, flow-through holding systems maintained at salinity levels consistent with collection conditions. Because salinity can influence physiology and behavior in estuarine fishes, all individuals were acclimated to a common salinity of 8 psu prior to experimentation. To minimize potential artifacts associated with capture stress and handling, all fish were held for a minimum of one month prior to behavioral testing. During this acclimation period, fish were fed locally collected grass shrimp (Palaemonetes spp.) ad libitum daily (McDonald et al., 2022) and monitored for health (lesions, injuries, etc.) and abnormal behavior. Only healthy fish were used in trials and mortality was <10% during the acclimation period. Collection and use of fish in experiments was conducted with IACUC approval #201710044.

Experimental trials

Behavioral avoidance trials followed the protocol described in Martin (2017). Trials were conducted in 38-liter aquaria (51 cm long x 27 cm wide x 32 cm tall) divided into two equal zones: one side containing clean sediment and the other side containing sediment mixed with oil. Approximately 3 L of sediment (1.5 L per side) was added to each tank and leveled to a uniform depth of ~18 mm. For choices involving oil, the oil was thoroughly mixed with the sediment prior to placement and then covered with ~2 cm of clean sediment to simulate shallow burial. This approach was used to prevent oil from floating to the sediment surface during trials and to maintain consistency with previous Fundulus avoidance studies (Martin, 2017; Martin et al., 2020b). Although burial may reduce diffusion of some volatile compounds, all treatments were subjected to identical sediment layering procedures, allowing direct comparison among treatments. For all trials, we used commercially available, rinsed play sand as the sediment source. This standardized substrate ensured that sediment characteristics were consistent across all experimental units and prevented any fish from a particular source location from encountering sediment that might resemble their native marsh habitat. All fish were starved for 24 hours prior to trials. Two oil treatments were tested: fresh oil and weathered oil, both applied at a concentration of 40 mL oil L⁻¹ sediment. This concentration was selected because previous studies demonstrated consistent behavioral responses at this exposure level (Martin, 2017; Martin et al., 2020b). Weathered oil was produced by aerating fresh crude in a fume hood to achieve 40% mass loss, mimicking conditions typical of oil that had undergone environmental weathering prior to marsh deposition (specific sediment and aromatic concentrations are reported in Martin, 2017).

For each trial, a single fish was introduced to the center of the tank and allowed to acclimate for five minutes. This duration was selected based on pilot trials in which individuals resumed normal swimming behavior (e.g., active exploration, absence of freezing or startle responses) within this time frame, and is consistent with acclimation periods used in previous Fundulus behavioral studies. Following acclimation, fish behavior was recorded using a GoPro digital camera for a 10-minute observation period. Fish position was recorded every 30 seconds, resulting in a total of 20 observations per trial. Occupancy was quantified as the proportion of observations in which a fish was located over oiled sediment. A value of 0.5 indicates no preference between oiled and unoiled habitats, values less than 0.5 indicate avoidance of oiled sediment, and values greater than 0.5 indicate preference for oiled sediment. For statistical analyses, the number of observations over oiled sediment and the number of observations over unoiled sediment were retained as binomial count data.

Side assignment (left or right) for oiled sediment was randomized across trials to control for any side bias. In addition to the oil treatments, we conducted clean:clean control trials in which both sides of the aquarium contained clean sediment processed identically but without oil. Fish showed no significant side preference in these control trials (p > 0.05), confirming that tank side did not influence occupancy. These controls were excluded from the main text for brevity but validated the absence of inherent side bias. Behavioral responses were analyzed using a generalized linear model with a binomial error distribution and logit link function. For each trial, the number of observations in which the fish was located over the oiled sediment was treated as the number of successes, while observations over the unoiled sediment were treated as failures. Source location, oil treatment (fresh vs. weathered), and their interaction were included as fixed effects. This approach preserves the factorial structure of the experimental design and avoids transformation of proportional occupancy data. Model fit was evaluated using residual deviance and Pearson residuals, and overdispersion-adjusted inference was used when appropriate. Each unique source location * oil treatment was replicated 12 times, and no single fish was used in multiple trials.

A binomial generalized linear model revealed a significant source location × oil treatment interaction on the probability of fish occupying the oiled side of the tank (χ² = 32.62, df = 3, p < 0.001), indicating that behavioral responses to oil differed among source locations and between fresh and weathered oil treatments. Fish were also less likely to occupy fresh oil sediments than weathered oil sediments overall (χ² = 40.41, df = 1, p < 0.001) (Figure 1B), and occupancy differed among source locations (χ² = 8.38, df = 3, p = 0.039). Because model diagnostics indicated moderate overdispersion, inference was evaluated using overdispersion-adjusted standard errors; the source location × oil treatment interaction remained significant.

Fish collected from Mississippi Sound (MS) and Terrebonne Bay (TB) exhibited the strongest avoidance of fresh oil, occupying the oiled side during only 23 ± 5% and 25 ± 4% of observations, respectively, compared to the 50% occupancy expected under random habitat use. In contrast, fish from Pensacola Bay (PB) and Cedar Key (CK) occupied the oiled side during 38 ± 3% and 45 ± 5% of observations, respectively, indicating substantially weaker avoidance. Occupancy of weathered oil sediments was generally near 50% across all source locations (MS: 49 ± 5%, TB: 52 ± 7%, PB: 43 ± 5%, CK: 45 ± 5%), suggesting little evidence of avoidance. Collectively, these results indicate that avoidance behavior was strongest in fish collected from western Gulf source locations when exposed to fresh oil, whereas weathered oil elicited little evidence of avoidance regardless of source location.

Our results demonstrate location-specific sensitivity to oil contamination, with western Gulf fish (MS, TB) displaying stronger avoidance responses than eastern fish (PB, CK). These findings reveal previously undocumented geographic variation in behavioral responses to oil contamination within F. grandis. While this pattern is consistent with differences in exposure history, local environmental conditions, developmental effects, or adaptive processes, our study was not designed to distinguish among these alternative explanations. This geographic variability in behavioral sensitivity suggests that specific traits, potentially influenced by multiple generations and natural selection to local environmental histories or potentially differential exposure to contaminants, may influence ecological resilience following oil spills. Additionally, the consistent lack of avoidance across all source locations to weathered oil highlights a reduction in perceived risk or detectability of hydrocarbons after environmental degradation, with important implications for exposure risk in marshes where weathered oil persists long after acute spill events.

The observed west-to-east gradient in avoidance responses is consistent with regional differences in historical hydrocarbon exposure (Michel et al., 2013), although our study was not designed to directly test this mechanism. It is important to emphasize that we did not directly measure contamination history, hydrocarbon concentrations, biomarkers of exposure, or other indicators that would allow this mechanism to be evaluated. Consequently, any connection between historical oil exposure and the behavioral patterns observed here should be considered speculative. The northern Gulf experienced substantial spatial variation in oiling during the DwH spill, with Louisiana, Mississippi, and Alabama marshes receiving considerably greater shoreline oiling than many areas farther east (Michel et al., 2013). In addition, western Gulf estuaries have historically experienced greater oil and gas extraction, transportation, and refining activity. These regional differences make exposure history a plausible hypothesis, but alternative explanations, including pre-existing geographic variation, local environmental conditions, developmental effects, or other unmeasured ecological factors, remain equally possible. Outside of major spills such as the Exxon Valdez (Peterson et al., 2003), Ixtoc (Soto et al., 2014), Arthur Kill (Burger, 1994), and Falmouth Harbor (Reddy et al., 2002), smaller spills are frequent - especially in the northern Gulf of Mexico. The United States Coast Guard National Response Center has over 20,000 spills reported annually (Martin et al., 2023) and in Louisiana alone the Louisiana Oil Spill Coordinator’s Office (LOSCO) responded to 44 spills in 2024 (https://data.losco.org). Previous exposure to hydrocarbons has been shown to alter olfactory detection, risk assessment, and overall behavioral responses in fishes (Martin et al., 2020b; Schlenker et al., 2019). Thus, fish from heavily oiled regions or areas with frequent leaks/spills, which likely experience sublethal exposures, may have retained or developed stronger avoidance behavior as an adaptive or conditioned response. Alternatively, these differences could arise from local adaptation or phenotypic expression driven by unrelated environmental variability (e.g., salinity, hypoxia, or anthropogenic pollutants; Hamilton et al., 2016), which may influence the development and expression of avoidance behavior across source locations. We note that these regional differences in historical hydrocarbon inputs may help explain the observed geographic variation in avoidance behavior, but we present this solely as a speculative hypothesis rather than a causal process as we only assessed four areas in this study.

An important consideration is that the present findings appear to contrast with our previous work showing that prior laboratory exposure to oil reduced avoidance behavior in F. grandis (Martin et al., 2020b). In that study, fish experimentally exposed to oil subsequently lost their preference for uncontaminated sediments, suggesting impairment of sensory or behavioral mechanisms involved in oil detection. In contrast, fish collected from western Gulf in the present study exhibited stronger avoidance of fresh oil than fish from the eastern Gulf. We propose that these patterns are not necessarily contradictory because they operate at different biological scales. The previous study examined short-term, individual-level responses following controlled laboratory exposure, whereas the present study compared wild fish that may differ due to long-term environmental histories, developmental experiences, genetic variation, or other ecological factors. One possibility is that acute exposure impairs avoidance behavior within individuals while fish inhabiting regions with greater historical hydrocarbon inputs have, over longer timescales, developed stronger baseline avoidance responses through selection, learning, or other mechanisms. Alternatively, the observed geographic variation may predate major oiling events and be unrelated to contaminant history. Distinguishing among these possibilities will require future studies integrating behavioral assays with exposure histories, population genetics, and common-garden experiments.

Our findings align with earlier work demonstrating that F. grandis avoid heavily oiled sediments in laboratory trials (Martin, 2017; Martin et al., 2020b), but add a new dimension by revealing that such responses are not uniform across the species’ range. This variation has important ecological implications: if fish from certain geographic areas are less likely to detect and avoid contamination, they may be more susceptible to sublethal effects on foraging, reproduction, and predator avoidance (Rice et al., 1987; Khursigara et al., 2021), thereby enhancing population- or community-level effects. Furthermore, source locations with reduced avoidance responses may be at greater risk of prolonged exposure to harmful compounds, even in weathered oil conditions that do not elicit immediate behavioral aversion but can still impair physiological function (Whitehead et al., 2011).

The observed lack of avoidance to weathered oil may yet have significant implications for F. grandis populations, given that weathering increases the proportion of high molecular weight polycyclic aromatic hydrocarbons (PAHs) in crude oil (Di Toro et al., 2007). Weathering is also associated with the preferential loss of lower molecular weight and more volatile hydrocarbons, including naphthalenes and other light aromatic compounds that readily evaporate or degrade following environmental exposure. Chemical characterization of the same oil source and weathering protocol used in this study demonstrated substantial reductions in these volatile fractions following weathering (Martin, 2017). Although we did not directly quantify the specific compounds responsible for avoidance behavior, the consistent lack of avoidance observed across source locations suggests that fish may rely on chemical cues associated with these more volatile components when detecting contaminated habitats. Their depletion during weathering could therefore reduce the ability of fish to recognize and avoid oiled sediments, even when residual contamination remains present. Most studies investigating toxicity of weathered versus fresh oil have focused on early life stages, finding weathered oil more harmful to fish embryos and larvae than fresh oil (Incardona et al., 2014; Esbaugh et al., 2016; O’Shaughnessy et al., 2018); but these results may still be relevant to later life stages. Although they did not compare to fresh oil, Lee et al. (2018) found significant immune dysfunction of rockfish (Sebastes schlegeli) after oral administration of weathered Iranian heavy crude oil. It might be that adults, however, are less susceptible to these effects or have escaped effects completely, possibly due to developmental stage-specific sensitivity of weathered oil.

Several caveats should be considered when interpreting these results. Laboratory avoidance assays necessarily simplify the structural, chemical, and ecological complexity of natural marshes, including the absence of vegetation, conspecifics, and hydrodynamic processes. However, laboratory experiments provide greater control over environmental conditions than field studies, allowing behavioral responses to be isolated from numerous co-occurring factors (Diamond, 1986). Although aquaria may constrain movement relative to natural habitats, this design has been widely used in previous Fundulus behavioral studies and was appropriate for assessing relative habitat occupancy. Additionally, acclimation under common laboratory conditions may have influenced behavior, and our occupancy-based metric does not capture more subtle responses such as changes in activity, ventilation, or exploratory behavior. While common-garden acclimation reduced short-term environmental variation among source locations, it may also have diminished some environmentally induced physiological differences present at the time of collection. An additional consideration is that we did not evaluate population structure or connectivity among source locations. Consequently, the behavioral differences observed among source locations should not be interpreted as evidence of genetically distinct or isolated populations, although previous studies have occasionally used collection location as a proxy for population identity (e.g., Reynolds et al., 2021). Rather, these findings indicate that fish collected from geographically separated estuaries differed in their responses to oil exposure. Future studies integrating behavioral assays with population genetic approaches would help determine whether these patterns reflect population structure, environmental history, developmental effects, or other mechanisms. Although population connectivity was not evaluated directly in this study, previous tagging studies have demonstrated high site fidelity in F. grandis, suggesting that individuals often remain associated with local estuarine habitats over extended periods (Vastano et al., 2017; Jensen et al., 2019; Nelson et al., 2014). Finally, these experiments assessed short-term behavioral responses to acute exposure; longer-term studies may reveal delayed or cumulative effects that are not detectable during brief laboratory trials (e.g., O’Shaughnessy et al., 2018; Vignet et al., 2019). Future work integrating behavioral assays with independently verified exposure histories, physiological measurements, and field observations would help clarify the mechanisms underlying the geographic variation observed here.

Overall, this study underscores the importance of incorporating geographic-level variation into assessments of oil spill impacts. Behavioral traits such as habitat avoidance may buffer some populations against acute contamination, while others may face higher exposure risk due to diminished sensitivity. Future studies should investigate the mechanistic basis of these differences, whether genetic, developmental, or experiential, and explore how they interact with other environmental stressors. Understanding this variability will be essential for predicting estuarine fish responses to future disturbances and informing site-specific conservation or restoration strategies.