Seasonal, environmental and individual effects on hypoxia tolerance of eastern sand darter (Ammocrypta pellucida)

Abstract

Metabolic rate and hypoxia tolerance are highly variable among individual fish in a stable environment. Understanding the variability of these measures in wild fish populations is critical for assessing adaptive potential and determining local extinction risks as a result of climate-induced fluctuations in temperature and hypoxic conditions. We assessed the field metabolic rate (FMR) and two hypoxia tolerance metrics, oxygen pressure at loss of equilibrium (PO2 at LOE) and critical oxygen tolerance (P_crit) of wild-captured eastern sand darter (Ammocrypta pellucida), a threatened species in Canada, using field trials (June to October) that encompassed ambient water temperatures and oxygen conditions typically experienced by the species. Temperature was significantly and positively related to hypoxia tolerance but not FMR. Temperature alone explained 1%, 31% and 7% of the variability observed in FMR, LOE, and P_crit, respectively. Environmental and fish-specific factors such as reproductive season and condition explained much of the residual variation. Reproductive season significantly affected FMR by increasing it by 159–176% over the tested temperature range. Further understanding the impact of reproductive season on metabolic rate over a temperature range is crucial for understanding how climate change could impact species fitness. Among-individual variation in FMR significantly increased with temperature while among-individual variation in both hypoxia tolerance metrics did not. A large degree of variation in FMR in the summer might allow for evolutionary rescue with increasing mean and variance of global temperatures. Findings suggest that temperature may be a weak predictor in a field setting where biotic and abiotic factors can act concurrently on variables that affect physiological tolerance.

Introduction

Species extinctions have increased in recent years due to anthropogenic effects and are predicted to further increase with climate change (Burkhead, 2012; Swan et al., 2018). In particular, freshwater organisms face heightened extinction risks because they are susceptible to modification of both terrestrial and aquatic landscapes (Ricciardi and Rasmussen, 1999; Jenkins, 2003; Olden et al., 2007). Anthropogenic factors such as deforestation, agricultural-driven nutrient and sediment loading, urbanization and climate change (Donohue and Garcia Molinos, 2009; IPCC, 2021) can all lead to increased water temperature fluctuations and the prevalence of hypoxic zones (low oxygen), both of which are known habitat stressors in aquatic ecosystems (Diaz, 2001; Diaz and Rosenberg, 2008; IPCC, 2013). Increases in temperature may induce behavioural changes that help avoid unfavourable conditions and can trigger population-level changes in abundance caused by temperature-dependent changes in mortality, growth and/or reproduction (Pörtner and Peck, 2010). Similarly, dissolved oxygen reductions can result in aquatic organisms having an oxygen deficiency in their tissues, which can lead to changes in locomotory activity, growth, reproduction and alter the dynamics of predator–prey interactions due to physiological and behavioural compensation (Robb and Abrahams, 2003; Landry et al., 2007; Pollock et al., 2007; Oldham et al., 2019). Therefore, understanding an organism’s metabolic rate and hypoxia tolerance in relation to changing temperature is necessary for determining the consequences of hypoxia on species survival.

Because fishes are ectotherms, temperature is the key factor controlling metabolic rate (Fry, 1947; McNab, 2002; Schulte et al., 2011). Measures of metabolic rate and hypoxia tolerance (critical oxygen tolerance and LOE) are correlated and generally increase with environmental temperature (Fry, 1947; Hlohowskyj and Wissing, 1987; McBryan et al., 2013; Rogers et al., 2016). However, some species have demonstrated metabolic compensation, allowing them to completely or partially maintain metabolic rate over a range of temperatures (Precht, 1958; Vetter, 1982; Sandblom et al., 2014; Abe et al., 2019). Metabolic compensation could be beneficial in the wild where individuals are exposed to fluctuating temperatures and would allow species to conserve energy and/or remain active in situations when temperatures impact locomotion, growth and predator–prey interactions (Fry, 1971; Vetter, 1982; Evans, 1990; Abe et al., 2019). Despite the importance of these principles, there remains a lack of information about metabolic rate and hypoxia tolerance in the wild where species face temperature changes on diurnal and seasonal cycles (Treberg et al., 2016).

Metabolic rate and hypoxia tolerances are also influenced by factors other than temperature (Metcalfe et al., 2016; Rogers et al., 2016), including food availability (Auer et al., 2015), salinity (Allan et al., 2006), turbidity (Hess et al., 2017) and fish mass (Clarke and Johnston, 1999). Under constant environmental conditions, and for fish of similar mass and sex, both routine metabolic rate and hypoxia tolerance show a large degree of individual variability (Burton et al., 2011; Metcalfe et al., 2016). Some of the among-individual variability can be explained by liver enzyme activity (Norin and Malte, 2012), social behaviour (Metcalfe et al., 1995) and cardiac performance (Joyce et al., 2016). Norin et al. (2016) showed a change in among-individual variability of standard metabolic rate (SMR) in barramundi (Lates calcarifer) in response to low salinity, high temperature and hypoxia. Low SMR-control individuals displayed a larger increase in metabolic rate at a higher temperature than high SMR-control individuals, leading to a decrease in among-individual variability at hotter temperatures (Norin et al., 2016). Such findings suggest that even within a species, fish are not uniformly impacted by changes in temperature for a given set of environmental conditions. At the population-level, among-individual variation in metabolic responses to temperature change implies some intrinsic buffering capacity, at least to low or moderate changes in environmental conditions. However, few studies have assessed among-individual variability in metabolic rate and hypoxia tolerance under natural conditions, raising questions regarding how fishes of conservation concern may respond to hypoxia and thermal stress.

Despite knowledge of how temperature and other abiotic and biotic variables influence metabolic rate and hypoxia tolerance for some freshwater fish species (e.g. Fry, 1947 ; Clarke and Johnston, 1999 ; Metcalfe et al., 2016 ; Rogers et al., 2016), there is a paucity of information about physiological responses under natural conditions where fish experience dynamically changing environments (Treberg et al., 2016). In the wild, fish experience changes in reproduction, food availability, and abiotic conditions (e.g. temperature, dissolved oxygen) simultaneously across the seasons. Additionally, transgenerational impacts (e.g. grandparent and maternal/paternal effects) and developmental plasticity can also influence physiological metrics (Schaefer and Walters, 2010; Le Roy et al., 2017). As both the mean and variance of water temperatures are expected to increase with climate change (Räisänen, 2002), investigating the effect of temperature-driven changes in oxygen availability and related fish responses in the wild will be important for understanding physiological plasticity, variation and determining local extinction risk.

To date, few studies have investigated the metabolic rate and hypoxia tolerance for fishes of conservation concern, yet knowledge of individual and species-specific responses will have important implications for evaluating threats and habitat suitability. We investigate these concepts for eastern sand darter (Ammocrypta pellucida), which is a small, benthic fish listed as threatened in Canada under the Species at Risk Act and is imperilled throughout its North American range (Grandmaison et al., 2004; Fisheries and Oceans Canada, 2012). Eastern sand darter is a habitat specialist that prefers clean sand and fine gravel substrate in rivers (Dextrase et al., 2014) and some lake areas. Numerous anthropogenic stressors have been associated with the decline of eastern sand darter, with siltation of preferred habitat suggested as a major source of population decline (COSEWIC, 2009). Siltation is thought to be linked to decreased oxygen availability in the substrate, which may lead to reduced burying behaviour and increased energy expenditure (Daniels, 1989; Drake et al., 2008; COSEWIC, 2009). However, like many imperilled fishes, there is uncertainty about the physiological response of eastern sand darter to habitat stressors, including its sensitivity to oxygen impairment. To overcome this limitation, the overall objective of this study was to determine how the hypoxia tolerance of eastern sand darter varied in response to seasonal temperature changes while accounting for the role of environmental and fish-specific factors (e.g. condition, reproductive season). Specifically, the study objectives were to determine if: (i) seasonal temperature influenced hypoxia tolerance (PO2 at LOE and P_crit) and metabolic rate in eastern sand darter; (ii) the inclusion of environmental turbidity, fish condition and reproductive season variables significantly improved the ability to explain residual variation in hypoxia tolerance and metabolic rate; and, (iii) seasonal temperature influenced among-individual variation in hypoxia tolerance (PO2 at LOE and P_crit) and metabolic rate.

Materials and Methods

Field sampling

Fish capture methods followed those described in detail in Firth et al. (2021). Briefly, eastern sand darter were collected monthly in the Grand River at Cockshutt Bridge, Brantford, Ontario (43.110°N, 80.245°W) from June to October 2019 (for a detailed description, see Gáspárdy and Drake, 2021) using an 9.14 m bag seine (1.8 x 1.8 x 1.8 m bag with 3.0-mm mesh) hauled downstream in sand-dominated depositional areas. All animal use and collection was conducted under a Species at Risk Act permit (#19-PCAA-00021) and approved by the University of Waterloo (AUP #40909) and Fisheries and Oceans Canada (AUP #1968/2068) Animal Care Committees. Field collections focused on adults, with all tested individuals measuring >50 mm in total length. As digestion impacts metabolism (i.e. specific dynamic action; SDA), fish are commonly fasted before testing as a means of allowing them to attain their routine/SMR (Jobling, 1993; Secor, 2009). Field-based testing, however, precluded the use of fasting. To reduce the effects of SDA on metabolic calculations in this study, all fish were collected before 10 am (except for three days in October when fish were collected until 12 pm). Darter species generally display diurnal feeding (e.g. Schenck and Whiteside, 1977; Chalupnicki and Johnson, 2016), with studies on least darter (Etheostoma microperca), banded darter (Etheostoma zonale; Cordes and Page, 1980), rainbow darter (Etheostoma caeruleum), fantail darter (Etheostoma flabellare; Adamson and Wissing, 1977) and blackbanded darter (Percina nigrofasciata; Mathur, 1973) demonstrating increased feeding activity during the day and decreased feeding after sunset. Studies have also reported the stomachs of darter species to be mostly empty in the morning (Mathur, 1973; Adamson and Wissing, 1977; Cordes and Page, 1980; Chalupnicki and Johnson, 2016).

Following capture, fish were held for 30 minutes to 4 hours prior to experimentation in a flow-through bin (56.5 × 39.3 × 30.4 cm), which was supplied with river-sourced water to ensure that field-related temperature acclimation remained consistent with natural river conditions.

Habitat measurements

At the center of each sampling site, water temperature (°C), dissolved oxygen (mg/L), conductivity (μS/cm), turbidity (NTU) and pH were collected ~0.2 m below the water surface using a YSI EX02 multiparameter sonde following Firth et al. (2021) and Gáspárdy and Drake (2021) (Figure 1).

Figure 1

Figure based on Firth et al. (2021) showing water temperature (°C; red) and dissolved oxygen (mg/L; black, data from Grand River Conservation Authority) taken over the sampling season in 2019. Red and Black dots represent a bi-weekly mean of the corresponding environmental variables. Blue dots represent single measurements of dissolved oxygen taken at the study site.

Respirometry protocol

There were three measurement endpoints in this study: field metabolic rate (FMR), critical oxygen tension (P_crit) and oxygen tension at loss of equilibrium (PO2 at LOE). FMR was defined as the baseline oxygen requirement for living of a post-absorptive, inactive fish plus the additional cost of maintaining posture and position in the water column and any additional unknown cost of specific dynamic action (Treberg et al., 2016) or postprandial increase in oxygen uptake. FMR was considered as the energy expenditure of a wild fish in natural conditions (Treberg et al., 2016). Thus, FMR can provide an improved understanding of natural environmental disturbances (e.g. thermal fluctuations, seasons, oxygen availability) on fish performance (Treberg et al., 2016). The P_crit and PO2 at LOE endpoints represent different aspects of an organism’s response to oxygen availability (Ultsch et al., 1978; Rogers et al., 2016). Critical oxygen tension being the point at which a species can no longer sustain oxyregulation (i.e. maintain oxygen consumption to support metabolic rate) in the face of decreasing environmental oxygen levels, and switches to oxyconforming where oxygen consumption is decreased in concert with decreasing environmental oxygen levels (Pörtner and Grieshaber, 1993). A lower P_crit suggests greater hypoxia tolerance as fish can maintain aerobic metabolism, and thus routine function, for a longer period in environments with decreased oxygen availability (Regan and Richards, 2017). In contrast, PO2 at LOE signifies the oxygen concentration at which delivery of oxygen to aerobic tissues can no longer be sustained. At PO2 at LOE a fish loses the ability to maintain its position in the water column and is likely to suffer mortality unless dissolved oxygen in the water increases (Fry, 1971; McBryan et al., 2013; Franklin, 2014; Rees and Matute, 2018).

Metabolic rate and hypoxia tolerance were measured streamside using closed respirometry. Each respirometry trial consisted of housing fish in an apparatus that included four chambers in a bin separated by barriers to prevent fish from seeing each other and to isolate them from external disturbances. Individual fish were placed in 60 ml cylindrical Plexi-glass chambers (length, 70 mm; diameter, 32 mm) with the ratio of fish mass to water volume averaging 1:42 (minimum 1:31 to maximum 1:85). Plexi-glass chambers were carefully inspected pre- and post-experimentation to ensure there was no leak or extra off gassing of oxygen (Stevens, 1992). A continuous flow of river water was provided to the bin and pumped through vinyl tubing to each chamber to maintain water temperature and oxygen levels throughout a ninety-minute acclimation period. Tests using rainbow darter have shown that a 90 minute acclimation period is sufficient to restore metabolic rate to routine levels after initial capture and handling (Mehdi et al., 2018; Hodgson et al., 2020). After acclimation, the respirometers were closed, and the oxygen was depleted in the chamber by fish respiration. Trial duration lasted 93 minutes on average (range: 27–201 minutes). Following Ultsch et al. (1978), Hlohowskyj and Wissing (1987), De Boeck et al. (2013) and Craig et al. (2014) chamber mixing during respirometry was achieved by fish movement and ventilation. During the entire trial, river water was continuously pumped into the holding bin to maintain ambient river water temperature. In all trials, temperature stayed within 1.5°C of the measured start temperature.

The rate of oxygen depletion was measured using fiber-optic oxygen sensors placed in the center of each respirometer (FireStingO₂, Pyro Science, Aachen, Germany), with one chamber left empty during every trial to measure background rates of biological oxygen demand. Oxygen concentration in the chambers was measured every second using Pyro Oxygen Logger Software (Pyro Science, Aachen, Germany) and a temperature probe in the bin allowed for software temperature compensation during recording. Recalibration of both oxygen (0 and 100%) and temperature probes occurred between each monthly set of trials. The rate of oxygen consumption was measured until LOE occurred, which was denoted by inverted swimming and operculum muscle spasms (Lutterschmidt and Hutchison, 1997). After LOE, the fish was immediately placed in a recovery bin with 100% oxygen saturation. Once the fish recovered, fish total length (mm) and mass (g) were measured (Table 1) and the fish was tagged with a visible implant elastomer tag (Northwest Marine Technology, Inc.) to prevent re-use in subsequent trials.

Sample size (n), mean ± standard error of fish mass (g), total length (mm), river temperature (°C), dissolved oxygen concentration (mg/L), measures of FMR (umol/hr-g), oxygen pressure at loss of equilibrium (PO2 at LOE; mg/L) and critical oxygen tension (P_crit; kpa) for eastern sand darter sampled in the Grand River, summarized by sample month

Table 1

Sample size (n), mean ± standard error of fish mass (g), total length (mm), river temperature (°C), dissolved oxygen concentration (mg/L), measures of FMR (umol/hr-g), oxygen pressure at loss of equilibrium (PO2 at LOE; mg/L) and critical oxygen tension (P_crit; kpa) for eastern sand darter sampled in the Grand River, summarized by sample month

Month (# days sampled)	n	Mass (g)	Total Length (mm)	Temperature (°C)	Dissolved Oxygen (mg/L)	FMR (umol/h-g)	PO2 at LOE (mg/L)	Pcrit (kpa)
June (4)	18	1.38 ± 0.054	63.06 ± 0.63	16.43 ± 0.50	8.88 ± 0.04	20.49 ± 1.15	0.64 ± 0.056	8.39 ± 0.93
July (4)	15	1.19 ± 0.096	58.40 ± 1.44	25.28 ± 0.26	6.75 ± 0.11	16.42 ± 1.57	1.14 ± 0.042	13.31 ± 1.52
August (3)	15	1.48 ± 0.048	63.73 ± 0.59	24.45 ± 0.24	6.95 ± 0.14	18.50 ± 3.04	0.89 ± 0.091	11.09 ± 1.59
September (4)	15	1.51 ± 0.046	62.40 ± 0.69	19.64 ± 0.28	8.50 ± 0.12	12.79 ± 1.57	0.81 ± 0.082	10.61 ± 1.61
Sept/Oct (3)	15	1.41 ± 0.054	63.47 ± 0.87	18.90 ± 0.32	8.39 ± 0.07	12.45 ± 1.55	0.63 ± 0.045	11.34 ± 1.05
October (4)	12	1.35 ± 0.053	61.83 ± 0.78	13.01 ± 0.29	11.12 ± 0.35	10.67 ± 0.89	0.50 ± 0.067	9.64 ± 1.06

The oxygen consumption rate was calculated over 5-minute intervals with the R package ‘respirometry’ (Birk, 2019) using the equation provided in (Rogers et al., 2016):

MO₂ = [(ΔV x ΔO₂)/(Δt x w)],

where ΔV is the respirometer volume (ml) minus the fish volume (ml; assumed to equal w), ΔO₂ is the measured change in oxygen content (mg/L), Δt is the time interval (minutes) that aligns with the measured ΔO₂ and w is fish mass (g). Similar to Regan and Richards (2017), to standardize FMR we calculated the mean oxygen consumption of each fish using the calculated 5-minute intervals that fell between 110% and 80% partial oxygen pressure in water. Fish (n = 19) with a partial oxygen pressure <80% in the first 5-minute interval were selected as potential outliers (n = 1 excluded). Measured oxygen consumption with slopes having low r² (< 0.5) were also considered outliers and were removed from the data set (n = 5, 5% of the data). The average r² of all metabolic rate calculations, including those used for calculation of P_crit, was 0.87.

The critical oxygen level (P_crit) was calculated with the R package ‘Segmented’ (Muggeo, 2003) using the curve of the 5-minute MO₂ intervals with the corresponding PO₂ for individual fish (Supplementary Fig. S1). During P_crit calculations, 12-34 MO₂ intervals were used per P_crit calculation depending on the length of the trial. The segmented procedure allows for the identification of multiple break points (Muggeo, 2003; Borowiec et al., 2020). When individuals demonstrated only 2 break points (n = 6), the lower PO₂ breakpoint was designated as P_crit following Borowiec et al. (2020). Fish (n = 10) for which there was no break point were removed from the P_crit data set used for analysis.

Statistical analysis

All data used in parametric statistical analyses were tested for normality and homogeneity of variance using the Shapiro–Wilk W and Levene’s homogeneity tests, respectively (Zar, 2018). General linear models were used to estimate the relationships between seasonal temperature and all dependent variables (FMR, PO2 at LOE, P_crit).

Multivariate linear models were developed and evaluated using the Akaike Information Criterion (AIC) adjusted for small sample sizes (AIC_c; Symonds and Moussalli, 2011) with turbidity, body condition, and reproductive season considered as potential explanatory variables. Turbidity (NTU) has been shown to increase metabolic rate in cinnamon clownfish (Amphiprion melanopus; Hess et al., 2017) and during our study turbidity increased during rain events (i.e. from 2.1 to 19.5 NTU). Fulton’s K, an index of body condition, was included as a predictor variable and used as a proxy for food availability since low ration corresponds to low body condition (Turko et al., 2020) and low food availability has been observed to decrease metabolic rate in brown trout (Salmo trutta; Auer et al., 2015, Archer et al., 2020). Reproductive season was included as a predictor variable since reproduction is energetically costly and can increase metabolic demand when individuals are reproductively active (Angilletta and Sears, 2000; Pollock et al., 2007; Holt and Jørgensen, 2015). Sex is difficult to distinguish in eastern sand darter with non-lethal methods and reproductive status is not clear based on visual inspection alone. Finch et al. (2012), however, have noted that in the Thames River, Ontario, the reproductive season occurs between late April and late June. The geographic proximity of the Thames and Grand Rivers (Wilcox et al., 1998; Lake Erie Source Protection Region Technical Team, 2008) suggests that reproductive season in the Grand River will be similar. Thus, here the reproductive season for experimental fish was defined as occurring from May 1st to June 30th using a binary variable; i.e. 1 = the period when eastern sand darter reproduction typically occurs and 0 = the reproductively inactive period. Models contained in the 95% confidence interval set, i.e. the set of models whose cumulative Akaike weights best approximate 0.95 (Burnham and Anderson, 2002; Symonds and Moussalli, 2011), were used to rank the importance of predictor variables in explaining each dependent variable (Symonds and Moussalli, 2011). All statistical analyses were performed in R Studio version 1.4.1717 (RStudio Team, 2021).

To test the effect of variation, all data were averaged for a given experimental testing day and the corresponding coefficient of variation for all dependent variables was regressed against seasonal temperature to determine the significance of relationships between seasonal temperature and the among-individual variation of each response variable.

Results

Eastern sand darter had the lowest mean ± standard error PO2 at LOE and FMR in October (0.50 ± 0.07 and 10.67 ± 0.89, respectively), P_crit in June (8.39 ± 0.93), the highest mean PO2 at LOE and P_crit in July (1.14 ± 0.04, 13.31 ± 1.52, respectively) and FMR in June (20.49 ± 1.15; Table 1). Both PO2 at LOE (regression F_1,88 = 41.15, r²_adj = 0.31, P < 0.001; Figure 2A) and P_crit (regression F_1,73 = 6.74, r²_adj = 0.07, P < 0.05; Figure 3A) were significantly and positively related to seasonal temperature, but FMR was not significantly related to seasonal temperature (regression F_1,82 = 1.85, r²_adj = 0.01, P = 0.18; Figure 4A). Temperature alone explained (r²_adj) ≤ 31% of the variation observed in any of the dependent variables modelled.

Figure 2

Relationship between oxygen pressure at loss of equilibrium (PO2 at LOE) and acclimation temperature (°C) of Eastern sand darter (A) and the coefficient of variation for PO2 at LOE measured for each sample day at a given acclimation temperature (°C; B). Significant regressions are shown as solid lines with associated 95% confidence intervals (grey shaded).

Figure 3

Relationship between critical oxygen tension (P_crit) and acclimation temperature (°C) of Eastern sand darter (A) and the coefficient of variation for P_crit measured for each sample day at a given acclimation temperature (°C; B). Significant regressions are shown as solid lines with associated 95% confidence intervals (grey shaded).

Figure 4

Relationships for eastern sand darter between (A) FMR (oxygen consumption, umol O₂/hr g) and acclimation temperature (°C), and (B) and FMR and acclimation temperature when distinguishing between reproductive (grey triangles) and non-reproductive (black circles) periods. Significant regressions are shown as solid lines with associated 95% confidence intervals (grey shaded)

For PO2 at LOE the single variable temperature model was ranked as the best model for explaining variation in the dependent variable. For P_crit, temperature combined with turbidity and reproductive season best explained variation in the dependent variable. For FMR, temperature combined with reproductive season and fish condition best explained variation in the dependent variable. In all cases there were equally plausible multivariate models (i.e. ∆AICc < 2, Table 2–4) for explaining variation in the dependent variable. Within the 95% confidence interval set of models, all of the considered predictor variables appeared in at least one model. Within the PO2 at LOE model confidence interval set, seasonal temperature was the only parameter with a high relative importance weight (0.942), being greater than 2× relatively more important than turbidity (0.433) and condition (0.412), and 4.6× relatively more important than reproductive season (0.203) for predicting PO2 at LOE. For PO2 at LOE, the top model was the same as the simple linear regression described above, with r²_adj = 0.31. Parameter importance for PO2 at LOE combined with the model analysis demonstrated that consideration of other variables did not improve explained variation. Within the P_crit model confidence interval set, the best model contained other variables. However, seasonal temperature was the only parameter with a high relative importance weight (0.937), being 1.8×, 2.2× and 4.3× relatively more important than turbidity (0.523), reproduction (0.433) and condition (0.218), respectively. The top model for P_crit (regression F_3,71 = 3.94, r²_adj = 0.11, P < 0.05) was significantly related to seasonal temperature (0.29, P < 0.05), but not turbidity (−0.29, P = 0.10) or reproductive season (−2.06, P = 0.12) and explained 1.48× more of the variation compared to the single temperature model (Figure 3B). The P_crit model analysis demonstrated that consideration of other variables did improve explained variation. However, the parameter importance for the variables was low and the variables in the top model were not significant, although both variables are important for adding predictive power to the model from an information theoretic point of view (Burnham and Anderson, 2002). When estimating FMR, however, the best model contained other variables and the relative importance weight for parameters in the FMR confidence set indicated that temperature (0.958) and reproductive season (0.958) were equally important for determining FMR, with condition (0.839) also being important. Turbidity was of marginal importance, with the top three variables being 3.1× more important than turbidity (importance weight = 0.229). For FMR, consideration of other variables helped improve the ability to explain variation in the data. The top model for FMR (regression F_3,80 = 9.65, r²_adj = 0.24, P < 0.001) was significantly related to seasonal temperature (0.51, P < 0.01), reproductive season (7.39, P < 0.001) and condition (−36.28, P < 0.05) and explained 24× more of the variation compared to the single temperature model. Being within the reproductive season was associated with a significant (P < 0.001) average increase of 7.39 umol/hr⋅g in FMR compared to not being within the reproductive season (Figure 4B). Further, as Fulton’s K condition factor increased by a unit of 1, FMR decreased by −36.28 umol/hr⋅g.

AIC_c and rankings of variables used to explain among-individual variation in eastern sand darter FMR. Bolded models are in the 95% confidence interval set. K defines the number of model parameters. w_i is the model Akaike weight that defines the probability that a given model is the best approximating model. Variables considered included: test water temperature (Temp), fish condition as defined by Fulton’s K (Condition), measured river turbidity in NTU (Turb) and reproductive season (Repro)

Table 2

AIC_c and rankings of variables used to explain among-individual variation in eastern sand darter FMR. Bolded models are in the 95% confidence interval set. K defines the number of model parameters. w_i is the model Akaike weight that defines the probability that a given model is the best approximating model. Variables considered included: test water temperature (Temp), fish condition as defined by Fulton’s K (Condition), measured river turbidity in NTU (Turb) and reproductive season (Repro)

Model	k	AICc	ΔAICc	wi
Temp + Repro + Condition	5	550.452	0.000	0.610
Temp + Repro + Turb + Condition	6	552.417	1.965	0.229
Temp + Repro	4	553.725	3.273	0.119
Temp + Repro + Turb	5	555.851	5.399	0.041
Temp + Condition	4	564.109	13.656	0.001
Temp + Turb + Condition	5	565.488	15.036	0.000
Temp	3	570.300	19.847	0.000
Temp + Turb	4	572.217	21.764	0.000

AIC_c and rankings of variables used to explain among-individual variation in eastern sand darter oxygen pressure at LOE. Bolded models are in the 95% confidence interval set. K defines the number of model parameters. w_i is the model Akaike weight that defines the probability that a given model is the best approximating model. Variables considered included test water temperature (Temp), fish condition as defined by Fulton’s K (Condition), measured river turbidity in NTU (Turb) and reproductive season (Repro)

Table 3

AIC_c and rankings of variables used to explain among-individual variation in eastern sand darter oxygen pressure at LOE. Bolded models are in the 95% confidence interval set. K defines the number of model parameters. w_i is the model Akaike weight that defines the probability that a given model is the best approximating model. Variables considered included test water temperature (Temp), fish condition as defined by Fulton’s K (Condition), measured river turbidity in NTU (Turb) and reproductive season (Repro)

Model	k	AICc	ΔAICc	wi
Temp	3	22.411	0.000	0.197
Temp + Turb	4	22.422	0.011	0.196
Temp + Condition	4	22.621	0.210	0.177
Temp + Turb + Condition	5	22.723	0.312	0.169
Temp + Repro	4	24.511	2.100	0.069
Temp + Repro + Turb	5	24.538	2.127	0.068
Temp + Repro + Condition	5	24.612	2.201	0.066
Temp + Turb + Condition + Repro	6	24.837	2.426	0.059

The coefficient of variation (CV) for PO2 at LOE and P_crit at each testing day did not vary significantly with temperature (PO2 at LOE: regression F_1,20 = 0.32, r²_adj = −0.03, P = 0.58; P_crit: regression F_1,18 = 0.07, r²_adj = 0.02, P = 0.26; Figures 2B and3C). For the FMR regression, the CV was significantly and positively related to seasonal temperature (regression F_1,20 = 8.03, r²_adj = 0.25, P < 0.05; Figure 5).

Discussion

We provide the first comparison of metabolic rate and hypoxia tolerance for eastern sand darter across seasonal temperature changes in a field setting. Eastern sand darter hypoxia tolerance metrics (P_crit and PO2 at LOE) were positively correlated with seasonal temperature over the study period. Temperature was also the best overall predictor of both PO2 at LOE and P_crit, although turbidity and reproductive season helped improve explanation of among-individual variation in P_crit. In contrast, for FMR temperature alone was not the best predictor, with more of the variation in FMR being explained when temperature, reproductive season, and condition were used jointly in a regression model. Furthermore, temperature did not provide strong explanatory power for either hypoxia tolerance or metabolic rate, suggesting that (1) eastern sand darter may be able to biochemically, molecularly and/or physiologically compensate across temperatures; and/or (2) the effect of temperature may be masked due to interactions with abiotic and biotic environmental factors (e.g. dissolved oxygen, ground water, species interactions). Among-individual variation in FMR did increase with temperature, whereas temperature did not influence among-individual variability in either PO2 at LOE and P_crit, possibly as a result of different individual thermal sensitivities. The result implies some intrinsic buffering capacity in metabolic rate to increased temperature changes under natural conditions. Overall, findings suggest that temperature may be a weak predictor of metabolic rate and hypoxia tolerance in a field setting where biotic and abiotic factors can act concurrently on physiological variables to affect tolerances.

As predicted, eastern sand darter PO2 at LOE and P_crit increased with increasing ambient temperature. The increase in PO2 at LOE and P_crit with increasing temperature has been observed in numerous fish species (He et al., 2015; McDonnell and Chapman, 2015; Borowiec et al., 2016; Rogers et al., 2016). However, closely related darter species have demonstrated varying hypoxia tolerance responses to changes in temperature. For example, with an increase in field temperature fantail darter PO2 at LOE did not change, rainbow darter PO2 at LOE decreased and greenside darter (E. blennioides) PO2 at LOE increased (Hlohowskyj and Wissing, 1987). Similarly, P_crit and oxygen concentration at death (the point past PO2 at LOE) increased in E. squamiceps and decreased in E. rufilineatum, E. flabellare, E. duryi, E. boschungi and E. fusiforme (Ultsch et al., 1978). The different relationships observed between temperature and hypoxia tolerance among darter species could be explained by the modifying effects of microhabitat preferences, as darters occupy a wide range of microhabitats (Ultsch et al., 1978; Chipps et al., 1994; Stauffer et al., 1996). For example, darters have been observed to segregate based on substrate, velocity and depth, which can impact the degree of temperature change darters experience on a daily and seasonal basis (Chipps et al., 1994; Stauffer et al., 1996). Ultsch et al. (1978) compared hypoxia tolerances with habitat preferences in six darters and found species with low tolerances (3.5–5.3 mg/L) were restricted to fast moving water while highly tolerant (0.76–1.8 mg/L) species occupied small and shallow streams dominated by still water. Eastern sand darter typically occupies low-velocity microhabitats, and showed a similar hypoxia tolerance to the highly hypoxia tolerant E. fusiforme that similarly occupies slow-moving waters (Ultsch et al., 1978). Eastern sand darter may have a high hypoxia tolerance compared to other darters because of their fossorial behaviour and preference for sand substrates (Daniels, 1989), which typically contain less than 2 mg/L of dissolved oxygen within the sediment (Barnucz et al., 2022).

AIC_c and rankings of variables used to explain among-individual variation in eastern sand darter critical oxygen tension (P_crit). Bolded models are in the 95% confidence interval set. K defines the number of model parameters. w_i is the model Akaike weight that defines the probability that a given model is the best approximating model. Variables considered included test water temperature (Temp), fish condition as defined by Fulton’s K (Condition), measured river turbidity in NTU (Turb) and reproductive season (Repro)

Table 4

AIC_c and rankings of variables used to explain among-individual variation in eastern sand darter critical oxygen tension (P_crit). Bolded models are in the 95% confidence interval set. K defines the number of model parameters. w_i is the model Akaike weight that defines the probability that a given model is the best approximating model. Variables considered included test water temperature (Temp), fish condition as defined by Fulton’s K (Condition), measured river turbidity in NTU (Turb) and reproductive season (Repro)

Model	k	AICc	ΔAICc	wi
Temp + Repro + Turb	5	440.545	0.000	0.209
Temp + Turb	4	440.847	0.302	0.180
Temp	3	440.938	0.393	0.172
Temp + Repro	4	441.116	0.571	0.157
Temp + Condition	4	442.367	1.822	0.084
Temp + Turb + Condition	5	442.814	2.269	0.067
Temp + Repro + Turb + Condition	6	442.833	2.288	0.067
Temp + Repro + Condition	6	442.952	2.407	0.063

Figure 5

The linear regression (solid line) and associated 95% confidence intervals (grey shading) for explaining the significant relationship between coefficient of variation (%) computed from the FMR data measured for individual eastern sand darter and temperature given as acclimation temperature (°C).

Although many studies have reported a strong correlation between metabolic rate and temperature, or significant shifts in metabolic rate as temperature increases (Fry, 1947; Clarke and Johnston, 1999; Gillooly et al., 2001; Allen and Gillooly, 2007; Borowiec et al., 2016), the relationship is by no means ubiquitous, with some species demonstrating biochemical, molecular, and/or physiological compensation of metabolic rate across temperatures (Bullock, 1955). Rosyside dace (Clinostomus funduloides) and mottled sculpin (Cottus bairdi) tested across seasons demonstrated higher metabolic rates in spring (10°C) compared to summer (15°C) due to gonadal recrudescence (Facey and Grossman, 1990). Similarly, with eastern sand darter, the FMR model including reproductive season better explained changes in FMR as temperature increased, although the overall explanatory power of the relationship remained weak (r²_adj = 0.24). The weak temperature-metabolic relationship observed in eastern sand darter may reflect partial metabolic compensation, which has been observed in a number of species. Metabolic compensation can be categorized into: complete compensation, where metabolic rate is maintained with changing temperature; no compensation, where metabolic rate increases with increasing temperature; and, partial compensation, between complete and no compensation where fish show some temperature effect (Precht, 1958). Partial metabolic compensation across natural, species-specific temperature ranges has been demonstrated in snow trout (Schizothorax richardsonii; Kamalam et al., 2019), African cichlid (Pseudocrenilabrus multicolor; McDonnell and Chapman, 2015) and rainbow trout (Oncorhynchus mykiss; Facey and Grossman, 1990) tested close to field conditions. Additionally, a longer acclimation period has been shown to lead to greater metabolic compensation in spiny chromis (Acanthochromis polyacanthus; Donelson and Munday, 2012), Nile perch (Lates niloticus; Nyboer and Chapman, 2017) and shorthorn sculpin (Myoxocephalus Scorpius; Sandblom et al., 2014). Wild zebrafish (Danio reino) appear to have greater metabolic compensation than lab fish, likely as a result of the need to maintain metabolic rate over a fluctuating range of temperatures in the wild (Morgan et al., 2022).

Eastern sand darter are exposed to diurnal temperature fluctuations of ~ 0.02–4.08°C from June to October, which on a monthly basis show average fluctuations of ~ 10°C, implying mean river temperatures change gradually across each month (Figure 1). Such gradual changes in mean temperature across the month from shifts in diurnal temperature fluctuations may result in wild fish undergoing longer acclimation, which, in turn, may lead to increased metabolic compensation in the wild. Partial compensation in metabolism would allow eastern sand darter to conserve energy and increase usable energy to maintain growth, locomotion and avoid predators across periods of diurnal and seasonal temperature changes (Evans, 1990).

The inclusion of turbidity and reproductive season, and fish condition and reproductive season, increased explanatory power in the P_crit and FMR models respectively, but not the PO2 at LOE model. Although turbidity and reproductive season increased explanatory power in the P_crit model, they were not significant in the regressions. Turbidity may have had only limited importance in the eastern sand darter models because its observed range was small (2.1–19.5 NTU) over the study, with only two days having NTU above 7. Turbidity has been shown to have no significant effect on routine metabolic rate in cinnamon clownfish until at/above 7.0 NTU (Hess et al., 2017). In contrast, turbidity has been shown to increase P_crit in imperilled pugnose shiner (Notropis anogenus) and bridle shiner (Notropis bifrenatus) at 7.31 ± 0.2 NTU (Gray et al., 2016).

Body condition contributed significantly to explaining among-individual variation in eastern sand darter FMR, despite the low variability (5% coefficient of variation) among experimental individuals, with FMR decreasing with increasing body condition. In contrast, blue catfish (Ictalurus furcatus) demonstrated no relationship between metabolic rate and body condition (Nepal et al., 2021). In our study, condition alone explained only 7% of the variation in FMR and interactive effects between fish weight, sex, seasonal changes in condition, and/or reproductive season could explain the significance observed (Weatherley, 1963; Hoeinghaus et al., 2006; Nepal et al., 2021).

Reproductive season significantly contributed to explaining among-individual variation in FMR of eastern sand darter, with metabolic rate increasing during the reproductive season. The overall influence of reproductive season on metabolic rate of oviparous fish species is poorly understood, particularly for eastern sand darter (Spreitzer, 1979; Johnston, 1989), but has been investigated in some fish species. Facey and Grossman (1990) attributed higher SMR in the spring and fall to reproduction in oviparous rosyside dace and mottled sculpin as most individuals showed signs of gonadal maturation during spring and fall. Similarly, two oviparous lizards, Sceloporus undulates and Sceloporus aeneus aeneus, demonstrated a 122% and 221% increase, respectively, in metabolic rate compared to non-gravid females (Guillette, 1982; Angilletta and Sears, 2000). Viviparous fishes have also shown significant increases in metabolic rate when reproductively active (Boehlert et al., 1991; Masonjones, 2001; Timmerman and Chapman, 2003). Well-known trade-offs between metabolic and reproductive activity may, however, limit the capacity of species to simultaneously adjust metabolic rate and maintain reproductive output under increased temperatures and/or hypoxic conditions (e.g. Fry, 1947 ; Healy and Schulte, 2012 ; Dupont-Prinet et al., 2013 ; Sandblom et al., 2014 ; Gilmore et al., 2019). For example, prolonged hypoxia has been observed to significantly reduce the gonadosomatic index of gulf killifish (Fundulus grandis;Landry et al., 2007) and common carp (Cyprinus carpio; Wu et al., 2003). If fish species, including eastern sand darter, are unable to sufficiently increase their metabolic rate during reproduction because of increased temperature and/or hypoxia, reduced energy allocation for reproduction could translate to reduced fitness and have negative consequences for population abundance and, ultimately, species conservation.

Among-individual variation increased as a function of temperature for FMR only. A linear change in among-individual variation across temperature implies that individuals have different thermal sensitives, as has been observed previously for metabolic rate but not for PO2 at LOE and P_crit (Nespolo et al., 2003; Careau et al., 2014; Norin et al., 2016; Drown et al., 2021). Our field study showed an increase in variation with increasing temperature, whereas lab studies often observe a decrease in variation with increasing temperature, e.g. juvenile barramundi (Norin et al., 2016) and mummichog (Fundulus heteroclitus; Drown et al., 2021). Similar to our study, wild zebrafish and wild-caught slimy salamander (Plethodon albagula) have both shown increased variation with increasing temperature (Careau et al., 2014; Morgan et al., 2022). In the wild, fish experience simultaneous seasonal and diurnal changes in environmental variables with temperature fluctuations tending to being greater in the summer (i.e. thermally heterogenous) than in the winter (i.e. thermally homogenous; Łaszewski, 2018). As thermal heterogeneity increases, variation in metabolic rate may have occurred for two reasons. First, individual fish may inhabit different micro-habitats with differing thermal characteristics (Kita et al., 1996; Schulte et al., 2011; Goller et al., 2014). Those inhabiting thermal refugia influenced by ground water may experience less immediate impacts of changes in water temperature, with the differences in among micro-habitat thermal characteristics becoming larger as temperatures rise (Snyder et al., 2015). In this case, fish would have larger variability in initial acclimation temperatures in the summer relative to the point measurements that were taken in the field to characterize the thermal status of the ecosystem. Second, individual fish may have different metabolic thermal sensitives. Highly sensitive individuals would have a larger thermal response to any given increase in temperature and likely acclimate more quickly to the hotter temperature of diurnal fluctuations in summer (Norin et al., 2016). Understanding field-based variation is difficult as there is a lack of control on the specific conditions an individual may have experienced prior to testing. Nonetheless, field-based studies can provide a practical assessment of physiological tolerances and related in-stream fish responses to environmental stressors, which may help to inform conservation decisions.

A large degree of individual variation was observed for all three metrics (FMR, PO2 at LOE, and P_crit) across the range of test temperatures. High among-individual variation in traits provides a portfolio effect by buffering a species from the potential negative consequences of environmental change, leading to a better overall chance of individual survival (Bolnick et al., 2011). If individuals survive the initial environmental change, and the environment continues to be habitable, a shift in the metabolic rate and hypoxia tolerance trait distributions beneficial for the species in the changed environment may occur in the next generation (e.g. Bolnick et al., 2011). For example, metabolic rate has been shown to be heritable in birds (Mathot et al., 2013) and mammals (Zub et al., 2012; Boratyński et al., 2013). A large degree of among-individual variation, in conjunction with the short generation time of eastern sand darter, therefore, could have positive ecological and evolutionary significance for the survival of this species based on the portfolio effect. However, it is important to note that heritability could be limited since the influences of developmental plasticity and transgenerational effects (i.e. the environment that previous generations were exposed to) may also influence the observed variation in adult eastern sand darter metabolic rate (Schaefer and Walters, 2010; Le Roy et al., 2017).

Our findings demonstrate the complexity of conducting physiological studies in the field and the effect that the lack of control on biotic and abiotic covariates can have on physiological metrics. Fry (1947) noted that temperature is the key controlling metabolic factor and subsequent authors have shown that temperature is a controlling factor in general for ectotherm physiology (Magnuson et al., 1979; McNab, 2002; Schulte et al., 2011). In contrast, our study shows that in a field setting, temperature is a weak controlling factor, with the reduced apparent importance of temperature being attributed to its interactions with other environmental variables that mask its effect. Although laboratory studies may facilitate accurate understanding of the effects of temperature alone on physiological metrics, as predictors of climate change impacts they are less reliable because of their failure to account for the simultaneously occurring action of those variables which field experiments cannot control (e.g. reproductive period, condition, turbidity) and which we have shown can influence key physiological metrics. Thus, while the conduct of field studies may be ‘messy’, they nevertheless provide a more realistic understanding of how changes in the environment will interact to affect the prospects for species survival.

Our field-based experiments indicated that eastern sand darter metabolic rate and hypoxia tolerance varied with season, with temperature being the most important but weak predictor of the traits overall. In the wild, diurnal and seasonal fluctuating temperatures may act in concert with other biotic and abiotic factors to determine fish physiological changes in metabolic rate and hypoxia tolerance. Our results demonstrate a large degree of individual variation across seasons, which could indicate that eastern sand darter has a greater chance of evolutionary adaptation to persistent increases in temperature and decreases in oxygen. However, a better understanding of the factors that influence this variation in the wild and determination of whether the underlying differences have a genetic basis is required to fully understand the ecological implications of our findings.

Funding

This work was supported by funding provided by a Natural Sciences and Engineering Research Council Discovery Grant to M.P., the Fisheries and Oceans Canada Species at Risk Program, and the Canadian Freshwater Species at Risk Research Network. B.F. was supported by a Natural Sciences and Engineering Research Council doctoral postgraduate scholarship.

Supplementary Data

Supplementary material is available at Conservation Physiology online.

Conflicts of Interest

The authors have no conflicts to declare.

Data Availability

The data presented in this article are available on reasonable request from the corresponding author.

Author Contributions

B.L.F. conceptualized and designed the study, conducted field work, completed statistical analysis and wrote the manuscript; P.M.C. reviewed and edited; D.A.R.D conceptualized and designed the study, reviewed and edited; M.P. conceptualized and designed the study, reviewed and edited.

Acknowledgements

The authors thank R. Gáspárdy, J. Barnucz, K. Smith, J. Burbank, M. Cowperthwaite, K. Gao, M. Hutchings, J. Skeath, S. Young and C. Timmerman who contributed to field sampling.