Hypoxia-seeking behavior, metabolic depression and skeletal muscle function in an amphibious fish out of water

Several animals enter a state of dormancy to survive harsh environmental conditions, including periods of prolonged cold or drought. During dormancy, animals remain inactive and have no external food supply; thus, a common feature of dormancy is the depression of metabolic rate to conserve limited endogenous energy reserves until the return of environmental conditions favorable for active life (for reviews, see Storey and Storey, 1990, 2012; Guppy and Withers, 1999). A true metabolic rate depression involves an active downregulation of cellular energy (ATP) turnover below the standard level (SMR; standard metabolic rate) (Storey, 2002), which supports only the essential homeostatic activities of cells and whole organisms (Hulbert and Else, 2004). Here, we use metabolic depression to describe a reduction in metabolic rate below the routine level (RMR; routine metabolic rate). The extent to which animals depress metabolic rate gives a proportional extension to the time they can survive using endogenous reserves (Storey and Storey, 1990; Boutilier et al., 1997). Consequently, a considerable metabolic depression is necessary for animals that remain dormant for many months (Hochachka and Guppy, 1987). For example, long-neck turtles (Chelodina rugosa) can survive seasonal droughts lasting 4–5 months by depressing metabolic rate by approximately 70% (Kennett and Christian, 1994), whereas African lungfish (Protopterus aethiopicus) can tolerate drought for years by depressing metabolic rate by 80–90% (Fishman et al., 1992).

Animals, both active and dormant, rely primarily on lipid and/or carbohydrate catabolism to satisfy energy (ATP) demands, while proteins are generally used as a fuel of last resort (Dark, 2005; McCue, 2010). For example, the common frog (Rana temporaria) was found to preferentially catabolize lipids during the first 40 days of hibernation, and progressively shifted to exclusive carbohydrate metabolism by 90 days (Donohoe et al., 1998). If dormant animals exhaust lipid and carbohydrate reserves, they must rely on protein catabolism to meet energy demands (Castellini and Rea, 1992; Caloin, 2004). However, extensive protein catabolism is problematic because it can result in skeletal muscle atrophy, which can be exacerbated by disuse during dormancy. Any muscle atrophy during dormancy is detrimental for animals upon arousal, when normal muscular activity must be resumed (Hudson and Franklin, 2002a,b). Impaired locomotor performance can threaten foraging ability, reproductive success, predator avoidance and, ultimately, survival.

Metabolic rate and fuel usage in dormant ectotherms will be affected by the microhabitat. For example, microhabitats with elevated temperatures can expedite substrate utilization and lead to skeletal muscle atrophy (Young et al., 2011). In contrast, microhabitat conditions that support metabolic depression may aid in skeletal muscle preservation. Hypoxic conditions, for example, have been reported to depress metabolism in a number of ectothermic animals (Boutilier et al., 1997; Hicks and Wang, 1999). For animals that remain dormant for extended periods of time, hypoxic hypometabolism would assist in skeletal muscle maintenance if it accentuates metabolic depression and protects endogenous energy reserves from premature exhaustion. Thus, do animals seek hypoxic microhabitats during dormancy that, in turn, maximally depress metabolic rate?

We used the amphibious mangrove rivulus (Kryptolebias marmoratus) to examine this question because they inhabit tropical mangrove swamps that desiccate for several months during the dry season. During seasonal droughts, K. marmoratus can survive out of water (emersed) for more than 60 days (Taylor, 1990; Turko et al., 2019), despite the inability to eat on land (Pronko et al., 2013). Although the definition of estivation is not clear cut, K. marmoratus remain largely inactive (Turko et al., 2017) and, like estivating animals, enter a hypometabolic state where RMR is depressed by more than 40% after 21 days in air (Turko et al., 2019). When on land, K. marmoratus are often found nestled within the termite tunnels of rotting logs (Taylor et al., 2008) that may be hypoxic (12–14% air saturation; Anderson and Ultsch, 1987). Several other ectothermic animals can similarly be found in potentially hypoxic habitats during periods of dormancy, including fish (Smith, 1931; Chew et al., 2004), amphibians (Ruibal and Hillman, 1981; Etheridge et al., 1990; Boutilier et al., 1997; Stewart et al., 2003; Booth 2006) and reptiles (Kennett and Christian, 1994; Markle and Chow-Fraser, 2017) that estivate/hibernate within subterranean burrows or beneath the ice of frozen lakes. Thus, we used K. marmoratus to test the hypothesis that animals seek hypoxic microhabitats during periods of dormancy that, in turn, accentuate metabolic depression. The hypoxic habitat hypothesis predicts that K. marmoratus will preferentially occupy hypoxic rather than normoxic microhabitats during periods of prolonged air exposure. Fish were placed in custom experimental choice chambers that maintained an O₂ gradient (normoxic to hypoxic) to determine the preferred O₂ level of fish out of water. If fish select hypoxic microhabitats when on land, then exposure to aerial hypoxia acutely should result in a more profound metabolic depression than results from aerial normoxia exposure. Thus, we measured the O₂ consumption rate of water- (control) and air-acclimated (21 days) fish in aerial normoxia, as well as the O₂ consumption rate of air-acclimated fish acutely exposed to aerial hypoxia. We then tested the hypothesis that chronic hypoxia acclimation in air would protect endogenous energy reserves and skeletal muscle integrity, thereby maintaining locomotor performance. The hypothesis predicts that K. marmoratus acclimated to aerial hypoxia will deplete energy stores more slowly, demonstrate less skeletal muscle atrophy and have better locomotor performance than fish acclimated to aerial normoxia, presumably owing to hypoxic hypometabolism. We measured the whole-body [glycogen] and lipid content in fish acclimated to water (control), air and aerial hypoxia for 21 days. The cross-sectional area of red and white muscle fibers was also measured, as well as terrestrial locomotor performance (tail-flip jumping).

All experimental fish (n=235, 0.088±0.002 g) were adult hermaphrodites of the self-fertilizing Kryptolebias marmoratus Poey 1880. We performed all experiments in duplicate on two isogenic strains: strain 1 (SLC8E, from St Lucie County, FL, USA) and strain 2 (50.91, from Twin Cayes, Papa Gabriel, Belize) (Tatarenkov et al., 2010). Prior to experiments, fish were individually maintained in 120 ml plastic holding cups (60 ml water, 15‰ salinity, 25°C) in the Hagen Aqualab at the University of Guelph on a 12 h:12 h light:dark cycle (Frick and Wright, 2002). Fish were fed live Artemia sp. nauplii three times weekly. All experimental procedures were approved by the University of Guelph Animal Care Committee (AUP 3891).

Three series of experiments were conducted: (1) a behavioral experiment to determine the environmental O₂ preference of K. marmoratus out of water during a 21 day air exposure period, (2) a physiological experiment to determine whether acute aerial hypoxia exposure after 21 days in air accentuates the metabolic depression in K. marmoratus, and (3) a physiological experiment to determine whether chronic aerial hypoxia acclimation (21 days) slows substrate utilization compared with aerial normoxia acclimation (21 days), thereby preserving skeletal muscle integrity and locomotor performance (Fig. S1). Following experimentation, fish were weighed and the standard length was measured for determination of body condition (Fulton's K) as described by Froese (2006). When the partial pressure of O₂ (P_O2) was not constant for the entire the experimental period (i.e. during choice chamber experiments and acute hypoxia exposure during O₂ consumption experiments), Fulton's K was not measured.

To determine whether K. marmoratus seek hypoxic microhabitats during periods of prolonged air exposure, we designed experimental choice chambers that maintained a P_O2 gradient through which fish could move freely (Fig. S2). We constructed the choice chambers from 10 ml serological pipettes to mimic the termite tunnels in rotting logs that K. marmoratus inhabit in the wild (Taylor et al., 2008). The serological pipettes were halved to create semi-circlular tubes (21 cm long, 1 cm diameter), which were then fitted with a mesh bottom. A strip of water-soaked (15‰) filter paper (0.5 cm wide) was placed inside each tube, on top of the mesh to give fish a moist substrate for travel within the tube.

We measured the P_O2 preference of K. marmoratus (strain 1, n=12; strain 2, n=23) within the choice chambers during the initial (days 1–2) and final (days 20–21) 48 h of a 21-day air exposure period. First, we randomly positioned fish in individual choice chambers, then introduced a P_O2 gradient (10.1±0.4 to 19.9±0.1 kPa). The P_O2 gradient was generated by placing the choice chambers over three side-by-side containers (1.5 litres) filled with water (15‰) (Fig. S2). We vigorously bubbled nitrogen into the container at one end to generate hypoxic air in the choice chamber above (‘low O₂ zone’=10.1±0.4 kPa). In the container at the opposite end, we vigorously bubbled air to generate normoxic air in the choice chamber above (‘high O₂ zone’=19.9±0.1 kPa). We bubbled both air and nitrogen into the middle container to generate an intermediate P_O2 in the air above (‘medium O₂ zone’=15.7±0.3 kPa). The P_O2 in each zone was continuously measured during experimental trials using O₂-sensing optodes and a fibre-optic probe (Loligo Systems WITROX 4, Tjele, Denmark). The optodes were calibrated prior to experimentation using humidified atmospheric air (21 kPa O₂) and 2 mol l⁻¹ sodium sulphite (0 kPa O₂). Fish were video recorded within the choice chamber using a webcam (Logitech C905w, Newark, CA, USA). Following exposure to the P_O2 gradient on days 1–2, the P_O2 gradient was stopped, and fish were maintained in their choice chamber under normoxic conditions until re-exposure to the P_O2 gradient during days 20–21. We did not reposition fish within the choice chambers before reintroduction of the P_O2 gradient to minimize disturbance. For analysis, we calculated the proportion of time fish spent in each O₂ zone during both recording periods.

All behavioral tests were carried out across seven 21-day experimental trials (five fish per trial on average). To ensure that all experimental trials were carried out consistently, we began all trials between 12:00 and 16:00 h, in the same position within the same climate-controlled room (25°C, 12 h:12 h light:dark). The choice chamber was illuminated in the day with room lighting (no exterior windows), and at night with red light to minimize disturbance to the fish. The direction of the O₂ gradient (hypoxia to normoxia) was randomly reversed (normoxia to hypoxia) between trials, and most of the trials contained fish from both strains. All videos were analysed by the same observer (G.S.R.).

A separate group of fish was acclimated for 21 days to one of two experimental treatments: water (control) or air. Control fish were maintained in water (15‰) and fed Artemia sp. nauplii three times weekly. Air acclimation was achieved by placing fish on moist filter paper (15‰) in 120 ml plastic holding cups, as previously described (Ong et al., 2007). The mass-specific rate of O₂ consumption was then measured in control (strain 1, n=10; strain 2, n=10) and air-acclimated (strain 1, n=8; strain 2, n=8) fish using closed-system respirometry. Fish were placed in custom glass micro-respirometry chambers (∼1 ml) containing a piece of wet filter paper that kept the environment moist during the experimental period. Fish were acclimated to the experimental chamber for a 2 h period in air, during which time the chamber was left open to maintain P_O2 at approximately 21 kPa. A 2 h acclimation has previously been shown to be sufficient for K. marmoratus to recover from handling stress (Blanchard et al., 2019). Following the 2 h acclimation period, the chamber was sealed and the decline in P_O2 was measured for 1 h using O₂-sensing optodes, which were calibrated weekly. To determine whether aerial hypoxia accentuates metabolic depression in K. marmoratus, an additional group of fish (strain 1, n=8; strain 2, n=8) held in air for 21 days was placed in micro-respirometry chambers that were continuously flushed with an air–N₂ mixture (approximately 9.3 kPa O₂; Wosthoff, Calibrated Instruments Inc., NY, USA) for 2 h prior to the 1 h measurement period. We attempted to measure O₂ consumption in fish acclimated to aerial hypoxia continuously for 21 days in preliminary experiments. However, we were unable to prevent re-oxygenation while transferring fish from their acclimation chambers to the micro-respirometry chambers, resulting in highly variable O₂ consumption measurements. We calculated O₂ consumption rates from the decline in P_O2 over the 1 h measurement period in three consecutive 20 min intervals, standardized to body mass. There was no significant difference in O₂ consumption across the technical replicates (repeated-measures ANOVAs; P>0.05), thus, we report the average of the three O₂ consumption measurements. The P_O2 in the O₂ consumption trials performed under normoxia started at 20.1±0.2 kPa and declined to 17.8±0.3 kPa by the end of the measurement period. The P_O2 in the O₂ consumption trials performed under aerial hypoxia started at 9.3±0.2 kPa and declined to 8.4±0.2 kPa. The P_O2 in all trials remained above the lower aerial critical oxygen tension (P_crit) for K. marmoratus strain 1 (SLC8E) and strain 2 (50.91) (Blanchard et al., 2019). All O₂ experiments were conducted at 25°C, and between 12:00 and 18:00 h to minimize the effects of diel metabolic rhythms (Rodela and Wright, 2006). Control fish were fasted for at least 48 h prior to use in experiments, which is more than sufficient time to ensure that metabolic rate returns to baseline levels post-feeding (Sutton et al., 2018). We also measured O₂ consumption in a group of fish that had been fasted for 21 days in water (fasted controls) to determine whether the inability to eat on land plays a significant role in metabolic depression in normoxic air.

An additional group of fish was exposed for 21 days to one of three experimental treatments – control, air or aerial hypoxia – for analysis of whole-body [glycogen] and lipid content. Aerial hypoxia acclimation was achieved by placing air-exposed fish in an incubator (Innova 4230, New Brunswick Scientific, NJ, USA) that was maintained at approximately 8.2 kPa O₂ using an air–N₂ mixture (25°C, 12 h:12 h light:dark). Following acclimation, fish were immediately euthanized via cold-water immersion, and either snap-frozen in liquid nitrogen for [glycogen] determination (strain 1, n=22; strain 2, n=21), or dried (48 h at 60°C) for lipid analysis (strain 1, n=22; strain 2, n=23). Whole-body [glycogen] was measured enzymatically (Bergmeyer et al., 1974), and lipid content was measured by chloroform extraction (Junior and Peixoto, 2013).

We measured the size (cross-sectional area) of red and white muscle fibers in another group of fish under control, air or aerial hypoxia conditions for 21 days (strain 1, n=22; strain 2, n=22). Following acclimation, fish were euthanized via immersion in MS-222 (500 mg l⁻¹). A ∼3 mm transverse steak immediately anterior to the dorsal fin was removed, coated in embedding medium (Shandon Cryomatrix™, Fisher Scientific, Hampton, NH, USA), frozen in liquid nitrogen-cooled isopentane and stored at −80°C. Frozen muscle steaks were then cut into 8 μm transverse sections in a cryostat (Leitz Cryostat Microtome, Labequip Ltd, Markham, ON, Canada) at −20°C, mounted on Superfrost Plus slides (Fisher Scientific) and stored at −80°C until staining. Red muscle staining and analysis was performed as previously described by Rossi et al. (2018). White muscle analysis was performed as described by Rossi et al. (2019a). Control and air-acclimated fish used for muscle phenotyping were previously used for O₂ consumption measurements.

We compared the terrestrial locomotor performance in a final group of control, air-acclimated and aerial hypoxia-acclimated fish (21 days; strain 1, n=23; strain 2, n=23), as previously described (Rossi et al., 2019a; McFarlane et al., 2019). Briefly, fish were encouraged to jump via gentle prodding with a clicker ballpoint pen until exhausted, i.e. unresponsive to prodding. The jumping trials were video recorded and analyzed to quantify the distance of the longest jump (in body lengths) and the number of jumps fish performed before reaching exhaustion (Brunt et al., 2016).

We determined the preferred O₂ zone of K. marmoratus during days 1–2 and days 20–21 using methods for compositional data, because the proportion of time fish spend in any one O₂ zone is dependent upon the time spent in other O₂ zones. All proportions were isometric log-ratio (ILR) transformed and then analysed using an ordinary least squares (OLS) regression. We subsequently back-transformed the coefficients for meaningful interpretation of the results (van den Boogrart and Tolosana-Delgado, 2013).

To determine the effects of strain and experimental treatment on O₂ consumption, body condition, lipid content, muscle phenotype and jumping performance, we used two-way ANOVAs. When a significant main effect of treatment was detected, a Tukey's post hoc test was used to identify which treatment groups differed (all reported P-values are adjusted for multiple comparisons). When a significant strain×treatment interaction was detected (glycolytic fiber size, number of jumps), the data were divided by strain and analysed using a one-way ANOVA, followed by a Tukey's post hoc test. The [glycogen] of strain 2 control and air-acclimated fish was compared using a two-sided t-test; the aerial hypoxia group was excluded from the analysis because the [glycogen] in five of six fish was below detectable limits. A one-way ANOVA was used to compare the [glycogen] of strain 1 fish across treatment groups, followed by a Tukey's post hoc test. Owing to the limitations of merely presenting P-values (Halsey et al., 2015; Halsey, 2019), we supplemented the P-values from all post hoc comparisons with an effect size calculation (Cohen's D; d-value) for each strain (strain 1=d₁; strain 2=d₂) to indicate the magnitude of the effect between treatment groups. When d<0.5, the treatment effect is relatively small and inconsequential, whereas d>0.5 indicates that the treatment effect is biologically meaningful (Sullivan and Feinn, 2012). Prior to analysis, we assessed all data for normality and homogeneity of variance; when data did not meet these parametric test assumptions, they were appropriately transformed. All analyses were performed using RStudio (version 1.1.447).

The proportion of time strain 2 fish spent in each O₂ zone during days 1–2 was equally distributed (OLS: P=0.63; Fig. 1). However, strain 2 fish showed a preference for the low O₂ zone by the end of the air exposure period (OLS: P=0.02): fish spent 53±7% of the time in the low O₂ zone, and only 26±5% and 21±6% in the medium and high O₂ zones, respectively (Fig. 1). The time strain 1 fish spent in each O₂ zone was equally distributed during days 1–2 (OLS: P=0.15) and days 20–21 (OLS: P=0.80) (data not shown).

The mass-specific rate of O₂ consumption in K. marmoratus was significantly influenced by the experimental treatment (two-way ANOVA: P<0.001) but not by strain (two-way ANOVA: P=0.76; Fig. 2A). We found that control fish had significantly higher O₂ consumption rates than both air- and aerial hypoxia-acclimated fish (Tukey: P<0.001, d₁=1.5, d₂=1.0; P<0.001, d₁=2.2, d₂=1.7). Air-acclimated fish also had higher O₂ consumption rates than aerial hypoxia-acclimated fish (Tukey: P=0.01, d₁=1.5, d₂=1.4). In contrast, the O₂ consumption rate between fasted controls and air-acclimated fish was not significantly different in either strain (two-way ANOVA: P=0.19, P=0.36; Fig. S3). We found that the experimental treatment affected body condition (two-way ANOVA: P<0.001) but the strain did not (two-way ANOVA: P=0.30; Fig. 2B). The body condition of air-acclimated K. marmoratus was lower than that of control and aerial hypoxia-acclimated fish (Tukey: P<0.001, d₁=0.7, d₂=0.8; P=0.01, d₁=0.5, d₂=0.6). We found no significant difference in the body condition of control and aerial hypoxia-acclimated fish (Tukey: P=0.39, d₁=0.3, d₂=0.3).

The [glycogen] of strain 1 fish differed significantly across treatment groups (one-way ANOVA: P<0.001; Fig. 3A). We found that control fish had a higher [glycogen] than those acclimated to air and aerial hypoxia (Tukey: P<0.001, d₁=1.2; P=0.004, d₁=0.9). Although the [glycogen] of air- and aerial hypoxia-acclimated fish did not differ significantly (Tukey: P=0.39), there was a meaningful trend towards reduced glycogen usage in aerial hypoxia-acclimated fish, which catabolized 17% less glycogen than fish acclimated to air (d₁=0.7). Similarly, in strain 2 fish, the control group had a higher [glycogen] than the air-acclimated group (t-test: P<0.001, d₂=3.3), but glycogen was undetected in the aerial hypoxia group (Fig. 3A). The lipid content of K. marmoratus was significantly influenced by the experimental treatment (two-way ANOVA: P=0.002) but not by strain (two-way ANOVA: P=0.20; Fig. 3B). We found that air-exposed fish had a lower lipid content than control fish (Tukey: P=0.001, d₁=0.9, d₂=1.5). There was no significant difference in the lipid content of control and aerial hypoxia-acclimated fish (Tukey: P=0.41), but the effect sizes imply that both strains catabolized lipids under aerial hypoxia (d₁=1.0, d₂=0.9). Similarly, there was no significant difference in the lipid content of air- and aerial hypoxia-acclimated fish (P=0.05), but effect size implies that the use of 23% fewer lipids by strain 2 fish acclimated to aerial hypoxia may be biologically important (d₁=0.4, d₂=0.8).

Both the experimental treatment and the strain affected the size of red muscle fibers in K. marmoratus (two-way ANOVA: P=0.007, P=0.04; Fig. 4A). We found that air-exposed fish had significantly larger red muscle fibers than control fish (Tukey: P=0.005, d₁=1.3, d₂=1.0). Although there was no significant difference in the size of red muscle fibers between air- and aerial hypoxia-acclimated fish (Tukey: P=0.12), in strain 1 fish there was a trend towards smaller fibers in aerial hypoxia than in air (d₁=1.3, d₂=0.3). Similarly, we found no significant difference in the size of red muscle fibers between control and aerial hypoxia-acclimated fish (Tukey: P=0.43), though strain 2 fish acclimated to aerial hypoxia had 15% larger fibers than control counterparts, which may be biologically meaningful (d₁=0.0, d₂=1.0). The size of the white muscle fibers in strain 1 and strain 2 fish was differentially affected by the experimental treatment, as evident by a significant strain×treatment interaction (two-way ANOVA interaction: P=0.01; Fig. 4B). The size of white muscle fibers did not differ across treatments in strain 1 fish (one-way ANOVA: P=0.12) but differed in strain 2 fish (one-way ANOVA: P=0.03). Strain 2 fish acclimated to air had significantly smaller white muscle fibers than control fish (Tukey: P=0.02, d₂=1.3). We found no significant difference in the size of white muscle fibers between control and aerial hypoxia-acclimated fish (P=0.70, d₂=0.4). We found no significant difference in the size of white muscle fibers between air- and aerial hypoxia-acclimated fish (Tukey: P=0.15), but there was a meaningful trend towards larger white muscle fibers in aerial hypoxia-acclimated fish, which were 24% larger than those of fish in air (d₂=1.2).

The longest jump K. marmoratus could perform was affected by the experimental treatment (two-way ANOVA: P=0.04), but strain had no effect (two-way ANOVA: P=0.06; Fig. 5A). The longest jump air-acclimated fish could perform was shorter compared with that of control fish (Tukey: P=0.03, d₁=1.5, d₂=1.3). There was no significant difference in longest jump length between control and aerial hypoxia-acclimated fish (Tukey: P=0.54), but in strain 1 the effect size implies that the longest jump fish could perform in aerial hypoxia was shorter than under control conditions (d₁=0.8, d₂=0.4). Similarly, we found no significant difference in the longest jump length between air- and aerial hypoxia-acclimated fish (Tukey: P=0.29), but there was a meaningful trend towards longer jumps by strain 2 fish in aerial hypoxia, which were 20% longer than jumps in air (d₁=0.2, d₂=1.1). We found that the number of jumps strain 1 and strain 2 fish could perform before exhaustion was differentially affected by treatment (two-way ANOVA interaction: P=0.05; Fig. 5B). Strain 1 fish jumped the same number of times regardless of treatment (one-way ANOVA: P=0.46). However, the number of jumps strain 2 fish performed differed across treatments (one-way ANOVA: P=0.01): air-acclimated fish jumped fewer times than control fish (Tukey: P=0.02, d₂=1.6). There was no significant difference between the number of jumps performed by air- and aerial hypoxia-acclimated fish (Tukey: P=0.05), but the effect size implies that the 20% more jumps performed by strain 2 fish acclimated to aerial hypoxia may be biologically important (d₂=1.6).

We used K. marmoratus to test the hypothesis that animals seek hypoxic microhabitats that, in turn, accentuate metabolic depression during dormancy. In support of our hypothesis, we found that one isogenic strain spent more than double the amount of time in hypoxia compared with normoxia after 21 days in air. Furthermore, prolonged air exposure in both strains resulted in lower O₂ consumption rates compared with active fish (35% depression), which was accentuated (51% depression) when fish were exposed to aerial hypoxia acutely. The fact that fish exposed to aerial hypoxia demonstrated lower O₂ consumption rates than air-exposed fish (which were fasted and inactive) is suggestive of active metabolic rate depression, particularly because O₂ consumption measurements were obtained above P_crit for K. marmoratus in air (Blanchard et al., 2019), indicating that this was not simply an oxy-conforming response. We also tested the hypothesis that chronic hypoxia exposure in air would protect endogenous energy reserves and skeletal muscle integrity, thereby maintaining locomotor performance, possibly owing to hypoxic hypometabolism. We found that air-acclimated fish from both strains were in poorer body condition relative to control fish and those acclimated to aerial hypoxia. Air-acclimated fish demonstrated considerable glycogen and lipid catabolism (both strains), white muscle atrophy (strain 2 only) and poorer terrestrial locomotion compared with control fish (strain 2 only). However, aerial hypoxia acclimation reduced glycogen usage by 17% in strain 1 fish, and spared lipids and white muscle protein in strain 2 fish by 23% and 22%, respectively. Furthermore, aerial hypoxia acclimation maintained the terrestrial locomotor performance of strain 2 fish compared with air acclimation (22% farther longest jump, 20% more jumps). Overall, our findings demonstrate that K. marmoratus seek microhabitats that, in turn, accentuate metabolic depression, and that microhabitat O₂ availability can have significant implications for substrate utilization, and the structure and function of skeletal muscle.

We showed that K. marmoratus (strain 2) seek hypoxic microhabitats after prolonged air exposure, and one possible benefit is an accentuation of metabolic depression. Our laboratory findings are consistent with the observed habitat preference of K. marmoratus in the wild, as fish are often found within decaying mangrove logs (Taylor et al., 2008), which are likely hypoxic (Anderson and Ultsch, 1987). Although strain 1 fish did not seek hypoxia on land, they catabolized less of their metabolic fuel stores (glycogen and some lipids) after 21 days in air compared with strain 2 fish (glycogen, lipids and muscle protein). This was an interesting finding because the absolute rate of O₂ consumption was not significantly different between strains after 21 days in air. One possible explanation for the differential fuel use relates to the rate of metabolic depression. For instance, if strain 1 fish depressed metabolic rate to the air-acclimated (21 day) value several days before strain 2 fish, this would likely result in considerable energy savings. Thus, strain 2 fish may face greater pressure to seek low O₂ microhabitats during seasonal droughts that accentuate metabolic depression and preserve their remaining energy stores. Alternatively, there are several other reasons why K. marmoratus may be found in hypoxic microhabitats in the wild, not just as a strategy to reduce O₂ consumption. For example, the decaying logs in which K. marmoratus are found may contain more moisture that would protect fish from dehydration and help survival. Overall, the microhabitat conditions that mediate microhabitat selection strategies in dormant taxa are poorly understood, but this remains an exciting avenue for future investigation.

Prolonged air exposure in both strains resulted in lower O₂ consumption rates compared with active fish, which was accentuated when fish were exposed to aerial hypoxia acutely. Microhabitat O₂ availability has similarly altered the extent of metabolic depression in other dormant animals. For example, hibernating frogs (Rana temporaria) exposed to chronic aquatic hypoxia exhibited a 75% suppression of metabolic rate (Donohoe and Boutilier, 1998), whereas those in normoxic water only depressed metabolism by approximately 50% (Donohoe et al., 1998). It is unknown whether hypoxic hypometabolism would persist in K. marmoratus with chronic hypoxia exposure in air. However, in many fishes acclimated to chronic hypoxia (weeks to months), O₂ consumption remained low over time (e.g. Kerstens et al., 1979; Timmerman and Chapman, 2004). We recently showed that the green-striped burrowing frog (Cyclorana alboguttata) also selects environmental hypoxia, which consistently accentuated metabolic depression by ∼30% throughout the first 7 weeks of estivation (Rossi et al., 2020). Taken together, we suggest that the microhabitats occupied by animals during dormancy can have a profound impact on the physiological strategies used to tolerate environmental extremes.

The pattern of metabolic fuel use differed between K. marmoratus strains. Strain 1 fish catabolized 75% of their glycogen stores after 21 days in air, but only 58% after 21 days in aerial hypoxia. Glycogen stores are considered to be the primary source of metabolic fuel for fish experiencing hypoxia (Hochachka and Somero, 1984). Thus, glycogen preservation under aerial hypoxia suggests that strain 2 fish may be reducing their overall ATP demand. In contrast, strain 2 fish catabolized glycogen, lipids and presumably protein (given the white muscle fiber atrophy) in air. Aerial hypoxia acclimation attenuated lipid usage in this strain by 23%, and spared muscle protein. These findings are consistent with that from overwintering goldfish (Carassius auratus), in which lipid oxidation is constrained under extreme hypoxia in favor of anaerobic glycolysis (Hochachka and Somero, 1984). Accordingly, the [glycogen] of most strain 2 fish was undetectable after aerial hypoxia acclimation. A recent study from our laboratory also demonstrated considerable glycogen and lipid catabolism in three isogenic strains of K. marmoratus (50.91, SLC8E and HON9) after 21 days of air exposure (Turko et al., 2019). Interestingly, both our study and that of Turko et al. (2019) found that strain 2 (50.91) fish catabolized 20–25% more of their lipid reserves than strain 1 (SLC8E) fish after air exposure despite the same rate of O₂ consumption between the two strains after 21 days in air. The rate at which metabolic depression is achieved between K. marmoratus strains may explain the discrepancy in substrate use, but this hypothesis warrants further investigation. Overall, our findings suggest that K. marmoratus strains can employ different behavioral and physiological strategies to prevent the premature exhaustion of endogenous energy reserves during prolonged air exposure. Furthermore, the differential responses between isogenic strains suggest that genetic factors contribute to phenotypic differences in the behavior and physiology of K. marmoratus.

The integrity of skeletal muscle in K. marmoratus can be critical for survival. Like most fish, K. marmoratus rely on their skeletal musculature to power the locomotor movements required for accomplishing essential daily tasks (e.g. predator avoidance, prey capture). Air exposure resulted in the hypertrophy of red muscle fibers in K. marmoratus, which we have previously demonstrated is the result of the higher O₂ availability in atmospheric air relative to water (Rossi et al., 2018). In contrast, strain 2 fish demonstrated atrophy of their white muscle fibers after air acclimation, suggesting that the greater substrate utilization by these fish may have led to protein catabolism. Consequently, strain 2 fish demonstrated poorer locomotor performance (shorter longest jump, fewer jumps performed) after air acclimation. In support of our hypothesis, the skeletal muscle integrity and locomotor performance of K. marmoratus was maintained after 21 days on land with chronic aerial hypoxia exposure. Previous studies have demonstrated that a profound metabolic depression during dormancy can preserve skeletal muscle structure and function (for review, see Hudson and Franklin, 2002a). For example, the frog C. alboguttata depresses metabolic rate by up to 80% during estivation (Kayes et al., 2009; Young et al., 2011) and exhibited no skeletal muscle atrophy after several months (Hudson and Franklin, 2002b). Any changes in the skeletal musculature of dormant animals that negatively impact locomotor performance can threaten foraging ability, reproductive success, predator avoidance and, ultimately, survival.

We demonstrated that some K. marmoratus strains seek hypoxic microhabitats during prolonged air exposure, and that acute hypoxia exposure accentuates metabolic depression on land. Interestingly, previous studies from our laboratory have showed that K. marmoratus avoids aquatic hypoxia by emersing (Regan et al., 2011; Livingston et al., 2018). How is this behavioral pattern regulated? In other words, what mechanisms allow animals to avoid hypoxia during active periods, but favor hypoxic microhabitats during dormancy? Elucidating the physiological mechanisms modulating behavioral responses to hypoxia in dormant animals will be a fruitful avenue for future investigation.

Although we focused on environmental hypoxia in this study, there are numerous other abiotic factors that may interact with hypoxia to enhance metabolic depression in dormant animals. For example, the subterranean burrows of hibernating hamsters (Mesocricetus auratus) can be extremely hypoxic and hypercapnic (Kuhnen, 1986). Similarly, in mangrove swamps, low O₂ environments are often accompanied by hydrogen sulphide (H₂S) (Rossi et al., 2019b). If the microhabitats occupied by K. marmoratus during the dry season are both hypoxic and H₂S-rich, we might expect a more profound metabolic depression, as H₂S inhibits oxidative phosphorylation in K. marmoratus (Cochrane et al., 2019). Although laboratory studies are valuable for evaluating the physiological mechanisms involved in metabolic depression during dormancy, understanding how microhabitat conditions alter physiological responses is critical.

We thank Mike Davies, Matt Cornish, Nicole Carpenter, Sarah Quereshi and numerous undergraduate volunteers for animal care. We also thank the reviewers for helpful commentary that improved the manuscript.