Snapping turtle (C. serpentina) eggs were collected from the wild in Minnesota, United States, and transported to the University of North Texas for incubation (altitude elevation = 70 m and barometric pressure 759–760 mmHg). Permission to collect the eggs was granted to DA Crossley by the Minnesota Department of Natural Resources (permit no. 21232). Two eggs from individual clutches were staged to determine age. Incubations lasted no more than 55 days and all eggs were maintained at the sex-determining temperature of 30°C, to ensure all embryos developed as females (Alvine et al., 2013). Eggs were embedded to their midpoint in vermiculite, inside plastic incubators (2.5-l Ziploc© Container, SC Johnson, Racine, WI, United States) that were stored in a walk-in Percival Environmental Control Room (model IR-912L5; Percival Scientific, Perry, IA, United States). The vermiculate was mixed in a 1:1 ratio with water, as previously described (Crossley and Altimiras, 2005).

At approximately 20% development (9–12 days after laying; determined by embryonic staging), eggs were randomly assigned to either normoxic (21% O₂, partial pressure 152 mmHg) or hypoxic (10% O₂, partial pressure 72 mmHg) cohorts. The experimental gas conditions were maintained in 76-l Ziplock© bags that were connected to gas supply of either normoxia or 10% oxygen, in the environmental chamber. The normoxic gas was supplied using air pumps (LT 11 Whitewater) passed through a rotameter flow controller. The hypoxic gas was generated using rotameters (Sho Rate Brooks Instruments Division, Hatfield, PA, United States), with compressed N₂ and air supplied by an air pump (Whisper AP 300 tetra products). Normoxic and hypoxic air was humidified using bubbling chambers and delivered to the bags at a rate of 2–4 L min^–1. Gas composition was monitored continuously using an oxygen analyzer (S-3AI, Ametek Applied Electrochemistry, Berwyn, PA, United States). Turtle hatch time was between 50 and 52 days, with no difference between experimental groups. Upon hatching, all turtles were housed in common, normoxic conditions (21% O₂), at 26°C, in a daily 12:12 light–dark cycle, and fed with dry crocodilian food (Mazuri, PMI Nutrition International, Brentwood, MO, United States) two to four times weekly. Mitochondrial experiments were performed when turtles were 8 months old.

Mitochondria were isolated according to previous protocols (Galli et al., 2013, 2016). Turtles were first induced into anesthesia in a sealed plastic box, containing cotton gauze soaked in isoflurane (Isoflo^®, Abbott Laboratories, North Chicago, IL, United States). Once the turtles were fully anaesthetized, they were euthanized by cranial and spinal pithing. Hearts were removed and weighed, and the ventricle was removed and cleared of connective tissue. The ventricle was then rinsed with ice-cold homogenization buffer (Table 1) and minced into small pieces with a razor blade. The tissue was resuspended in ice-cold homogenization buffer to remove any blood, and the suspension was homogenized in a 12-ml glass mortar using four passes of a loose-fitting Teflon pestle at 120 rpm. The homogenate was centrifuged in polycarbonate tubes at 600 g for 10 min at 4°C. The supernatant was filtered through gauze and centrifuged again at 8,500 g for 10 min. The pellet was then resuspended in fresh homogenization buffer and centrifuged again at 8,500 g for 10 min. Finally, the pellet was resuspended in ∼200 μl of fresh buffer and immediately analyzed for protein content with a Bradford assay, according to the manufacturer’s protocol (Bio−Rad Laboratories, Hercules, CA, United States). The mitochondrial suspension was kept on ice until assayed.

Respiration and H₂O₂ production were measured simultaneously with an Oxygraph O₂-k high-resolution respirometry system (Oroboros Instruments GmbH, Innsbruck, Austria), fitted with an O₂k-fluorescence LED2-module. Two identical respiration chambers (chamber A and chamber B) were held at the same temperature (25°C) and run in parallel for each experiment. O₂ electrodes were calibrated every morning with air-saturated respiration medium (Table 1). To measure H₂O₂ production, 10 μM Amplex^® UltraRed and 1 U ml^–1 horseradish peroxidase (HRP) were added to each chamber. Amplex^® UltraRed oxidizes in the presence of H₂O₂ and forms resorufin, using HRP as a catalyst. Amplex^® UltraRed was excited at 563 nm and emission was read at 587 nm; 5 U ml^–1 superoxide dismutase (SOD) was also added to the chambers to convert any extramitochondrial superoxide (O^–₂) to H₂O₂. At the beginning of each experiment, isolated mitochondria (9.7–29.4 μg protein ml^–1) were added to each chamber containing 2 ml of respiration medium, and the Amplex UltraRed signals were calibrated with known quantities of exogenously added H₂O₂.

We were interested in measuring three main variables commonly used to assess mitochondrial function. First, Leak respiration rate (Leak) is the amount of O₂ used to compensate for proton Leak across the mitochondrial inner membrane. Proton Leak can account for up 20–25% of routine metabolic rate in vertebrates, so a reduction in this parameter is a substantial energy-saving mechanism. Second, we measured OXPHOS respiration rate, which is the maximum rate of ADP-stimulated respiration and is an estimate of mitochondrial ATP production. Lastly, we used a protonophore, carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone (FCCP), to uncouple mitochondrial respiration and measure the maximal respiration rate of the electron transport chain, otherwise known as maximum electron-transfer (ET) capacity. We measured these three parameters with different substrate combinations, using a standard substrate inhibitor titration protocol (SUIT protocol), designed according to Pesta and Gnaiger (2012).

An original trace of the SUIT protocol is presented in Figure 1. First, pyruvate (5 mM), malate (2 mM), and glutamate (10 mM) were added to achieve Leak respiratory state with complex-I (CI) substrates in the absence of adenylates (Leak_N,CI). When O₂ consumption was stable, saturating ADP (5 mM) was injected to activate oxidative phosphorylation with CI substrates (OXPHOS_CI). Once ADP had been completely phosphorylated to ATP, isolated mitochondria enter the Leak respiratory state in the presence of adenylates with CI substrates (Leak_T,CI), otherwise known as state-IV respiration. Succinate (10 mM) was then added to assess the additive effects of complex-II (CII) substrates (Leak_T,CI+CII), and ADP was added again to assess OXPHOS with CI and CII substrates (OXPHOS_CI+CII). To uncouple mitochondria and assess ET with CI and CII substrates (ET_CI+CII), FCCP was titrated to a final concentration of 0.1–0.3 μM. Next, the CI inhibitor rotenone (0.5 μM) was added to assess ET_CII, with CII substrates only. To block the electron transport chain and assess residual non-mitochondrial O₂ consumption (ROX), the complex-III (CIII) inhibitor antimycin-A (2.5 μM) was added. To assess complex-IV (CIV) activity in isolation, the electron donor N, N,N’,N’-tetramethyl-p-phenylenediamine (TMPD; 0.5 mM) was added in combination with ascorbate (2 mM) to avoid autooxidation of TMPD. Lastly, the CIV inhibitor sodium azide (50 mM) was added to assess background non-mitochondrial O₂ consumption from the addition of TMPD.

Juvenile and adult turtles endure extensive periods of anoxia (up to 4 months) when they overwinter and engage in breath-hold dives (Jackson, 2002). In mammals, one of the main problems leading to anoxic cell death is a burst of mitochondrial ROS from complex I during reoxygenation, driven by the accumulation of succinate (Chouchani et al., 2016). Interestingly, recent work has shown that adult turtles limit oxidative damage at reoxygenation in the heart, by inhibiting the maximal capacity for mitochondrial ROS production via reverse electron transport (RET), in addition to limiting the accumulation of succinate (Bundgaard et al., 2019). Therefore, we reasoned that aspects of the RET pathway could be programmed by hypoxia during development, leading to improved cardiac anoxia tolerance in adulthood. To investigate this possibility, we measured H₂O₂ production in mitochondria that were incubated with succinate alone. Under these conditions, electrons are transported to the ubiquinone pool via CII and subsequently travel in reverse to CI, leading to the production of H₂O₂.

Mitochondrial O₂ affinity was investigated by measuring O₂ consumption during the transition into anoxia. Briefly, O₂ concentration in the chambers was reduced to ∼20% of air saturation by blowing nitrogen over the respiration media. The chambers were then closed, the mitochondria were added, and malate, pyruvate, glutamate, and succinate were injected. Saturating levels of ADP were then added and mitochondria subsequently consumed all the remaining O₂, thus, entering into anoxia, while remaining in the OXPHOS state. To measure mitochondrial O₂ binding affinity, the partial pressure of oxygen where mitochondrial respiration is half maximal (P₅₀) was calculated with Prism software (v9, GraphPad, San Diego, CA, United States), using the equation V_O2 = J_Max⋅[O₂]/(P₅₀+[O₂]); where V_O2 is normalized to O₂-consumption rate, and J_Max is maximal normalized O₂-consumption rate, and [O₂] is the concentration of O₂ in the chamber.

Respiration rate and H₂O₂ production were normalized to total mitochondrial protein content (Bradford assay, VersaMax spectrophotometer: Molecular Devices, Sunnyvale, CA, United States), and H₂O₂ production was also expressed relative to the rate of O₂ consumption. To estimate mitochondrial efficiency, the respiratory-control ratio (RCR) was calculated as OXPHOS_CI/Leak_T,CI, and the OXPHOS coupling-efficiency ratio was calculated as 1-Leak_T,CI/OXPHOS_CI. The OXPHOS-control ratio was calculated as OXPHOS_CI+CII/ET_CI+CII. We also calculated the ratio of phosphorylated ADP to the atoms of consumed O₂ (P:O ratio), by making linear extrapolations of O₂ concentration during OXPHOS (immediately after ADP addition) and Leak_T (immediately after ADP depletion), as described by Illingworth; the difference in O₂ concentration at the intercepts is the total O₂ uptake. Statistical significances were determined using generalized linear models (GLMs), with sequential Sidak post hoc tests, for pairwise comparisons. Developmental O₂ was the between-group factor, and the pooled mitochondrial samples were the random factors. Data were considered significant when P ≤ 0.05.

Exposure to hypoxia during embryonic development decreased body mass in juvenile turtles, while heart mass was unaffected, leading to a larger heart-to-body-mass ratio (Table 2).

Mitochondrial preparations were of good quality, as attested by high RCRs (21.5 ± 1.4) and OXPHOS coupling-efficiency ratios (0.95 ± 0.003), with CI substrates (malate, pyruvate, and glutamate). Furthermore, mitochondrial oxygen consumption and H₂O₂ production responded to substrates and inhibitors in the expected manner (Pesta and Gnaiger, 2012; Figure 1).

Compared to their normoxic counterparts, juvenile snapping turtles previously exposed to developmental hypoxia had significantly lower levels of mitochondrial O₂ consumption, under all respiratory states and substrate combinations (Figures 2A–H). Basal levels of H₂O₂ production were also significantly lower in turtles exposed to developmental hypoxia (Figure 3), but this effect was only statistically significant in two respiratory states: Leak_T and ET states, with CI and CII substrates (Figures 3C,F). These latter effects disappeared when H₂O₂ production was normalized to O₂ consumption (Figure 4), suggesting that the lower rates of ROS production in turtles from hypoxic incubations are due to a reduced respiratory rate. We also investigated mitochondrial capacity for H₂O₂ production with RET in the presence of succinate alone. Under these conditions, mitochondria from turtles that developed in hypoxia also had significantly lower H₂O₂ production (Figure 3I), and this effect disappeared when normalized to O₂ consumption (Figure 4).

Mitochondria from turtles exposed to developmental hypoxia were significantly more coupled (H10 RCRs were 20% higher than N21) and produced more ATP per molecule of O₂ (H10 P:O ratios were 30% higher than N21) (Figures 5A,B), indicative of an improved efficiency of ATP production. However, developmental hypoxia had no effect on mitochondrial O₂ affinity (no change in P₅₀) (Figure 5C).

Compared to their normoxic counterparts, juvenile turtles that developed under hypoxic conditions had significantly lower rates of mitochondrial respiration. A reduction in aerobic capacity is a common response to chronic hypoxia in a wide range of adult vertebrates (Johnston and Bernard, 1982; Donohoe and Boutilier, 1998; St-Pierre and Boutilier, 2001; Heather et al., 2012; Hickey et al., 2012), including humans acclimatizing to high altitude (Murray and Horscroft, 2016) and animals that are genetically adapted to hypoxia (Ali et al., 2012; Horscroft et al., 2017). While it might seem counterintuitive to reduce aerobic capacity when oxygen is limiting, this strategy is believed to limit ROS production (Murray, 2009; Murray and Horscroft, 2016). ROS production is a major problem in hypoxia, because electrons slip from the chain more often when O₂ is limited and the chain becomes more reduced (Brand, 2016). Therefore, a lower level of oxidative phosphorylation will reduce the amount of electrons in the electron transport chain and limit the production of superoxide (Murray and Horscroft, 2016). In support of this concept, we also observed reduced basal levels of H₂O₂ production in juvenile turtles from hypoxic incubations, and this effect disappeared when H₂O₂ was normalized to respiratory rate. Furthermore, developmental hypoxia reduced the rate of turtle H₂O₂ production under conditions that promote RET. This finding has ecological significance, because ROS production from RET mainly occurs when tissues are reoxygenated after a period of anoxia or ischemia (Chouchani et al., 2016). Given that turtles regularly engage in long breath-hold dives and overwinter in anoxia for several months, limiting ROS production from RET reduces the likelihood of oxidative stress when turtles resurface. Taken together, our data suggest that developmental hypoxia programs a mitochondrial phenotype in turtles that is better able to cope with oxidative stress.

Although it was not a focus of this study, our experimental design gives some insight into the mechanisms reducing respiratory rate in turtles from hypoxic incubations. Given that the reduction was observed under all respiratory states and substrate combinations, the underlying mechanism is likely to be something fundamental to the entire electron transport chain (Brand and Nicholls, 2011). One possibility is that developmental hypoxia reduced the volume or cristae density of turtle cardiac mitochondria. Indeed, a downregulation of mitochondrial biogenesis is a common response found in adult vertebrates exposed to chronic hypoxia at high altitude (Murray, 2009). Additionally, a widescale reduction in respiration rates can also occur via a reduction in the expression or activity of complex IV (cytochrome c oxidase), which would limit the respiratory capacity of the entire chain. However, this modification would also increase the reduction state of the electron transport chain, leading to higher rates of ROS production (Dawson et al., 1993). Lastly, several studies have shown that environmental stress during development can epigenetically program mammalian antioxidant-defense systems, which could theoretically alter basal levels of ROS production in adulthood (Strakovsky and Pan, 2012). In this respect, it is interesting to note that hypoxia tolerance is closely related to antioxidant capacity in ectothermic vertebrates (Bickler and Buck, 2007), including turtles (Storey, 1996). Clearly, further experiments are necessary to confirm the mechanism that program respiratory rate and ROS production in turtles from hypoxic incubations. Future research should be directed toward investigating the effects of developmental hypoxia on turtle mitochondrial morphology with electron microscopy, as well as the expression or activity of electron-transport-chain complexes and antioxidants. Furthermore, it would be interesting to measure the activity of anaerobic enzymes, such as lactate dehydrogenase, which are known to be altered by developmental hypoxia in some ectotherms (Matschak et al., 1998; Ton et al., 2003; Anderson and Podrabsky, 2014).

In order to offset the reduced aerobic capacity that accompanies chronic hypoxia, many adult animals increase the efficiency of mitochondrial ATP production by reducing proton leak and increasing the P:O ratio (Gnaiger et al., 2000; Semenza, 2007; Zhang et al., 2013; Murray and Horscroft, 2016). Interestingly, we found that developmental hypoxia programmed these same attributes in turtle ventricular mitochondria. The P:O ratio is a measure of the number of ATP molecules synthesized by oxidative phosphorylation for each oxygen atom reduced. Although oxygen consumption is routinely used as a proxy for ATP production, the amount of ATP generated per unit of oxygen consumed can vary significantly between individuals, and it is well established that environmental stress can alter this parameter (Salin et al., 2015). Improvements in the P:O ratio can have major implications for animal life-history traits, including growth, reproduction, and lifespan (Barros et al., 2004; Stier et al., 2014; Salin et al., 2015). While several factors can influence the P:O ratio, the most common mechanism that accounts for individual variation is proton leak. This phenomenon refers to the process whereby protons leak across the mitochondrial membrane without the generation of ATP (Divakaruni and Brand, 2011). Importantly, proton leak can account for up to 25% of an animal’s metabolic rate, so a reduction in this parameter can greatly increase the efficiency of ATP production. Therefore, the elevation in the P:O ratio we observed in turtles from hypoxic incubations could be driven by a decrease in proton leak. In support of this contention, we found that developmental hypoxia significantly lowered levels of leak respiration, leading to an increase in mitochondrial coupling (higher RCR values). Interestingly, we found the same result in juvenile alligators exposed to developmental hypoxia (Galli et al., 2016), suggesting that proton-leak pathways are a common target for developmental programming. Future research should, therefore, determine the mechanism driving the reduction in proton leak, which could include changes to the inner mitochondrial membrane, like the composition of the lipid bilayer, expression of the adenine nucleotide translocase, and/or suppression of uncoupling proteins (Zhao et al., 2019). Nevertheless, it should be noted that a decrease in proton leak in the absence of other changes will lead to an increase in ROS (Brand, 2000).

Several studies have shown that mitochondrial P₅₀ is lower in hypoxia-tolerant vs. hypoxia-sensitive species (Scott et al., 2011; Zhang et al., 2013; Lau et al., 2017), and it can be decreased in some adult vertebrates after acute and chronic episodes of hypoxia (Gnaiger et al., 1998b; St-Pierre et al., 2000b). A low P₅₀ allows the mitochondria to extract oxygen from the cytosol more effectively, which is particularly important during periods of hypoxia, when the concentration gradient for O₂ has been reduced (Gnaiger, 2001). Several mechanisms can account for these differences, including changes in the abundance and activity of complex IV (the site of oxygen reduction), altered flux through the electron transport chain relative to complex-IV capacity, and differences in the mitochondrial membrane potential (Bienfait et al., 1975; Verkhovsky et al., 1996; Gnaiger et al., 1998a). Nevertheless, in the present study, we did not find any differences in mitochondrial O₂-binding affinity between the developmental cohorts.

Adult vertebrates acclimatizing to hypoxic environments undergo mitochondrial remodeling to enhance oxygen delivery, maintain ATP balance, and limit oxidative stress. The mechanisms underlying this remodeling include a reduction in aerobic capacity and proton leak, a reduced basal ROS production, and an increase in the P:O ratio. For the first time, we show that these same attributes can be permanently programmed in turtles by exposure to hypoxia during embryonic development. The molecular mechanisms underlying this response remain to be investigated, but they could include hypoxia-inducible factor (HIF) signaling. Indeed, HIF-1 activation is known to coordinate mitochondrial remodeling in adult vertebrates exposed to hypoxia (Murray and Horscroft, 2016; Thomas and Ashcroft, 2019), and this pathway is activated in hypoxic zebrafish embryos, leading to enhanced hypoxia tolerance in adulthood (Robertson et al., 2014). In addition, mammalian mitochondria can be programmed during development via epigenetic alterations to the nuclear and mitochondrial genomes, changes to the mitochondrial–telomere axis, and the accumulation of mitochondrial DNA defects (Gyllenhammer et al., 2020). Whatever the mechanism, we suspect that mitochondrial programming in turtle embryos will improve hypoxia tolerance in adulthood, which would be beneficial for breath-hold diving and overwintering in anoxic environments (Jackson, 2000). Indeed, our recent study suggests turtles from hypoxic incubations have improved cardiac anoxia tolerance (Ruhr et al., 2019). Therefore, developmental hypoxia could represent an important environmental cue for turtles that primes their physiology for a future in low-oxygen environments.