Do naked mole rats accumulate a metabolic acidosis or an oxygen debt in severe hypoxia?

Although mammals are largely intolerant of hypoxia, a few species inhabit hypoxic niches (e.g. some fossorial species and species resident to high altitudes). These animals have evolved complicated suites of physiological and molecular adaptations that enable them to thrive in low O₂ environments (Dzal et al., 2015). The key to tolerating prolonged hypoxia is to match metabolic demand to reduced energy (O₂) supply (Buck and Pamenter, 2006, 2018; Hochachka, 1986; Hochachka et al., 1996), with most hypoxia-tolerant animals exhibiting robust decreases in metabolic rate when O₂ supplies are limited (Dzal et al., 2015; Guppy and Withers, 1999). Conversely, hypoxia-intolerant animals are generally unable to sufficiently reduce their metabolic rate during hypoxia to accommodate the reduced O₂ supply and instead increase ventilation in a largely futile attempt to maintain O₂ supply. This is an energetically expensive strategy, particularly when breathing air with reduced O₂ content (Dzal et al., 2015; Pamenter and Powell, 2016). As a result, and in the absence of sufficient metabolic rate depression in severe hypoxia, most mammals rely on anaerobic energy production to make up the gap between energy demand and their ability to produce energy aerobically (Buck and Pamenter, 2006; Hochachka, 1986).

Anaerobic metabolism (primarily glycolysis) usually leads to the accumulation of lactate and, upon reoxygenation, this accumulated lactate must either be oxidized to CO₂ and H₂O via the tricarboxylic acid cycle, converted to glucose/glycogen via gluconeogenesis in the liver and/or muscle, or utilized in protein synthesis (Warren and Jackson, 2008). Oxidation is the predominate end-point for lactate that accumulates during hypoxia in most mammals and it yields 15 molecules of ATP per molecule of lactate oxidized; however, this process also requires metabolic O₂. As a result, animals that utilize this pathway typically hyperventilate and exhibit a high rate of O₂ consumption (V̇_O2) during reoxygenation, reflecting repayment of the O₂ debt accrued during hypoxic exposure (Robin, 1980). This phenomenon has been observed in both individual organs and whole animals across a wide range of adult vertebrate species following exposure to hypoxia (Coffman, 1963; Lewis et al., 2007; Maxime et al., 2000; Plambech et al., 2013; Svendsen et al., 2012). In all of these species, the elevated metabolic rate upon reoxygenation correlates with the level of metabolic acidosis and elevated lactate concentration in the blood and/or tissues. Little is known regarding lactate accumulation and clearance in hypoxia-tolerant mammals, but animals that are capable of robust metabolic rate depression during acute hypoxia typically do not exhibit an O₂ debt upon reoxygenation, and lactate accumulation is minimal (Frappell et al., 1991). For example, in hypoxia-tolerant neonatal mammals, instead of metabolic rate overshooting baseline levels during reoxygenation from hypoxia (15% O₂), V̇_O2 quickly returns to pre-hypoxic levels and lactate accumulation is only observed in more severe hypoxia (10% O₂) (Frappell et al., 1991).

Naked mole rats (NMRs; Heterocephalus glaber Rüppell 1842) are the most hypoxia-tolerant mammal presently identified and tolerate minutes of anoxia, hours at 3% O₂ and days to weeks at 8% O₂ (Chung et al., 2016; Pamenter et al., 2015, 2018; Park et al., 2017). In acute severe hypoxia (3% O₂), the V̇_O2 of adult NMRs decreases up to 85% (Pamenter et al., 2018). Although this degree of V̇_O2 suppression is not remarkable among hypoxia-tolerant species (Guppy and Withers, 1999), it is important to note that other adult mammals that are capable of similar or more extreme metabolic rate suppression in severe hypoxia typically enter into a coma- or torpor-like state until O₂ levels increase (Guppy and Withers, 1999; Hayden and Lindberg, 1970). Conversely, NMRs remain awake and active (albeit to a reduced degree) in hypoxia (Houlahan et al., 2018; Ilacqua et al., 2017; Kirby et al., 2018). This combined metabolic and behavioural phenotype (robust metabolic rate depression combined with the retained ability to ambulate and interact within their environment) may be a unique characteristic of this remarkable species. Unfortunately, previous studies have not evaluated the physiological and biochemical mechanisms that support the NMR's remarkable ability to stay active in severe hypoxia, leaving an important question unanswered: do NMRs employ anaerobic metabolism to support behavioural activity during metabolic suppression?

In the present study, we comprehensively evaluated physiological and behavioural variables in NMRs across a wide range of hypoxic exposures and also during reoxygenation. We hypothesized that NMRs rely on anaerobic metabolism during hypoxia and thus would exhibit signs of an accumulated O₂ debt after hypoxia. We predicted that they would exhibit a metabolic acidosis and a deficit in energetic substrate pools (e.g. ATP) during hypoxia, and an overshoot in V̇_O2 during reoxygenation. To test our hypothesis, we exposed NMRs to progressive hypoxia (21–3% O₂), followed by a 1 h reoxygenation period, and measured breathing and heart rates (i.e. the primary components of O₂ supply), behavioural activity, body temperature (T_b) and metabolism (i.e. the primary consumers of O₂) throughout these exposures. Additionally, at the end of the hypoxic protocol, a subset of animals were euthanized to collect blood and tissues from which we analyzed markers indicative of metabolic acidosis and energy balance.

NMRs were group-housed in interconnected multi-cage systems at 30°C and 21% O₂ in 50% relative humidity with a 12 h:12 h light:dark cycle. Animals were fed fresh tubers, vegetables, fruit, and Pronutro cereal supplement ad libitum. Animals were not fasted prior to experimental trials. All experimental procedures were approved by the University of British Columbia Animal Care Committee (Protocol A13-0248) in accordance with the Animals for Research Act and by the Canadian Council on Animal Care. All experiments were performed during daylight working hours in the middle of the animals' 12 h:12 h light:dark cycle. NMRs that are housed within colony systems, as our experimental animals are, do not exhibit circadian rhythmicity of general locomotor activity (Riccio and Goldman, 2000b) and exhibit inconsistent rhythmicity of T_b and metabolic rate (Riccio and Goldman, 2000a); however, significant changes in these latter variables were only reported in animals during the nocturnal phase of their circadian cycle, with no significant changes observed during the daylight period of this cycle. Therefore, since we only ran experimental trials during the daylight period, we did not expect our results to be influenced by circadian rhythms. We examined physiological responses to environmental hypoxia in non-breeding NMRs that were 1–2 years old. Non-breeding (subordinate) NMRs do not undergo sexual development or express sexual hormones and thus we did not take sex into consideration when evaluating our results (Holmes et al., 2009).

Twenty-two adult, non-breeding male and female NMRs, weighing 39.0±4.6 g (mean±s.e.m.) were individually placed, unrestrained, inside a 450 ml Plexiglass experimental chamber (the animal chamber), which was in turn placed into an environmental chamber held at ∼28–29°C. The temperature of the animal chamber was recorded continuously throughout the experiment using multiple iButtons that recorded ambient temperature (T_a) at a frequency of one measurement min⁻¹ (Maxim Integrate, Chandler, QC, Canada). Animals were provided with a thin layer of bedding on the floor of the experimental chambers. The animal chamber was sealed and constantly ventilated with gas mixtures, set to the desired fractional gas composition by calibrated rotameters (Praxair, Mississauga, ON, Canada). The advantage of this open-flow system is that it prevents the depletion of O₂ and accumulation of metabolic CO₂ by flushing the animal chamber with fresh gas, and it allows for continuous and simultaneous monitoring of metabolic and ventilatory variables. Inflowing gas was provided at a flow rate of 110 ml min⁻¹, as assessed by a calibrated mass flow meter (Praxair Technologies Inc., Danbury, CT, USA). The analyzers were calibrated prior to each trial with 20.95% O₂ and 1.5% CO₂, both balanced with N₂, and with 100% N₂ gas mixtures. During experimentation, animal ventilation caused pressure fluctuations due to humidity and the warmth of air in each expired breath, which were compared to the pressure of a reference chamber that was identical to the animal chamber. Continuous monitoring by a differential pressure transducer (Validyne, Northridge, CA, USA) connected between the animal and reference chambers allowed for the detection of breaths. Breathing frequency (f_R) was determined by counting these ventilation-induced pressure oscillations. All ventilatory variables are reported at body temperature, pressure, saturated (BTPS).

Heart rate, arterial O₂ saturation (S_a,O2) and behavioural activity (active/inactive) were also measured in vivo using pulse oximetry collars (STARR Life Sciences Corp., Oakmont, PA, USA). Blood pH (pH_a), partial pressure of CO₂ (P_CO2; mmHg), partial pressure of O₂ (P_O2; mmHg), base excess, [HCO₃⁻] (mmol l⁻¹), total CO₂ in the blood (T_CO2, which includes free CO₂ in the plasma and bound to proteins, as well as in the form of bicarbonate, carbonic anions and carbonic acid), [Na⁺] (mmol l⁻¹), [K⁺] (mmol l⁻¹), intracellular [Ca^{2 +}] ([Ca²⁺]_i; mmol l⁻¹), [glucose] (mg dl⁻¹), haemoglobin concentration ([Hb]; mmol l⁻¹) and haematocrit (Hct; %) were analyzed ex vivo from blood at 30°C (which is approximately the NMR T_b at our experimental T_a in hypoxia) using CG8⁺ cartridges with the i-STAT VetScan Analyzer (Abaxis, Union City, CA, USA). Haematocrit was also calculated from the ratio of the volume of packed red blood cells to the total blood volume after centrifugation in micro-haematocrit tubes (∼10 µl). Haemoglobin concentration was also determined by the cyanmethaemoglobin method using Drabkin's reagent (Sigma-Aldrich, St Louis, MO, USA) and a haem-based extinction coefficient of 11 mmol⁻¹ cm⁻¹ at a wavelength of 540 nm.

In a separate experiment, T_b was measured from a group of eight NMRs held individually in cages at 28.4°C and 70% relative humidity and exposed to the same progressive hypoxia/reoxygenation treatment protocol as described above. In these experiments, T_b was measured using a handheld radio frequency identification (RFID) reader that scanned individual NMRs instrumented with subcutaneous RFID microchips (Destron Fearing, Dallas, TX) every 30 min, as described previously (Kirby et al., 2018). Measurements were taken when the body region containing the RFID microchip was not in contact with the floor to avoid skewed readings.

Animals were transferred to the animal chamber under normoxic conditions and baseline recordings were obtained for 1 h. The inflowing gas composition was then switched progressively to 9, 7, 5 and 3% O₂ for 1 h each. Finally, the inflowing gas concentration was switched back to 21% O₂ for 1 h. Following experimentation, animals were returned to their colonies. At the end of the hypoxic protocol, a subset of animals were euthanized by conscious cervical dislocation followed immediately by decapitation to assess blood and tissue parameters. Tissues were removed within 2 min of decapitation, snap-frozen in liquid nitrogen and then stored at −80°C for future analysis.

Frozen temporalis muscle, liver, heart and brain tissue were ground into a fine powder under liquid nitrogen using mortar and pestle. Tissues were aliquoted and used for intracellular pH (pH_i), creatine phosphate (CrP), ATP and glucose analysis. For pH_i, frozen tissue was mixed with a metabolic inhibitor solution [150 mmol l⁻¹ KF; 6 mmol l⁻¹ nitrilotriacetic acid (Sigma)] using a liquid-nitrogen-cooled 18-gauge needle, and vortexed. Samples were centrifuged at 10,000 g for 1 min, then the supernatant was removed and incubated at 30°C in a water bath for 15 min. Following the incubation period, pH_i was recorded using a calibrated semi-micro pH electrode (Cole-Parmer, IL). The pH probe was re-calibrated every four samples and held in a metabolic inhibitor solution between sample measurements.

Frozen tissue for CrP, ATP and glucose analysis was homogenized in 8% HClO₃ and stored on ice. Samples were neutralized with NaOH, and the supernatant was removed and aliquoted for either CrP and ATP, or glucose determination. CrP and ATP were determined using a modified technique from Bergmeyer (1965). A total of 100 µl of neutralized extract was mixed with 150 µl of assay buffer [7 ml buffered base solution (50 mmol l⁻¹ Tris base; 8 mmol l⁻¹ MgCl₂·6H₂O; 50 mmol l⁻¹ glucose; 2 mmol l⁻¹ EDTA·Na₂·H₂; mercaptoethanol; adjusted to pH 8.0); 10 mg βNADP; 20 µl glucose 6-phosphate dehydrogenase (3520.3 U ml⁻¹); Sigma]. Samples were incubated at 37°C for 5 min and absorbance was recorded at 340 nm (A₁). Hexokinase (150 U ml^–1; Sigma) was added (2 µl), and samples were then incubated at room temperature for 15 min before recording absorbance (A₂). A total of 5 mmol l⁻¹ adenosine diphosphate (ADP) (20 µl) was added, with samples left to incubate at room temperature for 5 min, and absorbance was measured (A₃). Finally, 2 µl of creatine kinase (10 mg ml⁻¹; Sigma) was added, and samples were incubated at room temperature for 20 min before recording absorbance (A₄). ATP was calculated as A₂–A₁ and CrP as A₄–A₃. Standards between 0–500 µmol l⁻¹ and 0–2000 µmol l⁻¹ were used for ATP and CrP, respectively.

Glucose was determined using 20 µl of neutralized sample and adding 100 µl of assay buffer [17 ml of 4 mmol l⁻¹ MgCl₂·6H₂O 83 mmol l⁻¹ NaH₂PO₄; 20.67 mg βNADP; 14.88 g ATP; 10 µl glucose 6-phosphate dehydrogenase (1.5 U ml⁻¹)]. Samples were incubated for 5 min at 37°C and absorbance was read at 340 nm (A₁). Hexokinase (2 µl; 2.25 U ml⁻¹; Sigma) was added to each sample, which was incubated for 20 min at 37°C, and absorbance was recorded (A₂). Glucose was calculated as A₂–A₁ using a standard curve between 0–5 mmol l⁻¹. Adenylate energy charge was calculated using [CrP], [ATP] and pH_i measurements. Estimates of inorganic phosphate (Pi) were determined as the inverse of CrP hydrolysis, assuming a normoxic Pi concentration of approximately 1 µmol g⁻¹ wet mass for each tissue (Portner et al., 1996). Assuming equilibrium of the creatine kinase and adenylate kinase reactions, and stoichiometry of creatine phosphate and Pi (Lawson and Veech, 1979), ADP and adenosine monophosphate (AMP) were estimated based on the formula described previously (Lanza et al., 2006). Equilibrium constants for creatine kinase and adenylate kinase were corrected for pH and temperature, while assuming a concentration of free Mg²⁺ of 0.001 mol l⁻¹ (Golding et al., 1995; Teague et al., 1996). Adenylate free energy charge was calculated as described previously (Atkinson, 1968).

T_b, incurrent and excurrent O₂ and CO₂ concentrations, and the ventilation-induced pressure signals were recorded and analyzed using PowerLab data acquisition hardware and LabChart software (AD Instruments Ltd, Colorado Springs, CO, USA). Pulse oximetry data were recorded using MouseOx Plus (STARR Systems). From the recorded signals, and calculated dependent variables, we determined average T_b, V̇_O2, V̇_CO2, RER, f_R, S_a,O2, heart rate and behavioural activity state for the last 10–15 min of each O₂ exposure (21, 9, 7, 5, 3% O₂), and for the last 1–2 min of each 10 min period of post-hypoxia recovery (10, 20, 30, 40, 50, 60 min reoxygenation). Inflowing gas concentrations were measured before and after each O₂ exposure. Upon reoxygenation, we were unable to measure V̇_O2 for ∼20 min due to a combination of the volume of the animal chamber, the slow rate of gas flow through the animal chamber (and thus the long washout period), and the large difference between O₂ levels in the incurrent gases supplied during hypoxia and normoxia treatments (3 and 21%, respectively). Conversely, incurrent gas CO₂ levels did not differ between hypoxia and normoxia, and so we were able to measure V̇_CO2 during the early reoxygenation phase as an indirect measure of metabolic rate.

Statistical analyses were performed using R (http://www.R-project.org/) and GraphPad Prism 7.0. We used linear mixed effects models (lme4 and lsmeans package) (Bates et al., 2015; Lenth, 2016) to account for repeated sampling of the same individual with changes in O₂ exposure, with individual treated as a random effect. All physiological and behavioural variables met the assumptions of normality, homogeneity of variances, linearity and independence. We entered the level of inspired O₂ and body mass as fixed effects in our models. We ran a Tukey's post hoc test using the lsmeans package to determine differences between inspired O₂ levels and to correct for multiple pairwise comparisons to test for differences between normoxia and hypoxia-treated samples. Blood and tissue data were analyzed using Student's t-tests. All results are presented as means±s.e.m., with statistical significance set as P<0.05.

S_a,O2, heart rate and behavioural activity were measured using a pulse oximeter. NMRs are very flexible and agile and, in most experiments, they were able to free themselves from the pulse oximeter collar or to bite through the electric lead. As a result, S_a,O2, heart rate, and activity data are presented only for the subset of animals exposed to progressive hypoxia from which pulse oximeter data were obtained throughout the experiment (i.e. for each level of O₂). Arterial blood was fully saturated with O₂ in normoxia (21% O₂) and saturation significantly decreased with progressive hypoxia (Fig. 1A, left panel; F_4,35=8.153, P<0.0001; n=8). S_a,O2 decreased significantly to ∼90% when NMRs were breathing 9% O₂, further decreased to 60–70% when animals were breathing 7–5% O₂, and then dipped significantly to ∼40% saturation when NMRs were breathing 3% O₂. Upon reoxygenation, S_a,O2 rapidly returned to baseline levels (within 10 min) (Fig. 1A, right panel).

Breathing frequency decreased by ∼25% (from ∼90 breaths min⁻¹ in normoxia) in 9% O₂ and then decreased further and plateaued at ∼60 breaths min⁻¹ at inhaled O₂ levels ranging from 7 to 3% O₂ (Fig. 2B, left panel; F_10,116=4.13, P<0.00001; n=22). Upon reoxygenation (within 10 min), f_R decreased further to ∼50% of the normoxic breathing rate (Fig. 1B, right panel). Over the ensuing 50 min of reoxygenation, f_R continued to increase back to the pre-hypoxia baseline and was not statistically different from normoxic levels after 50 min of reoxygenation.

Heart rate significantly decreased in progressive hypoxia. When NMRs were breathing hypoxic gas containing 9 and 7% O₂, heart rate decreased by 17% from the normoxic levels of ∼220 beats min⁻¹ (Fig. 1C, left panels; n=8), and then decreased by 30 and 40% of normoxic levels when animals were breathing 5 and 3% O₂, respectively (F_4,35=13.65, P<0.0001). Upon reoxygenation, heart rate did not fall further but increased more rapidly than f_R, such that heart rate was not different from pre-hypoxia normoxic levels after 30 min of reoxygenation (Fig. 1C, right panel).

Behavioural activity, as indicated by total time active min⁻¹, was not different at 9% O₂ relative to normoxic levels but decreased by ∼65% when animals were breathing 7–5% O₂ and by 85% of normoxic levels when animals were breathing 3% O₂ (Fig. 1D, left panel; F_4,35=20.1, P<0.0001; n=8). Following reoxygenation, NMRs appeared to enter into a fully comatose state for the first 15–20 min (Fig. 1D, right panel). Animals were completely motionless and, in a subset of experiments, we were compelled to open the chamber to check on the health of the animals. Animals were non-responsive to auditory, visual or mechanical stimuli. Breaths were difficult to detect visually. However, following ∼20–30 min of reoxygenation, all animals regained consciousness and ambulated normally.

In normoxia, NMR T_b was ∼3–4°C above T_a (Fig. 2; n=8). With the onset of acute hypoxia, T_b decreased by ∼2°C at 9% O₂, and then progressively decreased by ∼0.5°C at each further level of progressive hypoxia, finally reaching a minimal level that was within 0.5°C of T_a in animals breathing 3% O₂ (F_7,63=6.10, P<0.0001). Upon reoxygenation, T_b recovered very slowly and, after 1 h of reoxygenation, T_b had increased ∼1°C from the lowest hypoxic T_b (28.7±0.04°C) but was still ∼3°C below the pre-treatment normoxic temperature (32.2±0.02°C).

NMRs exhibited initial decreases in V̇_O2 and V̇_CO2 of 55–60% with the onset of hypoxia (9% O₂ inhaled; Fig. 3A,B, left panels; F_8,140=40.88, P<0.00001 for V̇_O2, F_10,165=32.08, P<0.00001 for V̇_CO2; n=22 for each). These changes increased in magnitude with each subsequent level of hypoxia such that the maximum suppression was observed from animals breathing 3% O₂ (87% decrease in V̇_O2 and 80% decrease in V̇_CO2).

Upon reoxygenation, we found that V̇_CO2 dropped further below the maximum suppression observed in 3% O₂ such that V̇_CO2 was suppressed by 96 and 90% following 10 and 20 min of reoxygenation, respectively (Fig. 3B, right panel). This apparent metabolic rate suppression decreased progressively through the reoxygenation period and V̇_O2 similarly returned slowly to pre-hypoxia baseline levels from 30 min of reoxygenation onward (Fig. 3B, right panel).

The NMR RER was 0.74 in normoxia (Fig. 4, left panel; n=22), indicating that they predominately use lipid substrates as a metabolic fuel source, with a lesser contribution from carbohydrates. With the onset of progressive hypoxia, NMRs immediately exhibited a shift towards total reliance on carbohydrate substrates (F_8,118=25.00, P<0.00001). Upon reoxygenation, NMRs expressed RER values that were well below 0.7 (Fig. 4, right panel), indicating that the animals were not in steady state; however, their RER values approached pre-hypoxic baseline levels towards the end of the reoxygenation period.

To better understand fuel switching following hypoxic exposure, we evaluated metabolic parameters in the blood, brain, heart, temporalis muscle and liver in a subset of NMRs at the end of the progressive hypoxic protocol and compared the resulting values to samples taken from NMRs not exposed to progressive hypoxia (Tables 1, 2). We found that blood glucose levels nearly doubled in acute hypoxia (Table 1), whereas liver glucose levels were markedly lower (Table 2). Together, these observations indicate mobilization of liver glucose stores to support enhanced carbohydrate metabolism when O₂ is low. Glucose levels were not changed in heart, brain or muscle.

Interestingly, we did not observe any signs of a metabolic acidosis in blood or tissues derived from hypoxic animals. Specifically, blood pH was elevated and P_CO2 was lower in hypoxic animals, whereas base excess and T_CO2 were unchanged with hypoxic treatment (Table 1). Brain pH_i was lower in hypoxic animals but pH_i was not altered in heart, liver or muscle between treatment groups (Table 2).

Importantly, the concentrations of ATP of brain, heart, muscle and liver were not different between normoxic and hypoxic animals (Table 2), indicating that NMRs maintained energy balance during their exposure to progressive hypoxia. However, CrP levels were higher in brain and muscle from hypoxic animals, while free adenylate charge was significantly elevated in brain, but not significantly altered in other tissues. Derangements of blood [Na⁺], [K⁺] and [Ca²⁺] levels were not observed, indicating that cellular integrity was maintained in the organism (Table 1). Haematocrit and [Hb] measurements were also similar between treatment groups (Table 1).

In the present study we examine the hypothesis that NMRs utilize a combined strategy of metabolic rate suppression and upregulation of glycolytic metabolism during acute hypoxic exposure. We predicted that NMRs would exhibit hallmarks of a metabolic acidosis during acute hypoxia and of O₂ debt repayment during the reoxygenation phase following the return to normoxia. We report several significant findings. First, during progressive hypoxia, the oxidative metabolism of NMRs (as measured indirectly through V̇_O2 and V̇_CO2) is markedly suppressed by as much as 85% of the normoxic rate. This metabolic rate suppression is accompanied by significant reductions in heart rate, f_R, T_b and behavioural activity. During reoxygenation, NMRs enter into a coma-like state lasting 15–20 min, during which they are entirely inactive, are largely non-responsive to external stimuli and exhibit further decreases in their metabolic rate (>95% suppression of V̇_CO2). Importantly, at the end of our hypoxic protocol, ATP concentrations in major organs are maintained and adenylate energy charge is sustained or increased in all tissues examined at normoxic levels, indicating that net energy production meets energy demand under hypoxic conditions. In addition, blood and tissue pH levels are largely unchanged after acute hypoxia, suggesting the absence of a metabolic acidosis. Finally, our RER data indicate that NMRs undergo a fuel substrate switch from a reliance on mixed fuel (predominately lipids) in normoxia towards complete reliance on carbohydrates in severe hypoxia. This conclusion is supported by our observations of the mobilization of glucose from the liver into the blood. Taken together, these data indicate that NMRs may not recruit anaerobic metabolic processes or accumulate an O₂ debt during acute hypoxia but may instead rely on substrate modifications to enhance the efficiency of energy production, and thereby support sustained consciousness and activity despite a remarkably deep suppression of metabolic rate.

These data are not unequivocal, however. An imbalance in the rates at which V̇_O2 and V̇_CO2 return to normoxic levels during reoxygenation hint at the possibility that NMRs do utilize anaerobic metabolism during hypoxia in combination with a tissue and/or blood buffering capacity that masks typical markers of metabolic acidosis, and prioritize the synthesis of glucose from lactate instead of lactate oxidation during recovery. Alternatively, NMRs may rely upon fructose metabolism in hypoxia, as they do in anoxia (Park et al., 2017), to support carbohydrate metabolism without developing a significant lactate accumulation. Finally, several mechanisms related to lactate handling in vivo may mitigate the effects of lactate accumulation during hypoxia. For example, changes in the expression of monocarboxylate transporters in anaerobically active tissues or upregulation of gluconeogenesis in the liver may play an important role (Choi et al., 2005); further experiments are warranted to test these possibilities.

NMRs exhibit a unique physiological and behavioural response to acute hypoxia and reoxygenation. Most hypoxia-intolerant adult mammals increase their activity in hypoxia in an attempt to escape to a more hospitable environment, whereas most hypoxia-tolerant vertebrates and small rodents (for the most part) enter into a torpor-like state when O₂ is limited (typically 3–8% O₂; Hayden and Lindberg, 1970). Physiologically, all adult mammals exhibit a hypoxic ventilatory response mediated by increased ventilation, and/or decreased metabolic rate, immediately following hypoxic exposure (Pamenter et al., 2014; Pamenter and Powell, 2016), which is typically accompanied by an increase in heart rate (Dzal et al., 2015). Upon reoxygenation, most animals continue to hyperventilate to blow off CO₂ as part of the repayment of the O₂ debt (i.e. as lactate is oxidised to CO₂+H₂O). Conversely, NMRs exhibit a behavioural pattern that is the inverse of this typical adult mammalian response: they reduce their physical activity in hypoxia but remain responsive to their environment. This behavioural decrease is accompanied by decreases in f_R and heart rate, and also a reduction in T_b to near T_a. These findings agree with previous behavioural and physiological studies from our laboratory (Chung et al., 2016; Houlahan et al., 2018; Ilacqua et al., 2017; Kirby et al., 2018; Pamenter et al., 2015). Upon reoxygenation, NMRs enter into a coma-like state (whereas other mammals emerge from a torpor-like state in this period), during which ventilation is further suppressed but heart rate increases.

The decreases in physiological function described above likely all reduce energy demands during hypoxia and support metabolic rate suppression. Indeed, a key finding of our study is that NMRs reduce their metabolic rate by a remarkable 85% in acute severe hypoxia, which is very near the threshold at which their energetic demands could be entirely met through anaerobic metabolism (∼90–95% suppression) (Mookerjee et al., 2017). However, upon reoxygenation, NMRs do not exhibit an overshoot in their metabolic rate, which is a hallmark of O₂ debt repayment (Coffman, 1963; Lewis et al., 2007; Maxime et al., 2000; Plambech et al., 2013; Svendsen et al., 2012). These findings suggest that they do not rely on anaerobic energy production to a large degree in hypoxia (but see discussion of fructose metabolism, below). In support of this, we do not observe signs of metabolic acidosis in heart, muscle, liver or blood. Specifically, pH values for these samples are not changed by hypoxic exposure, and neither are blood base excess, [HCO₃⁻] or T_CO2.

Recently, several groups have proposed that NMRs retain a neonatal phenotype into adulthood and that this neotenic phenotype may contribute to the hypoxia tolerance of this species (Pamenter et al., 2015; Penz et al., 2015; Peterson et al., 2012). However, the fact that NMRs maintain metabolic rate depression for a prolonged period following reoxygenation is in striking contrast to the phenotype of neonatal mammals, in which metabolic rate rapidly returns to normoxic levels (Frappell et al., 1991). Indeed, many neonates exposed to severe hypoxia or to ischemia develop a metabolic acidosis and exhibit the accumulation of an O₂ debt that is then repaid upon reoxygenation (Fahey and Lister, 1989; Frappell et al., 1991). However, the O₂ debt in neonates tends to be less than that observed in adults. Therefore, the neotenic response should be considered as a blunting of the O₂ debt relative to adults, rather than the absence thereof.

Another important finding of our study is the observation that NMRs exhibit a sustained shift of metabolic fuel substrate use from primarily lipids (with some contribution from carbohydrates) in normoxia to a greater reliance on carbohydrates in severe hypoxia. The mobilization of liver glucose stores and increase in blood [glucose] during hypoxic exposure supports an increased reliance on carbohydrate metabolism. This response is similar to that observed in other hypoxia-tolerant rodents, including murine species that live at high altitudes (Schippers et al., 2012). The ATP yield per mole of O₂ catabolized is 15–30% higher when derived from carbohydrates than from lipids (due to the higher energetic costs of breaking high-energy bonds in lipids) and thus a greater reliance on carbohydrate fuels would increase energetic efficiency during acute hypoxia (Hochachka, 1985). The stability seen in our energy charge measurements supports this efficiency, as the metabolic suppression by NMRs may outweigh reductions in ATP production. This strategy would be particularly useful in NMRs because they likely experience hypoxia transiently during intense exercise (e.g. when digging tunnels) and thus would have opportunities to replenish carbohydrate stores while in regions of their burrows with higher levels of O₂, when they rely more heavily on lipid energy stores (as indicated from our normoxic RER calculations). This fuel switch, along with a remarkable metabolic rate depression, may limit the need to recruit anaerobic metabolism and thus the need to pay off O₂ debt post-hypoxia.

NMRs mobilize both glucose and fructose in acute anoxia and are able to metabolize fructose in anoxia (Park et al., 2017); it is possible that fructose contributes to carbohydrate metabolism in severe hypoxia as well. However, the anoxia-induced elevation of glucose in NMRs in vivo is considerably greater than that of fructose (Park et al., 2017), and the source of metabolic fructose in NMRs exposed to anoxia is unclear. It is important to note that, although remarkable for a mammal, the ability of NMRs to tolerate a few minutes of anoxia pales in comparison to animals that are truly anoxia tolerant and able to survive in anoxia for days to months (Bickler and Buck, 2007). NMRs should not be considered anoxia tolerant but are more accurately able to resist the lethal effects of anoxia for a short period of time. In this context, it is notable that fructose metabolism also occurs in pathophysiological heart disease in humans (Mirtschink et al., 2015), and it is possible that the ability of NMRs to mobilize and metabolize fructose in acute anoxia reflects a ‘last-ditch’ effort to survive in such harsh environmental conditions. If this were the case, we would not expect NMRs to mobilize and metabolize fructose while in sublethal levels of hypoxia. Nonetheless, further experiments to assess changes in the fructose transporter Glut5 and liver fructokinase would be fascinating.

It is also important to note that lactate transport strategies may be beneficially modulated during acute hypoxia in this species, but our study does not address this question. Specifically, monocarboxylate transporters are transmembrane proteins that facilitate the transport of lactate across cell membranes. The expression of monocarboxylate transporter 4 in particular is sensitive to hypoxia and is differentially modulated following a day of hypoxia exposure in various tissues in rats and zebrafish (McClelland and Brooks, 2002; Ngan and Wang, 2009; Ullah et al., 2006). Clearance of lactate from NMR tissues that preferentially utilize anaerobic metabolism during hypoxia would support sustained glycolysis and limit the inhibitory effect of accumulated lactate on glycolytic throughput. Studies to date that have examined hypoxia-mediated changes of monocarboxylate transporter expression in vivo have largely found that this change follows several days of hypoxia exposure and not hours. Therefore, it is unlikely that significant insertion of this protein occurs within a few hours of acute hypoxia in NMR tissues. However, higher endogenous expression of these transporters may prime NMR tissues for prolonged anaerobic metabolism.

It is informative to consider lactate management strategies employed by other hypoxia-tolerant vertebrates. In goldfish, which are among the most anoxia-tolerant vertebrates identified and which exhibit marked metabolic rate suppression during periods of anoxia or hypoxia (Vanwaversveld et al., 1989), O₂ debt is not observed following up to 3 h of anoxia but manifests following longer anoxic exposures (Vandenthillart and Verbeek, 1991). In this species, the O₂ debt is hypothesized to be related to the recovery of glycogen pools (Mandic et al., 2008) rather than to the oxidation of lactate because goldfish produce ethanol as an alternative end-product to lactate and thus do not face the same metabolic challenges associated with lactate clearance as do most other vertebrates following exposure to hypoxia (Shoubridge and Hochachka, 1980). Similarly, western painted turtles, which are the most anoxia-tolerant vertebrates, rely entirely on anaerobic metabolism while overwintering in anoxic conditions and accumulate extremely high levels of plasma lactate (Ultsch and Jackson, 1982). However, upon reoxygenation these animals do not exhibit significant lactate oxidation and instead the majority of the accumulated lactate is utilized to restore glycogen pools depleted by the reliance on anaerobic metabolic processes for months (Jackson et al., 1996). Importantly, like NMRs in hypoxia, goldfish and turtles are able to maintain their tissue ATP concentrations during anoxic exposure (Buck and Hochachka, 1993; Mandic et al., 2008), thus reducing the need to regenerate stores of phosphorylated energetic substrates (and thereby the need to ramp up oxidative phosphorylation above pre-anoxic levels during reoxygenation).

Given these strategies of other hypoxia/anoxia-tolerant vertebrates, it is tempting to speculate that NMRs similarly convert lactate accumulated due to anaerobic metabolism in hypoxia to glucose and/or glycogen during hypoxia and also during the reoxygenation phase. Specifically, lactate generated by anaerobic tissues in hypoxia (e.g. muscle, brain, heart) might be transported via the blood to the liver, wherein it could be converted back to glucose via the Cori cycle (Choi et al., 2005; Proia et al., 2016). The resulting glucose would then be available to be transported back to active tissues to sustain further anaerobic metabolism. Importantly, this process is a net consumer of ATP as six molecules of liver ATP are required to convert a molecule of lactate to a molecule of glucose, whereas glucose metabolism by an active tissue only yields two ATP via glycolysis. However, although this conversion yields considerably less net energy than the oxidation of lactate to CO₂ and H₂O, it also confers the distinct advantage of conserving carbon molecules that would otherwise be lost through ventilation. In a nutrient poor environment such as fossorial habitats, the retention of this macromolecule may be a distinct advantage for this species. The sustainment of this process during reoxygenation would explain the lack of an increase in metabolic rate during this stage and might also underlie the nearly comatose behavioural profile of NMRs, as energetic production might be temporarily minimized to support more rapid re-establishment of glucose pools.

Given that we do not observe a metabolic acidosis, NMRs may also express significant buffering capacity in their blood and/tissues to compensate for the H⁺ load generated during anaerobic metabolism. In light of this hypothesis, it is notable that the NMR V̇_O2 recovers to pre-hypoxia levels far more rapidly than V̇_CO2 during reoxygenation; this imbalance may reflect carbon sequestration to re-establish HCO₃⁻ pools, which are depleted (albeit not significantly so) in the blood of hypoxic animals. In support of this possibility, it is interesting to note that NMRs do not exhibit systemic acidosis when inhaling CO₂ levels <10% (as measured with an abdominal probe in anaesthetized animals) (Park et al., 2017), which suggests that this species may indeed have adaptations to buffer their physiology in environments or conditions that would tend to induce acidosis. Further experiments are warranted to evaluate this hypothesis.

Taken together, these results suggest that either: (1) anaerobic metabolism is not strongly recruited in this species, (2) NMRs utilize anaerobic metabolism and also possess potent buffering capacities in their blood and/or tissues that allow them to focus on replenishing glucose stores during reoxygenation over eliminating accumulated lactate loads, or (3) endogenous priming of lactate clearance mechanisms from cells and upregulation of gluconeogenesis support anaerobic metabolism while minimizing the deleterious consequences of this pathway upon reoxygenation. In any case, the marked reductions in behavioural activity and physiological and thermoregulatory function during acute hypoxia, combined with the switch to fuel substrates that yield more energy per molecule of O₂, minimize the degree to which the animals must rely on anaerobic metabolism. We speculate that these responses are a unique adaptation in this species that enables them to conduct important energetically expensive activities, such as digging new tunnels through dense and hard soils, in severely hypoxic environments.

We thank the UBC animal care and veterinary services team for their assistance in animal handling and husbandry.