Interspecific variation in brain mitochondrial complex I and II capacity and ROS emission in marine sculpins

Oxygen use in eukaryotic cells is a ‘double-edged sword’. On one side, O₂ binding to cytochrome c oxidase (COX) serves as a critical step in the function of the mitochondrial electron transport system (ETS) and aerobic ATP production. On the other side, stray electrons from ETS redox centres can bind O₂ to form reactive oxygen species (ROS) (Hoffman and Brookes, 2009; Quinlan et al., 2013), which at low levels play a signalling role and are scavenged by cellular and mitochondrial antioxidant systems (Jones, 2006; Turrens, 2003). When ROS generation exceeds scavenging capacity, however, ROS can accumulate and lead to oxidative damage (Ott et al., 2007; Sies, 1997). The relative rates of mitochondrial respiration and ROS generation by the ETS depend upon the redox environment (Aon et al., 2010), whereby mitochondrial ROS production is minimal in respiring mitochondria with a more oxidized redox environment. Mitochondrial ROS production increases as the ETS complexes become more redox-reduced, as occurs under O₂-limiting conditions. The relationship between mitochondrial redox environment, respiration and ROS generation has been extensively studied in mammalian systems in O₂-related pathologies (e.g. ischemia–reperfusion injury; Aon et al., 2010), but much less is known about the natural variation in mitochondrial respiration and ROS metabolism across organisms that routinely experience hypoxia.

Hypoxia tolerance necessitates numerous modifications to the O₂ transport cascade that improve O₂ extraction from the environment and delivery to mitochondria (Mandic et al., 2009, 2013; Scott et al., 2011). Hypoxia tolerance has been associated with a higher mitochondrial respiratory capacity, as observed in intraspecific comparisons among high- and low-altitude native deer mice (Peromyscus maniculatus; Mahalingam et al., 2017) or in elasmobranchs inhabiting different intertidal environments (Hemiscyllum ocellatum versus Aptychoterma rostrata; Hickey et al., 2012). In these same species, the higher respiratory capacity was accompanied by lower rates of ROS generation compared with the less hypoxia-tolerant animals (Hickey et al., 2012; Mahalingam et al., 2017). Although it is tantalizing to speculate that hypoxia tolerance may be associated with an overall higher respiratory capacity leading to lower rates of ROS generation, the opposite has also been observed. Indeed, laboratory-selected lines of hypoxia-tolerant Drosophila sp. had a lower overall respiratory capacity compared with wild-type flies, which was concomitant with a decrease in superoxide generation (Ali et al., 2012). Thus, the precise relationship between hypoxia tolerance, mitochondrial respiratory capacity and ROS generation requires more in-depth analysis. Furthermore, because the mitochondria redox environment can strongly influence ROS emission through effects on ROS generation and antioxidant defense mechanisms (Aon et al., 2010), interspecific analysis of ROS metabolism must attempt to take mitochondrial redox into account.

The goal of the present study was to assess the relationship between hypoxia tolerance and mitochondrial complex flux capacity, respiration, mitochondrial H₂O₂ emission per O₂ consumed (H₂O₂/O₂) and redox environment. To address this goal, we isolated mitochondria from the brain of a well-characterized group of fish species, commonly called sculpins (Cottidae, Actinopterygii), that are distributed along the marine intertidal zone (Richards, 2011) and show interspecific variation in whole-animal hypoxia tolerance (Mandic et al., 2013). We chose to work on the brain because sculpins display complex behaviours throughout the daily tidal cycle (such as hypoxia-induced aerial respiration and air emergence; Mandic et al., 2009); thus, the brain must continue to function during hypoxia. The variation in whole-animal hypoxia tolerance among sculpins is correlated with P_crit (the critical oxygen tension at which O₂ consumption rate transitions from oxy-regulating to oxy-conforming; Mandic et al., 2013), and interspecific variation in P_crit is largely, but not completely, explained by modifications to the O₂ transport cascade to enhance O₂ uptake from the environment and delivery to mitochondria (Lau et al., 2017; Mandic et al., 2009). Furthermore, variation in hypoxia tolerance among sculpins is also associated with modifications to the mitochondrion, where COX of hypoxia-tolerant sculpins show a higher O₂ binding affinity than those from hypoxia-intolerant sculpins (Lau et al., 2017). Building upon this putatively adaptive variation in mitochondrial function across sculpins, we hypothesized that mitochondria from more hypoxia-tolerant sculpins would show more efficient O₂ use than those in less hypoxia-tolerant sculpins, showing higher phosphorylation efficiency and lower ROS emission. We addressed the main goal of this study in four ways: first, defining the interspecific relationship between complex I- and II-linked respiratory capacity and P_crit; second, determining whether interspecific differences in mitochondrial respiratory capacity were related to H₂O₂ emission; third, comparing interspecific differences in the relationship between H₂O₂ emission and mitochondrial redox environment as assessed by titration of isolated mitochondria with reduced glutathione (GSH); and fourth, examining the responses of mitochondria to in vitro anoxia recovery.

All species of sculpin used in this study were collected near Bamfield Marine Sciences Centre (British Columbia, Canada) at Ross Islets (48°52.4′N, 125°9.7′W) or Wizard's Rock (48°51.5′N, 125°9.4′W) using either handheld dipnets in tidepools or pole seines along nearshore beaches at the lowest point in the daily tidal cycle. Animals were transported to The University of British Columbia (UBC) and housed in a recirculating aquaculture system with artificial seawater and maintained on a diet of shrimp, Atlantic krill and bloodworms for at least 1 month before experimentation. Temperature in the recirculating aquaculture system was maintained at 12°C, which is the average ocean temperature measured near the collection sites and the lower end of the temperature measured in tidepools (see Richards, 2011). All experimental procedures were reviewed and approved by The University of British Columbia Animal Care Committee (A13-0309) and are in accordance with guidelines of the Canadian Council on Animal Care. The experimental analyses described below were performed within 4 months of transport to UBC and individuals of each species were selected at random to avoid bias over the course of analysis.

Individual fish from each of the following species, Oligocottus maculosus Girard 1856 (6.6±0.3 g), Artedius fenestralis (Jordan & Gilbert 1883) (38.2±5.3 g), Artedius lateralis (Girard 1854) (34.7±7.7 g), Leptocottus armatus Girard 1854 (91.3±18.4 g), Scorpaenichthys marmoratus (Ayres 1854) (160.5±40.0 g) and Blepsias cirrhosus (Pallas 1814) (30.0±3.6 g), were netted from their stock tank, weighed, stunned via concussion and euthanized by spinal severance, and the entire brain was quickly dissected for mitochondrial isolation. For fish <10 g (i.e. O. maculosus), tissue from four animals was pooled to obtain sufficient yield for experiments, and for fish between 10 and 40 g (e.g. A. lateralis), tissue from two animals was pooled. Each pooled sample is considered a single replicate for statistical analysis. We first assessed basic mitochondrial function using a substrate uncoupler inhibitor titration (SUIT) protocol (part I of our study, see below) on six species of sculpin. Subsequently, we assessed the relationship between mitochondrial respiration and H₂O₂ emission (part II), and the effects of an in vitro redox (part III) and anoxia recovery (part IV) challenges on mitochondrial respiration and H₂O₂ emission. These latter experiments (parts II to IV) were performed on O. maculosus, A. lateralis and S. marmoratus, which were available in sufficient numbers and showed variation in hypoxia tolerance.

Briefly, brains were roughly chopped on an ice-cooled surface with a razor blade in isolation buffer (250 mmol l⁻¹ sucrose, 5 mmol l⁻¹ Tris, 1 mmol l⁻¹ EGTA, pH 7.4) and transferred into a Potter-Elvehjem tissue grinder and homogenized manually with six to eight passes of a polytetrafluoroethylene pestle. The tissue homogenate was then centrifuged at 1000 g for 3 min at 4°C and the supernatant was filtered through four layers of cheesecloth. The filtrate was then centrifuged at 10,000 g for 10 min at 4°C and the resulting supernatant was removed. The mitochondrial pellet was resuspended in cold isolation buffer and centrifuged again at 10,000 g for 10 min at 4°C to generate the final mitochondrial pellet, which was resuspended in isolation buffer and kept on ice until analysis as described below. Protein content of the mitochondrial suspension was determined with the Bradford assay (Bradford, 1976). An aliquot of the isolated mitochondria suspension was frozen in liquid N₂ and stored at −80°C for the determination of mitochondrial ETS complex maximal activity.

Mitochondrial respiration was measured with an Oroboros oxygraph high-resolution respirometer (Innsbruck, Austria). The polarographic oxygen sensors were calibrated daily with air-saturated and anoxic MiR05 buffer (0.5 mmol l⁻¹ EGTA, 3 mmol l⁻¹ MgCl₂·6H₂O, 60 mmol l⁻¹ K-lactobionic acid, 20 mmol l⁻¹ taurine, 10 mmol l⁻¹ KH₂PO₄, 20 mmol l⁻¹ Hepes, 110 mmol l⁻¹ sucrose, 1 g l⁻¹ bovine serum albumin; pH 7.1; Gnaiger et al., 2000) at 18°C. We chose an assay temperature of 18°C to maximize the signal-to-noise ratio during the respiration and H₂O₂ emission measurements, which was based on preliminary experiments and represents a temperature that all species encounter in their natural environment. Other studies in fish show that mitochondrial respiration and ROS generation have Q₁₀ values between 2 and 4 across much broader temperature ranges (Banh et al., 2016; Hochachka and Somero, 2002), thus we do not believe the slightly warmer assay temperature would affect any conclusions from our interspecific comparisons. The mitochondrial assays described below were randomized on the day of experiment and separate experiments indicate that mitochondrial quality does not vary over the time course of an experiment (data not shown).

To assess mitochondrial respiration in the six species of sculpin, we employed a SUIT protocol. Briefly, isolated mitochondria (∼0.2 to 0.5 mg mitochondrial protein) were first introduced into the respirometry chamber containing MiR05 buffer at 18°C. Following the introduction of mitochondria, NADH-generating substrates to fuel complex I were added (5 mmol l⁻¹ pyruvate, 1 mmol l⁻¹ malate and 10 mmol l⁻¹ glutamate; PMG) to yield state 2 (PMG) respiration rate, after which 16 µmol l⁻¹ ADP was added to stimulate respiration to assess ADP/O. Once state 4 respiration was obtained following the addition of 16 µmol l⁻¹ ADP, 0.75 mmol l⁻¹ ADP was added to ensure that we achieved maximal complex I-fueled state 3 respiration rate, followed by 10 mmol l⁻¹ succinate to assess complex I and II fueled state 3 respiration rate (PMGS). State 4 respiration was estimated with 2.5 µmol l⁻¹ oligomycin (herein referred to as state 4_o). This was followed by titration of 0.5 µmol l⁻¹ steps of carbonyl cyanide 4-(trifluoromethyoxy)phenylhydrazone (FCCP) until mitochondria were fully uncoupled and respiration rate no longer increased with FCCP addition. Once fully uncoupled, 0.5 µmol l⁻¹ rotenone was added to inhibit complex I, followed by addition of 2.5 µmol l⁻¹ antimycin A to inhibit complex III and assess residual O₂ consumption (refer to Fig. S1 for an example respiration trace).

To assess mitochondrial respiration simultaneously with H₂O₂ emission in O. maculosus, A. lateralis and S. marmoratus, mitochondrial respiration was assessed using a SUIT protocol similar to the one described above and changes in H₂O₂ emission were assessed fluorometrically using Amplex Ultrared (excitation 568 nm/emission 581 nm; Invitrogen) with the Oroboros respirometer and fluorometer (excitation LED at 525 nm with filter set optimized for Amplex Ultrared from manufacturer). Isolated mitochondria (∼0.5–0.6 mg mitochondrial protein) were introduced into the chamber calibrated with MiR05 maintained at 18°C, followed by the addition of 5 µmol l⁻¹ Amplex Ultrared, 2 U horseradish peroxidase and 13 U superoxide dismutase (where 1 U is the amount of enzyme required to convert 1 µmol substrate to product per minute). To calibrate the H₂O₂ detection methods, four additions of H₂O₂ were sequentially added to the chamber to a total of 0.56 µmol l⁻¹ H₂O₂ while monitoring changes in fluorescence. In order to confirm that ROS was the primary reactive species being detected (Amplex Ultrared is known to be sensitive to several reactive species, including reactive sulfur species; DeLeon et al., 2016), we added three to five 10 µl injections of 40 mg ml⁻¹ catalase to catalyze the conversion of H₂O₂ to O₂ at the end of several experimental runs (after GSH titrations in experimental part III until there were no measurable effects on H₂O₂ measurement), which resulted in a 75 to 80% reduction in ROS mg⁻¹ protein in all species examined (data not shown). This analysis suggests that the 75 to 80% of the reactive species detected in our experiments were H₂O₂.

For this part of the study, isolated brain mitochondria were first incubated with 10 mmol l⁻¹ pyruvate (excess pyruvate was used to ensure oxaloacetate removal to avoid inhibition of complex II; Hickey et al., 2012), 1 mmol l⁻¹ malate and 10 mmol l⁻¹ glutamate until a stable complex I-fuelled state 2 respiration rate was achieved (<5 min). This was followed by 10 mmol l⁻¹ succinate to assess complex I and II fuelled state 2. Then, 0.75 mmol l⁻¹ ADP was added to stimulate maximal state 3 respiration rate, followed by 2.5 µmol l⁻¹ oligomycin to establish state 4_o respiration, and 0.5 µmol l⁻¹ rotenone to inhibit complex I in state 4_o conditions (representative trace in Fig. S3A).

To investigate the effects of redox environment on ROS generation, the glutathione redox environment was manipulated by titrating GSH (to increase the extra-mitochondrial ratio of reduced to oxidized glutathione). These manipulations allowed us to relate H₂O₂ emission to the measured mitochondrial GSH redox potential under state 2 conditions in O. maculosus, A. lateralis and S. marmoratus. Briefly, the oxygraph was set up as described in part II with Amplex Ultrared. Once the H₂O₂ calibration was complete, substrates for complexes I and II (PMGS) were added to establish stable state 2 respiration and H₂O₂ emission rates, after which the contents of the oxygraph chamber were removed and centrifuged at 10,528 g for 5 min. The resulting mitochondrial pellet was washed twice with MiR05, and the mitochondrial pellet was frozen in liquid nitrogen and stored at −80°C until analyses of GSH redox potential (see below). To adjust the extra-mitochondrial GSH redox potential and induce reductive stress, the oxygraph was set up as described above, but after state 2 (PMGS) respiration was established, four injections of 0.6 mmol l⁻¹ GSH were added to the chamber to a final concentration of 2.4 mmol l⁻¹ GSH. State 2 respiration rate and H₂O₂ emission were monitored simultaneously with the oxygraph as GSH was titrated (see representative trace in Fig. S3B). At the end of the titrations, the entire mitochondrial suspension was removed from the oxygraph and processed as described previously for analysis of mitochondrial pellet GSH redox potential. The supernatant from the last MiR05 wash was confirmed to have no detectable levels of GSH. Analysis of mitochondrial pellet GSH redox potential was not performed after each 0.6 mmol l⁻¹ GSH titration because of limited biological sample, but if nominal concentrations are taken into account (assuming a linear dose-dependent reduction in mitochondrial GSH redox potential with extra-mitochondrial GSH additions), we see a typical titration curve with H₂O₂ emission for each GSH addition (Fig. S4).

To assess potential interspecific differences in recovery from in vitro anoxia in O. maculosus, A. lateralis and S. marmoratus, the oxygraph was set up as described in part II with Amplex Ultrared. Once the H₂O₂ calibration was complete, substrates to fuel complexes I and II (PMGS) were added followed by 0.75 mmol l⁻¹ ADP to stimulate state 3 respiration. At this point, the chamber O₂ concentration was adjusted to ∼60 µmol l⁻¹ by passing a stream of N₂ gas over the surface of the mitochondrial suspension. The chamber was then sealed and mitochondria were allowed to deplete chamber O₂ to anoxia, at which point mitochondria were maintained in anoxia for 20 min. The chamber was opened to quickly reoxygenate the mitochondrial suspension, after which 0.5–1 mmol l⁻¹ ADP was added to stimulate state 3 respiration rate. State 4_o was then induced with 2.5 µmol l⁻¹ oligomycin, and finally 0.5 µmol l⁻¹ rotenone was added to inhibit complex I (representative trace in Fig. S3C).

Mitochondrial GSH redox potential was determined spectrophotometrically using a modified recycling method that relies on glutathione reductase to convert GSH to GSSG in the presence of NADPH (Rahman et al., 2006). Briefly, to prepare samples for the assay, frozen mitochondrial pellets were resuspended in 80 µl buffer with 100 mmol l⁻¹ KH₂PO₄ and 5 mmol l⁻¹ EDTA disodium salt at pH 7.5 (KPE buffer), with 0.1% triton X-100 and 0.6% sulfosalicylic acid and thawed in a room temperature water bath. The suspension was freeze–thawed twice and 50 µl of the resulting suspension was used to determine glutathione disulfide (GSSG; oxidized glutathione) and the remainder of the sample was used to determine total glutathione (both reduced and oxidized forms). The remainder of the spectrophotometric assay is detailed in Rahman et al. (2006). At a pH of 7.5 and standard/assay temperature of 25°C, the standard reduction potential of GSH/GSSG is −270 mV {=−240 mV+[(7.5–7.0)×–59]mV; Schafer and Buettner, 2001}. In order to calculate the mitochondrial GSH redox potential, we used E=−270 mV–30×log([GSH]²/[GSSG]), where [GSH] and [GSSG] are the measured concentrations of GSH and GSSG in the mitochondrial pellets (expressed as molar concentration and assuming mitochondrial matrix volume is 1 µl mg⁻¹ protein; Tretter et al., 2007).

To determine the maximal activity (V_max) of complexes I, II and V, frozen samples of isolated mitochondria were thawed on ice and centrifuged at 15,000 g for 10 min at 4°C. The pellet was resuspended in hypotonic medium (25 mmol l⁻¹ K₂HPO₄, 5 mmol l⁻¹ MgCl₂·6H₂O at pH 7.2) and freeze–thawed three times before proceeding with the assay protocols described in Galli et al. (2013). To determine the maximal activity of complex III, the frozen mitochondrial sample was thawed on ice and freeze–thawed twice before proceeding with the spectrophotometric assays. Briefly, rotenone-sensitive complex I was monitored with the reduction of 5,5′-dithiobis(2-nitrobenzoic acid) (DCPIP) at 600 nm, complex II activity was also monitored as the reduction of DCPIP in the presence of rotenone and antimycin A, complex III activity was monitored as the reduction of cytochrome c at 550 nm with ubiquinol as the electron donor in the presence of rotenone, and finally oligomycin-sensitive complex V activity was monitored as the oxidation of NADH at 340 nm in the presence of pyruvate kinase and lactate dehydrogenase. Details of the assays are described in Galli et al. (2013) and Birch-Machin and Turnbull (2001).

Using regression analysis, we first confirmed that there were no relationships between body mass and the measurements made in part I. Complex I flux capacity was calculated from the SUIT protocol as the difference in respiration rate between the maximal uncoupled respiration rate (taken after the final FCCP addition in the SUIT protocol) and the subsequent rotenone-inhibited step (see Figs S1 and S2), which was expressed relative to the FCCP-uncoupled respiration rate to represent complex I flux capacity as a percent of maximum ETS capacity. Complex II flux capacity was calculated from the SUIT protocol as the difference in respiration rate between the rotenone-inhibited and antimycin A-inhibited steps (see Figs S1 and S2) and expressed relative to the FCCP-uncoupled respiration rate as done for complex I flux capacity. Respiratory control ratio (RCR) was determined as the ratio of state 3 (both complex I and II linked) to state 4_o respiration rate. To calculate the ADP/O ratio, the O₂ consumed was calculated using linear extrapolations of O₂ concentrations under state 3 respiration rate after the addition of ADP, and at state 4 after ADP depletion. The difference in O₂ concentrations at these intercepts represents the total O₂ uptake, which is used to divide nmol of ADP to give the ADP/O ratio (nmol ADP nmol⁻¹ O; Chance and Williams, 1955). H₂O₂/O₂ (pmol H₂O₂ nmol^–1 O₂) was calculated as the ratio of H₂O₂ emission rate (pmol H₂O₂ s⁻¹ mg⁻¹ protein) to respiration rate (nmol O₂ s⁻¹ mg⁻¹ protein).

To determine whether there were interspecific differences in respiration rates and flux capacities, we performed one-way ANOVA followed by Tukey's multiple comparisons tests. To determine whether interspecific differences in flux capacity were related to variation in P_crit among species, we performed correlative analysis using P_crit values taken from Mandic et al. (2009), which we have verified in subsequent analysis for a different study (D. Somo and J.G.R., unpublished). Correlative analysis was performed following the protocol outlined in Lau et al. (2017) using both ordinary least squares (OLS) and phylogenetically generalized least squares (PGLS). Detailed results from this analysis are shown in Table S1. Two-way ANOVA was used to analyse the effect of GSH treatment and species on H₂O₂/O₂ and mitochondrial pellet GSH redox potential between the three species in part III. In part IV, the normoxic (or pre-anoxia) respiration rates were normalized to 100% and all post-anoxia respiration rates (states 3, 4_o and 4_o with rotenone) were expressed relative to this value. Repeated-measures two-way ANOVA was used to analyse the effect of species and anoxia treatment on respiration rate and H₂O₂ emission. Post hoc analysis consisted of a Holm–Šidák multiple comparisons test.

The relative rate of complex I-fuelled ADP-stimulated respiration to ETS capacity significantly differed between sculpins (one-way ANOVA; P=0.005) and was related to interspecific variation in P_crit (OLS; P=0.05; Fig. 1A, Table S1). Hypoxia-tolerant sculpins with lower P_crit had lower complex I-fuelled (PMG) ADP-stimulated respiration rate than less tolerant species (Fig. 1A). The relative rate of complex I- and II-fuelled (PMGS) ADP-stimulated respiration rate to ETS capacity did not differ between species (P=0.12) and was not correlated with P_crit (data not shown; P=0.45). In uncoupled mitochondria, there were significant differences in complex I flux capacity between species (P=0.0035), which were related to P_crit (OLS; P=0.03; Fig. 1B). There were no differences in complex II flux capacity between species (P=0.074), but there was a strong, negative interspecific relationship with P_crit (OLS; P=0.002; Fig. 1C). Overall, species with lower P_crit (O. maculosus) had lower complex I flux capacity (65% of ETS capacity) and higher complex II flux capacity (29%) than species with higher P_crit (B. cirrhosus), which had higher complex I flux capacity (86%) and lower complex II flux capacity (8%; cf. Fig. 1B and C; see Fig. S2 for respiration data expressed relative to mg protein). These interspecific variations in complex flux capacities were not correlated with differences in ADP/O, RCR or the V_max of complex I or II (Table 1 and Table S1).

To assess interspecific differences in ROS metabolism, we chose to work on three species of sculpin: O. maculosus, which has the lowest P_crit among the species investigated, and A. lateralis and S. marmoratus, both with higher and similar P_crit values.

Consistent with our previous findings in part I (Fig. S2), there were no significant differences in state 2 respiration rates (PMG or PMGS) among the three species tested in this part of the study (Fig. 2A). Oligocottus maculosus had lower state 3 respiration rate (one-way ANOVA, P=0.018; Fig. 2A) and lower state 4_o respiration rate (P=0.042) than S. marmoratus, with A. lateralis having intermediate values. There were no differences in rotenone-inhibited state 4_o respiration rates between the three species. Simultaneous measurements of H₂O₂ emission rates showed no differences in H₂O₂ emission when expressed relative to total mitochondrial protein (H₂O₂ mg⁻¹ protein) in each respiration state (Table S2), but when H₂O₂ emission rates were expressed relative to respiration rates (H₂O₂/O₂), O. maculosus in state 4_o showed between 1.4- and 1.5-fold higher H₂O₂/O₂ than the two other species (P=0.016; Fig. 2B).

We investigated the relationship between H₂O₂/O₂ and mitochondrial GSH redox potential in three species, O. maculosus, A. lateralis and S. marmoratus, by adding GSH to the extra-mitochondrial environment in an attempt to shift the intra-mitochondrial GSH redox potential. There were significant GSH treatment and species effects on mitochondrial GSH redox potential, with no treatment by species interaction (Fig. 3A, Table S3). The addition of 2.4 mmol l⁻¹ GSH shifted intra-mitochondrial GSH redox potential to a more reduced state (i.e. more negative) in all species, but the effect of GSH addition on mitochondrial redox was greatest in S. marmoratus and least in A. lateralis, with O. maculosus showing an intermediate effect. GSH addition significantly increased H₂O₂/O₂ (Fig. 3B) and H₂O₂ mg⁻¹ protein (Fig. S4) in all species examined, and there were no significant differences between species (Table S3).

We compared the responses of brain mitochondria from O. maculosus, A. lateralis and S. marmoratus to 20 min of in vitro anoxia followed by reoxygenation. A repeated-measures two-way ANOVA revealed a significant effect of anoxia–reoxygenation on state 3 respiration rate, but no effect of species (Fig. 4A, Table S4). There was also a significant interaction between species and anoxia-recovery treatment on state 3 respiration, where it was decreased in O. maculosus and A. lateralis, but not in S. marmoratus. There was no effect of anoxia–reoxygenation on state 4_o or rotenone-inhibited state 4_o respiration rate in any species examined. The H₂O₂/O₂ under state 3 conditions was not affected by anoxia–reoxygenation and did not differ between species, whereas there was a significant effect of anoxia–reoxygenation on state 4_o and rotenone-inhibited state 4_o H₂O₂/O₂ (Fig. 4B, Table S4). When H₂O₂ was expressed relative to mitochondrial protein, there were no effects of anoxia–reoxygenation on state 3, but state 4_o H₂O₂ mg⁻¹ protein was decreased in all three species (Fig. 4C, Table S4). There was a significant effect of anoxia–reoxygenation on rotenone-inhibited state 4_o H₂O₂ mg⁻¹ protein, which was diminished in mitochondria of the least hypoxia-tolerant species, S. marmoratus. There was no significant effect of species in any of the analysis performed on H₂O₂ mg⁻¹ protein.

The present study demonstrates that interspecific variation in hypoxia tolerance among sculpins is related to a lower complex I flux capacity in mitochondria from the brain of more hypoxia-tolerant species. This variation, however, was not related to improving mitochondrial phosphorylation efficiency or lowering ROS emission. In fact, we demonstrated that the hypoxia-tolerant O. maculosus generated more H₂O₂/O₂ under state 4_o conditions than the less tolerant A. lateralis and S. marmoratus. In response to an in vitro redox challenge, the three species studied varied in their ability to maintain mitochondrial GSH redox potential, but the degree of disturbance to mitochondrial GSH redox potential was not related to whole-animal hypoxia tolerance. In addition, under these redox challenges, all species had the same H₂O₂/O₂. Finally, mitochondria from all three species tested were able to lower H₂O₂ mg⁻¹ protein after in vitro anoxia recovery, but O. maculosus and A. lateralis appeared to achieve this by decreasing mitochondrial respiration whereas S. marmoratus showed a reduction in H₂O₂ generated, possibly from complex I. Overall, these outcomes suggest that hypoxia tolerance in sculpins is associated with brain mitochondria having a lower complex I and higher complex II flux capacities and higher H₂O₂ emission, but the species-specific differences in H₂O₂ emission are only revealed under oligomycin-induced state 4 conditions.

The strong interspecific relationships observed between complex I and II flux capacity and hypoxia tolerance suggest an underlying and potentially beneficial adaptation. Among the six species of sculpin examined, there were strong correlations between complex I and II flux capacities and P_crit, whereby hypoxia-tolerant species had lower complex I and higher complex II flux capacities compared with hypoxia-intolerant species (assessed in uncoupled mitochondria; Fig. 1B,C). This interspecific variation does not appear to be due to a difference in phosphorylation control, as we also observed a lower complex I-linked state 3 respiration rate in coupled mitochondria in the hypoxia-tolerant species compared with the hypoxia-intolerant species (Fig. 1A). Furthermore, the variation does not appear to be explained by complexes I and II V_max (Table I), which would tend to suggest that despite having the same enzyme capacity, i.e. V_max, the hypoxia-tolerant species utilize less of the capacity available, i.e. respiratory flux capacity.

Even though the present study is the first to show putatively adaptive variation in ETS substrate dependency across species that vary in hypoxia tolerance, previous studies in other species and systems have shown that acute hypoxia exposure can differentially affect complex I and II flux rates. In gerbil and rat brains, complex I-linked respiratory flux was inhibited to a greater extent in response to hypoxia exposure and was also slower to recover in normoxia when compared with complex II-linked respiratory flux, which was either inhibited to a lesser extent or not affected (Almeida et al., 1995; Gilland et al., 1998; Stepanova et al., 2018). This decrease in complex I activity was, in part, mediated by the transcription factor, hypoxia inducible factor 1 (HIF-1), which lowered the expression of several complex I subunits (reviewed in Fuhrmann and Brüne, 2017) and caused an inactivation of the complex (Galkin et al., 2009; Stepanova et al., 2018). Stepanova and colleagues (2018) proposed that the hypoxia-induced inhibition of complex I would serve to lower ROS production and limit hypoxia-induced cellular damage. In cells that have impaired complex I activity, Hawkins et al. (2010) demonstrated that the oxidation of the complex II substrate succinate maintained mitochondrial membrane potential during hypoxia, likely by saturating the mitochondrial inner membrane with partially or fully reduced coenzyme Q. Thus, it is possible that the higher complex II flux capacity observed in hypoxia-tolerant sculpins may aid in sustaining the proton gradient and thus ADP phosphorylation during O₂ limitation, but further investigation is needed before this possibility is confirmed in marine sculpins. Overall, however, these acute responses to hypoxia exposure are consistent with the innate interspecific differences observed in marine sculpins, suggesting that lower complex I and higher complex II flux capacities observed here in hypoxia-tolerant sculpins may be more broadly observed. Although variation in substrate dependency may not appear to benefit phosphorylation efficiency in hypoxia-tolerant sculpins (i.e. increase ADP/O; Table I), we hypothesize that this adaptive variation in substrate dependency may be part of a strategy to minimize ROS generation when hypoxic conditions are encountered.

Hypoxia-tolerant sculpins possess an O₂ transport cascade that presumably increases O₂ extraction and delivery to tissues (Lau et al., 2017; Mandic et al., 2009, 2013), thus signifying the importance of efficient O₂ use in determining whole-animal hypoxia tolerance. Mechanisms to increase the efficiency of O₂ use would presumably also minimize potentially wasteful and harmful ROS generation. Thus, we investigated whether hypoxia-tolerant sculpins would have lower H₂O₂/O₂, indicating that less O₂ is eventually emitted as H₂O₂ (converted from superoxide via superoxide dismutase).

We chose three species that varied in P_crit and complex capacities in part I to investigate the relationship between respiration and mitochondrial H₂O₂ emission. Assuming the same sites of electron entry into the ETS and substrate flux, O. maculosus had significantly lower state 4_o respiration rate than S. marmoratus (A. lateralis was intermediate; Fig. 2A), which is consistent with lower futile proton cycling and thus better mitochondrial coupling (although our six-species comparison in part I showed no relationship between P_crit and RCR). Increased mitochondrial coupling typically brings about a higher proton gradient, which can be associated with higher ROS generation (Brand, 2000; Brand and Esteves, 2005). Indeed, O. maculosus had higher H₂O₂/O₂ in state 4_o compared with A. lateralis and S. marmoratus, which was not consistent with our hypothesis that hypoxia tolerance is associated with lower ROS generation. This pattern of higher H₂O₂/O₂ with lower leak respiration only emerged under state 4_o conditions rather than in state 2, indicating that the H₂O₂ generation may be sensitive to phosphorylation state or there may be species-specific sensitivities to oligomycin. As these were in vitro analyses, it would be interesting to perform further studies to investigate whether similar responses would be observed in vivo, firstly, as neither of these leak respiration states (2 and 4_o) occur in vivo, and secondly, as our ROS measurements were performed under air-saturating conditions, which do not accurately reflect in vivo O₂ tensions.

Although we hypothesized that variation in complex I flux capacity may be associated with lower ROS emission, the interspecific differences in H₂O₂/O₂ were not reflected in state 4_o in the presence of rotenone, which could potentially suggest that there are no interspecific differences in ROS generated from the quinone-binding site on complex I (I_Q site; Fig. 2B) (Quinlan et al., 2013). However, this is not a definitive test for the contribution of complex I to overall ROS emission, and a more detailed study manipulating electron flux through complex I under different respiratory states would be required to assess potential differences in complex I ROS generation capacity.

Mitochondrial ROS generation is highly dependent on mitochondrial respiration state and redox environment (Aon et al., 2010; Cortassa et al., 2014; Munro and Treberg, 2017), both of which can be profoundly affected by the availability of O₂. Reductive stress, i.e. increase in reducing equivalents in redox couples such as GSH/GSSG, NADPH/NADP⁺ and NADH/NAD⁺, has been shown to cause oxidative damage (Niknahad et al., 1995; Shen et al., 2005; Yu et al., 2014; Zhang et al., 2012). Hypoxia leads to reductive stress, where lower ETS flux leads to a highly reduced redox environment and excess electrons in the ETS complex redox centres can favour ROS generation even though O₂ levels are low (Jones, 1985; Niknahad et al., 1995). The combined effects of increased ROS generation and overwhelmed scavenging capacity results in higher net ROS emission (Aon et al., 2010). We hypothesized that the more hypoxia-tolerant sculpin would emit less ROS under a stressful redox environment when compared with a less tolerant species.

To manipulate the mitochondrial redox environment, we altered the extra-mitochondrial GSH pool as it is a major redox couple within the cell and is also easily assessed by monitoring concentrations of GSH and GSSG (Rahman et al., 2006; Schafer and Buettner, 2001). Mitochondria contain two GSH pools, with one located in the intermembrane space and a second in the matrix. Although the communication between these two pools is not well understood (e.g. Kojer et al., 2012; Mari et al., 2013), it is generally recognized that mitochondria are not capable of synthesizing GSH, thus there must be a transport mechanism between the intermembrane space and the matrix. Furthermore, we were able to successfully change mitochondrial GSH redox potential by altering the extra-mitochondrial GSH/GSSG pool. Under these conditions, we predicted that the more hypoxia-tolerant sculpin would (1) accumulate lower H₂O₂/O₂ as GSH redox potential becomes more reduced, and (2) be able to better maintain mitochondrial redox status with changes in extra-mitochondrial GSH redox potential.

Our results clearly indicate that the three species examined did not differ in mitochondrial H₂O₂ emission when challenged with the same extra-mitochondrial GSH additions (Fig. 3B). However, under the identical GSH challenge conditions, the mitochondrial GSH:GSSG was more oxidized in A. lateralis (less negative; Fig. 3A) and appeared to be better buffered against extra-mitochondrial GSH:GSSG changes, as compared with S. marmoratus and O. maculosus. Thus, these results suggest that mitochondria of A. lateralis has a higher H₂O₂/O₂ for a given change in redox status than O. maculosus and S. marmoratus. Although the ability to resist changes in mitochondrial redox does not appear to be related to variation in hypoxia tolerance, this analysis does highlight the importance of considering the control of mitochondrial redox state as an important component of managing ROS emissions. One mechanism that may be involved is the increased GSH redox buffering capacity of A. lateralis, via the actions of glutathione peroxidase and glutathione reductase, to maintain glutathione redox status (Rahman et al., 2006). Future studies with simultaneous measurements of multiple redox couples and other variables are necessary in order to increase our understanding of mitochondrial redox regulation.

Injury is sustained from ischemia–reperfusion as a result of the surge of ROS generated in normoxic recovery following a period of O₂ deprivation (Chouchani et al., 2013). The oxidative damage caused by the transition from anoxia/hypoxia to normoxia results in incomplete recovery of respiration in mammalian mitochondria (Almeida et al., 1995; Du et al., 1998; Shiva et al., 2007). In fact, the interspecific difference in recovery from 20 min in vitro anoxia showed that the O. maculosus and A. lateralis brain mitochondria significantly decreased state 3 respiration rate, whereas S. marmoratus showed a complete recovery of state 3 respiration (Fig. 4). This reduction in state 3 respiration rate aligns with typical mammalian responses to similar protocols in O₂-sensitive tissues, and has been attributed to oxidative damage (Du et al., 1998), but in fact all three sculpin species decreased state 4_o H₂O₂ mg⁻¹ protein after anoxia. Although this appears to be achieved via a lowering of respiration rate in O. maculosus and A. lateralis, in S. marmoratus the reduction in H₂O₂ emitted upon anoxia recovery may be due to decreased H₂O₂ generated from complex I, as indicated by the lowered rotenone-inhibited ROS mg⁻¹ protein. Considering that this change in H₂O₂ emission from S. marmoratus complex I occurred quickly in response to an in vitro anoxia exposure, we posit that this may have been a result of a post-translational modification. Additional analysis on more species should be conducted to elucidate whether these interspecific differences relate to hypoxia tolerance.

Our analysis clearly demonstrates that variation in hypoxia tolerance among sculpins is related to differences in the capacities of complexes I and II in brain mitochondria, with the more hypoxia-tolerant species showing lower complex I and higher complex II flux capacity compared with less-tolerant species. The lower complex I capacity in the hypoxia-tolerant sculpins was not associated with higher phosphorylation efficiency. We thus investigated whether the differences in complex capacity may be related to ROS generation. Our analysis of ROS metabolism in sculpin mitochondria, however, paints a more complex and nuanced relationship between hypoxia tolerance and ROS than previous studies, which suggested that hypoxia-tolerant animals may lower mitochondrial ROS emission (e.g. Du et al., 2016; Hickey et al., 2012; Mahalingam et al., 2017). Our interspecific comparisons suggest that hypoxia tolerance may be associated with a higher ROS emission, but this was only revealed under pharmacologically (oligomycin) induced state 4_o conditions (Fig. 2B). In addition, our analysis indicates that the ability to maintain matrix redox status under extra-mitochondrial reductive stress differed between species, and despite these differences, the same H₂O₂/O₂ was emitted (Fig. 3). We also observed that in vitro anoxia-recovery exposure lowered mitochondria H₂O₂ emissions in all species examined and negatively impacted state 3 respiration rates in several species. Thus, although the strong, phylogenetically independent relationship between hypoxia tolerance and complex I and II flux capacities strongly suggests that a low complex I flux capacity is an adaptation for hypoxia, the role of variation in flux capacity and hypoxia tolerance in ROS metabolism must remain speculative. Furthermore, it will be important for future studies to also directly assess whether there is variation in mitochondrial ROS emission as a function of decreasing O₂ in species that differ in hypoxia tolerance.

We would like to thank Joshua Emerman, Monica Goldade, Dr Tammy Rodela, Derek Somo and Andrew Thompson for their help with fish collections.