Intraspecific variation and plasticity in mitochondrial oxygen binding affinity as a response to environmental temperature

Mitochondrial function has been suggested to underlie constraints on whole-organism aerobic performance and associated hypoxia and thermal tolerance limits, but most studies have focused on measures of maximum mitochondrial capacity. Here we investigated whether variation in mitochondrial oxygen kinetics could contribute to local adaptation and plasticity in response to temperature using two subspecies of the Atlantic killifish (Fundulus heteroclitus) acclimated to a range of temperatures (5, 15, and 33 °C). The southern subspecies of F. heteroclitus, which has superior thermal and hypoxia tolerances compared to the northern subspecies, exhibited lower mitochondrial O2 P50 (higher O2 affinity). Acclimation to thermal extremes (5 or 33 °C) altered mitochondrial O2 P50 in both subspecies consistent with the effects of thermal acclimation on whole-organism thermal tolerance limits. We also examined differences between subspecies and thermal acclimation effects on whole-blood Hb O2-P50 to assess whether variation in oxygen delivery is involved in these responses. In contrast to the clear differences between subspecies in mitochondrial O2-P50 there were no differences in whole-blood Hb-O2 P50 between subspecies. Taken together these findings support a general role for mitochondrial oxygen kinetics in differentiating whole-organism aerobic performance and thus in influencing species responses to environmental change.

mechanism underlying thermal tolerance limits in fishes 20,21 . However, given the relationship between ambient temperature and aerobic metabolism, prolonged thermal stress likely alters mitochondrial function and thus an investigation of temperature effects on Mito-P50 is necessary 3,4,6 .
To address this question, here we utilize F. heteroclitus, a eurythermal teleost found in estuarine salt marshes along a large latitudinal range that spans a steep thermal gradient [(Northern Florida, USA (mean monthly southern temperature range Sapelo Island, GA, USA: 11-30 °C 22 ) to Nova Scotia, Canada (mean monthly northern temperature range 3-11 °C Wells Inlet, ME, USA 22 )]. This species is highly tolerant of both thermal and hypoxic stress, experiencing considerable variation in these abiotic factors over diel as well as seasonal cycles [23][24][25] . Furthermore, northern and southern F. heteroclitus subspecies exhibit variation in thermal and hypoxia tolerance that is consistent with apparent adaptation to their local environments 24,25 . This species recruits a wide array of physiological responses to thermal stress, including altered mitochondrial function 12,15,[25][26][27] . Thus, this species is an ideal model in which to investigate the potential roles of mitochondrial kinetic properties in thermal acclimation and adaptation.
The objectives of this study were three-fold, (1) determine if intraspecific variation in Mito-P50 exists between putatively thermally adapted northern (Fundulus heteroclitus heteroclitus) and southern (Fundulus heteroclitus macrolepidotus) Atlantic killifish subspecies, (2) investigate the effects of thermal acclimation (5,15, and 33 °C) on Mito-P50 and, (3) characterize the acute thermal response of Mito-P50 between F. heteroclitus subspecies acclimated to different temperatures. We also assessed intraspecific variation and thermal acclimation effects on hemoglobin (Hb) O 2 -P50, as there is some evidence of intraspecific variation in this parameter in F. heteroclitus 28 and any variation in this parameter could result in changes in PO 2 gradients between the circulatory system and the mitochondrion.

Results
Whole-organism hypoxia tolerance (LOE hyp ). We measured time to loss of equilibrium in hypoxia (LOE hyp ) at 15 °C in fish acclimated to 15 °C to confirm subspecies differences in whole organism hypoxia tolerance 25 . The northern and southern subspecies of killifish acclimated to 15 °C differed in LOE hyp at T assay = 15 °C across a range of PO 2 (Fig. 1). Southern killifish exhibited greater hypoxia tolerance, as indicated by time to LOE hyp , when compared to northern killifish (p subspecies < 0.001). PO 2 also affected LOE hyp, which decreased with decreasing PO 2 (p partial pressure < 0.001), as did the difference between the subspecies (p partial pressure*subspecies < 0.05).
Hb and Mito-P50. In this study, we examined mitochondrial O 2 binding affinity using isolated mitochondria from the liver because high-quality mitochondria can be isolated from this tissue in this species 15,27 . These assays were conducted using a mixture of substrates at saturating levels to be representative of working or stressed states. For Mito-P50, there were significant effects of subspecies (p subspecies < 0.05), assay temperature (p assay < 0.0001), thermal acclimation (p acclimation < 0.0005), and the interaction between thermal acclimation and assay temperature (p acclimation*assay < 0.0001) on Mito-P50. No additional significant interaction effects were detected (p subspecies*acclimation = 0.181, p subspecies*assay = 0.144, p population*acclimation*assay = 0.273) ( Supplementary Fig. S1).

Figure 1.
Intraspecific differences in whole-organism hypoxia tolerance. Northern and southern F. heteroclitus were acclimated to and assayed at 15 °C. Hypoxia tolerance was estimated using a time to loss of equilibrium assay across a range of low O 2 partial pressures. Data are mean ± SEM, n = 9-10.
Because of the complex interactive effects of subspecies, acclimation temperature, and assay temperature on both Mito-P50 and Hb-P50, we tested a series of specific hypotheses regarding the effects of individual factors on both of these parameters.

Subspecies effects on mitochondrial and Hb-O 2 affinity.
We compared mitochondrial and Hb-P50 between subspecies when assayed at their acclimation temperature to test our prediction that southern killifish maintain lower O 2 P50 when compared to northern killifish (Fig. 2). Mito-P50 was lower in southern killifish compared to northern killifish when acclimated to 15 °C and assayed at 15 °C ( Fig. 2A; p subspecies < 0.005), but not in 33 °C acclimated killifish assayed at 33 °C ( Fig. 2B; p subspecies = 0.301). Hb-O 2 binding affinity did not differ between subspecies at either acclimation temperature ( Thermal acclimation effects on mitochondrial and Hb-O 2 affinity. We compared thermal acclimation effects (5, 15, and 33 °C) on mitochondrial (Fig. 3A) and Hb-P50 (Fig. 3B) between subspecies at a common assay temperature of 15 °C to test our prediction that acclimation to higher temperatures would result in lower O 2 P50 for both parameters. We used 5 and 33 °C as our experimental acclimation temperatures as these are temperatures at which effects on whole-organism aerobic metabolism are first observed 29 . In addition, we acclimated F. heteroclitus to 33 °C to avoid the induction of substantial breeding physiology at slightly lower temperatures (i.e., 25 to 30 °C). These acclimation treatments represent the majority of temperatures experienced by natural F. heteroclitus populations 22 .
We detected a significant effect of acclimation on Mito-P50 ( Fig. 3A; p acclimation < 0.001). This effect was driven by higher Mito-P50 in 5 °C acclimated killifish and lower Mito-P50 in 33 °C acclimated northern killifish. We also detected a significant effect of subspecies that was driven by greater Mito-P50 in northern killifish (p subspecies < 0.05). Significant interactions between subspecies and thermal acclimation effects were a result of subspecies differences being removed as T acclimation increased to 33 °C (p subspecies*acclimation < 0.05). We did not detect a significant effect of acclimation on Hb-P50 ( Fig. 3B; p acclimation = 0.087) or an effect of subspecies (p subspecies = 0.599) or an interaction between subspecies and acclimation temperature (p subspecies*acclimation = 0.140).

Acute thermal effects on mitochondrial and Hb-O 2 affinity.
We assessed the effects of thermal acclimation and local adaptation on Δ Mito-P50 (i.e., the difference in Mito-P50 between T assay = 15 and 33 °C) and the heat of oxygenation of Hb, ΔH (Fig. 4). We predicted that thermal acclimation and putative local adaptation would alter the acute thermal response for both parameters. Acute changes in temperature alter buffer pH, and this has the potential to affect both Mito-P50 and Hb-P50. But this variation in pH does not affect the interpretation of our specific predictions as most comparisons were made at the same assay temperature. The only exception is our assessment of acute thermal effects on O 2 binding affinity ( Fig. 4), which might be subject to an interaction between intraspecific variation or thermal acclimation effects and pH variation induced by the acute temperature shift between 15 and 33 °C.

Discussion
Here we demonstrate that Mito-P50 differs between putatively thermally adapted subspecies of killifish and is sensitive to thermal acclimation. These effects on Mito-P50 are consistent with intraspecific variation and the  effects of thermal acclimation on whole-organism thermal and hypoxia tolerance 25 . These data provide evidence for altered mitochondrial oxygen affinity as a potential mechanism for maintaining whole-organism performance under environmental stress, which may ultimately contribute to subspecies-specific differences in thermal tolerance.
There were clear intraspecific differences in Mito-P50 between northern and southern F. heteroclitus subspecies, with southern fish maintaining lower O 2 -P50 ( Fig. 2A, S1). We propose that variation in Mito-P50 partially underlies intraspecific variation in hypoxia tolerance ( Fig. 1 25 ). Intraspecific variation in Mito-P50 may be a consequence of variation in genes encoding cytochrome c oxidase (CCO) subunits. CCO is the terminal acceptor of the electron transport chain and the primary site of O 2 consumption in the mitochondrion. In addition, CCO subunits are encoded by both nuclear and mitochondrial genomes, the latter being subject to high mutation rates. Consequently, CCO function has been demonstrated to be a target of selection [30][31][32] . Genome sequencing efforts in F. heteroclitus 9 have revealed mixed evidence for the presence of functionally significant variation in CCO among F. heteroclitus populations, but at present there have been no comprehensive examinations of sequence variation between populations from the extremes of the species distribution. In addition, differences in CCO function could be a consequence of variation in mitochondrial membrane composition or post-translational modifications of the enzyme 33 .
In contrast to the variation between subspecies in Mito-P50, we did not observe significant differences in Hb-P50 between F. heteroclitus subspecies (Fig. 2C,D), consistent with previous observations in a hybrid F. heteroclitus population 28 . However, in this population individuals with differing LDH-B genotypes differed in Hb-P50 following exhaustive swimming, implicating a genotype-specific Bohr shift (i.e., low pH decreasing O 2 affinity) due to allosteric modification of Hb by ATP 28 . These differences are thought to result from differences in glycolytic metabolism due to variation in LDH-B 34 . Both subspecies maintain equivalent hematocrit ( Supplementary  Fig. S4 28 ), indicating that if Hb characteristics differentiate aerobic performance between subspecies it likely occurs through allosteric mechanisms.
Hb-P50 in killifish (approximately 0.4 kPa) is low relative to other fish species (e.g., Hb-P50 = 3.6 kPa in Oncorhynchus mykiss 35 ; 2.6 kPa in Kryptolebias marmoratus 36 ; and ranges between 3-8 kPa among intertidal sculpins 37 ) suggesting that selection may have acted on F. heteroclitus to maximize O 2 extraction from the environment. However, for Hb there is a trade-off between the ability to load O 2 at the gills and unload at the tissues. The fact that we do not observe much variation between subspecies in Hb-P50 may reflect this trade-off. Thus, variation in Mito-P50 could represent an adaptation to maximize tissue O 2 diffusion in the hypoxia tolerant southern subspecies. The hypothesis of a tissue O 2 diffusion limitation in this species is supported by electron microscope observations of the location of the mitochondrion in this species, which at least in muscle are localized immediately below the plasma membrane 26 . However, both Hb-P50 and Mito-P50 are subject to considerable regulation in vivo, and this is an effect that we are unable to account for in our assays. Nevertheless, the difference in O 2 binding affinity between Hb and the mitochondrion likely contributes to the shape of O 2 diffusion gradients at the tissue 38 .
In humans, there is an inverse relationship between basal metabolic rate and Mito-P50 32 . We therefore predicted that northern F. heteroclitus would have lower Mito-P50 than southern fish, because northern populations maintain higher routine metabolic rates than do their southern counterparts 29 . In contrast, we found that the southern subspecies has a lower Mito-P50, suggesting that the relationship between metabolic rate and Mito-P50 does not represent a functional constraint in F. heteroclitus. Alternatively, the inconsistent relationship between Mito-P50 and estimates of basal metabolism in humans and F. heteroclitus may be a consequence of the energetic demands imposed by endothermy when compared with ectothermy. The combination of low routine metabolic rate and Mito-P50 exhibited by the southern subspecies may represent a beneficial strategy in high temperature hypoxic environments, allowing for greater tissue O 2 extraction while also decreasing overall demand in the face of environmental hypoxia.
Organisms' thermal tolerance limits are thought to be shaped by temperature effects on aerobic metabolism that may, at least in part, be due to effects at the level of the mitochondrion 3,6,13 . However, these effects have mostly been examined in the context of mitochondrial respiratory capacity. Here we show that northern and southern F. heteroclitus subspecies that differ in thermal tolerance 24 also exhibit differences in Mito-P50. The lower Mito-P50 in the southern subspecies, which reflects a greater mitochondrial oxygen affinity, could aid in the maintenance of mitochondrial O 2 diffusion gradients, particularly at high temperatures, which could help to sustain aerobic metabolism at high temperatures. Similarly, low Mito-P50 has the potential to aid O 2 delivery during environmental hypoxia. Thus, in southern habitats where temperatures are higher and hypoxic events are more likely, low Mito-P50 could be favored. Alternatively, the relatively low temperatures and higher oxygen levels in northern habitats might result in relaxed selection on these traits, reducing the constraints on Mito-P50 in this subspecies. Taken together, our demonstration of intraspecific variation in Mito-P50 thus provides a candidate trait accounting for potential covariation of both whole-organism hypoxia and thermal tolerance.
Acclimation to both 5 and 33 °C resulted in clear changes in Mito-P50 that were subspecies-dependent (Fig. 3, S1). In contrast, Hb-P50 did not exhibit a clear response to thermal acclimation (Fig. 4, S2). This previously undescribed phenomenon identifies potential targets and mechanisms of thermal acclimation and provides support for a role for mitochondrial function in maintaining performance following prolonged thermal stress.
We predicted that acclimation to 33 °C would decrease Mito-P50 thereby compensating for the aerobic challenges associated with high temperature 3,8,15,39 . When compared at a common assay temperature of 15 °C, acclimation to 33 °C decreased Mito-P50 in northern but not southern F. heteroclitus (Fig. 3A). Decreases in Mito-P50 under these conditions may alleviate limitations on total O 2 flux suggested to occur with elevated temperatures 3 and is consistent with the maintenance of a PO 2 gradient to the mitochondrion. In contrast, southern F. heteroclitus exhibit low Mito-P50 (high oxygen affinity) at both intermediate and high temperatures, indicating a difference in strategy between the subspecies. However, when compared at a higher assay temperature of 33 °C, acclimation to 33 °C increased Mito-P50 in both subspecies (Supplementary Fig. S1). This paradoxical response could represent maladaptive acclimation as this would presumably decrease O 2 diffusion gradients to the tissues. Thus, it is possible that this response may instead be a consequence of sub-lethal effects associated with 33 °C acclimation. 33 °C is a non-lethal temperature to which F. heteroclitus can acclimate for prolonged periods 24 . However, acclimation to 33 °C is associated with decreases in whole body mass and routine O 2 consumption, perhaps indicative of trade-offs necessary to mitigate the effects of extremely high acclimation temperatures on energetic balance 8,29 . Thus, our observed change in Mito-P50 may be a consequence of other mitochondrial responses associated with high temperature acclimation (e.g., altered mitochondrial morphology and membrane composition 40,41 ). This raises interesting questions as to the mitochondrial responses of organisms under conditions of sub-lethal stress, which are likely to have a profound influence on species' fitness [42][43][44] .
In both subspecies, we observed higher Mito-P50 in fish acclimated to low temperature when assayed at 15 °C (Fig. 3A). Indeed, cold acclimation is associated with increases in mitochondrial respiratory capacity, mitochondrial volume density and lipid content in aquatic ectotherms 14,15,26,41 . Increases in lipid content increase O 2 solubility 45 which may alleviate potential mitochondrial O 2 limitations in systemic tissues at low temperatures that are brought on by decreased O 2 diffusion rates 46 . This 5 °C acclimation response may reveal a mechanism for life at low temperatures as it is consistent with the greater Mito-P50 exhibited by northern F. heteroclitus (Fig. 2).
Similar to our demonstration of thermal acclimation effects on Mito-P50 (Fig. 3A), prolonged exposure to hypoxia might be predicted to decrease Mito-P50 7,17 , if reductions in Mito-P50 are important for maintaining O 2 diffusion gradients. However, hypoxia acclimated northern F. heteroclitus exhibit no modification of Mito-P50 7 . This suggests that declines in Mito-P50 at high temperatures are not directly mediated by the associated environmental hypoxia or resulting hypoxemia. Thus, the effects of thermal acclimation that we observe may play a role other than maintaining oxygen diffusion gradients. Alternatively, during hypoxic acclimation this species could recruit other mechanisms for maintaining O 2 supply and demand balance, such reductions in demand 23 would reduce the necessity for decreased Mito-P50 following acclimation.
Although we detected intraspecific variation and thermal acclimation effects on liver Mito-P50, this effect might not be present in mitochondria from other tissues 47 . Indeed, thermal acclimation and intraspecific variation effects on F. heteroclitus mitochondrial respiratory capacity vary among the heart, liver and brain 8,15,27 , suggesting that mitochondria from different tissues may respond differently. The role of liver mitochondria in constraining whole-organism hypoxia and thermal tolerance is not clear, whereas other tissues such as the heart and brain may be more important 3,5,21 . Consistent with this idea, variation in Mito-P50 from brain mitochondria has been shown to be associated with evolutionary variation in hypoxia tolerance 19 whereas Mito-P50 from liver mitochondria does not change in response to hypoxic acclimation in F. heteroclitus despite changes in whole-organism hypoxia tolerance 7 . Thus, the role of the differences in liver Mito-P50 that we observe in setting whole organism thermal or hypoxia tolerance requires further investigation.
As assay temperature increased (assay temperature: 15-33 °C), both Hb and Mito-P50 increased (Figs S1, S2). The loss of Hb O 2 affinity associated with increasing assay temperature is often observed and is primarily a consequence of the exothermic nature of Hb oxygenation 48 . Decreased Hb O 2 binding affinity with acute increases in temperature might aid in unloading O 2 to the tissues but might also compromise O 2 loading at the gills. However, these effects are likely subject to considerable regulation in vivo. In contrast, decreased mitochondrial O 2 affinity with increased assay temperature has not been previously characterized and is presumably a result of temperature effects on enzyme stability. This decrease in mitochondrial O 2 binding affinity at high assay temperatures might cause a decline in mitochondrial ATP synthesis, perhaps revealing an aspect of declining mitochondrial function associated with acute increases in temperature.
We detected clear subspecies effects on the acute thermal sensitivity of O 2 binding affinity at all acclimation temperatures (Fig. 3). Northern F. heteroclitus exhibited greater Δ Mito-P50 and a more exothermic Hb ΔH (i.e., greater thermal sensitivity) when compared to the southern subspecies (T assay = 15 to 33 °C). These data indicate that northern F. heteroclitus exhibit a greater relative loss of O 2 binding affinity than the southern subspecies following acute increases in temperature. This loss of O 2 binding affinity at high acute temperatures may result in greater constraints on aerobic metabolism in this subspecies, which could be associated with the differences between subspecies in acute thermal tolerance limits 24 .
Thermal acclimation also altered the acute thermal sensitivity of Mito-P50 ( Supplementary Fig. S1). This is reflected by increased Δ Mito-P50 following 33 °C acclimation in both F. heteroclitus subspecies (Fig. 4A). This increase in sensitivity may be a consequence of sub-lethal effects associated with 33 °C acclimation and changes in mitochondrial morphology as discussed previously. In contrast, we did not detect significant thermal acclimation effects on hemoglobin thermal sensitivity (ΔH°, Fig. 4B), although such effects have been detected in other fish species. (e.g., Oncorhynchus mykiss 49 ).
In this study, we demonstrate clear effects of intraspecific variation and thermal acclimation on Mito-P50. This variation in kinetic properties is consistent with subspecies differences and the effects of thermal acclimation on hypoxia and thermal tolerance. These changes in Mito P50 may help to maintain oxygen diffusion gradients particularly at high temperatures. We thus propose that altered Mito-P50 is involved in differentiating organism-level aerobic and thermal performance between putatively adapted F. heteroclitus subspecies and in response to thermal acclimation and could represent a novel target for thermal adaptation.

Methods
Animals. All procedures were carried out at the University of British Columbia according to the University of British Columbia Animal Care Committee approved protocol #A01-0180. Northern (Fundulus heteroclitus macrolepidotus; Ogden's Pond Estuary, NS, 45°71′N; 61°90′W) and southern (Fundulus heteroclitus heteroclitus; Jekyll Island, GA, 31°02′N; 81°25′W) Atlantic killifish were collected in summer, 2014, and were transported to UBC's Aquatics Facility where they were kept in 190 L tanks with biological filtration. Fish were held at 15 ± 2 °C, and 20 ppt salinity, with a 12:12 L:D photoperiod. Animals were fed once daily to satiation (Nutrafin Max). In June 2015, fish were transferred to 114 L tanks with biological filtration. Temperature was held at 5, 15 or 33 ± 2 °C and all other conditions were maintained. After a minimum of four weeks of thermal acclimation, fish were fasted for 24 h and sampled as described below.
Loss of equilibrium in hypoxia assay. Brackish water (T a = 15 °C, 20 ppt salinity) was circulated in a 50 l plexiglass arena with bubblewrap at the water's surface to prevent O 2 diffusion. Fish were placed in individual containers in the arena (10 fish per trial, 1 fish per container), where the former allowed for full mixing of water while preventing access to the water's surface. PO 2 was monitored continuously using a fiber-optic oxygen probe (NEO-Fox, Ocean Optics, Dunedin, FL). Following 10 min of acclimation to the container, PO 2 was decreased over 30 min by bubbling N 2 gas to one of four PO 2 values (0.84, 0.44, 0.24, or 0.13 ± .02 kPa). When the desired PO 2 was reached, the trial began and time to LOE hyp was recorded. LOE hyp was determined as the time at which fish no longer responded to gentle movement of their containers following which fish were removed and placed in a recovery tank.
Hemoglobin O 2 P50 and hematocrit content. Fundulus heteroclitus were removed from the 5, 15, or 33 °C thermal acclimation tanks at 8:00 AM PST, euthanized by cervical dislocation and weighed. The caudal peduncle was severed and blood was collected from the incision using micro-hematocrit tubes. Micro-hematocrit tubes were centrifuged (10 min, 12,700 g, 25 °C) and hematocrit was measured as the volume percentage of packed RBCs within the total blood sample.
Micro-hematocrit tubes were separated at the boundary between RBCs and plasma. RBCs were re-suspended to approximately the same measured hematocrit content with buffered saline (50 mM HEPES with 100 mM NaCl, final osmolality = 390 Osm kg −1 , pH = 7.8 at 25 °C) for O 2 equilibria experiments 50 . Osmolality of the buffered saline solution was set based on plasma osmolality measurements (10 μL of whole blood measured using standard protocols with an osmometer, Vapro 5520, Wescor, Logan, UT) from five randomly selected 15 °C acclimated F. heteroclitus (396 ± 14 Osm⋅kg −1 mean ± sd).
Oxygen equilibrium curves were generated using protocols described by Lilly et al. 51 . Re-suspended RBCs (3 μL) were sandwiched between two sheets of low density polyethylene that were secured on an aluminum ring with two plastic O-rings. Blood samples were placed in a gas tight tonometry cell modified to fit into a SpectraMax 190 microplate reader (Molecular Devices, Sunnyvale, USA). Assay temperature was maintained at 15 or 33 °C, and blood from the three acclimation temperatures were assayed at both assay temperatures. Samples were equilibrated with pure N 2 for 30 min to achieve full Hb deoxygenation (deoxyHb). O 2 tension was increased by 6 to 12 increments of air (21% O 2 ) balanced with N 2 using a Wӧsthoff DIGAMIX gas mixing pump (H. Wösthoff Messtechnik, Bochum, Germany). Optical density (OD) was measured every 10 s at 390 nm, and 430 nm, which correspond to the isosbestic point (i.e., OD is independent of Hb-O 2 saturation), and maxima for deoxygenated Hb, respectively. Hb was assumed to be fully saturated (oxyHb) with O 2 (100% air) when no change in OD at 430 nm was detected after three equilibration steps.
Hb-O 2 saturation for each equilibration step was calculated using Eqn. 1. Where P50 is the PO 2 at which Hb is 50% saturated and is a measure of Hb-O 2 affinity, and n is the cooperativity (Hill) coefficient 51 . All curves used in the final analyses fit well to the data (r 2 > 0.99). We estimated the effects of acute thermal shifts (T assay = 15 to 33 °C) on Hb-O 2 affinity by calculating the apparent heat of oxygenation using the van't Hoff isochore (ΔH, Eqn. 3 52 ).
T T 50 1 1 2 Where R is the gas constant and T 1 and T 2 are the absolute temperatures 306 and 288 K respectively.
Liver mitochondrial isolation. Seven F. heteroclitus were removed from the 5, 15, or 33 °C thermal acclimation tanks and euthanized for liver mitochondrial isolation as described previously 15 . Liver tissue was excised and pooled into one aliquot of ice-cold homogenization buffer (250 mM sucrose, 50 mM KCl, 0.5 mM EGTA, 25 mM KH 2 PO 4 , 10 mM HEPES, 1.5% BSA, pH = 7.4 at 20 °C). Liver tissue was cut into approximately 1 mm 3 pieces and homogenized with 5 passes of a loose-fitting Teflon pestle followed by filtration through 1-ply cheesecloth. Crude liver homogenate was centrifuged at 4 °C for 10 min at 600 g. The fat layer was removed with aspiration and the remaining supernatant was filtered through 4-ply cheesecloth. Filtered supernatant was centrifuged at 4 °C for 10 min at 6000 g. The resulting pellet was washed twice and suspended in 800 μL of homogenization buffer and stored on ice until experimentation. Protein content was determined using a Bradford assay with BSA as a standard 53 . where  M O 2 is the mitochondrial respiration rate, J max is maximal respiration rate and P 50 is the P O 2 at which respiration rate is half of J max . We accounted for the time delay of the O 2 sensor, background O 2 consumption and internal zero drift at each assay temperature 18 .
Two ml of MiRO5 was air-equilibrated at each assay temperature (15, 33 and 37 °C) followed by the addition of liver mitochondria (0.5 mg protein). A saturating mixture of substrates (glutamate 10 mM, pyruvate 10 mM, malate 2 mM, succinate 10 mM, palmitoyl-carnitine 20 mM) was added to the chamber followed by ADP (2.5 mM) to fuel mitochondrial respiration. Mitochondrial respiration was allowed to proceed until all O 2 was consumed (i.e., anoxia). Anoxic conditions were maintained for 15 min, followed by the termination of the experiment.
The effects of acute temperature shifts on Mito-P50 (i.e., Δ Mito-P50) were estimated by calculating the difference in Mito-P50 between T assay = 15 and 33 °C for each treatment.

Statistics and calculations.
Statistical tests were completed using R software (v3.3.3). Data are presented as mean ± SEM and α was set at.05. We confirmed the presence of normal distributions and homogeneity of variance in our data using Shapiro-Wilk and Bartlett's tests respectively. Sample size is indicated (n) in the respective figure captions.
We compared the effects of subspecies, thermal acclimation, assay temperature using three-separate linear mixed effect models (individual as the random effect) to compare Hb and Mito-P50 and Hill coefficients.
We used separate t-tests to compare subspecies effects on Hb and mitochondrial O 2 binding affinity. Comparisons were made between subspecies assayed at their acclimation temperature (e.g., 33 °C acclimated northern and southern fish assayed at 33 °C). We accounted for an increased false discovery rate by adjusting α using a Benjamini-Hochberg correction for our specific test of subspecies effects.
Thermal acclimation effects (5, 15, and 33 °C) were assessed between subspecies at T assay = 15 °C using a two-way ANOVA. The effects of acute thermal shifts (T assay = 15 to 33 °C) on mitochondrial (Δ Mito-P50) and Hb-P50 (ΔH) were assessed between subspecies and among thermal acclimation treatments using a two-way ANOVA.
Thermal acclimation and subspecies effects on hematocrit content were assessed using a two-way ANOVA. The effects of subspecies and decreasing PO 2 on time to LOE hyp were assessed using a two-way ANOVA.
Data availability. The data generated during the current study are available from the corresponding author on reasonable request.