Low-O₂ acclimation shifts the hypoxia avoidance behaviour of snapper (Pagrus auratus) with only subtle changes in aerobic and anaerobic function

The prevalence of low oxygen (environmental hypoxia) has increased in coastal regions (Diaz and Rosenberg, 2008), and thus it is now more important than ever to understand how different fish species will respond to environmental change. Because hypoxia presents a significant metabolic challenge for most aquatic animals, physiological adaptations to chronic hypoxia generally involve enhancing the capacity for O₂ uptake and delivery (Richards, 2009). For example, many fish species studied to date increase red blood cell numbers and haemoglobin (Hb) concentrations to boost O₂ carrying capacity (Wells, 2009; Wells et al., 1989), restructure gill morphologies to enhance gas exchange (Sollid et al., 2003), increase Hb–O₂ binding affinities to increase O₂ uptake (Wood and Johansen, 1972; Wood et al., 1975) or even modify cardiac function to improve low O₂ performance (Petersen and Gamperl, 2010). When combined, cellular and tissue modifications should enhance the whole animal's capacity to tolerate hypoxia, and be evident as a lowering of the critical O₂ tension (P_crit), which defines the partial pressure of O₂ (P_O2) above which basal metabolic demand (maintenance) is satisfied (Timmerman and Chapman, 2004).

Fish engage an array of physiological mechanisms to combat low oxygen, but may also use behavioural strategies to counter deleterious effects (Richards, 2009). This is particularly true for species that have a limited capacity to adapt physiologically (Pichavant et al., 2003). Reductions in locomotory activity and early avoidance of low O₂ represent two notable behaviours that would undoubtedly help to ensure the survival of fish facing hypoxic conditions (Herbert et al., 2011; Poulsen et al., 2011). Some fish engage these strategies before encountering their respective P_crit (Herbert and Steffensen, 2005; Poulsen et al., 2011), but the New Zealand snapper (Pagrus auratus, Sparidae) does not show these responses (Cook et al., 2011). Snapper are late to leave hypoxic water because low O₂ avoidance commences below their P_crit (Cook et al., 2011). Moreover, snapper do not show any significant change in swimming speed during hypoxic exposure (Cook et al., 2011). The snapper used in this study are presumed to have never experienced low O₂, raising the possibility that behavioural responses and resulting physiological impacts could differ with previous exposure as a result of a loss of naivety and/or acclimatory responses (i.e. physiological adaptation).

To date, the effects of hypoxic acclimation on behaviour of fish in low O₂ conditions remain largely untested. Moreover, little direct evidence identifies how adaptations associated with hypoxia preconditioning influence fish behaviour. Therefore, the present study aimed to resolve whether the behavioural avoidance and swimming speed response of snapper would differ after acclimation to long-term hypoxia. The physiology of snapper was also investigated in detail using both novel and commonly applied measures of aerobic and anaerobic physiology to gauge how physiological changes (cellular, organ and whole animal) integrate with behavioural responses. We hypothesised that snapper would show one of two responses. Firstly, long-term hypoxia could convey a degree of low-O₂ experience to fish whereby they are simply less naive and employ a more cautious avoidance strategy well above P_crit limits. In this scenario, snapper would adjust their behavioural strategy and simply avoid low O₂ earlier without any major changes in physiology. Alternatively, long-term hypoxia could provide greater low-O₂ tolerance across a number of physiological levels (e.g. improved Hb–O₂ transport potential). In this scenario, fish might remain in hypoxia and avoid even lower levels of O₂, feasibly without enlarged levels of low O₂ stress. The present study therefore set out to resolve how physiological and behavioural changes might interact following chronic hypoxia.

Juvenile Pagrus auratus (Forster 1801) (Sparidae; common name snapper or red bream; 150–300 g) were captured by line and barbless hooks from coastal waters around Leigh (36°19′S, 174°48′E, Northland, New Zealand). Following capture, fish were housed in one of two 500 l tanks (maximum of 50 individuals per tank) at the Leigh Marine Laboratory. Fish were provided with a continuous flow of high-quality, aerated seawater for at least 6 weeks before experimentation. After this period, one tank was designated for hypoxia acclimation and the other as a normoxic control. A reduced P_O2 (10.2–12.1 kPa) was maintained in the hypoxia tank using an oxygen controller (Model PR514, PR Electronics, Rønde, Denmark) and an Oxyguard oxygen electrode (Mini probe, Technolab, Mornington, Tasmania, Australia) positioned in the centre of the holding tank. The P_O2 level of 10.2–12.1 kPa was deemed a significant level of hypoxia for this species because it reduces their aerobic scope by >50% at lower experimental temperatures (Cook et al., 2011). An on/off relay output from the controller actuated a solenoid-controlled flow of compressed nitrogen (BOC Gas Supplies, Auckland, New Zealand) through a fine bubble diffuser fitted to the floor of the holding tank. A second relay was programmed to actuate an auxiliary flow of compressed air if P_O2 values fell below the desired set-point of 10.2 kPa. This provided protection against any sudden drop in P_O2 (e.g. as a result of post-prandial metabolism). Normoxic conditions in the control tank (>95% saturation) were maintained by continually passing compressed air through a fine bubble diffuser fitted in a similar fashion to the hypoxia tank. The initial stage of hypoxia acclimation involved decreasing the P_O2 of the hypoxia tank at a rate of ~1 kPa day⁻¹ over a 10 day period until the desired minimum set-point range was established. Fish remained under these conditions for 6 to 12 weeks, enabling investigators to complete the sampling and observational phase of the experiment. Normoxic controls were subject to the same period of experimental holding. P_O2 in both tanks was confirmed regularly with a Cell-Ox probe connected to a WTW 3310 meter (Wissenschaftlich-Technische Werkstätten, Weilheim in Oberbayem, Germany). Water temperatures across the duration of holding and experimentation ranged from 20.1 to 21.8°C. Fish were fed a standard ration of squid and pilchard. All capture, holding and experimental techniques were performed under approval of The University of Auckland Animal Ethics Committee (approval: R711).

Behavioural responses to an avoidable progressive hypoxia stimulus (referred to as ‘escapable hypoxia’) were investigated in an oxygen choice chamber described elsewhere (Cook and Herbert, 2012; Cook et al., 2011). Individual fish occupied a behavioural arena (BA) that consisted of two flows of water separated by a divider. Fish could actively move between these two flows via a square port (10×10 cm) positioned centrally on the divider. Diffusers and baffles were used to create rectilinear flow in the BA (4800 l h⁻¹ combined flow rate) and each side of the BA received water from a large (400 l) gassing tower. P_O2 on either side of the BA was therefore manipulated by purging nitrogen gas (BOC Gas Supplies), or compressed air, through the towers. Oxygen set-points were controlled by two Oxyreg units (Loligo Systems, Tjele, Denmark) coupled to a DAQ device (miniLAB 1008, Measurement Computing, Norton, MA, USA) under the control of Labview software (v. 8.6, National Instruments, Austin, TX, USA). The behavioural activity of fish in the choice box was quantified using a digital camera (Fire-I, Unibrain, San Ramon, CA, USA) that streamed video to a PC running ‘Swistrack’ software (Correll et al., 2006). The movements of fish were sampled at a rate of 15 Hz and were used to resolve avoidance behaviour and activity variables (including swimming speed and spontaneous turning rates). Water temperatures within the choice chamber were actively maintained at 21.0±0.3°C (mean ± 95% CI).

Fish were transferred to the BA at least 18 h prior to experimentation. Following this period, behavioural variables were determined over a 1 h control period, during which any preference for side and normoxic swimming activity was determined. During this 1 h control period, all fish formed a strong preference for one particular water flow (i.e. >80% of time spent on one particular side of the divider) and excursions into the alternate channel were infrequent and never in excess of 15 s. This behavioural routine formation presented investigators with the opportunity to ‘drive’ fish from their preferred side and identify clear avoidance thresholds (Cook et al., 2011). Following the 1 h control period, the preferred channel was deoxygenated progressively at a linear rate of 1 kPa 5 min⁻¹. The P_O2 at which snapper avoided their preferred channel for 30 s was taken as the avoidance threshold. The experiment was terminated at this point; fish were immediately captured and then rapidly euthanized by brain ablation before blood was sampled via caudal venepuncture (<30 s post capture).

A heparinised syringe and needle was used to draw caudal mixed blood at the point of avoidance in the behavioural trial above (N=8 per treatment), as well as a separate resting control group (N=10) for comparison. Whole blood was placed in 75 mm capillaries and haematocrit (Hct; the percentage of red blood cells) was determined after 3 min of centrifugation (Haemocentaur, MSE, London, UK). In addition, 10 μl of whole blood was pipetted into 1 ml of modified Drabkin's reagent and haemoglobin concentration ([Hb]) was calculated after reading absorbance at 540 nm against a blank (Wells et al., 2007). Mean corpuscular haemoglobin concentration (MCHC) was estimated from the ratio of [Hb]:(Hct×100). The remaining blood was then centrifuged (14,000 g, 4°C), and the plasma was separated before being stored at −80°C for later analysis of plasma glucose and lactate using standard enzymatic techniques (described in Cook et al., 2011).

Muscle enzyme activity was determined according to methods detailed previously (Hickey and Clements, 2003). Pre-weighed samples of tissue (~50 mg) were diluted in ice-cold homogenisation buffer (in mmol l⁻¹: 25 Tris-HCl pH 7.8, 1 EDTA, 2 MgCl₂, 50 KCl, 0.5% Triton-X 100) to a ratio of 1:10 (w/v). Samples were homogenised (TissueLyser II, Qiagen, Auckland, New Zealand) and then centrifuged (20,000 g, 4°C), with the supernatant retrieved for analysis. Lactate dehydrogenase (LDH) activity was determined by adding 20 μl of appropriately diluted tissue homogenate to 180 μl of LDH assay mix (in mmol l⁻¹: 100 Tris-HCl pH 7.0, 1 EDTA, 2 MgCl₂, 1 dithiothreitol and 0.15 NADH). The oxidation of NADH was measured at 340 nm in a plate reader (Spectramax 340, Molecular Devices, Sunnyvale, CA, USA) after addition of 25 μl 1.5 mmol l⁻¹ pyruvate. Citrate synthase (CS) was measured by adding 30 μl of dilute tissue homogenate to 170 μl of CS assay mix [in mmol l⁻¹: 50 Tris-HCl pH 8.0, 0.1 acetyl coenzyme A and 0.2 5,5′-dithiobis-(2-nitrobenzoic acid) (DTNB)]. The reduction of DTNB was followed at 412 nm after the addition of 25 μl 5 mmol l⁻¹ oxaloacetate. LDH and CS enzyme activity were expressed as U mg⁻¹ but also in terms of their ratio to one another (LDH:CS). The use of the LDH:CS ratio accounted for individual variability in enzyme activities and, importantly, enabled us to resolve differences in the relative anaerobic and aerobic capacities of fish between the two treatments (Hochachka et al., 1983). All chemicals were sourced from Sigma-Aldrich (Auckland, New Zealand)

As all blood components had been previously washed from ventricular tissue within the modified Langendorff preparation (see below), ventricular myoglobin concentrations ([Mb]) could be determined using the cyanmetmyoglobin (cyanmetMb) method (Drabkin, 1950). Two hundred microlitres of supernatant (also utilised for enzyme activity analyses) was added to 800 μl of modified Drabkin's reagent. Absorbance at 540 nm was recorded in 1 cm wide cuvettes (Multiskan Spectrum, Thermoscientific, Vaanta, Finland). Tissue [Mb] was calculated using a cyanmetMb extinction coefficient of 10.36 mmol cm⁻¹ according to the methodology of Viriyarattanasak and colleagues (Viriyarattanasak et al., 2011).

Whole-blood binding properties and the fixed acid Bohr effect of Hb were investigated in venous mixed blood drawn from the caudal vein of rested fish in each treatment (N=7). Samples were prepared by suspending 50 μl of whole blood in 4 ml of 50 mmol⁻¹ Hepes salt solution (125 mmol l⁻¹ NaCl), buffered to a range of physiologically relevant pH levels (7.8, 7.6, 7.4, 7.2, 7.0 and 6.8). Upon dilution of whole blood, residual catecholamines that might influence blood O₂ binding properties were expected to degrade rapidly (Wells et al., 2003). Antifoam (2 μl) was added to the solution immediately before analysis. The binding properties of haemoglobin solutions at each pH were then investigated using a Haem-ox analyser (TCS Research Products, New Hope, PA, USA). Oxygen equilibrium curves (OECs), Hill's cooperativity coefficient (n) and the P_O2 at which Hb was half O₂ saturated (Hb-P₅₀) were calculated using Haem-ox analytical software (version 2.14, TCS Research Products). The Bohr coefficient (ϕ) was determined using the standard equation, ϕ=ΔlogHb-P₅₀/ΔpH. The Root effect was recorded as the decrease in maximal Hb saturation at each measured pH level relative to a reference value for maximal Hb saturation (Hb-P₁₀₀) at pH 7.8 and 21°C (Regan and Brauner, 2010; Wells and Dunphy, 2009).

The heart of resting, euthanised snapper was perfused in a modified Langendorff preparation. The bulbous arteriosis was cannulated and perfused with a gravity delivered flow of teleost Ringer's solution, which, in combination with spontaneous ventricular contractions, served to wash all blood from the tissue. The Ringer's solution was composed of (in mmol l⁻¹, unless otherwise indicated): 159.6 NaCl, 2.1 KCl, 0.99 MgCl₂, 1.30 CaCl₂, 10 glucose, 3 Hepes acid, 6.99 Hepes sodium salt, 0.30 Na-glutamate, 0.40 l-glutamate, 0.02 Na-aspartate, 0.05 dl-carnitine, 1 mg l⁻¹ insulin (porcine), 1 mg l⁻¹ thiamine pyrophosphate-co-carboxylase, pH 7.6 (Forgan and Forster, 2010). After 5–10 min of perfusion, the heart was removed from the perfusion setup and prepared for mitochondrial respiration assays. Permeabilised heart fibres were prepared as follows. Spongy myocardium excised from the perfused heart was placed into modified, ice-cold relaxation medium (BIOPS, in mmol⁻¹: 2.77 CaK₂EGTA, 7.23 K₂EGTA, 5.77 Na₂ATP, 6.56 MgCl₂·6H₂O, 20 taurine, 20 imidazole, 0.5 dithiothreitol, 50 K-MES, 15 Na-PCr and 50 sucrose, pH 7.1) and then teased apart into fibre bundles and transferred to permeabilising solution (BIOPS + 50 μg ml⁻¹ saponin). After gentle shaking (30 min, 0°C) the tissue was removed from the permeabilising solution and washed three times in a modified mitochondrial respiratory medium [MiRO5, in mmol l⁻¹ unless otherwise indicated: 0.5 EGTA, 3 MgCl₂·6H₂O, 60 K-lactobionate, 20 taurine, 10 KH₂PO₄, 20 Hepes, 160 sucrose and 1 g l⁻¹ bovine serum albumin (Gnaiger et al., 2000)]. Fibres were removed from MiRO5, blotted dry and weighed into bundles (~5–10 mg) for respirometric determinations. Isolated mitochondria were prepared by immersing 50–100 mg of cardiac tissue into ice-cold MiRO5 followed by manual mincing. After washing two to three times with MiRO5, the tissue was digested (MiRO5 + 5% trypsin) for 30 min and then filtered through fine muslin before re-suspension in 1000 μl MiRO5. Following centrifugation (700 g, 10 min, 4°C), the supernatant was transferred to a new 1.7 ml microcentrifuge tube and spun (8000 g, 10 min, 4°C). The supernatant was discarded and 100 μl of MiRO5 was added to resuspend the remaining pellet containing isolated mitochondria. Protein concentration was determined from an aliquot of the mitochondrial suspension using the BCA protein assay (ThermoFisher Scientific, Auckland, New Zealand) according to guidelines provided. All chemicals utilised were obtained from Sigma-Aldrich.

Fibres typically require super-saturation with oxygen in order to determine maximal flux capacities (Gnaiger, 2009). In general settings, such as with mammalian cardiac or skeletal muscles, this may be problematic as muscle fibres are typically perfused by capillaries, which are not functional in vitro. However, fish cardiomyocytes, in particular the spongy myocardium of the snapper, is perfused by luminal blood. This permitted us to test the oxygen dependence of flux (J_O2) of permeabilised fibres isolated from snapper ventricle, without concerns of experimental diffusion limitation.

Using an Oroboros Oxygraph-2K respirometer (Oroboros Instruments, Innsbruck, Austria), the mitochondrial function of heart tissue from the two acclimation treatments was evaluated by successive addition of tricarboxylic acid (TCA) cycle substrates (malate, glutamate, pyruvate then succinate) to stimulate Complex I, then Complex I+II, respiratory states (Gnaiger, 2009). Excess ADP (2.5 mmol l⁻¹) was added to stimulate oxidative phosphorylation (OXPHOS) (see Fig. 3). Following these determinations, chambers were equilibrated to ~21 kPa. Once fueled with further respiratory substrates (in mmol l⁻¹: 5 malate, 10 glutamate, 10 pyruvate, 10 succinate and 2.5 ADP), oxygen was allowed to deplete into anoxia with oxygen flux rates determined using Datlab 4.2.3.2.7 (Oroboros Instruments, Innsbrook, Austria). The characteristics of mitochondrial respiration during progressive O₂ depletion [oxygen-dependent flux (ODF)] were quantified by transforming exported data according to the following equation: ODF=(J_O2/O₂)/O₂. When plotted graphically, this transformed measure of respiration was observed to decrease in a linear fashion, enabling determination of slope characteristics – the gradient of this line is here termed the oxygen-dependent flux index (ODFI). ODFI values are greatest in tissues showing high oxygen dependence and lowest in tissues showing low oxygen dependence.

Isolated mitochondria from the two acclimation treatments were initially evaluated by successive addition of TCA cycle substrates (as above). As isolated mitochondria are not diffusion limited, the Mito-P₅₀ (i.e. the P_O2 at which mitochondrial respiration is 50% of maximum recorded levels) of isolated mitochondria was determined following respiratory depletion of oxygen from saturating conditions (~21 kPa) to anoxia.

The mitochondrial ultrastructure of cardiac tissues (ventricle) was investigated to identify whether hypoxic preconditioning changed the structural morphology of the heart. Excised, perfused sections of spongy myocardium were immersed in fixative buffer (10 mmol l⁻¹ Hepes, 250 mmol l⁻¹ sucrose, pH 7.1, 2.5% glutaraldehyde) and then stored at 4°C. Preparation involved washing samples three times (10 min wash⁻¹) with Sorensen's buffer (0.1 mol l⁻¹) before being fixed (1% osmium tetroxide in 0.1 mol l⁻¹ Sorensen's buffer). This was followed by a series of alcohol washes (30–100% ethanol, dry ethanol, and then two washes in 100% acetone). Samples were infiltrated with epoxy resin (1:1, 812 epoxy:acteone), later replaced with 100% epoxy resin and left overnight. Resin was removed the following morning and replaced with fresh 100% epoxy. After a further 6 h, tissue was transferred to moulds and kept at 60°C for 48 h. Following trimming, ultra-thin (80 nm) sections of cardiac tissue were then stained with 2% aqueous uranyl acetate. Images were acquired at the following magnifications, ×5600, ×8800 and ×66000, using a Phillips CM-10 transmission electron microscope (Eindhoven, The Netherlands). Analysis of mitochondrial surface density (% cover) and cristae density (cristae μm⁻¹) was performed in ImageJ software (v. 1.45, National Institutes of Health, Bethesda, MD, USA) using protocols described by Hickey and colleagues (Hickey et al., 2009).

Respirometric measures of standard metabolic rate (SMR), routine metabolic rate (RMR), maximum metabolic rate (MMR) and critical oxygen partial pressure (P_crit) were determined using intermittent closed-phase respirometry with an automated protocol (Cook et al., 2011). Fish were transferred from their holding tanks to a 5 l respirometry chamber supplied with fully saturated (normoxic) water at constant temperature (21±0.1°C). A computer running custom software controlled the flush, wait and measure cycles of the respirometry protocol chamber (4, 1 and 5 min intervals, respectively) and recorded the closed-phase decline in chamber O₂ saturation via a WTW CellOx probe connected to a 3310 meter (Wissenschaftlich-Technische Werkstätten, Weilheim, Germany). The mass-specific rate of oxygen consumption () was calculated from the decline in closed-phase O₂ (i.e. one measure of every 10 min) according to standard formulae, detailed previously (Cook et al., 2011). SMR was thus resolved over 40 h of normoxia using 240 measures of and the 15% quantile method (Chabot and Claireaux, 2008; Cook et al., 2011; Dupont-Prinet et al., 2010). RMR was defined as the mean throughout the duration of recording. MMR was defined as the peak measurement of observed during the first three recorded measurement cycles. Once SMR, RMR and MMR were resolved, the P_O2 of water supplying the respirometry chamber was reduced progressively to the following levels: 15, 12, 9, 7.5, 6 and 4 kPa. These steps were used to identify the P_crit where was reduced below SMR and fish were in an oxy-conforming state (Behrens and Steffensen, 2007). Defined P_O2 set-points were achieved by circulating water through a 40 l gassing tower that was intermittently purged with compressed nitrogen for deoxygenation. Three measures were recorded at each P_O2 level. A one-way ANOVA and a Tukey's post hoc test were used to identify levels of P_O2 at which fish could not maintain SMR (i.e. resting <SMR). The line of oxygen dependence for the animal was subsequently derived from a linear regression fitted to resting <SMR with a forced y-intercept of zero. The point of intercept between the extrapolated value of SMR and the line of oxygen dependence was taken as P_crit.

The behavioural responses to progressive low O₂ were investigated using a two-way repeated-measures non-parametric ANOVA (Shrier–Ray–Hare comparison). Behavioural observations at exposures <4 kPa were omitted from statistical analysis as some fish had avoided low O₂ prior to this observation period (resulting in poor statistical power as a result of reduced sample sizes). Comparisons between avoidance levels and the physiological state of treatments at the point of avoidance were performed using a Student's t-test. Between-treatment comparisons of descriptors of mitochondrial ultrastructure, enzyme activities and heart [Mb] were all performed using Student's t-tests, with data appropriately (e.g. log) transformed when variances were not equal. OEC data were compared using two-way repeated-measures ANOVA, with Tukey's post hoc tests incorporated where applicable. The Schrier–Ray–Hare test was performed in SPSS (v. 18, IBM, www.ibm.com/software/analytics/spss/), but all other analyses were performed in SigmaPlot (v. 11, Systat Software, www.systat.com). All results are presented as means ± 95% CI. Significance was accepted at P<0.05.

Individuals exposed to moderate hypoxia for 6–12 weeks avoided low-O₂ conditions at significantly lower P_O2 (3.3±0.7 kPa) than normoxia-acclimated individuals (5.3±1.1 kPa, F=12.20, P<0.05; Table 1), but at all levels of P_O2, the two treatments behaved similarly in terms of swimming speed, turning rate, percentage of time in lowest P_O2 and the frequency of hypoxic excursions (Fig. 1, refer to Table 2 for statistical summary). No treatment differences or interaction effects existed in any behavioural variable investigated.

The haematological profiles of both normoxia- and hypoxia-acclimated snapper were different to those of rested (unexposed) controls immediately after avoidance (see Table 1). Specifically, exposing normoxia- and hypoxia-acclimated snapper to hypoxia yielded a significant increase in [Hb] (t=−2.14, P<0.05 and t=−3.32, P<0.01, respectively), plasma lactate (t=−2.219, P<0.01 and t=−7.12, P<0.01) and plasma glucose (t=−3.32, P<0.01 and t=−8.90, P<0.01). A significant increase in blood Hct was also observed in hypoxia-acclimated individuals (t=−3.96, P<0.01), but normoxia-acclimated fish only showed a near-significant increase (t=−1.95, P=0.08). Despite hypoxia-acclimated fish being exposed to lower P_O2 (see above) there was no significant difference in [Hb] (t=1.29, P=0.22), Hct (t=−0.27, P=0.79), MCHC (t=−1.49, P=0.15), plasma lactate (t=−1.72, P=0.49) or plasma glucose (t=−0.30, P=0.77) between the two groups following avoidance. This lack of difference in physiology between the two treatments is particularly important and is discussed further below.

The haemoglobin of snapper exhibits typical responses to pH. For example, pH had a strong negative effect on Hb-P₅₀ (F=21.38, P<0.01, Fig. 2A), whilst exerting a positive effect on binding cooperativity (n-value, F=182.42, P<0.01; Fig. 2B). Snapper blood also revealed a modest Root effect according to the reduction in maximal Hb–O₂ saturation at pH<7.4 (F=32.94, P<0.01; Fig. 2C). In terms of treatment differences, the Hb-P₅₀ of hypoxia-acclimated snapper was on average 12% higher than normoxia-acclimated individuals at any level of pH (F=7.84, P<0.01) but the Bohr coefficient of normoxia-acclimated fish (ϕ=−1.02±0.12) was similar to that of hypoxia-acclimated individuals (ϕ=1.08±0.12; t=0.66, P=0.524). The sigmoidal shape of the oxygen binding curve (n-value), at any level of pH, was also not affected by treatment conditions (F=1.13, P=0.29).

Exposure to moderate hypoxia across a 6 week period led to other minor modifications in the resting physiology of snapper (for a summary of results, see Table 1). With respect to haematology, only [Hb] differed between treatments under resting conditions; the [Hb] of hypoxia-exposed individuals was 8% lower than that of normoxic individuals (t=−2.14, P<0.05). Furthermore, only LDH activity in the red skeletal muscle was different between hypoxia- and normoxia-acclimated fish; all other enzyme activities were similar (Table 3). It is important to note that when LDH was expressed relative to CS, the LDH:CS ratios were higher in cardiac and red skeletal muscle of hypoxia-treated fish. LDH activity and LDH:CS in white muscle tissue did not differ significantly. Snapper ventricular myoglobin concentrations were not affected by long-term hypoxia (~2.7 mg g⁻¹, pooled values).

Respiration flux of isolated cardiac mitochondria was not significantly affected by the hypoxia acclimation treatment (Table 4). Complex I- and II-supported oxidative phosphorylation (Complex I, OXPHOS and Complex I+II, OXPHOS) and ‘leak’ rates (synonymous with Complex I respiration) were all statistically similar between treatments. Despite these similarities, permeabilised cardiac fibres showed different oxygen-uptake dynamics: the respiration of fibres from the heart of hypoxia-acclimated fish appeared to be less affected by the diffusive oxygen gradient between the respiratory medium and active mitochondria. This was reflected by a lower conformance to P_O2 (i.e. low ODFI scores) in the hypoxia-acclimated fish (t=−2.560, P=0.03; Fig. 3). Snapper heart mitochondria occupied ~19% of myocyte volume with mean cristae densities of ~22 cristae μm⁻¹ (Fig. 4), and were not altered by 6 weeks exposure to moderate hypoxia (see Table 4 for statistical summary).

Long-term exposure to moderate hypoxia had no significant effect on the rates of O₂ consumption (Fig. 5). The SMR of the normoxia treatment group was 152.1±13.4 mg O₂ kg⁻¹ h⁻¹ compared with 155.4±10.4 mg O₂ kg⁻¹ h⁻¹ in the hypoxia treatment (F=0.46, P=0.51). RMR was 196.69±10.83 and 209.2±20.4 mg O₂ kg⁻¹ h⁻¹ for the normoxic and hypoxic treatment groups, respectively (F=1.99, P=0.18). MMR in the normoxia treatment group was 442.38±28.64 and 420.1±21.5 mg O₂ kg⁻¹ h⁻¹ in the hypoxia treatment group (F=1.29, P=0.28). P_crit values from the normoxic fish (6.93±0.61 kPa) were no different to those of the hypoxia-acclimated fish (7.01±0.55 kPa; t=−0.19, P=0.85).

Snapper exposed to moderate long-term hypoxia (10.2–12.1 kPa) avoid low O₂ at a significantly lower level of P_O2 than normoxic controls (3.1±0.7 vs 5.5±1.1 kPa; Table 1), signifying that hypoxia acclimation does not sensitise the behavioural response of this species as a result of low-O₂ experience. It has previously been reported that snapper avoid hypoxia below P_crit thresholds (Cook et al., 2011) and this was again observed with both groups in the present study. It therefore seems likely that long-term hypoxia does not encourage snapper to adopt a more cautious avoidance response by leaving low O₂ conditions at higher P_O2. In fact, quite the opposite was observed. Given that each treatment was exposed to a standardised and linear drop in ambient P_O2, hypoxia-acclimated individuals spent a longer period of time exposed to critically low O₂ that increased in severity (i.e. 2.4 kPa lower P_O2 for an extra 12 min). Yet surprisingly, hypoxia-acclimated fish did not experience higher levels of low-O₂ stress because levels of plasma lactate were comparable between treatments. The ability of fish to tolerate more severe hypoxia for greater periods of time, but show similar levels of physiological perturbation at the point of avoidance, would strongly suggest that hypoxia-acclimated fish gained a significant advantage in terms of improved low-O₂ tolerance. This was presumably the result of a change in physiology at low O₂ because fish did not show any downregulation of swimming speed or turning behaviour that could otherwise have reduced energetic expenditure and allowed fish to reside in hypoxia for longer periods (Fig. 1A,B).

Exposing fish to long-term hypoxia is typically considered to improve hypoxia tolerance through alterations in Hb. For example, elevated [Hb] enhances the blood O₂ carrying capacity (Lai et al., 2006; Petersen and Gamperl, 2011; Wells et al., 1989). Similarly, increased Hb–O₂ affinities are commonly reported to improve the O₂-uptake capacity of Hb during environmental hypoxia (Weber et al., 1976; Wood and Johansen, 1972). Both features can be viewed as beneficial (adaptive) to the survival of fishes in hypoxic waters. However, following 6 weeks of chronic hypoxia, snapper did not show many of the responses commonly observed in other studies. Indeed, both the concentration and O₂ affinity of Hb decreased by 8 and 12%, respectively, following hypoxia exposure, changes that would not typically be deemed adaptive. The drop in resting [Hb] should be detrimental, unless this change accompanied a decreased Hct, which could lower cardiac energetic requirements by decreasing blood viscosity (Gallaugher et al., 1995). However, no changes in Hct were observed, suggesting that the reduction in [Hb] would not have benefitted cardiac function. Because Hct, [Hb] and MCHC levels were equivalent in normoxia- and hypoxia-acclimated fish after acute low O₂ (Table 1), the differences observed in resting fish could have been due to altered red blood cell distribution (i.e. splenic storage). With respect to the O₂-binding characteristics of Hb, the observed drop in Hb–O₂ affinity may have provided an adaptive benefit as it would favour O₂ offloading at the tissues. Indeed, this condition could aid aerobic tissue function by elevating O₂ diffusion gradients while also potentially increasing the O₂ content of the venous return (Brauner and Wang, 1997; Wang and Malte, 2011) and the availability of O₂ to the snapper heart. Although generally considered atypical for fish in hypoxia, the observed drop in Hb–O₂ affinity of snapper is consistent with the small decrease in Hb–O₂ affinity of lowland mammals during moderate hypoxia (Lenfant, 1973). It could therefore be argued that exposing snapper to more extreme levels of hypoxia might have induced more ‘typical’ haematological responses (e.g. increased Hb–O₂ affinity). However, 10.2–12.1 kPa does not actually represent a moderate level of P_O2 for snapper. Instead, with a >50% reduction of metabolic scope at this level (Cook et al., 2011), it is unlikely that our fish would have retained adequate fitness characteristics at lower P_O2. This level of hypoxia was therefore chosen as challenging but ethically acceptable for long-term acclimation. Interestingly, the haematological response observed in snapper could potentially limit O₂ uptake potential, but no measures of whole-animal O₂ uptake (including P_crit) actually differed between the two groups (Fig. 5). No evidence exists to suggest that the drop in [Hb] and Hb–O₂ affinity hindered O₂ uptake during hypoxia. It therefore seems plausible that snapper possessed adequate capacities for O₂ uptake during long-term hypoxia, and that a physiological advantage was possibly gained with a prioritisation of O₂ delivery to tissues.

Measures of in situ O₂ utilisation identified improved low-O₂ performance in the cardiac mitochondria of hypoxia-acclimated snapper. Respiratory function in most teleost hearts is oxygen dependent, with a drop in O₂ decreasing respiratory function in a near-linear manner (Forgan and Forster, 2010). Although this is consistent with our findings, the respirational flux of cardiac tissue from the hypoxia-acclimated snapper was less influenced by decreasing [O₂] than that of normoxia-acclimated snapper. We therefore speculate that hearts from hypoxia-acclimated snapper may have been able to maintain ATP synthesis at lower O₂ levels. Although no statistical differences in mitochondrial flux capacities were observed between treatments, hypoxia-acclimated snapper trended higher in terms of maximal oxidative phosphorylation capacity (P=0.06; Table 4). This apparent increase in flux may act to increase oxygen gradients and enhance flux into cardiac muscles, and this effect is perhaps better observed in terms of the ODFI. Differences in mitochondrial oxygen-uptake kinetics could also be explained by alterations in the electron transport systems of mitochondria, including changes in the expression (amounts or isoforms) of cytochrome c oxidase, and/or changes in the positioning of mitochondria within cardiomyocytes (Gnaiger et al., 1998). Although no ultrastructural differences were immediately apparent within hearts, analyses could not determine the relative placement and/or arrangements of mitochondria within cardiomyocytes, which would alter diffusion distances.

The observed alterations in blood function and mitochondrial function following long-term hypoxia did not correspond with changes in whole-animal O₂ uptake (i.e. SMR, RMR and MMR), and neither did whole-animal P_crit shift with treatment. This response contrasts that of a lowered P_crit for the sailfin molly (Poecilia latipinna) following more severe low O₂ over comparable time periods (Timmerman and Chapman, 2004). However, because snapper did not increase the level of circulating erythrocytes or boost O₂-uptake capacity with an increase in Hb–O₂ affinity, it is perhaps unrealistic to expect a lowering of P_crit thresholds as a result of long-term hypoxia in this species. In fact, whole-animal O₂-uptake measures and P_crit may not actually provide the best estimate of low-O₂ tolerance because O₂ within the respiratory cascade is governed by the rate of O₂ uptake, O₂ supply and O₂ utilisation at the tissues. Indeed, improved hypoxia tolerance could feasibly occur without a change in P_crit if, for example, O₂ supply was enhanced through the lowering of Hb–O₂ affinity and/or improved efficiencies in the handling of O₂ by mitochondria, as seen in the present study (Fig. 2, Table 4). Although not investigated in the present study, changes in tissue capillarisation may have also contributed to improved low-O₂ tolerance in hypoxia-acclimated snapper. Such adjustments could partly explain why hypoxia-acclimated snapper occupied lower O₂ levels without a change in P_crit thresholds.

The red tissue groups of skeletal and cardiac muscle showed an increased capacity for anaerobic metabolism, indicated by relatively higher LDH:CS ratios (Table 3). Elevation of these ratios suggests that when snapper are chronically challenged by low P_O2 (10.2–12.1 kPa), anaerobic systems may compensate for constrained aerobic ATP synthesis. Similar increases in anaerobic potential have been observed in other species of fish exposed to periods of chronic hypoxia (Greaney et al., 1980; Johnston and Bernard, 1982; Ton et al., 2003; Yang et al., 1992; Zhou et al., 2000), but not in others (Driedzic et al., 1985; Martínez et al., 2006; Zhou et al., 2000). Whilst these adjustments in anaerobic enzyme activity were possibly important to snapper during moderate long-term hypoxia (10.2–12.1 kPa), hypoxia-acclimated fish in the behavioural trials also showed signs of improved low-O₂ tolerance at sub-critical P_O2 levels (<7 kPa), where fish would likely struggle to meet basal metabolic demands using aerobic pathways alone. It is therefore plausible that the anaerobic adjustments provided a dual benefit during long-term hypoxia. For example, anaerobic ATP supply and perhaps oxidative lactate metabolism would be enhanced during moderate hypoxia, but ATP generation at sub-critical P_O2 might also have been improved. Unfortunately, our study provides no real insight into the role of anaerobic function during either moderate or acute hypoxia because greater focus was placed on identifying aerobic rather than anaerobic mechanisms of adaptation. Further investigations are therefore required to gain a deeper understanding of the metabolic adaptations that improve both the chronic and acute low-O₂ tolerance of snapper and other fish species.

Snapper exposed to moderate levels of hypoxia for 6 to 12 weeks showed a significant shift in their hypoxia avoidance threshold from 5.3±1.1 to 3.3±0.7 kPa, a major alteration in behaviour that did not result in greater levels of physiological perturbation (stress) after avoidance. The ability to endure hypoxia for longer was not achieved through differences in swimming activity, but was likely a result of subtle adjustments in mitochondrial uptake efficiency and anaerobic capacity within the cardiac and skeletal red muscle, or other factors unidentified at this time. Although our findings are contrary to general hypotheses, we suggest that the decrease in [Hb] and Hb–O₂ binding affinity, in combination with improved mitochondrial uptake efficiencies, might have helped to improve low-O₂ tolerance by boosting tissue O₂ delivery without any loss in whole-animal O₂ uptake. This study therefore presents novel insight into how snapper might cope with low-O₂ challenge after long-term acclimation. Although the most common indicators of aerobic and anaerobic performance were screened in the present study, other physiological strategies may exist that improved the hypoxia tolerance of snapper. Despite this possibility, snapper are undoubtedly limited in their ability to adapt to low-O₂ conditions and do not demonstrate the same level of physiological plasticity as other more hypoxia-tolerant species (Timmerman and Chapman, 2004; Wood and Johansen, 1972; Yang et al., 1992; Zhou et al., 2000). This presents a cause for concern, should O₂ conditions within the habitat of snapper deteriorate.

Technical assistance and support provided by Adrian Turner during microscopic analyses is gratefully acknowledged. Technical staff Murray Birch and Peter Browne fabricated the experimental choice chamber and respirometry setups. John Atkins designed and built all respirometric software and hardware components.

 
• BA

  behavioural arena

 
• CS

  citrate synthase

 
• Hb

  haemoglobin

 
• Hct

  haematocrit

 
• J_O2

  mitochondrial O₂ flux

 
• LDH

  lactate dehydrogenase

 
• MCHC

  mean corpuscular haemoglobin concentration

 
• MMR

  maximum metabolic rate

 
• 

  metabolic oxygen consumption

 
• n

  Hill's cooperativity coefficient

 
• RMR

  routine metabolic rate

 
• ODF

  oxygen-dependent flux

 
• ODFI

  oxygen-dependent flux index

 
• OEC

  oxygen equilibrium curve

 
• OXPHOS

  oxidative phosphorylation

 
• Hb-P₅₀

  P_O2 at which haemoglobin is 50% oxygen saturated

 
• Hb-P₁₀₀

  P_O2 at which haemoglobin is 100% oxygen saturated

 
• Mt-P₅₀

  P_O2 at which mitochondrial respiration is 50% of maximal levels

 
• P_crit

  critical oxygen partial pressure

 
• P_O2

  partial pressure of oxygen₂

 
• SMR

  standard metabolic rate

 
• TCA

  tricarboxylic acid

 
• ϕ

  Bohr coefficient