Extreme blood-boosting capacity of an Antarctic fish represents an adaptation to life in a sub-zero environment

It is well documented that performance-enhancing strategies such as blood doping, the practice of boosting one's blood oxygen carrying capacity, are used by human athletes seeking a competitive advantage (Momaya et al., 2015; Tokish et al., 2004). The resulting improvements in oxygen delivery afford these athletes substantial increases in their aerobic capacity, which translates into an enhanced athletic performance (Momaya et al., 2015; Tokish et al., 2004). Increasing blood oxygen carrying capacity can be achieved naturally via high-altitude training or artificially via the administration of illicit substances (i.e. erythropoietin, which stimulates an increased production of erythrocytes) and autologous blood transfusions (i.e. reinfusing the donor's stored blood back into their bloodstream) (Momaya et al., 2015; Tokish et al., 2004; WADA, https://www.wada-ama.org). In most professional sports, the use of erythropoietin and autologous blood transfusions is prohibited as they provide an unfair competitive advantage (i.e. ∼5–25% increase in blood oxygen carrying capacity) and pose a significant risk to health (i.e. hypertension, myocardial infarction, pulmonary embolism and an increased risk of thrombosis) (Momaya et al., 2015; Tokish et al., 2004). Interestingly, the practice of boosting the oxygen carrying capacity of blood can also occur naturally via contraction of the spleen, which results in the ejection of stored erythrocytes into the bloodstream. While humans have a relatively small splenic reservoir and limited use of this strategy, it is much more developed in other members of the animal kingdom. For example, it has been demonstrated that a number of fish species (Axelsson, 2005; Wells and Weber, 1990; Yamamoto et al., 1980), marine mammals (Hurford et al., 1996; Zapol, 1987) and terrestrial mammals (Hannon et al., 1985; Turner and Hodgetts, 1959; Perrson et al., 1973a,b) are able to boost their blood oxygen carrying capacity by ∼40–60% in response to metabolically demanding situations. In fact, the spleen of Weddell seals (Leptonychotes weddellii) has been referred to as a natural ‘scuba tank’, as the release of oxygenated red blood cells tightly regulates the arterial oxygen content during diving (Hurford et al., 1996; Zapol, 1987). However, the most remarkable capacity for splenic blood boosting in vertebrates can be observed in an Antarctic notothenioid fish species, the bald notothen (Pagothenia borchgrevinki), which have been reported to more than double the oxygen carrying capacity of their blood during exercise (Davison et al., 1988; Franklin et al., 1993).

The opening of the Drake Passage (∼40 million years ago), followed by the Tasmanian Gateway (∼25 million years ago), allowed water to flow around the entirety of the Antarctic continent and ultimately led to the formation of one of the most thermally extreme and challenging environments that marine teleosts currently inhabit (Beers and Jayasundara, 2015; Clarke and Johnston, 1996; Eastman, 2005). The combination of stable water temperatures approaching the freezing point of sea water (−1.87°C), the presence of sea ice and the seasonality of primary production poses substantial ecological and physiological challenges for Antarctic notothenioid fishes, such as the bald notothen (Beers and Jayasundara, 2015; Clarke and Johnston, 1996; DeVries and Eastman, 1981; Eastman, 2005). As these fish are subjected to temperatures well below the equilibrium freezing point of their tissue fluids (−0.8°C), a necessity for survival was the evolution of antifreeze molecules that are secreted into the bloodstream (DeVries, 1988). However, the blood viscosity of these fish is substantially elevated as a result of the combination of a high concentration of antifreeze molecules, the extremely low temperatures and circulating erythrocytes, which cumulatively increase the workload on the heart (Axelsson, 2005; Farrell, 1991; Graham and Fletcher, 1985; Hemmingsen, 1991). The energetic cost and cardiovascular consequences (i.e. chronic hypertension, increased risk of thrombosis, increased risk of myocardial and cerebral infarction, increased cardiac wear and tear, and chronic ischaemia) of transporting highly viscous blood through the cardiovascular system may constitute important driving forces for the numerous cardiovascular cold adaptations present in Antarctic fishes (Axelsson, 2005; Beers and Jayasundara, 2015; Farrell, 1991; Hemmingsen, 1991; Lowe, 1992; Sidell and Driedzic, 2012; Willis and Patterson, 2013).

In order to reduce blood viscosity and cardiac workload, it has been proposed that an early cardiovascular cold adaptation in Antarctic fishes was the coupling of a low vascular resistance with substantial reductions in circulating erythrocytes (Axelsson, 2005; Buckley et al., 2014; Hemmingsen, 1991). In fact, a unique evolutionary development of the notothenioids is the complete loss of functional erythrocytes in adult members of the family Channichthyidae (icefish), while in red-blooded notothenioids, the number of circulating erythrocyte cells at rest is approximately half that found in temperate teleost species (Axelsson, 2005; Graham and Fletcher, 1985; Hemmingsen, 1991). While the complete loss or decrease of erythrocytes in notothenioids is associated with a substantial reduction in blood viscosity, it also coincides with similar reductions in blood oxygen carrying capacity, which has inevitably led to the development of a range of other cardiovascular modifications (Axelsson, 2005; Axelsson et al., 1992; Hemmingsen, 1991; Joyce et al., 2018, 2019). In the icefish group, these adjustments consist of larger blood volume, an enlarged heart and significantly reduced vascular resistance, whereas in the red-blooded notothenioids, the level of circulatory adjustments is believed to differ depending on the ecology of the various species (Axelsson, 2005; Axelsson et al., 1992; Buckley et al., 2014; Joyce et al., 2018, 2019). It has been proposed that the next step in the evolutionary sequence for nototheniod fishes was an increase in their capacity for systemic oxygen delivery (i.e. the product of cardiac output and blood oxygen content) (Axelsson, 2005; Axelsson et al., 1992; Buckley et al., 2014). For example, the bald notothen – a more active, cryopelagic red-blooded notothenioid – tends to have a higher cardiac output and relative ventricular/atrial mass than the relatively sedentary, benthic emerald rockcod, Trematomus bernacchii, which is able to perform daily activities with a reduced oxygen carrying capacity because of the high oxygen solubility of seawater/plasma and its low metabolic demand (Axelsson et al., 1992). In addition, we argue that the splenic blood-boosting strategy of bald notothens represents a cardiovascular cold adaptation for increasing systemic oxygen delivery, as it allows them to match their oxygen carrying capacity with their energetic demand (Axelsson, 2005; Davison et al., 1988; Franklin et al., 1993). For example, when in a resting and unfed state, these fish should be able to reduce blood viscosity, and thus cardiac workload, by sequestering a large proportion of their erythrocytes in the spleen (Davison et al., 1988; Franklin et al., 1993). However, during instances of elevated metabolic demand (e.g. exercise and acute stress), these fish are extremely adept at rapidly increasing arterial oxygen carrying capacity and venous oxygen reserves by releasing erythrocytes from the spleen (Davison et al., 1988; Franklin et al., 1993). This highly flexible strategy of being able to utilize a reserve of oxygen carrying capacity only when it is most beneficial should provide interspecific advantages with regards to competition for ecological resources and/or when escaping from predators. However, the metabolic and cardiovascular pros and cons of splenic blood boosting in the bald notothen have not yet been determined, and thus the potential selective advantages of this strategy remain speculative.

The overall aim of the present study was to quantify the magnitude, as well as the cardiorespiratory implications, of splenic blood boosting in bald notothens in response to metabolically demanding situations. Specifically, we hypothesized that bald notothens would substantially increase their blood oxygen carrying capacity via splenic blood boosting to match the metabolic demands associated with enforced exercise and/or ingestion of a meal. Furthermore, the increase in blood oxygen carrying capacity via splenic blood boosting in bald notothens was hypothesized to coincide with metabolic benefits (due to an increased capacity for systemic oxygen delivery) and cardiovascular tradeoffs (due to the physiological effects of an increased blood viscosity on cardiovascular performance). Therefore, we initially measured changes in the haematocrit of uninstrumented individuals following enforced exercise and/or ingestion of a meal. We then determined the cardiorespiratory implications of an elevated blood oxygen carrying capacity following enforced exercise by evaluating the differences between splenectomized individuals (which were unable to release erythrocytes from the spleen into the circulation to boost oxygen carrying capacity) and sham-operated individuals. By exploring the eco-physiological costs and benefits of the splenic blood-boosting strategy, we aimed to provide further insights into how the bald notothens were able to radiate into, as well as dominate, the seasonally highly productive and relatively underutilized cryopelagic zone of the waters surrounding the Antarctic continent (Beers and Jayasundara, 2015; Clarke and Johnston, 1996; Eastman, 2005).

Bald notothens, Pagothenia borchgrevinki (Boulenger 1902), were caught using line and hook through a hole in the sea ice at Evans Wall, Ross Island, McMurdo Sound, Antarctica (n=73, mean±s.d. body mass 87±26 g). Following capture, the fish were transported back to the Crary Lab at McMurdo Station, Antarctica (Fig. 1). Fish were randomly assigned to one of two identical 2000 l tanks containing fresh, aerated seawater at a temperature of −1.6°C and held at a natural photoperiod of 24 h daylight. Fish in one tank were fed (diet of fish flesh, ad libitum every second day), whereas fish in the other tank remained unfed. All fish were allowed >1 week to acclimate prior to experimentation. Animal care and all experimental procedures were performed in accordance with national regulations and were covered by an ethical permit (264-2011) issued by the regional ethical committee on animal research in Gothenburg, Sweden. Sample sizes for the treatment groups described in the following sections were based on our previous experience with similar types of comparative physiological experiments, as well as previously published data on bald notothens (Davison et al., 1988; Franklin et al., 1993).

To investigate the changes in blood oxygen carrying capacity of the bald notothen in response to exercise and feeding, we determined the haemotocrit (%, an indicator for blood oxygen carrying capacity), absolute spleen mass (g) and relative spleen mass (% body mass) of 40 uninstrumented fish (Fig. 2A–D). The fish were randomly divided into four groups, which consisted of (i) ‘unfed resting fish’ (i.e. fasted for 7 days and then sampled in a resting state, n=9, mean±s.d. body mass 55±17 g), (ii) ‘unfed exercised fish’ (i.e. fasted for 7 days and then sampled directly following enforced exercise as described in detail below, n=9, body mass 89±13 g), (iii) ‘fed resting fish’ (i.e. fed and then sampled 24 h later in a resting state, n=13, body mass 87±29 g) and (iv) ‘fed exercised fish’ (i.e. fed and then sampled 24 h later directly following enforced exercise as described in detail below, n=9, body mass 59±10 g). Fed fish were sampled 24 h after feeding, as a previous study indicated that bald notothens are still in a post-prandial state at this stage (e.g. 58±2% of the ingested meal still remained in the stomach of uninstrumented bald nothothens after 24 h; Sandblom et al., 2012).

Unfed resting fish and fed resting fish were captured from their respective holding tanks with a net and sampled immediately (Fig. 2A,B). Sampling consisted of fish being killed by a cranial blow, promptly sampled for blood via a caudal puncture with a heparinized syringe and a 20-gauge needle, and then weighed and measured. The sampling procedure took less than 1 min. In addition, the body cavity of the fish was opened and the spleen carefully dissected out and weighed to calculate relative spleen mass (% body mass). The blood samples were subsequently analysed for haematocrit (%), as the fractional red cell volume upon centrifugation of a subsample of blood in 80 µl microcapillary tubes at 10,000 rcf for 5 min. No food contents remained in the gut of unfed, resting fish, indicating that these fish were in a pre-prandial state, whereas the remaining food contents in the gut of fed, resting fish accounted for 6.6±1.4% of individual body mass (mean±s.e.m.), indicating that these fish were in a post-prandial state.

Unfed exercised fish and fed exercised fish were captured from their respective holding tanks with a net and placed individually into respirometers (Fig. 2C,D). Unfed exercised fish (Fig. 2C) were allowed to recover in the respirometers for 48 h prior to determining whole-animal oxygen uptake (Ṁ_O2) before and after a period of enforced exercise (see ‘Respirometry’, below, for details concerning respirometry techniques used to determine Ṁ_O2). The period of enforced exercise involved manually chasing the fish inside the respirometer for 10 min using plastic tubing inserted through an opening located at the top of the respirometer (Seth et al., 2013). Fish displayed a strong aversive behavioural response to the plastic tubing, which resulted in the fish frantically swimming back and forth inside the respirometers for the majority of the enforced exercise period. All individuals were visibly exhausted after 10 min of enforced exercise, as highlighted by a lack of response to an experimenter tapping the caudal fin of the fish with the plastic tubing (Clark et al., 2013; Seth et al., 2013). Ṁ_O2 measurements obtained from unfed exercised fish before and after the period of enforced exercise were used to estimate resting and maximum metabolic rate (see ‘Respirometry’, below, for metabolic rate calculations), which were subsequently used to evaluate whether the instrumented fish (described in the section below) had sufficiently recovered from the stressors associated with capture, handling and surgical instrumentation prior to experimentation. In contrast, fed exercised fish (Fig. 2D) were immediately subjected to the period of enforced exercise upon entry into the respirometers and thus Ṁ_O2 was only determined after exercise to estimate maximum metabolic rate. This was due to the fact that these fish did not voluntarily feed inside the respirometers and thus no attempt was made to determine resting Ṁ_O2 in a fed state. Following the period of enforced exercise, the fish were removed from the respirometers and immediately sampled in a similar fashion to that described above for the unfed and fed resting fish. No food contents remained in the gut of unfed, resting fish, indicating that these fish were in a pre-prandial state, whereas the remaining food contents in the gut of fed, resting fish accounted for 14.8±1.6% of individual body mass (mean±s.e.m.), indicating that these fish were in a post-prandial state.

To investigate the cardiorespiratory implications of the splenic blood-boosting strategy employed by the bald notothen during metabolically demanding situations, we measured Ṁ_O2, cardiac output, heart rate and ventral aortic blood pressure of ‘splenectomized’ (i.e. splenic vessels were ligated to prevent the exercise-induced increase in hematocrit, n=14, mean±s.d. body mass 106±19 g; Fig. 2E) and ‘sham operated’ (n=19, body mass 101±17 g; Fig. 2F) fish before and after a period of enforced exercise (see ‘Surgical instrumentation’, below, for details concerning surgical instrumentation and methodology for determining cardiorespiratory parameters). From the measured cardiorespiratory parameters, we estimated metabolic rate (see ‘Respirometry’, below), stroke volume (i.e. cardiac output divided by heart rate), total vascular resistance (i.e. ventral aortic blood pressure divided by cardiac output) and cardiac power index (i.e. an estimate of cardiac workload, calculated by multiplying ventral aortic blood pressure by cardiac output).

Fish from both groups were fasted for 7 days before being captured from the holding tank, anaesthetized and instrumented (i.e. fish from both groups were surgically instrumented with custom-made Doppler flow probes and catheters, whereas whether fish had their spleens surgically ligated or underwent a sham operation was randomly determined). After surgery, fish were individually placed into respirometers and allowed to recover for 48 h. All fish were visually observed to be in a calm, resting state following the recovery period. Resting levels of the cardiorespiratory parameters were determined by calculating means for each parameter in undisturbed fish for 15 min directly prior to the fish being subjected to enforced exercise. Once resting levels were determined, a 200 µl sample of blood was carefully withdrawn via the implanted catheter and subsequently analysed for haematocrit. Fish were then subjected to a period of enforced exercise, which involved manually chasing the fish inside the respirometers for 10 min (as above). Maximum cardiorespiratory parameters following enforced exercise were determined by calculating the mean of each parameter during the 5 min period directly after enforced exercise, as this period corresponded with the highest determinations of Ṁ_O2 (i.e. the steepest section of the decline in the partial pressure of oxygen in the water within the respirometers; see ‘Respirometry’, below). Another 200 µl of blood was then withdrawn for haematocrit analysis. At the end of the protocol, fish were removed from the respirometers, killed by a cranial blow, measured and weighed, and the spleen was dissected out and weighed in order to calculate relative spleen mass.

Fish were anaesthetized in seawater (−1.6°C) containing 100 mg l⁻¹ MS222 (ethyl-3-aminobenzoate methanesulphonic acid, Sigma-Aldrich Inc., St Louis, MO, USA) buffered with 200 mg l⁻¹ NaHCO₃. Anaesthetized fish were transferred to an operating table covered with soft, water-soaked foam. To maintain anaesthesia, the gills were continuously flushed with aerated seawater containing 50 mg l⁻¹ MS222 buffered with 100 mg l⁻¹ NaHCO₃ at 0°C.

To determine cardiac output, the ventral aorta was carefully dissected free, ensuring that the pericardium and nearby nerves and blood vessels remained intact (Axelsson et al., 1992). A 20 MHz Doppler flow crystal (Iowa Doppler Products, Iowa City, IA, USA) mounted in 1.3–2.0 mm cuffs (depending on the diameter of the artery) was placed around the vessel. The flow probe was secured with silk sutures to the skin near the area of placement and in front of the dorsal fin. Cardiac output was measured with a directional-pulsed Doppler flowmeter (model 545C-4, Iowa Doppler Products). The cardiac output signal, as well as the other signals described below, were relayed to a PowerLab 8/30 system (ADInstruments, Castle Hill, NSW, Australia) and data were collected on a PC using ADInstruments acquisition software LabChart™ 5 Pro v7.2.5, at a sampling rate of 10 Hz. Heart rate was determined by quantifying the number of blood flow peaks (i.e. cardiac systole) per minute in the acquisition software.

To determine ventral aortic pressure, the third afferent branchial artery was cannulated (Axelsson and Fritsche, 1994). A 4-0 silk suture was placed around the gill arch, and then the artery was punctured upstream from the ligation at the base of the gill filaments. A PE-31 catheter (Natsume, Tokyo, Japan) filled with heparinized saline (25 IU ml⁻¹) was advanced into the vessel toward the ventral aorta. The catheter was secured using the ligation suture, an additional suture placed around the gill arch ∼5 mm below the point of insertion, and a skin suture. The ventral aortic catheter was connected to a pressure transducer (model DPT-6100, Pvb Medizintechnik, Kirchseeon, Germany) and calibrated against a static water column with the water level in the experimental chamber as the zero reference. The blood pressure signal was amplified using a Senselab 4ChAmp amplifier (Somedic Sales, Hörby, Sweden).

For the splenectomized fish, a 2 cm mid-ventral incision was made into the body cavity to access the spleen, after which a silk suture was placed around the splenic vessels and tied firmly in order to prevent blood flow to and from the spleen (Franklin et al., 1993). Blood loss from the mesenteric vessels and from the body wall where the incision was made was minimal, and after a successful ligation, the incision was closed with interrupted silk sutures. For the sham-operated fish, the same surgical procedure was employed with the exception of tying of the splenic vessels. The total time taken for the surgical procedures (i.e. time from when the fish was initially anaesthetized till their placement in the respirometers) was ∼15–30 min.

Fish were individually housed in one of four identical custom-made Perspex respirometers (volume 3.1 l) that were submerged in a larger experimental tank with recirculating aerated seawater at −1.6°C for 48 h prior to determination of Ṁ_O2 to allow sufficient time to recover from preceding stressors. The partial pressure of oxygen in the water within the respirometer was measured continuously using fibre-optic oxygen meters calibrated in accordance with the supplier's manual (Oxy-4 Micro with oxygen Micro-optode PSt1, PreSens Gmbh, Regensburg, Germany). Because of the high demand for the fibre-optic oxygen meters at the Crary Lab in McMurdo Station during the limited research season, resting and maximum metabolic rate of bald notothens in the present study could only be estimated via determination of Ṁ_O2 directly before and after a period of enforced exercise, respectively. However, all fish were visually observed to be in a calm, resting state prior to estimating resting metabolic rate.

Statistical analyses were performed using SPSS Statistics 25 (IBM Corp., Armonk, NY, USA). All data used were assessed to ensure that they did not violate the assumptions of the specific models outlined below. F-, t- and P-values obtained from the statistical analyses are reported throughout the text and P<0.05 was considered statistically significant. Unless otherwise specified, all data are presented as means±s.e.m. All data supporting the paper are available in Tables S1–S5.

For the analyses regarding the haematocrit of uninstrumented fish, we used a two-way ANOVA. The model used feeding state (i.e. unfed versus fed) and exercise state (i.e. resting versus exercised) as fixed factors, as well as the interaction between these factors. If an interaction was detected from a significant interaction term between fixed factors and/or visual inspection of the means plot, we carried out an analysis of the simple main effects. If no interaction between fixed factors was detected, the main effects for the fixed factors produced by the model are reported. For analyses regarding the spleen mass of uninstrumented fish, we used a two-way ANCOVA. The model used feeding state (i.e. unfed versus fed), exercise state (i.e. resting versus exercised) and the interaction between these factors as fixed factors, with body mass as a covariate as it was linearly related to spleen mass. Linear regressions were used to assess the relationship between haematocrit and relative spleen mass in uninstrumented fish.

In order to statistically analyse changes in the metabolic rate of uninstrumented, unfed fish before and after a period of enforced exercise, whilst comparing these rates with those of sham-operated, unfed fish (for control purposes), we used a two-way mixed ANOVA. The model used instrumentation state (i.e. uninstrumented versus sham operated) as a between-subjects factor, exercise state as a within-subjects factor, and the interaction between these factors. To statistically analyse the differences in maximum metabolic rate between uninstrumented unfed and fed fish, we used a one-way ANCOVA (i.e. unfed, exercised versus fed, exercised fish), which used feeding state as a fixed factor and body mass as a covariate (because of the significant differences in body mass between these two groups).

For the analyses regarding the cardiorespiratory variables of instrumented, unfed fish, we used two-way mixed ANOVA. The model used spleen functionality (i.e. sham operated versus splenectomized) as a between-subjects factor, exercise state as a within-subjects factor, and the interaction between these factors. In order to meet the assumptions of the model, we applied a natural logarithmic transformation on cardiac output and cardiac power index data. For the analyses regarding the scope of the absolute increase in the cardiorespiratory variables of instrumented and unfed fish in response to enforced exercise, we used independent t-tests with spleen functionality as the between-subjects factor. Linear regressions were used to assess the relationship between haematocrit and blood pressure, as well as haematocrit and aerobic scope in instrumented fish.

A statistically significant interaction (F_1,36=15.457, P<0.001) was detected between feeding state (i.e. unfed versus fed fish) and exercise state (i.e. resting versus exercised) on the haematocrit of uninstrumented bald notothens. Analysis of simple main effects revealed that both exercise and feeding had a significant effect on the haematocrit of bald notothens (Fig. 3A). In a resting and unfed state, the haematocrit of uninstrumented fish was 8.6±1.6%. Haematocrit was ∼3.1-fold higher in unfed fish subjected to a period of enforced exercise (i.e. unfed resting fish versus unfed exercised fish, F_1,36=75.187, P<0.001; Fig. 3A) and ∼2.4-fold higher in fish that were fed but still in a resting state (i.e. unfed resting fish versus fed resting fish, F_1,36=41.801, P<0.001; Fig. 3A). In addition, haematocrit of fed fish increased further following a period of enforced exercise (i.e. fed resting fish versus fed exercised fish, F_1,36=13.099, P=0.001; Fig. 3A) to reach a level that did not significantly differ from that observed in exercised, unfed fish (i.e. unfed exercised fish versus fed exercised fish, F_1,36=0.367, P=0.549; Fig. 3A).

The interaction between feeding state and exercise state on absolute spleen mass of uninstrumented bald notothens when adjusted for body mass was not significant (F_1,35=1.375, P=0.249). The main effects of the model reveal that absolute spleen mass of fish subjected to a period of enforced exercise was significantly lower than that of fish in a resting state (0.22±0.02 versus 0.37±0.02 g, respectively, for a 74 g fish, F_1,35=24.836, P<0.001). This is also reflected in the statistically significant and strong negative correlations between relative spleen mass and haematocrit for both unfed and fed fish (unfed: F_1,16=19.950, P<0.001; fed: F_1,20=14.187, P=0.001; Fig. 3B,C). No significant differences in absolute spleen mass were observed between unfed and fed fish (0.31±0.02 versus 0.28±0.02 g, respectively, for a 74 g fish, F_1,35=1.399, P=0.245).

To evaluate whether the instrumented fish (see below) had sufficiently recovered from the stressors associated with capture, handling and surgical instrumentation prior to experimentation, we compared haematocrit and metabolic rate of unfed and uninstrumented fish with those from sham-operated fish. A statistically significant interaction (F_1,26=11.362, P=0.002) was detected between instrumentation state (i.e. uninstrumented versus sham operated) and exercise state (i.e. resting versus exercised) on the haematocrit of bald notothens. Analysis of simple main effects revealed that resting haematocrit in uninstrumented fish was significantly lower than that observed in sham-operated fish (8.6±1.6% versus 16.1±1.0%, F_1,26=17.721, P<0.001), while haematocrit levels following enforced exercise did not significantly differ between groups (∼26–27%, F_1,26=0.070, P=0.793). With regard to metabolic rate, the interaction between instrumentation state and exercise state was not significant (F_1,26=1.797, P=0.192). The main effects of the model revealed that although exercise significantly increased metabolic rate in both groups (F_1,26=179.368, P<0.001), no significant differences were observed in either resting or maximum metabolic rate between uninstrumented and sham-operated fish (F_1,26=0.322, P=0.575). In addition, when adjusted for body mass, no significant differences were observed in the maximum metabolic rate of uninstrumented bald notothens with regard to feeding state (i.e. unfed versus fed, F_1,15=1.842, P=0.195).

Statistically significant interactions were detected between spleen functionality (i.e. sham operated versus splenectomized) and exercise state on the haematocrit and metabolic rate of bald notothens (Table 1). Analysis of simple main effects revealed no significant differences between the groups in the resting levels of these parameters (haematocrit: F_1,31=2.856, P=0.101, metabolic rate: F_1,31=0.091, P=0.765; Table 1). As expected, haematocrit of sham-operated fish was substantially higher than that of splenectomized fish following a period of enforced exercise (25.9±1.1% versus 15.3±2.0%, F_1,31=24.901, P<0.001; Table 1), which coincided with a significantly higher maximum metabolic rate (164±7 versus 119±10 mg O₂ kg⁻¹ h⁻¹, F_1,31=24.901, P<0.001; Table 1). Consequently, sham-operated fish had a significantly higher aerobic scope than splenectomized fish (t₃₁=−4.388, P<0.001; Fig. 4). In addition, a statistically significant and strong positive correlation was observed between haematocrit and aerobic scope of instrumented bald notothens (F_1,31=38.517, P<0.001; Fig. 5A).

The interactions between spleen functionality and exercise state were not significant with regard to cardiac output, heart rate, stroke volume and total vascular resistance of instrumented bald notothens (Table 1). The main effects of the models revealed no significant differences between the groups with regard to these cardiovascular parameters (Table 1). As expected, exercise significantly increased cardiac output, heart rate and stroke volume whilst decreasing the total vascular resistance in bald notothens (Table 1, Fig. 4).

Interactions between spleen functionality and exercise state were detected in the ventral aortic blood pressure and cardiac power index of bald notothens (Table 1). Analysis of simple main effects revealed no significant differences in the resting levels of these cardiovascular parameters (ventral aortic blood pressure: F_1,31=0.012, P=0.915, cardiac power index: F_1,27=0.006, P=0.940; Table 1). Yet, both ventral aortic blood pressure and cardiac power index were significantly elevated in sham-operated fish when compared with splenectomized fish following exercise (F_1,31=4.379, P=0.045 and F_1,27=4.712, P=0.039, respectively; Table 1, Fig. 4). In addition, a statistically significant and strong positive correlation was observed between haematocrit and ventral aortic blood pressure in instrumented bald notothens (F_1,31=31.012, P<0.001; Fig. 5B).

Here, we reveal the unprecedented extent to which the bald notothen can elevate their blood oxygen carrying capacity via the splenic blood-boosting strategy when faced with metabolically demanding situations.

In a resting and unfed state, the haematocrit of uninstrumented fish (∼9%) was lower than previously reported resting levels for this species (∼15%) (Davison et al., 1988; Franklin et al., 1993). This is most likely due to the fact that fish in the present study were given a longer period of time to recover from capture and handling prior to experimentation (i.e. >1 week versus 48 h in the other studies) (Davison et al., 1988; Franklin et al., 1993; Wells et al., 1984). When subjected to a period of enforced exercise, the haematocrit of these fish substantially increased from ∼9% to ∼27%. Thus, in response to enforced exercise, the bald notothen are able to increase their blood oxygen carrying capacity by ∼207%, which is higher than previously reported exercise-induced increases of 110–136% for the bald notothen (Davison et al., 1988; Franklin et al., 1993). The exercise-induced elevation in the blood oxygen carrying capacity of bald notothens is also substantially higher than the relative increases reported in other notothenioid species in response to exercise, heat stress and hypoxia (i.e. approximately −5–76%: Davison, 2001; Davison et al., 1994, 1995; Egginton et al., 1991; Forster et al., 1998), as well as the exercise-induced increases typically reported for other vertebrates (i.e. ∼40–60%: Hannon et al., 1985; Hurford et al., 1996; Perrson et al., 1973a,b; Turner and Hodgetts, 1959; Wells and Weber, 1990; Yamamoto et al., 1980; Zapol, 1987).

We also demonstrate that the bald notothen utilizes the splenic blood-boosting strategy to meet the cardiorespiratory demands associated with the ingestion and processing of a meal (Sandblom et al., 2012; Secor, 2009), as fed fish in a resting state had a higher haematocrit than unfed fish (i.e. ∼20% versus ∼9%). Thus, to match the moderate and prolonged periods of energetic demand associated with processing a meal (Sandblom et al., 2012; Secor, 2009), bald notothens are able to increase blood oxygen carrying capacity by ∼131% to reach an intermediate level. Interestingly, in response to enforced exercise, fed fish were still able to further increase their haematocrit to reach a level that was not significantly different to that of unfed, exercised fish (i.e. ∼26–28%). This indicates that bald notothens in a postprandial state (i.e. following food ingestion) still retain enough scope to further increase blood oxygen carrying capacity for other metabolically demanding situations (e.g. when escaping predators) (Axelsson et al., 1992; Buckley et al., 2014; Franklin et al., 1993; Gallaugher et al., 1995).

The rapid exercise-induced increase in the haematocrit of bald notothens in the present study was clearly due to the substantial recruitment of erythrocytes from the spleen (Nilsson et al., 1996), as spleen mass was 41% lower in fish subjected to a period of enforced exercise when compared with that of fish in a resting state (demonstrating that the spleen contracted and ejected erythrocytes into the circulation). Although the substantial splenic release of erythrocytes is mainly responsible for the elevated haematocrit in the bald notothen following exercise, previous studies have shown that factors such as significant decreases in mean cell haemoglobin concentration (an indicator of cell swelling) and slight increases in plasma chloride concentration (an indicator for haemoconcentration as water moves out of the plasma) can also play minor roles (Franklin et al., 1993; Wells et al., 1984). Interestingly, although feeding clearly induced an increase in the haematocrit of bald nothothens, absolute spleen mass in fed fish was not significantly lower than that of unfed fish (i.e. ∼0.28 versus 0.31 g, respectively, for a 74 g fish). Although the underlying reason for this finding currently remains unknown and further research is warranted, it could be speculated that erythrocytes were ejected into the circulation in response to feeding, after which the spleen may have refilled with plasma during the 24 h between feeding and sampling.

Prior to determining the cardiorespiratory pros and cons of splenic blood boosting in the bald notothen, we compared the haematocrit and metabolic rate of uninstrumented fish with those of sham-operated fish. It was observed that resting haematocrit was significantly lower in uninstrumented fish when compared with sham-operated fish, which was most likely due to the stress-induced release of erythrocytes that commonly occurs during anaesthesia and surgery (Wells et al., 1984). However, even though significant differences in resting haematocrit were observed, no significant differences in resting or maximum metabolic rate were observed between the two groups, which indicates that surgical instrumentation did not have a detrimental impact on the aerobic scope of these fish (Clark et al., 2013). Yet, it should be noted that resting metabolic rate (∼71 mg O₂ kg⁻¹ h⁻¹ for a 104 g fish) reported in the present study is higher than previously reported values for similarly sized bald notothens (∼35–50 mg O₂ kg⁻¹ h⁻¹; Forster et al., 1987; Robinson and Davison, 2008). This is most likely related to the fact that the resting metabolic rate for each individual in the present study could only be determined from a single Ṁ_O2 measurement taken directly before the period of enforced exercise (because of our limited access to fibre-optic oxygen meters at the Crary Lab in McMurdo Station during the busy research season) instead of over a longer period of time as recommended in best practice guides (Clark et al., 2013). Despite the abovementioned differences in resting metabolic rate, the maximum metabolic rate induced by our enforced exercise protocol (∼164 mg O₂ kg⁻¹ h⁻¹ for a 104 g fish), as well as absolute aerobic scope (∼92 mg O₂ kg⁻¹ h⁻¹ for a 104 g fish) fell within the range of previously reported values for similarly sized bald notothens exercised in a swim tunnel (∼125–195 mg O₂ kg⁻¹ h⁻¹ and ∼91–146 mg O₂ kg⁻¹ h⁻¹, respectively; Forster et al., 1987; Robinson and Davison, 2008).

The present study is the first to demonstrate the substantially enhanced aerobic scope that splenic blood boosting provides for bald nothothens, as well as the significant cardiovascular trade-offs associated with the strategy.

As expected, the haematocrit in sham-operated fish was substantially higher than that of splenectomized fish following a period of enforced exercise (∼26% versus ∼15%). Consequently, the blood oxygen carrying capacity of sham-operated fish was ∼69% higher than that of splenectomized fish, which coincided with the former being able to increase their metabolic rate to a far greater extent than the latter. As systemic oxygen delivery is the product of cardiac output and blood oxygen content, it seems that the average difference in aerobic scope between sham-operated and splenectomized fish (∼103%) is largely due to the substantial increase in blood oxygen content (via the splenic release of erythrocytes), as no significant differences were observed in cardiac output (Axelsson, 2005; Farrell, 1991). Indeed, in the present study, haematocrit levels of bald notothens following a period of enforced exercise explained 55% of the variation in aerobic scope.

In humans, the performance-enhancing benefits of blood doping are well documented, as increases in blood oxygen carrying capacity of up to 25% can increase maximum oxygen uptake by up to 13% and significantly improve performance times on treadmill tests, cross-county skiing, long-distance running and cycling (Momaya et al., 2015; Tokish et al., 2004). In fish, by experimentally altering the blood oxygen carrying capacity of rainbow trout (Oncorhynchus mykiss), it was demonstrated that swimming performance and maximum oxygen uptake were lowest in anaemic individuals (i.e. haematocrit of 12–17%) and highest in polycythaemic individuals (i.e. haematocrit of 47–48%) (Gallaugher et al., 1995). Based on the collective findings from the present study and those of Franklin et al. (1993), the substantially enhanced aerobic scope of bald notothens following splenic blood boosting clearly provides performance-enhancing benefits during metabolically demanding activities, as splenectomized individuals were unable to complete an exercise regime and fatigued twice as fast as sham-operated individuals as a result of their reduced capacity for systemic oxygen delivery.

However, the abovementioned ecophysiological benefits of the splenic blood-boosting strategy in the bald notothen are also associated with some cardiovascular trade-offs, which may underlie why these animals only utilize this strategy during instances of elevated metabolic demand. On average, ventral aortic blood pressure of sham-operated fish was ∼12% higher than in splenectomized fish in response to enforced exercise. As no significant differences in cardiac output or total vascular resistance were observed between the two groups following enforced exercise, it seems that the increase in ventral aortic blood pressure is largely due to the increased blood viscosity associated with the substantial elevation in circulating erythrocytes (Axelsson, 2005; Farrell, 1991; Graham and Fletcher, 1985). Indeed, in the present study, haematocrit levels of bald notothens following a period of enforced exercise explained 50% of the variation in ventral aortic blood pressure. Consequently, cardiac power index (an estimate of cardiac workload) following enforced exercise was ∼30% higher in sham-operated fish. As cardiac metabolism is directly related to cardiac workload, our findings indicate that there is a considerable energetic cost associated with transporting the highly viscous blood through the cardiovascular system following the splenic release of erythrocytes (Axelsson, 2005; Farrell, 1991; Sidell and Driedzic, 2012). Finally, although not thoroughly investigated in fish, the adverse effects of pumping increasingly viscous blood around the body in humans are well documented (Momaya et al., 2015; Tokish et al., 2004), and thus this may, at least in part, explain why other Antarctic fish species do not utilize this strategy to the same extent as bald notothens.

The present study is the first to demonstrate the substantial metabolic benefits, as well as the cardiovascular trade-offs, associated with the splenic blood-boosting strategy of bald notothens inhabiting the thermally extreme Antarctic marine environment. The advantage of this remarkable strategy is that these fish are able to utilize a reserve of oxygen carrying capacity only when it is most beneficial, which undoubtedly improves a range of aerobic performance traits during metabolically demanding situations whilst minimizing the energetic and physiological costs associated with pumping highly viscous blood around the body during times of reduced oxygen demand. This in turn should incur a selective advantage for this species and may underlie their success in the predator-rich, yet seasonally productive and relatively underutilized cryopelagic zone of Antarctica.

We are sincerely grateful to the staff of McMurdo Station, Antarctica, for their hospitality and much appreciated technical support.