Exploring nature's natural knockouts: in vivo cardiorespiratory performance of Antarctic fishes during acute warming

The notothenioid suborder of fishes that are dominant in the frigid Southern Ocean include the remarkable haemoglobinless (Hb⁻) Channichthyids (icefishes) (O'Brien and Crockett, 2013). The Hb⁻ state renders icefishes unique amongst adult vertebrates (Ruud, 1954) and a suite of compensatory cardiovascular modifications accompanies the corresponding reduction in blood oxygen carrying capacity when compared with closely related red-blooded nototheniids (Axelsson, 2005), such as an enlarged heart (Holeton, 1970; Harrison et al., 1991) and greatly increased baseline vascular conductance (Hemmingsen and Douglas, 1977), which together permit a much greater cardiac output (Q̇) than that of other fishes (Hemmingsen et al., 1972; Hemmingsen, 1991; Axelsson, 2005).

It is generally believed that icefishes can tolerate the Hb⁻ condition in part because of the extremely cold and stable temperature (∼0°C) of the Southern Ocean, which not only increases the solubility of oxygen in seawater and the blood plasma (Ruud, 1954) but also results in a stable and low metabolic rate. Consequently, the fact that the area around the West Antarctic Peninsula, which is where many species of icefishes are found, is among the world's most rapidly warming biomes (Meredith and King, 2005; Clarke et al., 2007) is of considerable concern. In this region, average surface water temperatures have risen by approximately 1°C over the past 50 years, and this rate is expected to continue for the next century (IPCC, 2014), which could create a precarious situation for icefishes.

The lack of haemoglobin (and myoglobin) in icefishes is expected to be particularly problematic as the demand for oxygen transport increases during warming (Ruud, 1954; O'Brien and Crockett, 2013; Buckley et al., 2014; Beers and Jayasundara, 2015), but we do not understand how they are likely to respond. In fact, the only published study on oxygen consumption (Ṁ_O2) of Chaenocephalus aceratus discovered that Ṁ_O2 increased only between 1 and 4°C, but not with additional warming to 10°C (Hemmingsen and Douglas, 1972). Icefishes also show a lower critical thermal maximum (CT_max) than sympatric Hb⁺ nototheniids (Beers and Sidell, 2011). While these observations suggest that warming could create problems with internal oxygen transport, CT_max was insensitive to hyperoxia (Devor et al., 2016). This, together with the previous conflicting measurements of Q̇, raises the question about the flexibility of cardiovascular function in icefishes.

For example, previous indirect estimates of routine Q̇ in C. aceratus (using the Fick principle) vary almost 3-fold from 60 ml min⁻¹ kg⁻¹ (Holeton, 1970) to 100–150 ml min⁻¹ kg⁻¹ (Hemmingsen et al., 1972). Furthermore, this range for routine Q̇ estimates could be even wider given the direct measurement of blood flow in the two cranial pairs of gill arches (10–15 ml min⁻¹ kg⁻¹), which equates to routine Q̇ being 20–30 ml min⁻¹ kg⁻¹ by assuming equal blood flow in all branchial arteries (Hemmingsen et al., 1972). Despite this rather large range, the paradigm of a high routine Q̇ in icefishes (80–100 ml min⁻¹ kg⁻¹) relative to Hb⁺ species still persists in the literature (Hemmingsen, 1991; Tota et al., 1991; Zummo et al., 1995; Acierno, et al., 1997; Feller and Gerday, 1997; Sidell and O'Brien, 2006; Axelsson, 2005; Garofalo et al., 2009). Certainly, maximum Q̇ of icefishes determined in vitro with perfused heart preparations can be high (maximum Q̇ values between 100 and 300 ml min⁻¹ kg⁻¹ in Chionodraco hamatus at 3°C; Acierno et al., 1997; Tota et al., 1991). Nonetheless, our own recent work in vitro with perfused heart preparations found a lower maximum Q̇ in both C. aceratus (70 ml min⁻¹ kg⁻¹) and Chionodraco rastrospinosus at 1°C (126 ml min⁻¹ kg⁻¹) (S. E. Egginton, M. Axelsson, E. L. Crockett, K. M. O'Brien and A. P. Farrell, unpublished). Thus, if routine Q̇ for icefishes is close to maximum Q̇, an obvious concern is that regulatory capacity is necessarily limited and there can be little cardiovascular flexibility during environmental challenges.

The aim of the present study was to test the hypothesis that in vivo cardiovascular function in C. aceratus lacks regulatory flexibility during acute warming. Comparisons could then be made with cardiorespiratory regulatory capacity of Hb⁺ nototheniids (e.g. Axelsson et al., 1992, 1994; Franklin et al., 2007; Campbell et al., 2009; Egginton and Campbell, 2016).

Previous studies have concluded that when Antarctic fishes are acutely warmed, they primarily adjust their heart rate (f_H) to regulate Q̇ (Axelsson et al., 1992; Campbell et al., 2009), much like other fishes (Farrell, 2009; Farrell and Smith, 2017). Also, while some studies report a high routine vagal tone in Antarctic notothenioids (Axelsson et al., 1992; Egginton and Campbell, 2016), Hemmingsen et al. (1972) reported only a small increase in f_H after atropine injection (from 16 to 18 beats min⁻¹) that could indicate stress. Therefore, we tested the possibility that earlier in vivo cardiovascular studies were affected by stress, which would have an impact on other measured variables such as Q̇ and cardiac power output. We used a minimally invasive method [electrocardiogram (ECG) measurements during an acute thermal ramp with and without autonomic vagal blockade with atropine] to provide insight into routine f_H, the level of vagal tone and the contribution of f_H in setting upper thermal tolerance (Farrell, 2009, 2016). The results from these minimally instrumented animals were used as a baseline for the more heavily instrumented animals that were part of the first comprehensive assessment of in vivo cardiorespiratory response of C. aceratus during activity and acute (sub-critical) warming. By simultaneously measuring a number of key cardiorespiratory variables, we could directly assess the flexibility and interplay of autonomic regulation and oxygen transport and address whether the Hb⁻ condition reduces cardiac scope and thus the ability of C. aceratus to tolerate temperature changes.

Given our concerns about physiological measurements being made on overly stressed animals, the ECG measurements on blackfin icefish, Chaenocephalus aceratus (Lönnberg 1906) (N=12, body mass M_b=1.55±0.21 kg, mean±s.d.), were also performed on the closely related Hb⁺ black rockcod, Notothenia coriiceps Richardson (N=13, M_b=1.24±0.32 kg), to allow comparison with previously published data on other Hb⁺ Antarctic fishes. An additional group of C. aceratus were also instrumented for ECG measurements and used to ascertain autonomic tone (N=4, M_b=1587.5±387.0 kg). Finally, a separate group of C. aceratus (N=16, M_b=0.85±0.21 kg) were used for the more invasive cardiorespiratory study. Adult fish were captured at 100–200 m depth using an otter trawl and baited benthic pots (for N. coriiceps) deployed from the ARSV Laurence M. Gould in Dallmann Bay (64°10′S, 62°35′W) and off the south-western shore of Low Island (63°24′S, 62°10′W). Fish were held on the ship in circulating, aerated seawater tanks for up to 2 days before being transferred to Palmer Station, Antarctica, where they were held at ambient seawater temperature (0±1°C) for at least 72 h and up to 3 weeks before experiments commenced. Notothenia coriiceps were fed blocks of fish muscle fillet every other day, but not 24–48 h prior to surgery, whereas C. aceratus did not feed in captivity. All experiments involving live animals were approved by the University of Alaska, Fairbanks Institutional Animal Use and Care Committee (570217-9).

Instantaneous ECGs were recorded with minimal surgical intervention using laterally staggered bipolar electrodes (7-strand Teflon­-coated wire, 0.2 mm diameter; A-M Systems, Sequim, WA, USA). Fish were first anaesthetized in ice-chilled seawater containing MS-222 (100 mg l⁻¹) until unresponsive to touch but retaining opercular movements, then weighed and maintained in an anaesthetized state (gills irrigated with aerated MS-222, 50 mg l⁻¹) while electrodes were inserted through the ventral muscle overlying the pericardial cavity, taking care not to pierce the pericardial membrane (Campbell et al., 2004). To assess the influence of autonomic tone on thermal sensitivity, a catheter (PP10) was inserted through the skin at the base of the pectoral fins, fed into the pericardial region and externally secured to ECG wires. All surgery was performed in a cold room at ∼5°C, and completed within ∼10 min, with forced ventilation in a holding tank used to minimize the delay before regular ventilation was re-established. Fish recovered for 48–72 h in cradles covered with plastic netting suspended in darkened flow-through tanks until f_H was stable and consistently low. The cradles allowed space for fish to move without them turning around, and gave ready access to electrodes/cannulae in an unobtrusive manner.

CT_max was measured on individual fish (both C. aceratus and N. coriiceps) held in 70 l insulated experimental tanks and exposed to an acute thermal ramping treatment (3.3°C h⁻¹; Beers and Sidell, 2011) by recirculating water through an in-line heater (AquaLogic, San Diego, CA, USA). This warming rate was chosen to ensure that core body temperature had sufficient time to equilibrate with environmental temperature, rather than slower ramping to avoid the opportunity for acclimation (Beers and Sidell, 2011). CT_max is usually defined as the temperature when righting ability is lost, i.e. loss of neuromuscular coordination, but in this study we adopted a more cardiocentric endpoint of prolonged periods of ventricular asystole (defined as a period lasting at least 10–20 cardiac cycles based on the f_H immediately prior to asystole). Cardiac asystole was chosen as an index of CT_max because it is probably most comparable with behavioural indices (loss of righting reflex) when brain blood flow becomes compromised. Following the appearance of sustained asystole, animals were stunned and the spinal cord transected; blood-free atrial and ventricular mass were then measured. The responses of intact fish were compared with those of fish following vagal blockade (atropine, 1 mg kg⁻¹).

Separately, an index of sympathovagal balance was calculated (Egginton and Campbell, 2016) using f_H measurements after vagal blockade (atropine, 1 mg kg⁻¹) and then after complete autonomic blockade (1 mg kg⁻¹ atropine+1 mg kg⁻¹ propranolol) in four C. aceratus at ambient temperature (∼0°C). The acute effects of atropine (tachycardia, increased ECG amplitude) and propranolol (bradycardia, decreased ECG amplitude) were evident within 3–4 min of injection, and effects were sustained for 12–24 h (i.e. substantially outlasting the experiments). Autonomic tone (atropine+propranolol) has previously been measured using this method in N. coriiceps (Egginton and Campbell, 2016), so instead of repeating the experiment, we provide relevant comparisons with previously published data (Table 1).

ECG signals were collected using a PowerLab (AD Instruments, Oxford, UK) with an animal bio-amplifier (1 kHz sampling, 5 Hz high-pass filter), and analysed using LabChart (version 7, AD Instruments). Fish were sensitive to external noise and so routine f_H (based on 10–20 consecutive heartbeats) was recorded at 06:00 h (local time) when fish were quiescent and external disturbance was minimal.

Heart rate often changes in a curvilinear manner with temperature, hence the use of inter-beat interval (R–R), which provides a linear relationship (Vornanen et al., 2014). Heart rate variability (HRV) analysis reflects the influence of the autonomic nervous system on the sinus node, when the heart is in regular sinus rhythm. Tachograms and period histograms were used to ensure data blocks were sufficiently stable. In addition to f_H (expressed as R–R interval), HRV indices were calculated in both the time (standard deviation of inter-beat intervals, SDNN; root mean square of successive differences, RMSSD) and frequency domains [power spectral analysis (PSA) using Fourier analysis; low frequency:high frequency (LF:HF) ratio of power distribution around the midpoint frequency]. At rest, up to 512 continuous heartbeats were sampled, but during the ramp test, this number was reduced to 256 or 128 heartbeats, as appropriate to avoid shifting baselines; spectral leakage was minimized by use of a Hanning window. A Poincaré plot of R–R intervals as a function of the subsequent R–R interval was used as an index of long-term (diagonal axis from the origin) and beat-to-beat variations (normal to the axis maxima) and provides a graphic parallel to SDNN and RMSSD indices, respectively (Campbell and Egginton, 2007).

Chaenocephalus aceratus were anaesthetized in ice-chilled seawater containing MS-222 (140 mg l⁻¹) until unresponsive to touch but retaining opercular movements, and maintained in an anaesthetized state (gills irrigated with aerated MS-222, 70 mg l⁻¹) on the surgery table (on a foam pad covered with goatskin chamois to prevent desiccation and protect the scaleless skin). A ventral mid-line incision allowed placement of a 2.5 or 4 mm transit-time flow probe (model T420, Transonic Systems Inc., Ithaca, NY, USA) around the pericardium and ventral aorta immediately distal to the bulbus arteriosus to measure ventral aortic blood flow (13 fish). The fish were then placed on their right side to access the branchial arteries in the left third gill arch. The efferent branchial artery was occlusively cannulated using PE-50 tubing with a 2F polyurethane tip to measure dorsal aortic pressure (P_da) in 16 fish. The afferent branchial artery in the same gill arch was cannulated in seven fish to measure ventral aortic pressure (P_va) using a PE-50 cannula, which in smaller animals was tipped with heat-pulled PE-30. The left duct of Cuvier was non-occlusively cannulated in 5 fish to measure central venous pressure (P_cv) using PE-50 tubing tipped with 3F silicon (to prevent damage to the delicate venous vasculature), equipped with side-holes to maintain patency, which was advanced into the sinus venosus. The cannulae and electrical leads from the flow probe were sutured to the skin and the ventral incision was closed with 3-0 surgical silk. Surgery was performed in a cold room at ∼2–3°C, and completed within 45–120 min, after which the gills were irrigated with fresh seawater until spontaneous ventilation recommenced.

After surgery, the fish were placed in custom-made 12.4 l PET (polyethylene terephthalate) respirometers (https://skfb.ly/RMsV), triangular in cross-section to accommodate the protruding pelvic fins that characterize C. aceratus. The 700 l tank contained two identical respirometers that were covered in black plastic. Icefish were permitted a minimum 40 h post-surgical recovery before experiments started. Submersible pumps (Eheim, Deizisau, Germany) recirculated water in the respirometers to attain complete mixing. Oxygen concentration (measured as percentage air saturation) was measured with a P_O2 optode (FireSting, Aachen, Germany) placed in-line with the mixing pump. Ṁ_O2 was recorded as O₂ concentration decrease over a 10 min period, before a second Eheim pump started to flush the respirometers with fresh seawater for 15 min, thereby creating a 25 min cardiorespiratory measurement period. This regime ensured O₂ concentration never fell below 90% air saturation.

The flow probe was connected to a Transonic flow meter (T402; Transonic Systems, Inc.). The cannulae were attached via pinports (Instech Laboratories, Plymouth Meeting, PA USA) to pressure transducers (Medizintechnik, Kirchseeon, Germany), which were regularly calibrated against a static water column. The output signals from the pressure transducers were pre-amplified by a Senselab 4CHAMP amplifier (Somedic sales, Hörby, Sweden). The flow meter and 4CHAMP amplifier were connected to a PowerLab data acquisition system (ADInstruments, Castel Hill, NSW, Australia), which provided input to, as well as data storage on, a computer running LabChart Pro (version 7; ADInstruments, Bella Vista, NSW, Australia).

The cardiorespiratory response to acute warming was assessed using in-line heating (AquaLogic) from ambient temperature (0.8±0.1°C) to 4°C over approximately 3 h and then to 8°C over a further 3 h. At all three temperatures (maintained for at least 30 min each), routine cardiorespiratory variables were measured before inducing activity by agitating the respirometers for 1 min. For routine cardiorespiratory measurements, we ensured that the fish were not disturbed by external stimuli. Sustained activity is difficult to elicit in icefish (e.g. Egginton, 1997), but this method typically produced sustained tail flicking for several minutes without apparently exhausting the fish. Therefore, it is unlikely that this agitation method induced a maximum cardiorespiratory response.

Differences between C. aceratus and N. coriiceps were determined using a multi-factorial ANOVA, with Fisher's protected least-significant difference post hoc test to discriminate amongst group means (StatView 5.0, SAS Institute Inc., Cary, NC, USA). Broken-line regression analyses of f_H versus temperature were conducted for each fish using the segmented package in R (v.3.1.0; http://www.R-project.org/) to determine the cardiac breakpoint temperature (BPT), i.e. when the increase in f_H with temperature first attenuated. The regression line fitted to this phase of increasing f_H was tested for differences in slope and intercept values using an ANOVA.

Three C. aceratus were excluded from subsequent analysis because routine f_H at ambient temperature was appreciably higher than in other fish (P<0.001), suggesting stress, despite a similar peak f_H in both groups (P=0.16). BPT was significantly lower in the outlier group (P<0.05). Even so, the outlier group showed no overt signs of stress (e.g. banding of integument pigmentation or prolonged recovery from surgery) and therefore probably would have been considered unstressed without cardiovascular (ECG) data, potentially leading to erroneous conclusions (Fig. S1). In addition, one N. coriiceps was excluded because of apparent restlessness and periodic bouts of bradycardia throughout acute warming (note: BPT and HRV values also deviated from those of the remaining fish).

An unpaired t-test was used to investigate significant differences in routine f_H between C. aceratus instrumented with ECG electrodes and those with more extensive surgery for the cardiorespiratory measurements. The effect of temperature and activity on cardiorespiratory measurements in C. aceratus was investigated using a two-way repeated-measures analysis of variance (ANOVA), revealing the combined and isolated effects of warming and activity on each cardiorespiratory variable. We assumed acute warming to 8°C would impose an intolerable stress, which proved incorrect as revealed by subsequent fish responses, and so the first two fish tested were not agitated and therefore not included in the ANOVA model (to avoid Type II errors by using matched data for all measurement points). Nevertheless, these two fish are represented in the mean values in figures where data were obtained. The relative importance of f_H and V_S in meeting the increased oxygen demands following activity was elucidated at each temperature with a linear regression to correlate individual fish changes in Q̇, f_H and V_S with respect to changes in Ṁ_O2. Statistical analysis was performed with GraphPad Prism (v.7.0d). Statistical significance was assigned to α≤0.05 and data are presented as means±s.e.m.

Atropine methyl nitrate, propranolol hydrochloride and tricaine methanesulfonate (MS-222) were purchased from Sigma-Aldrich (St Louis, MO, USA).

Routine f_H was lower in C. aceratus than in N. coriiceps (P<0.05), an absolute difference that widened at peak f_H (P<0.001; Table 1). Consequently, the rise in f_H was slower and of a lower magnitude during a thermal ramp for C. aceratus (0 versus 10°C Q₁₀=3.30, Δf_H=21.2 beats min⁻¹) than for N. coriiceps (Q₁₀=2.74, Δf_H=27.3 beats min⁻¹). Periodic missed and ectopic beats were noted in C. aceratus around 8–10°C, while ventricular asystolic events (with P-wave presence) became more frequent and prolonged closer to BPT, suggesting a atrioventricular conduction failure (Fig. S2).

No significant difference in BPT was found between C. aceratus and N. coriiceps (Fig. 1) despite the large significant difference in peak f_H (38.2±4.2 and 54.8±2.2 beats min⁻¹, respectively; Table 1). Some C. aceratus individuals had a brief tachycardia after BPT, corresponding to the first observed change in fish behaviour. In C. aceratus, catastrophic cardiac failure rapidly ensued (Fig. 1), with prolonged ventricular asystolic periods (up to 8–10 s in duration) with a regular P-wave, interspersed with bursts of ventricular tachyarrhythmias of >40 beats min⁻¹, a situation that points to conduction failure (Fig. S2). In contrast, N. coriiceps displayed missed or ectopic beats 1–2°C above ambient temperature, with a transient increase in HRV around the mid-point of the ramp (5–10°C). Even so, regular sinus rhythm was re-established and maintained at a higher temperature than in C. aceratus. Again, heat-induced bradycardia was evident following the BPT, with cardiac failure preceded by progressively extreme periodicity (Figs S2 and S3).

Atropine nearly doubled routine f_H at 0°C in both species, and increased peak f_H by 25% (P<0.05) and 5% (not significant, n.s.) in C. aceratus and N. coriiceps, respectively, but decreased both Δf_H/ΔT (^oC) (Table 1) and Q₁₀ (2.42 and 2.03, respectively) (Table 2). Consequently, the difference in peak f_H between C. aceratus and N. coriiceps was reduced from ∼30% to ∼20% after atropine (P<0.05), but peak f_H was reached at similar temperatures (n.s.). In addition, atropine substantially increased the temperature at which the first change in the ECG occurred in both species, making it closer to BPT; while ventricular asystolic periods progressively increased in duration, no tachyarrhythmia was observed. Vagal blockade with atropine also reduced the difference in both BPT (Fig. 1) and, especially, CT_max (Table 1) between the two species.

The time- and frequency-domain indices of HRV showed post-surgery depression, were transiently elevated as anaesthesia wore off, and quickly stabilized within 12 h when HRV was proportional to R–R interval (Table S1). Although C. aceratus recovered from surgery more slowly than N. coriiceps (data not shown), interspecific differences in HRV were modest (Table 2; Table S1). Both short-range (RMSSD) and long-range (SDNN) indices of HRV generally declined with rising temperature to ∼10°C in C. aceratus and to ∼12°C in N. coriiceps. Hence, the Poincaré data cloud progressively narrowed with acute warming but widened as CT_max was approached (rapid collapse of C. aceratus after peak f_H precluded analysis at that point; Fig. S3). While total spectral power was heterogeneous among individuals, LF:HF ratio (a dynamic index of sympatho-vagal balance) progressively declined in both species to a temperature of 5–10°C, but increased markedly at higher temperatures and especially at CT_max, indicating an autonomic conflict consistent with signs of ECG disruption (Table 2; Table S1). Only two C. aceratus reached 15°C with sinus rhythm intact, at which point all HRV variables (e.g. SDDNN, RMSSD, spectral power) were markedly different from those at 5 or 10°C (Table S1). Such changes became accentuated between BPT and CT_max.

Vagal blockade not only eliminated differences in BPT and CT_max between C. aceratus and N. coriiceps but also resulted in extremely low spectral power during the thermal ramp for both species, without changing the species difference in peak f_H (Table 1). All HRV indices were dramatically reduced by atropine, until just before BPT. The mid-range elevation of LF:HF by 10°C in atropinized C. aceratus is consistent with greater thermal sensitivity (probably sympathetic in origin), whereas LF:HF was minimal at this point in N. coriiceps (Table 2).

Routine Q̇ in C. aceratus at 0.8°C was 26.6±2.1 ml min⁻¹ kg⁻¹ at a f_H of 11.5±0.9 min⁻¹. Routine f_H measured in C. aceratus equipped with only ECG electrodes for the thermal tolerance test was not significantly different (P=0.21; 9.0±1.1 min⁻¹; Table 1), suggesting that fish had recovered sufficiently from the more extensive surgical interventions to restore the high vagal tone characterizing unstressed fish.

Routine Ṁ_O2 (24.5±0.8 mg O₂ kg⁻¹ h⁻¹), as well as P_va (2.06±0.05 kPa) and P_da (1.81±0.06 kPa), were similar to previously reported values (Holeton, 1970; Hemmingsen and Douglas, 1972; Hemmingsen et al., 1972). P_cv has not previously been measured in C. aceratus, but the near-ambient central venous pressure (0.06±0.03 kPa) was consistent with very low caudal venous pressures reported previously (Hemmingsen and Douglas, 1972).

Acute warming of C. aceratus from 0.8°C to 8°C increased routine Ṁ_O2 by 2.4-fold to 59.5±2.4 mg O₂ kg⁻¹ h⁻¹ (Fig. 2A; P<0.001) and Q̇ by 2.6-fold to 69.9±4.7 ml min⁻¹ kg⁻¹ (Fig. 2B; P<0.001). The increase in Q̇ with temperature was entirely attributable to an increase in f_H to 29.1 beats min⁻¹ (P<0.001), which compensated for a significant but small decrease in V_S (P=0.02), which was most pronounced between 0 and 4°C (Fig. 2C). Importantly, f_H similarly increased from 9.0±1.1 at 0.6°C to 28.0±3.0 beats min⁻¹ at 8°C in ECG-instrumented C. aceratus (cf. values for N. coriiceps: 13.5±0.6 and 38.2±1.7 beats min⁻¹, respectively; Table 1), again suggesting that these fish had recovered sufficiently from extensive surgical intervention.

Despite the large increase in Q̇ in C. aceratus during warming, both P_da and P_va decreased with increasing temperature (Fig. 3; P=0.001 and P=0.002, respectively) because both G_sys (P<0.001) and G_branch (P=0.001) increased significantly with temperature and prominently at 8°C. P_cv tended to decrease with temperature (P=0.06), while CPO increased significantly (Fig. 3; P=0.02).

At ambient temperature, enforced activity was associated with significant (P<0.001) increases in Ṁ_O2, Q̇, f_H and V_S in C. aceratus (Fig. 2; Fig. S4). Although P_da decreased during activity (P=0.02), as with warming, P_va increased with activity (P<0.001) because G_sys (P<0.001) increased with activity but G_branch did not (P=0.68). There were no significant changes in P_cv from rest to activity, even though CPO increased (Fig. 3; P=0.04).

Individual changes in Q̇ and Ṁ_O2 with temperature showed a strong positive correlation as C. aceratus transitioned from rest to activity at all test temperatures (Fig. 4). At ambient temperature, a strong relationship (R²=0.87) existed between Δf_H and ΔṀ_O2, without a significant correlation between ΔV_S and ΔṀ_O2. At 4°C, however, the relationship between Δf_H and ΔṀ_O2 weakened and a significant relationship between ΔV_S and Ṁ_O2 developed. By 8°C, a strong correlation between ΔV_S and ΔṀ_O2 (R²=0.73) emerged while the association between Δf_H and ΔṀ_O2 had disappeared, reflecting the necessarily diminished potential change in f_H as peak heart rate was approached (Fig. 4).

In quiescent, undisturbed C. aceratus at 0.8°C, Q̇ was 26.6 ml min⁻¹ kg⁻¹. This value accords well with the partial measurements of flow inferred from Hemmingsen et al. (1972), but is much lower than the Fick estimate (100–150 ml min⁻¹ kg⁻¹; Holeton, 1970; Hemmingsen et al., 1972). This difference could be a consequence of well-vascularized icefish skin (Jakubowski, 1982), which provides an alternative pathway for oxygen uptake for which the Fick principle does not account (Farrell et al., 2014). Even though Hemmingsen et al. (1972) recognized this potential error, the considerably greater Fick-derived estimates were accepted without addressing the discrepancy with their direct flow measurements, a surprising conclusion considering that the same authors previously suggested that cutaneous oxygen uptake could constitute up to 40% of total Ṁ_O2 (Hemmingsen and Douglas, 1970). Similarly high (35%) cutaneous oxygen uptake has been independently measured in another scaleless Antarctic fish (Rhigophila dearborni; Wells, 1986). However, Holeton (1976) disputed such a significant role for a cutaneous oxygen supply in C. aceratus by arguing that branchial gas exchange must dominate because of a greater P_O2 diffusion gradient (i.e. between water and venous rather than arterial blood), a shorter blood-to-water distance, and gills receiving the entire Q̇ versus only a small portion to the skin. Thus, alternative factors, such as post-surgical stress, may explain the difference between current and earlier measurements of routine Q̇.

Cardiac rhythmicity, which was linked with ventilatory activity, was evident in both C. aceratus and N. coriiceps, thus demonstrating good surgical recovery (see Campbell et al., 2004). The resting f_H (16–18 beats min⁻¹) reported by Hemmingsen et al. (1972) was close to the maximum f_H we recorded during activity at a similar temperature, again indicating that post-surgical stress could explain the extremely high Fick-derived estimates of routine Q̇ for C. aceratus at ∼1°C (Hemmingsen et al., 1972). However, this concern would equally apply to the fish instrumented with electromagnetic flow probes in the same study and that yielded a lower routine Q̇. Our assumption of equal blood flow in all branchial arteries may therefore not be correct, underestimating the inferred Q̇, such that correspondence between our data and the estimated blood flows could be coincidental. Post-surgical stress was less likely to be an issue in Holeton's (1970) study, which reported the lowest Ṁ_O2 measured in the period 2–10 days post-surgery and a relatively low routine f_H (13.8 beats min⁻¹), while Q̇ was calculated as 61 ml min⁻¹ kg⁻¹ using the Fick principle. This f_H was lower than that of Hemmingsen et al. (1972) but, rather than being close to our routine value (11.5 beats min⁻¹), it is closer to our value immediately following the 10 min of enforced activity (14.7 beats min⁻¹). We therefore conclude that a combination of cutaneous O₂ supply skewing the Fick calculation, post-surgical stress, and perhaps other methodological differences account for the previous overestimations of routine Q̇ in C. aceratus. This is an important conclusion, as we now have clear evidence that the low routine Q̇ provides a large scope for change and that the cardiorespiratory regulation may permit physiological flexibility during acute temperature challenges, something that appeared unlikely based on earlier measurements of routine Q̇.

In vivo ECGs have not been reported previously for C. aceratus, and the similarity of the ECGs between C. aceratus and N. coriiceps allowed our detailed assessment of disturbances to cardiac conduction, which may be central to the failure of the heartbeat at high temperature in other fish species (Vornanen et al., 2014). Even though CT_max may vary with the extent of facilitated O₂ transport (Beers and Sidell, 2011), the current observations suggest that for acute warming this effect is modest, which is consistent with recent work showing that hyperoxia failed to increase the upper thermal tolerance of icefish (Devor et al., 2016).

Reduced HRV is often used as an index of diminished cardiovascular regulation in mammals, although changes in blood pressure may also induce changes in short-term HRV as a result of baroreflex control. Acute warming of both C. aceratus and N. coriiceps dramatically reduced HRV at around 10°C, before a marked elevation prior to BPT and CT_max, at which point ventricular asystole occurred, probably as a result of conduction failure. In combination, these data suggest a biphasic response to thermal stress. Using the LF:HF ratio as a simplified readout of sympathovagal balance, the pharmacological blockade shows it to be a reasonable index of autonomic regulation in these species, as seen in other comparative studies (Campbell et al., 2004). Our data demonstrate that C. aceratus uses vagal tone to suppress thermally induced tachycardia more so than N. coriiceps, but similar to other Hb⁺ nototheniids (Axelsson et al., 1992; Franklin et al., 2001). While this vagal reflex during acute warming could minimize cardiac Ṁ_O2, acidosis or oxidative stress, it is likely to incur a cost of reducing the limits to thermal tolerance (see Sandersfeld et al., 2015). Thus, acute cardiac stress may have chronic downstream effects that have not been explored to date. Poincaré plots represent a more global view of HRV, and at this level there appears to be little difference between species, highlighting that additional influences on cardiac performance need to be considered (see below). For example, C. aceratus is relatively unresponsive to adrenaline, consistent with a low synthesis rate, and attenuated stress-induced release of catecholamines (Whiteley and Egginton, 1999), thereby obviating a humoral arm of cardiovascular control that could be available to other notothenioids, albeit to a lesser extent than in temperate species.

Regardless, it is clear that C. aceratus acutely stressed to temperatures well beyond current environmental conditions can mount a robust cardiovascular response, one that is similar to that of Hb⁺ nototheniids. An important difference is that while N. coriiceps retains a modest scope for f_H changes at the BPT (26%), C. aceratus retains a significantly higher scope for f_H change (81%), leading to a lower peak f_H. Thus, the cardio-protective strategy of imposing an autonomic brake on f_H in C. aceratus appears to be more robust, failing only in extremis. Indeed, indices of HRV (e.g. RMSSD, an index of the integrity of vagus-mediated autonomic control) are consistent with this suggestion. The HRV response to elevated temperature was similar among individual fish and between species, e.g. RMSSD tracked the change in SDNN in both N. coriiceps and C. aceratus, consistent with maintaining tight regulation of cardiac rhythmicity even in the Hb⁻ fish.

As noted above, icefish are clearly capable of developing a robust cardiorespiratory response to acute warming. The initially high Q₁₀ probably reflects thermal history, as speciation in the extremely cold and stable aquatic environment is assumed to allow specializations that may blunt physiological plasticity (Hochachka and Somero, 2002). However, N. coriiceps displays greater vagal withdrawal in response to a temperature ramp, possibly because extreme sensitivity to metabolic acidosis accompanying the lack of proton buffering by Hb would otherwise drive a greater vagal suppression of thermally induced tachycardia in icefish. The reduction in Q₁₀ for f_H with atropine suggests a similarity in fundamental thermal sensitivity that is overridden by vagal tone during a temperature increase, revealing a similarity of intrinsic f_H seen among notothenioids of varying ecotype (Campbell et al., 2009).

In C. aceratus, both resting and active Ṁ_O2 increased almost linearly between 0.8 and 8°C, whereas Hemmingsen and Douglas (1972) reported that Ṁ_O2 only increased from 1 to 4°C and remained unchanged up to 10°C, consistent with stressed animals reaching maximum Ṁ_O2 earlier in a temperature ramp. However, we used fish that were rapidly warmed (i.e. over several hours) compared with the earlier study where fish were incrementally warmed by 0.3–1°C per day and maintained at a given temperature for at least 48 h prior to Ṁ_O2 measurements. Thus, fish may have become less stressed by the slower warming speed, which permitted an overall reduction in oxygen demand, as seen in some temperate species (Gräns et al., 2014). However, it is unlikely that the fish were acclimating to a new temperature over a few days, as N. coriiceps does not exhibit reduced Ṁ_O2 following acclimation (6–9.5 weeks) from 0 to 5°C (Egginton and Campbell, 2016; Joyce et al., 2018).

At 0.8°C, Q̇ reached 44.7 ml min⁻¹ kg⁻¹ during enforced activity, which is about 60% of the maximum value for Q̇ in the in situ perfused heart preparation (70 ml min⁻¹ kg⁻¹; S. E. Egginton, M. Axelsson, E. L. Crockett, K. M. O'Brien and A. P. Farrell, unpublished), an expected difference considering the relatively moderate activity imposed probably did not exploit maximal cardiac capacity. Also, we measured mean Q̇ concomitantly with Ṁ_O2 over a period of several minutes, while peak Q̇ was transitory and higher than the mean value (Fig. S4; peak Q̇ immediately following agitation was 52.9±4.5 ml min⁻¹ kg⁻¹). Importantly, peak f_H (17.6±0.7 min⁻¹) following agitation was identical to that measured in in situ perfused hearts at the same ambient temperature (S. E. Egginton, M. Axelsson, E. L. Crockett, K. M. O'Brien and A. P. Farrell, unpublished), an indication of a complete loss of vagal tone with just moderate activity.

The implication of these data is that C. aceratus at 0.8°C has a considerable cardiac scope and that the heart is not pumping near its maximum capacity when the fish is in a quiescent, unstressed state. While an abnormally high cardiac preload may be used with perfused hearts to generate maximum Q̇, the maximum Q̇ measured in vivo at high temperatures (86.3 ml min⁻¹ kg⁻¹) was close to the maximum Q̇ (98 ml min⁻¹ kg⁻¹) reported for perfused hearts at 4°C. Thus, we confirmed that the icefish heart is not only capable of pumping large volumes of blood at a near-freezing temperature but also retains a cardiac scope by maintaining vagal tone to increase f_H. Furthermore, the Hb⁻ C. aceratus can increase Q̇ with activity even at elevated temperature in vivo by elevating f_H. In fact, despite their similar routine Q̇, maximum Q̇ in C. aceratus clearly exceeded that of the more active Hb⁺ nototheniid Pagothenia borchgrevinki (∼60 ml min⁻¹ kg⁻¹; Franklin et al., 2007).

Whether fish primarily regulate V_S or f_H to change Q̇ during activity remains a subject of debate (Farrell, 1991; Altimiras and Larsen, 2000; Sandblom et al., 2005; Shiels and White, 2008; Farrell and Smith, 2017; Nelson et al., 2017). The present study informs this discussion because C. aceratus increased Q̇ during activity by elevating V_S and f_H to different degrees across temperatures. While ΔṀ_O2 was closely correlated with Δf_H, as described previously in N. coriiceps (Campbell et al., 2008), this occurred only at 0.8°C. In fact, V_S increased by approximately 1 ml kg⁻¹ irrespective of the increase in Ṁ_O2 across temperature, and became a primary determinant of the magnitude of ΔṀ_O2 with activity at 8°C. This suggests that at ambient temperature, f_H is finely tuned to preserve adequate cardiovascular oxygen transport in the face of elevated metabolic demands, but as f_H moves closer to peak capacity at high temperatures, there is less scope to increase pacemaker activity and Δf_H can no longer be closely correlated with ΔṀ_O2. Thus, the closer alignment of ΔV_S and ΔṀ_O2 with temperature and the fact that the change in Q̇ with activity became predominantly volume regulated is more a reflection of a limit on peak f_H rather than a change in the ability to regulate V_S. The present data also support the general finding that changes in f_H are prioritized over V_S when a fish is acutely warmed, which is an unresolved issue because quiescent fish with an artificially depressed f_H can elevate V_S in a compensatory fashion (Gamperl et al., 2011), yet exercising fish that are warmed can certainly increase V_S (present study; Steinhausen et al., 2008; Eliason et al., 2011, 2013). Thus, while an increase in f_H reduces cardiac filling time and thus limits the capacity to increase V_S (Altimiras and Axelsson, 2004), at high temperature in C. aceratus, and in exercising fishes more generally, this confounding problem must be overcome for V_S to increase and contribute to increasing Q̇.

During both warming and activity, we revealed surprisingly large (up to 5-fold) increases in G_sys, which indicates vasodilatation of the peripheral vasculature. A similar change has previously been reported in swimming P. borchgrevinki (Axelsson et al., 1992, 1994). This change in conductance can be attributed only in part to passive distension of the vasculature when accommodating greater Q̇ (e.g. Wood and Shelton, 1975) because P_da clearly decreased. Both activity and temperature increase the metabolic demand of tissues, which may release local vasoactive factors and induce a functional hyperaemia. Indeed, in humans it is believed that peripheral vasodilatation is the primary driver of the increased Q̇ during exercise (Bada et al., 2012).

In comparison with Hb⁺ Antarctic notothenioids, our measurements of routine Q̇ in C. aceratus (at 0.8°C, 26.6 ml min⁻¹ kg⁻¹) is not as divergent as previously reported. Indeed, routine Q̇ is lower (29 ml min⁻¹ kg⁻¹ at −0.5°C: Axelsson et al., 1992), or only marginally higher (24 ml min⁻¹ kg⁻¹ at 0°C: Sandblom et al., 2012; 22 ml min⁻¹ kg⁻¹ at −1.0°C: Franklin et al., 2007), than that previously reported in smaller specimens of the more active, cryopelagic P. borchgrevinki. In a more typically benthic species, Trematomus bernacchii, routine Q̇ was lower (17.6 ml min⁻¹ kg⁻¹; Axelsson et al., 1992). The greatest difference reported exists with another benthic species, N. coriiceps (6–7 ml min⁻¹ kg⁻¹: Egginton, 1997; Joyce et al., 2018) of similar body mass to C. aceratus. Thus, there is clearly considerable variation among notothenioid fishes. As a consequence of the lower routine Q̇, the estimated routine CPO (which now also includes P_cv) places C. aceratus squarely within the range of their Hb⁺ relatives (Table 3). This suggests that cardiac pumping in C. aceratus is not necessarily twice as metabolically expensive as in Hb⁺ nototheniids (cf. Sidell and O'Brien, 2006), and the loss of haemoglobin may not be costly in terms of cardiac work. Furthermore, if recent estimates of maximum Q̇ with in situ perfused hearts are accurate (S. E. Egginton, M. Axelsson, E. L. Crockett, K. M. O'Brien and A. P. Farrell, unpublished), factorial scope for Q̇ is about 4 in C. aceratus at ∼0°C, and is in line with that of other fish species such as rainbow trout (Oncorhynchus mykiss) and seabass (Dicentrarchus laborax) (Farrell and Smith, 2017).

Comparisons among related species provide valuable insight into the functional consequences of vicarious loss of individual gene products or wider disruption of gene expression, such as variation in cardiac performance within the Channichthyid icefish family, where six species lack not only haemoglobin but also myoglobin. While C. aceratus occupies a largely benthic habitat, other icefish species are more pelagic (Rutschmann et al., 2011). In C. aceratus, relative ventricular mass is ∼0.3% of body mass (Holeton, 1970; Robertson et al., 1998), whereas in other icefish species, C. rastrospinosus, C. hamatus and Champsocephalus gunnari, it is over 25% greater (0.38–0.4%) (Tota et al., 1991; Robertson et al., 1998; S. E. Egginton, M. Axelsson, E. L. Crockett, K. M. O'Brien and A. P. Farrell, unpublished). Although resting values have not been reported, maximum Q̇ in C. rastrospinosus (128 ml min⁻¹ kg⁻¹) and C. hamatus (300 ml min⁻¹ kg⁻¹) perfused hearts are also greater than those in C. aceratus (70 ml min⁻¹ kg⁻¹) (Tota et al., 1991; S. E. Egginton, M. Axelsson, E. L. Crockett, K. M. O'Brien and A. P. Farrell, unpublished). Chaenocephalus aceratus is among the six out of 16 icefish species that does not express cardiac myoglobin (Moylan and Sidell, 2000; Sidell et al., 1997), the absence of which has been suggested to limit cardiac power, particularly in the face of increased afterload (Acierno et al., 1997). Chionodraco rastrospinosus (Hb⁻Mb⁺) had a larger ventricle, a higher f_H and greater maximum V_S and Q̇ than C. aceratus, suggesting that cardiac Mb has functional relevance in supporting greater CPO in vivo (S. E. Egginton, M. Axelsson, E. L. Crockett, K. M. O'Brien and A. P. Farrell, unpublished).

Our measures show that routine Q̇ and CPO in C. aceratus under ambient temperatures are not as high as the values reported previously, but are still higher than those of benthic red-blooded (Hb⁺) notothenioids. In fact, routine Q̇ in C. aceratus is similar to Q̇ in P. borchgrevinki, a more active cryopelagic species. Furthermore, C. aceratus was also capable of increasing routine Q̇ severalfold to meet the increased Ṁ_O2 when metabolic demands were elevated during an acute temperature challenge up to +8°C and stress-induced activity. Also, there was no difference in cardiac BPT in Hb⁺ N. coriiceps and Hb⁻ C. aceratus during acute temperature challenges.

Hemmingsen and Douglas (1972) were able to maintain C. aceratus ‘in satisfactory condition’ for at least 2 weeks at 8°C and only encountered aberrant behaviour at 10°C. Further, the Patagonian Champsocephalus esox (which is also Hb⁻) occupies habitats in which water temperatures reach as high as 10°C, although this may be the edge of their thermal tolerance/ecological range because modelling studies conclude that at this temperature there is little scope for activity (Egginton et al., 2002).

Together, these data suggest that increased temperature – at least during acute warming – is surprisingly well tolerated by C. aceratus. However, it remains unresolved whether acute thermal tolerance leaves sufficient scope for sustained activity, digestion and reproduction, which are crucial for long-term species survival in a rapidly changing environment.

We are sincerely grateful to Amanda Biederman, Anna Rix and Elizabeth Evans for assistance with fishing and animal care. We are also indebted to the Masters and crew of the ARSV Laurence M. Gould, and the winter support staff at Palmer Station.