Coronary blood flow influences tolerance to environmental extremes in fish

Rapid changes in water temperature and O₂ levels are normal phenomena in aquatic ecosystems, but their incidence and magnitude are exacerbated from ongoing anthropogenic activities such as eutrophication, habitat degradation and climate change; trends that are projected to continue and worsen in the future (Breitburg et al., 2018; Diaz, 2001; Ummenhofer and Meehl, 2017; Woodward et al., 2010). To predict how fish and other aquatic organisms will be impacted, there is a need to understand the physiological mechanisms underlying tolerance to these environmental perturbations. Heat and hypoxia tolerance are often correlated in fish, indicating a functional relationship between the physiological processes that dictate these environmental boundaries (Anttila et al., 2013; Zhang et al., 2018).

The overall metabolism and thus O₂ demand increases in fish during warming, which is met by elevations in cardiovascular O₂ delivery and tissue O₂ extraction. This coincides with elevations in cardiac O₂ demand as cardiac output and cardiac workload (a product of cardiac output and blood pressure) increases along with a direct stimulatory effect of warming on cardiac cellular metabolism (Ekström et al., 2016; Eliason and Anttila, 2017). Similarly, cardiac work appears to increase within a certain hypoxia range in some fish species, e.g. rainbow trout (Oncorhynchus mykiss), as cardiac output is either maintained or elevated while ventral aortic blood pressure increases (Holeton and Randall, 1967; Perry and Desforges, 2006; Steffensen and Farrell, 1998; Wood and Shelton, 1980), whereas other species reduce cardiac work (see table in Stecyk, 2017). Even so, both warming and hypoxia may constrain the supply of O₂ to the heart as arterial blood oxygenation at the gills is impaired, tissue O₂ extraction increases and the partial pressure of O₂ (P_O2) in the venous blood returning to the cardiac lumen decreases (Clark et al., 2008b; Ekström et al., 2016; Steffensen and Farrell, 1998; Thomas et al., 1994). Reductions in P_O2 are well known to impair the contractility of cardiac muscle strips in vitro (Roberts and Syme, 2018), as well as perfused heart preparations (Davie and Farrell, 1991; Davie et al., 1992; Petersen and Gamperl, 2010), suggesting a linkage between cardiac O₂ supply, cardiac performance and environmental tolerance limits.

The hearts of many fishes are composed exclusively of spongy myocardium, which generally relies entirely on the luminal venous blood for O₂ supply. Other fishes (e.g. salmonids) have a coronary circulation, which delivers oxygenated arterial blood and is generally, although not exclusively, associated with myocardial compaction (Axelsson, 1995; Farrell et al., 2012; Farrell and Smith, 2017). The primordial fish heart was hypothesized to be fully vascularized and composed of a mix of spongy and compact myocardium. Yet, the adaptive trade-offs of this cardiac arrangement remain unclear because more than half of all extant fish species lack coronary arteries (Durán et al., 2015; Farrell et al., 2012). Still, hypoxia and cardiac work are thought to be the main evolutionary drivers for coronary circulations (Farrell et al., 2012), highlighting the possibility that exposure to physiological and environmental extremes that induce increased myocardial work and reduced cardiac luminal O₂ supply, have been important selection pressures in the evolutionary history of the coronary circulation. Consequently, coronary perfusion capacity can be hypothesized to play an important role in determining the ability of fishes to withstand environmental challenges.

Salmonid fishes have a coronary circulation that perfuses the compact myocardium of the ventricle, which comprises ∼20–50% of the total ventricular mass (Brijs et al., 2017; Ekström et al., 2017, 2019; Farrell et al., 2009; Farrell and Smith, 2017). Yet, chronic surgical ligation of the coronary artery has been shown to either have little influence on routine cardiac function (heart rate, stroke volume and cardiac output; Gamperl et al., 1994a) or slightly elevate heart rate. This latter response is likely a compensation for compromised cardiac contractility and stroke volume to maintain cardiac output and arterial blood pressure, but there is currently no information available on absolute cardiac output after coronary ligation to substantiate this (Ekström et al., 2018, 2017, 2019; Steffensen and Farrell, 1998). However, consistent with the idea that the adaptive benefits of the coronary circulation are predominantly evident when fish are exposed to environmental and physiological challenges, coronary blood flow in salmonids increases markedly during acute hypoxia (Axelsson and Farrell, 1993; Gamperl et al., 1994b, 1995) and warming (Ekström et al., 2017), as well as during exercise (Gamperl et al., 1995). The functional significance of this response is evident after coronary artery ligation, which reduces relative stroke volume and cardiac output in rainbow trout exposed to acute warming (Ekström et al., 2017; Ekström et al., 2019), and impairs ventral aortic blood pressure generation in rainbow trout swimming in hypoxia (Steffensen and Farrell, 1998). Moreover, coronary ligation reduced the critical thermal maximum (CT_max; Ekström et al., 2017; Ekström et al., 2019) and compromised cardiac and aerobic metabolic capacity in rainbow trout in normoxia (Ekström et al., 2018), as well as the maximum sustained swimming speed in chinook salmon (Farrell and Steffensen, 1987). However, there is currently no information detailing the influence of coronary flow on absolute cardiac output responses during hypoxia or warming in salmonids, or in governing tolerance to hypoxia in fish in general.

The specific mechanisms dictating hypoxia tolerance are multi-faceted and can be broadly related to various combinations of a fish's capacity to: (i) maintain aerobic metabolism during hypoxia, (ii) depress overall metabolism and/or (iii) sustain anaerobic metabolism (for review, see Mandic and Regan, 2018; Richards, 2011). Hypoxia tolerance is often quantified by the critical Pw_O2 (P_crit), defined as the Pw_O2 below which the fish cannot sustain the O₂ consumption rate (Ṁ_O2; a proxy for aerobic metabolic rate) needed for aerobic standard metabolic rate (SMR) in hypoxia (Negrete and Esbaugh, 2019; Ultsch and Regan, 2019). When the animal cannot sustain SMR below P_crit, a mismatch between tissue ATP supply and demand develops and ultimately leads to loss of equilibrium at a certain Pw_O2 (i.e. P_LOE; Claireaux and Chabot, 2016; Negrete and Esbaugh, 2019; Rogers et al., 2016). There are several cardiovascular adjustments thought to promote O₂ extraction, transport and delivery to tissues in hypoxia (Gamperl and Driedzic, 2009; Sandblom and Axelsson, 2011; Stecyk, 2017). Below a certain Pw_O2, obligate water-breathing fishes typically develop a vagal reflex inhibition of heart rate (i.e. hypoxic bradycardia; Farrell, 2007; Sandblom and Axelsson, 2005; Sandblom et al., 2009; Stecyk, 2017), which is initiated by stimulation of O₂-sensitive chemoreceptors in the gills (see Milsom, 2012). This bradycardia has been suggested to benefit cardiac performance in hypoxia by, for example, decreasing cardiac work and O₂ demand (Speers-Roesch et al., 2010), as well as improving the O₂ delivery to the myocardium by increasing end-diastolic volume and the residence time of blood in the heart lumen, which promotes O₂ diffusion into the spongy myocardium (for review, see Farrell, 2007; Stecyk, 2017). The prolongation of diastole with bradycardia also elevates stroke volume through increased diastolic filling and contractile force (i.e. Frank–Starling mechanism and the negative force–frequency relationship; for reviews, see Sandblom and Axelsson, 2007; Shiels et al., 2002). This allows cardiac output to be maintained, or even increase in hypoxia (Stecyk, 2017; Wood and Shelton, 1980). Interestingly, coronary blood flow may play an important role in this response, as coronary perfusion improves the contractility of isolated heart preparations during hypoxic conditions (Davie and Farrell, 1991; Davie et al., 1992). Furthermore, the hypoxic bradycardia likely enhances coronary blood flow by extending the diastolic period when coronary blood flow peaks (Axelsson and Farrell, 1993; Farrell, 2007). Taken together, these observations strongly indicate an important role of the coronary arteries for promoting cardiac performance in hypoxia, which likely extends to governing hypoxia tolerance in fish.

The aim of this study was to determine the role of coronary O₂ supply to the heart during exposure to acute hypoxia and warming, and to study the mechanisms dictating cardiac and whole-animal tolerance to these environmental extremes. We determined the effects of coronary ligation on cardiac (absolute cardiac output, heart rate and stroke volume) and respiratory performance (Ṁ_O2) in rainbow trout subjected to progressive acute hypoxia to determine P_crit and P_LOE, as well as a standardized thermal challenge protocol to determine CT_max. Specifically, we hypothesized that coronary ligation would impair cardiac performance and the ability to sustain aerobic metabolism in hypoxia and during warming, which would compromise tolerance to hypoxia (i.e. elevate P_crit and/or P_LOE) and heat (i.e. reduce CT_max). We also analysed the relative contribution of key cardiorespiratory traits underlying intra-specific variation in environmental tolerance limits, as well as the relationship between hypoxia and heat tolerance across experimental treatment groups to unravel potential common underlying cardiorespiratory mechanisms.

Rainbow trout (Oncorhynchus mykiss Walbaum 1792; see Table 1 for morphological details) of mixed sex where obtained from a commercial fish farm (Vänneåns Fiskodling AB, Halland, Sweden). The fish were acclimated to laboratory conditions for a minimum of 4 weeks and were held in 1000 litre tanks continuously supplied with aerated recirculating freshwater at 10±0.5°C under a 12 h:12 h light:dark photoperiod. The fish were fed pellets (7 mm, Protec Trout pellets, Skretting, Norway) twice a week until the start of the experiments, and were fasted for 3 days before the day of the surgical procedure to ensure they were in a post-absorptive state throughout the protocol. All experimental procedures were covered by ethical permit 165-2015, approved by the regional ethical committee in Gothenburg.

Individual fish were anesthetized prior to surgery in freshwater containing 150 mg l⁻¹ MS-222 (Tricaine methanesulfonate, Scan Aqua AS, Årnes, Norway) buffered with 300 mg l⁻¹ NaHCO₃. Once ventilatory movements ceased, indicating a surgical plane of anaesthesia, a 300 μl blood sample was drawn from the caudal vessel for analysis of haematological variables (haematocrit and [haemoglobin]). Fork length and body mass (M_b) were recorded, and the fish was then placed on its left side on a surgery table covered with wet foam. During the surgery, the gills were perfused with a continuous flow of aerated freshwater (10°C) containing 75 mg l⁻¹ MS-222 and 150 mg l⁻¹ NaHCO₃. A small incision was made in the isthmus and blunt dissection was used to expose the ventral aorta and the coronary artery. Care was taken not to damage surrounding vessels and nerves and to ensure that the pericardium remained intact. One group of fish had their coronary artery ligated with a 6-0 silk suture (coronary-ligated group), downstream from where the coronary artery branches from the hypobranchial artery. Another group of fish underwent an identical surgical protocol with the exception that the coronary artery was not ligated and left intact (sham-operated group). A 2.5 mm Transonic transit-time blood flow probe (L type; Transonic Systems, Ithaca, NY,USA) was then placed around the ventral aorta to allow recordings of ventral aortic blood flow (cardiac output). The probe was secured with 3-0 silk sutures inside the opercular cavity and in the skin just outside of the opercular cavity and a final (2-0) suture in front of the dorsal fin. Following instrumentation, the fish were placed individually in 10 litres Perspex respirometers submerged in a ∼120 liters experimental tank filled with aerated recirculating freshwater at 10.5±0.05°C. The fish were allowed to recover from surgery for ∼20 h, throughout which Ṁ_O2 was continuously recorded (see below for details). The experiments were run in pairs daily with one sham-operated and one coronary-ligated fish in each run. All fish were then subjected to a hypoxia challenge followed by a temperature challenge as described below.

Measurements of cardiorespiratory variables in normoxia commenced in the morning and were recorded for a minimum of 2 h before the start of the hypoxia challenge. Once steady state heart rate values had been obtained during the initial recording period, the hypoxia challenge commenced and was performed using closed circuit respirometry following previously described protocols (Negrete and Esbaugh, 2019; Reemeyer and Rees, 2019). Briefly, by turning off the inflow of normoxic water (Pw_O2 ∼21 kPa at 100% air saturation) into the respirometers, the O₂ consumption of the fish caused a gradual decrease in Pw_O2 to hypoxic levels until the fish reached P_LOE (Nilsson and Östlund-Nilsson, 2004), which was defined as the Pw_O2 where the fish lost equilibrium for longer than 10 s. Cardiac measurements were recorded at P_LOE for one more minute, after which the individual respirometer was reperfused with aerated water to bring the Pw_O2 inside the respirometer back to normoxic conditions. The duration of the hypoxia challenge averaged 61 min and ranged between 25 to 109 min. All fish recovered once normoxic conditions were restored.

Following >20 h of recovery from the hypoxia challenge, which was sufficient for all cardiorespiratory variables to return to routine values, baseline values were again recorded for a minimum of 2 h before the start of an acute heating challenge. The temperature was then increased from 10.5°C to 20°C in 5°C h⁻¹ increments, with the temperature ramp lasting approximately 30 min and the target temperature being maintained for another 30 min. When 20°C was reached, the temperature was increased further by 1°C steps every 30 min (2°C h⁻¹), where the ramping lasted approximately 10 min and the target temperature was maintained stable for 20 min. CT_max was determined as the temperature at which the fish lost the ability to maintain an upright body position for more than 10 s. Before each temperature increment, two Ṁ_O2 slopes were recorded by turning off the flush pump and letting the % air saturation decrease by ∼10% before reinitiating flushing. Once CT_max was reached, the fish were removed from the respirometer and euthanized via blunt trauma to the head, immediately followed by drawing a blood sample from the caudal vessels for analyses of haematological variables. To ensure that the coronary artery had been successfully ligated, the position and integrity of the suture was verified post mortem. Finally, the wet mass of the spleen and heart ventricle (after careful removal of atrium, bulbus and any remaining blood in the lumen) were determined. The ventricle was preserved in 70% ethanol for further analyses (see below).

The % air saturation inside the respirometer was continuously measured using an O₂ optode connected to a Firesting O₂ system (PyroScience, Aachen, Germany). Ṁ_O2 was determined using automated intermittent closed respirometry and calculated from the decline in % air saturation (i.e. slope) in the respirometer between flush cycles. Ṁ_O2 measurements at rest were typically performed in continuous 20 min cycles with the flush cycle set to 15 min. However, to ensure that water air saturation remained >85%, the cycles were reduced to 15 min cycles with 12 min flushing periods for some fish.

where V_r is the volume of the respirometer, V_f is the volume of the fish assuming that 1 g of fish equals 1 ml of water, Δ%Sat/t is the change in % O₂ saturation per unit time, α is the solubility coefficient of O₂ in freshwater and adjusted for the different experimental temperatures and M_b is the body mass of the fish (Clark et al., 2013). SMR was calculated as the lowest 20th percentile of the Ṁ_O2 values (Chabot et al., 2016) obtained throughout the ∼20 h overnight post-surgery recovery period up until the start of the hypoxia challenge. Routine metabolic rate (RMR) throughout the hypoxia protocol was calculated from the average slope at 10% air saturation intervals until the air saturation reached 50%, and then at 5% intervals until P_LOE (i.e. starting at 90, 80, 70, 60, 50, 45, 40, 35, 30, 25, 20, 15 and 10% air saturation). SMR values were then plotted against Pw_O2, and P_crit was determined by fitting a linear regression to RMR values during the hypoxia exposure that were below the normoxic SMR to identify the Pw_O2 where the regression line intercepted the SMR (P_crit) obtained in normoxia (see supplementary material, Fig. S1; McBryan et al., 2016). Prior to the temperature challenge, RMR at 10.5°C was obtained by calculating the average from four Ṁ_O2 slopes when the fish was in a steady state, as determined by a low and steady heart rate. During the temperature challenge, mean values were calculated from the two Ṁ_O2 slopes at each temperature step when the temperature had stabilized. Measurements of bacterial background respiration were performed following each trial after the removal of the animals from the respirometers and the acquired slopes were subtracted from the Ṁ_O2 slopes of the fish.

The Transonic flow probe was connected to a Transonic blood flow meter (model T206; Transonic Systems, Ithaca, NY) and the signals where recorded using a PowerLab system (ADInstruments, Castle Hill, Australia) at a sampling rate of 10 Hz using LabChart pro data acquisition software (v.7.3.2, AD Instruments, Castle Hill, Australia). To correct for any temperature effects on the Transonic flow probe readings, all probes were individually bench calibrated as specified by the user manual at 10.5, 15, 20, 22, 24, 26, 28 and 30.0°C. Briefly, the transonic flow probe was mounted on a calibration tubing submerged in a temperature controlled water bath. A flow of temperature-regulated water was pumped through the tubing using a peristaltic pump (Gilson 312 Minipuls 3, Villiers-Le-Bel, France) for 1 min. The outflowing water was collected and weighed to determine the flow gravimetrically. This process was repeated at increasing flow rates covering the flow range of the probe (1 to 100 ml min⁻¹). The values obtained from the transonic flow probe were plotted against the recorded gravimetric flow and the resulting regression equation was used to correct the obtained in vivo cardiac output values. The heart rate was calculated from the pulsating blood flow traces using the cyclic measurements module in LabChart Pro. Stroke volume was calculated as the quotient of cardiac output divided by heart rate.

The Pw_O2 at the onset of hypoxic bradycardia was determined via segmental linear regression analyses using GraphPad prism 8.3.0, according to established methods for identifying critical break points (Yeager and Ultsch, 1989). Briefly, this method fits two lines that intersect at the Pw_O2 where heart rate starts to decrease. The Pw_O2 and average value for peak stroke volume in hypoxia (peak SV_hypoxia), and cardiac output at peak SV_hypoxia (CO at peak SV_hypoxia) were determined for each fish. We also determined the temperatures and average peak values for cardiac output, heart rate and MO₂ during warming (peak CO_warming, peak heart rate and peak Ṁ_O2, respectively), as well as the average stroke volume at peak CO_warming (SV at peak CO_warming).

Statistical analyses were performed using SPSS statistics 24 for Windows (IBM Corp., Armonk, NY, USA). Differences between groups (i.e. sham-operated versus coronary-ligated) in morphological variables, environmental tolerance indices (i.e. CT_max, P_crit and P_LOE), specific cardiorespiratory indices (e.g. peak responses, bradycardia) and mean routine cardiorespiratory values (at ≥18.9 kPa and 10.5°C) were analysed using independent-sample t-tests, Welch t-test (if variances were unequal) or Mann–Whitney U-tests (if data were not normally distributed). Differences in haematological variables where analysed using a linear mixed model with an unstructured repeated covariance structure, with fish ID as subject variable and sampling time (initial versus final), treatment (sham-operated versus coronary-ligated) and their interaction as fixed factors. Differences in cardiorespiratory variables during the hypoxia and temperature challenges were analysed using a linear mixed model with a within-subjects factor (Pw_O2 or temperature), treatment and the interaction between the within-subject factor and treatment as fixed factors, and fish ID as subject variable. Either a first-order autoregressive or a heterogeneous first-order autoregressive repeated covariance structure was used, which provided best fit to the models (as indicated by the lowest Akaike's information criterion, AIC). Pw_O2 values ranging from ≥18.9 to 2.1 kPa were included in the models assessing the effects of hypoxia and values at P_LOE were excluded. Models assessing the effects of temperature included temperatures ranging from 10.5 to 22°C, as this was the lowest temperature at which individual fish peaked across all variables. If significant interactions were found, these were further explored with between- and within-treatments pair-wise comparisons, where confidence intervals were adjusted for multiple testing using Bonferroni correction. For the statistical analysis of the hypoxia challenge, cardiac output, heart rate and Ṁ_O2, as well as Ṁ_O2 during the temperature challenge, were transformed to their natural logarithm to comply with the assumption of homoscedasticity of the residuals. P_crit was transformed to its natural logarithm to comply with the assumption of normality.

Multiple regression models were used to determine the relative contribution of key variables on the variances in hypoxia (P_LOE) and temperature tolerance (CT_max). In these analyses, values for sham-operated and coronary-ligated treatments were pooled. Moreover, to determine potential relationships between temperature and hypoxia tolerance, additional regression analyses were carried out with CT_max as the dependent variable and P_LOE, P_crit and treatment (sham-operated versus coronary-ligated) as independent variables. The variables used in the multiple regression models were selected as they lacked multicollinearity and appeared to follow a linear relationship with the dependent variables (P_LOE and CT_max) when assessed by visual inspection of partial regression plots and fitted the assumptions of the model. Variables were subsequently gradually eliminated from the model using stepwise backward regression until only the variables that best explained the variability in the dependent variable remained. Only adjusted R² values are reported. Regression coefficients, standard errors and confidence intervals are included in Table S1. In addition, correlations between peak cardiac variables and cardiac morphological variables were analysed for each treatment using Pearson's correlations (P_corr). Statistical significance was accepted at P<0.05. All data are presented as means±s.e.m.

There were no differences in overall body characteristics (M_b, fork length and condition factor), cardiac morphology (relative ventricular mass and % compact myocardium), haematological variables or relative spleen mass between treatment groups (Table 1).

In normoxia, there were no differences in SMR between sham-operated control and coronary-ligated trout (58.6±4.5 versus 60.8±2.8 mg O₂ kg⁻¹ h⁻¹, respectively; Fig. S1). Similarly, there were no differences between control and coronary-ligated fish in routine cardiac output (13.4±1.2 versus 11.4±1.3 ml min⁻¹ kg⁻¹, respectively; Fig. 1B) in normoxia. Even so, heart rate was significantly elevated after ligation (57.1±1.9 versus 47.6±3.9 beats min⁻¹; t_11.888=−2.213, P=0.047; Fig. 1C), which coincided with a significantly reduced stroke volume relative to control trout (0.20±0.03 versus 0.29±0.02 ml kg⁻¹; t₂₁=2.461, P=0.012; Fig. 1D).

Given the methodology used, body mass affected the rate of decline in Pw_O2 during the gradual hypoxia exposure. However, there was no linear relationship between the duration of the hypoxia protocol or body mass with P_crit or P_LOE (as indicated by visual inspection of scatterplots; data not shown). Throughout the hypoxia challenge, there were no significant differences in Ṁ_O2 between treatment groups (Fig. 1A). Moreover, P_crit occurred at similar Pw_O2 levels in control and coronary-ligated trout (Table 2). However, P_LOE was significantly higher in coronary-ligated trout (t₂₁=−2.284, P=0.033; Table 2). As hypoxia progressed, an increasingly pronounced bradycardia developed in both groups (Fig. 1C), but the onset of bradycardia occurred at a significantly higher Pw_O2 by 2.6 kPa in coronary-ligated fish (t₁₇=−2.384, P=0.029; Table 2). Moreover, while cardiac output was maintained in hypoxia via a significant increase in stroke volume in the sham-operated control fish (Fig. 2B and D), the stroke volume in coronary-ligated trout was significantly lower during the hypoxia exposure (Fig. 1D). Consequently, the peak SV_hypoxia was significantly lower (t₂₁=4.747, P<0.001) and occurred at a higher Pw_O2 (U=102, P=0.021; Table 2), which coincided with a significantly lower cardiac output in coronary-ligated fish during the hypoxia exposure (Fig. 1B). Multiple regression analyses revealed that relative ventricular mass, percentage compact myocardium and peak SV_hypoxia combined, significantly explained part of the variation in P_LOE (F_3,19=4.896, P=0.011, R²=0.35).

There were no differences between control and coronary-ligated trout in normoxia at 10.5°C, following the recovery from the hypoxia exposure, with regards to routine Ṁ_O2 (53.0±4.6 versus 59.5±4.0 mg O₂ kg⁻¹ h⁻¹, respectively; Fig. 2A) and cardiac output (11.2±1.2 versus 10.0±1.0 ml min⁻¹ kg⁻¹, respectively; Fig. 2B). Moreover, although routine heart rate was not significantly different between groups (41.3±3.5 versus 50.8±3.0 beats min⁻¹, respectively), a trend towards a higher heart rate was observed in coronary-ligated trout (t₂₁=−2.003, P=0.058; Fig. 2C), which coincided with a significantly lower stroke volume (0.29±0.03 versus 0.20±0.02 ml kg⁻¹; t₂₁=2.248, P=0.035; Fig. 2D).

Ṁ_O2 did not differ between treatment groups and increased with warming at a similar rate across treatment groups (Fig. 2A). However, the overall temperature tolerance was significantly reduced in coronary-ligated trout as indicated by a 1.3°C reduction in CT_max (U=20, P=0.004; Table 3). Consistent with the increase in Ṁ_O2, cardiac output also increased with temperature in both groups, although the increase was greater in control fish from 21°C onwards (temperature effect_sham: F_4,51=12.528, P<0.001; temperature effect_ligated: F_4,51=3.947, P=0.007; Fig. 2C). The increased cardiac output was governed by elevations in heart rate in both groups (Fig. 2B,C). Stroke volume remained significantly lower in coronary-ligated trout throughout the thermal challenge relative to the control group (Fig. 2D). Consequently, the peak CO_warming was lower in coronary-ligated fish (t_12.275=3.696, P=0.003; Table 3), which was due to a decreased SV at peak CO_warming (t_11.902=3.476, P=0.005; Table 3). Moreover, the temperatures at peak heart rate, peak Ṁ_O2 and peak CO_warming were lower in the ligated compared with the sham-operated trout (U=16, P=0.002; U=23, P=0.047 and U=22, P=0.006, respectively; Table 3). Multiple regression analyses revealed that the peak CO_warming significantly explained part of the variation in CT_max (F_1,21=9.685, P=0.005, R²=0.28), such that CT_max was positively correlated with CO_warming (P_corr=0.548).

Haematocrit was significantly increased following the temperature challenge in both control and coronary-ligated fish (Table 1), whereas [Haemoglobin] only increased significantly in the control group (Table 1). There were no significant differences in MCHC between treatment groups at the end of the temperature challenge (treatment effect_afterCTmax: F_1,20=2.964, P=0.101; Table 1). However, MCHC was significantly reduced compared with pre-surgery values in coronary-ligated trout indicating pronounced red blood cell swelling, but this was not observed in sham-operated fish (Table 1).

A multiple regression was run to predict CT_max from P_crit, P_LOE and treatment. The model revealed that treatment (F_1,21=12.352, P=0.002, R²=0.34), P_LOE (F_1,21=8.485, P=0.008, R²=0.25) and treatment in combination with P_LOE (Fig. 3) significantly predicted CT_max. This indicates that individuals with lower P_LOE (and thus greater hypoxia tolerance) were also the most temperature tolerant, which strengthens the above findings that coronary ligation reduces tolerance to both environmental extremes.

A significant negative correlation was found in the coronary-ligated fish between the percentage compact myocardium and peak SV_hypoxia (P_corr=−0.698; P=0.008; Fig. 4A), and consequently CO at peak SV_hypoxia (P_corr=−0.609; P=0.027). Similarly, peak CO_warming (P_corr=−0.655; P=0.015; Fig. 4B) and SV at peak CO_warming (P_corr=−0.629; P=0.021) were also negatively correlated with percentage compact myocardium. Thus, the acute coronary ligation in individuals with a large proportion of compact myocardium (which was presumably devoid of coronary blood supply) resulted in lower peak stroke volume and cardiac output during exposure to both hypoxia and warming. These significant correlations were not observed in the sham-operated control group (Fig. 2A,B).

Similarly to previous studies, coronary ligation resulted in either a trend towards an elevated routine heart rate (prior to the heating protocol) or a significantly elevated routine heart rate (prior to hypoxia protocol) in rainbow trout (Ekström et al., 2018, 2017, 2019; Steffensen and Farrell, 1998). Here, we can show for the first time that this response serves to maintain O₂ consumption rate and cardiac output by compensating for a ∼44% decrease in stroke volume following the coronary ligation. The elevations in heart rate are most likely mediated via a reduced cholinergic (i.e. vagal) tone on the heart, presumably by a reduction in ventral aortic blood pressure inducing a barostatic reflex (see Sandblom and Axelsson, 2011). Indeed, Ekström et al. (2019) showed that treatment with the muscarinic antagonist atropine, which normally would abolish the cholinergic tone on the heart thus leading to an elevated heart rate, had no effect on heart rate in coronary-ligated rainbow trout suggesting that the ligation per se had already caused a release of cholinergic tone.

In both treatment groups, the bradycardic responses to hypoxia occurred at a Pw_O2 (7.5–10.1 kPa), which is within the range of previously reported values for this species (see table in Stecyk, 2017). The onset of bradycardia occurred at a Pw_O2 that was 2.6 kPa higher in the ligated group compared with the control group. It is possible that this was because the input from the branchial O₂-sensitive chemoreceptors in hypoxia overrode the barostatic response that may initially have kept heart rate elevated in the coronary-ligated fish. Even so, below a Pw_O2 of ∼8 kPa the heart rate was similar between treatment groups, suggesting that the hypoxic bradycardia response dominated. However, in contrast to the control trout where stroke volume increased substantially as the hypoxic bradycardia developed, the stroke volume was only marginally elevated in trout with ligated coronary arteries, which meant that cardiac output collapsed and was halved immediately prior to P_LOE relative to the cardiac output recorded in normoxia. Thus, the abolished O₂ delivery to the compact myocardium hampered the capacity to elevate stroke volume and maintain cardiac output at lower Pw_O2 levels, in contrast to trout with an intact coronary flow, which were far better able to maintain cardiac output at reduced ambient Pw_O2. Another possible contributing factor that requires further testing is that along with an insufficient myocardial oxygenation, ligation of the coronary circulation also prevents catecholamines from reaching and effecting the compact myocardium. These are normally released during acute hypoxia (Perry and Reid, 1994, 1992) and protect myocardial function from hypoxia, acidosis and hyperkalaemia (see Driedzic and Gesser, 1994; Hanson et al., 2006; Roberts and Syme, 2018).

Despite a cardiac output that was twice as high in the control relative to coronary-ligated trout immediately prior to P_LOE, coronary-ligated trout were only marginally less tolerant to acute hypoxia as their P_LOE was only elevated by 0.6 kPa (∼3% air saturation). Moreover, the P_crit for both treatment groups were within the range of previously reported values for rainbow trout (Williams et al., 2019; Wood, 2018), and while P_crit was numerically slightly higher in the ligated fish (by 0.8 kPa), this was not statistically significant. Thus, it seems likely that the coronary-ligated fish initiated some response in hypoxia that compensated for the severely reduced capacity for circulatory O₂ delivery to the tissues, which in turn minimized the negative impacts of the coronary obstruction on overall hypoxia tolerance. Indeed, both treatment groups maintained a similar Ṁ_O2 throughout most of the hypoxia protocol, suggesting a significantly increased tissue O₂ extraction in the coronary-ligated trout when cardiac output collapsed in hypoxia. Given that Steffensen and Farrell (1998) found no difference in venous P_O2 between sham-operated and coronary-ligated rainbow trout, and since blood O₂ carrying capacity (as indicated by a similar haematocrit) did not differ between treatments, one possibility is that there was a right (Bohr effect) and possibly downward shift (Root effect) in the haemoglobin O₂ dissociation curve in the coronary-ligated trout, thus augmenting O₂ unloading at the tissues (Harter and Brauner, 2017; Rummer and Brauner, 2015; Rummer et al., 2013). This shift could have been caused by an exacerbated acidosis in coronary-ligated fish in hypoxia, but unfortunately neither blood pH nor blood O₂ content was measured in the current or earlier studies to confirm this. Although the overall effects of acute coronary ligation on hypoxia tolerance where relatively mild, the effects of a sustained impaired coronary O₂ delivery over the long-term on organismal hypoxia tolerance remains to be explored.

Our study is the first to demonstrate the role of the coronary circulation in maintaining cardiac output and stroke volume along with respiratory performance during warming. Cardiac output is mainly elevated via heart rate increases in response to warming (Eliason and Anttila, 2017). Similarly to previous observations (Ekström et al., 2017; Ekström et al., 2019), the heart rate peaked at a ∼1.6°C lower temperature in coronary-ligated trout, although the peak heart rate was similar for both treatment groups. It is possible that the lower temperature for peak heart rate during warming was simply an effect of the slightly higher routine heart rate often observed in coronary-ligated trout (Ekström et al., 2018, 2017, 2019; Steffensen and Farrell, 1998). The peak CO_warming was driven by the peak heart rate in both control and ligated fish, which meant that the peak CO_warming occurred at 1.8°C lower temperature in trout with ligated coronary arteries. However, despite the similar peak heart rate response, the peak CO_warming was reduced by 45% in ligated fish. Similarly to the situation in hypoxia, this impaired capacity to elevate cardiac output with warming was explained by a severely constrained stroke volume (reduced by ∼44%) at peak CO_warming in coronary-ligated trout. This is consistent with the ∼50% reduction in peak relative cardiac output (recorded using Doppler flow probes) in coronary-ligated trout observed by Ekström et al. (2019). Despite the reduced cardiac output during warming, peak Ṁ_O2 was not statistically different between groups; again, however, the peak occurred at a lower temperature in ligated fish. Nevertheless, the earlier collapse of cardiorespiratory function in the coronary-ligated trout was associated with a lower CT_max, indicating a loss of heat tolerance (Ekström et al., 2017; Ekström et al., 2019).

Following thermal ramping, haematocrit and [haemoglobin] increased in the control fish, indicating splenic contraction (Perry and Kinkead, 1989), whereas haematocrit increased and MCHC was reduced in the coronary-ligated group, indicating erythrocyte swelling (Nikinmaa, 1982; Templeman et al., 2014). Both responses are consistent with an increase in circulating catecholamines, which commonly occurs following acute warming and increases blood O₂ carrying capacity (Templeman et al., 2014) and haemoglobin O₂ affinity by elevating erythrocytic intracellular pH (Nikinmaa, 1982). However, as the reduction in MCHC was more pronounced in the coronary-ligated trout, it is possible that this reflects a more pronounced increase of circulating catecholamines in this group in an attempt to mitigate the effect of blood acidosis on haemoglobin O₂ affinity and to increase blood O₂ carrying capacity in face of hypoxemia at elevated temperatures (Boutilier et al., 1986; Perry et al., 1989).

Although acute tolerance limits to both hypoxia and warming have been previously linked to cardiac function, it is less clear if there are common underlying mechanisms that confer tolerance to both environmental extremes. For example, intra-specific differences in whole-animal hypoxia tolerance in European seabass (Dicentrarchus labrax) were positively correlated with contractile force production of ventricular strip preparations under both normoxia and hypoxia (Joyce et al., 2016). Moreover, studies in various teleost species indicate that intra-specific warming tolerance is positively correlated with cardiac morphological traits such as ventricle size (Anttila et al., 2013; Ozolina et al., 2016) and cardiac myoglobin levels (Anttila et al., 2013). Anttila et al. (2013) also found a positive association between hypoxia and thermal tolerance, indicating that a better O₂ supply to the heart (i.e. due to increased myoglobin levels) and a capacity for maintaining a higher cardiac output (given the positive association between cardiac output and ventricle size; Franklin and Davie, 1992) could be a common trait associated with tolerance to both environmental drivers. In the current study, 33% of the variation in P_LOE could be explained by the combined variation in relative ventricular mass, percentage compact myocardium and peak SV_hypoxia, whereas peak CO_warming had the most pronounced influence on temperature tolerance and accounted for 26% of the individual variability in CT_max. This highlights not only the importance of cardiac function in determining overall hypoxia and temperature tolerance but also some of the underlying mechanisms that allow for such enhanced performance. Indeed, a relatively larger ventricle can generate a larger stroke volume for a given size fish (Franklin and Davie, 1992) and a larger percentage compact myocardium may allow for a greater pressure generating capacity, especially as hearts become larger and require a greater ventricular wall tension production (Laplace's law; Brijs et al., 2017; Farrell, 1991; Farrell et al., 2009). Similarly, ventricle size and percentage compact myocardium has been shown to positively correlate with other fitness traits such as migratory capacity in sockeye salmon (Oncorhynchus nerka), highlighting the relationship between these cardiac morphological traits and the ability of the fish to overcome the challenges they face in their upstream spawning migrations (Eliason et al., 2011). Our data also indicate that the impaired cardiac performance in hypoxia and during warming after coronary ligation was specifically related to an ischaemic compact myocardium, because there was a negative relationship between percentage compact myocardium and the peak SV_hypoxia and CO at peak SV_hypoxia, as well as between percentage compact myocardium and peak CO_warming and SV at peak CO_warming. We interpret this finding as individuals with a larger proportion of O₂-deprived compact myocardium suffered greater reductions in ventricular contraction force as luminal O₂ became increasingly limiting in hypoxic and warming conditions.

Although coronary arteries appear to have been present in the hearts of evolutionarily ancient fishes, they have been lost multiple times throughout their evolutionary history (Durán et al., 2015; Farrell et al., 2012), raising questions regarding the adaptive significance of their presence or absence (Farrell et al., 2012). Our results highlight that while coronary blood flow is not essential for the maintenance of routine cardiac output, its importance increases when fish are exposed to environmental extremes such as severe hypoxia and high temperatures. This allows fishes with high aerobic capacity like salmonids to increase cardiac work even during conditions and physiological states when cardiac luminal O₂ supply is severely constrained. As a corollary, our findings also highlight that pathological restrictions in coronary blood flow (e.g. coronary arteriosclerosis), which are known to be highly prevalent in both farmed and wild salmonids (Brijs et al., 2020; Farrell, 2002), may severely impact on cardiac capacity and animal performance traits. The coronary O₂ delivery is likely of particular adaptive benefit during episodes of elevated physical activity and high cardiac workloads (e.g. swimming during challenging spawning migrations), specifically when combined with environmental conditions that constrain luminal O₂ supply such as extreme temperatures and hypoxia as examined here.

We gratefully acknowledge Jeroen Brijs for the advice regarding the data analyses. The authors also acknowledge Julija Česnulaitytė, Helena Delgado, Laura Glasner, Frida Pallapies, Staffan Persson, Teemu Piippolainen, Álvaro Pradal, Saga Samuelsson, Frida Seger, Jaime Villaverde and Yannic Wocken for the practical assistance during some of the experiments.