Cardiorespiratory adjustments to chronic environmental warming improve hypoxia tolerance in European perch (Perca fluviatilis)

Aquatic hypoxia, a condition of reduced water O₂ partial pressure (Pw_O2), is a naturally occurring phenomenon in aquatic ecosystems (Diaz and Breitburg, 2009). However, hypoxic conditions have spread and become more severe in recent decades (Breitburg et al., 2018). For example, the frequency and extent of hypoxia in the Baltic Sea has increased severalfold in both the open sea and shallow coastal areas, which has led to fish kills and large uninhabitable areas (Carstensen et al., 2014; Carstensen and Conley, 2019; Casini et al., 2016). This expansion of hypoxia is mainly a result of increased anthropogenic nutrient input causing eutrophication and increased microbial O₂ consumption (Carstensen and Conley, 2019). While many water-breathing fishes have evolved a suite of physiological, biochemical and behavioral traits to cope with aquatic hypoxia (for reviews, see Farrell and Richards, 2009a; Mandic and Regan, 2018; Richards, 2011), species-specific tolerance limits are highly variable, and there is increasing concern that anthropogenically induced hypoxia may negatively impact numerous biological responses and performance traits (Sampaio et al., 2021), and even push species beyond their tolerance limits (McBryan et al., 2013). An additional problem is that elevated temperatures in a changing climate may exacerbate hypoxic conditions in the future, mainly because warming decreases the O₂ solubility of water, increases microbial O₂ consumption, and reinforces thermal stratification that prevents water mixing and re-oxygenation (Carstensen and Conley, 2019; McBryan et al., 2013; Oschlies et al., 2018; Perello et al., 2017; Shepherd et al., 2017).

If severe enough, acute hypoxia constrains blood oxygenation at the gills in fish, which may compromise tissue O₂ delivery and aerobic metabolism (Boutilier, 2001; Butler and Taylor, 1975; Chan, 1986; Holeton and Randall, 1967; Powell et al., 2000; Stecyk, 2017; Steffensen and Farrell, 1998). Most fish, however, are ‘O₂ regulators’, meaning that they adjust cardiorespiratory functions (cardiovascular and metabolic) to maintain a stable O₂ consumption rate (Ṁ_O2, a proxy for aerobic metabolic rate) over a range of species-specific hypoxic Pw_O2 values (Svendsen et al., 2019). At a specific critical Pw_O2 (P_crit), however, cardiorespiratory adjustments are no longer sufficient to compensate for increasingly lower O₂ diffusion gradients, and a transition from O₂ regulation to O₂ conformation occurs, which is marked by a progressive decline in Ṁ_O2 (Rogers et al., 2016; Ultsch and Regan, 2019). If low Pw_O2 levels persist below P_crit and the animal cannot compensate with increased anaerobic metabolism and/or reduce energetic demands (e.g. through metabolic depression), a progressive mismatch between tissue ATP supply and demand develops (Richards, 2011). This leads to compromised vital functions, loss of equilibrium (LOE) and eventually death (Boutilier, 2001; Claireaux and Chabot, 2016; Farrell and Richards, 2009b; Rodnick and Gesser, 2017; Speers-Roesch et al., 2013; Wood, 2018).

In addition to exacerbating aquatic hypoxia per se, acute warming elevates the Ṁ_O2 of fish (Eliason and Anttila, 2017), which typically reduces the ability to maintain aerobic metabolism in hypoxia as indicated by elevated P_crit (Borowiec et al., 2016; Butler and Taylor, 1975; Corkum and Gamperl, 2009; Ern et al., 2016; Fry and Hart, 1948; Gehrke and Fielder, 1988; Rogers et al., 2016), as well as elevated Pw_O2 at which LOE occurs (P_LOE) (Borowiec et al., 2016), and a reduced time to LOE in fish exposed to a constant level of severe hypoxia (He et al., 2015; McBryan et al., 2013, 2016). A key question is to what extent phenotypic plasticity or genetic adaptation of physiological functions following long-term (chronic) warming can alleviate the expected impaired hypoxia tolerance at elevated temperatures (McBryan et al., 2013).

The thermal sensitivity of hypoxia tolerance has so far only been studied in a limited number of fish species (Anttila et al., 2015; Chen et al., 2019; He et al., 2015; Jung et al., 2019; McBryan et al., 2016). Moreover, detailed characterizations of mechanisms underlying the thermal sensitivity of hypoxia tolerance have often been hindered by small body sizes precluding, for example, detailed cardiorespiratory measurements and repeated blood sampling. Nonetheless, available evidence suggests that warm acclimation can alleviate some of the detrimental effects of warming on hypoxia tolerance. Indeed, when fish are given time to acclimate to increased temperatures there is generally a reduction in P_crit and P_LOE, and/or an increased time to LOE in constant severe hypoxia (Anttila et al., 2015; Chen et al., 2019; He et al., 2015; McBryan et al., 2016). The mechanisms behind these responses are likely to be multi-faceted and may involve several plastic changes along the O₂ transport cascade. While warm acclimation can increase the gill surface area to improve branchial O₂ diffusion (Chen et al., 2019; McBryan et al., 2016), and increase anaerobic capacity (He et al., 2015), perhaps the most important adjustment is that the thermal sensitivity of aerobic metabolic processes is reduced, causing a reduction in animal O₂ demand (i.e. thermal compensation of Ṁ_O2; Seebacher et al., 2015).

In the present study, we examined thermal effects on hypoxia tolerance in European perch (Perca fluviatilis) subjected to long-term warming in the Biotest enclosure: a unique semi-natural ecosystem in the Baltic Sea that has been artificially warmed by ∼5–10°C for the past 40 years by cooling water effluents from a nuclear power plant (Huss et al., 2019; Sandblom et al., 2016a). This has significant relevance for this species in nature, as periods of days to months of severe (Pw_O2<2 mg l⁻¹) and moderate (Pw_O2=2–4 mg l⁻¹) hypoxia are well documented in shallow coastal Baltic Sea areas inhabited by perch (Conley et al., 2011). We recently showed that Biotest perch (field-acclimated to 23°C in summer) exhibit pronounced metabolic adjustments as indicated by significantly lower resting Ṁ_O2 relative to reference perch from an adjacent area (field-acclimated to 18°C) that were acutely warmed to 23°C (Sandblom et al., 2016a). This thermal compensation of Ṁ_O2 was associated with reduced cardiac output driven by depressed heart rate, the latter being a combined effect of reduced intrinsic cardiac pacemaker rate and increased autonomic inhibitory cholinergic (i.e. vagal) tone on the heart (Sandblom et al., 2016a,b). It is currently unknown, however, whether such thermal adjustments of cardiac autonomic control affect cardiovascular reflex responses to hypoxia (e.g. hypoxic bradycardia; Farrell, 2007). Moreover, very few studies have evaluated in detail how chronic warming affects integrated cardiorespiratory responses to hypoxia in fish, and how that is related to overall hypoxia tolerance.

To address these questions, we examined perch from the Biotest enclosure (∼23°C) and the adjacent reference area (∼16–18°C) to assess routine cardiovascular variables and Ṁ_O2 (Ṁ_O2,routine; Chabot et al., 2016), and define the influence of temperature on hypoxia tolerance (i.e. P_LOE) and cardiovascular performance. Specifically, we examined responses to acute hypoxia in perch from the two populations at their respective acclimation temperatures, as well as in reference perch acutely warmed to the Biotest temperature (∼23°C). We hypothesized that acutely warmed reference perch would exhibit elevated cardiorespiratory rates and reduced hypoxia tolerance relative to reference fish at the cooler temperature, while metabolic and cardiorespiratory thermal compensation in the chronically warmed Biotest perch would result in lower cardiorespiratory rates and mitigation of the impaired hypoxia tolerance expected to occur with acute warming. Perch are known to be O₂ regulators with a relatively low P_crit (Bilberg et al., 2010; Thuy et al., 2010). If Pw_O2 remains above P_crit, perch appear to be able to withstand prolonged periods of hypoxia (Horoszewicz, 1973); however, once Pw_O2 falls below P_crit, survival is very time limited (Alabaster et al., 1957). As it appears that hypoxia tolerance is reliant on the ability to maintain aerobic metabolism at low Pw_O2 (i.e. a low P_crit), we hypothesized that any thermal compensation of hypoxia tolerance during chronic warming would be linked to thermal compensation of P_crit.

We conducted this study in August and September 2018 at the Biotest enclosure, which is a man-made ∼1 km² artificial basin located in the Baltic Sea archipelago close to Forsmark, Sweden (60°25′41.5″N, 18°11′20.8″E) (for details, see Sandblom et al., 2016a; Huss et al., 2019). The Biotest enclosure receives heated cooling water from two reactors at the Forsmark nuclear power plant. Relative to the temperature of the surrounding Baltic Sea archipelago, this elevates the average water temperature by ∼5°C (summer) to ∼10°C (winter) (Huss et al., 2019) (see Fig. S1). The semi-natural ecosystem in the Biotest enclosure holds several species of teleost fish, including European perch (Perca fluviatilis, Linnaeus 1758) (Huss et al., 2019; Sandström et al., 1995).

Perch of mixed sex and age were caught by hook and line from sites close to the cooling water intake channel upstream of the power plant where the temperature is normal (reference population, ∼16°C), and from the Biotest enclosure (‘Biotest population’, ∼23°C; see Table 1 for fish size metrics). We then transferred the perch to 1200 l holding tanks continuously supplied with aerated seawater (∼5 ppt salinity) from the Baltic Sea or the Biotest enclosure. Thus, the temperature in the holding tanks closely matched the environmental temperature at the respective fish collection sites (see Fig. S1; Table 1). A 6 kW water heater (VB 6010L, Värmebaronen, Österslöv, Sweden) was used to adjust and maintain the temperature in the holding tanks, which were partly covered and located outdoors under a natural photoperiod. We acclimated the fish to the holding tanks for a minimum of 3 days prior to experimentation, during which the fish were not fed. All experimental procedures were approved by the animal ethics committee in Gothenburg (ethical permit no. 165-2015).

Perch were anesthetized in aerated Baltic Sea water containing tricaine methanesulfonate (MS-222, 100 mg l⁻¹), and placed left-side up on wet foam on a surgery table. During surgery, the gills were continuously irrigated with aerated Baltic seawater (10°C) containing MS-222 (50 mg l⁻¹). The fish were then surgically instrumented with a catheter in the third afferent branchial artery to record ventral aortic blood pressure, and a Transonic blood flow probe (Transonic Systems Inc., Ithaca, NY, USA) fitted around the aorta to record cardiac output as described in detail by Ekström et al. (2016a). Care was taken not to damage the pericardium or surrounding vessels and nerves during this procedure.

The fish were then placed in one of four cylindrical transparent Perspex respirometers (3.1 l) submerged in a 160 l experimental tank continuously supplied with water at the reference or Biotest temperature (see below for details), and allowed at least 24 h of post-surgical recovery. The tank was covered with opaque drapes and the fish were monitored via a camera to minimize external disturbances.

We assessed responses to hypoxia in reference (n=15) and chronically warmed Biotest (n=15) fish at their respective environmental acclimation temperatures (i.e. 17.8±0.2°C and 23.4±0.4°C, respectively), as well as in reference fish acutely exposed (24–28 h prior to the start of the protocol) to the Biotest temperature (reference acutely warmed, n=16, 22.8–23.9°C) (see Table 1 for fish size metrics). We examined a maximum of four fish each day, and the order of experimental treatment groups was randomized.

At the start of the experimental protocol, we first took a resting blood sample (∼0.3 ml) in normoxia (90–100% air saturation) from the catheter using a heparinized syringe. The collected blood volume represented less than 4% of the total blood volume (assuming a total blood volume of ∼3% of body mass) and was replaced with 0.3 ml saline (0.9% NaCl). The fish were then connected to the recording equipment, and routine cardiorespiratory variables (see below for details) in normoxia were recorded for at least 2 h. We initiated hypoxia exposure by ceasing the water flow to the respirometers, whereupon the O₂ consumption of the animals caused a gradual decline in Pw_O2 in the respirometers. We allowed the decline in Pw_O2 to continue to the point at which the fish could no longer maintain an upright body position for more than 10 s, which we here define as the Pw_O2 at LOE (P_LOE). The time from the start of the protocol to P_crit and P_LOE was determined in each individual fish. From the onset of the hypoxia exposure, reference perch, chronically warmed Biotest perch and acutely warmed reference perch reached P_crit within 33, 17 and 14 min, respectively, and P_LOE within 52, 35 and 28 min, respectively (see Fig. S2). This meant that the average hypoxia induction rate during the entire hypoxia protocol was 0.33, 0.49 and 0.59 kPa min⁻¹ for reference perch, chronically warmed Biotest perch and acutely warmed reference perch, respectively.

Immediately following P_LOE, we collected a second blood sample, and the fish were subsequently rapidly removed from the experimental chamber and euthanized by a cranial blow. If a final blood sample could not be collected from the catheter, we collected a blood sample from the euthanized fish by drawing blood from the caudal vessels using a heparinized syringe. This procedure was employed in five reference fish, four acutely warmed reference fish and five chronically warmed Biotest fish. Following termination of the experiments, the spleen and ventricle were dissected and their mass was determined. The respirometers and experimental tank were thoroughly cleaned, and the entire water volume within the experimental setup was replaced between each experimental run.

The ventral aortic blood pressure and cardiac output were recorded as described by Ekström et al. (2016a). The Pw_O2 in the respirometers was recorded using Firesting O₂ optodes connected to a Firesting fiberoptic O₂ meter (Pyroscience, Aachen, Germany), and sensors were calibrated by two-point calibration (i.e. at 0% air saturation in water saturated with sodium sulfite and at 100% air saturation in water vigorously bubbled with air). The water temperature was recorded continuously using a custom-built temperature logger (EW 7221, Crn Tecnopart, Barcelona, Spain). Analog signals from the recording equipment were relayed to a PowerLab system (ADInstruments, Sydney, Australia) connected to a laptop computer with LabChart Pro software (version 7.3.8; ADInstruments, Bella Vista, Australia).

We determined the mass-specific Ṁ_O2 (expressed as mg O₂ h⁻¹ kg⁻¹) using intermittent-closed respirometry and calculated the Ṁ_O2 from the decline in Pw_O2 inside the respirometer according to Clark et al. (2013). In normoxia (i.e. 90–100% air saturation), Ṁ_O2,routine was first determined by recording Ṁ_O2 slopes for at least 2 h where the decline in Pw_O2 inside the respirometer was recorded for 5 min every 15 min between automated 10 min respirometer flush cycles. This cycle duration ensured that the Pw_O2 inside the respirometers did not go below 80% air saturation when the respirometers were closed and the fish remained in a quiescent state. We calculated an average Ṁ_O2 during this period. During the hypoxia protocol, when the respirometer remained closed, Ṁ_O2,routine was assessed continuously by calculating 1 min slopes until the point at which each individual fish reached P_LOE. We then averaged Ṁ_O2,routine values spanning ∼5% air saturation sections at 90%, 80%, 70%, 60%, 50%, 45%, 40%, 35%, 30%, 25%, 20%, 15%, 10% and 5% air saturation (e.g. the average Ṁ_O2,routine at 90% air saturation represents the average value between ∼92.5% and 87.5% air saturation). We accounted for the background microbial respiration by subtracting the background Ṁ_O2 slopes measured in empty respirometers after each experimental run.

We determined the P_crit for each individual using a segmental linear regression analysis in GraphPad prism for Windows (version 8.0.2, San Diego, CA, USA), where the Ṁ_O2,routine, determined from consecutive 1 min declines in O₂ during the hypoxia protocol, was plotted against Pw_O2 to identify the intersection of two linear regression lines fitted to the data (Yeager and Ultsch, 1989) (Fig. S3). Similarly, segmental linear regression analyses were used to determine the Pw_O2 where cardiac output and heart rate started to progressively decline in hypoxia (i.e. P_{cardiac output} and P_{heart rate}), using the average values determined at the fixed Pw_O2 steps as described above. In these analyses, we excluded individual data points from fish that displayed transient activity inside the respirometer and were associated with elevated cardiorespiratory activity.

We used linear mixed models to assess the effects of hypoxia exposure on cardiorespiratory variables in perch from the different temperature regimes. Individuals were set as subject variables and the fixed Pw_O2 steps during hypoxia as the repeated variable, excluding the values at P_LOE. As fixed effects, the model included Pw_O2, experimental group (Exp.group; i.e. reference perch, chronically warmed Biotest perch or acutely warmed reference perch) and their interaction (Pw_O2×Exp.group). We found that a first-order autoregressive covariance structure provided the best fit to these data. We further explored significant main and interaction effects with either general (across the entire hypoxia exposure) or specific (i.e. at each Pw_O2 level) pair-wise comparisons of the responses between experimental groups. Significant main effects of the hypoxia exposure or interaction effects were further examined by comparing the values during hypoxia with the initial values at the onset of hypoxia exposure (i.e. 90% air saturation) across or within groups.

For all other comparisons among groups, we performed one-way ANOVA or a Welch's test for data displaying heterogeneous variances (i.e. length, body mass, P_heart rate and tissue O₂ extraction), and significant main effects were assessed by pair-wise comparisons. We used paired two-tailed t-tests to compare hematological variables in normoxia and after hypoxia. We transformed (log₁₀) variables that failed to meet the assumption of normality (Shapiro–Wilks test) or the assumption of homogeneity of variance (Levene's test). All P-values were Bonferroni adjusted for multiple testing, and statistical significance was accepted at P≤0.05. Statistical analyses were performed using SPSS (v.25, SPSS Inc., Chicago, IL, USA). We present all data as means±s.e.m.

There was no difference in body mass, length, condition factor or relative spleen mass among experimental groups (Table 1). However, chronically warmed Biotest perch had a lower relative ventricle mass compared with both reference (P=0.018) and acutely warmed reference perch (P=0.003; Table 1).

Acute warming from 18 to 23°C in normoxic reference perch resulted in increased Ṁ_O2,routine (105 to 163 mg O₂ kg⁻¹ h⁻¹, Q₁₀: 2.2, F₂=37.4, P<0.001; Fig. 1A), cardiac output (from 32 to 47 ml min⁻¹ kg⁻¹, Q₁₀: 1.7, F₂=6.1, P=0.042; Fig. 2A) and heart rate (from 60 to 101 beats min⁻¹, Q₁₀: 2.6, F₂=28.0, P<0.001; Fig. 2B), whereas stroke volume did not change (0.58 versus 0.47 ml beat⁻¹ kg⁻¹; Fig. 2C).

In the chronically warmed Biotest perch, considerable metabolic and cardiovascular thermal compensation was evident because Ṁ_O2,routine was only 134 mg O₂ kg⁻¹ h⁻¹, which was significantly lower than in the acutely warmed reference fish (P<0.001). However, the metabolic compensation was not complete as Ṁ_O2,routine was still significantly elevated relative to that of reference perch at the lower temperature (Q₁₀: 1.5, P=0.001; Fig. 1A).

The metabolic adjustments in chronically warmed Biotest perch were also reflected in pronounced thermal compensation of cardiovascular function. Routine cardiac output in Biotest perch (32 ml min⁻¹ kg⁻¹) was similar to that of reference fish measured at the lower temperature (Q₁₀: 0.8; Fig. 2A). This was a combined effect of significantly lower stroke volume (0.39 versus 0.47 ml beat⁻¹ kg⁻¹, F₂=6.7, P=0.002; Fig. 2C) and depressed heart rate relative to the acutely warmed reference fish (84 versus 101 beats min⁻¹, F₂=28.0, P<0.001; Fig. 2B), which meant that Q₁₀ for heart rate in Biotest perch was reduced to 1.6. There were no differences in ventral aortic blood pressure, cardiac power output or total vascular resistance among treatment groups in normoxia (Fig. 3). While tissue O₂ extraction was not different in normoxia (Fig. 2D), the chronically warmed Biotest perch had a higher blood oxygen carrying capacity as indicated by elevated Hct and [Hb] compared with reference perch (F₂=3.9, P=0.033 and F₂=4.8, P=0.016, respectively), but there was no difference in MCHC (Fig. 4). Acutely warmed reference perch showed an intermediate hematological response and did not differ significantly from either of the other groups (Fig. 4).

The patterns and differences in cardiorespiratory traits observed in normoxia were generally maintained over the higher Pw_O2 range of the hypoxia exposure (Figs 1A and 2A–C). Ṁ_O2,routine was significantly higher in chronically warmed Biotest perch compared with reference perch until 35% air saturation, and significantly lower than in acutely warmed reference perch until 30% air saturation. Importantly, however, the metabolic thermal compensation in Biotest perch was associated with an improved ability to maintain Ṁ_O2,routine in hypoxia compared with the acutely warmed reference perch. Thus, P_crit in Biotest perch was intermediate relative to that of acutely warmed reference perch (P=0.022) and reference perch (P=0.08, F₂=14.2, P<0.001; Fig. 1B). However, this pattern was only partially reflected in the absolute hypoxia tolerance because Biotest perch had a significantly lower P_LOE than both the reference perch (P=0.045) and the acutely warmed reference perch (P<0.001), while there was no difference in P_LOE between the two reference groups (Fig. 1C).

The thermal compensation of P_crit in Biotest perch was associated with an improved ability to maintain cardiovascular function in hypoxia as the decline in heart rate started at a significantly lower Pw_O2 (i.e. lower P_heart rate) compared with reference and acutely warmed reference perch (Welch_2.18=14.9, P<0.001; P=0.01 and P=0.01, respectively; Fig. 2E), and there was a tendency for a lower P_{cardiac output} compared with both reference (P=0.058) and acutely warmed reference perch (P=0.153; Fig. 2E). Even so, acute warming had no significant effect on P_{cardiac output} and P_heart rate in reference perch (Fig. 2E). If anything, there was also a tendency towards higher tissue O₂ extraction in the chronically warmed Biotest perch during hypoxia compared with both reference groups (P=0.06; Fig. 2D).

The decline in cardiac output in hypoxia were largely driven by reductions in heart rate because P_heart rate always coincided with (Biotest perch) or preceded P_{cardiac output} (both reference groups; Fig. 2E). While stroke volume generally increased from 35% air saturation to lower Pw_O2 across all groups, this increase did not fully compensate for the reduced heart rate (Fig. 2A–C). Nonetheless, the extent to which heart rate declined in absolute terms during hypoxia (i.e. hypoxic bradycardia) differed significantly among groups (Fig. S4A,B). The most pronounced bradycardia occurred in the acutely warmed reference perch (63 beats min⁻¹, 64% reduction), followed by chronically warmed Biotest perch (53 beats min⁻¹, 63% reduction) and, last, reference perch (40 beats min⁻¹, 65% reduction). Thus, the hypoxic bradycardia in relative terms was similar (Fig. S4B).

There were no differences among experimental groups in cardiac power output in hypoxia, which declined significantly from 40% air saturation in all groups (Fig. 3A). This was due to the reduced cardiac output because the ventral aortic blood pressure remained unchanged across intermediate Pw_O2 levels in all experimental groups (Fig. 3B). There was a general tendency (P=0.06) for differences amongst groups in vascular resistance, which increased across experimental groups from approximately 45% air saturation.

At the lowest Pw_O2 levels immediately preceding P_LOE, the ventral aortic blood pressure remained stable in chronically warmed Biotest perch, while both reference groups showed tendencies for a declining blood pressure.

After the hypoxia exposure, there were no significant differences among treatment groups for any of the hematological variables, but all groups showed significantly increased Hct relative to normoxic values (T=−10.5, P<0.001; T=−6.1, P<0.001; and T=−3.7, P=0.006 for reference, chronically warmed Biotest and acutely warmed reference, respectively). This was probably due to red blood cell swelling as MCHC was reduced across all experimental groups (T=4.5, P=0.003; T=8.5, P<0.001; and T=7.6, P<0.001, for reference, chronically warmed Biotest and acutely warmed reference perch, respectively; Fig. 4C), while [Hb] did not consistently change in hypoxia (Fig. 4B).

We show that European perch exposed to chronic warming (possibly across lifetime or generations) in an artificially heated ecosystem display considerable physiological thermal compensation. Relative to acutely heated reference perch, this was primarily reflected in depressed Ṁ_O2,routine along with improved capacity to maintain cardiovascular function and aerobic metabolism in hypoxia (i.e. reduced P_crit). The thermal compensation involved multiple changes along the oxygen transport cascade including elevated blood O₂ carrying capacity, reduced resting heart rate and cardiac output, as well as indications of increased tissue oxygen extraction. Interestingly, while chronic warming only elicited a partial thermal compensation of P_crit, Biotest perch had superior absolute hypoxia tolerance (i.e. the lowest P_LOE) relative to both groups of reference perch, irrespective of thermal regimes. This reveals a disconnect between P_crit and absolute hypoxia tolerance as measured by P_LOE across environmental temperatures, suggesting that factors beyond thermal compensation of Ṁ_O2,routine and P_crit are important for safeguarding hypoxia tolerance with environmental warming. Taken together, our results suggest that European perch, and possibly other eurythermal fish species, may be relatively robust to transient hypoxic challenges in a warmer future.

The thermal compensation of Ṁ_O2,routine in normoxic Biotest perch (Q₁₀: 1.5) is identical to previous observations by Sandblom et al. (2016a), and was associated with a reduced cardiac output (Q₁₀: 0.8) resulting from lower stroke volume and smaller relative ventricle mass, along with a moderate re-setting of routine heart rate (Q₁₀: 1.8). This last response is also consistent with previous findings and probably reflects an increased inhibitory cholinergic (i.e. vagal) tone on the heart in Biotest perch (Sandblom et al., 2016a,b). Although routine cardiac output was fully compensated in Biotest perch, they still had to sustain a slightly higher Ṁ_O2,routine. This was achieved through elevated blood O₂ carrying capacity (i.e. higher Hct and [Hb]) and probably elevated tissue O₂ extraction.

Two previous studies have estimated P_crit in European perch. Thuy et al. (2010) and Bilberg et al. (2010) reported P_crit values of 3.4 kPa (20°C) and 4.4 kPa (16°C), respectively, in hatchery reared perch. The higher P_crit reported here for reference perch under similar thermal conditions (6.9 kPa at ∼18°C), in part probably reflects that we defined P_crit as the breakpoint in Ṁ_O2,routine, while the previous studies estimated P_crit as the Pw_O2 where Ṁ_O2 declined below basal or standard Ṁ_O2. Another contributing factor may have been that surgical instrumentation in the current study elevated Ṁ_O2,routine (previous perch P_crit measures were performed on uninstrumented fish). Indeed, the resting Ṁ_O2 of perch from the same populations as the current study were lower in a previous study using uninstrumented fish (Sandblom et al., 2016a). Thuy et al. (2010) showed a significant positive correlation between P_crit and resting Ṁ_O2 in perch, so it is entirely plausible that instrumentation and a corresponding elevated Ṁ_O2,routine led to the elevated P_crit estimates observed here. Regardless, our experimental protocol provides a robust comparative approach to determine the influence of warming across different time scales on hypoxia tolerance, while combining assessment of hypoxia tolerance and in vivo cardiorespiratory responses in individual fish.

Acute warming increased P_crit in reference perch, which is consistent with findings in other fish species (Borowiec et al., 2016; Butler and Taylor, 1975; Corkum and Gamperl, 2009; Ern et al., 2016; Fry and Hart, 1948; Gehrke and Fielder, 1988; Nilsson et al., 2010; Rogers et al., 2016). The reduced P_crit in Biotest perch relative to the acutely warmed reference perch is consistent with our hypothesis that, through various metabolic and cardiorespiratory adjustments, fish subjected to long-term environmental warming have the capacity to alleviate at least some of the detrimental impacts of warming on the ability to maintain aerobic metabolism during acute hypoxia exposure. It is likely that the partial thermal compensation of P_crit is related to the partial thermal compensation of Ṁ_O2,routine in Biotest perch, because the normoxic aerobic metabolic rate is well known to be positively correlated with P_crit in perch (Thuy et al., 2010) and other fish species (McBryan et al., 2016; Nilsson et al., 2010; Rogers et al., 2016). Assuming that the hematological differences among groups in normoxia were also present around P_crit, it is also possible that an improved blood O₂ carrying capacity contributed to the lowering of P_crit in the chronically warmed Biotest perch.

Regardless of temperature, P_heart rate and P_{cardiac output} always occurred at higher Pw_O2 than P_crit, supporting the idea that the bradycardia and ensuing decline in cardiac output in hypoxia relate to the onset of P_crit as reported in other fish species (Gehrke and Fielder, 1988; McKenzie et al., 2009; Rantin et al., 1993; Speers-Roesch et al., 2010). However, reference perch at the lower temperature were better able to sustain Ṁ_O2,routine at Pw_O2 values below P_{cardiac output}, which was probably mediated by an increase in tissue O₂ extraction as hypoxia developed. By contrast, tissue O₂ extraction was already elevated in normoxia in the acutely warmed reference perch, which probably left less scope for compensation through increased extraction in hypoxia and a more rapid decline in Ṁ_O2,routine as heart rate and cardiac output started to plummet.

Biotest perch had an improved ability to maintain cardiac output in hypoxia, as P_heart rate and P_{cardiac output} occurred at lower Pw_O2, whereas these breakpoints occurred at very similar Pw_O2 in reference perch at the two test temperatures. These observations could indicate that hypoxic bradycardia in perch is elicited by internally oriented chemoreceptors located in one or several gill arches (Milsom, 2012; Reid and Perry, 2003), which are stimulated by reduced partial pressure of O₂ in the venous blood (Pv_O2) as hypoxia progresses (Perry and Reid, 1994; Steffensen and Farrell, 1998). Thus, the lower P_{heart rate} in Biotest perch could be attributed to the fact that chronic warming elevates Pv_O2 relative to reference perch (Ekström et al., 2016a), and therefore postpones the decline in Pv_O2 and the corresponding stimulation of branchial chemoreceptors during hypoxia. This would cause the bradycardia to occur at a lower ambient Pw_O2 relative to that in reference perch. Moreover, it was recently shown that acute warming does not significantly lower Pv_O2 in reference perch over the temperature interval tested here (i.e. 18–23°C) (Ekström et al., 2016a), which could explain why the bradycardia occurred at a similar Pw_O2 in reference perch across test temperatures.

The increased vagal tone assumed to be associated with the thermal compensation of routine heart rate (Ekström et al., 2016b) did not interfere with the capacity to induce a vagally mediated hypoxic bradycardia in perch. The consequent depression of cardiac power output and thus cardiac O₂ demand, which was similar among experimental groups, could have aided prevention of a mismatch between cardiac O₂ supply and demand, and could thus explain how the perch heart was able to maintain an elevated stroke volume in severe hypoxia (Farrell and Stecyk, 2007; Speers-Roesch et al., 2010; Stecyk, 2017). Even while cardiac output progressively declined with bradycardia, compensatory elevations in total vascular resistance resulted in the maintenance of ventral aortic blood pressure at diminishing O₂ levels beyond P_crit across groups.

Despite a substantially elevated P_crit, acute warming did not affect P_LOE in reference perch, which contrasts with reports in other fish species (Borowiec et al., 2016). The lack of an effect of acute warming on P_LOE is enigmatic, but may be associated with an increased transcription of heat shock proteins during the ∼20 h acute heat exposure. For example, the induction of heat shock proteins following an acute heat shock may have contributed to an improved ability to survive subsequent hypoxia in an intertidal fish (Oligocottus maculosus) (Todgham et al., 2005). The maintenance of P_LOE in acutely warmed perch may also reflect an improved anaerobic capacity, which has been identified as a driver of hypoxia tolerance in other fish species (Borowiec et al., 2016; Mandic and Regan, 2018; McArley et al., 2019; Richards et al., 2009). Indeed, acute warming may elevate the expression and/or the activity of the anaerobic glycolytic enzyme lactate dehydrogenase in fish, including perch (Ekström et al., 2016c; Iftikar et al., 2014).

As measured by P_LOE, chronically warmed Biotest perch not only maintained a greater level of hypoxia tolerance than the acutely warmed perch, which was expected, but also maintained a greater level of hypoxia tolerance than reference perch at the lower test temperature. Thus, if P_LOE is taken as a reliable measure of absolute hypoxia tolerance (Wood, 2018), then perch actually became more hypoxia tolerant with chronic warming. Similar to the lack of difference in P_LOE between acutely warmed and reference perch despite a pronounced difference in P_crit, the improved hypoxia tolerance (lower P_LOE) of Biotest perch relative to reference perch (18°C) was not impeded by a higher P_crit. This is not to say that thermal compensation of Ṁ_O2,routine and P_crit is unimportant for safeguarding hypoxia tolerance with chronic warming, only that the lack of a direct association between P_crit and P_LOE across treatment groups demonstrates that other factors make an equal or more important contribution to hypoxia tolerance. This disconnect between P_crit and absolute hypoxia tolerance as measured by LOE is not unique to the current study, as this has been reported in several other fish species (Borowiec et al., 2020; Mandic et al., 2013; McArley et al., 2020; Speers-Roesch et al., 2013).This points to the fact that the ability to maintain aerobic ATP production at low Pw_O2 as measured by P_crit, while undoubtedly important, is only one part of a species' total hypoxia tolerance strategy, and that P_crit does not necessarily reflect absolute hypoxia tolerance (Mandic et al., 2013; Wood, 2018). In regards to the current study, the mechanisms by which perch are able to improve P_LOE with chronic warming despite a higher P_crit remain to be elucidated. However, one possibility could be that prolonged warming stimulates an enhanced capacity for anaerobic ATP production (e.g. through increased anaerobic fuel stores such as glycogen and/or enhanced activity and expression of glycolytic enzymes) (Mandic and Regan, 2018; Mandic et al., 2013; Speers-Roesch et al., 2013). There may also be an important behavioral component of hypoxia tolerance in perch. During 4 h exposure to severe hypoxia (∼23% air saturation at 23°C), which was survivable, perch were reported to rest in a stationary position on the bottom of tanks (Douxfils et al., 2012). In the same study, despite the level of hypoxia exposure being substantially below the P_crit of Biotest perch (39% air saturation at 23°C) measured here, there was no evidence of lactate accumulation in white muscle following hypoxia exposure, suggesting anaerobic metabolism was not recruited. The quiescent behavior of the fish was probably important for reducing energetic demands, so that aerobic ATP production remained sufficient throughout hypoxia exposure. If there was an unmeasured difference in the behavioral response to hypoxia among the treatments in the current study, this could also have contributed to the lack of correlation between P_crit and P_LOE. Moreover, we recently showed that Biotest perch exhibited higher expression of mitochondrial genes nd4 and cox1, which encode subunits for complex I and IV, respectively, in the mitochondrial respiration chain, which is indicative of molecular adaptation in the Biotest perch (Pichaud et al., 2020). It is therefore possible that the differences between Biotest and reference perch in terms of their tolerance to hypoxia may be attributable to some, as yet unknown, genetically adapted trait(s) in the Biotest perch.

A legitimate criticism of the method used in the current study to assess P_crit and P_LOE is that the rate of hypoxia induction differed between treatment groups. This was the result of using closed system respirometry to generate hypoxia: a higher Ṁ_O2 in the warm treatments drove a faster rate of O₂ decline in the respirometer (0.58, 0.49 and 0.33 kPa min⁻¹ in the acutely warmed, Biotest and reference treatment groups, respectively). There was no evidence, however, of a faster rate of hypoxia induction relating to improved hypoxia tolerance (i.e. there was no correlation between the rate of hypoxia induction and P_LOE within treatment groups; Fig. S5). Thus, while we cannot rule out different findings using a common rate of hypoxia induction (i.e. through bubbling nitrogen to generate hypoxia), the unexpectedly low P_LOE of the acutely warmed reference and Biotest perch is not explained by the fact that these groups underwent a faster rate of hypoxia induction. Another potential caveat worth mentioning is that the closed respirometry approach to eliciting hypoxia employed here may have, to different extents, induced accumulation of metabolic by-products (e.g. CO₂) inside the respirometers (see review by Rogers et al., 2016). Whether or not this occurred here, or whether this affected the outcome of the current study is, however, unknown.

The hypoxia exposure used in the current study was highly acute in all treatment groups (total duration of 52, 34 and 28 min in the reference, Biotest and acutely warmed reference groups, respectively; Fig. S2). Such fast development of inescapable hypoxia would be unlikely to occur in the natural habitats of perch, if in any habitats occupied by fish. Even in shallow vegetated lakes and intertidal rock pools, both habitats with pronounced diel O₂ cycles, night-time hypoxia develops progressively over several hours (e.g. Andersen et al., 2017; McArley et al., 2019). Where tolerance to severe, acute hypoxia may be relevant to perch is when juveniles move into deeper water pelagic hypoxic zones to feed and avoid predation – a behavior known to occur in this species (Vejřík et al., 2016). Of course, when utilizing pelagic hypoxic zones, juveniles are not trapped and can migrate vertically to avoid hypoxia if it becomes too severe. Indeed, avoidance of hypoxia – moving from hypoxia to areas of higher Pw_O2 – has been demonstrated in perch (Alabaster and Robertson, 1961). Nevertheless, if the resilience to acute hypoxia following chronic warming in adult perch observed here is also present in juvenile perch, then this behavior may be robust to climate warming. Other than severe acute hypoxia exposure, perch could also face longer periods of severe to moderate hypoxia. In the Baltic Sea, periods of days to months of severe (Pw_O2<2 mg l⁻¹) and moderate (Pw_O2=2–4 mg l⁻¹) hypoxia are well documented in shallow coastal habitats important to perch (Conley et al., 2011). If such hypoxia occurs over widespread areas so that behavioral avoidance is prevented, then perch may have to tolerate prolonged episodic hypoxia exposure at warmer temperatures under climate change. We do not currently know what levels of hypoxia are survivable for Baltic Sea perch over prolonged time periods and how warming impacts tolerance to hypoxia of this nature. Furthermore, even when prolonged hypoxia is survivable, there may be important negative consequences for animal performance (e.g. growth, swimming and reproduction), which could be worsened by warming. Thus, developing a more complete picture of the resilience or susceptibility of Baltic Sea perch populations to hypoxia under climate warming will require examining responses to a range of ecologically relevant hypoxia regimes. Despite these caveats and the continued threat of worsening hypoxia due to climate change, the findings of the current study point to cautious optimism that climate warming may have somewhat less dire consequences for absolute tolerance to severe, acute hypoxia in perch and other eurythermal fishes than previous prognoses have predicted.

We extend our gratitude to the anonymous reviewers who provided us with valuable and insightful feedback on an earlier version of our manuscript.