Preferential intracellular pH regulation is a common trait amongst fishes exposed to high environmental CO₂

Freshwater environments are often characterized by high CO₂ (hypercarbia) (Cole et al., 1994; Hasler et al., 2016; Park et al., 1969; Raymond et al., 2013). Hypercarbia is often associated with various behavioral and physiological effects in fishes owing to the associated acidosis (Heuer and Grosell, 2014; Heuer et al., 2019; Ou et al., 2015; Schneider et al., 2018). Hypercarbia tolerance may have been selected for as fishes expanded into new hypercarbic habitats, which are particularly common in tropical systems (Heisler, 1984; Li et al., 2013; Rasera et al., 2013; Scofield et al., 2016). Aquatic environments with high CO₂ do not seem to have restricted the proliferation of fishes, as these environments are often very speciose (De Pinna, 2005).

Aquatic hypercarbia can arise by a number of factors, including presence of fishes and herbivores (Atwood et al., 2013, 2014), percolation of source water through carbonate deposits (e.g. limestone; Wetzel, 2002), microbial respiration (Heisler, 1984) and organic decay (Furch and Junk, 1997). The CO₂ partial pressure (P_CO2) in many aquatic water bodies can be significantly higher than the atmospheric P_CO2 of 0.04 kPa (409 ppm; NOAA September 2019, https://www.esrl.noaa.gov/gmd/ccgg/trends/full.html). The global average P_CO2 of stream and river systems is ca. 0.3 kPa (∼3100 μatm), approximately 8-fold above current atmospheric levels (Raymond et al., 2013). Some aquatic environments experience hypercarbia that greatly exceeds this; for example, P_CO2 values may reach values as high as 2–8 kPa (20,249–79,000 μatm) in tropical systems (Cole et al., 1994; Heisler, 1984; Rasera et al., 2013), 2–4 kPa in aquaculture ponds (Damsgaard et al., 2015), 0.5–2 kPa (10–40 mg l⁻¹) in recirculating aquaculture systems (Fivelstad, 2013; Mota et al., 2019), and 5−11 kPa (100–200 mg l⁻¹) where CO₂ is used as a barrier for invasive species (Kates et al., 2012). However, these environments are often heterogeneous, where fishes can experience large changes in CO₂ levels spatially or temporally. Thus, fishes that reside in these systems must be able to respond to these changes in CO₂ rapidly and effectively.

When exposed to severe acute hypercarbia, fishes may experience a large reduction in blood and intracellular tissue pH (Brauner and Baker, 2009; Brauner et al., 2019). As pH changes alter protein structure and function, and ultimately will affect whole-animal performance, acid–base regulation is one of the most tightly regulated physiological systems (Occhipinti and Boron, 2015). Consequently, hypercarbic freshwater environments represent an important system to investigate adaptations in acid–base regulation in fishes, and indeed, hypercarbia has been proposed to be an important driver not only of the evolution of air breathing in fishes, but also in the transition of vertebrates from water to land (Brauner and Baker, 2009; Ultsch, 1987).

Initial exposure to hypercarbia acidifies the blood, and the most common response observed in fishes is that extracellular blood pH (pH_e) is compensated by a net increase in plasma [HCO₃⁻] in exchange for Cl⁻ over the following 24–72 h, with the gills playing the primary role in compensation (Brauner and Baker, 2009). Changes in pH_e are often associated with qualitatively similar changes in intracellular pH (pH_i), as the two have generally been shown to be tightly coupled, referred to as ‘coupled pH regulation’ (Fig. 1A,C; Shartau et al., 2016a). Although pH_i recovery is more rapid than pH_e recovery, complete pH_i recovery takes hours to days and is usually associated with a significant degree of pH_e recovery (Baker et al., 2015; Larsen et al., 1997; Shartau et al., 2016a, 2019; Wood and LeMoigne, 1991) in coupled pH regulation. In some fishes, the increase in plasma [HCO₃⁻] (in exchange for Cl⁻) associated with pH_e recovery during acute hypercarbia appears to be limited to approximately 40 mmol l⁻¹, which corresponds with complete pH_e compensation at a P_CO2 of approximately 2 kPa (Shartau et al., 2019).

Importantly, some groups of fishes are capable of tolerating severe acute hypercarbia (>3–6 kPa P_CO2), well in excess of this P_CO2 value (i.e. ca. 2 kPa P_CO2) without achieving complete pH_e recovery. Complete pH_e compensation at 6 kPa P_CO2 would require an increase of plasma [HCO₃⁻] of ca. 100–140 mmol l⁻¹, which is not possible (Shartau et al., 2019). In these fishes, hypercarbia tolerance appears to be associated with the ability to completely and rapidly (minutes to hours) regulate pH_i in tissues such as the heart, brain, liver and muscle, despite a large, often uncompensated, reduction in pH_e. This strategy, termed ‘preferential pH_i regulation’, where pH_i is rapidly and tightly regulated despite a reduction in pH_e (Fig. 1B,C; Brauner et al., 2004; Shartau et al., 2016a), was first documented by Heisler (1982) in the marbled swamp eel (Synbranchus marmoratus) when he stated that ‘this particular strategy of acid–base regulation provides a constant milieu for the intracellular structures and demonstrates the prevalence of intracellular over extracellular acid–base regulation’. He also observed a similar pattern in salamanders exposed to aquatic hypercarbia (Heisler et al., 1982). Later, this strategy of preferential pH_i regulation was observed in armored catfish (Pterygoplichthys pardalis; Brauner et al., 2004), white sturgeon (Acipenser transmontanus; Baker et al., 2009a) and striped catfish (Pangasianodon hypophthalmus; Sackville et al., 2018), and was hypothesized to be associated with acute hypercarbia tolerance (Brauner and Baker, 2009). However, it is not known whether preferential pH_i regulation is restricted to a few specialized fishes, or whether it represents a more common pattern among fishes.

In this study, we address the hypothesis that preferential pH_i regulation is relatively widespread among fishes. To investigate this, we conducted a survey of the presence or absence of preferential pH_i regulation amongst 10 phylogenetically separated fish species, spanning six orders, eight families and 10 genera (Betancur-R et al., 2013) endemic to three continents (North America, South America and Africa). Initial experiments were conducted on a coupled pH regulator, rainbow trout (Oncorhynchus mykiss) and a preferential pH_i regulator, white sturgeon, to determine the rate of P_CO2 increase used to assess acute hypercarbia tolerance in fishes used in the following hypercarbia exposures. Thereafter, 10 fish species were exposed to severe acute hypercarbia ranging from 1.5 to 6 kPa P_CO2, depending upon the acute hypercarbia tolerance, and the strategy for acid–base regulation was examined by terminal sampling. As preferential pH_i regulation was hypothesized to be rare (Brauner et al., 2004), species selected for pH measurements were chosen from primarily hypercarbic-prone environments to increase the chance of finding fishes using this strategy of pH regulation; species from various orders were selected to assess the phylogenetic distribution of preferential pH_i regulation. The results demonstrate the widespread usage of preferential pH_i regulation amongst fishes, revealing that this physiological strategy is common and likely associated with tolerance of a hypercarbia-induced respiratory acidosis in fishes.

In this study, 11 species of fish were used and experiments were conducted as follows. From September 2012 to November 2013, measurements were made on the following species at the University of British Columbia (UBC), Vancouver, BC, Canada: rainbow trout (Oncorhynchus mykiss; 250–400 g) from Miracle Springs Inc. (Mission, BC, Canada) and tilapia hybrid (Oreochromis niloticus×mossambicus×hornorum; ∼300–400 g) from Redfish Ranch (Courtenay, BC, Canada). Experiments with white sturgeon (Acipenser transmontanus; ∼200 g) were conducted at the International Centre for Sturgeon Studies, Vancouver Island University (VIU), Nanaimo, BC, Canada; these fish were progeny of wild-caught brood stock from the Fraser River reared at the facility since 1991.

Experiments were conducted at the Instituto Nacional de Pesquisas da Amazônia (INPA) (Manaus, AM, Brazil) with fishes caught in the Rio Negro near Manaus and transferred to a holding facility at INPA in 2008: tamoatá (Hoplosternum littorale; ∼70 g), matrinxã (Brycon amazonicus; ∼57 g), tambaqui (Colossoma macropomum; ∼61 g) and oscar (Astronotus ocellatus; ∼27 g). In 2013, the following fishes were obtained from local fish farms and transferred to facilities at INPA: tambaqui (∼100 g) and oscar (∼100 g).

Experiments with the following farm-reared species were conducted at the South Farm Aquaculture research facility at Mississippi State University (MSU) (Starkville, MS, USA) in March 2013: American paddlefish (Polyodon spathula; ∼150–400 g), alligator gar (Atractosteus spatula ∼400−1000 g) and channel catfish (Ictalurus punctatus; ∼100–200 g). Experiments with the following species were conducted at the University of North Texas (UNT) (Denton, TX, USA) with fishes caught from nearby lakes/rivers in November 2012: spotted gar (Lepisosteus oculatus; ∼300–700 g) and channel catfish (∼50–200 g).

Typically, animals were kept for at least 2 weeks in tanks prior to experiments under a natural photoperiod with flow-through water at the following temperatures depending on the location (UBC: 12°C for rainbow trout, 22°C for tilapia hybrid; VIU: 12°C; MSU: 18°C; UNT: 26°C; INPA: 28°C). Water used in this study at the various locations depended on the facility where experiments were performed. As we were only interested in the presence or absence of preferential pH_i regulation in these fishes, as opposed to investigating how water quality affects pH regulation, these water quality differences were not of specific concern in the experimental design. Water quality parameters for each site are listed in Table 1.

Fishes were fed to satiety three times per week. Feed was withheld for at least 48 h before the start of experiments.

To determine the rate of CO₂ increase to use in hypercarbia experiments, we first used model species that were previously known to be either a coupled pH regulator (rainbow trout; Wood and LeMoigne, 1991) or a preferential pH_i regulator (white sturgeon; Baker et al., 2009a) to determine the optimal rate of hypercarbia increase for use in subsequent experiments to investigate presence or absence of preferential pH_i regulation.

Fishes were randomly selected from the holding tank and placed in individual black plexiglass boxes (24 liters) with aeration in a re-circulating system (flow rate ∼3 l min⁻¹ per box, 15°C; total water volume of system ∼320 liters) overnight prior to experiments. Fishes were then exposed to progressively increasing levels of hypercarbia at a rate of 1, 2 or 4 kPa h⁻¹. Water P_CO2 was monitored to ensure P_CO2 increased at the desired rate for the duration of exposure using a thermostated (15^oC) Radiometer P_CO2 electrode (E5036) (output, Radiometer PHM 73). Fishes were continuously observed until the desired endpoint was reached – loss of equilibrium (LOE), which indicates the fish's acute CO₂ tolerance. Immediately following this, fishes were transferred to a recovery tank for at least 48 h.

The strategy of acid–base regulation used during exposure to severe acute hypercarbia in fishes was determined at a P_CO2 of 1.5, 3 or 6 kPa. The P_CO2 tension used was based upon an initial CO₂ exposure trial at a rate of 2 kPa h⁻¹. This allowed us to identify the highest P_CO2 test exposure that could be used to observe changes in pH_i during maximal pH_e depression; as pH_i only changes by approximately one-third of pH_e (owing to a lower starting pH_i value and greater tissue buffer value), larger reductions in pH_e allow for a more accurate determination of whether fishes tightly and completely regulate pH_i consistent with preferential pH_i regulation. Once the P_CO2 tension that the fish species of interest could tolerate was determined, we performed the following experiment.

Fishes were acclimated individually for 24 h in black plexiglass boxes in the system described above; this period is sufficient to allow recovery from handling stress in sturgeon (Baker et al., 2005; Barton et al., 2000). Normocarbic fish (n=6–8) were terminally sampled immediately (see below) following this acclimation period (control group). Other fish were then exposed to 3 h hypercarbia (1.5, 3 or 6 kPa P_CO2, depending on the CO₂ trial above). We also examined the response of an extremely hypercarbia-tolerant species, tambaqui, to even more severe hypercarbia of 20 kPa P_CO2 owing to their CO₂ tolerance being extremely high in the CO₂ tolerance trial.

In this study, we were interested in assessing the presence or absence of preferential pH_i regulation in fishes. As such, we specifically chose 3 h as the sampling time point because we wanted to measure pH_i when the reduction in pH_e was at its maximum. At 3 h, all fishes (whether coupled or preferential pH_i regulators) will experience a reduction in pH_e during hypercarbia; this has been demonstrated in several species including white sturgeon (Baker et al., 2009a), Pacific hagfish (Eptatretus stoutii; Baker et al., 2015), armored catfish (Brauner et al., 2004), rainbow trout (Tovey and Brauner, 2018) and striped catfish (Damsgaard et al., 2015). Fishes using coupled pH regulation will experience a reduction in tissue pH_i values at this time point as pH_i recovery takes considerably longer than 3 h (Shartau et al., 2016a). In contrast, fishes using preferential pH_i will exhibit no change or an increase in pH_i values at 3 h of hypercarbia relative to controls despite the large reduction in pH_e at this time point. It should be emphasized that at 3 h of hypercarbia, minimal pH_e recovery would be expected in any fish species, and this protocol was used to assess the ability to rapidly and completely regulate pH_i as predicted by preferential pH_i regulation. This protocol does not allow for the interpretation of pH_e regulatory ability, which was not the focus of this study.

Within the recirculating system, target CO₂ tensions were achieved by bubbling a reservoir tank with preset rates of air and 100% CO₂ using Sierra Instruments mass flow controllers. Water was then pumped from the reservoir tank to a header tank and then gravity-fed into the experimental tanks where water P_CO2 was measured with a P_CO2 electrode to confirm target CO₂ tensions, which was reached within approximately 10 min once the trial started; water O₂ levels remained >80% saturation. Water from the experimental tanks drained into the reservoir tank. Fishes were allowed access to normocarbic air during hypercarbia.

At the time of sampling, each box was isolated from the recirculation and anesthetic (MS-222; tricaine methanesulfonate; 0.3 g l⁻¹ buffered with NaHCO₃) was added to the water while maintaining the respective water P_CO2. Once ventilation ceased (<3 min), each fish was turned ventral side up, while gills remained submerged in the aerated water and blood (∼2–3 ml) was drawn from the caudal vein into a syringe (3 ml syringe, 23 G 1¼ inch needle) flushed with heparinized saline (10 i.u. ml⁻¹, lithium heparin; Sigma-Aldrich H0878) and placed on ice. By rapidly taking blood samples with the gills submerged and disturbances to the fish minimized, caudal puncture is a valid method for assessing blood acid–base status (Brauner et al., 2019).

Following this procedure, fishes were killed via cephalic concussion and cervical dislocation and tissues (0.5–1.0 g) were removed within 2–3 min, wrapped in aluminium foil and immediately flash-frozen in liquid N₂ and stored at −80°C until pH_i was determined. Tissues were sampled in the following order: heart (gently squeezed and patted dry to remove any excess blood), liver, dorsal white muscle (left side, just posterior of the dorsal fin; skin and red muscle removed), and brain. Blood was divided into two aliquots. Blood pH and hematocrit were measured from one aliquot; the other aliquot was centrifuged (3 min at 11,200 g) and plasma was removed for measurement of [Cl⁻].

Blood pH was measured using a Radiometer PHM 84 (Copenhagen, Denmark) connected to a thermostatted (at the temperature at which the fish was sampled) Radiometer Analytical SAS pH electrode (GK2401C, Cedex, France). Red blood cell (RBC) pH_i was measured using the freeze–thaw method as described by Zeidler and Kim (1977). Tissue pH_i was measured using the metabolic inhibitor tissue homogenate method (MITH) (Pörtner et al., 1990) and validated for use in fishes by Baker et al. (2009b). Plasma Cl⁻ (HBI model 4425000; digital chloridometer) and osmolarity (model 5520; Westcor Vapor Pressure Osmometer) were measured; not all samples were measured owing to loss of plasma samples during transport.

Data were compared by Welch's two-tailed t-test, or where multiple treatments were evaluated, data were analyzed by ANOVA, followed by Tukey's post hoc test. If the data did not meet the assumptions of normality (Shapiro–Wilk normality test) or equal variance (Bartlett's test), a Kruskal–Wallis test followed by Dunn's multiple comparison test was used (P<0.05). Cohen's d was calculated in R to determine effect size using the effsize package (https://CRAN.R-project.org/package=effsize); this provides information about the size of differences between group means. GraphPad Prism (v.5) and R version 3.5.1 (https://www.r-project.org/) were used for statistical analyses and preparation of figures. All values are expressed as means±s.e.m. throughout.

For inter-specific data analysis, phylogenetically informed statistical analyses were used to account for phylogenetic non-independence. First, a composite, time-calibrated phylogeny was generated using divergence time obtained from the literature (Aschliman et al., 2012; Betancur-R et al., 2013, 2015; Blair, 2005). A Pagel test for binary character evolution was used to test whether the evolution of preferential pH regulation affected the evolution of air-breathing (Pagel, 1994).

To infer the evolutionary history of pH regulatory strategy, we first compared two different evolutionary models for the strategy for pH regulation using the two states ‘coupled pH regulation’ and ‘preferential pH_i regulation’. Here, we fitted a continuous-time Markov model using maximum likelihood using either the same forward and reverse transition rates between the two states (lnL=−9.98, AICc=22.2, AICw=0.68) or different forward and reverse transition rates between the two states (lnL=−9.72, AICc=24.2, AICw=0.32). This was done using the fitContinuous()-function in the geiger package (Harmon et al., 2007). Based on AICw, the first model was used to infer the evolutionary history of the strategy for pH regulation. Next, we generated 10,000 stochastic character maps of the evolution of pH regulatory strategy on the phylogeny using the make.simmap function in the phytools package (Bollback, 2006; Revell, 2011), and summarized the probability of the pH regulatory strategy being preferential pH_i regulation in all internal branches in the phylogeny using the describe.simmap function in the phytools package (Bollback, 2006; Revell, 2011).

Within species, the P_CO2 at loss of equilibrium (LOE; defined as the inability of fishes to maintain an upright position within the water column) recorded for fishes exposed to the highest rate of increase in environmental hypercarbia (4 kPa h⁻¹) was lower than that of fish exposed to lower rates (P<0.01); however, there was no difference between the rates of 1 and 2 kPa h⁻¹ for either species. In rainbow trout, mean P_CO2 values at LOE were 5.5±0.3, 4.8±0.3 and 2.7±0.2 kPa at exposure rates of 1, 2 and 4 kPa h⁻¹, respectively. In white sturgeon, mean P_CO2 values at LOE were 22.1±2.2, 14.6±2.3 and 6.3±1.8 kPa, respectively (Fig. 2). Comparison between species at the different rates of hypercarbia increase indicated that white sturgeon had a higher P_CO2 at LOE for 1 and 2 kPa h⁻¹, but not 4 kPa h⁻¹ (P<0.01). No mortalities occurred in the 72 h following LOE in fishes allowed to subsequently recover in normocarbia.

Exposure of rainbow trout, a strict coupled pH regulator (Shartau et al., 2016a), to 3 kPa P_CO2 for 3 h resulted in the expected reductions in pH_e, RBC pH_i and pH_i of heart, liver, brain and white muscle (Table 2, Fig. 3). Hypercarbia exposure for 3 h to 1.5 kPa P_CO2 in American paddlefish, 4 kPa P_CO2 in tamoatá, matrinxã and tambaqui, and 6 kPa P_CO2 in spotted gar, alligator gar, channel catfish and tilapia hybrid reduced both pH_e and RBC pH_i as expected, but not in oscar exposed to 4 kPa P_CO2, although the statistical power of this analysis was low owing to a limited sample size (for oscar only, n=2). In contrast to rainbow trout, all other species investigated exhibited an increase or no statistically significant change in pH_i of the non-RBC tissues during 3 h of hypercarbia, indicative of preferential pH_i regulation. The pH_i of the heart increased in spotted gar, tamoatá, matrinxã, tambaqui and oscar (P<0.05). Other tissues also exhibited increases in pH_i: liver of alligator gar, brain of channel catfish, and brain and white muscle of tilapia hybrid (Table 2, Fig. 3); no other statistically significant changes were observed in other tissues despite the severity of the extracellular and RBC acidosis. Similarly, exposure of tambaqui to 20 kPa P_CO2 severely reduced pH_e and RBC pH_i, yet induced no changes in heart, liver or brain pH_i, although white muscle pH was reduced (P<0.05; Table 2, Fig. 3).

Hypercarbia exposure did not alter plasma Cl⁻ or osmolarity (when measured) (Table 3). Hematocrit remained unchanged except in American paddlefish and rainbow trout. In American paddlefish, hematocrit was reduced from 27±2% to 18±2% following 3 h exposure to 1.5 kPa P_CO2. Rainbow trout hematocrit was elevated from the control value of 27±2% to 35±2% following 3 h exposure to 1.5 kPa P_CO2 and to 51±4% following 3 h exposure to 3 kPa P_CO2 (Table 3). No changes in plasma osmolarity were observed (Table 3).

The phylogenetic model indicates that preferential pH_i regulation was the most likely strategy for pH regulation amongst the earliest actinopterygian fishes (Bayesian posterior probability of 0.55, 0.67 and 0.87 for the most recent common ancestor of vertebrates, jawed vertebrates and ray-finned fishes, respectively; Fig. 4), and it was lost by three species. Preferential pH_i regulation was not specifically associated with air-breathing fishes (P=0.36).

The objective of this study was to investigate the presence or absence of preferential pH_i regulation in phylogenetically separated fishes. Amongst the basal and derived bony euteleostomi fishes investigated, nine out of 10 species studied exhibited preferential pH_i regulation. Fishes using preferential pH_i regulation were, with the exception of American paddlefish, tolerant of hypercarbia levels well above the value at which pH_e compensation can be achieved in prolonged exposure (ca. 2 kPa P_CO2; Shartau et al., 2019), indicating it may represent a ubiquitous strategy of acid–base regulation in hypercarbia-tolerant fishes (Figs 3 and 4).

During hypercarbia, all species examined except rainbow trout tightly regulated pH_i despite a pH_e reduction (Fig. 3); these species are considered to be preferential pH_i regulators. Most previous studies measuring both pH_i and pH_e demonstrate that following hypercarbia exposure, pH_i and pH_e are initially reduced, and both are compensated in the subsequent hours to days with recovery of pH_i occurring faster (Shartau et al., 2016a; coupled pH regulation). In the present study, all nine preferential pH_i regulating species experienced no reduction in heart, liver or brain following 3 h exposure to water P_CO2 ranging from 1.5 to 20 kPa. White muscle was also well regulated, with reduced white muscle pH_i only occurring in tilapia hybrid exposed to 6 kPa P_CO2 and tambaqui exposed to 20 kPa P_CO2; white muscle was not reduced in tambaqui exposed to 4 kPa P_CO2 (Table 2, Fig. 3). Notably, in heart, liver and brain there were several observed increases in pH_i despite the reduction in pH_e; it is unclear why these tissues overcompensated pH_i, but it may be that their ability to finely regulate pH_i is less well developed than other tissues, and it has been observed in other studies (e.g. Baker et al., 2009a). The reduction in white muscle pH_i may be a consequence of the large size of this tissue and the fact that pH_i reductions in white muscle are less damaging than those to heart, liver or brain. In contrast to those four tissues, RBC pH_i was reduced in all species. This reduction in RBC pH_i has been observed in other species, including those demonstrating preferential pH_i regulation (Baker et al., 2009a; Brauner et al., 2004; Shartau et al., 2016a); these reductions in RBC are expected given their limited capacity for pH regulation. Although many teleosts have the capacity to rapidly regulate RBC pH_i using beta-adrenergic Na⁺/H⁺ exchange (β-NHE) (Berenbrink et al., 2005), this ability is short-lived and provides limited protection during severe acidoses such as hypercarbia (Shartau et al., 2019), as is observed in rainbow trout RBC pH_i in particular, which is well studied in relation to RBC β-NHE.

Most studies where fishes have been exposed to acute hypercarbia have used P_CO2 <2 kPa and observe concurrent pH_e and pH_i reductions; there are only a few studies at higher P_CO2 levels where both pH_e and pH_i have been measured throughout CO₂ exposure (Brauner and Baker, 2009; Shartau et al., 2016a). Species that utilize coupled pH regulation appear unable to completely compensate blood pH_e during acute exposure to water P_CO2 >2 kPa (Heisler, 1984), possibly to prevent hypochloremia associated with net increase in plasma HCO₃⁻ in equimolar exchange for plasma Cl⁻ (Baker et al., 2015; Brauner and Baker, 2009). This relationship between pH_e compensation and plasma Cl⁻ during hypercarbia exposure was assessed in the present study. Not surprisingly, as pH_e recovery was absent, there were no changes to plasma Cl⁻ (Table 3) in those species measured. This is consistent with findings from armored catfish (Brauner et al., 2004), but differs from those from white sturgeon, which exhibit significant reductions in plasma Cl⁻ following 6 h exposure to 1.5, 6 and 12 kPa P_CO2 despite no pH_e compensation (Baker and Brauner, 2012; Baker et al., 2009a).

The capacity for pH_e compensation is influenced by exposure time and water quality, as fishes chronically exposed to gradual hypercarbia can compensate pH_e beyond their capacity during acute exposure (McKenzie et al., 2003; Smart et al., 1979). Furthermore, the rate and degree of pH_e compensation during acute hypercarbia exposure is also affected by water salinity (Tovey and Brauner, 2018), ion composition (Larsen and Jensen, 1997) and pH (Sackville et al., 2018). However, survival during severe acute hypercarbia may depend on protecting pH_i (Shartau et al., 2017a; Vandenberg et al., 1994; Wood et al., 1983; Yoshikawa et al., 1994) more so than pH_e. Mortality in marine fishes during severe acute hypercarbia may be due to reduced heart pH_i (Hayashi et al., 2004). Reduced cardiac pH_i is hypothesized to lower cardiac contractility, which reduces cardiac output and O₂ supply (Vandenberg et al., 1994). Protection of cardiac pH_i may guard against reduced cardiac performance, as an in situ perfused heart preparation demonstrated cardiac output, power output and stroke volume did not change during acute hypercarbia (5 kPa P_CO2) in the preferentially pH_i-regulating armored catfish (Hanson et al., 2009) and white sturgeon (Baker et al., 2011). The ability to protect pH_i in various tissues may allow those species to be resilient to a variety of respiratory and metabolic acidoses (Harter et al., 2014; Shartau et al., 2017a) and provide them with a window for pH_e to recover following the development of an acidosis as time and/or conditions permit. Nevertheless, preferential pH_i regulation may not be sufficient in itself to confer tolerance to severe acute hypercarbia. This is demonstrated in American paddlefish, which fully protected pH_i despite their relative sensitivity to hypercarbia based on an initial assessment of CO₂ tolerance, and were only exposed to a P_CO2 of 1.5 kPa when assessed for pH regulation owing to them experiencing LOE at 2.7±0.5 kPa P_CO2; the basis for this sensitivity is unknown.

Although preferential pH_i regulation may be important for tolerating acute exposure to hypercarbia levels at which pH_e compensation would be limited, several fish species that exhibit preferential pH_i regulation also possess the ability to regulate pH_e. During exposure to 4 kPa P_CO2, complete or partial pH_e recovery occurs over 24 h in the preferentially pH_i-regulating tamoatá, matrinxã, tambaqui and oscar (data not shown). This is similar to the response observed in the striped catfish of the Mekong, which compensate pH_e by 48 h at ca. 4 kPa P_CO2 (Damsgaard et al., 2015) while preferentially regulating pH_i (Sackville et al., 2018). In the striped catfish, the degree of pH_e regulation was significantly reduced at lower water pH despite complete preferential pH_i regulation (Sackville et al., 2018), indicating that water pH also has a direct effect on the ability to compensate pH_e. Thus, preferential pH_i regulation does not preclude pH_e regulation, but results in a far more rapid and robust pH_i regulatory response at times when pH_e is depressed during initial exposure to hypercarbia.

The cellular and molecular mechanisms involved in this ability for pH_i regulation remain largely unknown; however, in tissues that preferentially regulate pH_i, there must be a net transport of acid–base equivalents across the plasma membrane into the extracellular fluid surrounding the tissues. Tissue buffering may play a role in minimizing any reduction to pH_i immediately following the onset of the acidosis; the relationship between the non-bicarbonate tissue buffer capacity and preferential pH_i regulation is uncertain. Preferential pH_i regulation is dependent on active regulation of acid–base equivalents; the specific transporters are not known, but are assumed to involve common acid–base transporters such as sodium proton exchangers (NHEs) and Cl⁻/HCO₃⁻ transporters (e.g. AE). The putative mechanisms of acid–base regulation in response to CO₂-induced respiratory acidosis are reviewed in Brauner et al. (2019) and remain an exciting area of future research.

Preferential pH_i regulation is observed in a phylogenetically separated group of fishes, ranging from basal to derived, that includes both water and air breathers found in habitats ranging from temperate to tropical (Fig. 4). Reconstruction of the evolution of preferential pH_i regulation on a vertebrate phylogeny indicates that preferential pH_i regulation was likely the ancestral state for pH regulation in adult actinopterygians (Fig. 4); however, this conclusion should be interpreted cautiously given the limited number of species investigated in a very diverse group. It would be interesting to investigate the presence or absence of preferential pH_i regulation more thoroughly in more basal vertebrate and invertebrate groups (i.e. Chondrichthyes, Dipnoi, Cephalochordata, Hemichordata, Urochordata and Echinodermata) to provide a more comprehensive picture of the early evolution of preferential pH_i regulation.

Recently, preferential pH_i regulation has been hypothesized to be an embryonic trait that is either retained or lost in adults (Shartau et al., 2016a), as studies in embryonic reptiles indicate use of preferential pH_i regulation during development with a gradual loss to the adults (Shartau et al., 2018, 2016b). In embryonic/larval fishes, acid–base regulation is not well studied, but early-stage zebrafish embryos exposed to hypercarbia can rapidly regulate pH_i, suggesting that some degree of preferential pH_i regulation may occur (Molich and Heisler, 2005). Fishes inhabiting environments prone to severe acid–base challenges could retain the embryonic preferential pH_i regulatory strategy to successfully tolerate these challenges as adults; this may extend to fishes that experience severe metabolic acid–base disturbances as armored catfish (Harter et al., 2014) and white sturgeon (Shartau et al., 2017a) protect pH_i during severe exercise. In contrast, fishes inhabiting less stressful environments may lose this trait, which may explain, at least in part, the absence of preferential pH_i regulation (and use of coupled pH regulation) in temperate species such as rainbow trout and Coho salmon, which are unlikely to encounter severe hypercarbia (Ou et al., 2015) (Fig. 4); this remains to be investigated further.

Hypercarbia tolerance may have been a selective pressure for adult fishes expanding into certain CO₂-rich environments. Additionally, Ultsch (1987, 1996) has postulated that aquatic hypercarbia was a selective pressure, along with aquatic hypoxia, for the evolution of air breathing. Aerial respiration in fishes is often associated with increased blood P_CO2 because (1) hypoxic waters are often simultaneously hypercarbic, and (2) CO₂ release still largely occurs at the gills, but gill ventilation is typically reduced during air breathing (Shartau and Brauner, 2014). CO₂ tolerance conferred by preferential pH_i regulation, which leads to rapid and complete pH_i regulation, may have been a beneficial adaptation that was co-opted for the evolution of air breathing, and thus could be considered an exaptation for this important evolutionary transition in vertebrates.

In summary, prior to this study, only four fish species were known to exhibit preferential pH_i regulation; we identified an additional nine species using this strategy, and now there are a total of 13 species that span a broad phylogenetical range, from the basal actinopterygians to derived teleosts, known to use preferential pH_i regulation. Here, we demonstrate that preferential pH_i regulation may be a widely used strategy to survive and tolerate CO₂ tensions ranging from 1.5 to 20 kPa P_CO2. As acid–base regulation is intimately associated with routine physiological functioning and, ultimately, survival, understanding how preferential pH_i regulation allows fishes, and other vertebrates, to tolerate severe acid–base disturbances may provide insight into key evolutionary transitions in vertebrates, such as the evolution of air breathing.

We thank Fernanda Dragan and Nazare Paula da Silva at the Instituto Nacional de Pesquisas da Amazônia, Mack Fondren and the staff of the Mississippi State University South Farm Aquaculture Facility, Dave Switzer and Gord Edmondson at the International Center for Sturgeon Studies at Vancouver Island University, as well as the numerous individuals who assisted with fish care, and provided assistance with experiments and feedback. We also thank the two anonymous reviewers for comments that greatly improved the manuscript.