In all vertebrates studied to date, CO2 excretion depends on the enzyme carbonic anhydrase (CA) that catalyses the rapid conversion of HCO3− to CO2 at the gas-exchange organs. The largest pool of CA is present within red blood cells (RBCs) and, in some vertebrates, plasma-accessible CA (paCA) isoforms participate in CO2 excretion. However, teleost fishes typically do not have paCA at the gills and CO2 excretion is reliant entirely on RBC CA – a strategy that is not possible in icefishes. As the result of a natural knockout, Antarctic icefishes (Channichthyidae) are the only known vertebrates that do not express haemoglobin (Hb) as adults, and largely lack RBCs in the circulation (haematocrit <1%). Previous work has indicated the presence of high levels of membrane-bound CA activity in the gills of icefishes, but without determining its cellular orientation. Thus, we hypothesised that icefishes express a membrane-bound CA isoform at the gill that is accessible to the blood plasma. The CA distribution was compared in the gills of two closely related notothenioid species, one with Hb and RBCs (Notothenia rossii) and one without (Champsocephalus gunnari). Molecular, biochemical and immunohistochemical markers indicate high levels of a Ca4 isoform in the gills of the icefish (but not the red-blooded N. rossii), in a plasma-accessible location that is consistent with a role in CO2 excretion. Thus, in the absence of RBC CA, the icefish gill could exclusively provide the catalytic activity necessary for CO2 excretion – a pathway that is unlike that of any other vertebrate.
The first scientific investigation of an Antarctic icefish, less than a century ago (Ruud, 1954), overthrew the common perception that haemoglobin (Hb) was a necessity to sustain vertebrate life. In fact, an entire family of teleosts, Channichthyidae within the suborder Notothenioidei (Perciformes) and comprising 16 species, do not express Hb as adults (Ruud, 1954; Eastman, 1993) and largely lack red blood cells (RBCs) in their circulation; residual haematocrit (Hct) is typically <1% (Egginton, 1994). The implications for cardiovascular gas transport are tremendous. In the absence of Hb, icefish blood has a 10-fold lower O2-carrying capacity compared with that of red-blooded notothenioids (Holeton, 1970), and without RBCs icefish lack the important pool of carbonic anhydrase (CA; and also aquaporins and Cl−/HCO3− exchangers) that facilitates CO2 transport and excretion in all vertebrates (Tufts and Perry, 1998). Those adaptations that address the dramatic impairment of O2 transport in icefishes are largely known, such as a lower metabolic rate, a higher blood volume, a higher cardiac output that is generated by a hypertrophied heart and a higher production of nitric oxide at the tissues (Hemmingsen and Douglas, 1970, 1972; Holeton, 1970; Sidell and O'Brien, 2006); however, adaptations needed to resolve the associated problem of CO2 excretion are not.
Most vertebrates transport the majority of CO2 that is produced in tissues as dissolved HCO3− in the blood plasma. In this regard, icefishes are no exception, as indicated by venous blood pH and PCO2 values (7.84 and 0.3 kPa in Chaenocephalus aceratus; Hemmingsen and Douglas, 1972) that are in line with those found in other fishes. Under these conditions, and because of the low apparent pK of the CO2–HCO3− reaction of ∼6.3 (Boutilier et al., 1984), blood plasma is an effective sink for CO2. While this greatly increases the capacitance for CO2 transport in blood (Tufts and Perry, 1998; Henry and Swenson, 2000), it also requires a rapid conversion of CO2 to HCO3− at the tissues and the reverse reaction at the gills for CO2 excretion. However, the spontaneous rates of these reactions are slow relative to the residence time of blood at the respiratory surfaces and tissue capillaries, and these rates further slow with decreasing temperature. At physiological temperatures in icefish, around −1.9°C (Littlepage, 1965), the halftime of spontaneous HCO3− dehydration to CO2 is ∼300 s (Kern, 1960; Heming, 1984) and thus exceeds the residence time of blood at the gills (∼1–3 s) by two orders of magnitude (Cameron and Polhemus, 1974; Hughes et al., 1981). Based on the arterial–venous differences in PCO2 and pH in C. aceratus (Hemmingsen and Douglas, 1972), it can be estimated that in resting, normoxic icefish, about 68% of CO2 excretion must depend on HCO3− dehydration in the branchial vasculature, while the remainder is from physically dissolved CO2 in the plasma. During aerobic exercise, where blood pH is largely maintained, HCO3− concentration may increase by 50% (Brauner et al., 2000) and the residence time at the gills will be reduced further, because of a higher cardiac output (Randall, 1982); an increase in cardiac output following exercise has recently been shown for C. aceratus (Joyce et al., 2018). Clearly, the uncatalysed rate of HCO3− dehydration is simply not rapid enough to support CO2 excretion in any adult vertebrate, but in particular in icefishes at these low temperatures.
The rate limitation of CO2–HCO3− reactions in the blood of vertebrates is largely alleviated by the catalytic activity of CA. The major CA pools are: (i) RBC intracellular CA (Maren, 1967) – plasma HCO3− has functional access to this CA pool via rapid Cl−/HCO3− exchange across the RBC membrane (Romano and Passow, 1984); (ii) soluble CA isoforms in the plasma (Henry et al., 1997b); and (iii) plasma-accessible CA (paCA) isoforms that are anchored to the apical membranes of the endothelium (Henry and Swenson, 2000). At the tissue capillaries, paCA is typically present and ensures a rapid conversion of CO2 to HCO3− (Henry et al., 1997a). However, at the gas exchange surface, the contribution of different CA pools to CO2 excretion varies largely among the major vertebrate groups. On one end of the spectrum are the basal hagfishes (Esbaugh et al., 2009) and Chondrichthyes (Gilmour et al., 2002, 2007), which rely on RBC CA, soluble CA in the plasma and paCA at the gills for CO2 excretion. All Euteleostomi lack soluble CA activity in the plasma, and thus most tetrapods rely on RBC CA and, to a lesser degree (<10% of total CO2 excretion), on paCA at the gas exchange surface (Bidani et al., 1983; Zhu and Sly, 1990; Stabenau and Heming, 2003). And finally, on the other end of the spectrum are teleost fishes that have also lost paCA activity at the gills (for review, see Harter and Brauner, 2017). Therefore, in teleosts, HCO3− dehydration is shifted entirely into the RBCs (Perry et al., 1982; Wood et al., 1982; Desforges et al., 2001, 2002; for review see Perry and Gilmour, 2002), creating a strong coupling between O2 and CO2 transport – a hallmark of teleost gas exchange (Brauner and Randall, 1996). This strategy is clearly not available to icefishes, which are teleosts but lack RBCs. Thus, with a clear need to catalyse HCO3− dehydration, some other CA pool must be present in icefishes to compensate for the loss of RBC CA.
Previous studies on gill homogenates from icefishes have provided biochemical evidence for a higher activity of membrane-associated CA when compared with red-blooded notothenioids (Feller et al., 1981; Maffia et al., 2001). Tufts et al. (2002) further characterised the branchial CA isoform distribution of notothenioids and found biochemical markers for the presence of a membrane-bound Ca4 isoform in the gills of an icefish species but, surprisingly, also in the gills of a red-blooded notothenioid. A critical detail, the cellular orientation of putative paCA isoforms remains unexplored and therefore the potential involvement of a Ca4 isoform in CO2 excretion remains unresolved for icefishes. Building on these previous findings, we hypothesised that icefishes express a membrane-bound CA isoform at the gill that is accessible to the blood plasma where it would catalyse CO2 excretion in the absence of RBC CA. To this end, biochemical, molecular and immunohistochemical techniques were used to compare the CA isoform distribution in the gills of the icefish Champsocephalus gunnari and the red-blooded Notothenia rossii. The obtained results shed new light on a divergent strategy of CO2 excretion in icefishes, which is unlike that found in any other adult vertebrate.
MATERIALS AND METHODS
Specimens of Notothenia rossii Richardson 1844 and Champsocephalus gunnari Lönnberg 1905 (average mass 343.4±17.2 and 644.3±70.1 g, and length 38.8±0.7 and 37.7±1.1 cm) were captured using otter trawls or baited pot traps deployed from the US ARSV Laurence M. Gould at Low Island (63°30′S, 62°37′W) and North Dallmann Bay (63°55′S, 62°43′W), Antarctica. Animals were stunned by a sharp blow to the head. Blood was drawn from the caudal vein and mixed with 3.2% sodium citrate (9:1 for N. rossii and 4:1 C. gunnari). All samples were centrifuged at 5300 g for 10 min and plasma was decanted. Blood cells and plasma were frozen in liquid nitrogen and stored at −70°C. After blood sampling, animals were euthanised by severing the spinal cord and brain pithing. Gills and hearts were perfused with notothenioid Ringer (in mmol l−1: 260 NaCl, 2.5 MgCl2, 5 KCl, 2.5 NaHCO3 and 5 NaH2PO4, pH 8.0) and tissues were frozen at −70°C or fixed in 10% buffered formalin for 24 h and then transferred to 70% ethanol. Fixed tissues were shipped on ice and frozen tissues were shipped on dry ice to The University of British Columbia (UBC), in Vancouver, Canada. All samples were collected opportunistically and in strict compliance with the guidelines of The Institutional Animal Care and Use Committee (IACUC Protocol no. 14-L-004, Ohio University).
Biochemical analysis of CA activity
Approximately 2 g of gill lamellae were homogenised (Polytron PT1200, Lucerne, Switzerland) in 8 ml of assay buffer on ice (in mmol l−1: 225 mannitol, 75 sucrose and 10 Tris base, adjusted to pH 7.4 with 10% phosphoric acid). Differential centrifugation was at 4°C (see Henry, 1988; Henry et al., 1993): (i) 800 g for 20 min; (ii) 8500 g for 20 min (Allegra 64R, Beckman Coulter, Brea, CA, USA); (iii) 100,000 g for 90 min (Beckman L8-70M) to produce a microsomal pellet containing plasma membranes and a supernatant containing the cytosolic fraction. Pellets were re-suspended in 3 ml of assay buffer, by vortexing and mild sonication (5 W for 3 s). Protein concentration was measured spectrophotometrically at 595 nm using the Bradford assay (Sigma B6916, St Louis, MO, USA) and bovine serum albumin standards (Bio-Rad Quickstart 5000206, Hercules, CA, USA).
The activity of CA in cellular fractions was measured using the electrometric ΔpH assay (Henry, 1991). Reactions were in 6 ml of assay buffer in a thermostatted vessel at 4°C using 100 μl CO2 saturated water as a substrate. The reaction kinetics were assessed as the time for a 0.15 unit pH change, with a GK2401C electrode and PHM84 meter (Radiometer, Copenhagen, Denmark). Uncatalysed reaction rates (without sample addition) were subtracted from the enzymatic rates and absolute enzyme catalytic rates were calculated from the buffer curve of the assay buffer over the tested pH range (determined in separate titrations).
Membrane pellets were washed by an additional step of ultracentrifugation (100,000 g for 90 min) and re-suspended in 3 ml of fresh buffer. Washed pellets were incubated with 1 IU phosphatidylinositol-specific phospholipase C (PI-PLC; Invitrogen P6466, Carlsbad, CA, USA), an enzyme that cleaves the common glycosylphosphatidylinositol (GPI) membrane anchor, or with assay buffer as a control, for 90 min at 21°C. CA inhibition kinetics were assessed by: (i) adding 0.005% sodium dodecyl sulfate (SDS) to the assay buffer; (ii) titrations with 0.6–6 nmol l−1 acetazolamide (Az) (see Easson and Stedman, 1936; Dixon, 1953); and (iii) adding 100 µl of plasma from either C. gunnari or N. rossii to the assay buffer. RBC lysates were produced from 50 µl packed RBCs from N. rossii, diluted 50-fold in distilled water and frozen in liquid nitrogen twice; CA activity was measured on 5 µl of lysate.
Plasma protein concentration was measured in both species as described above. In addition, Hb concentration in plasma samples from N. rossii was measured spectrophotometrically at 540 nm using the cyanomethaemoglobin method with human Hb dilutions as standards (Sigma H7379). The concentration of protein from Hb was then subtracted from total protein concentration measured in the plasma. The plasma non-bicarbonate buffer capacity (βplasma) was measured with an automated titrator (TIM865, Radiometer). Plasma aliquots of 200 µl were added to 4.5 ml of deionised water in a magnetically stirred glass titration vessel (4°C) that was continuously sparged with N2. All results represent upward titrations from pH 4 to 9 with 0.01 mol l−1 NaOH. βplasma was calculated from the change in pH that corresponded to individual steps of base addition (10 µl) over the physiologically relevant pH range in notothenioids of pH 7.4–8.2 (Acierno et al., 1997); βplasma is given as the mean value over the tested pH range.
Localisation of Ca4 in the gills of C. gunnari and N. rossii was carried out with a custom rabbit polyclonal antibody raised against rainbow trout (Oncorhynchus mykiss) Ca4, which has been described in detail (Gilmour et al., 2007) and has been successfully used in rainbow trout and spiny dogfish (Squalus acanthias). The antigenic sequence (TRRTLPDERLTPFTFTGY) corresponds to amino acids 57–74 of the rainbow trout Ca4 (GenBank AAR99330), which is 73% conserved in Notothenia coriiceps. The immunohistochemical results were later replicated using a custom chicken polyclonal antibody raised against Ca4 of three Chondrichthyes (Squalus acanthias, DQ092628.1; Rhincodon typus, XM_020514262.1; Callorhinchus milii, XP_007894777.1). The antigenic peptide sequence for S. acanthias Ca4 was GSEHTIDGEQYPMELHIVH (amino acids 125–144), and the sequence in the notothenioid Ca4 is 100% conserved. Other sections were immunolabelled with a rabbit anti-Ca2 antibody (ab191343, Abcam, Cambridge, UK); the cytosolic Ca2-like isoform in fishes was recently reclassified as Ca17 (Ferreira-Martins et al., 2016). Ca4 and Ca2 antibodies were tested by western blot analysis using cytosolic and microsomal fractions of gill homogenates from both species. Subsamples containing 20 µg of protein were separated by SDS-PAGE using 10% polyacrylamide gels (with 4% stacking gel). Proteins were then wet-transferred onto 0.2 µm PVDF membranes (Immun Blot, Bio-Rad), rinsed and air dried. Transfer was assessed using total protein staining with 0.5% Ponceau S in 1% acetic acid and then imaged. Blots were rinsed with TTBS (Tris-buffered saline with 0.05% Tween 20, pH 7.4) and blocked with 5% blotto in TTBS overnight at 4°C. Thereafter, one membrane was probed with a 1:1000 dilution of the rtCa4 and the other with a 1:2500 dilution of the Ca2 antibody, overnight at room temperature on a rotisserie (Lab QuakeII, Thermo Scientific, Waltham, MA, USA). Protein size was determined using a Precision Plus Protein Dual Color ladder (Bio-Rad 1610374). All membranes were rinsed three times with TTBS and incubated with a 1:25,000 dilution of a goat anti-rabbit secondary antibody conjugated to horseradish peroxidase (HRP; Genscript, Piscataway, NJ, USA), for 1 h at room temperature. Finally, membranes were rinsed with TTBS and proteins were visualised using a chemiluminescent HRP substrate (Clarity, Bio-Rad). Images were acquired using the Azure C300 imaging system and software (Azure Biosystems, Dublin, CA, USA).
To localise Ca4 and Ca17 in the gills, fixed tissues were stepwise dehydrated in ethanol, cleared in xylene and embedded in paraffin. Thin sections (5 µm) were cut on a microtome (Leica RM2500, Wetzlar, Germany) and mounted on aminopropylsilane (APS)-coated microscope slides. A hydrophobic barrier (SuperPAP, Sigma) was created around the sections that were incubated in a blocking buffer (Bløk™, Millipore, Burlington, MA, USA) for 15 min. Incubation with the primary antibody (rtCa4 or Ca2, 1:200) in blocking buffer was overnight at 4°C in a humidified chamber. Negative controls were incubated with blocking buffer alone, or with normal rabbit serum. Detection of the primary antibody was done with a goat anti-rabbit IgG conjugated to Alexa 488 (Jackson ImmunoResearch, West Grove, PA, USA). Sections were then rinsed three times with 0.1 mol l−1 phosphate buffered saline (PBS) for 5, 10 and 15 min and incubated with secondary antibody in a humidified chamber for 1 h at 37°C. DAPI (4′,6′-diamidino-2-phenylindole) was added to the second wash step to visualise cell nuclei. Coverslips were mounted with 1:1 PBS glycerol containing 0.1% NaN3 and imaging was done with a fluorescence photomicroscope (Leica DM5500, Orca Flash 4, Hamamatsu, Japan).
Sequencing and expression of ca4
Total RNA was extracted from approximately 100 mg of gill and ventricle tissue in 1 ml of Trizol, following the manufacturer's protocol (Invitrogen 15596018). Ventricles were used as a control tissue, in which the presence of Ca4 has been confirmed in several teleost species (Georgalis et al., 2006; Alderman et al., 2016). Tissues were homogenised with a Bullet Blender 24 with ∼10 zirconium oxide beads (Next Advance, Averill Park, NY, USA). The resulting RNA samples were treated with DNAse I (Thermo Scientific EN0521). RNA concentrations were measured using a Nanodrop ND-2000 spectrophotometer (Thermo Scientific). First strand cDNA was synthesised from 2 µg of RNA using a high capacity reverse transcription kit (Applied Biosystems 4368814, Foster City, CA, USA) and the cDNA product was diluted 3-fold with molecular grade DEPC-treated deionised water (Invitrogen 46-2224).
Degenerate PCR primers were designed by aligning available fish ca4 sequences using the Clustal Omega web service (http://www.clustal.org), and identifying conserved sections among the sequences (primer sequences were forward 5′-GGA GAG CAG TAY CCC ATG G-3′ and reverse 5′-TGG GCT TCT CAA ACA MRG TCC-3′). PCR products (40 cycles; 94°C for 2 min, 94°C for 30 s, 72°C for 1 min) were purified on a 1% agarose gel with a 1 kb ladder. Sections of the gel containing the 323 bp PCR product were removed and purified using a GeneJet gel extraction kit (Thermo Scientific K0691). Purified PCR products were ligated into Topo2.1 plasmids and transformed in One-Shot Topo10 competent cells (Invitrogen C404010) following the manufacturer's protocol. Plasmids were extracted from 10 different bacterial colonies with a GeneJet MiniPrep plasmid kit (Thermo Scientific K0503). Purified plasmids were sequenced at the UBC Nucleic Acid and Protein Service core facility (NAPS, Vancouver, Canada).
Primers for real-time quantitative PCR (qPCR) analysis were designed by aligning the obtained partial coding sequences (CDS) for C. gunnari ca4a with the N. coriiceps ca4a-like mRNA sequence (XM_010775657.1). The generated primers were used on gill and ventricle tissues of both species (primer sequences were forward 5′-GGG AAG CAG AGA AGT GTT GC-3′ and reverse 5′-TTT CAG ACG CAG AGG GAG TT-3′). Primers for the ef1α control gene were those reported by Urschel and O'Brien (2008), designed for three notothenioid species, and all results are reported relative to the expression of ef1α. qPCR amplifications were with the SybrGreen kit (Applied Biosystems 4309155) on a Bio-Rad CFX96 RT-PCR Detection System with the following cycling conditions: 40 cycles, 95°C for 10 min, 95°C for 15 s, 55°C for 1 min; melt curve over 65–95°C at 0.5°C s−1. No-amplification controls (no reverse transcriptase in the cDNA synthesis reaction) were run for each sample and showed no detectable amplification. Standard curves were run on each plate by serially diluting (1:5) pooled sample cDNA with molecular grade water in five steps. Primer pair efficiencies were within 100–120% and R2>0.99 for all samples. To confirm the identity of the amplified products, qPCR products were processed with a GeneJet PCR purification kit (Thermo Scientific K0701). The purified qPCR products were cloned and plasmids were extracted as described above. Purified plasmids from 10 colonies were sequenced using M13 forward and reverse primers (UBC NAPS).
Data analysis and statistics
The results for the immunohistochemical localisation of Ca4 and Ca17 protein in the gills of C. gunnari and N. rossii are shown in Fig. 1. In C. gunnari, reactivity for Ca4 protein was observed as a clear ring, lining the entire blood space of the secondary lamellae. This staining pattern was associated with the apical membrane of pillar cells and the basolateral membrane of lamellar epithelial cells (Fig. 1A, asterisks). In contrast, in N. rossii, reactivity for the Ca4 antibody was a diffuse staining pattern associated with the intracellular space of lamellar pillar cells and epithelial cells and absent from the lamellar blood space (Fig. 1B). Nonetheless, reactivity for the cytosolic Ca2 antibody showed a similar pattern for the two species, with staining confined to the cytosol of all lamellar cell types and RBCs that remained in un-perfused areas within the lamellae (marked with asterisks in Fig. 1C,D). The specificity of the antibodies was confirmed by the western blotting results. Probing of immunoblots from C. gunnari microsomal pellets with the Ca4 antibody revealed one band at ∼37.5 kDa that was not observed in the pellets of N. rossii or in the supernatants of either species. The Ca2 antibody showed immunoreactivity against a band at ∼25 kDa in the cytosolic fractions from both species, but not in the pellets.
Homology cloning yielded a CDS for C. gunnari ca4a of 323 bp (uploaded to GenBank: MG561387) that was blasted against the stickleback (Gasterosteus aculeatus) and cod (Gadus morhua) genomes from the Ensembl genome browser (http://www.ensembl.org). BLAST results returned a ∼90% sequence homology with the ca4a gene in both stickleback and cod (e-values were 1e−29 and 4e−9, respectively). The CDS of C. gunnari codes for a deduced protein of 103 amino acids, most closely resembling Ca4 and sharing 95% identity with N. coriiceps Ca4 (XM_010775657.1) and 67% with O. mykiss Ca4 (XP_021479942.1).
Control gene expression of ef1α did not differ between ventricles and gills of C. gunnari (P=0.338) or N. rossii (P=0.203), and the expression of ca4a is reported relative to that of the control gene in Fig. 2. The relative expression of ca4a mRNA in the gills of C. gunnari was not different from that of the ventricle (P=0.610), a tissue in which ca4 expression has been reported in other teleosts (Georgalis et al., 2006; Alderman et al., 2016), and expression values were comparable to the expression of the control gene. Likewise, in the ventricle of N. rossii, ca4a was expressed at levels comparable to the control gene; however, expression in the gills was significantly lower than that in the ventricle (P=0.044).
Biochemical analysis of CA activity
All cellular fractions obtained by differential centrifugation of gill homogenates showed significant CA activity. In both species, CA activity was highest in the supernatant containing the cytosolic fraction, compared with microsomal pellets that contain plasma membranes, and averaged over species, values were 529±50 and 99±28 µmol H+ mg−1 min−1, respectively. The effects of washing on microsomal CA activity for both notothenioids are shown in Fig. 3. Washing significantly increased CA activity in pellets of C. gunnari (when expressed per unit protein), from 17±1 to 109±9 µmol H+ mg−1 min−1 (P<0.001), whereas washing significantly decreased CA activity in pellets of N. rossi, from 181±27 to 31±2 µmol H+ mg−1 min−1 (P=0.002). However, washing significantly reduced total CA activity in both species (when expressed per volume of fraction), from 244±17 to 149±11 µmol H+ ml−1 min−1 in C. gunnari (P<0.001) and from 453±69 to 94±9 µmol H+ ml−1 min−1 in N rossii (P=0.003).
To assess whether microsomal CA isoforms were membrane bound by a GPI anchor, membrane pellets were incubated at 21°C for 90 min, in the absence (control) or presence of PI-PLC. In the control incubations of both species, CA activity per unit of protein was reduced by about half compared with initial values (67±9 and 15±12 µmol H+ mg−1 min−1 for C. gunnari and N. rossii, respectively). The effects of PI-PLC on microsomal CA activity of both notothenioids are shown in Fig. 4. Treatment of C. gunnari pellets with PI-PLC significantly decreased CA activity compared with control values, to 20±1.5 µmol H+ mg−1 min−1 (P=0.031), and a corresponding increase was observed in the CA activity of the supernatant (P=0.002). Likewise, a significant effect of PI-PLC was detected on CA activity in pellets of N. rossii, which decreased to 10±1 µmol H+ mg−1 min−1 (P<0.001); however, no significant change in CA activity was observed in the supernatant (P=0.450).
Fig. 5 shows the inhibitory effect of SDS, a surfactant, on the CA activity in cellular fractions of both notothenioids. In the pellets of C. gunnari, CA activity was unaffected by SDS (0.7±2.5% inhibition; P=0.813), whereas CA activity in the supernatant was significantly inhibited by 55.1±6.6% (P=0.003). In contrast, in N. rossii, SDS significantly inhibited CA activity in the pellets by 38.5±4.4% (P=0.003) and in the supernatant by 44.8±3.9% (P=0.001). In addition, titrations with increasing concentrations of Az resulted in inhibition constants (ki) that were significantly different (P=0.032) between CA isoforms derived from microsomal pellets or supernatants of C. gunnari gills, and the average ki was 0.74±0.11 and 1.18±0.13 nmol l−1, respectively.
As expected, RBC lysates from N. rossii had a high CA activity of, on average, 39±1 µmol H+ mg−1 min−1 (despite the high non-CA protein content of this fraction) and the inhibitory effects of 100 µl plasma from either species are shown in Fig. 6. The addition of plasma from N. rossii significantly inhibited CA activity in the RBC lysate, by 92.3±2.8% (P<0.001). Likewise, the addition of plasma from C. gunnari significantly inhibited CA activity in the RBC lysates of N. rossii, by 81.7±3.7% (P<0.001). However, the CA activity in microsomal pellets of C. gunnari gills was unaffected by the presence of endogenous plasma (0.9±1.1% inhibition; P=0.809), whereas CA activity in pellets of N. rossii was significantly inhibited by the presence of endogenous plasma, by 77.2±4.3% (P<0.001). In both species, CA activity in the supernatant was significantly inhibited by the addition of endogenous plasma, by 73.1±6.3% in C. gunnari (P<0.001) and by 90.2±3.5% in N. rossii (P<0.001).
Plasma protein concentration was significantly higher (P=0.001) in C. gunnari compared with N. rossii (18.3±1.4 and 10.8±0.3 mg ml−1, respectively). Fig. 7 shows βplasma for both notothenioids over the physiological pH range. A significant species effect in the regression analysis indicated that βplasma of C. gunnari was significantly (P<0.001) higher than that of N. rossii, and average values were 5.19±0.26 and 4.32±0.15 mmol H+ l−1 pH−1, respectively. Plasma protein concentrations in N. rossii were corrected for Hb protein from RBC lysis. Hb concentration was typically low in all samples and close to the detection limit of the assay (25 µg ml−1). However, a single sample had an elevated Hb concentration of 5.49 mg ml−1, and this sample was excluded from the analysis of βplasma (open symbols in Fig. 7A).
Taking advantage of the natural Hb-knockout model provided by Antarctic icefishes, we tested the hypothesis that, in the absence of RBC CA, icefish gills express a paCA isoform that can provide the catalytic activity necessary for CO2 excretion. To determine the cellular orientation of the putatively plasma-accessible Ca4 isoform in the gills of notothenioids (Tufts et al., 2002), gill sections of the icefish C. gunnari and the red-blooded N. rossii were immuno-labelled with an antibody raised against rainbow trout Ca4 (Gilmour et al., 2007). In gills of C. gunnari, a clear immunohistochemical signal (Fig. 1A, asterisks) placed Ca4 protein in association with the apical plasma membranes of pillar cells and the basolateral membrane of some lamellar epithelial cells. Thus, Ca4 appears to line the entire lamellar blood space, a pattern that is consistent with a plasma-accessible orientation of the enzyme and which has not been observed previously in a teleost. A similar pattern has been described in the gills of dogfish, an elasmobranch (Gilmour et al., 2007), where the presence of Ca4 has been linked to functional measurements that infer a role of the enzyme in CO2 excretion (Gilmour et al., 2001). Western analysis revealed a Ca4 protein of ∼37.5 kDa in C. gunnari that matches closely the size of dogfish Ca4 of ∼40 kDa (Gilmour et al., 2007). Our immunohistochemical finding was corroborated by the pattern of gene expression in the gills of C. gunnari, where expression of ca4a was detected at high levels, comparable to those in the ventricle (Fig. 2). This is unlike the situation in other teleosts, such as rainbow trout, where ca4 is expressed in the ventricle but not in the gills (Georgalis et al., 2006). Surprisingly, the gills of N. rossii also showed detectable expression of ca4a, albeit at a significantly lower level compared with that in the ventricle (Fig. 2), and without a corresponding immunohistochemical signal. Reactivity for the Ca4 antibody was clearly absent in the lamellar blood space of N. rossii (Fig. 1B, asterisks), but some intracellular reactivity was detected. These immunohistochemical results were later confirmed using a second antibody, raised against Ca4 in three Chondrichthyes, for which the antigenic peptide sequence of the notothenioid Ca4 was 100% conserved (data not shown). It is possible that the low expression of ca4a mRNA in gills of N. rossii is not translated into a protein, or perhaps that it is translated into a small pool of protein that is anchored to intracellular membranes and does not undergo the post-translational modifications required for export to a plasma-accessible location (Waheed et al., 1996). This issue was clarified by the biochemical characterisation of this CA pool as follows.
To characterise the CA isoform distribution in gills of C. gunnari and N. rossii, gill homogenates were fractionated by differential centrifugation and CA activity was measured in the supernatant, comprising the cytosolic fraction, and in microsomal pellets that contain plasma membranes. In both species, CA activity was highest in the supernatant versus the microsomal pellets (averaged over species 529±50 and 99±28 µmol H+ mg−1 min−1, respectively). This result is in line with immunohistochemical data showing reactivity for a soluble Ca17 protein in the gills of both species that is clearly confined to the cytosol of both pillar and lamellar epithelial cells, although more abundant in the latter (Fig. 1C,D). This prevalence of cytosolic over membrane-associated CA activity is consistent with previous findings on the CA isoform distribution in the gills of notothenioids (Maffia et al., 2001) and other fish species (Harter and Brauner, 2017) and highlights the importance of this CA pool for ionoregulation and acid–base regulation and the sensing of CO2 and pH in neuro-epithelial cells (for review, see Gilmour, 2012). However, this soluble cytoplasmic CA is not plasma accessible and thus cannot participate in plasma HCO3− dehydration and CO2 excretion.
Membrane-bound Ca4 isoforms were identified by using four common biochemical markers: (i) resistance to washing of microsomal pellets, (ii) liberation of CA by PI-PLC, (iii) resistance to SDS and (iv) resistance to plasma CA inhibitors. Washing significantly reduced the CA activity in the pellets of N. rossii but, when expressed per unit of protein, washing increased CA activity in the pellets of C. gunnari (Fig. 3). This was probably due to the washout of non-CA proteins from the microsomal fraction, and washing significantly reduced total CA activity in the pellets of both species. Importantly, after washing, the pellets of C. gunnari retained a 3-fold higher CA activity than those of N. rossii (109±9 and 31±2 µmol H+ mg−1 min−1, respectively). PI-PLC treatment significantly reduced CA activity in the microsomal pellet of C. gunnari, releasing CA activity into the supernatant (Fig. 4); this is a clear indication of the presence of a GPI membrane-bound Ca4 and/or Ca15 isoform in the icefish. A statistically significant, but numerically small effect of PI-PLC was also detected on CA activity in the pellets of N. rossii, but without a corresponding increase in supernatant CA activity. This is in line with the data of Tufts et al. (2002), who found a significant effect of PI-PLC on CA activity in microsomal pellets of C. aceratus and N. coriiceps, while other studies have found no effect of PI-PLC in non-notothenioid teleosts (Gilmour et al., 2001, 2002). In combination, these results corroborate the finding of a CA isoform that is linked to membranes by a GPI anchor in the gills of the icefishes C. gunnari and C. aceratus, and provide equivocal indications for the presence of a similar, but perhaps less abundant, isoform in red-blooded notothenioids that may be restricted to intracellular membranes or associated with epithelial cells and, thus, is membrane associated but not plasma accessible.
To further determine whether the observed CA activity in the gills of C. gunnari was derived from Ca4 protein, microsomal pellets were treated with SDS. Mammalian studies show that CA4 isoforms have two additional disulfide bonds that stabilise the enzyme against denaturation by SDS (Waheed et al., 1996) and, thus, SDS-resistant CA activity is often described as Ca4-like in fishes and other non-mammalian vertebrates (Gervais and Tufts, 1998; Gilmour et al., 2002, 2007; Stabenau and Heming, 2003; Esbaugh et al., 2009). CA activity in the pellet of C. gunnari was unaffected by SDS (Fig. 5), while CA activity was significantly reduced in pellets of N. rossii (by 38.5±4.4%). As expected, cytosolic CA activity in the supernatant of C. gunnari and N. rossii, which typically contains SDS-sensitive, soluble CA isoforms, was significantly reduced in the presence of SDS (by 55.1±6.6% and 44.8±3.9%, respectively). These findings corroborate previous data that indicate Ca4-like enzyme activity in gill membranes of the icefish C. aceratus, but not in those of N. coriiceps or in the supernatants of either species (Tufts et al., 2002).
The inhibition characteristics for Az, a common sulfonamide CA inhibitor, are well studied in mammals and allow further differentiation among CA isoforms (Baird et al., 1997). Tufts et al. (2002) found no difference between the inhibition constant (ki) for Az in pellets and supernatants of C. aceratus, indicating similar CA isoforms in the two fractions. However, in C. gunnari, the ki for Az was 0.74±0.11 nmol l−1 in gill microsomal pellets compared with 1.18±0.13 nmol l−1 in the supernatant, a significant difference, indicating that different CA isoforms are present in the two fractions. The discrepancy with previous data may indicate the presence of two isoforms with similar ki in C. aceratus, or perhaps that a low number of replicates in the earlier study (N=4; Tufts et al., 2002) was insufficient to resolve the small numerical difference observed here.
An intriguing finding was the discovery of a CA inhibitor in the plasma of the icefish, C. gunnari. In fact, the CA activity in RBC lysates from N. rossii was significantly reduced in the presence of 100 µl of plasma from either N. rossii (by 92.3±2.8%; Fig. 6) or C. gunnari (by 81.7±3.7%), providing strong evidence that both species possess a CA inhibitor in their plasma. The putative role of plasma CA inhibitors is to either inactivate or recycle CA from RBC lysis (Henry and Heming, 1998), but neither role would be relevant for icefishes that largely lack RBCs. A plasma CA inhibitor has also been described in the icefish C. aceratus (Tufts et al., 2002) and because the phylogenetic distance between C. aceratus and C. gunnari spans nearly the entire clade of Channichthyidae (Near et al., 2003), it is plausible that plasma inhibitors of CA are present in all icefishes. Whether the plasma CA inhibitor in icefishes is an evolutionary relic from a red-blooded ancestry or whether its role should include the scavenging of cytoplasmic CA shed by the lysis of other cell types remains unclear. Regardless, the presence of an endogenous plasma CA inhibitor can be used as a powerful diagnostic test for Ca4 that, in mammals, is largely unaffected by the inhibitor, and this safeguards its function in plasma-accessible locations (Hill, 1986; Heming et al., 1993). A critical finding, thus, was that CA activity in pellets of C. gunnari was unaffected by the presence of endogenous plasma (Fig. 6), indicating a Ca4 isoform that is not susceptible to inhibition by the plasma CA inhibitor in icefishes. In contrast, CA activity in pellets of N. rossii was significantly inhibited by 77.2±4.3% in the presence of endogenous plasma, indicating that this pool of CA is probably not a Ca4 isoform. Also, the supernatant of both species was significantly inhibited by plasma addition (in C. gunnari by 73.1±6.3% and in N. rossii by 90.2±3.5%), as expected for soluble CA isoforms. Thus, CA activity in membranes of C. gunnari displays Ca4-like characteristics that are not seen in membranes of N. rossii or in those fractions containing soluble CA isoforms.
As noted above, four biochemical criteria are commonly used to characterise membrane-bound Ca4: (i) resistance to washing of pellets, (ii) liberation by PI-PLC, (iii) resistance to SDS and (iv) resistance to plasma CA inhibitors. CA activity in the pellets of C. gunnari conformed to all four criteria and this was supported by the expression of ca4a mRNA at the gills and the immunohistochemical detection of Ca4 protein of the predicted size, in a subcellular location that indicates a plasma-accessible orientation. CA activity in the pellets of N. rossii was largely removed by washing and inhibited by SDS. A significant effect of PI-PLC and low levels of ca4a expression may indicate the presence of some Ca4 protein that appears to be localised to intracellular membranes. Regardless of the isoform identity, the fact that microsomal CA activity in N. rossii was susceptible to the plasma CA inhibitor prohibits this CA pool from participating in HCO3− dehydration in the plasma. In combination, these data support the hypothesis that C. gunnari possess plasma-accessible Ca4 at the gills that should catalyse CO2 excretion, while gills of N. rossii appear to lack a CA pool that could participate in this role. Thus, in the absence of RBC CA, icefish may be the only adult vertebrate in which CO2 excretion is driven exclusively by the paCA activity provided by the gill.
Why most other teleosts lack paCA activity at the gills, despite its potential benefit for CO2 excretion, is still debated. One powerful argument relates to the evolution of highly pH-sensitive Hbs that required the active regulation of RBC intracellular pH to safeguard branchial O2 uptake during a blood acidosis (Nikinmaa et al., 1984; Berenbrink et al., 2005); this protective mechanism requires an absence of CA activity in the plasma (Jacobs and Stewart, 1942; Motais et al., 1989; Rummer and Brauner, 2011). A functional relationship between paCA and RBC pH regulation is supported by lamprey, which also maintain RBC intracellular pH above equilibrium and appear to lack paCA at the gills (Henry et al., 1993), whereas chondrichthyans and tetrapods that do not regulate RBC intracellular pH have retained paCA at the gas exchange surface where it plays a role in CO2 excretion (Tufts and Perry, 1998; Stabenau and Heming, 2003). If the presence of paCA at the teleost gill was functionally constrained by the characteristics of teleost Hb and RBC function, perhaps these constraints were released in icefishes, which lack both. Assessing the presence of paCA in the gills of the closest red-blooded relatives of the Channichthyidae (the Bathydraconidae; Near et al., 2004) and confirming the absence of paCA in other notothenioid families would strengthen the functional link between the loss of Hb and the expression of paCA. In addition, teleost plasma is an unfavourable medium to support a high CA activity, mainly due to its low buffer capacity (βplasma), as HCO3− dehydration requires equimolar amounts of H+ (Bidani and Heming, 1991; Gilmour et al., 2002; Szebedinszky and Gilmour, 2002). In the presence of RBCs with fast Cl−/HCO3− exchange, an abundant pool of CA and buffers on Hb, paCA activity may be largely inconsequential for CO2 excretion in teleosts (Desforges et al., 2001). Notably, it is those fishes with the highest βplasma that also have paCA activity, conditions that, to varying degrees, contribute to CO2 excretion in Squalus acanthias (Lenfant and Johansen, 1966; Graham et al., 1990; Gilmour et al., 2001) and Eptatretus stoutii (Esbaugh et al., 2009).
The βplasma in C. gunnari was 5.19±0.26 mmol H+ l−1 pH−1, significantly higher than that in N. rossii (of 4.32±0.15 mmol H+ l−1 pH−1; Fig. 7). Previous studies that measured βplasma in other icefish species over the same pH range reported 3.4±0.2 mmol H+ l−1 pH−1 in Pagetopsis macropterus (Wells et al., 1988) and 9.7±0.9 mmol H+ l−1 pH−1 in Chionodraco hamatus (Acierno et al., 1997). While these values vary largely between studies and species, the βplasma values reported here exceed, by about 2-fold, typical teleost values (2–3 mmol H+ l−1 pH−1; Tufts and Perry, 1998), perhaps with the exception of some catfishes (Cameron and Kormanik, 1982; Szebedinszky and Gilmour, 2002). The high βplasma in the icefish may be critical to support a high CA catalytic activity at the gill, by providing the H+ necessary for HCO3− dehydration in the plasma. Not surprisingly, hagfish and dogfish also have a high βplasma that correlates with the presence of paCA at the gills, and that in dogfish contributes significantly to whole-animal CO2 excretion (Gilmour et al., 2001; Esbaugh et al., 2009). Plasma proteins in Channichthys rhinoceratus, another icefish, are rich in imidazole-based histidine, a residue capable of reversibly binding H+, which probably contributes to the high βplasma in this species (Feller et al., 1994). Similarly, histidine-rich proteins (in this case albumins) appear to underlie the unusually high βplasma in Ameiurus nebulosus, a catfish (Szebedinszky and Gilmour, 2002). In fact, plasma protein concentration in C. gunnari was 18.3±1.4 mg ml−1 and significantly higher than that in N. rossii (10.8±0.3 mg ml−1), a finding that may correlate with the higher βplasma in the icefish. The protein concentration in N. rossii plasma conforms with the range typically reported in teleosts (Acierno et al., 1997); however, values in C. gunnari are lower compared with those in other icefishes studied (Egginton, 1994; Acierno et al., 1997; Feller and Gerday, 1997); the reason for this discrepancy is unknown.
Because of the large blood volume of icefishes, about 7.6% of body mass (Hemmingsen and Douglas, 1970) compared with 2–3% in other teleosts (Thorson, 1961; Houston and DeWilde, 1969), and their low Hct (<1% compared with >25% in other teleosts; Holeton, 1970), the total volume of plasma in icefishes is at least 3 times higher than that in most teleosts (Feller et al., 1994). The plasma of red-blooded teleosts contributes 20–40% to whole blood buffer capacity, typically <10 mmol H+ l−1 pH−1, which is largely determined by the buffer capacity of Hb (Wood et al., 1982; Tufts and Perry, 1998; Gilmour et al., 2002; Szebedinszky and Gilmour, 2002). Although the measured βplasma in C. gunnari is only half that of typical teleost whole blood, this is clearly overcompensated by the 3-fold higher plasma volume of icefishes. Thus, per unit of animal mass, icefishes have a greater capacity to buffer metabolically produced H+ in their blood compared with most teleosts, despite lacking Hb. In combination with a low metabolic rate (Hemmingsen et al., 1969), and hence a lower release of CO2 and H+ to the plasma, βplasma in icefishes would seem adequate to sustain arterial–venous pH homeostasis and HCO3− dehydration (as supported by experimental data; Hemmingsen and Douglas, 1972), which is catalysed by the paCa4 isoform at the gill.
The evolutionary time course over which RBCs were lost from the circulation in the common ancestor of Channichthyidae, and whether this coincided with the loss of transcriptionally active Hb genes, is presently unknown. However, Hb is the largest H+ buffer within the RBC cytosol (in teleosts largely through the Bohr–Haldane effect) and the absence of Hb will have severely restricted the functional significance of RBC CA. Thus, the time course over which icefishes had to acquire paCA at the gill, to compensate for the reduction of RBC CA function, may have corresponded closely to the loss of Hb. The molecular mechanism by which icefishes catalyse HCO3− dehydration in the plasma is analogous to that in all other non-teleost vertebrates, where Ca4 is GPI anchored to the apical membrane at the gas exchange organs. Therefore, it seems likely that paCA was never ‘lost’ at the teleost gill, but functional constraints related to the pH sensitivity of teleost Hb prevented a significant expression of the trait, until the loss of Hb in icefishes simultaneously released functional constraints and created a need to catalyse HCO3− dehydration in the plasma. Possible scenarios include: (i) natural selection favoured phenotypes with higher paCA activity at the gill, which requires that there was standing variation in this trait in the common ancestor of Channichthyidae; (ii) phenotypic plasticity induced an up-regulation of ca4a gene expression at the icefish gill, which may be supported by the presence of the transcript in N. rossii; or (iii) neoteny allowed for branchial paCA to be retained throughout icefish ontogeny, a mechanism that underlies other adult characters in notothenioids (Montgomery and Clements, 2000), and which would place branchial paCA as an embryonic trait in teleosts – a scenario that could be tested experimentally.
In conclusion, the natural knockout of Hb in Antarctic icefishes had profound consequences for cardiovascular O2 transport and resulted in fascinating adaptations that compensate for the reduction in O2-carrying capacity of the blood. In addition, results from the present study show that the reduction of RBCs and the associated loss of CA catalytic activity in the blood of icefishes led to a divergent strategy of CO2 excretion. While paCA is functionally absent at the gills of all other teleosts, icefishes may have re-expressed this trait and, unlike the situation in any other vertebrate studied to date, in icefishes, the CA catalytic activity required for CO2 excretion is provided exclusively by the gills. Therefore, the study of Antarctic icefishes may reveal a previously unidentified evolutionary plasticity in the vertebrate CO2 excretion pathway and perhaps provides a framework to address more general questions on the evolutionary dynamics of vertebrate gas exchange.
We thank Kristin O'Brien and Lisa Crockett for generously collecting the samples for us and for valuable comments on the manuscript, and Elizabeth Evans for assistance in sample collection. We thank the Masters and crew of the ARSV Laurence M. Gould and the support staff at the US Antarctic Research Station, Palmer Station. Finally, thanks are due to Rick Taylor and Christian Damsgaard for providing critical feedback on the manuscript.
Conceptualization: T.S.H., M.A.S., A.P.F., C.J.B.; Methodology: T.S.H., J.M.W., D.C.M., A.J.E.; Validation: T.S.H.; Formal analysis: T.S.H., M.A.S., J.M.W., D.C.M., A.J.E.; Investigation: T.S.H., M.A.S., J.M.W., D.C.M., A.J.E.; Resources: A.P.F., C.J.B.; Data curation: T.S.H.; Writing - original draft: T.S.H.; Writing - review & editing: T.S.H., M.A.S., J.M.W., D.C.M., S.E., A.J.E., A.P.F., C.J.B.; Visualization: T.S.H., J.M.W.; Supervision: C.J.B.; Project administration: C.J.B.; Funding acquisition: C.J.B.
This study was supported by Natural Sciences and Engineering Research Council of Canada (NSERC) Accelerator Supplement (446005-13) and Discovery Grant (261924-13) to C.J.B.
The partial coding sequence for C. gunnari ca4a (323 bp) has been uploaded to GenBank: MG561387.
The authors declare no competing or financial interests.