An investigation of the role of carbonic anhydrase in aquatic and aerial gas transfer in the African lungfish Protopterus dolloi

Among the numerous species of air-breathing fishes, only three extant genera of lungfish (Neoceratodus, Lepidosiren and Protopterus) exploit true lungs for aerial gas transfer(Graham, 1997). The Australian lungfish Neoceratodus forsteri is a facultative air-breather whereas the South American lungfish Lepidosiren paradoxa and the African lungfish (genus Protopterus) are obligate air breathers that cannot survive without access to air (Burggren and Johansen, 1986; Graham,1997). The four species of Protopterus share, along with their capacity for pulmonary gas transfer, the ability to withstand extended conditions of water deprivation by entering into a state of aestivation(Fishman et al., 1986). Typically, the aestivating state is characterized by formation of a mucus cocoon, metabolic depression, total reliance on aerial gas transfer and an increase in hepatic ureagenesis (Smith,1930; Fishman et al.,1986; Greenwood,1986).

Although the lung is the primary site of O₂ uptake in Protopterus, the gill (and/or the skin) is believed to be the major route for CO₂ excretion(Lenfant and Johansen, 1968; Lahiri et al., 1970; McMahon, 1970; Babiker, 1979). Thus, even though African lungfish are bimodal breathers, the gill appears to be unimportant in O₂ uptake and the lung relatively unimportant in CO₂ excretion. The reduced reliance on the lung for CO₂excretion under resting conditions is also a feature of Australian(Lenfant et al., 1966) and South American (Johansen and Lenfant,1967) lungfish. In comparison to most fish, the gills of lungfish have small surface areas and large blood-to-water diffusion distances, and are ventilated by low volumes of water(Burggren and Johansen, 1986). Because of the much higher (∼30-fold) capacitance of the water for CO₂ relative to O₂, the low ventilation volumes may be adequate to sustain normal rates of CO₂ excretion but inadequate for similar rates of O₂ uptake(Graham, 1997). It was recently demonstrated that the overall contribution of the lung of Lepidosiren to CO₂ excretion can increase markedly with increasing metabolic rate (Amin-Naves et al., 2004), indicating that the lungfish lung can effectively excrete CO₂ under the appropriate conditions.

CO₂ transfer across the teleost gill has been demonstrated to behave as a diffusion limited system(Bindon et al., 1994; Greco et al., 1995; Julio et al., 2000) owing to chemical equilibrium limitations (Julio et al., 2000; Desforges et al.,2002). Essentially, the chemical equilibrium limitations reflect the relatively slow rate of conversion of blood HCO₃^- to CO₂ during the brief period of gill transit. Indeed, recent models for branchial gas transfer in teleosts(Perry and Gilmour, 2002)suggest that CO₂ transfer is more apt to be affected by diffusion limitations than is O₂ transfer. Thus, the fact that the lungfish gill is likely to display pronounced diffusion limitations implies that there might be other factors (in addition to low ventilation volumes) to explain why the reduced lungfish gill does not transfer O₂ while excreting CO₂ effectively.

In the present study, two possible mechanisms for the preferential excretion of CO₂ by the gill of African lungfish were investigated. First, we hypothesised that, as in dogfish, extracellular branchial membrane-associated carbonic anhydrase (CA) aids gill CO₂ excretion by allowing the catalysed dehydration of HCO₃^- within the plasma (Gilmour et al.,2001; Gilmour and Perry,2004). Assuming a high enough buffer capacity in the plasma(Desforges et al., 2001; Gilmour et al., 2004), this would allow CO₂ excretion to occur without a requirement for red blood cell (RBC) Cl^-/HCO₃^- exchange, which is the rate-limiting step in CO₂ excretion(Perry, 1986) and the origin of chemical equilibrium limitations(Gilmour et al., 2004). To test this hypothesis, aerial and aquatic gas transfers were measured before and after inhibition of extracellular CA using a relatively impermeant CA inhibitor, benzolamide (BZ). A second possible explanation for the preferential role of the gill in CO₂ excretion is that the high P_CO2 that is characteristic of lungfish blood(Lenfant and Johansen, 1968)obviates the need for catalysed dehydration of plasma HCO₃^-, because blood-to-water P_CO2 partial pressure gradients are large enough to sustain CO₂ excretion by diffusion alone. While potentially effective at the gill, a similar scenario might be less likely at the lung owing to a much smaller blood-to-gas diffusion gradient, reflecting the loss of CO₂ at the gill and the mixing of inspired and expired gas within the lung. To test this hypothesis, aerial and aquatic gas transfers were measured before and after inhibition of total CA activity using a permeant CA inhibitor, acetazolamide (AZ). In a final series of experiments,fish were injected with exogenous CA to test the hypothesis that CO₂ excretion in Protopterus is constrained by inadequate catalysis of HCO₃^- dehydration. All experiments were performed on Protopterus dolloi, a species that, to our knowledge,has not been used previously in studies investigating the mechanisms of gas transfer in African lungfish.

Experiments were performed both in Singapore (National University of Singapore) and Ottawa (University of Ottawa) using lungfish derived from a common source. Adult specimens of African lungfish Protopterus dolloiBoulenger, weighing 159±9 g [mean ± standard error of the mean(s.e.m.), N=25], were imported from Central Africa through a local fish farm in Singapore. Specimens were maintained in plastic aquaria filled with dechlorinated water containing 2.3 mmol l^-1Na⁺, 0.54 mmol l^-1 K⁺, 0.95 mmol l^-1 Ca²⁺, 0.08 mmol l^-1 Mg²⁺, 3.4 mmol l^-1 Cl^- and 0.6 mmol l^-1 mol l^-1 HCO₃^-, at pH 7.0 and at 25°C in the laboratory, and the water was changed daily. No attempt was made to separate the sexes. The specimens were acclimated to laboratory conditions for at least 1 month. During the acclimation period, specimens were fed frozen mosquito larvae (blood worms). Food was withdrawn 96 h prior to experiments, which gave sufficient time for the gut to be emptied of all food and waste.

Fish were shipped by air cargo in partially filled bags of oxygenated water contained within insulated containers to Ottawa. Upon their arrival, fish were placed individually into covered plastic containers containing 2–3 l of dechloraminated city of Ottawa tapwater warmed to 25°C. To maintain this water temperature, the air temperature was held at 25°C. The tank water was changed on alternate days or sooner if there was obvious fouling of the medium. The fish were kept in a sealed room under an artificial photoperiod of 10 h:14 hlight:dark. Fish were fed on alternate days with frozen blood worms or pieces of rainbow trout flesh. Fish were allowed to acclimate to these conditions for at least 1 month prior to beginning experiments.

Fish were anaesthetized in a solution of MS-222(ethyl-p-aminobenzoate; 1.0 g l^-1) adjusted to neutral pH with NaHCO₃ (2 g l^-1). After cessation of breathing movements, the fish were transferred to an operating table where they were draped with paper towels soaked with anaesthetic solution. In this way, they were kept moist and deeply anaesthetized for the duration of the surgery. To allow periodic blood sampling, a cannula (Clay-Adams PE 50 polyethylene tubing) was inserted into the dorsal aorta (DA) according to standard surgical procedures (Axelsson and Fritsche,1994). Briefly, a lateral incision (∼2 cm in length) was made at the level of the vent approximately 3 mm below the lateral line. The DA was exposed and the cannula was inserted via a small incision and advanced at least 3 cm in the anterior direction. The incision was sutured using a running stitch and the cannula was then secured to the body wall with silk ligatures, filled with heparinized (100 units ml^-1) saline(140 mmol l^-1 NaCl), and heat-sealed. Fish were retuned to their containers where they were allowed to recover from surgery for ∼24 h.

Fish were placed into customized respirometry chambers approximately 2 h prior to beginning an experiment. The chambers were filled with water (1.5 l)except for an adjustable (60 ml maximum volume) air space at one end of the respirometer. Once aware of the presence of the air space, the fish would typically position themselves just underneath it and begin breathing air at regular intervals. Initially, both compartments were provided with continually flowing media (water or air). However, during measurements of rates of O₂ consumption(Ṁ_O2) or CO₂ excretion(Ṁ_CO2), the water was recirculated using a peristaltic pump and the air chamber was sealed. O₂ and CO₂ electrodes inserted into the air chamber were used to measure aerial P_O2 and P_CO2. To measure aquatic P_CO2, water was pumped via a peristaltic pump into the sample compartment of a CO₂ electrode(Cameron Instruments, Port Aransas, TX, USA) housed within a thermostatted(25°C) cuvette (Radiometer, Copenhagen, Denmark) and returned to the respirometer. To measure aquatic P_O2, an O₂ electrode was inserted into the recirculating water loop.

After establishing that the fish was positioned correctly under the air chamber and breathing air at regular intervals, the flows of freshwater and air were stopped, the water and air chambers were sealed, and the water was re-circulated to provide mixing. Changes in aerial and aquatic P_CO2 and P_O2 as well as air breathing frequency were monitored for approximately 30 min or until stable rates of CO₂ accumulation and O₂ depletion were achieved. Preliminary experiments were performed to determine the extent of gas transfer across the air–water interface. To determine the potential for aerial CO₂ to be transferred to the water and thus be misinterpreted as CO₂ excretion into the water, mixtures of CO₂ and air were added incrementally to the air chamber. After sealing the air chamber, the water P_CO2 was monitored without a fish in the respirometer. To determine the potential for aquatic CO₂ to be transferred to the air space and thus be misinterpreted as aerial CO₂ excretion, fish were euthanized by lethal injection while in the respirometer after aerial and aquatic P_CO2 had been allowed to rise naturally to high levels. The air-chamber was then flushed with air to lower P_CO2 and after re-sealing the chamber, the P_CO2 of the air was monitored while the water P_CO2 remained at a high level.

Initial experiments revealed substantial Ṁ_O2 and Ṁ_CO2 in the absence of a fish in the respirometer, or during trials using recently euthanized fish. This background gas transfer presumably reflected cutaneous metabolism or metabolism arising from microorganisms within the water or attached to the fish. Thus, all experiments were terminated by euthanizing the fish (followed by intra-arterial injection of saturated KCl to stop the heart) in situ and performing a final respirometry trial for a further 30–60 min. For each fish, the values for Ṁ_O2 and Ṁ_CO2 were corrected for background metabolism by subtracting the rates determined on such trials.

The standard experimental protocol consisted of monitoring water/air P_O2 and P_CO2 in addition to air breathing frequency for three consecutive 30 min intervals. The first interval consisted of a control period, after which a blood sample(0.6 ml) was withdrawn from the DA cannula and P_O2, P_CO2 and pH immediately analyzed. The remaining blood was centrifuged (1 min × 10 000 g) and the pellet was combined with 0.6 ml of distilled water prior to being flash-frozen in liquid N₂. Red blood cell lysates were stored at –80°C for subsequent analysis of CA activity.

After the control period, the air chamber was flushed with air and re-sealed; the water compartment was occasionally flushed and re-sealed depending upon the extent of CO₂ accumulation during the control run. Once baseline rates of O₂ and CO₂ transfer had been re-established, fish were injected via the DA cannula with BZ (1.2 mg kg^-1 using an injection volume of 1.0 ml kg^-1), a CA inhibitor that is relatively slow to cross RBC membranes(Travis et al., 1964). Respirometry was performed for 30 min and blood samples were withdrawn at 10 and 30 min and treated as described above. After flushing and re-sealing, a final 30 min respirometry period ensued following injection of AZ (30 mg kg^-1 using an injection volume of 1.0 ml kg^-1); blood samples were withdrawn at 10 and 30 min. Fish were then euthanized to determine background Ṁ_O2 and Ṁ_CO2. In a separate group of fish (N=5), gas transfer and blood gases were monitored for 2 and 24 h, respectively, after AZ injection.

After a 30 min control period of respirometry and blood sampling, fish were injected via the DA cannula with bovine CA (25 mg kg^-1using a 0.5 ml injection volume) dissolved in 140 mmol l^-1 NaCl. A second 30 min period of respirometry was conducted, after which a final blood sample was withdrawn. The blood samples in this series were treated as above except that a RBC lysate was not prepared; instead, the plasma was collected,frozen in liquid N₂ and stored at –80°C for subsequent determination of plasma CA activities.

Aquatic hypercapnia was established by replacing the air supplying a water/gas equilibration column with mixtures of CO₂ and air provided by a gas mixing pump (Wösthoff, Bochum, Germany). The desired water P_CO2 of 40 mmHg was pre-set by adjusting the rate of water and/or gas flow through the column. To rapidly achieve aquatic hypercapnia within the respirometer, the bulk of water was removed and replaced with water derived from the water/gas equilibrium column. After 30 min, aquatic normocapnia was re-established and 30 min of aerial hypercapnia was imposed by gassing the air chamber with a mixture of 5% CO₂ and 95% air to yield a final nominal inspired P_CO2of 37.5 mmHg. Owing to the very high P_CO2gradients between air and water, which resulted in transfer of CO₂across the air–water interface, no attempts were made to measure Ṁ_CO2 during aquatic or aerial hypercapnia.

Blood samples (0.6 ml) were withdrawn from the DA cannula after 30 min of normocapnia, aquatic hypercapnia and aerial hypercapnia. These samples were analysed for pH and total CO₂(C_CO2).

In series 1–3 experiments, arterial blood P_O2 (Pa_O2) was measured using an Ocean Optics (Dundedin, FL, USA) fibre-optic O₂sensing system. A fibre-optic O₂ sensor (FOXY AL300; Dundedin, FL,USA) was inserted into a blood-filled syringe. Arterial blood P_CO2 (Pa_CO2) was measured by injecting blood into the sample compartment (maintained at 25°C) of a P_CO2 (Cameron Instruments)electrode that was connected to a three-channel blood gas analyser (Cameron Instruments). Blood pH (pHa) was determined using a blood gas analyser(Medica, Bedford, MA, USA) that was thermostatted to 37°C. The measured pH was then adjusted for the temperature differential (12°C) between the analyser and fish using a correction factor of –0.018 pH units°C^-1 (Reeves,1977).

In series 1–3 experiments, Pa_O2, Pa_CO2 and pHa were monitored by injecting blood into the sample compartments (maintained at 25°C) of P_O2, P_CO2(Cameron Instruments) and pH (Metrohm combination 6.0204.100; Herizau,Switzerland) electrodes that were connected in series; the electrodes were linked to a three-channel blood gas analyser (Cameron Instruments). In series 4 experiments, plasma pH was measured using a Cameron pH electrode contained within a temperature controlled chamber and connected to a Radiometer blood gas analyser. Plasma C_CO2 was determined in duplicate on 50 μl samples using a Capnicon V total CO₂ analyzer(Cameron Instruments). Plasma P_CO2 and[HCO₃^-] were calculated using the Henderson–Hasselbalch equation and physiochemical constants for trout plasma (Boutilier et al.,1984).

In both Singapore and Ottawa, the P_CO2 and P_O2 within the air compartment were monitored continually by inserting and sealing CO₂ (Cameron Instruments) and fibre-optic O₂ (Ocean Optics AL300) electrodes into the air space of the respirometer. The P_O2 of the water(Pw_O2) was monitored continually by inserting a fibre-optic O₂ electrode (Ocean Optics AL300) into the tubing through which water was recirculating. The Pw_O2 was monitored by moving water via a peristaltic pump through the sample compartment of a CO₂electrode (Cameron Instruments) housed within a thermostatted (25°C)cuvette (Radiometer) before returning it to the respirometer. The P_CO2 electrodes were connected to a second three-channel blood gas analyzer (Cameron Instruments) that was customized to accept two CO₂ inputs. Output from the blood gas analyzers was converted to digital data and stored by interfacing with a data acquisition system (Biopac Systems Inc., Goleta, CA, USA) using Acknowledge™ data acquisition software (sampling rate set at 30 Hz) and a Pentium™ PC. Output from the fibre-optic O₂ electrodes was displayed using Ocean Optics software and the data were compiled as text files for later importation into spreadsheet software. The O₂ electrode for water measurements was calibrated by immersing the electrode in a zero solution (2 g l^-1 sodium sulphite) or air-saturated water, until stable readings were recorded. The O₂ electrode for air measurements was calibrated in situ by flowing humidified N₂ gas (zero) or air continuously through the air chamber. The CO₂ electrodes for water and air measurements were calibrated in a similar manner using mixtures of 0.5 and 1.0% CO₂ in air that were provided by a gas mixing flowmeter(GF-3/MP, Cameron Instruments). The pH electrode was calibrated using precision buffers; all electrodes were calibrated prior to each experiment.

Aerial gas transfer was determined from the slopes of the relationships between inspired gas tensions and time, over the period that the air chamber was sealed; the solubility coefficients of O₂ and CO₂ in air at 25°C were obtained from Boutilier et al.(1984). Rates of aquatic gas transfer were determined using the slopes of the relationships between water gas tensions and time over the interval that the respirometer was re-circulating. The solubility coefficient for O₂ in water at 25°C was obtained from Boutilier et al.(1984). The capacitance coefficient of CO₂ in water at 25°C was determined experimentally to be 0.041 μmol l^-1 mmHg^-1. This was accomplished by measuring the changes in water total CO₂concentration (Cameron Instruments Capnicon V total CO₂ analyzer)over the range of P_CO2 values encountered in the respirometry trials. Air breathing frequency was assessed visually throughout each experiment.

The CA activity of plasma or RBC lysates was measured using the electrometric ΔpH method of Henry(1981). In brief, samples containing CA activity were added to 6 ml of reaction buffer (in mmol l^-1, 225 mannitol, 75 sucrose, 10 Tris base, adjusted to pH 7.40 with 30% phosphoric acid) held in a temperature-controlled (4°C) vessel,and the reaction was initiated by the addition of 200 μl of CO₂-saturated water. The rate of change of pH was measured over a fall of approximately 0.15 pH units. To obtain the true catalyzed rate, the uncatalysed rate (addition of CO₂-saturated water in the absence of any CA source) was subtracted from the observed rate. A pH electrode(Radiometer GK2401C or Metrohm 6.0204.100) and PHM 84 pH meter (Radiometer)linked to a data acquisition system (Biopac Systems Inc.) were used to measure pH.

In general, 50 μl samples of plasma or diluted RBC lysate were used to assay CA activity. CA activity in RBC lysates was also titrated with increasing volumes of 5 μmol l^-1 acetazolamide (AZ). The resulting data were plotted according to the presentation of Easson and Stedman (1937) to yield an inhibition constant (K_i) for lungfish RBC CA against acetazolamide.

To determine levels of CA inhibitors in RBCs, lysates were thawed and then heated to 100°C for 10 min to denature endogenous CA. After heating, the samples were spun at 1000 g for 15 min and the supernatant was assayed for inhibitor content using the micro-method of Maren(1960), in which a fixed amount of human RBC CA is used to test the inhibitory activity of an unknown sample. Known concentrations of the appropriate inhibitor were used to generate a standard curve of catalysis time vs inhibitor concentration and this was used to back-calculate the RBC inhibitor concentration in vivo, taking into account appropriate dilution factors.

The data are reported as means ± 1 standard error of the mean(s.e.m.). Treatments were compared using one-way repeated measures analysis of variance (RM ANOVA) or paired Student's t-tests (when two means were compared). If significant differences (P<0.05) were found for multiple comparisons, a post-hoc multiple comparisons test(Bonferroni t-test) was applied.

Preliminary experiments were performed to ensure that the measured rates of aerial and aquatic gas transfer were not influenced by artefacts arising from diffusion of gases across the air–water interface of the respirometer;representative data are depicted in Fig. 1. Increasing aerial P_CO2 in discrete steps to mimic the breath-by-breath addition of CO₂ was without effect on water P_CO2 during a 30 min period of simulated respirometry (Fig. 1A). These data demonstrated that there was likely to be negligible CO₂ transfer from air to water. Fig. 1B presents the results of a typical experiment designed to assess the potential for contamination of the air with CO₂ derived from the water. While the fish was alive there were linear increases in both air and water P_CO2. After euthanasia and flushing of the air chamber, the water P_CO2 continued to rise yet air P_CO2 remained constant, indicating insignificant transfer of CO₂ from water to air. The continued rise in water P_CO2 after the death of the fish indicated that factors other than fish metabolism per se were contributing to aquatic CO₂ accumulation during respirometry trials. Fig. 1C illustrates that a similar situation applied to O₂ depletion from the water. While the fish was alive, the rates of P_O2decline in water and air were linear and roughly equivalent. After its death,the P_O2 of the water continued to decrease whereas the air P_O2 was stable.

The data shown in Fig. 2summarise the partitioning of gas transfer between the air (via the lung) and water (via the gills and/or skin). The lung was responsible for 91.0±2.9% of Ṁ_O2 and 76.0±6.6% of Ṁ_CO2(N=11). Mean respiratory exchange ratios RER(Ṁ_CO2/Ṁ_O2)for total gas transfer (sum of air plus water) and aerial gas transfer were 0.56±0.03 and 0.45±0.02, respectively. It was not possible to calculate RERs for aquatic gas transfer using data from individual fish because in many cases, there was no gas transfer. However, using the mean data(Fig. 2A), the RER for aquatic gas transfer was 1.63.

Casual observation suggested that air-breathing frequency tended to be higher in fish held in respirometers compared to their holding aquaria. Furthermore, it was apparent that in some fish, air-breathing frequency varied greatly during a single experiment. Several approaches were used to determine whether the partitioning of CO₂ transfer between water and air was influenced by breathing frequency. First, regression analyses revealed that the relative proportion of CO₂ excreted via the gills/skin was unrelated to pulmonary breathing frequency between 6 and 48 breaths min^-1 (r²=0.03; P>0.05; N=19). Second, we were able to exploit several instances where fish dramatically changed their pulmonary breathing frequency. Fig. 3 illustrates two such cases where air-breathing frequency was suddenly decreased or increased. A sudden decline in air breathing frequency(Fig. 3A) was accompanied by a marked reduction in aerial CO₂ excretion, yet the rate of CO₂ accumulation in the water was unchanged. Conversely, a sudden increase in air breathing frequency (Fig. 3B) was accompanied by an obvious increase in aerial CO₂ excretion without any change in aquatic CO₂accumulation.

Analysis of RBC lysates demonstrated that the levels of BZ in RBCs were undetectable (detection limit=45 nmol l^-1) in BZ-injected fish(Table 1). In agreement, RBC CA activity was unaffected within 30 min(Table 1). The low sample sizes for these analyses (N=2–3) reflect the loss of some samples that thawed during shipment between labs. Blood acid–base status, air breathing frequency and the rates of aerial and aquatic gas transfer were unaffected by BZ treatment (Table 2; Fig. 4). The measured levels of AZ in RBCs were 0.4–1.5 mmol l^-110–30 min after injection of AZ(Table 1). These values are several orders of magnitude greater than the K_i that was determined for P. dolloi RBC CA (2.1±0.2 nmol l^-1; N=3). Consistent with these data, the injection of AZ resulted in essentially complete inhibition of RBC CA activity within 10 min(Table 1). Blood acid–base status, air-breathing frequency and the rates of aerial and aquatic gas transfer were unaffected for 30 min after AZ treatment(Table 2; Fig. 4). In a separate group of fish (N=5), Ṁ_CO2 was monitored for 2 h after AZ treatment; total (air plus water) Ṁ_CO2 was 0.37±0.12 and 0.35±0.06 mmol kg^-1 h^-1before and 2 h after AZ, respectively.

Injection of bovine CA caused a marked increase in plasma CA activity at 30 min from 39±36 to 9642±2179 μmol CO₂ml^-1 min^-1 (N=5). The mean CA activity in the plasma of pre-injected fish reflected an unusually high value in a single sample (184 μmol CO₂ ml^-1 min^-1); in three of five samples, CA activity was undetectable. The rise in plasma CA activity was associated with a slight, but statistically significant, decrease in Pa_CO2 and an increase in breathing frequency(Table 3). Although Ṁ_O2 was unaffected, CA treatment was associated with a significant increase in Ṁ_CO2 that largely resulted from increased CO₂ excretion by gills/skin(Fig. 5). The apparent rise in Ṁ_CO2 at the lung was not statistically significant (P=0.641).

Aerial hypercapnia (37.5 mmHg) caused a marked respiratory acidosis within 30 min. Pa_CO2 increased from 12.1±0.5 to 20.9±1.7 mmHg (N=6) and pHa decreased from 7.57±0.02 to 7.41±0.03. Aquatic hypercapnia (40 mmHg) did not affect arterial blood acid–base status.

Accepted models of gas transfer partitioning in lungfish (e.g. Burggren and Johanssen, 1986)identify the gill/skin as the major route of CO₂ excretion and the lung as the principal route for O₂ uptake. This general consensus is based on the results of several prior studies that examined the three genera of extant lungfish including the African genus, Protopterus.To our knowledge, however, only a single species of Protopterus(P. aethiopicus) has been studied. The upper estimates of aerial CO₂ excretion in P. aethiopicus are that approximately 30%of total CO₂ transfer occurs via this route (reviewed by Burggren and Johansen, 1986). Recently, however, Amin-Naves et al.(2004) demonstrated that the lung of the South American lungfish Lepidosiren paradoxa, while excreting only a small fraction of the total CO₂ under normal conditions, became the major site of CO₂ excretion when animals were acclimatised to a 20°C increase in temperature. Thus, under conditions of increased metabolic rate, the lungfish lung may play a more significant role in overall CO₂ excretion.

Although the principal aim of the present study was to investigate the potential roles of CA in aerial and aquatic CO₂ transfer, it quickly became apparent that gas transfer partitioning in P. dolloiwas markedly different than in other lungfish because the lung appeared to be the major site of both O₂ and CO₂ transfers. In this respect, P. dolloi resembled Lepidosiren acclimated to elevated ambient temperatures (Amin-Naves et al., 2004). Owing to the pronounced differences between this and previous studies on African lungfish, a series of experiments was performed to validate the basic respirometry protocol.

Several comprehensive reviews have been published that discuss some common problems associated with performing respirometry on aquatic animals(Steffensen, 1989; Kaufmann and Forstner, 1989; Cech, 1990). Arguably, the most serious problem is the potential for modification of water P_O2 and P_CO2 by bacteria or other extraneous sources. The potential for bacterial metabolism to contribute to the measured changes in gas partial pressures is greatest in recirculating respirometers at high water temperatures(Kaufmann and Forstner, 1989). Because antibiotics are not recommended to eliminate bacterial metabolism during respirometry trials (Kaufmann and Forstner, 1989), the protocol employed in the present study was to subtract background metabolism during blank runs conducted using dead fish. Thus, we were unable to distinguish microbial metabolism in the water, per se, from microbial metabolism occurring on the surface of the fish. Clearly, however, bacterial metabolism was a major source of O₂depletion and CO₂ accumulation in the water during closed system respirometry. Indeed in many cases, there were no differences in the rates of water gas changes after euthanasia. Because all blank runs were performed with a fish in the respirometer, it is conceivable that cutaneous metabolism, in addition to microbial activity, may have contributed to the background metabolism. Regardless of its source, the true rates of aquatic Ṁ_O2 and Ṁ_CO2 could only be determined after taking into account these significant background components. It is unclear whether previous respirometry studies on P. aethiopicus utilised corrections for background metabolism.

Closed system respirometry on bimodal breathers is particularly challenging because of the potential for gases to diffuse across the air–water interface of the respirometer (Graham,1997). For example, the entry of exhaled CO₂ from the air chamber into the water, if occurring, could mistakenly be attributed to CO₂ excretion across the gills/skin. Because of the high capacitance of water for CO₂, the potential for contamination of the water with exhaled CO₂ is likely to be significantly greater than that for O₂ entry from water into the air. In the present study, we were unable to detect any transfer of gas across the air–water interface. Regardless, to minimize diffusion gradients between the air and water, the chambers were flushed at regular intervals.

The results obtained in this study clearly implicate the lung as the primary route of both O₂ and CO₂ transfer in P. dolloi, but also demonstrate that the gill/skin is more suited to CO₂ excretion than to O₂ uptake. The former finding is in marked contrast to conclusions derived from previous studies on African lungfish (Lenfant and Johansen,1968; McMahon,1969; Lahiri et al.,1970). Because the data demonstrating the lung as the principal site of CO₂ excretion were unexpected and in opposition to accepted models, we considered it important to corroborate the respirometry data using independent methods. This was accomplished by assessing the impact of aerial vs aquatic hypercapnia on blood acid–base status. In support of the respirometry data, only aerial hypercapnia elicited an arterial blood respiratory acidosis. This result plainly indicated that pulmonary CO₂ excretion was being inhibited and that the gill was ineffective at completely clearing the accumulating CO₂. However, the fact that Pa_CO2 remained lower than inspired P_CO2 may have reflected an increased involvement of branchial CO₂ excretion. Alternatively, the arterial blood may not have yet reached equilibrium with inspired P_CO2. The most likely explanation for the absence of an effect of aquatic hypercapnia on blood acid–base status is that inhibition of the small component of CO₂ excretion occurring at the gills/skin or any CO₂ entry from the water was simply compensated for by increased pulmonary CO₂ excretion.

As reported in other studies (e.g. Johansen and Lenfant, 1968), pulmonary breathing frequency was highly variable within, and between, individuals. Moreover, while not quantified, it appeared that the average breathing frequency (∼20 min^-1) during respirometry trials was higher than that in fish being held in aquaria. Given previous reports that pulmonary hyperventilation in P. aethiopicus is associated with a lowering of Pa_CO2(Lahiri et al., 1970) or increased pulmonary Ṁ_CO2(McMahon, 1970), we considered the possibility that CO₂ excretion into the water was artificially low in the present experiments because of pulmonary hyperventilation. This is unlikely, however, based on the absence of any correlation in this study between breathing frequency and aquatic CO₂ excretion over a wide range (6–48 breaths min^-1). Moreover, there were no obvious changes in the rates of CO₂ accumulation in the water during periods of pulmonary apnoea.

Although CO₂ transfer into the water comprised only a minor component of overall CO₂ excretion, the relative proportion of aquatic CO₂ transfer (24% of overall CO₂ excretion) was greater than for aquatic O₂ transfer (9% of overall O₂uptake). These data confirm previous conclusions that the gill/skin of lungfish is more suited to CO₂ excretion than O₂ uptake(see reviews by Burggren and Johansen,1986; Graham,1997). The possibility that extracellular catalysis of plasma HCO₃^- dehydration underlies this phenomenon is discussed below.

It would be presumptuous to assume that previous conclusions concerning gas transfer partitioning in lungfish are incorrect. Discrepancies between the present, and other studies, may simply reflect species differences. Why P. dolloi should differ from other species of Protopterus, however,is unclear. Alternatively, by analogy to Lepidosiren(Amin-Naves et al., 2004), it is possible that the fish used in the present study had increased metabolic rates (compared to fish used in previous studies) and hence were more reliant on the lung as a route for CO₂ excretion. Interestingly, we (S. F. Perry, R. Euvermann, S. F. Chew, Y. K. Ip and K. M. Gilmour; unpublished data)have recently completed a series of experiments on P. annectens and found that, as in previous reports, the gill was the predominant site of CO₂ excretion whereas the lung was the major site of O₂uptake.

Initially, the present study was conceived to test the idea that the reputed preferential excretion of CO₂ by the lungfish gill(compared to the lung) might reflect the participation of extracellular branchial CA. As is the case in dogfish(Wood et al., 1994; Henry et al., 1997; Gilmour et al., 2001; Gilmour and Perry, 2004; Perry and Gilmour, 2002), it was hypothesized that a membrane-associated plasma-facing CA isoform could participate in the dehydration of plasma HCO₃^-, thereby aiding CO₂ excretion. Such a scheme might be particularly beneficial to Protopterus owing to its low rate constant for RBC Cl^-/HCO₃^- exchange(Jensen et al., 2003), the step normally considered to be rate limiting in CO₂ excretion(Perry, 1986; Tufts and Perry, 1998). In light of the results of the initial respirometry experiments, which showed the gill to be a minor route for CO₂ elimination, clearly the original hypothesis was no longer appropriate. Thus, the hypothesis was modified post hoc to test instead the idea that excretion of CO₂ at either site of gas exchange (gill/skin or lung) is aided by extracellular CA.

It is likely that all air-breathing animals possess membrane-associated endothelial pulmonary CA that is oriented toward the vascular lumen and thus able to catalyze extracellular reactions(Stabenau and Heming, 2003). However, its role in CO₂ excretion is constrained by the low buffering capacity of plasma that limits the supply of substrate(H⁺) for the HCO₃^- dehydration reaction(Bidani and Heming, 1991). It is generally accepted that less than 10% of overall excreted CO₂ is derived from the extracellular catalysis of HCO₃^-(Bidani, 1991; Cardenas et al., 1998; Henry and Swenson, 2000). Based on the present results obtained using the functionally impermeant CA inhibitor, BZ, pulmonary CA (if present in lungfish), as in other air-breathers, is clearly not playing a significant role in aerial CO₂ excretion. Although the use of BZ as an impermeant CA inhibitor has been challenged (Supuran and Scozzafava, 2004), results of this and previous studies(Gilmour et al., 2001) clearly demonstrate that low doses of BZ do not sufficiently penetrate RBCs of lungfish, trout or dogfish within 30 min to significantly influence HCO₃^- dehydration. Currently, studies are underway to determine if a membrane-associated CA isoform does actually exist in the lung of Protopterus.

CO₂ excretion into the water also was unaffected by BZ treatment, and thus the apparent preferential transfer of CO₂ by the gill/skin in comparison to O₂ presumably does not reflect extracellular catalysis of HCO₃^-. To date, the only fish for which extracellular CA has been shown to significantly contribute to CO₂ excretion is Squalus acanthias(Gilmour et al., 2001; Gilmour and Perry, 2004), a species known to possess high plasma buffering capacity(Lenfant and Johansen, 1966; Gilmour et al., 2002).

Results obtained using the permeant CA inhibitor, AZ, demonstrate that CO₂ excretion in P. dolloi can occur normally in the absence of RBC CA activity. The absence of any effect of AZ on CO₂excretion or blood acid–base status could not be attributed to insufficient inhibition of CA, based on independent measures of RBC CA activity, and AZ concentrations in treated fish compared with the K_i for lungfish RBC CA against AZ; this K_i value for lungfish was also comparable to the corresponding value for rainbow trout(Henry et al., 1993), where a similar dose of AZ elicits a profound respiratory acidosis (e.g. Hoffert and Fromm, 1973). Moreover, the AZ solution that was used in Singapore was brought back to Canada and injected into two rainbow trout Oncorhynchus mykiss;arterial P_CO2 increased over 30 min by approximately 2.72 and 2.22 mmHg in the two trout (S. F. Perry and K. M. Gilmour, unpublished observations). To our knowledge, P. dolloi is the only fish species examined to date that is unaffected by AZ treatment despite total inhibition of RBC CA activity. We must emphasise, however, that this result does not necessarily exclude a role for RBC CA in CO₂excretion under normal conditions. The results simply show that RBC CA activity is not a prerequisite for normal rates of CO₂ excretion. Examples of air breathing species for which AZ was found to inhibit CO₂ excretion or raise arterial P_CO2include Xiphister atropurpureus(Daxboeck and Hemimg, 1982), Trichogaster trichopterus(Burggren and Haswell, 1979), Blennius pholis (Pelster et al.,1988) and Lepisosteus oculatus(Smatresk and Cameron, 1982). The likeliest explanation for the absence of any effect of CA inhibition on CO₂ excretion or blood acid–base status in P. dolloiis that CO₂ excretion, at both the gills/skin and lung, is occurring in the absence of appreciable catalyzed HCO₃^-dehydration during gill/skin or lung transit. As discussed by Brauner and Val(1996) to explain CO₂ excretion in Arapaima gigas, the high P_CO2 values that are characteristic of lungfish blood,coupled with low rates of metabolism, may negate the requirement for catalysed HCO₃^- dehydration during the passage of blood through the respiratory organs. Nevertheless, it is puzzling that, given the presence of RBC CA activity and the likelihood of RBC Cl^-/HCO₃^- exchange (based on its presence in the related P. aethiopicus; Jensen et al., 2003), there would appear to be no obligate requirement for the catalysed dehydration of plasma HCO₃^- within the gills or lungs. One possible explanation is that the RBC CA in lungfish may be a slow-turnover isoform and thus unable to keep pace with the rate of substrate delivery by the Cl^-/HCO₃^- exchanger(Tufts et al., 2003). In support of this idea, we (L. Kenney, S. F. Perry, S. F. Chew, Y. K. Ip and K. M. Gilmour, unpublished data) have recently cloned and sequenced two CA isoforms from Protopterus blood, both of which are most similar to CA XIII, an enzyme with only moderate catalytic activity(Lehtonen et al., 2004). Because injection of bovine CA increased gill/skin CO₂ excretion(without increasing O₂ uptake), it would indeed appear that the extent of catalysed HCO₃^- dehydration is a constraint on aquatic CO₂ transfer in P. dolloi.

The costs of this research were supported by NSERC grants to S.F.P. and K.M.G. and NUS R-154-000-226-112 grant to Y.K.I. C.S.F. wishes to thank NIE for an Oversea Attachment Program grant to support the visit of S.F.P. to Singapore. We are grateful to Mrs Wong Wai Peng (Singapore) and Ms Cheryl Szebedinszky (Ottawa) for technical assistance and to Mr Bill Fletcher for caring for the fish while in Ottawa. We are indebted to an `anonymous' referee for encouraging us to enter into a series of constructive interactive debates and for agreeing to agree on disagreeing.