An Evaluation of Factors Limiting Carbon Dioxide Excretion by Trout Red Blood Cells in Vitro

In most fish, carbon dioxide is excreted across the gill into the ventilatory water owing to the rapid dehydration of plasma bicarbonate (HCO₃⁻) to molecular (gaseous) CO₂ (for reviews, see Cameron and Polhemus, 1974; Randall and Daxboeck, 1984; Wood and Perry, 1985; Perry, 1986; Randall, 1990; Perry and Wood, 1989; Perry and Laurent, 1990). The dehydration reaction is catalyzed by the enzyme carbonic anhydrase contained within the red blood cells (RBCs) (Maren, 1967; Swenson and Maren, 1987; Rahim et al. 1988). The plasma HCO₃⁻ gains access to the intracellular carbonic anhydrase during the brief period when blood flows through the branchial vasculature because of the presence of a rapid anion exchanger (band 3 protein) on the RBC membrane (Romano and Passow, 1984; Hubner et al. 1992). During the process of CO₂ excretion, the anion exchanger extrudes Cl⁻ from the RBC in exchange for plasma HCO₃⁻, a process termed ‘the chloride shift’ (Cameron, 1978; Obaid et al. 1979). The protons for the HCO₃⁻ dehydration reaction are provided largely by the dissociation of haemoglobin buffers. In theory, oxygenation of the haemoglobin should accelerate the rate of HCO₃⁻ dehydration, and hence overall CO₂ excretion, owing to the liberation of Bohr protons (for a review, see Jensen, 1991).

It has generally been assumed (e.g. Perry, 1986) that the process of RBC Cl⁻/HCO₃⁻ exchange is the rate-limiting step in teleost CO₂ excretion because the velocities of the other steps (catalyzed HCO₃⁻ dehydration and CO₂ diffusion) are considered to be too rapid to limit overall CO₂ excretion. Surprisingly few studies, however, have evaluated experimentally the variety of potential factors limiting CO₂ excretion. This may reflect, at least in part, the inherent methodological problems associated with monitoring CO₂ excretion in vivo and the inappropriateness of the traditional in vitro techniques (e.g. Haswell and Randall, 1976) utilized to assess RBC CO₂ excretion. Recently, a radioisotopic assay has been developed (Wood and Perry, 1991; Perry et al. 1991) to assess RBC CO₂ excretion quantitatively in vitro; it utilizes physiological levels of plasma HCO₃⁻ and has sufficient sensitivity to detect subtle or rapidly occurring changes in the rate of HCO₃⁻ dehydration. In the present study, we have used the newly developed assay to evaluate several factors that could potentially limit CO₂ excretion by rainbow trout RBCs. Specifically, we have assessed the velocity of Cl⁻/HCO₃⁻ exchange, the activity of intracellular carbonic anhydrase, the plasma HCO₃⁻ concentration and the availability of protons (for HCO₃⁻ dehydration) as variables potentially limiting carbon dioxide excretion in fish.

Rainbow trout [Oncorhynchus mykiss (Walbaum)] weighing between 175 and 250 g (experimental N=251) were obtained from Linwood Acres Trout Farm (Campbellcroft, Ontario) and were transported in oxygenated water to the University of Ottawa.

Fish were maintained on a 12h:12h L:D photoperiod in large fibreglass aquaria supplied with flowing, aerated and dechlorinated City of Ottawa tapwater ([Na⁺]= 0.10mmol l⁻¹, [Cl⁻]=0.15mmol l⁻¹, [Ca²⁺]=0.35–0.40mmol l⁻¹, [K⁺]=0.03mmol l⁻¹, pH7.7–8.0). Fish were acclimated to these conditions for at least 4 weeks before experimentation. Water temperature in holding and experimental facilities varied between 10 and 12°C during the course of the experiments (May–September). Trout were fed daily to satiation using a diet of commercial trout pellets; food was withheld for 48 h prior to experimentation.

Trout were anaesthetized in a 0.1 gl⁻¹ solution of ethyl-m-aminobenzoate (MS 222; Sigma Chemical Company) adjusted to pH7.5 with NaHCO₃ and then placed onto an operating table to allow continuous retrograde irrigation of the gills with anaesthetic solution. To permit blood sampling, an indwelling cannula was implanted into the dorsal aorta (Soivio et al. 1975) using flexible polyethylene tubing (Clay-Adams PE 50; internal diameter 0.580mm, outer diameter 0.965mm). Trout were revived on the operating table by irrigation of the gills with aerated water, then transferred to individual opaque acrylic experimental chambers (volume 3l) supplied with aerated, flowing water, where they were allowed to recover from the effects of anaesthesia and surgery for at least 48 h before experimentation commenced.

The technique for assessing CO₂ excretion within trout RBCs has been described in detail in a previous paper (Wood and Perry, 1991) and is therefore only briefly reiterated here. Approximately 20ml of whole blood was required for a typical single experimental run (i.e. N=1). Thus, it was necessary to use pooled blood obtained by slow withdrawal from the dorsal aortic cannulae of 4–5 fish (3–4ml per fish). The blood sampling was stopped if fish showed signs of agitation or struggling. Although not measured in the present study, this type of sampling protocol is known to avoid endogenously elevated catecholamine and lactate levels. Using an identical protocol, Wood and Perry (1991) reported that total plasma catecholamine (adrenaline plus noradrenaline) levels in pooled blood were always below 7nmol l⁻¹. 1ml samples of the pooled blood (stored on ice) or separated plasma were added to glass scintillation vials (20ml), which were then stoppered and gassed with a humidified gas mixture to yield a of 0.45kPa (3.38mmHg) and a of 20.7kPa (155mmHg), remainder N₂ for 2h at ambient water temperature in a shaking water bath. The gas mixture was provided by a gas-mixing pump (Wösthoff model M 301a/f). In experiments where haematocrit was varied (nominal range 0–35%), the pooled blood was centrifuged (5900 g for 2min; 4°C) and appropriate volumes of homologous plasma were added or removed; separated plasma was also obtained in this manner. In all cases, the actual haematocrits were measured (see below); throughout the paper, the nominal values are reported because there was never significant deviation from the measured values.

74kBq (10 μl of 7400kBqml⁻¹) of sodium [¹⁴C]bicarbonate (in teleost Ringer; Wolf, 1963) was added to each 1ml of blood or plasma. The vial was then immediately sealed with a rubber septum, from which was suspended a plastic well containing a filter paper trap (150 μl hyamine hydroxide) for CO₂, and shaking was started. After exactly 3min of shaking, the filter was removed and assayed for ¹⁴C activity. Whole-blood or plasma pH was determined and the remaining blood centrifuged (12000 g for 2min). The pellet was utilized to determine RBC pHi according to the freeze–thaw method (Zeidler and Kim, 1977). Samples of true or separated plasma were assayed for ¹⁴C activity (50μl) and total CO₂ (⁠⁠; 50 μl) to determine plasma HCO₃⁻ specific activity (disintsmin⁻¹ μmol⁻¹). The CO ₂ excretion rate for each assay vial was calculated by dividing filter paper ¹⁴C activity by plasma specific activity and time.

This preliminary series of experiments was designed to evaluate RBC Cl⁻/HCO₃⁻ exchange and intracellular carbonic anhydrase levels as potential limiting factors in RBC CO₂ excretion. The experiments were performed using either separated plasma or whole blood at a constant haematocrit of 20%. The CO₂ excretion assay was performed using the naturally occurring levels of HCO₃⁻ in the pooled blood (4–6mmol l⁻¹). The carbonic anhydrase inhibitor acetazolamide (Sigma; final concentration 10⁻⁴ mol l⁻¹) dissolved in 140mmol l⁻¹ NaCl was added (50 μl) to the blood/plasma 30min prior to the assay to allow it sufficient time to penetrate across the RBC membrane. The Cl⁻/HCO₃⁻ exchange blocker SITS (4-acetamido-4′-isothiocyanatostilbene-2,2′-disulphonic acid; Sigma), dissolved in 2% dimethylsulphoxide (DMSO), or bovine carbonic anhydrase (2500Wilbur-Anderson units per mg) was added to blood/plasma immediately before the assay to yield final nominal concentrations of 10⁻⁴ mol l⁻¹/0.1% DMSO and 0.25–6.0mgml⁻¹, respectively. Preliminary experiments as well as a previous study (Perry et al. 1991) demonstrated that addition of 50 μl of 2% DMSO was without effect on RBC CO₂ excretion; therefore, the controls in this series consisted of blood or plasma to which had been added 50 μl of 140mmol l⁻¹ NaCl.

In this series of experiments, the effects of haematocrit and acutely altered plasma HCO₃⁻ concentration were evaluated. Four final concentrations of HCO₃⁻ were utilized; the naturally occurring concentration (i.e. 4–6 mmol l⁻¹, hereafter referred to as 5 mmol l⁻¹ HCO₃⁻ blood/plasma), 10mmol l⁻¹ HCO₃⁻, 15mmol l⁻¹ HCO₃⁻ or 30mmol l⁻¹ HCO₃⁻. Plasma HCO₃⁻ levels were elevated abruptly at the start of the assay by ‘spiking’ the blood/plasma with 10–30 μl of sodium [¹⁴C]bicarbonate (7400kBqml⁻¹) prepared in teleost Ringer containing 500mmol l⁻¹ NaHCO₃ instead of the usual teleost Ringer. Each HCO₃⁻ concentration (N=6 for each concentration) was tested at eight different haematocrits ranging from zero (plasma) to 35% (0, 2.5%, 5%, 10%, 15%, 20%, 25%, 35%). The nominal HCO₃⁻ levels were verified by analyzing the true plasma ⁠.

In this series, the activity of the RBC membrane Cl⁻/HCO₃⁻ exchanger and the levels of intracellular carbonic anhydrase were evaluated as limiting factors in RBC CO₂ excretion. This was accomplished by comparing the rates of CO₂ production in (i) control blood (N=6), (ii) lysed blood (N=6) and (iii) lysed blood containing an additional 3 mgml⁻¹ bovine carbonic anhydrase (N=6). The blood cells were haemolyzed by subjecting the blood to high-frequency sonication for 10s using a micro ultrasonic cell disrupter (Kontes). Preliminary experiments demonstrated that this protocol was effective at causing total cellular lysis (i.e. no intact RBCs remained). The cell lysis or the addition of carbonic anhydrase to the lysed blood was performed immediately before the CO₂ excretion assay commenced. This series of experiments was performed using four haematocrits (0, 5%, 20%, 35%) and four levels of plasma HCO₃⁻ (5, 10, 15, 30mmol l⁻¹).

In this series, the effects of rapid oxygenation of the blood on RBC CO₂ excretion were assessed. The aim of these experiments was to simulate the situation in vivo whereby prebranchial deoxygenated blood is rapidly oxygenated as the blood flows through the gill vasculature. In theory, the oxygenation process should facilitate CO₂ excretion owing to the associated liberation of Bohr protons from haemoglobin (the Haldane effect) and the resultant acceleration of RBC HCO₃⁻ dehydration. In order to test this theory experimentally, it was necessary to modify the RBC CO₂ excretion assay so as to operate under open conditions in which the appropriate gases were continually flowing in and out of the assay vessel. This was achieved by connecting a second CO₂ trap [a 20ml glass scintillation vial containing 1.5ml of a CO₂-absorbing solution (Carbo-Trap 2; Baker)] in series with the primary reaction vessel (see Fig. 1). Thus, in these experiments RBC CO₂ excretion was calculated using the summed ¹⁴C activities from the filter paper and the Carbo-Trap solution. Pilot experiments demonstrated that a single additional CO₂ trap was sufficient to remove all of the ¹⁴CO₂ from the flowing gas provided that the flow of gas was kept at 120mlmin⁻¹. At higher gas flow rates, the single additional trap was not capable of removing all of the ¹⁴CO₂. Thus, it was critical in these experiments to monitor the gas flow rate continually using flow meters (see Fig. 1).

The experimental design involved pre-incubating 1ml samples of pooled blood (as above) with a gas mixture (0.5% CO₂, 2.0% O₂, remainder N₂; that was intended to simulate in vivo venous blood gas tensions in rainbow trout. One group was subjected to rapid oxygenation at the commencement of the 3min CO₂ excretion assay by switching the inflowing gas to pure O_2, while the other group was kept deoxygenated by switching the inflowing gas to 2% O₂ in N₂ at the beginning of the assay. CO₂ was purposely omitted from these gases to simulate the washout of CO₂ occurring at the gill.

To try to distinguish between the rapid effects of oxygenation in supplying H⁺ for intracellular HCO₃⁻ dehydration and the possible effects of the RBC pHi change per se (or other long-term effects associated with oxygenation), an additional set of experiments was performed in which RBC CO₂ excretion was assessed in blood pre-equilibrated with hypoxic (⁠⁠,⁠) or normoxic (⁠⁠) gases. In these experiments, the unmodified RBC CO₂ excretion assay was used.

Haematocrit was determined by centrifuging approximately 80 μl of blood in a heparinized capillary tube for 10min at 5000 g. Whole-blood or plasma pH (pHe) and RBC intracellular pH (RBC pHi) were determined with a Radiometer micro-capillary pH electrode (G299A) maintained at the experimental temperature (10–12°C) in a Radiometer BMS3 Mk2 blood micro-system. Plasma was determined on 50 μl samples using a Corning model 965 CO₂ analyzer. Plasma HCO₃⁻ levels were calculated using the Henderson–Hasselbalch equation and the appropriate constants listed in Boutilier et al. (1984).

Plasma and filter paper ¹⁴C activities were determined by liquid scintillation counting (Packard TR 2500) and automatically corrected for quenching. Plasma (50 μl), Carbo-Trap (1.5ml) and filter papers were counted using a commercial scintillation cocktail. The plasma and filter papers were counted in 10ml of ACS II (Amersham), whereas the Carbo-Trap was counted in 18ml of OCS II so as to reduce colour quenching.

All values shown are means ± 1 standard error of the mean (S.E.M.). For multiple comparisons, the results have been statistically analyzed using factorial analysis of variance followed by Fisher’s LSD multiple-comparison test. For two-sample comparisons, the unpaired two-tailed Student’s t-test was used. In both cases, 5% was taken as the fiducial limit of significance.

Incubation of blood with the carbonic anhydrase inhibitor acetazolamide caused a 72 % reduction in the rate of CO₂ production (Fig. 2A) to yield a rate that was equivalent to that of separated plasma; acetazolamide did not influence the rate of CO₂ production in the absence of RBCs. Treatment of blood with the Cl⁻/HCO₃⁻ exchange blocker SITS caused a 91% reduction in overall CO₂ production (Fig. 2A). Surprisingly, the rate of whole-blood CO₂ production after SITS treatment was lower than the rate observed using separated plasma. SITS did not affect the rate of CO₂ production in separated plasma (data not shown).

The addition of bovine carbonic anhydrase to separated plasma caused a dose-dependent elevation of CO₂ production between 0 and 1.0mgml⁻¹; at levels greater than 1.0mgml⁻¹ there were no further increases in the rate of CO₂ production (Fig. 2B). In all subsequent experiments employing carbonic anhydrase, a dose of 3mgml⁻¹ was utilized.

Increasing the haematocrit caused an elevation of the rate of CO₂ production, although this effect was apparent only between 0 and 15% haematocrit; the rate of CO₂ production was unaffected by further increases in haematocrit between 15 and 35% (Fig. 3A).

Although there was considerable overlap of data (compare 10 and 15mmol l⁻¹ HCO₃⁻), it is nevertheless clear that acutely elevating plasma HCO₃⁻ levels increased the rate of CO₂ production in whole blood (Fig. 3A). This stimulatory effect of elevated [HCO₃⁻] was absent in separated plasma while being most pronounced at the higher haematocrits (Fig. 3A). Fig. 3B illustrates the result of a separate group of experiments in which HCO₃⁻ levels were varied between 5 and 30mmol l⁻¹ in (i) high-haematocrit blood (35%), (ii) separated plasma, and (iii) separated plasma containing 3mgml⁻¹ carbonic anhydrase. In the whole blood, the rate of CO₂ production was a positive linear function of the plasma HCO₃⁻ concentration. However, CO₂ production was not correlated with HCO₃⁻ levels in separated plasma either with or without exogenous carbonic anhydrase.

Haemolysis of the blood caused significant increases in the rate of CO₂ production at all haematocrits (5%, 20%, 35%) and extracellular HCO₃⁻ levels (5, 10, 15 and 30mmol l⁻¹; Fig. 4). The stimulatory effect of haemolysis on CO₂ production was greatest at the highest haematocrits and HCO₃⁻ levels (compare panels A and D in Fig. 4). The addition of 3mgml⁻¹ bovine carbonic anhydrase to previously lysed blood caused a pronounced stimulation of CO₂ production at all HCO₃⁻ levels in the low-haematocrit (5%) blood (Fig. 4). At higher haematocrits, there was a slight stimulation (Fig. 4A), no effect or a slight inhibition (Fig. 4C,D).

The rapid oxygenation of whole blood (20% haematocrit) caused a pronounced increase in the rate of CO₂ production in comparison to the rate of CO₂ production from partially deoxygenated blood (approximately 50% haemoglobin O₂-saturation; Fig. 5A). Rapid oxygenation of deoxygenated blood elicited a decrease in RBC pHi from 7.30±0.05 to 6.95±0.05 (Fig. 5B). The rate of CO₂ production from separated plasma was unaffected by oxygen status (Fig. 5A).

In contrast, the rate of CO₂ production from blood that had been oxygenated prior to the assay was consistently lower than that from deoxygenated blood at the various haematocrits tested (Fig. 6A). In all cases, the RBC pHi was markedly lower in the oxygenated blood (Fig. 6B).

The goal of this study was to evaluate as potentially limiting factors several of the multiple components of the CO₂ excretion pathway in fish. Included within the analysis were (i) the rate of RBC Cl⁻/HCO₃⁻ exchange, (ii) the rate of the catalyzed dehydration of HCO₃⁻ to CO₂ by carbonic anhydrase and (iii) the availability of the substrates HCO₃⁻ and H⁺ for the dehydration reaction. This was accomplished in vitro by using a sensitive radioisotopic assay (Wood and Perry, 1991; Perry et al. 1991) that monitors the evolution of [¹⁴C]CO₂ from [¹⁴C]HCO₃⁻. As discussed in detail by Wood and Perry (1991), this technique offers several significant advantages over other commonly used methods (Heming and Randall, 1982; Tufts et al. 1988) because it avoids non-physiological buffer solutions, inappropriate acid–base values and abnormally high HCO₃⁻ gradients between the plasma and the RBC interior. The use of a high-capacitance CO₂ trap within the assay vessel (see Fig. 1 in Perry and Wood, 1991) maintains the gradient between the blood compartment and the gas phase and simulates the situation of blood flowing through the branchial vasculature in vivo where the ventilatory water is nearly zero. However, this assay, like all in vitro methods for determining the rates of CO₂ transfer between tissue, fluid and gas compartments, probably does not exactly simulate the kinetics of CO₂ exchange in vivo (see review by Klocke, 1988). This is probably a consequence of inadequate convection of the blood, leading to formation of unstirred layers which, in turn, may restrict the diffusion of CO₂. In addition, the 3min assay time is considerably longer than the residence time of blood within the gill (0.5–2.5s). Despite the relatively long duration of the assay, plasma [HCO₃⁻] is reduced only by approximately 1mmoll⁻¹ (Wood and Perry, 1991), which is similar to the reduction in plasma [HCO₃⁻] during CO₂ excretion in vivo (Perry, 1986). Thus, it is unlikely that extracellular [HCO₃⁻] or intracellular [Cl⁻] (based on a 1:1 stoichiometric exchange of Cl⁻ for HCO₃⁻) is a limiting factor for HCO₃⁻ dehydration during the in vitro assay.

The original assay of Perry and Wood (1991) was designed to operate in a sealed reaction vessel so as to prevent loss of evolved [¹⁴C]CO₂ into the atmosphere. Consequently, the assay (as designed) could not be used to evaluate the effects of rapid oxygenation of the blood (and the associated liberation of Bohr protons) on RBC CO₂ excretion owing to the requirement for a fully open system. Thus, for this component of the study, the assay was modified to operate under open conditions. This was achieved by adding a second CO₂ trap in series with the original reaction vessel (see Fig. 1). Gas was allowed to flow in the system without the loss of [¹⁴C]CO₂ to the atmosphere because the capacitance of the second trap (1.5ml Carbo-Trap solution) was sufficient to remove all of the [¹⁴C]CO₂ from the inflowing gas, provided that the flow rate of the gas was kept at, or below, 120mlmin⁻¹. Interestingly, the rate of whole-blood CO₂ excretion was approximately doubled under open conditions compared with closed conditions (compare Figs 5 and 6). This probably reflects an increase in the delivery (owing to convection) of the evolved CO₂ to the two traps rather than the addition of the second trap per se, because Wood and Perry (1991) demonstrated that the capacity of the single trap for CO₂ was well in excess of the rate of RBC CO₂ production. It is likely, therefore, that the open system provides a better estimate of the true rate of RBC CO₂ excretion. However, in the present study, no comparisons of data generated by the two different techniques were required and, indeed, it would be incorrect to compare the absolute levels of CO₂ excretion in the two methods.

Numerous previous studies have provided indirect (Wood et al. 1982; Heming and Randall, 1982) or direct (Cameron, 1978; Obaid et al. 1979; Perr et al. 1982, 1991; Tufts et al. 1988) evidence for the involvement of RBC Cl⁻/HCO₃⁻ exchange in CO₂ excretion in fishes. Owing to the slow velocity of Cl⁻/HCO₃⁻ exchange in relation to other steps in overall CO₂ excretion (CO₂ diffusion, HCO₃⁻ dehydration), it has largely been assumed that RBC Cl⁻/HCO₃⁻ exchange is a rate-limiting process in piscine CO₂ excretion (Perry, 1986; Piiper, 1989) as it is in mammals (see review by Bidani and Crandall, 1988).

The present study provides the first experimental data to support the theory of RBC Cl⁻/HCO₃⁻ exchange as a rate-limiting factor in fish CO₂ excretion. The near total inhibition of RBC CO₂ production with the Cl⁻/HCO₃⁻ exchange blocker SITS (Fig. 2) confirmed the requirement of this pathway to provide substrate (HCO₃⁻) for intracellular H₂CO₃ dehydration. Although this experiment, in itself, did not demonstrate that Cl⁻/HCO₃⁻ exchange is rate-limiting, it was nevertheless crucial to show that the production of CO₂ in this particular radioisotopic assay was indeed derived from the dehydration of plasma HCO₃⁻via Cl⁻/HCO₃⁻ exchange. When the need for this step was eliminated, by ultrasonic disruption of the RBC membrane, there was a marked stimulation of CO₂ excretion (Fig. 4). With a few exceptions (see below), the levels of CO₂ excretion obtained after membrane disruption were identical to those obtained after the addition of saturating levels (3mgml⁻¹) of bovine carbonic anhydrase, indicating that maximal rates of CO₂ production had been achieved. These data clearly indicate that the velocity of the Cl⁻/HCO₃⁻ exchange pathway was not sufficient to allow maximal rates of CO₂ production and that this pathway must be a rate-limiting factor under the conditions of the in vitro assay. It is also important to note that non-HCO₃⁻ buffers, normally present in high concentrations in the RBC, would be released into the plasma during lysis and would still supply H⁺ for HCO₃⁻ dehydration (see Bidani and Heming, 1991). Haemolysis of the blood also served to eliminate the RBC membrane as a barrier to CO₂ diffusion. It is extremely unlikely, however, that the increased rate of CO₂ excretion in haemolyzed blood was a result of the elimination of this diffusion barrier, because it is generally accepted that the diffusion of CO₂ across tissue barriers is extremely rapid and not rate-limiting (Piiper, 1989; Klocke, 1988).

It seems likely that the results of this in vitro study can be extrapolated to the in vivo situation. First, in vivo the ‘effective’ diffusing capacity (overall diffusive conductance) of the gill for CO₂ is less than that predicted from the CO₂ diffusing capacity (Piiper, 1989). Second, the catalytic capacity of the dogfish (Squalus acanthias) RBC carbonic anhydrase is about 100 times greater than the measured rate of CO₂ excretion (Swenson and Maren, 1987). These apparent diffusion limitations for branchial CO₂ transfer (Malte and Weber, 1985) are most readily explained by the slow velocity of Cl⁻/HCO₃⁻ exchange. Cameron (1978) determined the time constant of Cl⁻/HCO₃⁻ exchange to be 400ms in rainbow trout blood. The transit time of blood in the gill is probably 0.5–2.5s, so it is conceivable that RBC Cl⁻/HCO₃⁻ exchange continues in the post-branchial arterial blood.

The addition of acetazolamide to blood lowered the rate of CO₂ excretion to values observed using separated plasma (Fig. 2), confirming the essential role of intracellular carbonic anhydrase in these in vitro experiments. More importantly, the results of this study suggest that under certain conditions the activity of RBC carbonic anhydrase may be rate-limiting for overall CO₂ excretion. First, the rate of CO₂ excretion in haemolyzed blood was significantly greater (with the exception of the 5mmol l⁻¹ HCO₃⁻ blood) at the higher haematocrit values of 20 and 35% than at 5% (Fig. 4). Second, the addition of 3 mgml⁻¹ bovine carbonic anhydrase to previously haemolyzed blood significantly increased the rate of CO₂ excretion consistently at the low haematocrit value of 5 % (Fig. 4). These results suggest that the activity of carbonic anhydrase may be rate-limiting in CO₂ excretion, but only at the unusually low haematocrits that might occur during severe anaemia (e.g. Wood et al. 1982). This potential limitation by intracellular carbonic anhydrase activity of CO₂ excretion at low haematocrits is consistent with the notion that fish RBCs contain significantly less carbonic anhydrase than do mammalian RBCs (Maren, 1967). It is unlikely, however, that CO₂ excretion is limited by RBC carbonic anhydrase activity at normal haematocrits (20–30% in rainbow trout).

In a preliminary series of experiments, Wood and Perry (1991) showed, using an identical assay, that the relationship between RBC CO₂ excretion and haematocrit was curvilinear but that acute elevation of plasma [HCO₃⁻] could further increase CO₂ excretion rates. In the present study, we have elaborated on those initial observations by performing a detailed analysis of the interactive effects of haematocrit and plasma [HCO₃⁻] on CO₂ excretion. The results confirm the curvilinear relationship between haematocrit and CO₂ excretion and the stimulatory effects of acute elevation of plasma [HCO₃⁻] (Fig. 3). The stimulatory effect of increasing haematocrit between 0 and 15 % can be explained by the combined effects of the increasing levels of carbonic anhydrase in the blood and the greater numbers of Cl⁻/HCO₃⁻ exchangers. Although carbonic anhydrase levels continue to increase at haematocrits above 15%, there was no further stimulatory effect on CO₂ excretion because the levels were presumably already in excess. Despite excessive carbonic anhydrase activity and increasing numbers of Cl⁻/HCO₃⁻ exchange sites, the rate of CO₂ excretion remained constant as haematocrit was raised above 15%. The absence of any further stimulation of CO₂ excretion at haematocrits above 15% may be related to insufficient quantities of HCO₃⁻. Interestingly, the addition of HCO₃⁻ to plasma (with or without bovine carbonic anhydrase) did not affect CO₂ excretion, which suggests that the single mechanism underlying the stimulatory effect of increased [HCO₃⁻] was an increase in the velocity of Cl⁻/HCO₃⁻ exchange and that mass action effects were insignificant. These results demonstrate that rapid elevations of plasma HCO₃⁻ levels in vivo may transiently affect CO₂ excretion. The effect would be expected to be of short duration because the normal HCO₃⁻ gradient across the RBC membrane would be quickly re-established.

In contrast to the present study, Tufts et al. (1988) demonstrated a linear relationship between haematocrit and CO₂ excretion in vitro with no attenuation at the higher haematocrits. This may reflect the unusually high levels of HCO₃⁻ added to the blood at the onset of the assay (200mmol l⁻¹). It is likely, therefore, that the velocity of Cl⁻/HCO₃⁻ exchange was not limited at the higher haematocrits by insufficient substrate, unlike in the present study. Similarly, Tufts et al. (1988) reported that the addition of SITS caused CO₂ excretion to be inhibited by only 41% (compared to 91% in the present study); this difference may also reflect the high levels of HCO₃⁻ used by Tufts et al. (1988).

The conversion of HCO₃⁻ to CO₂via the dehydration of H₂CO₃ requires a supply of protons. The protons are derived from the dissociation of non-HCO₃⁻ buffers (see Bidani and Heming, 1991, for a discussion of the importance of non-HCO₃⁻ buffers in CO₂ excretion). In addition to the buffering process, which is independent of oxygen, there is also a supplementary source of H⁺ (Bohr protons) derived during oxygenation of the haemoglobin molecule (the Haldane effect). In mammals, this oxylabile component may constitute as much as 50% of overall CO₂ excretion (Klocke, 1988). Teleost fish are known to possess large Haldane effects (see review by Jensen, 1991); thus, the oxylabile component of CO₂ excretion is expected to be significant.

The results of this study permit for the first time a partitioning of O₂-independent and O₂-dependent CO ₂ excretion in trout. The rapid oxygenation of partially deoxygenated blood (to simulate blood arriving at the gills) caused a 37% increase in the rate of CO₂ excretion in vitro. Thus, it would appear that the oxylabile component is approximately 40% of overall CO₂ excretion in this in vitro assay and that, by extrapolation, the O₂-dependent linked supply of Bohr protons is an important factor limiting CO₂ transfer in vivo.

Unlike the situation in mammals (see Klocke, 1988), the O₂-dependent component of CO₂ excretion in fish probably does not include a significant contribution from oxylabile carbonate because carbamino CO₂ in fish is considered to be less important than in mammals (Jensen, 1991). However, the radioisotopic assay used in these experiments only measures the excretion of CO₂ derived from plasma HCO₃⁻ and not from other potential sources such as oxylabile carbamate bound to haemoglobin.

In order to differentiate between the O₂-dependent release of Bohr protons and other effects of oxygenation, such as changes in RBC pHi, a series of experiments was performed in which CO₂ excretion was determined in blood that had previously been deoxygenated or oxygenated. These results showed that prior oxygenation of the blood caused a pronounced reduction in the rate of CO₂ excretion in association with large decreases in RBC pHi (Fig. 6). These results suggest that one or more steps in the net conversion of plasma HCO₃⁻ to CO₂ (Cl⁻/HCO₃⁻ exchange or H₂CO₃ dehydration) is influenced by intracellular pH.

This study has identified several variables that can affect the rate of CO₂ excretion, including haematocrit, plasma HCO₃⁻ concentration, velocity of Cl⁻/HCO₃⁻ exchange and the extent of haemoglobin oxygenation. Under normal physiological conditions, in vivo, it is unlikely that haematocrit or plasma HCO₃⁻ concentration is a limiting factor for CO₂ excretion. During severe anaemia (Hct<5%), however, the low activity of carbonic anhydrase in the blood may limit CO₂ excretion. On the basis of extrapolation of the present results, it would seem that the two factors limiting CO₂ excretion in vivo are the velocity of RBC Cl⁻/HCO₃⁻ exchange and the supply of Bohr protons during the oxygenation of haemoglobin.

This work was supported by NSERC of Canada operating and equipment grants to S.F.P. K.M.G. was the recipient of an E. B. Eastburn Postdoctoral Fellowship. We thank Ms Dianne Randall for technical assistance far above and well beyond the call of duty.