A New in Vitro Assay for Carbon Dioxide Excretion by Trout Red Blood Cells: Effects of Catecholamines

Carbon dioxide excretion at the gills of teleost fish follows the standard vertebrate scheme whereby (i) plasma HCO₃^- enters the red blood cell (RBC) in electroneutral exchange for Cl^-; (ii) the HCO₃^- is dehydrated to CO₂ by erythrocytic carbonic anhydrase; and (iii) the resultant CO₂ diffuses from the RBC through the plasma and respiratory epithelium to the external environment (reviewed by Perry, 1986; Perry and Laurent, 1990). The flux is driven by the gradient from RBC to water. However, after strenuous exercise, fish routinely exhibit large increases in without a corresponding decrease in P ⁠, (e.g. Turner et al. 1983; Milligan and Wood, 1986; McDonald et al. 1989). The response is curious because a decrease in would be expected if the phenomenon were due to a simple diffusive or convective limitation on gas exchange. To explain this observation, several years ago we proposed that the mobilization of catecholamines into the acidotic bloodstream after exhaustive exercise produces an inhibition of the HCO₃^- entry step, thereby causing CO₂to back-up in the system until the rise in ⁠, is sufficient to restore the flux (Wood and Perry, 1985).

Since that time, this ‘CO₂retention theory’ has proved controversial. On the one hand, chronic intra-arterial infusions of adrenaline have been shown to cause simultaneous elevations of and (Perry and Vermette, 1987; Vermette and Perry, 1988) and intra-arterial injections of carbonic anhydrase have been shown to reduce the post-exercise increase in without altering (Perry and Wood, 1989). On the other hand, neither Steffensen et al. (1987) nor Playle et al. (1990) were able to demonstrate a transient fall in relative rate of carbon dioxide production (i.e. a decrease in the respiratory exchange ratio, RE) after exercise or catecholamine injections, in contrast to the predictions of the theory. Furthermore, Tufts et al. (1988) were unable to replicate a key observation on which the theory was based, namely that catecholamines cause an inhibition of CO₂excretion by intact trout RBCs in vitro.

Both the original observation (unpublished data of S. F. Perry and T. A. Heming presented by Wood and Perry, 1985) and the contrasting results of Tufts et al. (1988) were based on modifications of the manometric ‘boat’ assay for carbonic anhydrase, first used for RBCs by Booth (1938). In the assay, the RBCs are exposed to unphysiological buffer systems and pH values, and massive HCO₃^- pulses. Booth concluded that it was impossible to use this assay to measure CO₂ excretion rates in intact erythrocytes, and over the years its use for that purpose with fish RBCs has been plagued with controversy (e.g. Haswell and Randall, 1976; Obaid et al. 1979; Cameron, 1978; Heming and Randall, 1982). Unfortunately, other available methods involve similar disturbances, such as the use of artificial buffers, large pulses of acid or base, filtration separation of the RBCs, cell lysis or the addition of carbonic anhydrase to the extracellular fluid (e.g. Dirken and Mook, 1931; Piiper, 1969; Lambert and Lowe, 1978; Crandall et al. 1978; Maren and Couto, 1979). We therefore decided that an entirely different approach was required, one in which net CO₂ excretion from the unmodified plasma through the intact RBCs could be quantified under conditions closely approximating those in vivo.

In the present paper we describe and characterize a sensitive new method, based on the use of [¹⁴C]bicarbonate, for measuring net CO₂ excretion by trout RBCs in their normal plasma in vitro. The gradients and acid-base conditions are similar to those encountered by blood passing through the gills in vivo. Using this technique, we have confirmed the adrenergic inhibition of RBC CO₂ excretion, and identified the conditions under which it occurs. The companion paper (Perry et al. 1991) presents a detailed analysis of the mechanism of this inhibition.

Rainbow trout [Oncorhynchus mykiss (Walbaum); 150–400 g] of either sex were obtained from Thistle Springs Trout Farm (Ashton, Ontario). Fish were maintained indoors in large rectangular fibreglass tanks supplied with flowing, vigorously aerated, dechlorinated City of Ottawa tapwater ([Na⁺]=0.12mmoll^-1; [Cl^-]=0.15 mmol l^-1; [K⁺]=0.03mmoll^-1; pH 7.5–8.0; temperature 9–12°C; photoperiod 12h light: 12h dark). Fish were fed daily ad libitum on floating commercial trout pellets (Purina), but were not fed for 48 h prior to experimental procedures.

To avoid endogenously elevated catecholamine and lactate levels, all blood was drawn from chronically cannulated fish. Trout were anaesthetized in a 1:10000 (w/v) solution of ethyl-m-aminobenzoate (MS 222; Sigma) adjusted to pH 7.5 with NaHCO₃ and then placed onto an operating table which permitted continuous retrograde perfusion of the gills. An indwelling cannula (Clay-Adams PE 50) filled with modified Cortland saline ([HCO₃^-]=5mmoll^-1; Wolf, 1963) was implanted into the dorsal aorta (Soivio et al. 1975). After surgery, fish were placed in darkened Perspex boxes (volume 31) served with continuously flowing acclimation water at the experimental temperature (10±l°C). Fish were allowed to recover for at least 48 h after surgery before blood sampling. Cannulae were flushed daily with 0.2–0.3ml of heparinized saline (10i.u.ml^-1 ammonium heparin; Sigma).

Trout were initially infused with 0.5ml of heparinized (50i.u.ml^-1) saline; blood was then withdrawn slowly until the fish showed the first signs of struggling or disorientation, at which point sampling ceased. This procedure generally yielded 2–3 ml per fish, though occasionally large trout yielded up to 5 ml. Depending on the size of the experiment, blood from at least three, and up to 12, fish was pooled, additionally heparinized (150i.u. ml^-1), and stored on ice for 1–2 h prior to use. At this point, the pooled blood was routinely assayed for wholeblood pH (pHe=7.8-8.0), true plasma carbon dioxide content ⁠, plasma catecholamines (adrenaline plus noradrenaline always below 7 nmol l^-1) and haematocrit (15–25 %).

In preparation for the CO₂ excretion assay, the pooled heparinized blood was split into 1.0 ml samples in individual 20 ml glass scintillation vials. In experiments where haematocrit or plasma HCO₃^- levels (long-term changes) were manipulated, the adjustments were performed immediately prior to the splitting. For the former, the pooled blood was very lightly centrifuged (500g for l min), and an appropriate volume of homologous plasma added or removed. Plasma was also obtained in this fashion for ‘plasma only’ assays. Adjustments of plasma [HCO₃^-] were performed by very gradual additions of 140mmoll^-1 HC1 or 140mmoll^-1 NaHCO₃ to whole blood, care being taken to avoid haemolysis. Addition of 140 mmol I^-1 NaCl served as a control. The vials were then stoppered with rubber septa, gas-equilibrated in an open, flow-through system, as described by Walsh et al. (1990), and shaken for 2h in a constant-temperature bath at 10±l°C. The normal gassing medium was a humidified mixture with (1.91 mmHg), (155mmHg), balance N2 provided by a Wösthoff model M301a/f gas-mixing pump. This was virtually identical to resting measured in the same batch of trout (Wood et al. 1990). Gas equilibration was continued until the start of the CO₂ excretion assay.

The assay is illustrated diagrammatically in Fig. 1, and is based on the addition of [¹⁴C]bicarbonate to the whole blood (i.e. to the extracellular compartment) and collection of the resultant ¹⁴CO₂ in the overlying gas phase. Theoretically, this ¹⁴CO₂ may be generated by two routes: (i) dehydration of [¹⁴C]bicarbonate in the plasma (presumably at the uncatalysed rate); and (ii) dehydration of [¹⁴C]bicarbo-nate in the RBCs (presumably catalysed by erythrocytic carbonic anhydrase) subsequent to entry via C1^- /HCO₃^- exchange. The net CO₂ flux is driven by the gradients between the respective blood compartments and the gas phase. The blood is initially equilibrated to normal in vivo⁠, but in the assay vial it is exposed to a gas phase in which is close to zero, owing to the presence of a CO₂ trap; the situation is similar to that of blood passing through the gills, where the external water is dose to zero.

In practice, 2μCi (10μl of 200μCiml^-1) of sodium [¹⁴C]bicarbonate (in 5 mmol I^-1 HCO₃^- Cortland saline) was added to each 1ml of pre-equilibrated blood (or plasma). The vial was then immediately resealed with a new rubber septum from which was suspended a plastic well containing a trap for CO₂ (Fig. 1; Walsh et al. 1988), returned to the shaker bath, and timing commenced. Assays were routinely run for exactly 3min. The trap consisted of a fluted filter paper (Whatman GF/A 2.4cm glass microfibre filter) impregnated with 150μl of Imoll^-1 hyamine hydroxide in methanol. At the termination of the assay, the filter was immediately removed and assayed for ¹⁴C activity, and the blood drawn into a Hamilton gas-tight syringe. Whole blood (or plasma) pHe was measured, and the remaining blood immediately centrifuged (12000g for 2 min). The plasma was decanted anaerobically for determination (100 μl) and ¹⁴C counting (50 μl); the packed red cell pellet was frozen in liquid N₂ for later determination of RBC pHi by the freeze-thaw lysate method (Zeidler and Kim, 1977).

The CO₂ excretion rate for each assay vial was calculated by dividing filter paper ¹⁴C activity by plasma specific activity and time. Note that specific activity was measured at the end, rather than at the start, of the assay. The calculated rate was then corrected for the efficiency of CO₂ trapping by the hyamine hydroxide filter. Trapping efficiency was measured over a wide range of CO₂ evolution rates by acidifying various 1 ml HCO₃^- standards (¹⁴C-labelled, in plasma) with 100 μl of 35 % HCIO₃, and then running the assay for 3 min. Trapping efficiency was always 74%, regardless of the CO₂ evolution rate.

The following drugs were used in various tests: acetazolamide sodium U.S.P (Diamox; Lederle); bovine erythrocytic carbonic anhydrase (2500 Wilbur-Anderson units per mg; Sigma) (1 Wilbur-Anderson unit causes the pH of 0.012 mol I^-1 veronal to drop from 8.3 to 6.3 in 1 min at 0°C); L-adrenaline bitartrate (Sigma); and L-noradrenaline bitartrate (Sigma). Drug additions to the assay vials were made in 50 μlsamples of 140 mmol I^-1 NaCl, the vehicle alone being added to control vials.

Haematocrit was determined by centrifuging 80 μl of blood in a heparinized capillary tube for 10 min at 5000g. RBC pHi and pHe (whole blood or plasma) were determined with a micro-capillary pH electrode (G299A) thermostatted to the experimental temperature in a BMS3 Mk2 blood micro-system, and displayed on a PHM-71 acid-base analyzer (all Radiometer). Plasma was determined in a few early experiments by the method of Cameron (1971), and in later experiments using a Corning 965 CO₂ analyzer. Plasma and [HCO₃^-] were calculated using the Henderson-Hasselbalch equation and appropriate constants listed in Boutilier et al. (1984). Adrenaline and noradrenaline levels were determined on alumina-extracted plasma samples using high performance (pressure) liquid chromatography with electrochemical detection according to the method of Woodward (1982).

Plasma and filter paper ¹⁴C activities were determined by liquid scintillation counting (LKB Rackbeta) and automatic quench correction. Plasma (50 μl) was counted in 10 ml of commercial cocktail (ACS II; Amersham), and filter papers in 10 ml of a customized cocktail containing 2.0 g of PPO plus 0.1g of POPOP dissolved in 0.81 of toluene plus 0.21 of 95 % ethanol.

Results are reported as means±l S.E.M. (N), or representative experiments, as appropriate. Significant differences were detected using factorial analysis of variance followed by Fisher’s LSD multiple-comparison test; 5 % was taken as the fiducial limit of significance.

Initial trials were run at a range of assay durations. The cumulative CO₂ excretion of whole blood increased in an almost linear fashion with time up to about 10 min, and thereafter deviated only gradually from linearity (Fig. 2A). This basic linearity up to 10 min was seen both in tests run with the endogenous HCO₃^- levels in the equilibrated blood (about 5 mmol I^-1) and in tests where the plasma [HCO₃^-] was acutely doubled at the start of the assay so as to elevate the CO₂ excretion rates (see below).

Fig. 2B illustrates the change in acid-base status of the blood plasma over time in the assay, as measured at the immediate termination of each run. The fell progressively and plasma [HCO₃^-] declined as CO₂ was excreted. The trap kept the gas-phase close to zero, so the fall in blood ⁠, reflected equilibration with the gas-phase. When plotted on a pH-HCO₃^- diagram, the change in acid-base status virtually paralleled the non-HCO₃^- buffer line, confirming that it was due to decline alone. The decrease in gradient from blood to air (the net driving force for the CO₂ excretion process) undoubtedly contributed to the slight deviation from linearity at longer durations. The efficiency tests demonstrated that it was not due to saturation of the trap. Based on these results, a standard assay duration of 3 min, in the strictly linear region, was selected for all further trials.

The rate of CO₂ excretion increased with haematocrit in tests run at both endogenous and acutely elevated levels of plasma [HCO₃^-] (Fig. 3). The increase was not directly proportional to haematocrit, but tended to attenuate at higher RBC levels. At a normal haematocrit of 25 % and normal blood acid-base status (plasma [HCO₃^_]=5mmoll^-1, pHe=7.9), the CO₂ excretion rate of the whole blood was about 14μmol ml^-1 h^-1 or about 5 times the rate of true plasma.

The addition of 0.2 mg of bovine carbonic anhydrase (500 Wilbur-Anderson units) to 1 ml of true plasma raised its CO₂ excretion rate to a value slightly greater than that of whole blood (Fig. 4). The addition of 10⁻⁴ mol I^-1 acetazolamide (Diamox) either to whole blood or to plasma plus carbonic anhydrase lowered their CO₂ excretion rates back to those of plasma alone. Finally, as a check that catecholamines did not affect carbonic anhydrase activity directly, 10⁻⁶moll^-1 adrenaline was added to plasma plus carbonic anhydrase. The CO₂ excretion rate remained equal to that of plasma plus carbonic anhydrase alone. From these tests we conclude that the assay is responsive to changes in CO₂ excretion mediated either by the erythrocytes or by artificial intervention, and that the plasma rate can be subtracted from the whole blood rate to yield the rate of CO₂ excretion mediated by the RBCs alone.

The effects of alterations in plasma [HCO₃^-] at constant differed greatly, depending on whether the blood had been pre-equilibrated at the various HCO₃^- levels or exposed to acute changes. Fig. 5 illustrates an experiment where the plasma HCO₃^- levels were adjusted by addition of HC1 or NaHCO₃, and then the blood was equilibrated at a normal for 2h prior assay. Under these circumstances, the CO₂ excretion rate was markedly unresponsive to the plasma HCO₃^- concentration. Indeed, over a 45-fold range in [HCO₃^-] (0.6–27.3 mmol I^-1), and accompanying 1.9 unit increase in pHe, the whole-blood CO₂ excretion rate increased by only 35 %. Within a more physiological range of plasma HCO₃^- concentrations (2-12 mmol I^-1) there was no detectable effect in several experiments.

In contrast, the RBC CO₂ excretion rate was very responsive to acute changes in plasma [HCO₃^-], as illustrated by Fig. 3. Here the plasma [HCO₃^-] was abruptly elevated at the start of the assay from the equilibration level of 4.5 mmol I^-1 to 9.0 mmol I^-1 (at normal ⁠), thereby creating a non-equilibrium situation. RBC CO₂ excretion rate increased in almost direct proportion to the rise in plasma [HCO₃^-], and the response was related to the haematocrit.

The CO₂ excretion rate was also very responsive to the to which the blood had been equilibrated prior to assay. The data of Table 1 are from four separate experiments with pools of blood of slightly different haematocrits. Nevertheless, it is clear that increases in the equilibration greatly elevated the CO₂ excretion rate. Quantitative interpretation was complicated by the fact that the greater the starting ⁠. the greater was the fall in during the 3 min assay, so this subject was not pursued further.

In a series of preliminary experiments run at normal acid-base status, the addition of adrenaline or noradrenaline (10⁻⁸-10⁻⁶mol l^-1) to the blood 30–120 min prior to assay had small and inconsistent effects on CO₂ excretion rates. In general, the effects were inhibitory, but they were neither consistent nor statistically significant in most trials. However, in vivo the inhibitory effect after strenuous exercise is hypothesized to occur at the time of catecholamine mobilization, which is also a time of severe metabolic acidosis (see Introduction). Furthermore, recent studies on another trout RBC response, the regulation of RBC pHi via adrenergic activation of Na⁺/H⁺ exchange, have shown that this phenomenon becomes markedly sensitized at acidotic pHe values (see Discussion). Therefore, an experiment was performed to evaluate the possible time-dependency of both the RBC pHi and CO₂ excretion responses, using blood acidified to about pH7.4 with HC1 to simulate the metabolic acidosis component recorded in trout immediately after exhaustive exercise (Turner et al. 1983; Milligan and Wood, 1986).

Under these conditions, adrenaline (10⁻⁶moll^-1) caused a rapid rise (0.1 unit) in RBC pHi and fall (0.4unit) in pHe, measured 3 min after addition (Fig. 6C,D). The peak pHi response occurred at 3min, and the peak pHe response at 8min, after which both declined. By 33 and 63 min, the pHi effect had disappeared, though the fall in pHe was still highly significant.

Adrenaline (10⁻⁶moll^-1) had no immediate effect on RBC CO₂ excretion rate (measured 0–3 min after addition; Fig. 6A). However, by 5-8min, CO₂ excretion rate had declined significantly by 35 %, coincident with the maximum reduction in pHe—pHi. At 15-18min, the inhibition, while still significant, had attenuated to 12%. At 30–33min and 60–63 min there was no longer any significant effect, despite the persistence of pHe depression. The cumulative influence of inhibited CO₂ excretion was clearly reflected in the plasma levels measured at the end of each 3 min assay (Fig. 6B).

These results indicated that adrenaline caused a clear, though transient, inhibition of RBC CO₂ excretion under acidotic conditions. The results further suggested that the effect might in some way be associated with the pHe—pHi response. All subsequent experiments were performed with similarly acidified blood (pHe=7.40±0.02, [HCO₃^-] = 1.90±0.07mmoll^-1 at ⁠), with assays conducted at 5–8min after catecholamine addition. Both adrenaline and noradrenaline caused a concentration-dependent reduction in pHe—pHi over the range 10⁻⁸-10⁻⁶moll^-1. In individual samples (treated with either catecholamine at 10⁻⁸-10⁻⁶moll^-1 or a control saline addition) there was a highly significant correlation (r=0.76, P<0.001, N=62) between pHe-pHi and the RBC CO₂ excretion rate (Fig. 7). The greater the effect of catecholamines on pHe-pHi, the greater was the inhibition of CO₂ excretion.

[¹⁴C]bicarbonate-based techniques are widely used in algal physiology for assessing the role of carbonic-anhydrase-mediated HCO₃^- dehydration in photosynthesis (e.g. Tsuzuki et al. 1980; Shiraiwa and Miyachi, 1985). However, to our knowledge, they have not previously been used to monitor analogous processes in vertebrate RBCs. In developing a new in vitro assay for RBC CO₂ excretion, our prime concerns were that the assay be sensitive and reproducible, that the assay conditions be as representative of the in vivo situation as possible, and that the assay should yield rates that are physiologically realistic. The present assay fulfills all three criteria.

By virtue of the use of radiotracer, the assay can measure rates in 3 min or less with a coefficient of variation (CV=s.D./mean) generally less than 10% for multiple assays run on a common pool. In our experience, the largest source of error is the accuracy of timing with respect to insertion and removal of the CO₂ trap. When particular care is taken with this step, the CV can be reduced below 5 %. In contrast to other techniques (see Introduction), the assay is run with intact RBCs in their normal plasma (i.e. whole blood) under in vivo conditions of pHe, and plasma [HCO₃^-]. The exposure to a gas phase of zero simulates the conditions encountered by blood as it passes through the gills. At a normal haematocrit of 25 %, the CO₂ excretion rate of whole blood in the assay at 10°C was about 14 umol ml^-1 h¹. Assuming a normal resting cardiac output of 18 ml kg^-1 min^-1 for the rainbow trout at this temperature (Cameron and Davis, 1970; Kiceniuk and Jones, 1977), the estimated whole-animal would be about 15 mmol kg^-1 h^-1. Measured in vivo was about 20% of this figure (Wright et al. 1986; Lin and Randall, 1990). Given the nature of the extrapolation, the probable greater diffusion barriers in vivo (e.g. unstirred layers) and the fact that the equilibrating water cannot be truly zero in vivo, owing to regional ventilation-perfusion mismatch, the agreement is not unreasonable.

We cannot entirely eliminate the possibility that there is some component of diffusion limitation in the present assay. However, given the fact that the CO₂ excretion rate of whole blood in the assay appeared to be somewhat higher, rather than lower, than in vivo rates, this is unlikely to be a serious problem. The assay can measure experimentally induced elevations in rate well above normal control levels (e.g. Table 1, Fig. 3), as well as inhibitions below control levels (Figs 6, 7; Perry et al. 1991). We have also used the assay to measure large changes in CO₂ excretion rate of blood resulting from changes in the physiological condition of the fish from which the blood was sampled (C. M. Wood, unpublished results).

It is of interest to compare the performance of the present assay with that of the ‘boat’ assay, which has been commonly employed until now for trout blood (Haswell and Randall, 1976; Heming and Randall, 1982; Tufts et al. 1988). The boat assay itself can be run in as little as 1 min, but only after a 3–10 min settling period for initial pressure equilibration. As CO₂ is evolved into the gas phase, the back-pressure on the system progressively rises, so the assay quickly becomes nonlinear. The influence of this back-pressure at the gas/fluid interface probably contributes to the foaming artefact reported to block CO₂ excretion by some workers (Haswell and Randall, 1976; Heming and Randall, 1982), but not by others (Tufts et al. 1988) when the assay is run with plasma or whole blood. The coefficient of variation has not been reported, but would appear to be about 10 % based on recalculation of published data. The most serious drawback of the boat assay is that the acid-base status of the sample is grossly perturbed. The blood is initially incubated in a pH 6.8 phosphate buffer (HCO₃^- -free and Cl^- -free) followed by acute exposure to 100mmoll^-1 NaHCO₃ plus lOmmoll^-1 NaOH. Normal plasma is no longer present, pH and must be greatly altered, and substrate concentration ([HCO₃^-]) is suddenly dramatically elevated. In the light of the present findings on the effects of acute HCO₃^- elevation on RBC CO₂ excretion rates (Fig. 3, discussed below), it is not surprising that the CO₂ excretion rate reported for trout blood of 25 % haematocrit by Tufts et al. (1988) was about 1000 μmol ml^-1 h^-1, 70 times greater than the present values and 350 times greater than in vivo rates.

In the light of the large amounts of carbonic anhydrase normally present in erythrocytes (Maren, 1967; Perry and Laurent, 1990), it was initially surprising that whole-blood CO₂ excretion rates were only about fivefold higher than plasma rates. Nevertheless, this is in accord with findings of the boat assay for trout blood when wetting agents were used to overcome the foaming artefact (Heming and Randall, 1982), and also with the haematocrit versus relationship established by Perry et al. (1982),in vivo. The presence of large amounts of carbonic anhydrase in the RBCs is not expressed as vastly elevated whole-blood CO₂ excretion rates relative to that of plasma. This is because the rate of net HCO₃^- entry into the RBCs (via C1^-/HCO₃^- exchange), rather than carbonic anhydrase activity itself, is thought to be the rate-limiting step in CO₂ excretion (Perry, 1986).

Nevertheless, if this were the only factor involved, we would expect a linear relationship between haematocrit and CO₂ excretion rate, whereas this clearly did not occur (Fig. 3). Instead, rates tended to saturate at higher haematocrits. The only other examination of this relationship in fish blood, by Tufts et al. (1988) using the boat assay, found exactly the opposite, a virtually exponential increase with haematocrit (the rate increased more than eightfold as haematocrit rose from 10 to 30 %). Although it is difficult to account for such an exponential relationship, there are several possible explanations for the saturating relationship seen in the present study.

One obvious possibility is substrate limitation - i.e. depletion of plasma [HCO₃^-] over the course of the assay by the greater number of RBCs. However, this appears to be unlikely, because measured plasma HCO₃^- levels at the end of the 3 min assay were essentially the same at all haematocrits within a pool. In any event, under equilibrium conditions the blood CO₂ excretion rate was only marginally dependent on plasma [HCO₃^-] (Fig. 5). A more likely explanation is some sort of physical blocking phenomemon associated with more concentrated RBCs - e.g. interference with substrate access to C1^-/HCO₃^- exchange sites, a greater volume occupied by unstirred layers on the RBC surfaces, viscosity effects on convective mixing or lower diffusion rates through the RBCs or at the gas phase/blood interface (Gros and Moll, 1971; Klocke, 1988). It would be interesting to know whether such effects occur in vivo or are an artefact of the assay conditions.

The dependence of the CO₂ excretion rate on the pre-equilibration of the blood (Table 1) was the expected result, since the gradient from RBC to gas phase provides the net driving force for the CO₂ excretion process in the assay, as in vivo. The relative independence of the CO₂ excretion rate from the preequilibration plasma [HCO₃^-] (Fig. 5), but its sensitivity to acutely altered plasma [HCO₃^-] (Fig. 3), is more interesting. In the blood of resting trout (i.e. no adrenergic stimulation) under equilibrium conditions, HCO₃^- and CC (and H⁺) are probably passively distributed across the RBC membrane according to a Gibbs-Donnan equilibrium (Wood et al. 1982; Heming et al. 1986; Nikinmaa et al. 1987). Thermodynamically, there will be no driving force for net HCO₃^- entry, regardless of the plasma [HCO₃^-], until the assay is actually started by exposure to a gas phase of zero, thereby lowering intracellular [HCO₃^-] by catalysed conversion to CO₂. At this point, the HCO₃^- distribution ratio will no longer be dictated by the membrane potential, and net HCO₃^- entry/Cl^- exit will occur because of the electrochemical disequilibrium. This disequilibrium will tend to be slightly greater with higher extracellular [HCO₃^-], explaining the shallow dependence of CO₂ excretion rate on plasma [HCO₃^-] (Fig. 5).

In contrast, when the plasma [HCO₃^-] is acutely raised at the start of the assay, the assay starts under non-equilibrium conditions. Immediately, there is a large driving force for net HCO₃^- entry, and the CO₂ excretion rate is very sensitive to this non-equilibrium plasma HCO₃^- level (Fig. 3). This phenomenon undoubtedly accounts for the very high CO₂ excretion rates in the boat assay (see above), where the blood is acutely exposed to 100 mmol I^-1 NaHCO₃.

When trout blood was acidified to simulate the metabolic acidosis normally seen after exhaustive exercise (⁠⁠; Turner et al. 1983; Milligan and Wood, 1986), the addition of adrenaline or noradrenaline (10⁻⁸-10⁻⁶mol^-1) caused a clear inhibition of RBC CO₂ excretion (Figs 6, 7). This is in accord with the original observation (unpublished data of S. F. Perry and T. A. Heming presented by Wood and Perry, 1985), and supports the ‘CO₂ retention theory’. The time course of the inhibitory effect (maximal at 5–8 min after catecholamine addition, persistent at 15–18 min and over by 30–60 min; Fig. 6) is not dissimilar to the time course of elevation in vivo after exhaustive exercise in the rainbow trout (Wood and Perry, 1985).

The negative results of Tufts et al. (1988) are probably attributable to a combination of factors: the limitations of the boat assay discussed earlier, the transitory nature of the response (Fig. 6), and the general lack of response at normal blood pH (see also Perry et al. 1991, for additional data on this point). Our original reasons for tests at reduced pHe were three-fold: (i) this is the relevant situation in vivo when ‘CO₂ retention’ occurs; (ii) the original experiments of S. F. Perry and T. A. Heming (unpublished results) reported by Wood and Perry (1985) were conducted with trout RBCs in Cortland saline, which probably had a lower pHe; and (iii) it is now known that adrenergic activation of RBC Na⁺/H⁺ exchange (the RBC pHi regulatory response) is markedly sensitized at acidotic pHe values (Nikinmaa et al. A98T, Borgese et al. 1987; Cossins and Kilbey, 1989; Milligan et al. 1989).

The magnitude of the inhibition of RBC CO₂ excretion by catecholamines was clearly correlated with the magnitude of the pHe—pHi response (Fig. 7). It is possible that the two responses share a common pathway, and there are theoretical reasons why the two could be mechanistically related. However, the time courses were certainly not identical (i.e. the pHe—pHi effect lasted long after the inhibition of RBC CO₂ excretion had disappeared; Fig. 6), and correlation (Fig. 7) does not prove causation. The mechanism of the inhibitory response is addressed in the subsequent paper (Perry et al. 1991).

This research was supported by NSERC operating and equipment grants to C.M.W. and S.F.P. We are grateful to Dr R. Motais for helpful discussions.