Two different oxygen sensors regulate oxygen‐sensitive K+ transport in crucian carp red blood cells

The O2 dependence of ouabain‐independent K+ transport mechanisms has been studied by unidirectional Rb+ flux analysis in crucian carp red blood cells (RBCs). The following observations suggest that O2 activates K+–Cl− cotransport (KCC) and deactivates Na+–K+–2Cl− cotransport (NKCC) in these cells via separate O2 sensors that differ in their O2 affinity. When O2 tension (PO2) at physiological pH 7.9 was increased from 0 to 1, 4, 21 or 100 kPa, K+ (Rb+) influx was increasingly inhibited, and at 100 kPa amounted to about 30% of the value at 0 kPa. This influx was almost completely Cl− dependent at high and low PO2, as shown by substituting Cl− with nitrate or methanesulphonate. K+ (Rb+) efflux showed a similar PO2 dependence as K+ (Rb+) influx, but was about 4–5 times higher over the whole PO2 range. The combined net free energy of transmembrane ion gradients favoured net efflux of ions for both KCC and NKCC mechanisms. The KCC inhibitor dihydroindenyloxyalkanoic acid (DIOA, 0.1 mm) abolished Cl−‐dependent K+ (Rb+) influx at a PO2 of 100 kPa, but was only partially effective at low PO2 (0–1 kPa). At PO2 values between 0 and 4 kPa, K+ (Rb+) influx was further unaffected by variations in pH between 8.4 and 6.9, whereas the flux at 21 and 100 kPa was strongly reduced by pH values below 8.4. At pH 8.4, where K+ (Rb+) influx was maximal at high and low PO2, titration of K+ (Rb+) influx with the NKCC inhibitor bumetanide (1, 10 and 100 μm) revealed a highly bumetanide‐sensitive K+ (Rb+) flux pathway at low PO2, and a relative bumetanide‐insensitive pathway at high PO2. The bumetanide‐sensitive K+ (Rb+) influx pathway was activated by decreasing PO2, with a PO2 for half‐maximal activation (P50) not significantly different from the P50 for haemoglobin O2 binding. The bumetanide‐insensitive K+ (Rb+) influx pathway was activated by increasing PO2 with a P50 significantly higher than for haemoglobin O2 binding. These results are relevant for the pathologically altered O2 sensitivity of RBC ion transport in certain human haemoglobinopathies.

J Physiol 575.1 1992). In all species studied so far, including man, O 2 has opposing effects on the two RBC transport systems, activating KCC and deactivating NKCC (Gibson et al. 2000). Because the effects of O 2 are rapid, they appear to depend on constitutive signal transduction pathways, rather than on hypoxia-induced changes in gene expression (López-Barneo et al. 2001). However, despite continuous efforts, the nature of the O 2 sensor(s) and the transduction pathway(s) modulating RBC ion transport are still unknown.
Studies on rainbow trout RBCs -which appear to lack the NKCC but express high activities of KCC and a β-adrenergically activated Na + -H + exchange system (NHE) -have suggested that haemoglobin (Hb) is a component of the O 2 sensor transducing the effects of molecular O 2 on NHE and KCC activity (Motais et al. 1987;Borgese et al. 1991;Nielsen et al. 1992). Subsequently, Hb has also been discussed as the O 2 sensor modulating NKCC and KCC in mammalian and avian RBCs, where the O 2 affinities for KCC activation or NKCC deactivation approximately match the O 2 affinity of Hb inside RBCs (Speake et al. 1997;Muzyamba et al. 1999;Flatman, 2005). However, O 2 affinity of KCC in rainbow trout RBCs is much lower than that of Hb, casting doubt on a direct role of bulk Hb oxygenation in modulating KCC in this species (Berenbrink et al. 2000). It is presently not known whether the occurrence of O 2 sensors with Hb-like and non-Hb-like (low) O 2 affinity is species-specific, or whether they may even occur simultaneously in RBCs of a single species.
Here we use crucian carp RBCs as a model that expresses significant levels of both KCC and NKCC, and dissect their O 2 affinities. RBCs from hypoxia-tolerant carp species are especially suited for distinguishing between Hb-like and non-Hb-like (low) affinity O 2 sensors, because of an unusually high Hb O 2 affinity, which is coupled with approximately threefold lower normoxic resting arterial P O 2 values as compared with rainbow trout, birds and mammals (P O 2 for half-maximal blood O 2 saturation and for normoxic arterial blood at rest in carp species is typically 0.4-0.7 and 3.2-4.0 kPa, respectively; Prosser, 1950;Burggren, 1982;Knudsen & Jensen, 1998). Our results suggest that crucian carp RBCs simultaneously express two different O 2 sensors, one with Hb-like O 2 affinity that governs NKCC, as found in mammalian and avian RBCs, and another with significantly lower O 2 affinity that governs KCC, as found in rainbow trout RBCs. These experiments characterize a model system for the coordinated regulation of differentially O 2 -sensitive ion transport systems in a single cell type, and are relevant for understanding pathologically altered O 2 sensitivities of RBC ion transporters in human sickle cell disease, and α and β thalassaemia (Olivieri et al. 1994;Gibson et al. 1998;Drew et al. 2004).

Animals
Crucian carp (Carassius carassius) were obtained either from a pond in the area of Turku, Finland (mass 0.7-1.8 kg, total length 31-40 cm, n = 20), or from a pond near Ipswich, UK (20-123 g, 12.1-18.3 cm, n = 25). Rainbow trout (Oncorhynchus mykiss, 130-214 g, 24.5-28.0 cm, n = 4) were purchased from a commercial fish farm near Turku, Finland. Both species were kept indoors at 15 • C in running, dechlorinated tap water for at least 1 week prior to experimentation at the fish holding facilities of A bo Akademi University, Turku, Finland, or of the School of Biological Sciences, University of Liverpool, UK. Experiments were performed from July to early December to minimize seasonal variations in the magnitude of RBC ion-transport pathways (Berenbrink & Bridges, 1994). As no differences in K + (Rb + ) influx were apparent between the RBCs of the two crucian carp stocks, despite marked difference in their size, results were combined.

Chemicals and solutions
Inorganic salts, and dimethyl sulfoxide (DMSO), d-glucose, imidazole, ouabain and perchloric acid (PCA) were obtained from Merck, Darmstadt, Germany. Bumetanide, ethyl-m-aminobenzoate (MS 222), Hepes, methanesulphonic acid, N -methyl-d-glucamine (NMDG), sodium heparin and Tris were purchased from Sigma-Aldrich Chemical Company, while Triton X-100 was from Serva, Heidelberg, Germany. The radioactive tracer 86 Rb + (as RbCl) was obtained from NEN Life Science Products, Belgium, and the KCC inhibitor DIOA (dihydroindenyloxyalkanoic acid) was from Research Biochemicals, Natick, MA, USA. Stock solutions (10 mm) of DIOA and ouabain were prepared in ethanol and DMSO, respectively, and the chemicals were used at a final concentration of 0.1 mm. Bumetanide (10 mm) was prepared in ethanol and used at final concentrations of 1, 10 and 100 μm. The final volume of the respective solvents did not exceed 1% of the volume of RBC suspensions. Standard fish saline for RBCs of crucian carp and rainbow trout consisted of (mm): 125.5 NaCl, 3 KCl, 1.5 MgCl 2, 1.5 CaCl 2 , 5 d-glucose and 20 Hepes, adjusted with NaOH to pH 7.97 at 15 • C (Berenbrink et al. 2000;Völkel et al. 2001). The pH of standard saline was varied between 6.9 and 8.4 by adding NaOH or HCl. In Cl − -free salines, Cl − salts were replaced either by the respective nitrate salts or, alternatively, by the respective cation hydroxides. In the latter case, pH was adjusted using methanesulphonic acid, creating a Cl − -free saline with methanesulphonate as the principal anion. Na + -free saline was prepared by equimolar replacement of NaCl with NMDG. In this case pH was adjusted with HCl.

Blood sampling and preparation of RBCs
Fishes were normally killed by a sharp blow on the head and immediate exsanguination by caudal venipuncture using heparinized hypodermic syringes. Large crucian carp were immersed in an overdose of anaesthetic (1 g l −1 MS 222, neutralized with Tris salt) until all movement ceased, before exsanguination. Procedures were carried out in accordance with national ethical committee guidelines. RBCs were washed three times in 3-5 vols of ice-cold standard saline, each time removing the buffy coat. The resulting RBC suspensions were adjusted to half the original blood haematocrit value, oxygenated by contact with air and stored at 5 • C for at least 16 h to allow for stabilization of cell volume and cellular ion content.

Experimental procedure
Immediately before experimentation, RBCs were washed in ice-cold standard saline. In the case of small crucian carp, RBCs from three or four animals were pooled to obtain a sufficiently large volume. For ion replacements, RBCs were washed three times with a 10-fold excess of Cl − -or Na + -free saline, allowing 5 min for ion equilibration at room temperature after each wash. RBCs were then resuspended at the original blood haematocrit value in the respective saline, and subjected to 45 min standard pre-equilibration at 15 • C in shaking glass tonometers (Eschweiler, Kiel, Germany) with a water-vapour-saturated gas mixture of 5% air-95% N 2 , provided by mass flow controllers (Gf-3MP Cameron Instruments, Port Aransas, TX, USA) or a precision gas-mixing pump (Wösthoff KG, Bochum, Germany). Standard pre-equilibration at the resulting P O 2 of 1 kPa is frequently used to deactivate O 2 -dependent KCC in fish RBCs with minimal effects on cellular ATP levels (Nielsen et al. 1992;Berenbrink et al. 1997Berenbrink et al. , 2000Völkel et al. 2001). After 45 min, the haematocrit value was determined by microcentrifugation (Micro-Compur M110, Compur Elektronik, München, Germany), and equilibration continued in humidified gases at P O 2 values of 0, 1, 4, 21 or 100 kPa. After 10 min of experimental equilibration, RBC suspensions were diluted with 9 vols pre-equilibrated standard saline adjusted to the desired pH range, for determination of unidirectional K + ( 86 Rb + ) fluxes, Hb O 2 saturation and final pH.

Determination of K + fluxes
Unidirectional K + fluxes were determined in the presence of 0.1 mm ouabain using 86 Rb + as a tracer, substituting for K + , as previously described (Berenbrink et al. 2000). Briefly, for influx measurements, 11.1-18.5 kBq ml −1 86 Rb + was added to the diluted RBC suspensions (standard salines with or without ion replacements, 2-3% haematocrit) and after predefined time points 200 μl aliquots of the suspension were removed. RBCs were immediately washed three times by centrifugation and resuspension in ice-cold wash solution (100 mm MgCl 2 , 10 mm imidazole or Hepes, pH 7.97 at 15 • C). After final centrifugation, the supernatant was removed and the RBC pellet lysed (0.5 ml 0.05 vol% Triton X-100) and deproteinized (0.5 ml 0.6 m PCA). Cellular 86 Rb + activity was determined by Cerenkov radiation. K + (Rb + ) influx was calculated by linear regression from the rate of increase in cellular 86 Rb + activity with time and the extracellular 86 Rb + activity per extracellular K + concentration (3 mm). Influx is expressed in millimoles of K + (Rb + ) per hour and per litre of cells. RBC volume was determined from haematocrit measurements at the end of the pre-equilibration period. For efflux measurements, RBCs at high haematocrit (30-40%) were loaded with 86 Rb + for ≥ 3.5 h at 20 • C (37 kBq ml −1 suspension). After standard pre-equilibration, P O 2 was changed to the experimental value and 10 min later RBCs were diluted 20-fold in saline pre-equilibrated with the same gas mixture. The accumulation of 86 Rb + activity in the extracellular medium (standard saline) was followed for 15 min by centrifugation of aliquots at predefined time points and processing the supernatant as described above. As in our previous study (Berenbrink et al. 2000), 86 Rb + release appeared linear with time and K + (Rb + ) efflux was determined from the initial 86 Rb + release rate and the initial cellular 86 Rb + activity per cellular K + concentration. The latter was 102.2 ± 1.3 mmol per litre of RBCs (mean value ± s.e.m., n = 6 animals) in cells processed the same way as for K + (Rb + ) efflux determinations and measured by atomic absorption spectrometry (Perkin-Elmer 2380). Efflux is expressed in mmol K + (Rb + ) per hour and per litre of cells.

Hb O 2 saturation and pH measurements
Hb O 2 saturation was determined according to the method of Tucker (1967) in dilute RBC suspensions under the same experimental conditions as for K + (Rb + ) flux measurements. Experimental procedures and calculations were identical to those used in our previous study on rainbow trout RBCs (Berenbrink et al. 2000). The pH of final RBC suspensions used for Hb O 2 binding studies and K + (Rb + ) flux measurements was checked using a thermostatted (14.9-15.1 • C) capillary glass electrode with calomel reference (Radiometer BMS 3 Mk 2) and a pH meter (Radiometer PHM 72). The pH electrode assembly was calibrated at the experimental temperature with precision buffers (Radiometer). The final pH of RBC suspensions diluted with four different pH-adjusted salines was 6.942 ± 0.014, 7.387 ± 0.005, 7.931 ± 0.017 and 8.375 ± 0.036 (means ± s.e.m., n = 3). For readability, experiments at these pH values are referred to as pH 6.9, 7.4, 7.9 and 8.4 experiments. Oxygenation-induced J Physiol 575.1 pH changes of the extracellular medium amounted to maximally 0.07 pH units, as estimated by comparing the pH difference between O 2 -and N 2 -equilibrated RBCs in the four different salines.

Intracellular ion concentrations and net free energy for cotransport
Intracellular Na + , K + and Cl − concentrations ([Na + ] i , [K + ] i and [Cl − ] i ) were measured in RBCs that had been pre-equilibrated for 45 min under standard conditions and subsequently exposed to a P O 2 of 21 kPa for 10 min. Hence values refer to the time point where K + (Rb + ) flux determinations were started. RBC ion and water content were determined as described before (Berenbrink & Bridges, 1994). Briefly, RBCs were separated from the extracellular medium by rapid centrifugation in narrow 400 μl capacity Eppendorf tubes and quickly frozen in liquid nitrogen. Frozen tubes were cut with a razor blade 2 mm below the boundary between supernatant and pellet. The remaining pellet was weighed and deproteinized by addition of 200 μl 0.6 m PCA. After centrifugation (5 min, 10 000 g), ion concentrations in the supernatant were determined, Na + and K + by atomic absorption spectrometry (Perkin-Elmer 2380), and Cl − coulometrically using a chloride titrator (CMT 10, Radiometer, Copenhagen, Denmark). In parallel samples, pellets were preweighed and then dried to constant weight (≥ 40 h at 80 • C) for determination of cellular water content. This allowed the intracellular ion concentrations to be expressed in millimoles per litre cell water for calculation of the net free energy in transmembrane ion gradients.
The combined net free energy change, G, for the transport of 1 mol Na + , 1 mol K + and 2 mol Cl − into the cell was calculated according to Russell (2000): where R is the gas constant (8.314 J mol −1 K −1 ), T is absolute temperature (288 K) and subscripts 'i' and 'o' refer to intra-and extracellular ion concentrations, respectively. The net free energy change for the transport of 1 mol K + and 1 mol Cl − into the cell was calculated according to Lauf (1985): In both cases, positive values for G indicate that ion gradients favour a net outward direction of transport.

Data analysis and representation
Values are expressed as means ± s.e.m. of n experiments on RBCs of separate animals, or on RBCs pooled from separate groups of animals. In contrast to classical model organisms, animals were not from inbred laboratory lines and K + (Rb + ) influx and efflux values were somewhat variable between individuals. A similar degree of variability in ion transport activity is also evident in RBCs from other species, including humans (Ellory et al. 1985). In yet other studies, interindividual variation may be obscured by the common presentation of mean values and error bars of multiple measurements in a single, representative, experiment.
Statistical differences between treatments were assessed using one-way or two-way analyses of variance, as appropriate, followed by Tukey's test for pairwise comparisons (SigmaStat version 2.03). Data deviating from a normal distribution were transformed before further analysis according to x = log(x + 1) or x = 1/(x + 1) (Sachs, 1988). Statistical significance was accepted at P < 0.05.
P O 2 values for half-maximal K + (Rb + ) influx (P 50 ) were calculated by non-linear curve fitting (SigmaPlot version 8) using equations for simple saturation curves of the form or simple hyperbolic decay curves of the form The P 50 for Hb O 2 binding was calculated from non-linear curve fits according to a simple saturation curve where y is fractional Hb O 2 saturation, x is P O 2 , and b is P 50 . Using more complex equations for these fits, e.g. by introducing sigmoidicity constants (Berenbrink et al. 1997(Berenbrink et al. , 2000, did not improve the curve fits. The fractional amount of deoxygenated Hb was calculated as 1 -(fractional amount of oxygenated Hb).

Results
K + transport pathways were characterized using unidirectional 86 Rb + tracer fluxes. All K + (Rb + ) fluxes were measured in the presence of 100 μm ouabain to inhibit the Na + -K + -ATPase, and thus refer to ouabain-insensitive fluxes. Figure 1A shows the O 2 sensitivity of K + (Rb + ) influx at pH 7.9, which is close to arterial blood pH in normoxic, resting crucian carp at 15 • C (Van den Thillart & Van Waarde, 1990). K + (Rb + ) influx was moderate at P O 2 values between 4 and 100 kPa, but increased by about 150% at lower O 2 tensions of 1 and 0 kPa. These fluxes were Cl − dependent over the whole P O 2 range studied, as K + (Rb + ) influx was nearly completely abolished when Cl − was replaced by either nitrate or methanesulphonate in the incubation medium.
The major Cl − -dependent K + transport pathways in RBCs of higher vertebrates are NKCC and KCC (Lauf et al. 1992). In human RBCs, 100 μm DIOA inhibits KCC without affecting NKCC (Garay et al. 1988  K + (Rb + ) influx almost completely and to a similar extent as Cl − replacement (Fig. 1A). At lower P O 2 values DIOA inhibition became progressively less complete, revealing a low-P O 2 -activated, relatively DIOA-insensitive K + (Rb + ) influx pathway, which showed a significantly higher activity than the small basal flux observed in the absence of Cl − . J Physiol 575.1 Replacing external Na + by NMDG yielded somewhat reduced K + (Rb + ) influx values at low P O 2 values, although the effect was statistically not significant (Fig. 1B). Independence from extracellular Na + has frequently been used as supporting evidence for excluding NKCC as the mechanism for K + tracer fluxes (e.g. Gillen et al. 1996;Muzyamba et al. 1999;Mercado et al. 2000). Importantly, however, this cannot be taken as evidence against the involvement of NKCC in Cl − -dependent K + (Rb + ) tracer fluxes because in the absence of external Na + , the NKCC can perform a partial reaction and operate in a K + (Rb + ) self-exchange mode in many systems, provided some internal Na + is present (Lauf et al. 1987;Lytle et al. 1998).
In most cell types, the combined net free energy in the physiological transmembrane ion gradients greatly favours net efflux of cotransported ions via the KCC (Lauf et al. 1992). In contrast, the NKCC in RBCs is close to equilibrium and may be operating in a net inward or outward direction, depending on physiological plasma K + concentrations (Duhm & Göbel, 1984). Unidirectional K + (Rb + ) efflux showed a similar O 2 dependence as K + (Rb + ) influx, but was about 4-5 times higher over the whole P O 2 range, predicting a net outward direction of ouabain-insensitive K + transport pathways at low and high P O 2 (Fig. 1C). Figure 2 shows that at P O 2 21 kPa the net free energy of the measured ion gradients across the crucian carp red cell membrane was compatible with a net efflux via both NKCC and KCC.
Below a P O 2 of 21 kPa, the predicted driving force for net efflux via NKCC is even higher, because as in mammals and birds, teleost fish RBCs possess a powerful anion exchange system (band 3; Jensen & Brahm, 1995). This transporter equilibrates Cl − passively across the RBC membrane (e.g. Berenbrink & Bridges, 1994). At low P O 2 values, Hb deoxygenation and the associated neutralization of . Net free energy for coupled Na + -K + -2Cland K + -Cl − cotransport in crucian carp red blood cells Lines show the calculated relationships between intracellular Cl − concentration and G for electroneutral cotransport via K + -Cl − cotransport (KCC) and Na + -K + -2Cl − cotransport (NKCC) (stoichiometries of 1:1 and 1:1:2, respectively) in air-equilibrated crucian carp red blood cells. Intracellular Na + and K + concentrations under these conditions were 16 ± 1 and 131 ± 5 mmol (l cell water) −1 , respectively (means ± S.E.M., n = 3). Extracellular Na + , K + and Cl − concentrations (136, 3 and 135 mM) refer to concentrations in the medium, which was used for diluting RBC suspensions at the start of K + (Rb + ) flux measurements. Filled circles and error bars indicate values at the measured intracellular Cl − concentration of 68 ± 6 mmol (l cell water) −1 (mean ± S.E.M., n = 3). negative charges on Hb by Bohr protons induce a Cl − shift from the external medium into the cytosol (Hladky & Rink, 1977). This effect is expected to be even larger in modern teleost fishes because of their much stronger Bohr effect as compared to mammalian and bird Hbs (Berenbrink et al. 2005). Thus, the increase of [Cl − ] i at low P O 2 (Fuchs & Albers, 1988) will create even more favourable conditions for net ion efflux via both NKCC and KCC. In mammalian RBCs, NKCC activity is strongly stimulated by alkaline pH (Flatman, 1991), whereas KCC is stimulated by acidification (e.g. Speake et al. 1997). In contrast, KCC in fish RBCs is strongly activated by alkaline pH (Berenbrink et al. 2000;Völkel et al. 2001). K + (Rb + ) influxes in crucian carp RBCs at low and high P O 2 fundamentally differed in their pH dependence (Fig. 3A). Between P O 2 values from 0 to 4 kPa, pH did not significantly affect K + (Rb + ) influx. However, at P O 2 values of 21 and 100 kPa, K + (Rb + ) influx was significantly reduced by acidification below pH 8.4 and reached close to baseline levels at pH 7.4 and 6.9. This was similar to the inhibition of KCC by acidification in rainbow trout RBCs measured under the same experimental conditions (Fig. 3B). In rainbow trout, K + (Rb + ) influx was minimal at P O 2 values of 1 and 0 kPa, independent of pH.
DeoxyHb has been implicated as a transducer for the effects of O 2 on membrane ion transport in RBCs of several species (e.g. Motais et al. 1987;Flatman, 2005). Therefore the fraction of deoxyHb in crucian carp RBCs was measured under the same experimental conditions as in Fig. 3A. Figure 3C shows that between pH 6.9 and 7.9, activation of K + (Rb + ) influx by lowering P O 2 closely matched the increase of deoxyHb. Figure 4 illustrates the respective P O 2 values at which half-maximal K + (Rb + ) influx and half-maximal Hb deoxygenation were achieved (P 50 values). At both pH 6.9 and 7.9, P 50 for Hb and K + (Rb + ) influx were in the same range and not significantly different. Figure 3D shows that the three-dimensional pH and P O 2 profile of K + (Rb + ) influx in crucian carp RBCs can, in principle, be explained by the action of two different

Figure 3. Effects of O 2 and pH on K + (Rb + ) influx and Hb deoxygenation
After standard pre-equilibration (see Fig. 1), RBCs were exposed to P O 2 values between 0 and 100 kPa. After 10 min they were diluted in salines with pH values between 6.9 and 8.4, and these were pre-equilibrated with the same experimental P O 2 , and K + (Rb + ) influx (A) and fractional Hb deoxygenation (C) were measured. For clarity, mean values are shown with S.E.M. in one direction only (circles and bars, respectively). A, two-way analysis of variance with P O 2 and pH as factors revealed a highly significant interaction between the two factors (P ≤ 0.001). Thus, the effect of P O 2 on K + (Rb + ) influx (n = 3-10) depended on pH. a Significantly different from values at 0 kPa at the same pH; b significantly different from values at pH 8.4 at the same P O 2 . B, comparable data on rainbow trout RBCs (n = 3-4), partly taken from Berenbrink et al. (2000). C, fractional deoxyHb in crucian carp RBCs, measured under identical experimental conditions as for K + (Rb + ) influxes (n = 3). D, reconstruction of the basic shape of P O 2 and pH-dependent K + (Rb + ) influx as seen in A. At each x, y coordinate, the fraction of deoxyHb from C was added to the fractional K + (Rb + ) influx taken from B, and the sum is plotted as the z-value. This demonstrates that fluxes in crucian carp RBCs (A) can be modelled in principle by the sum of a deoxyHb-activated flux and a deoxyHb-independent flux, the latter being similar to KCC in rainbow trout RBCs. See text for further details. Note the logarithmic scale for P O 2 . K + (Rb + ) influx pathways. One pathway is taken to have an activity which depends directly on the amount of deoxyHb as obtained from Fig. 3C. The other component is taken to have the same P O 2 and pH dependence as the KCC of trout (Fig. 3B). The diagram shows for each J Physiol 575.1 pair of x, y co-ordinates (P O 2 , pH) the sum of the fractional amounts of these two putative K + (Rb + ) influx components (z-axis). This yields the same basic shape as observed in Fig. 3A, and it can be seen that allowing for the differences in pH dependency of K + (Rb + ) influx at high P O 2 between crucian carp and rainbow trout RBCs (Fig. 3A and B; Völkel et al. 2001) would even further increase the similarity between Fig. 3A and D.
To test the idea of two different K + (Rb + ) transport pathways, which are differentially modulated by O 2 , further experiments were carried out using bumetanide at pH 8.4, where K + (Rb + ) influx was maximally activated at both high and low P O 2 . At low concentrations, bumetanide specifically inhibits NKCC in various vertebrate and invertebrate tissues with a 50% inhibitory concentration (IC 50 ) of about 0.1 μm (Russell, 2000). At higher concentrations bumetanide also inhibits KCC, albeit with a more than 1000-fold higher IC 50 (∼180 μm for KCC1 and ∼900 μm for KCC4; Mercado et al. 2000). Figure 5A shows that K + (Rb + ) influx at low P O 2 was already significantly inhibited by more than 50% in the presence of 1 μm

Figure 4. O 2 affinity of K + (Rb + ) influx and Hb at different pH values
O 2 affinity is expressed as the P O 2 at which half-maximal change in K + (Rb + ) influxes or Hb O 2 binding occurred (P 50 , means ± S.E.M). At each pH, P 50 values for Hb (filled bars) were calculated by fitting hyperbolic saturation curves to data from three independent experiments (same data as in Fig. 3C). P 50 values for K + (Rb + ) fluxes (n = 5) were calculated by fitting hyperbolic saturation curves (open bars) or hyperbolic decay curves (hatched bars) to the data, as appropriate. These curve fits allowed for a variable, basal O 2 -independent K + (Rb + ) influx component. At pH 6.9, all flux data from Fig. 3A were used. At pH 7.9, only flux data from animals where K + (Rb + ) influx at P O 2 0 kPa was at least two times higher than at 100 kPa were used (5 out of the 8 independent experiments incorporated into Fig. 3A). At pH 8.4, total K + (Rb + ) influx could be separated into a 1 μM bumetanide-sensitive component and a 100 μM bumetanide-resistant component (see Fig. 5). n.s., P 50 for Hb and total flux did not differ significantly from each other at pH 6.9 and 7.9. At pH 8.4, P 50 for bumetanide-sensitive K + (Rb + ) influx and Hb also did not differ significantly (n.s.). However, both were significantly smaller than the P 50 of bumetanide-resistant K + (Rb + ) influx ( * * ).
bumetanide. Increasing the concentration to 10 μm led to significant further reductions, and 100 μm essentially abolished K + (Rb + ) influx at low P O 2 . With IC 50 values greater than 100 μm, any KCC isoform should be less than 50% inhibited at this bumetanide concentration. Therefore the near elimination of K + (Rb + ) influx by 100 μm bumetanide suggests that the flux at P O 2 0-1 kPa is almost entirely due to NKCC. At an intermediate P O 2 of 4 kPa, significant inhibition of K + (Rb + ) influx required 10 μm bumetanide. At 21 and 100 kPa P O 2 , only 100 μm bumetanide caused a significant reduction of K + (Rb + ) influx. Thus, at higher P O 2 , an increasing fraction of K + (Rb + ) influx was carried by a more bumetanide-resistant pathway such as KCC. However, the moderate, but consistent reductions of K + (Rb + ) influx by low bumetanide concentrations (1 and 10 μm) even at high P O 2 values suggest that NKCC was still partially active at 21 and 100 kPa.
A change in the characteristics of K + (Rb + ) transport between low and high P O 2 is also indicated by the biphasic shape of the O 2 profile of the control flux. Thus, in the absence of bumetanide, K + (Rb + ) influx at a P O 2 of 4 kPa was significantly lower than at 0 kPa, whereas the fluxes at higher P O 2 values did not differ significantly from those at 0 kPa. In contrast, in the presence of bumetanide at 1, 10 and 100 μm, fluxes at 100 kPa P O 2 were always significantly higher than at 0 kPa (Fig. 5A).
In the presence of 100 μm bumetanide, NKCC is expected to be completely inhibited, and this allowed the determination of the P 50 value for the O 2 -activated, bumetanide-resistant flux mechanism (KCC). In addition, calculation of the 1 μm-bumetanide-sensitive flux in Fig. 5B allowed the determination of the P 50 value for the activation of NKCC by low P O 2 . Figure 4 shows that the P 50 for the relative bumetanide-resistant K + (Rb + ) influx pathway (KCC) was significantly higher than the P 50 values for NKCC and Hb. In contrast, P 50 values for Hb and NKCC were not significantly different.

Discussion
Previous studies aimed at unravelling the mechanism of O 2 sensing in RBCs have tacitly assumed that there is one mechanism by which molecular O 2 governs the activity of RBC ion transport pathways. Here we suggest that two O 2 sensors, which differ significantly in their O 2 affinity, modulate two different O 2 -dependent K + (Rb + ) transport pathways in crucian carp RBCs.

Identification of two separate O 2 -dependent K + (Rb + ) transport pathways
We propose that the first K + (Rb + ) flux pathway, whose activity positively correlated with the fraction of deoxyHb in crucian carp RBCs, is carried by NKCC. At low P O 2 , where this mechanism was maximally activated, K + (Rb + ) influx was virtually abolished by replacement of extracellular Cl − with nitrate or methanesulphonate. The mechanism was highly sensitive to the specific NKCC inhibitor bumetanide, with 1 μm of the drug causing more than 50% inhibition, and 100 μm bumetanide almost abolishing the flux. The mechanism was progressively activated by decreasing P O 2 , similar to NKCC in turkey, chicken, human and ferret RBCs (Muzyamba et al. 1999;Drew et al. 2004;Flatman, 2005).
The second K + (Rb + ) influx mechanism was maximally activated by high P O 2 and showed the characteristics of KCC. It was abolished by replacement of Cl − with nitrate or methanesulphonate, or by 100 μm DIOA. In contrast to NKCC, it showed low bumetanide sensitivity, with 50% inhibition requiring 100 μm bumetanide or more, like the KCC in other systems (Ellory et al. 1985;Mercado et al. 2000). Similar to KCC in normal human, horse and rainbow trout RBCs, the mechanism was virtually silent at a P O 2 of 1 kPa or lower (Nielsen et al. 1992;Speake et al. 1997;Berenbrink et al. 1997;Gibson et al. 1998). Contrary to KCC in mammalian RBCs (e.g. Speake et al. 1997), the K + transport mechanism at high P O 2 was also strongly activated by alkaline extracellular pH, similar to KCC in Oxygen tension (kPa) [Bumetanide] (μM) Figure 5. O 2 dependence of bumetanide-sensitive and -insensitive K + (Rb + ) influxes K + (Rb + ) influxes (means ± S.E.M., n = 5) were measured as described in Fig. 1, but at pH 8.4 and in the presence of 0, 1, 10 or 100 μM bumetanide, as indicated (A). Two-way analysis of variance with P O 2 and [bumetanide] as factors revealed that there was a highly significant interaction between the two factors (P ≤ 0.001). Thus, the effects of bumetanide differed significantly depending on the P O 2 that was present. * Significantly different from the control value in the absence of bumetanide at the same P O 2 . †Significantly different from control and from the value in the presence of 1 μM bumetanide at the same P O 2 . §Significantly different from the value at P O 2 0 kPa at the same bumetanide concentration. B, 1 μM bumetanide-sensitive K + (Rb + ) influx, calculated as the difference between fluxes in the absence and presence of 1 μM bumetanide. Mean values not labelled by the same letter differed significantly from each other. Note the logarithmic scale for P O 2 . rainbow trout RBCs (Nielsen et al. 1992;Berenbrink et al. 2000). The partial inhibition of K + (Rb + ) fluxes at low P O 2 by the KCC inhibitor DIOA was unexpected, because the high bumetanide sensitivity of the flux strongly suggests that it is carried by NKCC. However, partial inhibition of the NKCC by 100 μm DIOA has previously been noted in human RBCs (18 ± 6% inhibition, mean value ± s.e.m, n = 3; Culliford et al. 2003), calling for a cautionary interpretation of inhibitory effects of DIOA (Berenbrink et al. 2000).
Ultimate proof of both KCC and NKCC requires demonstration that concentration changes of each of the respective cotransported ions influence the transport of the other ions in the predicted way. This has been reported only in a few cases, partly because of the high anion exchange activity and anion conductance of RBCs, which presents a difficulty in varying Cl − independently on both sides of the membrane (Lauf, 1985;Russell, 2000). In the absence of such evidence, we follow previous studies and use Cl − dependence and low bumetanide sensitivity provisionally to ascribe the O 2 -activated K + (Rb + ) flux pathway in crucian carp RBCs to a KCC mechanism.
Differential O 2 sensitivity of KCC and NKCC of cellular O 2 sensing in general and in particular for the altered O 2 dependence of RBC ion transport in certain haemoglobinopathies.