Introduction

Changes in arterial pCO2 induce variations in pulmonary ventilation, with hypercapnia promoting an increase in ventilation. In mammals CO2 levels are primarily sensed at central chemoreceptor neurons and although the primary function of the carotid body is to respond to changes in pO2, it also contributes to CO2 detection [16, 39]. Chemosensitive central neurons have been defined as those found in brainstem areas that are capable of responding to variations in CO2/H+ levels to signal ventilation changes [15]. Examples of those regions are the retrotrapezoid nucleus (RTN), the nucleus of the solitary tract, the locus coeruleus (LC) and caudal raphe nuclei [33]. Changes in CO2/H+ levels are transduced into changes in the firing response of chemoreceptor neurons [23, 39] and a great deal of work is being done to find the primary sensors involved. Depolarization, and therefore increase excitability of central pH-sensitive neurons has been shown to be linked to K+ conductance inhibition and these channels are prime candidates to act as CO2/H+ sensors [31].

At least two types of K+ channels have been proposed to be part of the cellular chemosensing apparatus. They are members of the background K2P K+ channel family [14] or they belong to the inwardly rectifying group of K+ channels known as Kir [19]. Kir5.1 and Kir4.1 are prominently expressed in several brainstem nuclei involved in cardio-respiratory control [48] and coassembled, they are known to form channels inhibited by hypercapnia, an effect mediated by intracellular acidification [49]. The possible role of Kir5.1/Kir4.1 channels in the respiratory response to hypercapnia has been explored through genetic inactivation of Kir5.1 in mice. Chemosensitive neurons of the LC from Kir5.1 KO animals become unresponsive to changes in cytoplasmic pH suggesting that Kir5.1 may be involved in the response to hypercapnic acidosis [13]; however, a study in the same animals shows that Kir5.1 channels are dispensable for functional central and peripheral respiratory chemosensitivity [44]. TASK subfamily K2P channels TASK-1 and TASK-3, and TALK channel TASK-2 have been considered owing to their high sensitivity to extracellular pH. Simultaneous deletion of TASK-1 and TASK-3 in the mouse abolishes the acid-sensitive background K+ current of raphe neurons, but the central ventilatory chemosensory response remains unaffected [32]. The involvement of TASK-1 or TASK-3 channels in the control of ventilation by peripheral chemoreceptors is a matter of controversy, with some authors implicating an important role for TASK-1 in carotid body response to hypoxia and hypercapnia [43], a notion not confirmed by more recent work [37].

TASK-2 [9], a TALK subfamily K2P channel, is highly expressed in kidney proximal tubule cells where it plays a role in HCO3 − reabsorption by virtue of its activation by extracellular alkalinization [47]. TASK-2 has a circumscribed brain expression notably in CO2 chemosensitive RTN, and based on work with a TASK-2 KO mouse it has been proposed to be important in central CO2 and O2 chemosensitivity [17]. Interestingly, in addition to its extracellular pH sensitivity, TASK-2 has been shown to be regulated by intracellular pH thus offering a possible way to couple channel activity to CO2 levels [35].

Fish are also able to respond to hypoxia and hypercapnia by regulating their ventilation rate that in this case corresponds to regulating water flow through the gill openings. The gill is also the main site of CO2 sensing in fish and it has been hypothesized that gill chemoreceptors sensitive to external CO2 rather than protons are capable of initiating a reflex mechanism leading to cardiovascular and respiratory adjustments (reviewed in [38]). Neuroepithelial cells (NECs) have been identified in the zebrafish gill and originally shown to respond to a decrease in O2 level with inhibition of background-type K+ channels and depolarization [24]. The same cells are sensitive to rather low levels of CO2 [40], so that small increases in CO2 inhibit background-type K+ channels of similar characteristics as those sensitive to O2, leading to CO2-dependent NEC depolarization and presumably increased excitability. The response of NECs to CO2 is only partially reduced (<50 %) by carbonic anhydrase inhibition [40]. Although this result suggests an effect of CO2 secondary to its hydration and the predicted secondary acidification, the fact that a substantial CO2 response persists could be interpreted to imply a direct effect. The same issue awaits a definite resolution in mammalian chemoreception, although it is known that the ventilatory response to hypercapnia is greater than that to metabolic acidosis and that central neurons exhibit a greater firing response to hypercapnic acidosis than isocapnic acidosis where the CO2 levels remain constant and bicarbonate concentration and pH are decreased [46]. Also current clamp and intracellular pH recordings in neonatal rat brainstem slices show that the major signal in the chemosensitive response of LC neurons is pHi rather than pHo or CO2 level [15].

If pH-sensitive K2P channels are to take part in CO2 chemoreception either in fish or mammals, it would appear that they will have to be inhibited by intracellular acidification or perhaps directly by increased CO2 levels. The requirement for pHi sensitivity discards TASK-1 and TASK-3 that are solely modulated by pHo [14]. TASK-2, already proposed to have a function in central chemoreception in the mouse [17], on the other hand, would fulfill the pHi-gating requirement [35]. In the present report we first identify a TASK-2 orthologue in zebrafish gill and provide a functional characterization. We then compare the response to CO2 of TASK-2 channels from zebrafish and mouse origins, zfTASK-2 and mTASK-2. Both channels respond very similarly to CO2, and the results suggest that part of the response is secondary to intracellular acidification but some pHi-independent CO2 sensitivity also forms part of the response. A difference in pK 1/2 for extracellular pH gating between zf and mTASK-2 might represent an adaptation to differing blood pH in fish and mammalian species.

Methods

Cloning and site directed mutagenesis

Danio rerio TASK-2 (zfTASK-2) was cloned from zebrafish gill by RT-PCR. Anesthesia by immersion in tricaine methanesulfonate (250 mg/ml) until cessation of opercular movements was followed by decapitation. Animal procedures were approved by the Institutional Animal Care and Use Committee of the Centro de Estudios Científicos. After dissection of the gill, total RNA was isolated with Trizol and 5 μg was reverse-transcribed using the SuperScript II system (Invitrogen) and oligo(dT) primer. Specific primers were designed to obtain the open reading frames (ORF) of two mTASK-2 orthologues accessible in the GenBank: zfTASK-2 (NP_001032478.1): 5′- AAGCTTACCATGGTGGACAAGGGACCTC-3′ and 5′-GGACAACACTGACATTTAA CTCGAG; and zfTASK-2b (NP_956927.1): 5′- AAGCTTGTATTATGGCAGATAAAGGACC-3′ and 5′-GGAAAAGCTTTGGTTGAAGGACTCGAG- 3′. Data in bold are the start and stop codons. The HindIII and XhoI restriction sites (underlined) were incorporated to clone the ORFs in the mammalian expression vector pCR3.1. Mus musculus TASK-2 (K2P5.1, GenBank accession no. AF319542) was that previously cloned from mouse kidney [34].The TASK-2 mutants were generated by the quick-change method using KOD polymerase (Merck). mTASK-2-TASK-3 chimera (244TASK-2-245TASK-3) and the truncated mTASK-2Δ273 had been obtained before [35]. All the constructs were confirmed by sequencing.

Cell culture and transfections

HEK-293 cells were cultured in Dulbecco's modified Eagle's medium and F12 Ham's medium at 1:1 ratio, supplemented with 7 % fetal bovine serum at 37 °C in 5 % CO2. For electrophysiological experiments, transient transfections of zfTASK-2, murine TASK-2 and mutants were done in HEK-293 cells as described previously using CD8 cotransfection to identify effectively transfected cells [8]. The CD8 antigen was revealed with microspheres (Dynabeads) coated with an anti-CD8 antigen.

Electrophysiology

When studying pHo dependence, cells were seeded on 35-mm dishes and continuously superfused with a bathing solution containing 67.5 mM Na2SO4, 4 mM KCl, 1 mM K-gluconate, 2 mM CaCl2, 1 mM MgCl2, 10 mM HEPES/Tris pH 7.5. Osmolality was adjusted with sucrose to a final value of 310 mOsm. Pipette solution contained 132 mM K-gluconate, 8 mM KCl, 1 mM MgCl2, 10 mM EGTA, 1 mM Na3ATP, 0.1 mM GTP, 10 mM HEPES/Tris pH 7.5, 300 mOsm. Standard whole-cell patch clamp recordings were done as described elsewhere [34]. Experiments were performed at room temperature using a holding potential of −70 mV. Currents were measured at several potentials, but data reported are generally those obtained at 0 mV.

In experiments designed to measure the extracellular pH dependence of the currents, bathing solutions contained HEPES (4-(2-hydroxyethyl) piperazine-1-ethanesulfonic acid) for pH 7.0, 7.5 and 8.0; AMPSO (N-(1,1-dimethyl-2-hydroxyethyl)-3-amino-hydroxypropanesulfonic acid) for pH 8.5 and 9.0; CAPS (3-(cyclohexylamino)-1-propanesulfonic acid) for pH 9.5, 10, 10.5 and 11; or MES (2-(N-morpholino)-ethanesulfonic acid) for pH 6.0 and 6.5. The intracellular pH dependence of the currents was measured using the pHi clamp method modifying the transmembrane concentration gradient for acetate as described previously [50]. The composition of the solutions is exactly as previously described [35]. A pulse of 10 mM NH4Cl was used to produce intracellular alkalinization.

To measure CO2 dependence a low buffering power pipette solution contained 132 mM K-gluconate, 8 mM KCl, 1 mM MgCl2, 10 mM EGTA, 1 mM HEPES/Tris, 1 mM Na3ATP, 0.1 mM GTP, pH 7.5. For a highly buffered pipette solution, 100 mM KCl was replaced by 100 mM HEPES-K and pH 7.5 was reached with H2SO4. Additionally, 20 μM EZA (6-ethoxyzolamide, Sigma) was used in some experiments to inhibit intracellular carbonic anhydrases. These solutions were adjusted to pH 7.5. Control bathing solution contained 135 mM Nagluconate, 4 mM KCl, 1 mM K-gluconate, 2 mM CaCl2, 1 mM MgCl2, 30 mM HEPES/Tris pH 7.5. CO2 containing bath solutions were made by replacing 6, 33 or 150 mM Nagluconate with NaHCO3. To obtain a pH of 7.5 they were bubbled with 1 %, 5 % or 25 % CO2/O2 mixtures, respectively. To measure a possible modulation of the currents by O2, an HCO3 −-free bathing solution bubbled with 100 % O2 was used.

Intracellular pH measurements

HEK-293 cells were mounted in an inverted microscope (IX70 Olympus) equipped with a 40× oil immersion objective, a Xenon lamp, an Optoscan Monochromator and a photomultiplier system (Cairn Research). Cells were dialyzed with a pipette solution containing 100 μM of 2′,7′-bis-(2-carboxyethyl)-5-(and −6)-carboxyfluorescein (BCECF) free acid in the whole cell configuration. Excitation wavelengths were alternated between 440 and 490 nm and emission was collected at 535 nm. The demultiplexed signal was acquired with an analog–digital interface (Digidata 1200; Axon Instruments) and monitored using the pCLAMP 10 software (Axon Instruments). Intracellular pH calibration curves (Fig. S1a) were obtained at the end of experiments by the high potassium nigericin calibration method as described previously [50]. When this was not possible, intracellular pH was calculated using a one-point calibration method [4, 6] as follows: cells were exposed to a pH 7.0 calibration solution containing 140 mM K-gluconate and 10 μM nigericin. The ratio 490/440 from the entire experiment was divided by the pH 7.0 490/440 ratio and intracellular pH was calculated using the parameters from the calibration curves.

Simultaneous measurements of currents and pHi

Standard whole cell patch clamp experiments were performed using a pipette solution containing 100 μM BCECF free acid. HEK-293 cells expressing mTASK-2 or zfTASK-2 were exposed to bathing solutions containing 1 %, 5 % or 25 % CO2. K+ currents were monitored simultaneously with intracellular pH. Pipette bathing solutions contained 1 or 100 mM HEPES to obtain a low or a high intracellular buffering power, respectively. Alternatively, 20 μM EZA was used in the pipette solution in order to block intracellular carbonic anhydrases.

Results

Zebrafish TASK-2 orthologues from gill tissue

A database search for zebrafish orthologues of mTASK-2 revealed two target sequences which are probably paralogues. We obtained both cDNAs by RT-PCR amplification of gill total RNA using specific primers. The predicted polypeptide sequences were 513 (zfTASK-2) and 448 (zfTASK-2b) amino acids long. Figure 1 shows an alignment of both zebrafish sequences along with mTASK-2. The overall identity between zebrafish paralogues is 46 %, but it increases to 68.3 % if amino acids 1–240 that encompass the four transmembrane domains (TM) and the two pore segments (P) are considered. Comparison of mTASK-2 (502 amino acids) with zfTASK-2 gave an overall identity of 49.4 % increasing to 73.3 % for amino acids 1–240. The same comparison done with zfTASK-2b gave respective identities of 40.7 % and 64.6 %. In the distal intracellular segment, from amino acid 241 to the end of the sequence, identity decreases to 28.6 % and 18.6 % when comparing mTASK-2 with zfTASK-2 and zfTASK-2b, respectively. Mammalian TASK-2 is gated both by extra- and intracellular pH and arginine 224 and lysine 245 have been identified as respective sensors in these processes [35, 36]. Both sensor amino acids are conserved in zfTASK-2, but in zfTASK-2b the lysine 245 is changed to asparagine.

Fig. 1
figure 1

Alignment of deduced amino acid sequences of zfTASK-2 paralogues and mTASK-2. Identical amino acids are shown in gray. Amino acids forming the extracellular helical cap helices are identified by the upper discontinuous line (−−). The extracellular and intracellular pH sensors R224 and K245 are indicated by arrows. Notice the K245N change in zfTASK-2b. Also highlighted (*) are zfTASK-2 non-conserved amino acids that when mutated to the mouse equivalents partially reversed the alkaline shift observed for the pK 1/2 for pHo-gating of zfTASK-2, and a region (~) that appears essential for the CO2 effect

Full size image

Attempts to express zfTASK-2b in HEK-293 cells failed to induce any currents and the mechanism for this lack of function has not been explored further. Expression of zfTASK-2, on the other hand, was accompanied by the appearance of voltage-independent instantaneous K+ currents accompanied by a further voltage-dependent component (Fig. 2a and b) typical of K2P TASK-2 channels described before. The steady-state IV relations approached open rectification (Fig. 2c). Figure 2d shows the response of zfTASK-2 currents recorded at 0 mV to graded changes in extracellular pH. A large activation was seen upon switching from pH 7.5 to 11.0, which was reversed gradually as the pH was again decreased with current inhibition at pH 6.0. This activation by alkalinization, typical of TALK K2P family members, occurred with a pK 1/2 of 8.82 ± 0.07 and a Hill coefficient of 0.81 ± 0.08 (means ± SEM, n = 5). This last figure did not differ from unity (t-test, P = 0.0764). A series of concurrently done measurements with mouse TASK-2 gave pK 1/2 of 8.03 ± 0.11 (mean ± SEM, n = 6) that does not differ from previously published figures [34, 41]. Despite conservation of mouse pHo-sensing R224 the sensitivity of zfTASK-2 to pHo is alkaline-displaced almost a full pH unit. As argued before, change in the environment of the sensor might strongly influence its pK a and we have mutated some non-conserved amino acids of zfTASK-2 that might affect R224 to the residues present in mTASK-2 [36]. The following mutations of zfTASK-2 were without effect on the pHo dependence: D77N, V78M, K78Q and P220N. Double mutation D214N-E218N and single I104V change shifted zfTASK-2 pK 1/2 to 8.59 ± 0.06 (n = 4) and 8.36 ± 0.03 (n = 3), respectively.

Fig. 2
figure 2

zfTASK-2 potassium channels are activated by extracellular alkalinization. Currents were recorded in HEK-293 cells after transfection with zfTASK-2 cDNA by the whole-cell mode of the patch-clamp technique. a, b Current traces obtained at 5 and 140 mM extracellular [K+], respectively, in a cell containing 140 mM intracellular [K+]. Square voltage pulses delivered from holding voltages of −70 mV (a) and 0 mV (b) ranged from −100 to 80 mV in 20-mV steps. c Current–voltage relationships corresponding to experiments in a and b based on measurements at the end of the pulses. d Time course of zfTASK-2-mediated current recorded at 0 mV with [K+]i/[K+]o 140/5 mM in response to changes in extracellular pH values. e Average extracellular pH dependence curve of zfTASK-2 (mean ± SEM, n = 5). The line is a fit of the Hill equation and was constructed by using the average of fitted parameters of the individual experiments. Parameters were 8.82 ± 0.07 and 0.81 ± 0.08 for pK 1/2 and nH, respectively (means ± SEM, n = 5). Respective values previously measured for mTASK-2 were 8.03 ± 0.11 and 0.85 ± 0.06 [36]

Full size image

We have previously shown that mTASK-2 is also pHi-gated with a pK 1/2 of 8.0 [35]. Using intracellular acetate and BAPTA to buffer Ca2+, it is possible to change pHi predictably by changing extracellular acetate. Figure 3a shows the effect of clamping pHi at values between 7.0 and 9.0 on K+ current of a zfTASK-2-expressing HEK-293 cell. This was achieved by changing [acetate−]o between 1.26 and 126 mM with [acetate−]i of 50 mM. Current mediated by zfTASK-2 was greatly increased by alkalinization from to pHi 7.4 to 9.0 with a graded decrease in the current as pHi was lowered to 7.0, the lowest pH attainable with this set of acetate concentrations. Using [acetate−]i values of 10 and 130 mM, it is possible to extend the range of pHi values achieved down to 6.3 and up to 9.5, respectively. This approach allowed composing Fig. 3b out of 27 experiments. Fit of a Hill equation to the data yielded a pK 1/2 value of 7.9 ± 0.05 and n H of 0.8 ± 0.09 (parameters ± SEM, obtained from the fit); this last value is not different from unity (t-test, P = 0.147). These values were not different from those of the mouse orthologue, pK 1/2 of 8.0 ± 0.07 and n H of 0.9 ± 0.04 [35].

Fig. 3
figure 3

Effect of intracellular pH on zfTASK-2-mediated currents. a Time course of zfTASK-2-mediated current recorded at 0 mV at various indicated pHi values. The cell had 50 mM intracellular acetate and extracellular acetate concentrations were changed from 1.26 to 126 mM to yield pHi values ranging from7.0 to 9.0 [35]. Extracellular pH was 7.4 throughout. b summary of experiments examining the pHi dependence of zfTASK-2 (means ± SEM, n = 9−27). The full range of pH values shown were attained using intracellular solutions with 10 or 130 mM acetate to reach pHi values down to 6.3 and up to 9.5, respectively. To analyze together the three sets of data, each group was normalized to the value of current attained with pHi 7.0. The line shows a fit of a Hill function to the data that gave a pK 1/2 of 7.9 ± 0.05 and n H 0.8 ± 0.09 (parameters ± SEM, obtained from the fit)

Full size image

CO2 sensitivity of mouse and zebrafish TASK-2

TASK-2 channels have been proposed to be important in mediating central O2/CO2 chemosensitivity in mice [17] while a background K+ channel, conceivably zfTASK-2 cloned here, is involved in CO2 sensing in NEC cells of zebrafish gills [40]. CO2 exerts an effect on mTASK-2 that has been attributed to an intracellular acidification secondary to its permeation into the cell and subsequent carbonic anhydrase-catalyzed hydration [35]. A direct effect of CO2 on the channel cannot, however, be discarded. We have examined the effect of increasing levels of CO2 on mouse and zfTASK-2 (Fig. 4a and b), both of which responded with graded inhibition that was reversed as the solution was returned to nominally CO2-free solution. These experiments were carried out using CO2/O2 gas mixtures, but O2 by itself was found to have no effect on TASK-2 (Fig. S2a). Changes in cell volume regulate the activity of TASK-2 [34], but no such changes occur during CO2 challenge of TASK-2-expressing cells (Fig. S2b).

Fig. 4
figure 4

TASK-2 channels are inhibited by CO2 in a dose-dependent manner. Time course of currents recorded at 0 mV at the indicated CO2 concentrations in a mTASK-2-expressing HEK-293 cells (n = 6) and in b zfTASK-2-expressing HEK-293 cells (n = 5). The extracellular pH was maintained at 7.5 in 1 %, 5 % and 25 % CO2 saturated solutions with bicarbonate concentrations of 6, 33 and 150 mM HCO3 −, respectively. The results are means ± SEM In all experiments the intracellular solution was weakly buffered with 1 mM HEPES

Full size image

Separate experiments using intracellular BCECF show that exposure of HEK-293 cells to 1 %, 5 % and 25 % CO2 decreased pHi from 7.37 ± 0.12 to 6.95 ± 0.05, 6.43 ± 0.05 and 6.10 ± 0.06, respectively (means ± SEM; see also Fig. S1c). The initial pHi found experimentally lies below the pK 1/2 value for intracellular gating of the TASK-2 channels, nevertheless around 25 % of the maximal activity is predicted, leaving room for further inhibition by acidification. It could therefore be surmised that the CO2 inhibition observed in mouse and zfTASK-2 is the exclusive consequence of intracellular acidification.

To approach the question of whether CO2 effect is due solely to changes in pHi we have used strongly buffered intracellular solutions. Figure 5a compares the effect of extracellular superfusing the weak base NH4Cl on cells that contain 1 mM intracellular HEPES or strongly buffered cells containing 100 mM HEPES. Influx of the highly permeable NH3 ought to lead to an increase in pHi when intracellular NH4 + is formed again by NH3 capture of protons [42]. An immediate increase in K+ current following a NH4Cl pulse was observed in cells containing 1 mM intracellular HEPES, but the response was greatly diminished in cells containing 100 mM buffer. This suggests that the NH4Cl pulse was able to alkalinize cells and that strong buffering with 100 mM intracellular HEPES largely attenuated the change in pHi. Figure 5b shows a similar experiment but using CO2 to acidify intracellularly. Although the response to the predicted decrease in pHi is reduced when strongly buffered pipette solution was used, we observed a sizable effect that could be attributed to a direct effect of CO2 if no change in pHi were taking place. Neutralization of lysine 245 of mTASK-2 abolishes pHi sensitivity [35]. A chimera where C terminus amino acids starting at residue 244 of TASK-2 are replaced by the C-terminal residues 245–365 of the pHi-insensitive TASK-3 channel [244TASK-2-245TASK-3, 35] was unresponsive to CO2. A construct we have termed TASK-2Δ273, which lacks C-terminal amino acids from residue 273 down and shows unaltered response to changes in pHi [35], responded to CO2 as the non-mutated TASK-2 (Fig. 5c). The stretch from aa 245 to 273 therefore appears essential for CO2 sensing in TASK-2. Silencing of the pHi sensor itself as in mutant TASK-2-K245A only decreased the effect of CO2 to the level observed with strong pHi buffering (Fig. 5b).

Fig. 5
figure 5

Effect of increased buffering capacity on TASK-2 response to ammonium and CO2 pulses. a Time course of currents recorded at 0 mV in mTASK-2-expressing HEK-293 cells. During the time indicated, the solution bathing the cells was switched to one containing 10 mM NH4Cl to elicit intracellular alkalinization. The activation of the channel activity observed when the pipette solution was buffered with 1 mM HEPES (circles, n = 8) was strongly inhibited by increasing HEPES concentration in the pipette solution 100 mM (triangles, n = 7). b An analogous experiment but using a bath solution saturated with 5 % CO2 and containing 33 mM HCO3 −. The inhibition observed when the pipette solution contained 1 mM HEPES (circles, n = 8) was partially impeded when the intracellular buffer capacity was increased using 100 mM HEPES (triangles, n = 8). Squares show the results obtained with 1 mM intracellular HEPES in cells expressing the mTASK-2-K245A pHi-insensitive mutant (n = 5). Extracellular pH was 7.5 throughout and the results are means ± SEM. c Comparison of the effect of 5 % CO2 on currents mediated by mTASK-2 (circles, n = 8), a mutant truncated in its C terminus end mTASK-2-∆273 (squares, n = 3) and 244TASK-2-245TASK-3 chimera, that has the C terminus of pHi-insensitive TASK-3 (triangles, n = 3)

Full size image

To explore further the possibility of an effect of CO2 that is separate from its predicted intracellular acidification response, we measured TASK-2-mediated current and pHi simultaneously in weakly or strongly buffered cells, we additionally inhibited carbonic anhydrase and utilized a TASK-2 mutant with pHi and pHo sensors disabled by mutation.

Measurements of K+ currents and pHi are summarized in Fig. 6a and b. In Fig. 6a, a graded increase in CO2 is shown to produce a marked inhibition in current in a weakly buffered cell expressing mTASK-2 which coincides with a marked intracellular acidification (Fig. 6b). To attempt to impede the drop in pHi, cells were strongly buffered with 100 mM HEPES and were additionally dialyzed with 20 μM ethoxyzolamide (EZA) to inhibit carbonic anhydrase. Under these conditions the CO2-induced acidification was abolished, yet a sizable inhibition of current still took place (Fig. 6a and b). In Fig. 6c, these data are replotted, together with an experiment carried out in zfTASK-2-expressing cells, to emphasize the fact that under strong pHi buffering and additional carbonic anhydrase inhibition there was an up to 50 % inhibition of K+ current by CO2 without any measurable change in pHi.

Fig. 6
figure 6

Simultaneous measurement of ion current and intracellular pH reveals pHi-independent inhibition of mTASK-2 by CO2. a mTASK-2 current inhibition by increasing CO2 measured with low intracellular buffer capacity (circles, n = 5) or with strong buffering with 100 mM HEPES and additional inhibition of carbonic anhydrase with 20 μM intracellular EZA (triangles, n = 3). b Simultaneously measured intracellular pH using fluorescent pH indicator BCECF. c Replottting the data in a and b as current versus pHi to show that using 100 mM HEPES and 20 μM a EZA sizable change in current occurred without concomitant change in pHi. Data for zfTASK-2 obtained using an intracellular solution containing 100 mM HEPES and 20 μM EZA are also shown (n = 7). Results are means ± SEM

Full size image

To ascertain whether the pHi-independent effect exerted by CO2 on TASK-2 activity requires the pH-sensing machinery of the channel, we used a mutant, mTASK-2-R224A-K245A, in which both extracellular and intracellular pH sensors are disabled. In Fig. 7, the effect of increasing CO2 from nominally zero to 1 %, 5 % and 25 % is seen to produce an effect on mTASK-2-R224A-K245A-mediated current that can be superposed on that elicited by CO2 under strongly buffered intracellular pH in wild-type mTASK-2-expressing cells.

Fig. 7
figure 7

Effect of CO2 on pHo- and pHi-insensitive mTASK-2-R224A-K245A mutant. Data (triangles) for the mutant mTASK-2 were collected using 1 mM intracellular HEPES and are means ± SEM (n = 6). Data for wild-type mTASK-2 with 1 or 100 mM intracellular HEPES are also shown for comparison

Full size image

Discussion

We have identified a zebrafish (D. rerio) orthologue of K2P TASK-2 K+ channel isolated from gill tissue. zfTASK-2 is 49 % identical to its mouse equivalent and, like mTASK-2, is gated both by extra- and intracellular pH. NECs in zebrafish, and presumably other fish, are capable of sensing O2 and CO2 levels, and do so through the modulation of a conductance exhibiting the characteristics of K2P background K+ channels [1, 24, 30]. TASK-1, TASK-3 and TASK-2 are K2P channels that have been considered as candidates to play roles in CO2 sensing in central and peripheral chemosensory neurons in mammals [17, 32, 37, 43]. Of these, only TASK-2 is sensitive to intracellular pH, being inhibited by acidification and further activated by alkalinization with a pK 1/2 of 8.0 [35], making it a plausible candidate to sense CO2 levels through its hydration and consequent intracellular acidification. Mammalian K2P channels sharing the property of being regulated by intracellular pH, and also present in zebrafish [18], are TREK-1 and TREK-2, but they are activated by acidification [27, 29], and TRAAK that is not inhibited by acidification [25]. TWIK-1and −2 are inhibited by 100 % CO2 but there is no evidence for a direct effect of pHi [7, 26]. This leaves TASK-2, and eventually its relatives TALK-1 and −2, as sole K2P candidates able to sense a CO2-dependent acidification with inhibition leading to depolarization and increased excitability.

Functional expression of zfTASK-2 yields currents with similar time- and voltage-independence as those observed for its mouse orthologue. A voltage-dependent component of otherwise instantaneous currents might be attributed to voltage-related changes in occupancy of the selectivity filter affecting pore opening. Indeed, the pK 1/2 of extracellular pH-gating has been seen to be dependent on K+ concentration and voltage in mTASK-2 [9]. Zebrafish TASK-2, like mTASK-2, is inhibited by intra- and extracellular acidification and activated by intra- and extracellular alkalinization. zfTASK-2 pHi dependence, with a pK 1/2 of 7.9, is quite similar to that of mTASK-2 [35]. There are no data on intracellular pH in adult zebrafish, but measurements with a fluorescent indicator in early embryos give a figure of 7.55 [30]. This is 0.3–0.4 pH units more alkaline than pH measurements in muscle tissue of adult fish of various species (cited in [30]). It must be said that these last measurements lacked cellular resolution.

In contrast to the similarity in pHi-gating between zebrafish and mouse TASK-2, the dependence of zfTASK-2-currents on pHo occurred with a pK 1/2 of 8.82, which compares with a value of 8.0 observed for mTASK-2 [36]. This large alkaline displacement in pHo dependence in zfTASK-2 correlates well with the relatively alkaline plasma pH observed in fish compared with that of mammals. Again, in zebrafish embryos, an "interstitial" pH value of 8.08 has been reported [30], while examples of measurements of adult fish aortic plasma pH give values of 7.91 in carp [10] and 8.08 in rainbow trout [21]. We have explored the possible reason for the alkaline displacement in zfTASK-2 pHo dependence and have some inkling that differences in amino acids near the pHo sensor might be responsible. Regardless of the mechanism involved, the change in pK 1/2 could be an adaptation to the markedly more alkaline extracellular pH in fish compared to mammals.

The question of whether a putative CO2 K+ channel sensor fulfills its signaling function through a change in activity in direct response to a change in CO2 levels, or does so secondarily through intracellular acidification, remains open. Given the possibility that mTASK-2 and zfTASK-2 play such a role in mammalian chemoreceptor neurons and gill NECs, respectively, we have tried to discern if CO2 sensitivity is direct or a function of changes in pHi.

Increasing CO2 led to a rapid, concentration-dependent inhibition of mouse and zfTASK-2. Perhaps surprisingly the response of zfTASK-2 to graded increases in CO2 level was not much different from that of mTASK-2, considering that air breathers have arterial blood levels of CO2 that are around one order of magnitude higher than fish [38]. This would appear to contradict a role of zfTASK-2 in NEC CO2 sensing, but the participation of a different component complementing the channel to form a molecular sensing unit cannot be discarded. In fact, carbonic anhydrases have often been proposed to interact functionally with membrane transport proteins in so called metabolons, as first proposed for the association of anion exchanger and carbonic anhydrase in red cells [45] and recently in zebrafish gill cells involved in Na+ transport [22].Footnote 1 Establishing whether such interaction between carbonic anhydrase and zfTASK-2 might be important in CO2 sensing will require experimentation with native NECs.

The effect of CO2 on mTASK-2 and zfTASK-2 appear indistinguishable. Additional dissection of the effect allows resolving two components to CO2 inhibition: one related to channel inhibition by intracellular acidification and a second that takes place in the absence of acidification and must pertain to a more direct type of CO2 effect. This view is based on three observations. (1) Use of strongly buffered intracellular solutions that abolish any change in pHi does not preclude CO2-dependent inhibition of TASK-2-mediated currents. (2) Neutralization of lysine 245 of mTASK-2, that abolishes pHi-sensitivity of the channel [35], did not abolish the effect of CO2 but just decreased to the level observed with strong pHi-buffering. (3) Use of a carbonic anhydrase inhibitor did not abolish the CO2 effect on TASK-2 activity but reduced it to the levels observed when changes in pHi were impeded by using highly buffered intracellular solutions.

The mechanism by which TASK-2 of either murine or zebrafish origin sense CO2 will have to be the subject of future investigation. The effect of CO2 might be a direct action on the channel. The other possibility is the participation of a different sensing partner that couples functionally to the channel as has been proposed for K+ channels involved in O2 sensing [28]. A direct effect of CO2 has been proposed to regulate an unidentified K+ inwardly rectifying channel of Hela cells, but this is activated by CO2 [20]. Enzymes using CO2 as substrate must have binding sites allowing CO2–protein interaction. An example of these is phosphoenolpyruvate carboxykinase that catalyzes the carboxylation of phosphoenolpyruvate to oxaloacetate and whose crystal structure has revealed CO2 hydrogen bonding with basic amino acid side chains [11]. Indeed, examination of a range of proteins has established a so-called CO2 formatics that suggests how proteins bind CO2 and in which basic amino acids feature prominently [12]. The apparently direct effect of CO2 on TASK-2 described here requires the presence of a stretch of amino acids in the C terminus going from residue 245 to 273. Whether this portion of the protein is necessary to CO2 sensing by TASK-2 or for interaction with a sensing partner is not known at present. Interestingly this stretch of C terminus sequence is rich in basic amino acids that have been proposed to be important determinants of TASK-2 G protein subunit binding and regulation of the channel [3]. That CO2 might interact with that process easily comes to mind, particularly since in preliminary experiments we have been unable to see an effect of CO2 added to the intracellular aspect of inside-out membrane patches (Fig. S2c). Independently of the mechanism involved, however, our data indicate that CO2 can inhibit TASK-2 background K+ channels by a possibly direct effect not connected to intracellular acidification. This observation, in addition to representing a novel type of regulation for K+ channels, might be of importance for the eventual roles played by TASK-2 in chemoreception in mouse and zebrafish.