Introduction

Carbon monoxide (CO) is a potent activator of large conductance, calcium-dependent potassium channels (alternately known as BKCa, maxi-K, or slo1) when heterologously expressed in human embryonic kidney 293 (HEK 293) cells [9, 28] or when natively expressed in vascular myocytes [8, 26, 27] and carotid body glomus cells [19]. Cellular CO production emanates primarily from the enzymatic catalysis of heme by two of the three known catalytically active isoforms of hemeoxygenases [13, 15, 23]. This process has the strict co-requirement for nicotinamide adenine dinucleotide phosphate (NADPH) and molecular oxygen (O2). The dependence on O2 of hemeoxygenase suggested this enzyme might act as a cellular O2 sensor and may provide a link between hypoxia and BKCa channel closure during periods of O2 deprivation. Such a notion was strengthened by the observation that one isoform of hemoxygenase (hemeoxygenase-2; HO-2) is a protein partner closely associated with the BKCa channel complex [28]. Furthermore, in excised, inside–out membrane patches of both recombinant and native O2-sensitive systems, co-application of NADPH, heme, and O2 evoked channel activation that was rapidly reversed upon removal of O2 [28].

Although much information is now available about how BKCa channels may sense acute changes in O2, the mechanism(s) by which CO can modulate BKCa channel activity is less clear, with proposals falling neatly into two camps. In the first is the proposal that CO activates BKCa channels via histidine-dependent binding directly to the channel protein complex—the direct binding hypothesis [27]. In the second is the idea that CO activates the channel via an intermediate heme binding step—the indirect binding hypothesis [9]. These two hypotheses are supported by data from studies in both native tissues and recombinant expression systems. Thus, the direct binding hypothesis was first posited by Wang et al. [27] to explain their earlier observation of endothelium-independent relaxation of systemic arteries by CO [24]. Further probing of the mechanism by employing chemical modification of the channel protein using diethyl pyrocarbonate suggested that CO interacted directly with a specific extracellular histidine residue [26]. The indirect binding hypothesis was developed by Jaggar et al. [8] also working in arterial smooth muscle. In this model, CO evokes an increase in Ca2+-sensitivity via an indirect binding step involving cellular heme [9]. This suggestion is very attractive because it unifies the two observations that CO binds avidly to heme and that heme binds directly to BKCa channel at an intracellular heme-binding pocket (containing a conserved histidine) located in the C-terminal domain of the α-subunit [7, 22].

As regulation by CO of BKCa channels is emerging as a widespread and physiologically important phenomenon that is intimately involved in the control of smooth muscle contractility (both systemic and pulmonary) and excitability of neurosecretory and neuronal cell populations, it is crucial that the mechanism of activation is thoroughly investigated in a system where both the molecular identity of the BKCa channel and the subunit composition is certain. To this end, we present herein a systematic study of the mode of activation of the α-subunit of human BKCa channel recorded from inside–out patches of HEK 293 cells expressing KCNMA1. The role of heme in the process of acute channel activation by CO was tested using redox manipulations of wild-type BKCa channels. The possibility that direct interaction of CO with specific histidine residues in either the C-terminal or channel core might account for channel activation was investigated using single and double histidine mutations. Furthermore, engineering a chimeric channel (containing hslo1 core and S9–S10 module of the C terminal of mslo3) allowed investigation of the structural basis of CO activation. Our new data suggest that neither the direct nor the indirect binding hypotheses can completely explain CO activation of the human BKCa α-subunit. They also show that a motif in the C-terminal tail is essential for CO activation and present the possibility that CO alters channel gating independently of its potential interactions with redox-sensitive proteins or compounds that are associated with the channel complex.

Materials and methods

Cell culture

HEK 293 cells that stably express human BKCa α-subunit alone (KCNMA1—GenBank accession number NM_002247) were a kind gift of Professor M. J. Ashford (University of Dundee, Tayside, UK). Cells were maintained in Earle’s minimal essential medium (containing l-glutamine) supplemented with 10% fetal calf serum, 100 U/ml penicillin G, 100 μg/ml streptomycin sulfate, 0.25 μg/ml amphotericin B, 1% nonessential amino acids, 50 mg/ml gentamycin, and 1 mg/ml G418 sulfate (Gibco BRL, Paisley, Strathclyde, UK) in a humidified incubator gassed with 5% CO2/95% air. Cells were passaged every 5–7 days in a ratio of 1:25 using Ca2+- and Mg2+-free phosphate-buffered saline. Wild-type HEK 293 cells express neither BKCa channel subunit messenger RNA (mRNA; not shown, but see [10]) nor functional BKCa channels (not shown, but see [12]).

Site-directed mutagenesis

The Quikchange site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) was used to replace the histidine residue within the heme-binding domain (H616R) and a histidine residue near the pore domain (H254R) with arginine (numbering from the BKα zero construct accession number NM_002247). A double histidine mutant construct was also generated by performing sequential reactions with the same primer sets. The primer sets used were H616Rfor 5′-CTGTGATGTCATCACGACAGGCCTTGCAG-3′, H616Rrev 5′-CTGCAAGGCCTGTCGTGATGACATCACAG-3′ and H254Rfor 5′-CAGCCGGGTTCATCCGTTTGGTGGAGAATTC-3′, H254Rrev 5′-GAATTCTCCACCAAACGGATGAACCCGGCTG-3′. The primers were designed using PrimerX software for site-directed mutagenesis (http://bioinformatics.org/primerx/cgi-bin/DNA_1.cgi) and purified by polyacrylamide gel electrophoresis before use. The following sample reactions were set up in a final volume of 50 μl; 50 ng of template DNA, 125 ng of each primer, 1 μl of deoxynucleotide triphosphate mix, 5 μl of 10× reaction buffer, and 2.5 Units of PfuTurbo DNA polymerase. Tubes were transferred to a thermal cycler and subjected to the following cycling parameters: 1 cycle at 95°C for 30 s followed by 18 cycles of 95°C for 30 s, 55°C for 1 min, and 68°C for 7 min. Reactions were chilled on ice for 2 min before the addition of ten units of DpnI restriction enzyme, followed by incubation for 1 h at 37°C. DpnI-treated DNA was used to transform XL-1 blue competent cells. Single colonies were selected, and plasmid DNA was isolated for sequencing to verify that individual histidine residues were correctly replaced by arginine.

Generation of the chimeric channel, Slo1C/Slo3T(S9–S10)

mSlo3 [21] was cloned from mouse testis mRNA. The Slo1C/Slo3T(S9–S10) construct consisted of the N-terminal region of hSlo1 up to lysine 684 and the S9–S10 module of the C-terminal tail of mSlo3 starting at methionine 687 [17]. The DNA construct was generated by standard overlap polymerase chain reaction (PCR) techniques using the following primers: hSlo1Nt Forward, 5′-ATATGCGGCCGCGATCCCAAGATGGATGCGCTC-3′; hSlo1Nt Reverse, 5′-CTTCACATTGGAGTCCATGTTG-3′; mSlo3Ct Forward, 5′-ATGGACTCCAATGTGAAGATGCTGGACAGCAGTGGCAT-3′; mSlo3Ct Reverse, 5′-AAGGTACCTTGCCAACTCGCTGACCACTAAG-3′.

The chimeric PCR product was then cloned into pcDNA3.1(−) for transient expression in HEK 293 cells.

Using an Amaxa Nucleofector (Amaxa, Cologne, Germany) according to the manufacturer’s instructions, pcDNA3.1 containing the channel constructs was transiently transfected into HEK 293 cells 1–3 days before patch-clamp experiments.

Patch-clamp recording

Cells were grown for 1–3 days on glass coverslips before being transferred to a continuously perfused (5 ml min−1) recording chamber (volume approximately 200 μl) mounted on the stage of an inverted microscope (Olympus CK41, Olympus UK, London, UK) equipped with phase-contrast optics. For inside–out, excised patch clamp recordings, the standard K+-rich bath solution was composed of (in mM): 10 NaCl, 117 KCl, 2 MgCl2, 11 N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES; pH 7.2) with [Ca2+] adjusted to 10, 100, and 336 nM and 1 and 10 μM using ethylene glycol tetraacetic acid (EGTA) and CaCl2 in appropriate ratios. The 100-μM solution contained no EGTA. As free [Ca2+] of more than 1 μM is effectively out of the buffering range of EGTA, the free [Ca2+] of the 10- and 100-μM solutions was measured (and adjusted if necessary) using a calibrated Ca2+ electrode from Sentek (Braintree, Essex, UK). The voltage output of this ion-selective electrode is linear from 5 × 10−6 to 100 M and is, therefore, ideally suited to measuring solutions that contain 10 μM Ca2+ and above. The K+-rich pipette solution was composed of (in mM): 10 NaCl, 117 KCl, 2 MgCl2, 2.0 CaCl2, 11 HEPES (pH 7.2).

Normoxic, hypoxic, and hyperoxic solutions were bubbled for at least 30 min with medical air, 100% N2(g) and 100% O2(g), respectively. This produced no shift in pH. Bath pO2 and temperature were continuously monitored using a polarized carbon fiber electrode [16] and microthermister, respectively. Although bubbling of solutions with gases resulted in a reduction in temperature at the holding chamber, there were no significant differences in bath temperature, as solutions were switched, for example, from normoxia to hypoxia (maximum shift ∼0.25°C). CO was delivered by bath perfusion of the CO donor molecule, tricarbonyldichlororuthenium (II) dimer ([Ru(CO3)Cl2]2—Sigma-Aldrich) as previously described [28]. RuCl2(DMSO)4 was used as a control for [Ru(CO3)Cl2]2 and was synthesized according to the method of Evans et al. [4] and as recently described in detail [28].

All K+ currents were recorded at a bath temperature of 22 ± 0.5°C. Current recordings were made using an Axopatch 200A amplifier and Digidata 1320 A/D interface (Axon Instruments, Forster City, CA, USA). To evoke K+ currents in inside–out membrane patches, one of four protocols was employed (all voltages are reported with respect to the inner membrane leaflet, i.e., −Vp): (1) activating step, 0 mV holding potential, voltage steps from holding to between +30 and +100 mV in 10 mV increments (250 ms, 0.1 Hz); (2) deactivating step, 0 mV holding potential, voltage step from holding to 50 mV (50 ms) followed by steps in the range −30 to −130 mV in 10 mV increments (250 ms, 0.1 Hz); (3) continuous gap-free recording, holding potentials of between −60 to +60 mV for up to 20 min at each voltage; and (4) tail currents, 0 mV holding potential, activating voltage step from holding to +100 mV for 100 ms, stepped back to between −50 and +110 mM for 250 ms.

Data handling

Data analyses were performed using the PClamp 9 suite of software. The product of channel number and open state probability (NPo) at any voltage was calculated from 30-s sections of current recording using equation NPo = I/i, where I is the mean patch current and i is the single channel amplitude at that voltage (determined from all-points histograms): All data are reported as mean±SEM. Statistical comparisons were made using paired Student’s t-test or analysis of variance with Tukey post hoc test. Activation and deactivation parameters were determined from iterative fitting using a standard exponential routine (ClampFit 9). G/G max curves were generated from the steady-state tail currents. The CO concentration–response curve and G/G max curves were generated from each complete data set by iterative fitting to the Hill and Boltzmann equations, respectively.

Results

Channel activation by CO

We have previously shown that, when KCNMA1 is co-expressed with KCNMB1 (β1-subunit of BKCa), CO is a potent channel activator [28]. BKCa channels in patches excised from cells expressing only the α1-subunit were activated in an analogous fashion (Fig. 1). This effect could be attributed specifically to activation of BKCa α-subunits by CO, as treatment with 30 μM of the donor product [RuCl2(DMSO)4], which is stable and does not release CO into solution, was completely without effect (Fig. 1a,d [inset]). Activation by CO evoked a significant leftward shift in the voltage at which G/G max was 0.5 (V0 × 5) of around −20 mV (control mean V0 × 5 = 78.5 ± 4.3 mV; CO donor mean V0 × 5 = 60.8 ± 3.8 mV, n = 4, P < 0.05—Fig. 1b). Exposing patches to increasing concentrations of the CO donor evoked a concentration-dependent increase in patch activity, which was fully reversible upon wash-out (Fig. 1c). The mean EC50 was 37.9 ± 4.3 μM with a mean Hill coefficient of 3.04 ± 1.12 (35 observations on 7 separate patches—Fig. 1d).

Fig. 1
figure 1

Activation of BKCa α1-subunits by carbon monoxide. a Exemplar continuous gap-free recording of channel activity during perfusion with normal K+-rich intracellular solution (Control) and solutions pre-equilibrated with 30 μM of either the donor product or the CO donor, as shown in the bars above the top recording. The top trace shows a 10-min recording at −Vp = +40 mV ([Ca2+]i = 336 nM). Below the main trace are shown four 400-ms sections of the same recording but on a faster time-base. Each section is expanded from the points indicated by the arrows above the main trace and the conditions prevailing are shown to the right of each expanded section. b. G/G max relationships for single channel currents recorded in macro-patches in the absence (open symbols) and presence (closed symbols) of 30-μM CO donor. Data were obtained using the final 10 ms of the steady-state tail currents. Curves were fitted to the Boltzmann equation by a nonlinear iterative fitting routine. c Typical continuous gap-free recording of channel activity during perfusion with normal K+-rich intracellular solution pre-equilibrated with concentrations of the CO donor shown above the trace. The initial 90 s and final 7 min represent perfusion with normal K+-rich intracellular solution (Control and Wash, respectively). Recording was performed at a holding potential −Vp = +40 mV with [Ca2+]i buffered to 336 nM. Below the main trace are sections of 1 s duration expanded from the control and wash regions. d Normalized concentration–response data for CO activation of the channel that were derived from seven continuous traces similar to that exemplified in (b). The inset shows the mean effects of addition of 30 μM of either the product control or the CO donor itself, as indicted under each bar

Figure 2a shows fast time-base, paired current recordings at a range of holding potentials and in the presence and absence of 30 μM of the CO donor. In symmetrical 140 mM K+ but with quasi-physiological Ca2+ concentrations ([Ca2+]o = 2.0 mM, [Ca2+]i = 336 nM), CO evoked a dramatic increase in patch activity (Fig. 2a) that was not reflected in a significant alteration in channel conductance (Fig. 2b, n = 14). Thus, in control solution, the mean single channel conductance of BKCa α1-subunits was 187.0 ± 4.8 pS, and this remained unaltered at 173.9 ± 6.9 pS in the presence of 30 μM CO donor (Fig. 2b inset, n = 14). However, CO treatment did augment the mean NPo in a voltage-dependent manner, with statistically significant (P < 0.025) increases at activating voltages between −10 and +40 mV, inclusive (Fig. 2c). Taken together, these observations extend previous studies on native BKCa channels and show that: (a) β-subunits are not required for activation by CO of BKCa channels; (b) CO is able to activate BKCa channels at much lower concentrations than have been previously tested; and (c) channel activation occurs independently of changes in its single channel conductance.

Fig. 2
figure 2

Effect of CO on single channel conductance and NPo. a Exemplar, paired, fast time-base recordings taken from the same trace of BKCa at the voltages shown to the right. Traces on the left were recorded in the absence of the CO donor (Product Control). Traces on the right were recorded after 90 s of perfusion with 30 μM CO donor. Symmetrical K+-rich solutions, [Ca2+]i = 336 mM. b Mean single channel current–voltage relationships derived from data sets exemplified in (a). Data are from channels perfused with normal control solution (Control—open symbols) and after application of 30 μM CO donor (closed symbols). The inset shows mean, calculated single channel conductances under the two conditions. c Mean NPo versus voltage relationships derived from data sets exemplified in (a). Data are from channels perfused in normal control solution (Control—open symbols) and after application if 30 μM CO donor (closed symbols). Asterisk indicates the voltages at which statistically significant differences between Control and CO conditions were observed, P < 0.025

Kinetics of BKCa activation by CO

Figure 3 shows the effects of CO on channel activation and deactivation kinetics. Similar to previous studies using BKCa channels composed of α11-subunits [12], all activating and deactivating currents evoked in BKCa α1-subunit channels were most simply described by a single exponential function. Application of 30 μM of the CO donor increased the rate of channel activation at all test potentials (Fig. 3a and c) in addition to increasing the steady-state current amplitude. Proportionally, this increased rate of activation was most pronounced at lower activating voltages. In contrast, deactivation was unaffected by CO at all test potentials (Fig. 3b and d).

Fig. 3
figure 3

Effect of CO on channel activation and deactivation kinetics. a Typical effect of treatment with CO donor on the activation of BKCa α1-subunit channels. The traces show excised macro-patch activity evoked from a holding potential of −40 mV to an activating potential of +60 mV in the absence and presence of 30 μM CO donor; [Ca2+]i = 336 nM. b Typical effect of treatment with CO donor on deactivation of BKCa α1-subunit channels. Channel activity was evoked from an activating potential of +60 mV to a deactivating potential of −120 mV in the absence and presence of 30 μM CO donor. [Ca2+]i = 336 nM. c Mean time constant versus voltage relationships of channel activation in the absence (open symbols) and presence (closed symbols) of CO donor. Data were derived from traces similar to those exemplified in (a). d Mean time constant versus voltage relationships of channel deactivation in the absence (open symbols) and presence (closed symbols) of CO donor. Data were derived from traces similar to those exemplified in (b)

Calcium and voltage dependence of the CO activation

At the low [Ca2+]i of 10 nM (not shown) and 100 nM (Fig. 4a), CO was able to increase patch activity in a voltage-dependent manner. At the moderate [Ca2+]i concentrations of 336 nM (see Figs. 1 and 2) and 1 μM (Fig. 4b), CO was a potent activator at all voltages. As [Ca2+]i was increased to 10 μM (Fig. 4c) and 100 μM (not shown), the ability of CO to activate channels waned, even at low voltages. Thus, it appears that the CO activation might be a direct consequence of increasing the Ca2+-sensitivity of the channel. However, such an effect is clearly much more subtle than merely shifting the Ca2+ concentration–response curve to the left. Such a supposition is supported by the data shown in Fig. 5. At all activating voltages (from +20 to +100 mV), increasing [Ca2+]i from 10 to 100 μM did not result in any further increase in patch activity (Fig. 5 a–e, open symbols). In other words, the channel was saturated and could not be activated further simply by increasing the [Ca2+]i. However, treating the patches with CO at sub-maximal [Ca2+]i evoked an increase in channel activity to levels which could not be achieved even by addition of a 100-fold increase in [Ca2+]i—i.e., from 100 nM to 10 μM (Fig. 5a–c). This ability to “super-stimulate” diminished as a function of voltage, and CO was without significant effect at voltages exceeding +100 mV. Thus, although a proportion of the CO activation could be attributed to an increase in Ca2+ sensitivity, a significant component was Ca2+ independent.

Fig. 4
figure 4

Ca2+ and voltage dependence of BKCa α1-subunit activation by CO. Exemplar continuous current recordings from patches excised into K+-rich intracellular solution in which [Ca2+]i had been buffered to 100 nM (a), 1 μM (b), and 10 μM (c). The first half of the recordings was made in the absence of CO, whereas the second half was made in the presence of 30 μM CO donor, as indicated in the box above the trace. The membrane voltage was stepped from 0 to +100 mV in 20 mM increments every 90 s, as indicated at the bottom of (c)

Fig. 5
figure 5

Ca2+ dependence of CO activation at different voltages. Mean patch currents versus [Ca2+]i relationships in the absence (open symbols) and presence (closed symbols) of 30 μM CO donor when voltage was held at +20 mV (a), +40 mV (b), +60 mV (c), +80 mV (d), and +100 mV (e). All recordings were carried out in the presence of K+-rich pipette and bath solutions. Date were derived from continuous, gap-free current recordings

Site of CO binding and mechanism of action

The indirect binding hypothesis was tested using redox and structural modifications of the α1-subunit stably and transiently expressed. In confirmation of earlier studies [6, 22], the human BKCa α1-subunit was inhibited by low concentrations of heme. Thus, application of oxidized heme (hemin) significantly reduced mean NPo from 0.35 ± 0.11 to 0.06 ± 0.02 (n = 14, P < 0.02 – Fig. 6a,b). A previous report of the irreversibility of heme inhibition [22] was confirmed in a separate series of experiments (data not shown). Concurrent CO treatment overcame the heme-dependent inhibition and increased mean NPo to 0.84 ± 0.35 (n = 14, P < 0.05). As hemin does not bind CO, it seemed likely that CO was acting at a locus not previously identified. However, to test the hypothesis that CO might be interacting with pre-bound heme, we assayed the ability of CO to activate channels after reduction and oxidation of the channel and its local environment. Modest reducing conditions (lowering the partial pressure of oxygen—pO2—from 150 to <10 mmHg) caused a reduction in mean NPo from 0.40 ± 0.14 to 0.03 ± 0.01, as previously reported [12, 28] and this environment supported robust CO activation (to 1.95 ± 0.8, n = 6—Fig. 6c,d). Hyperoxia (raising bath pO2 to >600 mmHg) did not significantly alter mean NPo (0.21 ± 0.17 versus 0.07 ± 0.04 Fig. 6e,f). Importantly, and consistent with other data herein, the oxidizing condition of relative hyperoxia was not able to dampen significantly the activation evoked by concurrent CO application (mean NPo from 0.07 ± 0.04 to 1.18 ± 0.54, n = 6—Fig. 6e,f).

Fig. 6
figure 6

Redox state dependence of CO activation. Exemplar current recordings and mean NPo values after the addition 30 μM CO donor in the absence and presence of 10 μM hemin (a, b), hypoxia (c, d), and hyperoxia (e, f). Additions are indicated by the boxes above the trace or below the bar graph. Scale bars apply to all panels and represent 20 pA/2 min. −Vp = +40 mV [Ca2+]i = 336 nM

To investigate the potential role of the two specific histidine residues that have been previously implicated in CO binding, a site-directed mutagenesis approach was employed. Mutation of the conserved histidine in the intracellular heme-binding pocket of the α1-subunit C-terminal (H616 in the human subunit, accession number NM_002247) has been shown by two groups independently to remove the heme-binding ability of the channel (see [9, 22]). In confirmation, the human H616R mutant was not inhibited by application of 10-μM hemin (mean NPo 0.0015 ± 0.0009 versus 0.0039 ± 0.0024, n = 8—Fig. 7a,b), showing that this specific domain could no longer inhibit channel activity via heme binding. In fact, 10-μM hemin slightly activated this mutant channel, although this increase in NPo was not statistically significant (P > 0.2, n = 8). However, although heme was unable to bind and inhibit, CO activation persisted (raising mean NPo from 0.0007 ± 0.0003 in the breakdown product to 0.0140 ± 0.0047 after CO addition, n = 8—Fig. 7a,b). An extracellular histidine has also been implicated in CO/channel interaction [25, 27]. However, mutation of this conserved histidine (H254R) either singly (mean NPo from 0.006 ± 0.002 to 0.82 ± 0.53, n = 9—Fig. 7c,d) or in combination with the H616R (the H254R/H616R mutant) was completely ineffective at attenuating CO activation and CO increased mean NPo from 0.003 ± 0.002 to 0.178 ± 0.08 (n = 9—Fig. 7e,f).

Fig. 7
figure 7

The role of histidine residues in CO activation. Exemplar current recordings and mean NPo plots from cells transiently expressing H616R mutant in the absence and presence of 30 μM CO donor, 30 μM product or 10 μM hemin (a, b). Exemplar currents in the absence and presence of 30 μM CO donor made from cells transiently expressing the H254R mutant (c, d) and the H616R/H254R double mutant (e, f). Additions are indicated by the boxes above the traces or below the bar graph. Scale bars apply to all panels and represent 5 pA/1 min. −Vp = +40 mV, [Ca2+]i = 336 nM

Lastly, the role of the S9–S10 module on the C-terminal (which contains the “Ca2+ bowl”) in CO sensing by the BKCa α subunit was interrogated by generating a chimeric BKCa α-subunit, designated Slo1C/Slo3T(S9–S10). In this chimera, the tail of the hSlo1 protein was replaced only with the S9–S10 module of mSlo3 tail, as originally performed by Moss and Magleby [17]. As described previously [17], this channel was not activated by [Ca2+]i (data not shown). Although previously untested, we show that the activity of this chimera is also not dependent upon intracellular pH (Fig. 8a,b), a novel observation that demonstrates that the S9–S10 module of mSlo3 is not the pH sensor. Importantly, the application of CO failed to activate Slo1C/Slo3T(S9–S10) at either pH 7.2 (mean NPo of 0.93 ± 0.27 versus 0.85 ± 0.26, n = 7, P > 0.5) or pH 8.0 (mean NPo of 0.81 ± 0.29 versus 0.83 ± 0.28, n = 6, P > 0.5—Fig. 8a,b). Furthermore, even when this chimeric channel could be activated by 10 μM of the BKCa channel opener 1,3-dihydro-1-[2-hydroxy-5-(trifluoromethyl)phenyl]-5-(trifluoromethyl)-2H-benzimidazol-2-one (NS1619—mean NPo increased from 1.24 ± 0.45 to 1.98 ± 0.49, n = 10, P < 0.02), subsequent CO application was completely without effect (mean NPo unchanged at 1.42 ± 0.48 to 1.33 ± 0.44, n = 10, P > 0.35—Fig. 8c,d). That the channels were not running down was demonstrated by the fact that addition of NS1619 after treatment with CO did not dampen the effect of NS1619 to activate the chimeric channels, in all patches so tested (four out the ten patches, e.g., Fig. 8c). These data show that the S9–S10 module of the C-terminal of Slo1 mediates the activation of the BKCa channel by CO, either by interacting directly with CO or by binding to a CO receptor protein/compound that associates with this module of the C-terminal tail of the channel.

Fig. 8
figure 8

The role of the S9–S10 module of the C-terminal tail in CO activation. Typical current recordings and NPo plots of the Slo1C/Slo3T(S9–S10) chimeric channel in the absence and presence of CO donor at pHi of 7.2 (a, b) and pHi 8.0 (b) and in the absence and presence of the CO donor after previous activation of the channel with the BKCa channel opener, NS1619 (c, d). Reapplication of NS1619 is shown in the trace (c). Additions are indicated by the boxes above the traces or below the bar graph. Scale bars apply to both panels and represent 5 pA/1 min. −Vp = +40 mV, [Ca2+]i was chelated nominally to zero with 5 mM EGTA (no added Ca2+)

Discussion

In rodent vascular smooth muscle, it has been suggested that CO modulates native BKCa channels by increasing the coupling of Ca2+ sparks to channel activation [8]. This has not been tested robustly in human recombinant systems, at pertinent [Ca2+]i or with low levels of CO. Indeed, where data are available, they are often from activation by very high concentrations of CO gas and/or [Ca2+]i. In our study, we have shown that the EC50 is around 38 μM of the CO donor (which equates to approximately 20 μM CO [5]—Fig. 1). Reliable measurement of physiological levels of cellular CO are not currently available, but as cellular heme (the hemeoxygenase substate) is often quoted as micromolar, it seems reasonable to assume that CO within cells may also be in that range, especially in the locale of the channel. Thus, it is probable that CO, emanating principally from the breakdown of heme by the co-localized HO-2, is a tonic physiological stimulant of BKCa channel activity. Furthermore, as the system is potentially working below EC50 under physiological conditions, extrinsic CO treatment would still evoke large channel activation, reinforcing the notion that sub-lethal doses of this gas may have therapeutic benefit in the treatment of disease (see, for example, [3]).

The activation by CO exhibits some similarities to the effect of O2 on this channel. Thus, both O2 (or, at least, the cessation of hypoxia [12]) and CO evoke large increases in channel activation rate without any change in deactivation kinetics (Fig. 2). Furthermore, both gases have a component of the activation that can be attributable to increases in Ca2+ sensitivity (Figs. 3, 4, and 5). As suggested by Jaggar et al. [8], there is undoubtedly an important component of the CO activation mechanism that relies on the ability of CO to increase Ca2+ sensitivity of the channel. However, there are clearly other mechanisms involved, and the idea that CO activation is a subtle combination of a number of different influences is suggested by the demonstration of “super-stimulation” by CO and the inability of CO to modulate channel deactivation, as would be expected for a process that modulates only Ca2+ sensitivity.

This study was designed to test several models that have been proposed over the last decade to account for the observations that CO can rapidly and reversibly interact with members of the BKCa channel subfamily. Central to the most recent and, undoubtedly, most attractive proposal is the strict requirement for heme in the channel activation process. In this model (indirect binding hypothesis), CO binds to a heme molecule that is linked to a histidine residue within a heme-binding pocket in the C-terminal of the BKCa subunit. This heme-binding pocket was first described by Hoshi and colleagues [22] and is common to many heme-binding motifs, including that found in cytochrome c. It contains a conserved histidine residue (at position 616 in the human α1-subunit homologue and 611 in the murine subunit) that is crucial to the heme-binding property of this peptide run. Binding of heme at this histidine residue inhibits BKCa channel activity almost irreversibly. This heme inhibition has been confirmed by data from several other studies, including those contained herein (Fig. 6a versus 7a). Although the proposal of the indirect binding model [9] was based on elegantly performed studies, it does not sit entirely well with data that accumulated before its proposal. Specifically, the model states that CO activates the BKCa channel only when bound to reduced heme. However, many previous studies have shown that CO is able to activate both native and recombinant BKCa channels in excised patches bathed in normoxic solutions that contain no reducing agents [19, 28]. If the model was to hold true, CO would only activate channels after a reducing step had been introduced into the protocol that, almost unanimously in these studies, it had not. This suggests, therefore, that either the redox state of heme is unimportant to the CO activation mechanism, that heme remains reduced even in atmospheric O2, or that heme is not required at all. Our present study has addressed each one of these possibilities. The effect of oxidizing and reducing both “intrinsically” bound heme and externally applied heme using several partial pressures of O2 was tested using gassing protocols (Fig. 6). Addition of oxidized heme (hemin) or manipulating the oxygenation of the channel did not affect the ability of CO to activate the channel (Fig. 6). Further evidence for lack of heme involvement in the CO activation process comes from the present structural studies. Mutation of histidine 616 resulted in loss of heme inhibition (indicating that heme was no longer able to bind at that residue) but did not affect CO activation (Fig. 7). Further, mutation of histidine 254 (singly or in combination with H616) again resulted in no significant suppression of the CO activation profile (Fig. 7). Taken together, these data strongly suggest that heme is not the only CO sensor and that previous proposals implicating either of these two specific histidine residues in CO activation need further refining. Such a notion is strengthened by our observation that CO sensitivity is actually conferred upon the channel by structural motifs (the S9–S10 module) within the C-terminal tail of the α-subunit. This notion is reinforced by the fact that even when NS1619 is able to activate the chimeric channel, CO addition to the patch is without affect (Fig. 8). The exact nature of the CO-binding motif is currently unknown, although we speculate that it is responsible for either binding a transition metal-containing α-subunit protein partner or that its cysteines may bind CO in an Fe2+-dependent manner (as suggested very recently by Mann and Motterlini [14]).

CO is a potent modulator of many different biological processes in several different tissues. This is especially true in cardiovascular and neuronal tissue where CO appears to have both a physiological role and therapeutic potential due to its ability to activate BKCa channels (see [2, 11, 14, 20] for recent reviews). The present data have implications for many different physiological systems, but they have particular impact on the current model of O2 sensing by BKCa channels in the carotid body. CO is a rapid modulator of carotid body function [1, 18]. Hemeoxygenase-2 is a protein partner of the BKCa α-subunit and constitutively produces CO in the presence of O2. Heme is an irreversible inhibitor of BKCa, but the co-localization of hemeoxygenase-2 and its enzymatic activity may provide a method whereby the channel/heme interaction can be reversed in an O2-dependent manner. Concurrently, CO activates the channel (independently of its binding to channel-associated heme) and amplifies the activation. In effect, BKCa channel activity can be coupled to ambient O2 levels via the oxygenase-dependent removal of heme and co-ordinate production of CO.