Identification of a Thiol/Disulfide Redox Switch in the Human BK Channel That Controls Its Affinity for Heme and CO*

Heme is a required prosthetic group in many electron transfer proteins and redox enzymes. The human BK channel, which is a large-conductance Ca2+ and voltage-activated K+ channel, is involved in the hypoxic response in the carotid body. The BK channel has been shown to bind and undergo inhibition by heme and activation by CO. Furthermore, evidence suggests that human heme oxygenase-2 (HO2) acts as an oxygen sensor and CO donor that can form a protein complex with the BK channel. Here we describe a thiol/disulfide redox switch in the human BK channel and biochemical experiments of heme, CO, and HO2 binding to a 134-residue region within the cytoplasmic domain of the channel. This region, called the heme binding domain (HBD) forms a linker segment between two Ca2+-sensing domains (called RCK1 and RCK2) of the BK channel. The HBD includes a CXXCH motif in which histidine serves as the axial heme ligand and the two cysteine residues can form a reversible thiol/disulfide redox switch that regulates affinity of the HBD for heme. The reduced dithiol state binds heme (Kd = 210 nm) 14-fold more tightly than the oxidized disulfide state. Furthermore, the HBD is shown to tightly bind CO (Kd = 50 nm) with the Cys residues in the CXXCH motif regulating affinity of the HBD for CO. This HBD is also shown to interact with heme oxygenase-2. We propose that the thiol/disulfide switch in the HBD is a mechanism by which activity of the BK channel can respond quickly and reversibly to changes in the redox state of the cell, especially as it switches between hypoxic and normoxic conditions.

Large-conductance Ca 2ϩ and voltage-activated K ϩ channels, also known as Slo or BK channels, play important roles in many physiological phenomena, including oxygen sensing, vasodilation, synaptic transmission, and hormone secretion (1,2). In the carotid body, the BK channel was found to be involved in the hypoxic response (3,4). BK channels (for big K ϩ ) have the largest single-channel conductance of all K ϩ -selective channels and are activated synergistically by intracellular Ca 2ϩ and membrane potential (1,5). When open, K ϩ efflux by the BK channel hyperpolarizes the cell membrane potential. Like other voltage-gated K ϩ channels, the human BK channel complex is composed of four pore-forming ␣ (Slo1) subunits that contain transmembrane and cytoplasmic regions (6). Each ␣ subunit includes a seven-helix transmembrane segment (S0) and a voltage-sensing domain (S1-S4) and contributes one-fourth of the ion conduction pore (S5-S6) (7). At the C-terminal end of the BK channel, there is a large more than 700-residue cytoplasmic region, which contains two homologous domains termed "regulators of conductance of potassium" (RCK1 and RCK2) 2 that form a gating ring that is essential for Ca 2ϩ activation of the channel (8,9). RCK1 also serves as a H ϩ sensor (9,10). RCK domains also exist in other potassium channels, such as MthK (Methanobacterium thermautotrophicum), KtrAB (Vibrio alginolyticus), and EcoliTrkA2 (Escherichia coli) K ϩ transporters (11)(12)(13).
The BK channel has recently been identified as a hemoprotein (14 -17). Heme binding under hypoxic conditions strongly inhibits the channel, whereas under normoxic conditions, CO, which is generated by heme oxygenase-2 (HO2) during heme degradation, activates the BK channel. Thus, the substrate (heme) and the product (CO) of heme degradation by HO2 are implied in regulating channel activity. Furthermore, co-immunoprecipitation experiments in HEK293 cells overexpressing the BK channel indicate that HO2 can form a complex with the BK channel (15). In addition, knocking out HO2 expression significantly reduces channel activity (15).
One of the aims of this work is to further explore the interactions between the BK channel and HO2, heme, and CO. This is significant because both heme and CO serve as signaling molecules in various physiological processes in cells, including circadian rhythm, oxygen utilization, and T cell activation (18,19). Although heme degradation in mammals is catalyzed by both the inducible HO1 and the constitutive HO2 isozymes (20), HO2 is the major isozyme in neural tissues and is the sole CO source in neurons (21).
The BK channel appears to be unique as a heme-regulated ion channel. A segment that links RCK1 and RCK2 has been implicated in heme binding (14,16). This linker region contains a CXXCH (X is any amino acid) motif that is highly conserved in all BK channels, and replacement of the Cys or His residues by Ser or Arg, respectively, abolish the sensitivity of the BK channel to heme and CO (14). CO also binds to the complex between heme and a synthetic 23-residue heme-binding peptide (HBP) containing residues 601-623 of the BK channel and including * This work was supported, in whole or in part, by National Institutes of Health Grant R21HL089837. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental "Methods," equations, and Figs. S1-S7. 1  the CXXCH motif (17). Mutagenesis studies demonstrate that an Asp and two His residues within the RCK1 domain are required for the stimulatory action of CO on the BK channel, suggesting another layer of regulation (22). In our work described here, a soluble 134-residue linker region, which we refer to as the HBD, between RCK1 and RCK2 was cloned, expressed, and purified to address the heme binding properties of the human BK channel. We characterized the HBD using EPR and UV-visible spectroscopy. Our results demonstrate that His 616 in the 612 CXXCH 615 motif serves as the axial heme ligand. The Cys residues were shown to form a thiol/ disulfide redox switch that regulates the affinities of the channel for heme and CO. In this paper, "oxidized" and "reduced" refer to the state of the disulfide bonds of the Cys 612 -Cys 615 ; the redox state of the iron in the heme will be explicitly identified as Fe 3ϩ -heme or Fe 2ϩ -heme. Reduction of the disulfide bond in the CXXCH motif enhances the affinity of HBD for heme by ϳ14-fold. In addition, replacement of the Cys residues by Ser decreases the CO affinity by over 50-fold. We propose that thiol/disulfide redox modulation of the affinity for heme and CO is one way that the BK channel can rapidly respond to the switch between hypoxic and normoxic conditions.

EXPERIMENTAL PROCEDURES
Cloning, Overexpression, and Purification of the HBD of Human BK Channel-Biochemical studies were performed with a construct that contains residues 584 to 717 of the cytoplasmic C-terminal end of the human BK channel (GI: 119574985) (23). This HBD consists of the linker segment between the intracellular RCK1 and RCK2 domains. The HBD fragment was cloned into the NdeI and EcoRI restriction sites of the pET28a (Novagen, Gibbstown, NJ) vector, which fuses the HBD to a His 6 tag.
We attempted, but were unsuccessful, in expressing and purifying the entire cytoplasmic region of the human BK channel in either Pichia pastoris or E. coli cells because the protein appeared to aggregate in inclusion bodies. We attempted the expression in yeast using a pPICZ␣A vector and in E. coli using either a pGEX4T-2 vector or a pET28-sumo vector to anchor the glutathione S-transferase or sumo fusion tags, respectively, to the protein. None of these approaches yielded a soluble protein. Because our primary interests were in the effects of heme and redox on the channel and because the RCK domains can also be found in other non-heme-regulated potassium channels, we finally focused on expressing the HBD, which contains the characteristic CXXCH motif and connects RCK1 and RCK2 in the human BK channel.
BL21 cells carrying the HBD-pET28a plasmid were cultured at 37°C in 1 liter of Luria-Bertani (LB) medium containing 100 g/ml of kanamycin. When the optical density (A 600 ) reached 0.8 -1.0, 300 l of 1 M isopropyl ␤-D-thiogalactopyranoside (Gold Biotechnology, Inc., St. Louis, MO) was added to induce HBD expression. The cells were incubated for another 20 h at 18°C and harvested by centrifugation at 7,000 ϫ g for 10 min at 4°C in a J2-HS centrifuge (Beckman, Palo Alto, CA).
A 23-residue HBP containing residues 601-623 of the BK channel and including the CXXCH motif was also used in our studies. The HBP was designed based on previous research (14) and synthesized by EZBiolab (EZBiolab, Carmel, IN).
HO2 and its F253W mutant used in our studies is a truncated form (HO-2⌬289 -316, denoted as HO2 and F253W here, respectively) that lacks the C-terminal membrane-binding region. The generation and purification of HO2 and its F253W variant were performed as previously described (24,25).
Site-directed Mutagenesis of HBD-A number of variants were generated to determine the functional role(s) of the CXXCH motif in the human BK channel, including C612S, C615S, H616A, C612S/C615S, C628S/C630S, and F610W. All mutations were generated using the QuikChange site-directed mutagenesis protocol from Stratagene (La Jolla, CA). The HBD-pET28a plasmid was the template for the PCR using primers from Integrated DNA Technologies. All these variants were purified by the same procedure as that described above for the wild-type enzyme.
Quantification of Heme Binding by the Pyridine Hemochrome Assay-HBD-heme complexes were prepared by incubating 50 M purified HBD with a 10-fold molar excess of Fe 3ϩ -heme from a stock that was freshly prepared in 20% dimethyl sulfoxide to prevent aggregation. After incubation at 4°C for 10 min, excess free heme was removed by chromatography on a Biospin 6 column (Bio-Rad). The HBD concentration in the HBD-Fe 3ϩ -heme complexes was determined using the Bradford method. Then, HBD-heme complexes were diluted to 2.0 or 2.5 M and the amount of heme bound to HBD in the diluted HBD-Fe 3ϩ -heme complex was calculated using the pyridine hemochrome assay with a difference extinction coefficient at 556 nm of 28.32 mM Ϫ1 cm Ϫ1 (26).
Determination of Free Thiol Groups-Free thiol quantification of HBD was conducted by the 5,5Ј-dithiobis(2-nitrobenzoic acid) (DTNB) assay basically as described previously (27). HBD and variants were added to a reaction buffer containing 100 mM Tris-HCl buffer, pH 8.0, and 100 M DTNB at room temperature for 15 min. Then, the difference absorption spectrum was recorded and the difference absorption value at 412 nm was used to calculate the free thiol concentration in the protein. The DTNB titration was performed in the presence of 8 M urea to expose all the thiol groups. When the thiol groups in DTT-reduced HBD were measured, the DTT in the protein solution was removed using the Bio-spin 6 column (Bio-Rad) under anaerobic conditions before reacting with DTNB.
Electron Paramagnetic Resonance (EPR) Spectroscopic Studies-EPR measurements were performed at 10 K on a Bruker EMX spectrometer, operating with microwave frequencies between 9.376 and 9.379 GHz, and equipped with an Oxford ITC4 temperature controller, a Hewlett-Packard model 5340 automatic frequency counter, and a Bruker gaussmeter (Bruker Biospin Corp., Billerica, MA). The EPR samples were prepared in 50 mM phosphate buffer, pH 7.5. The EPR spectrum of the low-spin heme in HBD-Fe 3ϩ -heme complex was simulated using the Bruker SimFonia program (Bruker Biospin Corp., Billerica, MA).
Determination of the Affinity of HBD for Fe 3ϩ -Heme, Fe 2ϩ -Heme, and Fe 2ϩ -Heme-CO-The heme binding affinity of each of the HBD variants was determined as previously described in an Olis-updated CARY-14 double beam spectrophotometer (24). For measurements with the reduced (dithiol-containing) protein, the isolated protein was incubated with a 100-fold molar excess of DTT in an anaerobic chamber (Vacuum Atmospheres Co., Hawthorne, CA) for 30 min and, then, DTT was removed by performing chromatography on a Bio-spin 6 column twice. The titration was performed under anaerobic conditions in serum-stoppered cuvettes. The Fe 3ϩ -heme and Fe 2ϩ -heme stock solutions were freshly prepared as previously described (24).
To determine the binding affinity of the HBD-Fe 2ϩ -heme complex for CO, the HBD-Fe 2ϩ -heme complex was purified using chromatography on a Bio-spin 6 column in an anaerobic chamber. Then, freshly prepared CO solutions were added to the reference and sample (containing the HBD-Fe 2ϩ -heme complex) cuvettes and the difference spectra from 350 to 750 nm were measured. The CO concentration in the stock solution was calculated by titration against a known amount of myoglobin, using extinction coefficients of 121 mM Ϫ1 cm Ϫ1 (deoxy-Mb) at 435 and 207 mM Ϫ1 cm Ϫ1 (MbCO) at 423 nm (28). The titration experiments were performed under anaerobic conditions.
Intrinsic Fluorescence Quenching Analysis-Strong intrinsic tryptophan fluorescence is not detectable in the isolated HBD; therefore, we engineered a Phe to Trp substitution near the CXXCH motif in HBD. This F610W variant exhibits intrinsic tryptophan fluorescence that decreases upon adding HO2, providing a straightforward and direct measure of the interaction between HBD and HO2. The excitation wavelength was set at 285 nm, and the intensity of fluorescence emission at 340 nm of the F610W (1 M) variant was measured with a Shimadzu RF-530 1 PC spectrofluorophotometer (Columbia, MD) at 20°C. Fluorescence quenching experiments were also performed by titrating the HBD containing zinc protoporphyrin (ZnPP) (Sigma) with Fe 3ϩ -heme-bound HO2 using an excita-tion wavelength of 415 nm and recording the emission intensity at 588 nm.
Data Fitting-The heme titration data were plotted and fit to a one-site binding model to determine the dissociation constants (Equations 1 and 2). ⌬A is the absorbance difference between sample and reference cuvettes. ⌬⑀ is the difference extinction coefficient between bound and free ligand and EL is the concentration of the protein-ligand complex. The EL concentration is determined using the quadratic Equation 2, in which E O is the total protein concentration, L O is the total ligand concentration, and K d is the dissociation constant. The quadratic binding equation was used because the K d value was near the concentration of protein in the assay.
The intrinsic fluorescence quenching data were plotted and fit using the same basic equation with a slight modification, e.g. substituting emission intensity for absorbance, etc. (see details under supplemental "Methods"). In intrinsic fluorescence quenching experiments, the EM (emission) intensity was corrected for the inner-filter effect (29).

Determination of the Midpoint Reduction Potential of the Thiol/Disulfide Redox Couples in HBD-
The midpoint reduction potential of the thiol/disulfide redox couples in HBD was calculated using MAL-PEG 5000 (Laysan Bio, Arab, AL) alkylation as previously described (25). The data were fit with the Nernst equation (Equation 3). In Equation 3, E is the ambient potential in solution; E 0 is the midpoint reduction potential of the thiol/disulfide couple in the HBD; R is the universal gas constant: r ϭ 8.31 JK Ϫ1 mol Ϫ1 ; T is the absolute temperature; F is the Faraday constant (9.65 ϫ 10 4 coulombs/mol); and z is the number of electrons transferred in the reaction.

RESULTS
Evaluation of Heme Binding to the CXXCH Motif in HBD-Previous in vivo electrophysiological studies indicated that the human BK channel binds heme (14); however, quantitative heme binding studies of the channel had not been performed. Because our primary interests concerned the effects of heme and redox on the channel, we finally focused on expressing the HBD, which is a soluble and well behaved protein that contains the characteristic CXXCH motif and connects RCK1 and RCK2 in the human BK channel.
We successfully cloned and expressed the His 6 -tagged HBD in E. coli cells and purified the 20-kDa protein to 85% homogeneity, based on densitometry measurements of the bands observed in SDS-PAGE experiments (supplemental Fig. S1). The UV-visible spectra of the Fe 3ϩ , Fe 2ϩ , and Fe 2ϩ -CO states of the isolated protein, in which cysteines were present in their disulfide states (Table 1), demonstrate that HBD specifically binds ferric and ferrous heme with a His residue as the axial ligand (Fig. 1). The absorbance maximum for the Soret peak of the Fe 3ϩ -heme is at 406 nm. The Fe 2ϩ -heme exhibits a Soret peak at 426 nm and two long wavelength peaks at 530 and 560 nm, whereas the Fe 2ϩ -heme-CO adduct shows a Soret peak at 419 nm and two long wavelength peaks at 535 and 568 nm.
The axial EPR spectra of the isolated HBD with g values at 5.92 and 1.99 indicate a predominantly six-coordinate highspin Fe 3ϩ -heme ( Fig. 2A), because the spectra of five-coordinate high-spin Fe 3ϩ -heme systems (30), generally are rhombic.
The g values are indicative of ligation by histidine and water, as observed in myoglobin (30). Interestingly, in addition to the predominantly six-coordinate high-spin Fe 3ϩ -heme species, there are minor low-spin component with g-values (g 1 ϭ 2.96, g 2 ϭ 2.25, and g 3 ϭ 1.53) that are characteristic of His/His axial ligation (31) (Fig. 2A, insets). The EPR spectra of the H616A variant of HBD are very weak, consistent with a low level of bound heme when its axial heme ligand is lost. The minor signals observed in H616A suggests an unknown nitrogenous ligand replaces His 616 (Fig. 2C). Thus, the UV-visible and EPR spectroscopic analysis indicate that the linker segment between the two RCK domains in the human BK channel is the HBD with His 616 in the CXXCH motif serving as the protein-based heme ligand.
Cys 612 and Cys 615 Undergo Thiol/Disulfide Redox Interconversion-Four highly conserved Cys residues are located in the HBD of the human BK channel; two are in the CXXCH motif (Cys 612 and Cys 615 ) and the other two are located at positions 628 and 630. To determine the redox states of these residues, we measured the number of free thiol groups in HBD and variants in their oxidized and DTT-reduced states using the DTNB assay (Table 1). In all cases, DTT was removed before the DTNB assay. This assay was performed with urea-denatured HBD to ensure that all Cys residues were accessible to the DTNB. Only 0.20 and 0.21 thiol groups were detected for the HBD and H616A variant, respectively; however, four thiols per mol of protein were measured for the DTT-reduced HBD (4.1) and the H616A variant (4.0). The C612S and C615S variants contain a single (0.86 and 0.89, respectively) free thiol group in the oxidized state, whereas, in the DTT-reduced state, they contain three (2.5 and 2.7, respectively). Correspondingly, the C612S/C615S variant contains no (0.19) free thiol groups in the oxidized state, but two (2.2) in the DTT-reduced state. Thus, the combined results of the DTNB titrations strongly indicate that the isolated HBD contains two intramolecular disulfide bonds; one is between Cys 612 and Cys 615 , the other is between Cys 628 and Cys 630 .  Stoichiometry of Heme Binding to the HBD-To measure the heme stoichiometry, a 10-fold molar excess of heme was incubated with the oxidized and reduced HBD. Unbound heme was removed by gel filtration chromatography and the heme content was measured by the pyridine hemochrome assay. Oxidized HBD, at 2.0 M, was found to bind 1.6 (Ϯ 0.1) M Fe 3ϩheme and reduced HBD, at 2.5 M, bound 2.3 (Ϯ 0.1) M Fe 3ϩheme (Fig. 3, A and B). The results indicated that both oxidized and reduced HBD saturate at a 1:1 complex with Fe 3ϩ -heme. Furthermore, the EPR spectrum of the reduced HBD is indistinguishable from that of the oxidized protein (Fig. 2, A and B), indicating a predominantly six-coordinate high-spin state heme. In addition, the absorption spectra of the Fe(III)-, Fe(II)-, and Fe(II)-CO states of DTT-reduced HBD are identical to those of the corresponding states of the oxidized protein (Fig.  1). Thus, the heme ligands in the ferric, ferrous, and ferrous-CO states are independent of the redox state of cysteines in the HBD. There is no evidence for heme binding to any of the four Cys residues in the HBD.
Determination of the Redox Potential of Cysteines in HBD-To test the hypothesis that the redox switch constituted by Cys residues in HBD is physiologically relevant, the midpoint reduction potential of the thiol/disulfide redox couples was determined. Because the C628S/C630S variant is very unstable, only the overall redox potential of the cysteines in HBD could be determined. After the isolated HBD samples were equilibrated with solutions having a gradient of ambient redox potentials from Ϫ133 to Ϫ252 mV, the proteins were precipitated and alkylated with MAL-PEG 5000 and run on nonreducing SDS-PAGE. MAL-PEG 5000 traps free thiol residues as an adduct that increases the mass of the protein by 5 kDa per thiol modification. Thus, the oxidized non-modified disulfide is the lowest band on the non-reducing gel and the modified proteins are the other bands of increasing size (Fig. 4A). There are four bands besides that of the fully oxidized HBD (the lowest band) existing in the non-reducing gel, which results from the graded alkylation during the reaction with MAL-PEG 5000. Then, the ratio of fully reduced/fully oxidized protein at each ambient redox potential was quantified and fitted to the Nernst equation. The midpoint reduction potential of the thiol/disulfide couples in HBD was determined to be Ϫ184 (Ϯ2) mV (Fig. 4B). This midpoint potential represents a cumulative value corresponding to both the Cys 612 /Cys 615 and Cys 628 /Cys 630 couples. Because the data fit a single titration curve and because bands corresponding to the modification of four Cys residues are observed, we conclude that both thiol/disulfide pairs are redox active and have similar redox potentials. Furthermore, because the intracellular ambient redox potential ranges from Ϫ170 to Ϫ325 mV (32, 33), the measured midpoint potential indicates that both thiol/disulfide redox couples (Cys 612 /Cys 615 and Cys 628 /Cys 630 ) in the HBD are poised to respond to the redox state within the cell.
Effect of Redox State of the Cys 612 /Cys 615 on Heme Binding-The effect of the redox state of the Cys 612 /Cys 615 thiol/disulfide  redox couple on heme binding was determined. As described above, HBD binds only 1 heme/mol of protein in both oxidized and reduced states. However, the Fe 3ϩ -heme titration curves of the oxidized and reduced HBD are noticeably different, indicating different affinities of the oxidized and reduced HBD for Fe 3ϩ -heme (Fig. 5A). Thus, the dissociation constants (K d ) were measured for the oxidized and reduced states of HBD and for the CXXCH variants.
Dissociation constants (K d values) for the complexes between HBD (and HBD variants) and Fe 3ϩ -heme were calculated by fitting the data to a quadratic one-site binding model ( Table 2). The K d values of all the oxidized HBD proteins were very high, varying from 1.50 to 13.5 M, indicating a very weak Fe 3ϩ -heme affinity. However, the Fe 3ϩ -heme binding affinity of reduced HBD (K d ϭ 0.21 M) was 14-fold higher than that of the oxidized protein (K d ϭ 2.8 M) (Fig. 5A). Similar results were observed with the HBP (Table 2 and supplemental Fig. S2), which also contains the CXXCH motif. For the oxidized HBP (generated by treatment with diamide), the K d value for the ferric heme is 14.5 Ϯ 4.6 M, which is 90-fold higher than that of the reduced HBP (K d ϭ 0.16 Ϯ 0.05 M). In addition, the oxidized and reduced C612S/C615S variant exhibited a similar weak Fe 3ϩ -heme affinity, with K d values of 1.5 and 1.7 M in the oxidized and reduced states, respectively (supplemental Fig. S3). The H616A variant exhibited very large K d values (K d ϭ 13.5 M), as expected if His 616 is the heme ligand in HBD (supplemental Fig. S4). These results strongly indicate that Cys 612 and Cys 615 may act as a redox switch that controls ferric heme binding to HBD. Considering the overall midpoint reduction potential of the cysteines in HBD is Ϫ184 mV, this switch might have physiological relevance in cells.
Effect of Substitutions in Cys 612 and Cys 615 on Fe 2ϩ -Heme-CO Binding-Heme binding to the BK channel inhibits channel activity, whereas CO can activate the channel. Thus, we measured how a substitution in the Cys 612 /Cys 615 couple affects CO binding to the Fe 2ϩ -heme (Table 2). We were unable to perform these titrations with both redox states of the HBD because reductants that reduce Fe 3ϩ -heme also reduce the disulfide bonds in HBD. The native HBD and the C612S/C615S variant have similar affinities for Fe 2ϩ -heme, with K d values of 0.29 and 0.49 M, respectively ( Fig. 5B and supplemental  Fig. S3C). However, the HBD-Fe 2ϩ -heme complex has a 50-fold higher affinity for CO (K d ϭ 0.05 M) (Fig. 5C) than the C612S/C615S variant (K d ϭ 2.7 M) (Fig. 5D). The H616A variant has even weaker affinity for CO (K d ϭ 13.8 M), presumably due to loss of the heme ligand (supplemental Fig. S4D).
HBD Interaction with HO2-HO is the only heme degradation enzyme identified so far in mammals. In vivo studies indicated that HO2 acts as an oxygen sensor to form a complex with human BK channel, regulating channel activity (15). Because the HBD was identified as the heme regulatory domain in the BK channel, we formed the hypothesis that HBD is the key domain involved in the interaction with HO2. This hypothesis was addressed using intrinsic tryptophan fluorescence quenching experiments with a F610W HBD variant (Fig. 6). We performed this substitution at position 610 because, if Cys 612 and Cys 615 are part of a loop or helix, an aromatic residue at position 610 would not be expected to interfere with the disulfide bond. Although sequence comparisons reveal that Trp 609 and Tyr 610 naturally occur in several organisms, the F610W variant exhibits intense intrinsic fluorescence, but the F609W variant does not. The F610W variant has similar affinities for Fe 3ϩ -heme, Fe 2ϩ -heme, and Fe 2ϩ -heme-CO as does the native HBD (Table  2 and supplemental Fig. S5), indicating that this Phe to Trp replacement does not significantly change the properties of HBD. The Fe 3ϩ -heme binding affinity of oxidized F610W (K d ϭ 2.0 M) was 5-fold lower than that of reduced F610W (K d ϭ 0.42 M). Furthermore, the affinities of reduced F610W for the Fe 2ϩ -heme (K d ϭ 0.18 M) and Fe 2ϩ -heme-F610W complexes for CO (K d ϭ 0.03 M) were comparable with those of reduced HBD, indicating that the Phe to Trp mutation does not undermine the binding abilities of protein for Fe 2ϩ -heme and CO. When excited at 285 nm, the F610W variant exhibited an emission spectrum centered at 338 nm. Fluorescence quenching was observed when either apo-HO2 or heme-bound HO2 was titrated into the F610W variant of HBD, yielding K d values of 0.25 Ϯ 0.08 and 0.43 Ϯ 0.04 M, respectively (Fig. 6, A and B). The interaction between HO2 and HBD was further investigated by titrating HBP into the F253W HO2 variant, which has been shown to exhibit quenching of intrinsic Trp fluorescence in earlier ligand binding and redox titration experiments (25). Fluorescence quenching was also observed when apo-HBP was titrated into F253W, indicating the CXXCH motif and its nearby region may be involved in its interaction with HO2 (supplemental Fig. 6).
In related experiments, ZnPP was used to monitor interactions between HO2 and HBD. To ensure that heme/ZnPP exchange between HO2 and HBD did not occur during the experiments, it was important to first establish whether HO2 or HBD binds ZnPP most tightly. It was established that oxidized HBD has a significantly higher affinity for ZnPP (K d ϭ 0.22 M) than for Fe 3ϩ -heme (supplemental Fig. S7A). On the other hand, HO2 binds Fe 3ϩ -heme (K d ϭ 0.03 M, (24)) 15-fold more tightly than ZnPP (K d ϭ 0.45 M, supplemental Fig. S7B). Thus, we could monitor quenching of fluorescence from the ZnPP-HBD complex, which has its emission peak at 588 nm when excited at 415 nm. When the ZnPP-HBD complex was titrated with heme-bound HO2, significant fluorescence quenching was detected (K d ϭ 0.16 Ϯ 0.06 M) (Fig. 6C). These fluorescence quenching studies complement the intrinsic Trp fluorescence quenching experiments and strongly indicate that the HBD not only binds heme, but also is the site in the BK channel that interacts with HO2. Our results indicate that this interaction can occur in both apo-and heme-bound states.

DISCUSSION
The rapid physiological response to hypoxia is mediated by the carotid body, which is an arterial O 2 chemoreceptor organ (34). The rapid inhibition of ion channel activity in glomus cells

TABLE 2 Heme affinity of HBD
Fe 3ϩ -heme and Fe 2ϩ -heme titrations were performed with the oxidized and reduced states of HBD (and HBD variants) under aerobic and anaerobic conditions, respectively. A freshly prepared heme solution was titrated into a 1 M HBD (or HBP) solution, and the absorbance spectra were recorded. The absorbance value at the Soret peak (415 nm for Fe 3ϩ -heme titration and 426 nm for Fe 2ϩ -heme titration) was plotted versus heme concentration. K d was calculated by using the quadratic equation with a one-site binding model (see details under "Experimental Procedures"). All the measurements were done in triplicate.  (35). Under normoxic conditions, the glomus cell membrane is hyperpolarized due to the pumping of K ϩ through the open BK channel, whereas inhibition of channel activity initiates a wave of depolarization that results in increased ventilation by the lungs (34). The involvement of the human BK channel in the physiological responses to hypoxia and the regulation of these responses by heme and CO have been well documented (14, 36 -38). Both ferric and ferrous heme inhibit the BK channel, whereas CO stimulates channel activity (16). In addition, co-immunoprecipitation experiments and immunocytochemical analyses indicate that HO2 binds directly to the BK channel (15). However, the mechanism(s) by which heme inhibits and CO activates the BK channel in response to hypoxic and normoxic conditions, respectively, remains unclear. Mechanisms by which the functions of HO2 and the BK-channel could be coupled include the direct activation of the BK-channel by CO, regulation of heme availability by HO2 and protein-protein associations. Our results indicate that a thiol/disulfide redox switch in the HBD of the BK channel affects binding of both heme and CO as well as HO2-HBD interactions.
Regarding CO activation, two contrasting mechanisms have been recently proposed. Based on the findings that CO affects the binding of Fe 2ϩ -heme, but not Fe 3ϩ -heme, to the HBD and that mutation of Cys 615 and His 616 in the CXXCH motif abolish both heme-dependent inhibition and CO-induced activation of the channel, it was proposed that the effects of CO are mediated by binding to heme (17). On the other hand, because substitutions of residues (two His and an Asp) in the high-affinity RCK1 high-affinity Ca 2ϩ sensor domain disrupt CO-dependent activation of the BK channel (using the CO donor [Ru(CO) 3 Cl 2 ] 2 ) without affecting the inhibition by heme, and because substitutions in the CXXCH motif (of His and Cys) disrupt heme binding without abrogating activation by CO, it was suggested that CO directly interacts with the His and Asp residues in the RCK1 domain in a heme-independent manner (22).
Our results demonstrate that a thiol/disulfide redox switch in the human BK channel regulates binding of both heme and CO. Previous studies have indicated that Cys residues can affect the regulation and/or activity of the human BK channel. Substitution of Cys 911 with Gly decreased the sensitivity of the channel to a CO donor (the tricarbonyldichlororuthenium (II) dimer) by 2-fold; furthermore, cysteine oxidation (39,40) and reaction of some cysteine-specific reagents, e.g. DTNB, 2-aminoethyl methanethiosulfonate, and p-chloromercuribenzoate (41), has been reported to decrease BK channel activity.
Here we have shown that residues Cys 612 and Cys 615 in a CXXCH motif between the two RCK domains interconvert between the thiol and disulfide states according to a midpoint reduction potential of Ϫ184 mV, whereas the His residue (His 616 ) serves as an axial heme ligand. This arrangement allows robust control of heme binding, such that the dithiol(ate) state of either the HBP or the HBD binds Fe 3ϩ -heme 90-or 14-fold tighter, respectively, than the oxidized disulfide form. The K d value for the complex between the reduced (disulfide) state of the HBD and Fe 3ϩ -heme is 0.21 M, which is similar to the K i value (ϳ0.1 M) for inhibition of the BK channel by Fe(III)-heme (14). Importantly, the K d value for the oxidized FIGURE 6. Titration of HBD with HO2. Fluorescence quenching experiments were performed to examine the interaction between HBD and HO2. The data were corrected by subtracting the inner filter effect as described (29). A, apo-F610W was titrated with apo-HO2. The experiments were performed by exciting at 285 nm and following the emission intensity at 338 nm. B, apo-F610W was titrated with Fe 3ϩ -heme-HO2, with the excitation wavelength set at 285 nm, and the emission intensity was measured at 338 nm. C, ZnPP-HBD was titrated with Fe 3ϩ -heme-HO2, with excitation at 415 nm and emission at 588 nm. HBP of 14.5 M (or 2.8 M for the HBD) is well outside the range of intracellular free heme concentrations of 20 -100 nM (42)(43)(44)(45)(46) (furthermore, free heme levels above 1 M are toxic (46)). Because the ambient intracellular redox potential ranges from Ϫ170 to Ϫ325 mV (32,33), the CXXC motif in the HBD should undergo redox interconversion under physiologically relevant conditions, e.g. normoxia and hypoxia. Therefore, when the HBD converts to the oxidized state, the BK channel would be expected to release its bound heme.
Heme and CO elicit opposite effects on activity of the BK channel. Our results indicate that CO interacts directly with heme in the HBD, forming a high affinity Fe 2ϩ -CO complex (K d ϭ 0.05 M), which could be considered to favor the former of the two mechanisms of CO activation described above (17). In comparison, the hemes in NPAS2 (K d ϭ 1 M) and a bacterial CO sensor (CooA) (K d ϭ 11 M) bind CO with significantly lower affinity (47,48). If the HBD is the sole CO binding site within the BK channel, one possible explanation for the importance of the RCK1 domain in the CO response is that allosteric effects may couple binding of Ca 2ϩ and CO to separate sites. Because it is highly unlikely that CO can bind directly to the implicated amino acid residues (His 365 , His 394 , and Asp 367 ) in the RCK1 domain in a metal-independent manner, one possibility is that the His and Asp residues in RCK1 ligate a lowvalent metal ion, e.g. Fe 2ϩ or Cu 1ϩ , which forms a CO binding site. There is evidence that Zn 2ϩ can activate the BK channel and that His 365 and Asp 367 are required for this activation, suggesting that these residues can bind metal ions. A large class of metalloenzymes, e.g. oxygenases, contains a 2-His-1-carboxylate facial triad motif in which a single metal ion (Zn 2ϩ or Fe 2ϩ ) is coordinated by two His residues and one Asp (or Glu) in a motif that often consists of HXDX n H (X can be any residue and n represents a variable number of residues) (49,50), which matches the sequence in RCK1 (His 365 -X-Asp 367 -X 22 -His 394 ). Although we are unaware of CO binding to the metal center of a protein containing the 2-His-1-carboxylate facial motif, the structure of isopenicillin N synthase has been solved with bound NO (51), a CO analog. If this site does indeed bind Zn 2ϩ , being a d 10 metal ion, it would not be able to bind CO; however, there are examples of Zn 2ϩ or Fe 2ϩ competing for binding to a metal binding site, e.g. carbonic anhydrase (52). Thus, further studies are required to clarify the relationship between CO and the potential metal binding sites in the RCK1 domain.
As was observed for heme binding, affinity of the HBD-Fe 2ϩheme for CO appears to be regulated by the CXXC motif. Surprisingly, the affinity of Fe 2ϩ -heme in the C612S/C615S variant, which was expected to mimic the reduced protein, is 50-fold weaker than that of the native protein (K d ϭ 2.68 M). How Cys 612 and Cys 615 affect CO binding is unclear. Perhaps one of the Cys residues undergoes ionization, which would alter the microenvironment of the heme in a way that Ser could not. Besides the Cys residues in the CXXCH motif, Cys 628 and Cys 630 form the other intramolecular disulfide bond in HBD. We have been unable to address the properties of these residues because the single and double (C628S/C630S) variants were unable to be purified, suggesting that these residues may play a structural role.
The results of our intrinsic fluorescence quenching experiments indicate that the HBD interacts directly with HO2. These results support earlier co-immunoprecipitation and immunocytochemical analyses indicating that HO2 binds directly to the BK channel (15). This interaction could allow HO2 to locally influence both the heme and CO concentrations to which the BK channel is exposed. The influence of the CXXC motif on heme and CO binding suggests that this motif functions as a thiol/disulfide redox switch. Interestingly, a similar redox regulatory mechanism has been recently identified in human HO2; however, with an opposite redox regulatory modality (24,25). HO2 has low affinity for heme when the Cys residues in a heme response motif are in the reduced (free thiol) state; however, when they switch to the oxidized (disulfide) state, the affinity for heme markedly increases (24,25). Thus, it is most significant that redox switches in both HO2 and the BK channel respond in a seemingly coordinated way to redox changes, with HO2 releasing heme under redox conditions that optimize heme binding to HBD and HBD releasing heme under oxidizing conditions that stimulate heme degradation by HO2. Fig. 7 describes a working hypothesis for how HO2 and the BK channel system may coordinately respond to altering intracellular redox conditions and to the related shifts between hypoxia and normoxia. Under hypoxic conditions, where the redox potential would decrease, the thiol/disulfide switches in HO2 and the BK channel would convert to the dithiol states, with HO2 having low affinity and BK channel having high affinity for Fe 3ϩ -heme. The combined effects of low heme affinity and limited O 2 concentration would decrease heme degradation activity by HO2, resulting in a local increase in heme and decrease in CO concentrations. Furthermore, the decrease in intracellular redox potential associated with hypoxia would promote binding of heme to the BK channel. Thus, these coordinated effects of high heme and low CO would promote clos- FIGURE 7. Proposed redox regulatory mechanism of human BK channel and HO2 under hypoxic/normoxic condition. This figure describes a working hypothesis for how HO2 and the BK channel system may coordinately respond to intracellular redox changes, e.g. hypoxia and normoxia. Under hypoxic or more reducing conditions, the thiol/disulfide redox switches in HO2 and the BK channel would be in the dithiol states, with HO2 having low affinity and BK channel having high affinity for Fe 3ϩ -heme. The combined effects of low heme affinity and low O 2 concentration would decrease HO2 activity, resulting in an increase in heme and decrease in CO concentrations. Increase in heme affinity of the BK channel, combined with high heme and low CO levels would promote closing of the channel, and depolarization of the cell membrane. Under more oxidizing and/or normoxic conditions, the redox active thiols of HO2 and the BK channel would exist in oxidized states, with HO2 having high affinity and the BK channel having low affinity for Fe 3ϩ -heme. Furthermore, the increased O 2 levels would stimulate HO2 activity to generate CO, which would activate the channel, leading to hyperpolarization of the cell membrane.
ing of the BK channel. Under normoxic conditions, the redox active thiols of HO2 and the BK channel would exist in oxidized states, with HO2 having high affinity and BK channel having low affinity for Fe 3ϩ -heme. Furthermore, the increased O 2 levels would stimulate degradation of heme to generate CO, which would activate the channel. It is likely that CO binding to the HBD would be one of the most rapid responses to the re-establishment of normoxic conditions, which would be followed by release of heme from the BK channel, due to its low heme affinity when the CXXC motif is in the oxidized state. The redoxbased control of BK channel proposed here would of course be interfaced with the many other inputs into regulation of ion channel function, including the major roles of Ca 2ϩ and membrane potential and other modulatory properties of the RCK domains.
In conclusion, our research provides strong evidence for a thiol/disulfide switch in the BK channel that involves Cys residues in the CXXCH motif of a HBD. This switch regulates affinity of the HBD for heme and CO and for HO2. We speculate that this redox-linked alteration in properties of the HBD has important physiological roles, including the ability to regulate the activity of the human BK channel in response to intracellular hypoxic/normoxic conditions.