Hemoxygenase-2 Is an Oxygen Sensor for a Calcium-Sensitive Potassium Channel

See allHide authors and affiliations

Science  17 Dec 2004:
Vol. 306, Issue 5704, pp. 2093-2097
DOI: 10.1126/science.1105010


Modulation of calcium-sensitive potassium (BK) channels by oxygen is important in several mammalian tissues, and in the carotid body it is crucial to respiratory control. However, the identity of the oxygen sensor remains unknown. We demonstrate that hemoxygenase-2 (HO-2) is part of the BK channel complex and enhances channel activity in normoxia. Knockdown of HO-2 expression reduced channel activity, and carbon monoxide, a product of HO-2 activity, rescued this loss of function. Inhibition of BK channels by hypoxia was dependent on HO-2 expression and was augmented by HO-2 stimulation. Furthermore, carotid body cells demonstrated HO-2–dependent hypoxic BK channel inhibition, which indicates that HO-2 is an oxygen sensor that controls channel activity during oxygen deprivation.

Large-conductance, Ca2+-sensitive potassium (BK) channels are strongly implicated in the acute O2 signaling cascade of a number of cellular systems. In carotid body chemoreceptors (1, 2), low arterial pO2 is detected by BK channels, and the resulting depolarizing signal is ultimately transduced into increased ventilation. BK channels in pulmonary arteriolar myocytes may contribute to both persistent prenatal (3) and acute postnatal hypoxic pulmonary vasoconstriction (3, 4). Hypoxic inhibition of BK channels in perinatal adrenomedullary chromaffin cells is necessary for the catecholamine secretion crucial for preparing the newborn's lung to breathe air (5). Hypoxic depression of BK channel activity in neurons of the central nervous system (68) may also contribute to the excitotoxicity that results from increased neuronal excitability. As O2 supply becomes compromised, BK channels are acutely and reversibly inhibited (2, 710), resulting in cell depolarization. Subsequent voltage-gated Ca2+ influx induces hypoxia-dependent neurotransmitter release (11). In the carotid body, this ultimately results in increased ventilation. However, the molecular nature of the O2 sensor that regulates BK channels has not been determined. Acute regulation by O2 of both native (2, 8, 12) and recombinant (10) BK channels is variably retained (2, 8, 6, 12), which indicates that the O2 sensor is either cell-specific or developmentally regulated, or both. Because human recombinant BK channels consisting of both BKα subunits and BKβ subunits (BKαβ) demonstrate O2 sensitivity in inside-out membrane patches (10), the O2 sensing machinery must be closely associated with the channel protein complex (8, 10, 11).

To identify proteins associated with recombinant human BK channels, human BKα1 (KCNMA1) and BKβ1 (KCNMB1) were stably expressed in human embryonic kidney (HEK293) cells (13). Proteins were immunoprecipitated with a BKα-specific antibody from lysates of HEK293 cells and separated by two-dimensional (2D) (Fig. 1A, right) and 1D (Fig. 1B, right) gel electrophoresis. Parallel immunoprecipitations were performed on untransfected cells (Fig. 1A, left, and Fig. 1B, left). Of the unique proteins that immunoprecipitated with BKα, peptide mass mapping with mass spectroscopy of trypsin digests consistently identified gamma glutamyl transpeptidase (GGT) and hemoxygenase-2 (HO-2) as associated proteins (table S1). Although GGT associates directly with BKα, it is not involved in hypoxic inhibition of BKαβ channels (14). Western blot analyses of cell lysates revealed constitutive expression of HO-2 in the absence or presence of BKαβ (Fig. 1C). The inducible form of hemoxygenase (HO-1) could not be identified in either case (15). BKα and HO-2 coimmunoprecipitated only from lysates of BKαβ cells (Fig. 1D). Despite the abundant expression of endothelial nitric oxide synthase (eNOS) and the α subunit of the Na+- and K+-dependent adenosine triphosphatase (Na+,K+-ATPase) (Fig. 1C), BKα did not immunoprecipitate in a complex with either protein (Fig. 1D). Like HO-2, eNOS is constitutively expressed, is an NADPH (reduced form of nicotinamide adenine dinucleotide phosphate)–dependent enzyme, and produces a short half-life gas. The α subunit of the Na+,K+-ATPase is a plasma-membrane protein. Colocalization of BKα with HO-2 was confirmed immunocytochemically by confocal microscopy (Fig. 1E).

Fig. 1.

HO-2 as a functional BKαβ channel–associated protein. (A) 2D gel electrophoresis of proteins immunoprecipitated with a BKα antibody from wild-type (wt) and BK αβ-expressing HEK293 cells. Boxed area indicates location of protein spots selected for matrix-assisted laser desorption/ionization–time-of-flight (MALDI-TOF) analysis. (B) SDS–polyacrylamide gel electrophoresis (SDS-PAGE) of immunoprecipitates from wt and BKαβ cells. Bands removed for MALDI/TOF analysis are indicated by the asterisk. Linear pH gradients and/or molecular weight markers (in kD) are shown. (C) Western blot analyses from lysates of wt and BKαβ cells show that HO-2, eNOS, and α subunit of Na+,K+-ATPase (pump) are constitutively expressed. (D) Western blot identification of BKα and HO-2 following immunoprecipitation (IP) with the antibodies shown to the right (top two blots). Neither eNOS nor the pump immunoprecipitated with BKα (lower two blots). Pump Western blot and IP were not performed on wt cells. (E) Confocal images of BKαβ cells showing colocalization (yellow) of BKα (red) and HO-2 (green) with specific antibodies. Scale bar is 20 μm and applies to all. (F) Exemplar current recording from an inside-out membrane patch excised from a BKαβ cell. Periods of application of 30 μM of CO donor and 30 μM of its control (product) shown above trace. Arrows indicate 10-s sections that have been expanded and shown in (G), (H), and (I). (J) Mean NPo plot showing effect of 30 μM of CO donor and product on BKαβ channel activity (n = 13 patches). Comparisons between groups are indicated by P values above bars and are from analysis of variance/Bonferroni post hoc test; NS indicates no significant difference. (K) Current-voltage relationships showing lack of voltage-dependence of CO activation. (L) Mean NPo plot showing effect of 10 μM biliverdin (n = 12 patches) and additive effects of 30 μM CO donor and 10 μM biliverdin (n = 5 patches). Patch potential (–Vp) = +20 mV; [Ca2+]i = 335 nM.

In the presence of O2 and NADPH, hemoxygenases catalyze the breakdown of heme to biliverdin, iron, and CO (16). Under normoxia (pO2 ≈ 150 mmHg), BKαβ channel activity was reversibly activated by the chemical CO donor, [Ru(CO)3Cl2]2. The breakdown product of this compound, RuCl2(DMSO)4, which does not release CO, did not affect channel activity, which indicates that CO activates BKαβ channels in inside-out membrane patches (Fig. 1, F to L). The CO donor produced a normalized NPo (product of the number of channels and their open-state probability) 15 times that of the control (Fig. 1, F and J), and this effect was apparent at all activating potentials (Fig. 1K and fig. S1). Biliverdin increased BK channel activity by a factor of 4 (Fig. 1L). In inside-out membrane patches treated sequentially with biliverdin and the CO donor, activation was additive, with the CO donor causing a further increase to 28 times as much as the control (Fig. 1L). Wild-type HEK293 cells do not display BK currents (10), and no channel activation could be evoked by the CO donor (17).

Consistent with earlier reports (10, 14), hypoxia (acute reduction in pO2 of the intracellular bathing solution to between 15 and 25 mm Hg) resulted in a modest depression in NPo of inside-out patches excised from BKαβ cells (Fig. 2, A to D). Under normoxia, addition of the HO-2 cosubstrates, heme (1 nM) and NADPH (1 μM), evoked an increase in patch NPo (Fig. 2, E to H). The 1 nM heme alone does not modulate recombinant BKα channel activity (18) or BKαβ channel activity, because NPo in the absence (0.25 ± 0.24) or presence (0.31 ± 0.23) of 1 nM heme were not significantly different from each other (19). In the continued presence of the HO-2 cosubstrates, hypoxia evoked a decrease in NPo of more than 70%, which suggests that the enzymatic activity of HO-2 enhances the O2 sensing ability of the HO-2/BKαβ channel protein complex (Fig. 2, I to L). Thus, O2 sensing by recombinant human BKαβ channels consists of two components, of which the HO-2–dependent element is the larger of the two.

Fig. 2.

Hemoxygenase substrates augment BKαβ channel activity and hypoxic inhibition. Exemplar traces and mean NPo plots indicating modest hypoxic channel inhibition in untreated patches (A to D) (n = 14 patches), increased baseline channel activity by 1 nM heme plus 1 μM NADPH (E to H) (n = 15 patches), and augmentation of the hypoxic inhibition in the continued presence of heme plus NADPH (I to L) (n = 10 patches). All traces are from inside-out membrane patches from BKαβ cells. Comparisons between groups are indicated by P values above bars and are from Student's t test. Patch potential (–Vp) = +20 mV; [Ca2+]i = 335 nM; normoxic pO2 ≈ 150 mmHg; hypoxic pO2 ≈ 15 to 25 mmHg.

Selective knockdown of HO-2 expression was achieved by RNA interference. Cells were transfected for 48 hours with Cy3-labeled small interfering (si) RNA designed against either a scrambled human glyceraldehyde phosphate dehydrogenase (GAPDH) coding sequence or the human HO-2 coding sequence. Fluorescence microscopy (Fig. 3A) was used to identify siRNA-positive cells prior to patch clamp. No knockdown of HO-2 immunoreactivity was observed with the scrambled siRNA. In contrast, almost total loss of HO-2 immunoreactivity was achieved with the specific HO-2 siRNA. This was confirmed by Western blot analysis (Fig. 3B). The NADPH- and heme-dependent hypoxic suppression seen in untreated cells (Fig. 2, I to L) was maintained in control siRNA-treated cells (Fig. 3, C to F). Following suppression of HO-2 expression with HO-2 siRNA, mean patch NPo was depressed and NADPH- and heme-dependent hypoxic suppression was absent (Fig. 3, G to J). However, the CO donor rescued this loss of function in all membrane patches tested (Fig. 3, K to N).

Fig. 3.

Modulation of heme- and NADPH-dependent hypoxic inhibition of BKαβ channels after knockdown of HO-2 expression by siRNA. (A) Cy3-labeled transfected cells are shown in (Ai) and (Aiv). HO-2 immunostaining with a fluorescein isothiocyanate–labeled secondary antibody shows the persistent expression of HO-2 after scrambled siRNA treatment (Aii) and knockdown of HO-2 expression after HO-2 siRNA treatment (Av). (Aiii) and (Avi) show the merged images. Scale bar in (Av) = 20 μm and applies to all. (B) Western blot of BKαβ cells transfected with scrambled siRNA (left) and HO-2 siRNA (right) shows ∼90% knockdown of HO-2 expression by HO-2 siRNA treatment. Exemplar traces and mean NPo plots of NADPH- and heme-dependent hypoxic BKαβ channel inhibition in scrambled siRNA-treated patches (C to F) (n = 10 patches), almost complete loss of channel activity in HO-2–treated patches (G to J) (n = 7 patches), and rescue of channel activity by the CO donor in HO-2–treated patches (K to N) (n = 3 patches). The siRNA-positive cells were selected by Cy3 fluorescence before patch clamp. Comparisons between groups are indicated by P values above bars and are from Student's t test; NS indicates no significant difference. Patch potential (–Vp) = +20 mV; [Ca2+]i = 335 nM; normoxic pO2 ≈ 150 mmHg; hypoxic pO2 ≈ 15 to 25 mmHg.

BK channel activity present in inside-out membrane patches excised from rat carotid body glomus cells was modestly inhibited by hypoxia (Fig. 4, A to C and G), an effect that is likely to be overestimated because rundown occurred in the absence of HO-2 substrates (Fig. 4G). As is the case with the recombinant system, supplying the channel complex with heme and NADPH (Fig. 4G) or the addition of the CO donor (20) under normoxic conditions increased patch NPo. Furthermore, hypoxic inhibition was augmented in the presence of heme and NADPH, which suggests that the HO-2–dependent O2 sensing system is fully operable in native carotid body glomus cells (Fig. 4, D to G).

Fig. 4.

Augmentation of carotid body glomus cell BK channel activity by hemoxygenase substrates. Exemplar traces indicating the modest hypoxic channel inhibition observed in untreated patches (A to C) (n = 7 patches), increased baseline channel activity by 1 nM heme plus 1 μM NADPH (C and D) (n = 7 patches), and augmentation of the hypoxic inhibition in the continued presence of heme/NADPH (D to F) (n = 7 patches). Corresponding mean NPo values are shown in (G). Comparisons between groups are indicated by P values above bars and are from Student's t test. All traces are from inside-out membrane patches excised from carotid body glomus cells. Patch potential (–Vp) = +20 mV; [Ca2+]i = 335 nM; normoxic pO2 ≈ 150 mmHg; hypoxic pO2 ≈ 15 to 25 mmHg.

HO-2 is highly and constitutively expressed in neuronal and chemosensing tissues, including carotid body glomus cells (16, 2123), whereas HO-1 is not. Functional interaction between HO-2 and BKαβ channels is intact in excised membrane patches, which suggests that their physical interaction is membrane delimited, whether direct or indirect. Either way, a colocalization of BKαβ with HO-2 is necessary for both basal and O2-dependent activity. Channel activation by CO gas has also been reported in glomus cells, further supporting the notion that HO-2 activity is crucial to native BK channel regulation (2). The presence of HO-2 in the BK channel complex provides a molecular explanation for the observation that hemoxygenase inhibition results in carotid body excitation (23). Our study supports a model in which O2 sensing is conferred upon the BK channel by colocalization with HO-2. In normoxia, tonic HO-2 activity generates CO and biliverdin, both of which maintain the open-state probability of the channel at a relatively high level. CO and biliverdin together evoke BK channel activation that is more than additive, representing a means by which the normoxic signal is amplified. However, because biliverdin is rapidly broken down to bilirubin, it seems more likely that the physiological messenger is CO. Whatever the molecular nature of the CO effect, cellular CO levels are reduced during a hypoxic challenge as HO-2 substrate (O2) becomes scarce, and they rapidly fall below the critical threshold for the maintenance of BK channel activity at the tonically high level. Thus, HO-2 functions as a sensor of acute reduction in environmental O2 by suppressing both native and recombinant BK channel activity, primarily through the production of CO.

Supporting Online Material

Materials and Methods

Fig. S1

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article