Mechanistic Insight into the Heme-independent Interplay between Iron and Carbon Monoxide in CFTR and Slo1 BK_Ca Channels

Abstract

Ion channels have been extensively reported as an effector of carbon monoxide (CO). However, the mechanisms of heme-independent CO action are still missing. Because most of ion channels are heterologously expressed on human embryonic kidney cells which are cultured in Fe³⁺-containing media, CO may act as a small and strong iron chelator to disrupt a putative iron bridge in ion channels and thus to tune their activity. In this review CFTR and Slo1 BK_Ca channels are employed to discuss the possible heme-independent interplay between iron and CO. Our recent studies demonstrated a high-affinity Fe³⁺ site at the interface between the regulatory domain and intracellular loop 3 of CFTR. Because the binding of Fe³⁺ to CFTR prevents channel opening, the stimulatory effect of CO on the Cl⁻ and HCO₃⁻ currents across the apical membrane of rat distal colon may be due to the release of inhibitive Fe³⁺ by CO. In contrast, CO repeatedly stimulates the human Slo1 BK_Ca channel opening possibly by binding to an unknown iron site because cyanide prohibits this heme-independent CO stimulation. Here, in silico research on recent structural data of the slo1 BK_Ca channels indicates two putative binuclear Fe²⁺-binding motifs in the gating ring in which CO may compete with protein residues to bind to either Fe²⁺ bowl to disrupt the Fe²⁺ bridge but not to release Fe²⁺ from the channel. Thus, these two new regulation models of CO, iron releasing from and retaining in the ion channel, may have significant and extensive implications for other metalloproteins.

Introduction

Carbon monoxide (CO) is emerging an intriguing role in actively regulating ion channels such as voltage-gated L-type Ca²⁺ (Ca_v) channels^1–2, background tandem P domain K⁺ (K_2p) channels³, large-conductance, voltage- and Ca²⁺ -activated K⁺ (BK_Ca) channels⁴, epithelial Na⁺ channels (ENaC)^5–6, P2X receptor⁷, voltage-gated K⁺ (K_v) channels⁸, voltage-gated sodium (Na_v 1) channels⁹, and T-type Ca²⁺ channels¹⁰. These ion channels are generally heterologously expressed on human embryonic kidney (HEK) cells. Their exposure to a CO-releasing molecule (CORM-2), tricarbonyl dichlororuthenium (II) dimer ([Ru(CO₃)Cl₂]₂ 2), stimulates or inhibits channel activity^11–12. Because the HEK cells are cultured in Fe³⁺-containing media such as Dulbecco’s modified Eagle’s medium (DMEM) and there are some trace amount of Fe²⁺ or Fe³⁺ in cells under physiological conditions, Fe²⁺ or Fe³⁺ may be involved in these events. However, most of experiments were done by using the whole-cell configuration, CO may act as a signal molecule (cellular messenger) to modulate channel activity via several intracellular signal pathways. For example, the products of heme catalyzed by heme oxygenase (HO) have not only CO but also Fe²⁺ and biliverdin. Which product specifically acts on ion channels are unknown. In addition, CORM-2 may also directly regulate channel activity and have nothing to do with the CO gas. Until now only Slo1 BK_Ca channels were studied at detail by using the inside-out membrane patch and the CO gas. Because repeated activation of this Slo1 channel by CO is prohibited not by the removal of heme-binding site but by cyanide^{4, 13}, and the non-bonding lone pair of electrons in carbon can render strong transition metal binding to CO, it is possible that CO may bind to a putative Fe site but not to release the Fe ion from the channel. On the other hand, CORM-2 has been reported to evoke the Cl⁻ and HCO₃⁻ currents across the apical membrane of rat distal colon¹⁴, CO and the cystic fibrosis transmembrane conductance regulator (CFTR) anion channel in colon may be involved. Because our recent studies indicated an inhibitive Fe³⁺ site with a high affinity for CFTR^15–16, CO may stimulate the CFTR activity by chelating and releasing Fe³⁺ from the CFTR. Thus, these two case studies on the potential heme-independent interplay between iron and CO in BK_Ca and CFTR channels may reveal two new regulation modes of CO which will prompt mechanistic understanding of heme-independent action of CO on other ion channels and metalloproteins.

CFTR channel

CFTR, a unique and important member of ATP binding cassette (ABC) superfamily, serves as not an ATP-driven pump but an ATP-gated anion channel. It regulates fluid and electrolyte across epithelial surface and resultant mucociliary clearance in multiple organs including lung, pancreas, liver, sweat duct, intestines and reproductive tract. Genetic mutation-induced dysfunction of human CFTR at the respiratory epithelium causes cystic fibrosis (CF), the most common monogenic inherited disorder among Caucasines. This genetic disease brings about many adult CF patients to be chronically infected with Pseudomonas aeruginosa (Pa) in the CF lung so that resultant biofilm formation enhances antibiotic resistance of Pa by acquiring iron. For example, the most common CF-causing mutation F508del increases iron release by human airway epithelial cells and thus promotes the exuberant formation of antibiotic-resistant biofilms while Fe³⁺ chelation with deferoxamine or deferasirox or hypothiocyanite prevents Pa biofilm formation and eliminates established biofilms on CF airway cells overexpressing F508del^17–19. Because Corr-4a increases F508del-CFTR-mediated chloride secretion in the plasma membrane and reduces biofilm formation without changing the iron concentration in the apical medium²⁰, limiting iron acquisition and increasing CFTR activity on airways may be an attractive therapeutic strategy to reduce Pa biofilm formation.

Canonic ABC transporters consist of two copies of a transmembrane domain (TMD) followed by a cytoplasmic nucleotide binding domain (NBD). ATP binding-induced dimerization of two NBDs triggers an inward-to-“outward” gating reorientation of two TMDs to pump a drug out^21–22. However, CFTR has a unique unstructured regulatory (R) domain linking NBD1 with TMD2. The early gold-labelled electron microscopy (EM) structure indicated that the R domain is located beside intracellular loops (ICLs) of TMD2²³. In contrast, recent metal-free cryo-EM structure of human CFTR revealed that the R domain is surrounded by NBD1 and NBD2 and ICLs²⁴. Furthermore, electrophyiological studies showed that the R domain inhibits channel opening not only by preventing the ATP binding-induced NBD1-NBD2 dimerization^25–26 but also by interacting with ICL3 via the H-bond, the Fe³⁺ bridge and the salt bridge^{15–16, 27–28}. Upon phosphorylation by PKA, the R domain promote channel opening by releasing from ICL3 and the NBD1-NBD2 interface for binding to the ICL1-ICL4 interface and the N-terminal of CFTR^{29, 30–33}. Thus, the unphosphorylated R domain may surround TMD2 and NBD2 for channel closure while the phosphorylated R domain may circle both TMDs and NBDs for channel opening (Figure 1). Since there is amount of trace Fe³⁺ in the cell under the physiological condition, further Fe³⁺-bound structures of human CFTR are expected. In addition, because Fe³⁺ inhibits channel opening of CFTR, a Fe³⁺- chelating potentiator is also helpful to rescue CF mutants. Although curcumin is such a potentiator^{16, 31}, the low bioavailability of curcumin limits its clinical application. On the other hand, although orkambi (lumacaftor 200 mg/ivacaftor 125 mg) has been approved by US Food and Drug Administration (FDA) to exploit VX-809 (lumacaftor) to rescue the surface expression of F508del-CFTR in the apical membrane of airway epithelial cells and VX-770 (ivacaftor) to promote channel opening of F508del-CFTR^34–35, VX-770 actually diminishes F508del-CFTR cellular stability and the efficacy of VX-809^36–37. More importantly, iron-required Pa reduces VX-809-stimulated F508del-CFTR Cl⁻ secretion by airway epithelial cells but VX-770 cannot reverse this reduction³⁸. Because iron chelators such as deferoxamine or deferasirox or hypothiocyanite can prevent Pa biofilm formation while VX-770 cannot^{17–19, 38}, VX-770 may be a weak Fe-chelator and the resultant clinical efficacy of orkambi is weakened. In this regard, more cost-effective Fe-chelating potentiators are necessary to more specifically increase F508del-CFTR chloride secretion by airway epithelial cells and to reduce Pa infection.

Figure 1. The tentative interplay between Fe³⁺ and CO in the human CFTR channel.

Fe³⁺ prevents channel opening by bridging the C-terminal part of the R domain (R2) with intracellular loop 3 (ICL3) and allowing the N-terminal part of the R domain (R1) to block the conductance pathway and to prevent the ATP binding-induced NBD1-NBD2 dimerization. In the non-phosphorylated state, the Fe³⁺ ligands include C832 and D836 in the R2 domain and H954 in ICL3 while the PKA site S768 (not shown) in the R2 domain H-bonds with H950 from ICL3. Upon phosphorylation of the R domain by PKA, phosphorylated S768 (not shown) and H775 (not shown) in the R2 domain and H950 are also bound to Fe³⁺ at the R2-ICL3 inteface^15–16. Once Fe³⁺ is removed by CO, the phosphorylated R domain can bind to the ICL1/ICL4 interface via a putative cation-π interaction of the R/K in a PKA site of the R domain with Y161 and F1078 for channel opening by ATP (not shown)³¹. (A), The closed and open states were prepared with on the cryo-electron microscopy structure of human CFTR and the crystal structures of bacterial transporters Sav1866, respectively^{22, 24}. The individual domains or residues are colored uniquely: cyan, TMD1 and NBD1; light gray,TMD2 and NBD2; red, Lasso motif; blue, unphosphorylatyed R1 and R2 domains; red, phosphorylated R1, R2 or R domain; forest green, Y161 and F1078; magenta, H950 and H954; orange, Fe³⁺. (B), The closed and open states were based on the cryo-electron microscopy structure of human CFTR and the crystal structure of the peptidase-containing ATP-binding cassette transporter (PCAT) 1 from Clostridium thermocellum, respectively^{24, 33}.

Early studies demonstrated that hemin-induced active chloride secretion in basolaterally depolarized epithelia of colonic tumor cell line Caco-2 is inhibited by NPPB, a nonspecific anion blocker. Because short hemin treatment is without effect on Cl⁻ secretion, it may result from CO, a product of heme degration by heme oxygenase 1 (HO-1)³⁹. Later in 2011, CORM-2, a CO donor, was also found to evoke chloride and HCO₃⁻ currents across the apical membrane of rat distal colon, which is blocked by another nonspecific anion blocker glibenclamide¹⁴. Given that HO-1 is ubiquitously distributed and apical CFTR is present in rat distal colon^40–43, endogenous CO might be a physiological modulator of CFTR channel currents. It is interesting to employ CFTR-specific inhibitor CFTRinh172 to examine whether the CO gas increases CFTR activity by releasing inhibitive Fe³⁺ from the ICL3-R interface and whether the CO effect is repeated and whether CO enhances Cl⁻ secretion of CF-causing mutants by airway epithelial cells (Figure 1)⁴⁴. Moreover, because recent studies showed that CO can attenuate Pa biofilm formation and the HO-1/CO pathway plays a multitasking role in activating cytoprotective pathway and CFTR expression in airway epithelial cells^45–46, increasing intracellular CO levels could be a promising therapy against lung infection and inflammation in patients with CF.

On the other hand, as the inhibitive Fe³⁺ site in CFTR has been well defined after the channel is expressed in HEK cells which are cultured in the Fe³⁺-containing medium, an Fe²⁺ or Fe³⁺ site may also be involved in other CO-sensitive ion channels such as Slo1 BK_Ca when expressed in an Fe²⁺/Fe³⁺-containing physiological environment.

Slo1 BK_Ca channel

The large-conductance voltage- and calcium-activated human Slo1 BK_Ca channel is a critical member of the high-conductance Slo family of K⁺ channel^47–49. A coupling of electrical signaling to Ca²⁺-mediated events allows this channel to regulate muscle contraction, hormone secretion and neuronal excitability^50–51. Recent Mg²⁺/Ca²⁺-free and Mg²⁺/Ca²⁺-bound cryo-EM structures of the full-length Slo1 channel from Aplysia californica demonstrated that the BK_Ca channel is a tetramer with the transmembrane domain (TMD) on top, the first regulator of K⁺ conductance (RCK1) domain in the middle and the RCK2 domain at the bottom (Figure 2)^52–53. The TMD includes S0 helix, a voltage-sensor domain (VSD) with S1–S4 helices and a pore domain (PD) with S5–S6 helices. Because the VSD of one subunit does not interact with the PD of the other neighboring subunit but the TMD of one subunit interacts with the RCK1 of the other neighboring subunit, the VSD is not domain-swapped with the PD while the TMD is swapped with the RCK1. The homology model of the human Slo1 BK_Ca channel based on these two structures reveals two Ca²⁺ -binding sites and an Mg²⁺ -binding site in the RCK1 and RCK2 domains which form a gating ring. The Ca²⁺ site consists of D367, R514, G533, E535 and E604 in the RCK1 domain, the Ca²⁺ bowl site is composed of Q889, D892, D895, D897 from one RCK2 domain and N449 from the other neighboring RCK1 domain, and the Mg²⁺ site comprises of N172, E339, E374, T396, E399 between the VSD and the RCK1 domain. The Ca²⁺ bowl has the strongest affinity for Ca²⁺ while the Mg²⁺ site has the weakest affinity⁵⁴. Because R514 in the RCK1 domain forms an ionized H-bond with E902 and a cation-π interaction with Y904 in the RCK2 domain, the two Ca²⁺ sites may cooperatively interact with each other to activate the Slo1 BK_Ca channel. In addition, the RCK1 N-lobe, which is shared by both Ca²⁺ binding sites via N449 or D367, tethers the C terminus of S6 via a polypeptide chain linkage or specifically interfaces with the voltage sensor/S4–S5 linker. Therefore, a rigid body titling of the RCK1 N-lobe may directly or indirectly drive a movement of S6 for channel opening in response to Ca²⁺ binding to the RCK1 or the RCK2 while the movement of S4 helices in VSD can in turn exert an effect on the gating ring and the apparent Ca²⁺ binding affinity^52–53. Recent electrophysiological measurements also support the notion that the Ca²⁺-sensitive gating ring facilitates the effective coupling between the VSD and the PD⁵⁵.

Figure 2. The structural assembly of the Slo1 BK_Ca channel.

The models were based on the cryo-EM structure of the Slo1 BK_Ca channel from Aplysia californica⁵³. Four subunits with different colors form a tetramer. The domains with different segments were added from left to right. The voltage sensor domain (VSD) is not swapped with the pore domain (PD) while RCK1 is swapped with TMD (S0+VSD+PD) and RCK2 is swapped with RCK1. The K⁺ conductance pathway centers in the TMD. From a side view, four Mg²⁺ form the first metal layer, four Ca²⁺ at the RC1-RCK2 interface shape the second metal layer and four Ca²⁺ in RCK1.generate the third metal layer. From a bottom view, the second metal layer has the largest ring while the ring sizes for the first and second metal layers are similar.

The Slo1 BK_Ca channel has a conserved heme-binding sequence motif ⁶¹²CXXCH⁶¹⁶ in the RCK1-RCK2 linker⁵⁶. The binding of both heme and hemin to H616, the axial ligand of this human Slo1 channel, increases P_o at negative voltages but decreases P_o at more positive voltages possibly by disrupting both Ca²⁺- and voltage-dependent gating via the gating ring^57–59. Further studies defined heme-sensitive C612 and C615 as a thiol/disulfide redox switch which controls its affinity for heme and CO and HO-2 and thus tunes channel gating in response to changes in the redox state of the cell under hypoxic and normoxic conditions⁶⁰. However, Hou and coworkers found that, in the absence of Mg²⁺ and Ca²⁺, the channel opening of Slo1 BK_Ca by CO is inhibited not by the heme-insensitive mutation H616R or C615S/H616R or internal H₂O₂ or acidification pretreatment but by histidine modifier diethyl pyrocarbonate (DEPC) or the heme-sensitive mutation H365R or H394R or H365R/H394R or H365A/H394A or D367A in the RCK1 domain. They further observed that the CO sensitivity is prohibited not by Mg²⁺-insensitive mutation E374A or E399A but by intracellular Ca²⁺ at the saturating concentration (120 μM)⁴. Thus, the heme-independent CO stimulation found in the Ca²⁺-free environment is Ca²⁺-independent.

On the other hand, Kemp and coworkers reported a different case. Although activation of the Slo1 BK_Ca channel by CO is partially dependent on [Ca²⁺]_i in the physiological range (0.1–1 μM), in the presence of 2 mM Mg²⁺ and 300 nM Ca²⁺ it is not completely prohibited by single or double mutation of H365 and H394 but by 5 mM EGTA or saturating [Ca²⁺]_i (>10 μM) or cyanide or the replacement of the Ca²⁺ -sensitive S9–S10 module of the C-terminal of hSlo1 with the Ca²⁺-insensitive S9–S10 module of the C-terminal of mSlo3^61–62. They also reported that the mutant H365R/H394R in RCK1 domain actually enhances CO sensitivity when [Ca²⁺]_i is in the physiological range. Although CO still potentiates the activity of the heme-bound Slo1 BK_Ca channel at a higher Ca²⁺ concentration (10 μM) by binding to heme^59–61, CO actually stimulates the hemin-insensitive H616R mutant in the presence of 336 nM Ca²⁺⁶². Because of the rapid reversibility of CO and cyanide binding, they proposed that CO activates the Slo1 BK_Ca channel by binding to a transition metal cluster⁶². As all these experiments were carried out by using HEK-293 cells which were cultured in the Fe³⁺-containing medium, and C911G in the RCK2 partially weakens the CO sensitivity⁶², it is possible that C911 may dynamically bind to an Fe site for channel closure and CO may open the BK_Ca channel by competing off C911 from the Fe site. Supporting this dynamic binding of C911 to the Fe site, Tang and collaborates found that C911 can be oxidized by H₂O₂ or MTSEA⁶³. This case is similar to the Zn²⁺-dependent redox switch at the intracellular dynamic T1-T1 interface of a Kv4.1 channel in which Zn²⁺ is tightly bound to H104, C131 and C132 in one T1 domain but weakly bound to C110 in the other neighboring T1 domain⁶⁴. It is interesting that modification of C911 by H₂O₂ or MTSEA still inhibits the BK_Ca channel current⁶³. Thus, oxidation C911 by H₂O₂ or MTSEA may still confer H-bonds to the Fe ligands. More importantly, because the CO effect was still observed in the presence of 11 mM metal chelator EGTA or EDTA and repeated after a washout⁴, Fe ions may not be released from the high-affinity site by EGTA or EDTA or CN⁻ or CO. In other words, the binding stability constant (logK) of Fe ions for the Slo1 BK1 channel may be more than 35 because logK6 of Fe(II) and Fe(III) for cyanide are 35 and 42, respectively and logK of Fe(II) and Fe(III) for EGTA are 11.8 and 20, respectively^65–66. These results are different from those observed in the Kv4.1 channel in which the disulfide bond between C132 and C110 or the H-bond between H104 and oxidized C110 can be formed unless Zn²⁺ is removed from C131, C132 and H104⁶⁴. Thus, the putative abnormal heme-independent Fe site in the Slo1 BK_Ca channel may not be coordinated by canonical protein residues such as cysteine, aspartate or glutamate as found in CFTR but by hydrophobic aromatic residues which may protect Fe ions from releasing by acidification or chelators such as EDTA, EGTA, CO and CN⁻.

Sequence alignment of Slo BK channels demonstrates a conserved motif ³⁹¹F_XXH/F³⁹⁴ and a highly-conserved motif ⁵²⁴WXXXY/F⁵²⁸ across the species (Figure 3), which may function like water-insoluble cis-cyclopentadienyl iron (II) dicarbonyl dimer to form a binuclear Fe²⁺ bowl to interact with CO or CN⁻. Recent structural models of the Slo1 BK_Ca channels showed that a conserved motif ³⁶⁵HXDR³⁶⁸ is 8 Å distant from the former Fe²⁺-binding motif and 4.5 Å away from R514/E902/Y904 near the Ca²⁺ RCK1 site while a conserved motif ⁹⁰⁷QXXXC⁹¹¹ is 6 Å distant from the latter Fe²⁺-binding motif and 3.5 Å away from the Ca²⁺ bowl^{54–55, 67–68}. Thus, two Fe²⁺ can be sandwiched by F391 and H394 via cation-π interactions to form the first putative binuclear Fe²⁺ bowl to which H365/D367 or CO or CN⁻ can bind regulating channel gating via the non-swapping interactions of R514 with E902 and Y904 (Figure 4). In the meanwhile, another pair of Fe²⁺ can be sandwiched by W524 and Y528 via cation-π interactions to produce a second binuclear Fe²⁺ bowl to which Q907/C911 or CO or CN⁻ can bind modulating channel gating via the swapping interactions of N449 with the Ca²⁺ bowl (Figure 5).

Figure 3. Sequence alignment of Slo family members.

Residues discussed in the text are bolded and colored. Numbers above and below the sequences refer to the human Slo1 and the chicken Slo2.2, respectively^74–75. For the human Slo1 BK_Ca channel, both the non-swapping interactions of R514 with E902 and Y904 and the swapping interactions of N449 with Q889, D892, D895 and D897 may facilitate a positive RCK1-RCK2 cooperation and channel opening^52–53. The former may be coupled to the first putative Fe²⁺ bridge between H365/D367 and the first putative Fe²⁺ bowl formed by F391/H394 while the latter may be coupled to the second putative Fe²⁺ bridge between Q907/C911 and the second putative Fe²⁺ bowl formed by W524/Y528. CO may target two putative Fe²⁺ bowls to disrupt two inhibitive Fe²⁺ bridges while CN⁻ may compete off CO and form a prohibitive salt bridge with counterion R368 or K1030.

Figure 4. The tentative heme- and Ca²⁺-independent CO stimulation pathway of the human Slo1 BK_Ca channel.

The in silico models were based on the crystal structure of the Slo1 BK_Ca channel from Aplysia californica⁵³. F391 and H394 in the RCK1 may form a high-affinity binuclear Fe²⁺ bowl via cation-π interactions to prevent a small or large iron chelator to access and to release iron ions. The binding of H365 and D367 to the Fe²⁺ bowl may disrupt the electrostatic interaction of R514 with E902 and the cation-π interaction of R514 with Y504 and thus promote channel closure. However, CO can compete off H365 and D367 from the Fe²⁺ bowl and thus reestablish the non-swapping interactions of R514 with E902 and Y904 for channel opening. CN⁻ may compete off CO from the Fe²⁺ bowl and form an electrostatic attraction with R368 and thus enhance channel closure by disrupting the nonswapping interactions of R514 with E902 and Y904. In any case, the binding of Ca²⁺ to D367 and R514, G533, E535 and E604 would alleviate the CO or CN⁻ effect by weakening the Fe²⁺ bridge between H365/D367 and F391/H394.

Figure 5. The tentative heme-independent but Ca²⁺-dependent CO stimulation pathway of the human Slo1 BK_Ca channel.

The in silico models were based on the crystal structure of the gating ring of the human Slo1 BK_Ca channel⁵². W524, Q525, Y527 and Y528 in the RCK1 may form a high-affinity binuclear Fe²⁺ bowl to prevent a small or large iron chelator to access and to release iron ions. Two Fe²⁺ may be sandwiched by W524 and Y528 via cation-π interactions while Q525 may render an additional coordination with two Fe²⁺ ions to hold the metal in the bowl. Y527 may protect Fe²⁺ from access of a large chelator. At a lower Ca²⁺ concentration (<10 μM), the binding of Q907 and C911 in the RCK2 domain to the second Fe²⁺ bowl in RCK1 domain may disrupt the swapping Ca²⁺ bridge between N449 and the Ca²⁺ bowl and thus promote channel closure. CO may bind to Fe²⁺ to release Q907 and C911 from the second Fe²⁺ bowl and thus to reestablish the intersubunit Ca²⁺ bridge for channel opening. CN⁻ may compete off CO from the second Fe²⁺ bowl and form an electrostatic attraction with K1030 in the RCK2 domain. The resultant disruption of the swapping RCK1-RCK2 Ca²⁺ bridge may promote channel closure. However, at a higher Ca²⁺ concentration (>10 μM), Ca²⁺ bridging N449 with the Ca²⁺ bowl may pull Q907 and C911 or K1030 away from the second Fe bowl so that no CO or CN⁻ effect would be observed. On the other hand, because Mg²⁺ is coordinated by 6 oxygen atoms⁷¹, the binding of Mg²⁺ to the Ca²⁺ bowl may disrupt both the swapping interaction between N449 and the Ca²⁺ bowl and the non-swapping interactions of R514 with E902/Y904 as a result of the RCK1-RCK2 cooperation^{52–53, 72}. In this case, neither CO nor CN⁻ can regulate channel activity.

In the absence of Ca²⁺ and Mg²⁺, N449 and the Ca²⁺ bowl may be disconnected, the binding of either Q907/C911 or CO or CN⁻ to the second Fe²⁺ bowl would not alter channel activity. However, the binding of H365/D367 to the first Fe²⁺ bowl may facilitate channel closure possibly by disrupting the non-swapping interactions of R514 with E902 and Y904. CO may compete off H365/D367 from the first Fe²⁺ bowl to promote channel opening by rebuilding the non-swapping interactions of R514 with E902 and Y904⁴. In this case, CN⁻ is expected to reverse the CO stimulation possibly by forming a salt bridge with R368 (Figure 4). The mutation of H365A/R or H394A/R or D367A/R may disrupt the first inhibitive Fe²⁺ bridge and thus fail to evoke the CO effect⁴. A mutation of F391A may also be expected to render the channel insensitive to CO in the absence of Ca²⁺. However, the mutation of Q907A/C911A may not suppress the Ca²⁺- independent CO stimulation. In addition, acidic activation of the human Slo1 BK_Ca channel by protonation of H365, D367 or H394 may disrupt the first inhibitive Fe²⁺ bridge⁶⁹. Finally, activation of Slo1 BK_Ca channel by Zn²⁺ binding to H365/D367/E399 may also result partly from the disconnection of the first inhibitive Fe²⁺ bridge⁷⁰.

In the presence of 2 mM Mg²⁺ and 5 mM EGTA⁶², the free Mg²⁺ concentration is about 40 μM (logK of Mg²⁺ for EGTA is about 5⁶⁶). Because Mg²⁺ is coordinated with six oxygen atoms⁷¹, the binding of Mg²⁺ to the Ca²⁺ bowl may leave N449 away from the Ca²⁺ bowl⁷². In the meanwhile, the Mg²⁺-bound Ca²⁺ bowl may also disrupt the interactions of R514 with E902 and Y904 possibly as a result of a cooperation between two Ca²⁺ sites^52–53. In this case, the release of H365/D367 from the first Fe²⁺ bowl and the separation of Q907/C911 from the second Fe²⁺ bowl by CO or CN⁻ would not modulate channel gating (Figure 5)⁶².

At a low Ca²⁺concentration (<0.1 μM), the binding of Ca²⁺ to the Ca²⁺ bowl may break up both the swapping interaction of N449 with the Ca²⁺ bowl and the non-swapping interactions of R513 with E902 and Y904 so that the channel is irresponsive to CO or CN⁻⁶¹. In the physiological range of Ca²⁺ (0.1–1 μM), even if the mutation of H365R/H394R damages the first Fe²⁺ bridge, CO may still release Q907 and C911 from the second Fe²⁺ bowl and promote Ca²⁺ bridging N449 with the Ca²⁺ bowl for channel opening while CN⁻ may reverse the CO stimulation possibly by forming an inhibitive salt bridge with K1030 and thus disrupting the Ca²⁺ bridge of N449 with the Ca²⁺ bowl (Figure 5)⁶². When C911 is modified by H O 63 2 2 or MTSEA, the introduced adduct may form a putative H-bond with the Fe²⁺ ligand Y527 or Y528 to prevent the stimulatory Ca2+ bridge between N449 and the Ca²⁺ bowl. When the RCK2 domain of the human Slo1 channel is replaced with that of mSlo3, this chimeric channel is not activated by Ca²⁺⁶¹. Therefore, even if Ca²⁺ binds to E880 and E886 at the Ca²⁺ bowl, N449 is separated from the Ca²⁺ bowl while R514 may not interact with S893 and H895 possibly as a result of RCK1-RCK2 cooperation^52–53. What is more, Q907/C911 have been substituted by T899/T903 in this chimeric channel and thus the second Fe²⁺ bridge is broken (Figure 3). In this case, even if CO and CN⁻ can compete off H365/D367 from the first Fe²⁺ bowl, they may not regulate channel gating in either Ca²⁺ independent or Ca²⁺-independent manner. At a saturation Ca²⁺ concentration (>10 μM), the strong binding of Ca²⁺ to the Ca²⁺ bowl and the Ca²⁺ RCK1 site may disconnect both inhibitive Fe²⁺ bridges and thus the binding of CO or CN⁻ to either Fe²⁺ bowl would no longer change channel gating^4,61. However, CO can still drive channel opening by binding to heme once bound to Slo1 BK_Ca⁵⁹.

Miranda and coworkers used patch-clamp fluorometry and observed the 30° rotation of RCK2 against RCK1 with channel activation⁷³, which is not shown in recent structural models of the BK channel^{52–53, 67–68}. Thus, the Ca²⁺/Mg²⁺-free Slo1 BK channel may Ca not be fully closed possibly as a result of the absence of two inhibitive Fe²⁺ bridges⁵³. In the presence of 10 mM Mg²⁺ or 200 mM Ba²⁺, the binding of Mg²⁺ or Ba²⁺ to the Ca²⁺ bowl, as required by six coordination oxygen atoms⁷¹, may expand the Ca²⁺ bowl and shrink the RCK1-RCK2 linker but may not disrupt two Fe²⁺ bridges⁷². In the presence of 100 μM Cd²⁺, the possible binding of Cd²⁺ to H365 and D367 and E399, as Zn²⁺ does⁶⁸, may only shrink the RCK1-RCK2 linker and disrupt the first Fe²⁺ bridge⁷¹. However, the binding of Ca²⁺ to the Ca²⁺ bowl and the Ca²⁺ RCK1 site may not only disrupt two Fe²⁺ bridges but also enhance the swapping interactions of N449 with the Ca²⁺ bowl and the non-swapping interactions of R514 with E902/Y904 so that the gating ring may be dramatically rearranged (Figure 6)^52–53,73. In other words, the Fe²⁺ bowl may contribute to the large changes in the FRET signal directly or indirectly^72–73. In this regard, CO is expected to induce a large conformation change in the gating ring and CN⁻ can reverse it (Figure 6).

Figure 6. The tentative heme-independent gating regulation of the human Slo1 BK_Ca channel by Ca²⁺ or CO or CN⁻.

The in silico models were based on the crystal structures of the gating ring of the human Slo1 BK_Ca channel^52–53. Two intracellular tandem C-terminal RCK1 and RCK2 from each of four channel subunits form a gating ring. Eight pairs of Fe²⁺ ions may always be bound to the channel. Either the swapping interactions between N449 and the Ca²⁺ bowl or the non-swapping interactions between R513 and E902/Y904 near the Ca²⁺ bowl may be required for channel opening^52–53. In the absence of Ca²⁺, the swapping bridges between N449 and the Ca²⁺ bowl are broken. The Fe²⁺ bridge between H365/D367 and F391/H394 in RCK1 may also disrupt the non-swapping interactions of R514 with E902 and Y904 so that the channel is fully closed. CO may disrupt this first Fe²⁺ bridge at the Ca²⁺ RCK1 site and thus enhance channel opening while CN⁻ can reverse the CO effect. When Ca²⁺ (<10 μM) is added, the Ca²⁺ bridge between N449 and the Ca²⁺ bowl may further promote channel opening because CO has disrupted the second Fe²⁺ bridge between Q907/C911 and W524/Y528 near the Ca²⁺ bowl. On the other hand, if Ca²⁺ (<10 μM) is first applied, the binding of Ca²⁺ to N449 at the Ca²⁺ bowl or D367 at the Ca²⁺ RCK1 site may weaken both Fe²⁺ bridges for the stimulatory interactions between N449 and the Ca²⁺ bowl and between R514 and E902/Y904. CO may completely disrupt both Fe²⁺ bridges so as to promote maximal channel opening. At saturation concentration of Ca²⁺ (>10 μM), once Ca²⁺ bridges N449 with the Ca²⁺ bowl, the binding of Ca²⁺ to the Ca²⁺ RCK1 site may drive a gating rotation of RCK1 against RCK2 and TMD. This rotation may expand both RCK1 and RCK2 but shrink the RCK1-RCK2 linker and finally turn TMDs for maximal channel opening⁷³.

Taken together, the first putative Fe²⁺ bridge may be coupled to the non-swapping interactions of R514 with E902 and Y904 while the second putative Fe²⁺ bridge may be coupled to the swapping interactions of N449 with the Ca²⁺ bowl. Although the disruption of the first and second Fe²⁺ bridges by CO may stimulate channel opening in both Ca²⁺-independent and Ca²⁺-dependent manners, respectively, both non-swapping and swapping interactions may facilitate a positive RCK1-RCK2 cooperation (Figure 6). In addition to the stimulatory swapping Mg²⁺ binding sites at the TMD-RCK1 interface, the stimulatory Ca²⁺ RCK1 site, the stimulatory swapping Ca²⁺ bridge at the RCK1-RCK2 interface and the inhibitory heme site at the RCK1-RCK2 linker, two non-swapping inhibitory Fe²⁺ sites, one at the RCK1-RCK2 interface and the other in the RCK1 domain, are to be identified for full channel closure and a large movement of the gating ring and gating regulation by CO and CN⁻.

It is also exciting to examine the heme-independent interplay between Fe²⁺ and CO in other Slo channels. Figure 3 suggests that most of the Slo1 BK_Ca channels have a potential of the Ca²⁺-dependent or independent CO stimulation even if the heme site is removed because they have not only two highly-conserved Fe²⁺ binding motifs but also two kinds of highly-conserved RCK1-RCK2 interactions. Since the hSlo3 BK_Ca channel is activated by Ca²⁺ but the mSlo3 BK_Ca channel is not⁷⁴, hSlo3 is expected to retain the first Fe²⁺ binding site and the Ca²⁺ RCK1 site and the swapping interactions of N438 with N894 and H896 so that the heme- and Ca²⁺-independent CO stimulation is expected (Figure 3). However, both hSlo3 and mSlo3 BK_Ca channels may exhibit a heme-dependent CO stimulation because they maintain the heme-binding motif⁷⁴. For Slo2 channels, although they may have the first Fe²⁺ bridging motif, the non-swapping RCK1-RCK2 interactions are absent. On the other hand, although they may have a second Fe²⁺ bowl, no residue bridges with it. What is more, they have not a heme binding motif⁷⁵. Accordingly, no CO stimulation would be observed.

Conclusions

Heme-independent activation of the CFTR channel and the Slo1 BK_Ca channel by CO may represent two different modes of gating regulation of ion channels by CO. In the case of CFTR, Fe³⁺ is bound to the R-ICL3 interface^15–16. This interfacial Fe³⁺ site is formed by traditional coordination ligands such as H950 and H954 from ICL3 and C832 and D836 from the R domain (Figure 1). This Fe³⁺ binding motif may be 832CXXXDXnHXXXH954. Once Fe³⁺ is released by CO, a washout cannot reverse the channel activity so that the CO effect would not be repeated. In the case of the human Slo1 BK_Ca channel, Fe²⁺ may be always located in a bowl with a putative motif of 524WXXYY/F⁵²⁸ or 391FXXHF395 no matter whether the channel is opened or closed by a change in voltage or Ca²⁺ or Mg²⁺ concentrations. However, the dynamic binding of H365/D367 to the first Fe²⁺ bowl or the dynamic binding of Q907/C911 to the second one may promote channel closure. Thus, CO may compete off H365/D367 and Q907/C911 from the first and second Fe²⁺ bowls for channel opening in Ca²⁺- independent and Ca²⁺-dependent manners, respectively (Figure 4–6). In this case, a washout followed by the CO stimulation can reverse the channel activity and thereby the CO effect can be repeated. Which regulation mode of heme-independent CO action works in ion channels may depend on whether the CO effect can be repeated or not upon a washout. More experiments by using inside-out membrane patches of HEK cells expressing ion channels may further define these two modes and enrich the interplay between iron and CO in ion channels. In the case of heme-dependent CO regulation of ion channels, if heme can be washed out from ion channels such as ENaC⁷⁶, the potential CO effect on channel activity cannot be repeated. However, if heme cannot be washed out from ion channels such as Slo1 BK⁵⁷ Ca, the possible CO effect on channel activity may still be repeated once heme is bound. Therefore, the heme-independent regulation of ion channels by CO should be defined without a heme binding site. Given that both apical CFTR and Slo1 BK_Ca channels in airway epithelial cells play a critical role in airway hydration and volume homeostasis^77–79, their CO stimulation may protect CF patients against lung infection and inflammation. Since CO may target other metalloproteins to which transition metals are bound, these two heme-independent regulation modes of CO may have significant and extensive implications for those metalloproteins.

Abbreviations

ABC

ATP-binding cassette

CF

cystic fibrosis

CFTR

cystic fibrosis transmembrane conductance regulator

CO

carbon monoxide

CORM

CO-releasing molecule

DMEM

Dulbecco’s modified Eagle’s medium

EM

electron microscopy

ENaC

epithelial Na⁺ channel

HO

heme oxygenase

FDA

Food and Drug Administration

FRET

fluorescence resonance energy transfer

HEK

Human embryonic kidney

DEPC

histidine modifier diethyl pyrocarbonate

ICL

intracellular loop

MTSEA

2-aminoethyl methanethiosulfonate hydrobromide

NPPB

5-nitro-2-(3-phenylpropylamino)benzoate

NPPB-AM

5-nitro-2-(3-phenylpropylamino)benzamide

NBD

nucleotide binding domain

PCAT

peptidase-containing ATP-binding cassette transporter

PD

pore domain

PKA

protein kinase A

Pa

pseudomonas aeruginosa

R

regulatory

RCK

regulator of K⁺ conductance

K_2p

tandem P domain K⁺

TMD

transmembrane domain

VSD

voltage-sensor domain