Bilirubin Oxidation End Products Directly Alter K + Channels Important in the Regulation of Vascular Tone

Abstract

The exact etiology of delayed cerebral vasospasm following cerebral hemorrhage is not clear, but a family of compounds termed ‘bilirubin oxidation end products (BOXes)’ derived from heme has been implicated. As proper regulation of vascular smooth muscle tone involves large-conductance Ca²⁺- and voltage-dependent Slo1 K⁺ (BK, maxiK, K_Ca1.1) channels, we examined whether BOXes altered functional properties of the channel. Electrophysiological measurements of Slo1 channels heterologously expressed in a human cell line and of native mouse BK channels in isolated cerebral myocytes showed that BOXes markedly diminished open probability. Biophysically, BOXes specifically stabilized the conformations of the channel with its ion conduction gate closed. The results of chemical amino-acid modifications and molecular mutagenesis together suggest that two specific lysine residues in the structural element linking the transmembrane ion-permeation domain to the carboxyl cytosolic domain of the Slo1 channel are critical in determining the sensitivity of the channel to BOXes. Inhibition of Slo1 BK channels by BOXes may contribute to the development of delayed cerebral vasospasm following brain hemorrhage.

Keywords

bilirubin maxi-K channels Slo1 BK channels vasoconstriction

Introduction

Large-conductance Ca²⁺- and voltage-dependent Slo1 K⁺ (BK, maxi-K, K_Ca1.1) channels open in response to an increase in intracellular Ca²⁺ concentration and/or depolarization, generally providing a negative feedback influence on cellular excitability (Salkoff et al, 2006). Physiologic roles of Slo1 BK channels are well documented in many phenomena, including regulation of smooth muscle tone, determination of action potential duration, action potential frequency adaptation, neurotransmitter release, and neurovascular coupling (reviewed in Salkoff et al (2006)). Mice with the gene for the Slo1 channel pore-forming subunit (Kcnma1) or the predominantly vascular auxiliary subunit β1 (Kcnmb1) disrupted exhibit various smooth muscle-related phenotypes, including hypertension (Brenner et al, 2000; Sausbier et al, 2005), erectile dysfunction (Werner et al, 2005), and urinary incontinence (Meredith et al, 2004; Werner et al, 2005) along with other neuronal phenotypes (Salkoff et al, 2006). Although compensatory responses for the gene disruption may occur (Sprossmann et al, 2009), the phenotypes observed in the genetically modified mice are generally consistent with the concept that activation of Slo1 BK channels promotes vascular relaxation and inhibition of the channels facilitates vascular constriction (Nelson et al, 1995).

Some human diseases may be associated with the dysregulation of Slo1 BK channels. For instance, a mutation in one of the putative Ca2⁺-sensor regions of the Slo1 channel is linked to epilepsy and paroxysmal movement disorder (Du et al, 2005). A form of hypertension has been reported to involve downregulation of the vascular Slo1 BK channel complex (Amberg and Santana, 2003), and a gain-of-function mutation in the auxiliary subunit gene β1 may offer protection against diastolic hypertension (Fernández-Fernández et al, 2004). Furthermore, pharmacological openers of Slo1 BK channels have been suggested to have a neuroprotective role following stroke (Gribkoff et al, 2001).

Subarachnoid hemorrhage (SAH) is a serious condition that claims many lives, and even those patients who survive the initial incidence often face delayed high mortality and exhibit delayed neurologic deterioration (Macdonald et al, 2007). Subarachnoid hemorrhage is often followed by delayed cerebral vasospasm, long-lasting narrowing of cerebral blood vessels, typically occurring 1 week after the initial hemorrhage (Macdonald et al, 2007). Although a definitive single etiology has not been established, heme and heme breakdown products released from hematomas may have critical roles in the vasospasm (Macdonald and Weir, 1991). In particular, a class of heme breakdown products known as bilirubin oxidation end products (BOXes; Figure 1A), small pyrroline derivatives, has been implicated (Clark et al, 2002; Clark and Sharp, 2006; Clark et al, 2008; Kranc et al, 2000; Pyne-Geithman et al, 2005). Bilirubin oxidation end products are present in cerebral spinal fluids (CSFs) of cerebral hemorrhage patients with vasospasm at concentrations of up to 1 μmol/L in the extracellular compartment (Pyne-Geithman et al, 2005) and potentiate contraction of isolated porcine carotid arteries (Clark et al, 2002; Kranc et al, 2000; Pyne-Geithman et al, 2008). Therefore, the evidence supports the idea that BOXes are important in the development of delayed cerebral vasospasm; however, the effectors or molecular targets of BOXes are only beginning to be revealed. Thus far, protein kinase C and the small GTPase Rho A have been suggested as direct or indirect effectors of BOXes (Pyne-Geithman et al, 2008). It is not known whether BOXes affect other cell constituents.

Figure 1

BOXes progressively inhibit Slo1 currents in a poorly reversible manner. (A) Structures of BOX A, BOX B, and 4-methyl-3-vinylmaleimide (MVM). (B) Representative native BK openings recorded in response to pulses from 0 to 50 mV (top) or 20 mV (bottom) from a mouse cerebral artery myocyte in the inside-out configuration. The openings were identified as Slo1 channel openings based on their large unitary current size, Ca²⁺ sensitivity, and K⁺ selectivity. Similar inhibitory effects were observed in three experiments. In each panel, a voltage pulse pattern and a representative current sweep are shown. (C) Inhibition of heterologously expressed human Slo1 channels by BOXes. Representative Slo1 current sweeps measured before (a), during treatment with BOXes (b), and after washout (c) (top), and a representative time course of Slo1 current inhibition by BOXes (bottom). Currents were elicited by pulses to 160 mV and then to −80 mV from the holding voltage of 0 mV in the inside-out configuration. At 160 mV, G/G_max is ∼0.5. BOXes (10 μmol/L assuming the average molecular weight of 200 g/mol) were applied to the cytoplasmic side for 400 seconds and then washed out. Peak outward current amplitudes at different times were scaled to the average peak outward current size before the application of BOXes and plotted as a function of time. The smooth curve represents an exponential fit to the data with a time constant of 79.5 ± 2.3 seconds and the fractional steady-state level of 0.27 ± 0.004. (D) Mean time courses of the fractional current inhibition by four different concentrations of BOXes (0.1, 1, 10, and 100 μmol/L). The currents were elicited as described in panel C. The shaded areas represent s.e.m. n = 3 to 8. (E) Sunlight-exposed BOXes are not effective in inhibiting the Slo1 channel current. Mean scaled current size in the presence of sunlight-exposed BOXes (10 μmol/L). Currents were elicited as described in panel C. The shaded area represents s.e.m. n = 4. The sample was exposed to sunlight for > 6 hours. (F) Application of BOXes (10 μmol/L) to the extracellular side is also effective. The currents in the outside-out configuration were elicited by pulses to 100 mV. n = 3. The intracellular solution contained no Ca²⁺. BOXes, bilirubin oxidation end products.

The importance of Slo1 BK channels in the regulation of vascular smooth muscle tone (Nelson et al, 1995) and the vasoconstrictive ability of BOXes (Clark et al, 2002; Kranc et al, 2000; Pyne-Geithman et al, 2008) together raise the possibility that BOXes may regulate functional properties of Slo1 BK channels. If BOXes inhibit the activity of Slo1 BK channels in vascular smooth muscle cells, the inhibition could promote vascular constriction. In this study, we therefore examined whether BOXes alter functional properties of Slo1 BK channels using electrophysiological and molecular biological methods.

Materials and methods

Human large-conductance Ca²⁺- and voltage-gated Slo1 K⁺ channels (U11058/AAB65837) were heterologously expressed in human embryonic kidney (HEK) cells using FuGene 6 (Roche, Indianapolis, IN, USA) (Horrigan et al, 2005). In some experiments, Slo1 was cotransfected with mouse β1 (AAH13338) or human β1 (AAS20193) (1:1 weight ratio). Other constructs used are noted in the figure legends. Mutations were introduced to Slo1 using a standard PCR-based mutagenesis method (Agilent, Santa Clara, CA, USA) and the sequences were verified (University of Pennsylvania DNA Sequencing Facility, Philadelphia, PA, USA).

Basilar arteries and their main branches were collected from ∼2-month-old C57BL/6 mice using the protocol of Jaggar (2001), except that dithiothreitol replaced dithioerythritol. The animals were killed according to a protocol approved by the University of Pennsylvania Animal Care and Use Committee. Smooth muscle cells were identified by their elongated appearance.

Electrophysiological experiments were performed and the results were analyzed as described previously (Horrigan et al, 2005). The inside-out configuration of the patch-clamp method was used primarily to record ionic currents at room temperature using an Axopatch 200A or 200B amplifier (Molecular Devices, Sunnyvale, CA, USA). The output of the amplifier was digitized (ITC16, HEKA, Bellmore, NY, USA) and stored for later analysis. For macroscopic current measurements, the recording pipettes made of borosilicate glass (Warner, Hamden, CT, USA) had a typical resistance of 0.8 to 1 Mohms when filled with the solution described below and ≥ 50% of the initial pipette resistance was electronically compensated. Native BK channel currents were recorded with the extracellular solution containing (in mmol/L): 134 NaCl, 6 KCl, 1 MgCl₂, 2 CaCl₂, 10 glucose, 10 HEPES, pH 7.4 with NMG (N-methyl-D-glucamine). For measurements of heterologously expressed channels, the extracellular solution contained (in mmol/L): 140 KCl, 2 MgCl₂, 10 HEPES, pH 7.2 (NMG). The ‘zero’ Ca²⁺ intracellular solution contained (in mmol/L): 140 KCl, 11 EGTA, 10 HEPES, pH 7.2 (NMG). The solutions with [Ca²⁺]_i ≥100 μmol/L did not contain any exogenous chelator. Macroscopic capacitative and leak currents were subtracted using a P/6 protocol. Bilirubin oxidation end products were perfused to the recording chamber (∼180 μL in volume) only after currents were stable.

The electrophysiological results were analyzed using custom routines implemented in IgorPro (WaveMetrics, Lake Oswego, OR, USA). To generate macroscopic conductance-voltage (GV) curves, tail currents, excluding the initial ∼180 microseconds to avoid capacitative transients, at −80 mV following prepulses to different voltages were fitted with a single exponential function, and the extrapolated amplitudes at t = 0 second were in turn fitted as a function of voltage with a simple Boltzmann equation as a data descriptor. Each GV curve was scaled using the maximal conductance estimated from the extrapolated tail current amplitudes following prepulses to very depolarized voltages (G_max; typically at ≥230 mV), so that each data set was described by the equation:

\frac{G (V)}{G_{max}} = \frac{1}{1 + e^{[\frac{- Q_{app} F (V - V_{0.5})}{R T}]}}

where Q_app (in e₀) describes the steepness of the curve, V_0.5 (half-activation voltage) represents the voltage at which the scaled macroscopic conductance is 0.5, Vis the voltage, F Faraday constant, R gas constant, and T temperature. Macroscopic current kinetics was fitted with a single exponential, excluding the initial ∼ 180 microseconds at ≤ 100 mV and 250 microseconds at ≥140 mV. The voltage dependence of the time constant was in turn fitted with the equation τ(V) = τ₀exp(qFV/RT), where τ(V) is the time constant at voltage V, τ₀ the time constant value at 0 mV, q (in e₀) the partial charge value describing the steepness of the voltage dependence, F Faraday constant, V voltage, R gas constant, and T the temperature.

Single-channel open probability at negative voltages was estimated from all-point amplitude histograms using the half-amplitude criterion and corrected for the number of channels present in each patch. The number of channels was determined from the peak macroscopic current size recorded in the same patch at ≥160 mV before application of BOXes. The single-channel current–voltage curve (i(V)) and the voltage dependence of absolute open probability (P_o(V)) of the wild-type channel measured in our previous study (Horrigan et al, 2005) determined the number of channels present using the equation I(V) = N·P_o(V) i(V), where I(V) is the peak macroscopic current size at voltage V and N the number of channels.

The statistical results are presented as mean ± s.e.m. (n). Statistical evaluations, typically paired Student's t-test, were performed using IgorPro or DataDesk (DataDescription, Ithaca, NY, USA). Multiple comparisons of means were performed by ANOVA (analysis of variance), followed by the Bonferroni test with DataDesk. The equation fit parameter values are presented as mean ± 95% confidence interval as implemented in IgorPro.

Bilirubin oxidation end products were synthesized and purified according to the protocol of Kranc et al (2000). Three members are present in the samples of BOXes used, namely BOX A, BOX B, and MVM (4-methyl-3-vinylmaleimide) (Kranc et al, 2000). Separation of BOXes to the individual components is difficult and drastically decreases the yield to <5% (Kranc et al, 2000), and the mixture containing all three components was used in the experiments presented herein because this is the ratio (2:2:1) found in CSF (BOX A, BOX B, and MVM, respectively). However, the preparations of BOXes were essentially free of bilirubin and other bilirubin derivatives. Bilirubin oxidation end products were applied with minimal illumination, typically to the cytoplasmic side and washed out. The effects of BOXes were slow to reverse (see below) and most of the results shown are comparisons of those obtained before and after treatment with BOXes followed by a washout. When the kinetics of the inhibitory effect of BOXes was examined, repeated pulses to a single voltage, at which the open probability was ∼0.5, were applied.

Simulations using the Horrigan and Aldrich (HA) model of gating of the Slo1 channel were performed as described previously (Horrigan et al, 2005). To simulate the results in the absence of Ca²⁺, the kinetic model (Figure 5A, bottom; Horrigan et al, 1999) was used to generate ionic currents with the parameter values shown in Table 1. The simulated ionic currents were then analyzed as described above to generate GV curves and the voltage dependence of the macroscopic time constant. The GV curves at different concentrations of Ca²⁺ were analytically simulated (Figure 5A, top) using a custom routine implemented in IgorPro (Horrigan et al, 2005).

Table 1

HA model parameter values

	Control	After BOXes	% Change
α	1,500/s	1,500/s	0
β	35,370/s	35,370/s	0
q_α	0.275 e₀	0.275 e₀	0
q_β	−0.275 e₀	−0.275 e₀	0
J = α(0)/β(0)	0.04241	0.04241	0
δ₀	0.11/s	0.0117 ± 0.006/s	−90 ± 5
δ₁	1.045/s	0.111 ± 0.05/s	−90 ± 5
δ₂	4/s	0.42 ± 0.2/s	−90 ± 5
δ₃	22/s	2.3 ± 1.1/s	−90 ± 5
δ₄	27/s	7.6 ± 0.7/s	−72 ± 2.5
q_δ0–4	0.3	0.3	0
q_δ0–4	−0.17	−0.17	0
L₀ = δ₀(0)/γ₀(0)	5 × 10⁻⁵	5.3 × 10⁻⁶ ± 2.5 × 10⁻⁶	−90 ± 5
D	9.8	9.8	0
C	8	8	0
E	2.4	2.4	0
K _D	11 μmol/L	11 μmol/L	0

BOXes, bilirubin oxidation end products; HA model, Horrigan and Aldrich model.

Changes in the HA model parameter values. The control values are taken from the study by Horrigan et al (2005), except that the constant D was lowered to 9.8 to better fit the data set in this study. The errors associated with L₀ and δ_0–3 were taken from the P_o measurements at −150 mV (see Figure 3) and that with δ₄ was estimated from the activation time constant at 230 mV (see Figure 2).

Figure 2

BOXes alter gating properties of Slo1 in the absence of Ca²⁺. (A) Representative Slo1 currents recorded at different voltages before and after treatment with BOXes. The channels were treated with BOXes (10 μmol/L) for 6 minutes and washed. (B) Peak IV curves before and after treatment with BOXes. (C) Representative normalized GV curves estimated from tail currents before and after treatment with BOXes. The results were obtained from the same patch as shown in panel A. (D) Mean GV parameters, V_0.5 (top) and Q_app (bottom), before and after treatment with BOXes; n = 6. (E) Scaled representative Slo1 currents recorded at different voltages before (thin sweeps) and after (thick sweeps) treatment with BOXes (10 μmol/L for 6 minutes) to illustrate the differential effects on deactivation and activation. (F) Voltage dependence of macroscopic Slo1 current kinetics before (open circles) and after (filled circles) treatment with BOXes. The time constant values at ≥170 mV were fitted with an exponential function. The time constant at 0 mV and the partial charge in the control group were 53 ± 21 milliseconds and 0.3 ± 0.05 e₀, respectively. The respective values after treatment with BOXes were 63 ± 0.2 milliseconds and 0.2 ± 0.0001 e₀. n = 7. BOXes, bilirubin oxidation end products.

Figure 3

BOXes decrease P_o at extreme negative voltages but do not change the mean open duration in the absence of Ca²⁺. (A) Representative Slo1 openings at 50 mV (top) and −150 mV (bottom) before (left) and after (right) treatment with BOXes (10 μmol/L for 5 minutes). The openings are upward at 50 mV and downward at −150 mV. The data at + 50 mV are from a patch with ∼ 50 channels and those at −150 mV are from another patch with ∼1,000 channels. (B) Voltage dependence of P_o was calculated from single-channel records before (open circles) and after (filled circles) treatment with BOXes (10 μmol/L for 5 to 10 minutes). The P_o values were determined using the macroscopic currents measured at positive voltages in each patch. (C) Mean open durations before (open circles) and after (filled circles) treatment with BOXes (10 μmol/L for 5 to 10 minutes). n = 4 to 5. BOXes, bilirubin oxidation end products.

Figure 4

Slo1 channels treated with BOXes remain Ca²⁺ sensitive. (A) Representative Slo1 macroscopic currents at 140 mV with the concentrations of Ca²⁺ indicated after treatment with BOXes (10 μmol/L for 6 minutes). (B) GV curves at different [Ca²⁺]_i from one representative patch. The control GV curve obtained in the absence of Ca²⁺ is shown using open circles, and those obtained after treatment with BOXes (10 μmol/L for 6 minutes) are shown using filled symbols. The numbers indicate [Ca²⁺]_i in μmol/L. The smooth curves represent Boltzmann fits. The V_0.5 values for the curves shown are, from left to right, 73, 119, 169, and 197 mV. The Q_app values for the curves show are 1.2, 1.2, 1.1, and 0.98 e₀. (C) Ca²⁺ dependence of V_0.5 before (open circles) and after treatment with BOXes (filled circles). (D) Changes in V_0.5 before (open circles) and after treatment with BOXes (filled circles) by different [Ca²⁺]_i normalized to the respective values without Ca²⁺. The two values at 200 μmol/L are nearly identical and the symbols overlap. n = 5. BOXes, bilirubin oxidation end products.

Figure 5

Simulations using the HA model. (A) The equilibrium HA model of Slo1 channel gating (top) and the HA model kinetic scheme in the absence of Ca²⁺ (HCA model; bottom). The ion conduction gate of the tetrameric Slo1 channel can be closed (C) or open (O) and the equilibrium is described by the equilibrium constant L. One channel contains four voltage sensor domains and each voltage sensor can be at rest (R) at negative voltages or activated (A) at depolarized voltages. The voltage-dependent equilibrium is described by the equilibrium constant J. For simplicity, each subunit is assumed to have a single Ca²⁺-sensor domain, which can be unoccupied (X) or occupied with a Ca²⁺ ion (X Ca²⁺). The equilibrium is described by the equilibrium constant K. D, C, and E describe the allosteric interactions among the gate, voltage sensors, and Ca²⁺ sensor domains, respectively. For each voltage sensor activated, the equilibrium constant L increases by D-fold (LD), thus leading to an increase in open probability. When all four voltage sensors are activated, the effective equilibrium constant is LD⁴. Similarly, activation of one Ca²⁺-sensor domain increases L by C-fold (LC). When all four Ca²⁺ sensors are occupied, L increases by C⁴-fold (LC⁴). Activation of all four voltage-sensor domains and of all four Ca²⁺-sensor domains together increases L by D⁴C⁴-fold (D⁴C⁴). The bottom panel shows the gating model in the absence of Ca²⁺ using rate constants. C⁰, C¹, C², C³, and C⁴ refer to the closed states with 0, 1, 2, 3, and 4 voltage sensors activated, respectively. O⁰, O¹, O², O³, and O⁴ represent the open states with 0, 1, 2, 3, and 4 voltage sensors activated, respectively. The ratio δ₀/γ₀ defines the equilibrium constant L and the ratio α₀/β₀ at 0 mV defines the equilibrium constant J. f² = D. The model parameter values are given in Table 1. (B) Macroscopic GV curves in the absence of Ca²⁺ simulated by the HA model. (C) Voltage dependence of kinetics of simulated macroscopic currents in the absence of Ca²⁺. (D) Ca²⁺ dependence of V_0.5. In each graph, the results simulated for the control condition (thin sweep) and those after treatment with BOXes (thick sweep) are shown. The shaded areas represent s.e.m. based on the experimental measurements. BOXes, bilirubin oxidation end products; HA model, Horrigan and Aldrich model.

Chloramine-T, DEPC (diethyl pyrocarbonate), PGO (phenylglyoxal), dimethyl adipimidate (DMA), and iodoacetamide were obtained from Sigma (St Louis, MO, USA). Both MTSEA (2-aminoethyl methanethiosulfonate) and MTSET (2-(trimethylammonium) ethyl methanethiosulfonate) were acquired from Biotium (Hayward, CA, USA). The reagents were prepared immediately before use and applied to the intracellular side. The reagents were washed out of the chamber and the sensitivity to BOXes was examined.

Results

Application of BOXes (Figure 1A; 10 μmol/L) to the cytoplasmic side of a membrane patch excised from a mouse cerebral artery myocyte markedly decreased the frequency of Slo1 BK channel openings (Figure 1B). To elucidate the biophysical and molecular mechanisms of the inhibitory action of BOXes in detail, we heterologously expressed human Slo1 channels in HEK cells. The opening and the closing of the ion conduction gate of the Slo1 channel are allosterically controlled by its transmembrane voltage and cytoplasmic Ca²⁺ sensors (Figure 5A, top), and a full description of the voltage- and Ca²⁺-dependent gating of the channel requires a large number of conformational states (Horrigan and Aldrich, 2002). To simplify interpretations of the electrophysiological results, our initial experiments were performed in the virtual absence of Ca²⁺, such that the channel opening is allosterically controlled only by the voltage sensors (Horrigan et al, 1999; Figure 5A). As found with the native myocyte BK channels (Figure 1B), application of BOXes (10 μmol/L) to the cytoplasmic side of an excised cell-free membrane patch containing Slo1 channels progressively decreased the current elicited by depolarization roughly following an exponential time course with a time constant value of ∼80 seconds (Figure 1C). Extensive wash, up to ∼10 minutes, did not fully reverse the inhibition (Figure 1C), indicating a stable interaction between BOXes and the channel protein. Similar measurements in multiple experiments showed that the kinetics of inhibition became faster with greater concentrations of BOXes (Figure 1D). For ≥10 μmol/L, the steady-state fractional inhibitions observed within a typical lifetime of a patch (∼15 minutes) were similar. Lower concentrations of BOXes (0.1 and 1 μmol/L) appeared less effective (Figure 1D), but it was not clear whether the inhibition reached steady state within the limited measurement duration. The inhibition time courses with BOXes were indistinguishable whether the holding voltage was 0 or −100 mV (data not shown).

A previous study has identified that one of the notable characteristics of BOXes is that their vasoconstrictive ability is diminished by exposure to sunlight (Kranc et al, 2000). Consistent with this report, application of sunlight-exposed BOXes was much less effective in altering the current size (Figure 1E).

Application of BOXes to the extracellular side was also effective (Figure 1F), but the inhibition kinetics was often slower than that observed with application of BOXes to the intracellular side. The faster action of BOXes applied to the cytoplasmic side may signify that the sites on the channel required for the interaction with BOXes are more readily accessed from the intracellular side. The results presented in the remainder of this article were obtained using application of BOXes to the intracellular side using cell-free inside-out membrane patches. Bilirubin oxidation end products were typically applied at 10 μmol/L, because the inhibition kinetics with this concentration was rapid enough to obtain sufficient results using the patch-clamp method.

Single-channel current–voltage curves before and after treatment with BOXes (10 μmol/L for 10 minutes) were indistinguishable, showing that BOXes had no effect on the open-channel ion-conduction properties (Supplementary Figure 1). In contrast, voltage-dependent steady-state activation of the Slo1 channel was markedly altered by BOXes (Figure 2A). The Slo1 currents were inhibited more noticeably at less depolarized voltages (e.g., 120 versus 200mV in Figure 2B). The inhibition was accompanied by a ∼35 mV shift in half-activation voltage (V_0.5) of macroscopic conductance (GV) to the positive direction (Figures 2C and 2D; 160 ± 4 mV versus 194 ± 3mV; P< 0.0001). The slope of the GV curve was also shallower; the apparent charge movement estimated from Boltzmann fits decreased by ∼15% from ∼1.10 ± 0.04 to 0.96 ± 0.04 e₀ (P < 0.005; Figure 2D).

Treatment with BOXes also altered the kinetics of the channel. Bilirubin oxidation end products noticeably slowed the activation kinetics at positive voltages at which the opening transitions dominate and open probability (P_o) is high (Figures 2E and 2F). For example, at 230 mV, the activation time constant increased by 3.60 ± 0.38-fold (Figure 2F). In addition, after treatment with BOXes, the activation time constant was less voltage dependent, corresponding to a ∼ 30% decrease in the partial charge movement (Figure 2F). In contrast, at negative voltages at which the closing transitions dominate and P_o is low (≤0 mV), treatment with BOXes failed to affect the deactivation kinetics of the channel (Figures 2E and 2F).

At extreme negative voltages (≤ −100 mV), the primary voltage sensors of the Slo1 channel largely assume the resting conformation (Figure 5A) (Horrigan et al, 1999). As voltage-sensor activation is not strictly required for the ion conduction gate of Slo1 to open, single-channel currents can be observed at such extreme negative voltages, although infrequently (Figures 3A and 5A) (Horrigan et al, 1999). Treatment with BOXes decreased P_o at such negative voltages (Figures 3A and 3B); at −150 mV, P_o decreased to 10.6% ± 5.0% of the control level. In contrast with the marked decreases in P_o, the mean open durations determined by the closing transition(s) (Figure 5A, bottom) remained unaltered by BOXes at these negative voltages (Figure 3C).

The results presented thus far show that treatment with BOXes leads to marked alterations in the functional properties of the Slo1 BK channels in the absence of Ca²⁺, reflecting the reciprocal and allosteric interactions between the gate and voltage sensors of the channel (Figure 5A). As physiologic activation of the Slo1 BK channel often involves intracellular Ca²⁺ (Nelson et al, 1995), we examined whether BOXes altered the Ca²⁺ sensitivity of the channel. After treatment with BOXes (10 μmol/L), Slo1 currents remained dependent on [Ca²⁺]_i, increasing in size with higher concentrations (Figures 4A and 4B). At a given Ca²⁺ concentration, BOXes shifted the voltage dependence of activation as measured by V_0.5 to the positive direction by 30 to 50 mV (Figure 4C). However, when normalized to the respective V_0.5 values in the absence of Ca²⁺ (Figure 4D), the control channels and the channels treated with BOXes had similar Ca²⁺ dependence of activation. For example, in both groups, 1 μmol/L C^a2+ caused a −70 mV shift of V_0.5, and 200 μmol/L Ca²⁺, a near saturating concentration for the high-affinity Ca²⁺ sensors of the Slo1 channel (Yang et al, 2008), caused a −200 mV shift in V_0.5 (Figure 4D).

Steady-state activation properties of the Slo1 BK channel are described by the model of Horrigan and Aldrich (2002) (HA model), and the kinetic characteristics in the absence of Ca²⁺ are well approximated by a subset of the HA model (HCA model; Horrigan et al, 1999) as depicted in Figure 5A. To gain biophysical insight, the HA model was used to simulate the action of BOXes on the Slo1 channel gating by varying the model parameter values in the following manner. Without Ca²⁺, at extreme negative voltages at which voltage-sensor activation is negligible, P_o is largely determined by the rate constants δ₀ and γ₀ (Figure 5A, bottom). The ratio δ₀/γ₀ defines the equilibrium constant L in the HA model (Figure 5A, top). The closing rate constant γ₀ is the primary determinant of the mean open duration and of the macroscopic deactivation kinetics at negative voltages. As treatment with BOXes decreased P_o by ∼10-fold but failed to alter the mean open duration or the deactivation kinetics at negative voltages, it is reasoned that BOXes decrease the equilibrium constant L0 defined as δ₀/γ₀ at 0 mV by ∼10-fold by decreasing the opening rate constant δ0 without altering the closing rate constant δ0. The 10-fold decrease in δ0 necessitates the same fractional changes in δ_1–4 to maintain microscopic reversibility (Figure 5A, bottom), including the opening rate constant δ4, which dominates the activation kinetics at positive voltages. Experimentally, however, only an approximately three-fold increase in the activation kinetics at positive voltages was observed after treatment with BOXes (Figure 2); the decrease in δ4 alone was constrained accordingly to simulate the kinetic changes at positive voltages. The changes in the parameter values are summarized in Table 1. With these changes incorporated, the model predicts the kinetic and steady-state properties that are in general agreement with the experimental results (Figures 5B–5D).

To explore which amino-acid residues of the Slo1 protein are required for the inhibitory effect of BOXes, we pretreated the channels with different amino-acid-modifying reagents and then, after washout, assessed the sensitivity of the modifier-treated channels to BOXes. Pretreatment with the methionine-preferring modifier chloramine-T, the imidazole modifier DEPC, the arginine modifier PGO, the cysteine modifiers iodoacetamide, MTSEA, and MTSET failed to alter the sensitivity of the channel to BOXes (data not shown). In contrast, pretreatment with the lysine modifier DMA (Duszyk et al, 1998) noticeably decreased the sensitivity to BOXes (Figures 6A–6D). The DMA pretreatment itself often decreased currents; however, the DMA-treated channels were clearly less sensitive to BOXes (Figures 6A versus 6B). Bilirubin oxidation end products (10 μmol/L for 400 seconds) decreased the current size only by 25% in DMA-treated channels compared with the ∼75% reduction observed in nontreated channels (Figures 6B and 6D).

Figure 6

Lysine residues may be important in the action of BOXes. (A) Sample current sweeps and a representative time course of inhibition of wild-type Slo1 channels by BOXes (10 μmol/L) at 160 mV. (B) Sample current sweeps and a representative inhibition time course at 160 mV in Slo1 channels pretreated with DMA. The patch was treated with freshly prepared DMA (10 μmol/L) in the recording solution for 10 minutes and washed for 10 minutes. (C) Sample current sweeps and a representative inhibition time course in the R329A:K330A:K330A channel (‘RKK mutant’) at 70mV. The voltage dependence of the RKK mutant is shifted to the negative direction by ∼100mV and G/G_max is ∼0.5 at 70mV. In panels A, B, and C, representative currents before and after treatment with BOXes are shown (top). (D) Fractional currents remaining in the wild-type channel, the wild-type channel treated with DMA and the RKK mutant after treatment with BOXes. The results of the DMA-treated and RKK mutant groups are significantly different from those in the control wild-type group (P≤0.0017, Bonferroni, n = 8 to 10). (E) Structural organization of a Slo1 subunit (left) and a cartoon representation of a tetrameric Slo1 channel viewed from the cytoplasmic side based loosely on the electron cryomicroscopy structure by Wang and Sigworth (2009). BOXes, bilirubin oxidation end products; DMA, dimethyl adipimidate.

The ability of the lysine-modifying agent DMA to antagonize the inhibitory effect of BOXes suggests an involvement of Lys residues. Each Slo1 subunit (U11058/AAB65837) possesses 59 Lys residues and the majority of them are located in the large cytoplasmic C-terminal area. Of the numerous Lys residues, we focused on those in the sequence ³²⁹RKK³³¹ in the linker segment between the transmembrane segment S6 and the cytoplasmic RCK1 domain (Figure 6E). This sequence has been implicated in the direct action of phosphatidylinositol 4,5-bisphosphate (PIP2); the triple mutation R329A:K330A:K331A (referred to as the RKK mutation hereafter) has been reported to diminish the stimulatory effect of PIP2 by 50% (Vaithianathan et al, 2008). Prompted by the purported importance of the ³²⁹RKK³³¹ sequence, we tested whether the sensitivity of the channel to BOXes was also altered by the mutation. Indeed, the RKK mutation significantly diminished the sensitivity of the channel to BOXes (Figures 6C and 6D). After treatment with BOXes (10 μmol/L), ∼60% of the current remained in the RKK mutant compared with ∼30% remaining in the wild-type nontreated channel (P≤0.0017). The fractional currents remaining in the RKK mutant channel and the DMA-treated wild-type channel were indistinguishable (P = 0.086). The RKK mutant channel was also largely insensitive to treatment with DMA (data not shown).

The transmembrane segments and the cytoplasmic RCK1 domains of the Slo1 channel together coordinate Mg²⁺ ions, which also promote the opening of the channel (Yang et al, 2008). As the sequence ³²⁹RKK³³¹ is located in the S6-RCK1 linker segment (Figure 6E), we examined whether treatment with BOXes altered the Mg²⁺ sensitivity of the channel. Intracellular Mg²⁺ (10 μmol/L) produced a ∼100mV shift in V_0.5 to the negative direction in channels treated with BOXes (Supplementary Figure 3), which was indistinguishable from that in the control group.

Native vascular BK channel complexes include both the pore-forming Slo1 and auxiliary β1 subunits (Salkoff et al, 2006). Consistent with the observation that BOXes inhibited native myocyte BK channels (Figure 1B), BOXes remained effective when Slo1 and β1 were heterologously coexpressed together (Supplementary Figure 3). In addition to Slo1 BK channels, vascular smooth muscle cells contain other K⁺ channels, including Kv1.5 (Brevnova et al, 2004). Bilirubin oxidation end products (10 μmol/L) did not alter currents through human Kv1.5 channels expressed in HEK cells (Supplementary Figure 4).

Discussion

Bilirubin oxidation end products, heme breakdown products, are found in the CSF of SAH patients and have been implicated in the development of delayed cerebral vasospasm following SAH (Clark and Sharp, 2006; Clark et al, 2008; Kranc et al, 2000). Although the molecular and cellular targets of BOXes are yet to be fully revealed, a recent study has shown that BOXes exert their action in part by regulating protein kinase C and GTPase Rho A (Pyne-Geithman et al, 2008). Our results here show that BOXes profoundly inhibit the Slo1 BK channel activity in a lysine-dependent manner and suggest that the inhibitory effect of BOXes on the channel could contribute to the vasoconstrictive effect of BOXes.

Gating of the Slo1 channel involves reciprocal and allosteric interactions among the ion conduction gate of the channel, transmembrane voltage sensors, and cytoplasmic ligand-binding domains (Horrigan et al, 1999; Horrigan and Aldrich, 2002). After treatment with BOXes, the voltage dependence of activation of the channel is shallower and shifted to the positive direction, and the activation kinetics is also slower. These changes in gating can be readily accounted for by postulating that BOXes specifically slow the opening process of the ion conduction gate of the channel. A 70% to 90% decrease in each of the four opening rate constants (δ_0–4) in the HA model of Slo1 gating provides a good approximation of the channel's gating after exposure to BOXes.

The inhibitory effect of BOXes is somewhat reminiscent of that of heme, which also alters the Slo1 channel activity (Tang et al, 2003). Heme and its oxidized form hemin inhibit the channel at positive voltages in part by shifting the voltage dependence of activation to the positive direction (Tang et al, 2003). In addition, the activation kinetics is noticeably slower in the presence of heme/hemin (Horrigan et al, 2005). Despite the superficial resemblances, the underlying mechanisms of BOXes and hemin are markedly different. Bilirubin oxidation end products decrease the opening rate constant of the channel's gate (δ₀) as shown in this study, whereas hemin increases the same rate constant (δ₀) and weakens the allosteric coupling between the voltage sensors and the gate (D) (Horrigan et al, 2005) (see Figure 5A). The differential effects of hemin and BOXes are particularly evident at negative voltages at which the primary voltage sensors of the channel are at rest; hemin increases but BOXes decrease P_o, suggesting that hemin and BOXes could have opposing physiologic effects.

The different biophysical mechanisms of the actions of hemin and BOXes on the Slo1 channel are reflected in the distinct amino-acid residues of Slo1 important for the actions of these related molecules. A short sequence containing two Cys and a His residue located in the cytoplasmic linker segment between the RCK1 and RCK2 domains of the Slo1 channel has an important role in the sensitivity of the channel to hemin (Jaggar et al, 2005; Tang et al, 2003). In contrast, the full sensitivity of the channel to BOXes requires the sequence ³²⁹RKK³³¹ located downstream of the S6 transmembrane segment in the primary sequence.

The sequence ³²⁹RKK³³¹ is also important in the direct action of the membrane-bound signaling molecule PIP₂ on the channel; contemporaneous mutation of the three residues to alanine has been reported to diminish the effect of PIP₂ by 50% (Vaithianathan et al, 2008). Given the putative location of the sequence ³²⁹RKK³³¹ downstream of the S6 segment, it may be speculated that BOXes position themselves between the transmembrane helices and the cytoplasmic RCK1 domain, and stabilize the closed conformation of the ion conduction gate. The interaction of BOXes with the channel may be further stabilized by the probable ability of hydrophobic and planar BOXes to partition well in membranes and/or to intercalate nonpolar areas of the channel protein, as suggested by the finding that BOXes from either side of the membrane are effective. The relatively hydrophobic characteristic of BOXes should also contribute to the poor reversibility by simple washing. These membrane-permeable and light-sensitive characteristics of BOXes somewhat resemble those of the voltage-gated Ca²⁺ channel modulators dihydropyridines (Meyer et al, 1984). The diminished inhibitory effect of the BOXes following sunlight exposure is consistent with an earlier report (Kranc et al, 2000). Such light exposure has been postulated to photooxidize BOX A and BOX B to MVM (Clark and Sharp, 2006). Thus, the amide group, present in BOX A and BOX B but absent in MVM (Figure 1A), may be required to inhibit the Slo1 BK channel.

In addition to the sequence ³²⁹RKK³³¹, there may be other determinants of the inhibitory effect of BOXes because the RKK mutation leaves ∼35% to 40% of the inhibition intact. Mechanistically, the sequence ³²⁹RKK³³¹ in the S6-RCK1 linker could function as a binding site component. Alternatively, the RKK mutation may alter the accessibility of BOXes to the binding sites located elsewhere. Our results do not directly distinguish these two possibilities.

The exact structural location of the ion conduction gate in the Slo1 channel is yet to be revealed, and consequently, elucidation of the physicochemical basis of the preferential closed-state stabilization by BOXes requires further investigation. The linker connecting the S6 segment and the cytoplasmic RCK1 domain, which includes the sequence ³²⁹RKK³³¹, has been envisioned to work like a passive mechanical spring to influence voltage-dependent gating of the Slo1 channel (Niu et al, 2004). However, it may be noted that our results do not show that BOXes alter the strongly voltage-dependent transitions mediated by the transmembrane voltage sensors.

Multiple types of K⁺ channels are expressed in vascular smooth muscle cells and alterations in any one of the channels could contribute to vascular dysfunction, including delayed cerebral vasospasm (Wellman, 2006). A transcript profiling study using a canine model of SAH involving autologous blood injection documented the downregulation of Kv2.2 and upregulation of Kir2.1 but no change in the expression of Slo1 or Cav1.2 (Aihara et al, 2003). Using the same canine model of SAH, Jahromi et al (2008) showed that the BK channel activity in acutely dissociated basilar artery smooth muscle cells is unaltered; however, the concentrations of BOXes were not measured in the study. How these findings using native cells harvested from the experimental SAH animals relate to our finding that BOXes regulate Slo1 is not clear. It is possible that the inhibitory effect of BOXes was somehow removed during the cell dissociation process, or that native BK channels possess or are exposed to cellular factors that protect them from BOXes or facilitate their unbinding.

SAH leads to the generation of multiple hemoglobin breakdown products, the concentrations of which in the CSF change with time. One of the early breakdown products is oxyhemoglobin, which has been suggested to have a role in SAH-induced cerebral vasospasm. The role played by oxyhemoglobin, however, may, some argue, be limited to early vasoconstrictive effects owing to its relatively short time in the hemolysate and also in part to its ability to scavenge the vasodilator nitric oxide (Windmuller et al, 2005). Thus oxyhemoglobin is an antidilatory agent as opposed to a procontractile agent (Aoki et al, 1994; Windmuller et al, 2005). The role of oxyhemoglobin in vasoconstriction during delayed vasospasm at 3 to 7 days is therefore unclear compared with other chemical species resulting from the hemorrhage. Bilirubin oxidation end products, in contrast, may serve as a procontractile agent (Pyne-Geithman et al, 2008), and the inhibitory effect of BOXes on the Slo1 channel may contribute to the development of delayed cerebral vasospasm following SAH.

Footnotes

The authors declare no conflict of interest.

Notes

Supplementary Information accompanies the paper on the Journal of Cerebral Blood Flow & Metabolism website ()

References

Aihara

Jahromi

Yassari

Nikitina

Agbaje-Williams

Macdonald

(2003) Molecular profile of vascular ion channels after experimental subarachnoid hemorrhage. J Cereb Blood Flow Metab 24:75–83

Amberg

Santana

(2003) Downregulation of the BK channel β1 subunit in genetic hypertension. Circ Res 93:965–71

Aoki

Takenaka

Suzuki

Kassell

Sagher

Lee

(1994) The role of hemolysate in the facilitation of oxyhemoglobin-induced contraction in rabbit basilar arteries. J Neurosurg 81:261–6

Brenner

Perez

Bonev

Eckman

Kosek

Wiler

Patterson

Nelson

Aldrich

(2000) Vasoregulation by the β1 subunit of the calcium-activated potassium channel. Nature 407:870–6

Brevnova

Platoshyn

Zhang

Yuan

(2004) Overexpression of human KCNA5 increases I_K(V) and enhances apoptosis. Am J Physiol Cell Physiol 287:C715–22

Clark

Reilly

Sharp

(2002) Oxidation of bilirubin produces compounds that cause prolonged vasospasm of rat cerebral vessels: a contributor to subarachnoid hemorrhage-induced vasospasm. J Cereb Blood Flow Metab 22:472–8

Clark

Sharp

(2006) Bilirubin oxidation products (BOXes) and their role in cerebral vasospasm after subarachnoid hemorrhage. J Cereb Blood Flow Metab 26:1223–33

Clark

Loftspring

Wurster

Pyne-Geithman

(2008) Chemical and biochemical oxidations in spinal fluid after subarachnoid hemorrhage. Front Biosci 13:1806–12

Bautista

Yang

Diez-Sampedro

You

Wang

Kotagal

Luders

Shi

Cui

Richerson

Wang

(2005) Calcium-sensitive potassium channelopathy in human epilepsy and paroxysmal movement disorder. Nat Genet 37:733–8

10.

Duszyk

Shu

Man

(1998) Activation of bovine tracheal chloride channels by amino group-specific reagents. J Physiol (Lond) 512:129–36

11.

Fernández-Fernández

Tomás

Vázquez

Orio

Latorre

Sentí

Marrugat

Valverde

(2004) Gain-of-function mutation in the KCNMB1 potassium channel subunit is associated with low prevalence of diastolic hypertension. J Clin Invest 113:1032–9

12.

Gribkoff

Starrett

Jr Dworetzky

Hewawasam

Boissard

Cook

Frantz

Heman

Hibbard

Huston

Johnson

Krishnan

Kinney

Lombardo

Meanwell

Molinoff

Myers

Moon

Ortiz

Pajor

Pieschl

Post-Munson

Signor

Srinivas

Taber

Thalody

Trojnacki

Wiener

Yeleswaram

Yeola

(2001) Targeting acute ischemic stroke with a calcium-sensitive opener of maxi-K potassium channels. Nat Med 7:471–7

13.

Horrigan

Aldrich

(2002) Coupling between voltage sensor activation, Ca²⁺ binding and channel opening in large conductance (BK) potassium channels. J Gen Physiol 120:267–305

14.

Horrigan

Cui

Aldrich

(1999) Allosteric voltage gating of potassium channels I. mslo ionic currents in the absence of Ca²⁺. J Gen Physiol 114:277–304

15.

Horrigan

Heinemann

Hoshi

(2005) Heme regulates allosteric activation of the Slo1 BK channel. J Gen Physiol 126:7–21

16.

Jaggar

(2001) Intravascular pressure regulates local and global Ca²⁺ signaling in cerebral artery smooth muscle cells. Am J Physiol Cell Physiol 281:C439–48

17.

Jaggar

Parfenova

Liu

Umstot

Dopico

Leffler

(2005) Heme is a carbon monoxide receptor for large-conductance Ca²⁺-activated K⁺ channels. Circ Res 97:805–12

18.

Jahromi

Aihara

Zhang

Weyer

Nikitina

Yassari

Houamed

Macdonald

(2008) Preserved BK channel function in vasospastic myocytes from a dog model of subarachnoid hemorrhage. J Vasc Res 45:402–15

19.

Kranc

Pyne

Tao

Claridge

Harris

Cadoux-Hudson

Turnbull

Schofield

Clark

(2000) Oxidative degradation of bilirubin produces vasoactive compounds. Eur J Biochem 267:7094–101

20.

Macdonald

Pluta

Zhang

(2007) Cerebral vasospasm after subarachnoid hemorrhage: the emerging revolution. Nat Clin Pract Neurol 3:256–63

21.

Macdonald

Weir

(1991) A review of hemoglobin and the pathogenesis of cerebral vasospasm. Stroke 22:971–82

22.

Meredith

Thorneloe

Werner

Nelson

Aldrich

(2004) Overactive bladder and incontinence in the absence of the BK large conductance Ca²⁺-activated K⁺ channel. J Biol Chem 279:36746–52

23.

Meyer

Bossert

Wehinger

Scherling

(1984) Chemistry of nitrendipine and its metabolites. In: Nitrendipine. ( Scriabine

Vanov

Deck

, eds). Baltimore: Urban & Schwarzenberg, 1–9

24.

Nelson

Cheng

Rubart

Santana

Bonev

Knot

Lederer

(1995) Relaxation of arterial smooth muscle by calcium sparks. Science 270:633–7

25.

Niu

Qian

Magleby

(2004) Linker-gating ring complex as passive spring and Ca²⁺-dependent machine for a voltage- and Ca²⁺-activated potassium channel. Neuron 42:745–56

26.

Pyne-Geithman

Morgan

Wagner

Dulaney

Carrozzella

Kanter

Zuccarello

Clark

(2005) Bilirubin production and oxidation in CSF of patients with cerebral vasospasm after subarachnoid hemorrhage. J Cereb Blood Flow Metab 25:1070–7

27.

Pyne-Geithman

Nair

Caudell

Clark

(2008) PKC and Rho in vascular smooth muscle: activation by BOXes and SAH CSF. Front Biosci 13:1526–34

28.

Salkoff

Butler

Ferreira

Santi

Wei

(2006) High-conductance potassium channels of the SLO family. Nat Rev Neurosci 7:921–31

29.

Sausbier

Arntz

Bucurenciu

Zhao

Zhou

Sausbier

Feil

Kamm

Essin

Sailer

Abdullah

Krippeit-Drews

Feil

Hofmann

Knaus

Kenyon

Shipston

Storm

Neuhuber

Korth

Schubert

Gollasch

Ruth

(2005) Elevated blood pressure linked to primary hyperaldosteronism and impaired vasodilation in BK channel-deficient mice. Circulation 112:60–8

30.

Sprossmann

Pankert

Sausbier

Wirth

Zhou

Madlung

Zhao

Bucurenciu

Jakob

Lamkemeyer

Neuhuber

Offermanns

Shipston

Korth

Nordheim

Ruth

Sausbier

(2009) Inducible knockout mutagenesis reveals compensatory mechanisms elicited by constitutive BK channel deficiency in overactive murine bladder. FEBS J 276:1680–97

31.

Tang

Reynolds

Garcia

Heinemann

Hoshi

(2003) Haem can bind to and inhibit mammalian calcium-dependent Slo1 BK channels. Nature 425:531–5

32.

Vaithianathan

Bukiya

Liu

Asuncion-Chin

Fan

Dopico

(2008) Direct regulation of BK channels by phosphatidylinositol 4,5-bisphosphate as a novel signaling pathway. J Gen Physiol 132:13–28

33.

Wang

Sigworth

(2009) Structure of the BK potassium channel in a lipid membrane from electron cryomicroscopy. Nature 461:292–5

34.

Wellman

(2006) Ion channels and calcium signaling in cerebral arteries following subarachnoid hemorrhage. Neurol Res 28:690–702

35.

Werner

Zvara

Meredith

Aldrich

Nelson

(2005) Erectile dysfunction in mice lacking the large-conductance calcium-activated potassium (BK) channel. J Physiol (Lond) 567:545–56

36.

Windmuller

Lindauer

Foddis

Einhaupl

Dirnagl

Heinemann

Dreier

(2005) Ion changes in spreading ischaemia induce rat middle cerebral artery constriction in the absence of NO. Brain 128:2042–51

37.

Yang

Shi

Zhang

Yang

Delaloye

Cui

(2008) Activation of Slo1 BK channels by Mg²⁺ coordinated between the voltage sensor and the RCK1 domains. Nat Struct Mol Biol 15:1152–9

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.05 MB

0.07 MB

0.11 MB

0.05 MB

0.03 MB

0.32 MB