Connexin-43 Hemichannels Opened by Metabolic Inhibition*

The cause of altered ionic homeostasis leading to cell death during ischemia and metabolic inhibition is unclear. Hemichannels, which are precursors to gap junctions, are nonselective ion channels that are permeable to molecules of less thanM r 1000. We show that hemichannels open upon exposure to calcium-free solutions when they are either heterologously overexpressed in HEK293 cells or endogenously expressed in cardiac ventricular myocytes. In the presence of normal extracellular calcium, hemichannels open during metabolic inhibition. During ischemia and other forms of metabolic inhibition, activation of relatively few hemichannels will seriously compromise the cell’s ability to maintain ionic homeostasis, which is an essential step promoting cell death.

The cause of altered ionic homeostasis leading to cell death during ischemia and metabolic inhibition is unclear. Hemichannels, which are precursors to gap junctions, are nonselective ion channels that are permeable to molecules of less than M r 1000. We show that hemichannels open upon exposure to calcium-free solutions when they are either heterologously overexpressed in HEK293 cells or endogenously expressed in cardiac ventricular myocytes. In the presence of normal extracellular calcium, hemichannels open during metabolic inhibition. During ischemia and other forms of metabolic inhibition, activation of relatively few hemichannels will seriously compromise the cell's ability to maintain ionic homeostasis, which is an essential step promoting cell death.
Ischemia, hypoxia, and other forms of metabolic inhibition cause rapid disturbances in ionic homeostasis including intracellular Na and Ca gain and K loss that contribute to cellular injury and death (1); however, the mechanism is uncertain. Recent studies using dye uptake assays in immortalized cells have shown that some types of hemichannels open in low extracellular Ca (2). Hemichannels are the precursors to gap junctions (for a review, see Ref. 3); two hemichannels aligned end to end form an intercellular communicating channel. If they are activated in nonjunctional plasma membrane, hemichannels are nonselective and exchange intracellular K for extracellular Na or Ca due to the asymmetrical intracellular/extracellular distribution of these cations.
We have previously described a nonselective current that is activated by metabolic inhibition in isolated ventricular myocytes (4). The purpose of this study was to investigate whether activated nonjunctional hemichannels might be responsible for this current. The findings are positive, indicating that nonjunctional hemichannels may be involved in the pathogenesis of ionic disturbances during myocardial ischemia and hypoxia.

EXPERIMENTAL PROCEDURES
Molecular Biology-Plasmid encoding GFP 1 (CLONTECH, Palo Alto, CA) was fused to the C terminus of Cx43 using the polymerase chain reaction overlap extension method (4,5). The chimeric product was subcloned and sequenced. Cx43 provided by Dr. Bruce Nicholson (State University of New York, Buffalo, NY) was subcloned into pCDNA3. Both native and chimeric constructs used the cytomegalovirus promoter. Transfection was carried out using the calcium phospate method (6).
Microscopy-The expression of Cx43-GFP was determined by examining cell monolayers grown on glass coverslips using a Nikon microscope fitted with a Xenon lamp and the appropriate filters (excitation bandpass, 450 -480 nm; emission cutoff, Ͻ515 nm). Cardiac myocytes were examined for calcein and dextran-fluorescein loading using the same filters. Magnification was ϫ400 in the original slides, which were then scanned and assembled using an Adobe Photoshop/Macintosh G3 combination.
Electrophysiology-Whole cell and single channel currents were recorded with an Axopatch 200A clamp amplifier and a Digidata 1200 data acquisition system using pClamp software (Axon Instruments, Foster City, CA). For the whole cell clamp experiments, the patch pipette contained 140 mM KCl, 1 mM MgCl 2 , 5 mM NaCl, and 10 mM HEPES, pH 7.4, for HEK293 cells or 115 mM cesium glutamate, 30 mM TEA-OH, 10 mM HEPES, 3 mM NaCl, 1 mM MgCl 2 , 1 mM NaH 2 PO 4 , 5 mM sodium pyruvate, and 1 mMNa-ADP for isolated ventricular myocytes. The bath solution contained 140 mM NaCl, 1 mM MgCl 2 , 5.4 mM KCl, 1.8 mM CaCl 2 , and 10 mM HEPES, pH 7.2, with or without 10 mM dextrose for HEK293 cells; CsCl was substituted for KCl for isolated ventricular myocytes. For the 0 Ca solution, CaCl 2 was removed, and 2 mM EGTA was added (estimated free Ca, 1 nM). Metabolic inhibitors (which were dissolved in dimethyl sulfoxide as appropriate) or LaCl 3 was added directly to the bath solution. For halothane, the appropriate bath solution was mixed with halothane and vigorously shaken. Excess halothane was suctioned off, and the halothane-saturated solution was covered with mineral oil throughout its use. Bath solutions were exchanged using a rapid solution exchanger with a 90% exchanger time of Ͻ500 ms (7). For single channel recordings, the patch pipette contained the appropriate bath solution described above.
Dye Uptake-HEK293 monolayers were washed twice with Ca-free Tyrodes containing 2 mM EGTA (EGTA-Tyrodes) and then incubated in EGTA-Tyrodes containing 1% Lucifer Yellow for 30 min at room temperature. Cells were extensively washed with normal Tyrodes containing 1.8 mM Ca before imaging. Isolated adult cardiac myocytes (8) were washed by resuspension and centrifugation in EGTA-Tyrodes and then incubated in normal Tyrodes or EGTA-Tyrodes containing 150 M calcein or 1% dextran-fluorescein for 30 min at room temperature. They were washed in normal Tyrodes three times before imaging.

RESULTS
HEK293 cells were transfected with wild-type Cx43 or Cx43 linked in frame at the C terminus to the green fluorescent protein (Cx43-GFP). In monolayers transfected with Cx43-GFP, the pattern of expression showed long lines of fluorescence that were consistent with the formation of gap junctions in regions of contact between adjacent transfected cells, in addition to some perinuclear and plasma membrane fluorescence (Fig. 1, D and F). These long fluorescent lines were not observed between adjacent nontransfected and transfected cells (Fig. 1D), suggesting that the endogenous connexin-43 in the nontransfected HEK293 cells (9) either did not pair with the Cx43-GFP or was present at too low a level (9) to form detectable fluorescent structures with the exogenous Cx43-GFP. Cells transfected with GFP alone showed a homogenous pattern of fluorescence (Fig. 1B) that did not transfer to adja-cent nontransfected cells, which is consistent with the size permeation characteristics of gap junctions (10).
To examine whether Cx43-GFP formed functional plasmalemmal hemichannels, individual HEK293 cells from nonconfluent monolayers were patch-clamped in the whole cell mode (Fig. 2, A and B). Consistent with previous studies (2,(11)(12)(13)(14)(15)(16), the removal of extracellular Ca rapidly induced a large current, with a linear current-voltage relationship that reversed at Ϫ5 Ϯ 1 mV (n ϭ 4). Current amplitude at Ϫ80 mV averaged 838 Ϯ 173 pA. No voltage-or time-dependent inactivation was evident when voltage clamps were used in place of the voltage ramp (data not shown). Replacement of extracellular Na and K with N-methyl-D-glucamine (M r 195) shifted the reversal potential in the negative direction by only Ϫ7 Ϯ 1 mV (data not shown), indicating a relatively nonselective current with significant permeability to large cations. The current was reversibly blocked by 68 Ϯ 6% using the gap junction blocker halothane (Refs. 2 and 17; Fig. 2, A and B). In nontransfected cells, removing extracellular Ca had a much smaller effect on the current-voltage relationship (Fig. 2D). Current amplitude at Ϫ80 mV increased from 26 Ϯ 6 to 179 Ϯ 48 pA, which was presumably attributable to either the activation of endogenous hemichannels or the nonspecific effects of Ca removal on membrane conductance.
To determine whether linking GFP to the C terminus of Cx43 artificially promoted hemichannel opening at low extracellular Ca, we co-transfected HEK293 cells with wild-type Cx43 and GFP fused to the ␤ subunit of Na,K-ATPase (␤ NaKP-GFP) as a marker for identifying transfected cells. Individual fluorescent cells examined in the whole cell patch clamp mode produced similar linear nonselective currents in response to the removal of extracellular Ca, as with Cx43-GFP. Currents were blocked by La and, to a lesser extent, by halothane ( Fig. 2D), possibly indicating that GFP linkage may modify halothane sensitivity.
Single channel currents were recorded from HEK293 cells expressing Cx43-GFP (Fig. 2C). With 1.8 mM Ca in the patch pipette, no single channel currents were observed (five cells). However, with low free Ca (1 nM) in the patch pipette, single channel currents were detected in six of nine cells transfected with Cx43-GFP, but in none of the nontransfected cells. Furthermore, halothane inhibited the channels, with the NP o decreasing from 0.28 Ϯ 0.02 to 0.13 Ϯ 0.03 (n ϭ 3). The single channel conductance of the fully opened channel averaged 120 Ϯ 25 pS. Several substate conductance levels were also observed, although their dwell times were too inconsistent to analyze quantitatively.
Dye transfer experiments confirmed that the current observed in low extracellular Ca was due to functional hemichannels (Fig. 3). HEK293 cell monolayers containing either all nontransfected cells (Fig. 3, A and B) or a mixture of transfected and nontransfected cells (Fig. 3, C-F) were exposed to nominally Ca-free media containing 1% Lucifer Yellow (M r 522). After 30 min, Lucifer Yellow was washed out with Cacontaining media, and the cells were imaged. Cells expressing Cx43-GFP showed the brightest dye uptake (Fig. 3, C and D). Contacting nontransfected cells also showed a dimmer fluorescence, which is consistent with dye transfer from the adjacent transfected cells via gap junctions. No dye uptake occurred in nontransfected cells that were not close to transfected cells, suggesting that they did not take up Lucifer Yellow. The entry of Lucifer Yellow and the subsequent transfer to contacting cells indicate that the putative Cx43 hemichannels opened by low extracellular Ca are nonselective and are large enough to allow molecules of at least M r 522 to enter, which is consistent with the known properties of gap junctions (18). Comparable results were obtained when wild-type Cx43 was transfected alone (Fig. 3, E and F), although the successfully transfected cells could not be unequivocally identified.
These findings establish that Cx43 forms functional hemichannels with typical properties of gap junctions. Linkage to GFP did not fundamentally alter hemichannel properties. We therefore proceeded to test the effects of metabolic inhibition. HEK293 cells expressing Cx43-GFP were patch-clamped (whole cell mode) with 1.8 mM [Ca] o and exposed to the metabolic inhibitors carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP; 10 M) and iodoacetate (IAA; 1 mM) to inhibit oxidative and glycolytic metabolism, respectively (Fig. 2, E and F). After ϳ15 min, a large linear current developed that was similar to that induced by low [Ca] o but ϳ2-fold larger (1720 Ϯ 282 pA at Ϫ80 mV; reversal potential, Ϫ7 Ϯ 1.5 mV; n ϭ 5). This current was reversibly inhibited by 73 Ϯ 8% by halothane and irreversibly inhibited by 97 Ϯ 1% by 1 mM [La] o (Fig. 2, E  and F). In contrast, in nontransfected cells, only a small increase in current from 39 Ϯ 15 to 132 Ϯ 49 pA was observed during metabolic inhibition (Fig. 2H).
To rule out a nonspecific effect of the metabolic inhibitors, we tested alternative combinations of metabolic inhibitors: 5 M rotenone (a mitochondrial inhibitor rather than an uncoupler like FCCP) ϩ 1 mM IAA or 10 M FCCP ϩ 10 mM 2-deoxyglucose (a glycolytic inhibitor). Although the activation time course differed, a similar current developed in transfected cells (Fig. 2H). The combination of FCCP ϩ IAA also activated the hemichannel current in cells co-transfected with wild-type Cx43 and ␤NaKP-GFP (Fig. 2H). Halothane and La inhibited the putative hemichannel currents (Fig. 2H), although to a slightly lesser degree.
We also recorded single channel currents from cell-attached patches on Cx43-GFP-transfected cells during metabolic inhibition with FCCP ϩ IAA (Fig. 2G) with normal extracellular Ca (1.8 mM) in the patch pipette and bath. After a delay of 8 -20 min, noisy-appearing currents were observed. Single channel amplitude was similar to that induced by low [Ca] o (5 Ϯ 1.2 versus 4.5 Ϯ 0.9 pA at Ϫ40 mV). However, the kinetics differed, due to a greater substate occupancy during metabolic inhibi- tion. These are likely to be Cx43-GFP hemichannels, because the NP o decreased significantly from 0.33 Ϯ 0.13 to 0.21 Ϯ 0.11 (n ϭ 3; Fig. 2G) upon superfusion with halothane, and channel activity was never observed during metabolic inhibition in nontransfected cells.
We have previously described a nonselective La-sensitive current activated by metabolic inhibition in isolated cardiac ventricular myocytes (4,8). Because Cx43 is the most abundant gap junctional protein in these cells (19), we investigated whether this current could be explained by the activation of Cx43 hemichannels. To detect whether functional hemichan-nels were present, isolated myocytes were incubated for 30 min at room temperature (22°C) in the absence of extracellular Ca with 150 M calcein (M r 623) present. After washing out calcein with Ca-containing solution, ϳ50% of the rod-shaped cells were fluorescent, indicating calcein uptake (Fig. 4A, bottom panel), as compared with only ϳ3% of rod-shaped myocytes after incubation with calcein in media containing 1.8 mM Ca (Fig. 4A,  top panel). Dead myocytes took up dye under both conditions. No rod-shaped cells became fluorescent after exposure to dextran-linked fluorescein (M r 1500 -3000) in either the absence or presence of Ca (Fig. 4B), which is consistent with the known

FIG. 2. Activation of hemichannel currents by low extracellular Ca (A-D) and by metabolic inhibition (E-H) in HEK293 cells expressing Cx43-GFP or wild-type Cx43.
A, time course activation of whole cell hemichannel currents by low extracellular Ca (0Ca) in a patch-clamped HEK293 cell transfected with Cx43-GFP. Current amplitude was measured at Ϫ80 mV. The current was reversibly blocked by the application of halothane (H). B, current-voltage relationships at the selected time points (a-d) indicated in A. Currents were recorded during a voltage ramp from Ϫ100 to ϩ60 mV at 0.1 mV/ms. Selected time points were as follows: a, control with 1. , cells transfected with Cx43-GFP or wild-type Cx43 (WT) and exposed to the metabolic inhibitors FCCP ϩ IAA developed a large current at Ϫ80 mV that was partially blocked by halothane (H) and completely blocked by La (n ϭ 5). Similar results were obtained in cells transfected with Cx43-GFP subjected to different combinations of metabolic inhibitors, either rotenone (Rot) ϩ 2-deoxyglucose (DOG; n ϭ 2) or FCCP ϩ DOG (n ϭ 2).
Electrophysiological evidence also supported the existence of functional hemichannels in isolated myocytes. After the calcein dye uptake protocol, brightly fluorescent rod-shaped myocytes were patch-clamped (whole cell mode). The removal of extracellular Ca (with or without EGTA present) rapidly induced a current with a linear voltage relationship that reversed near 0 (n ϭ 6; Fig. 5, A-C). Current amplitude at Ϫ80 mV averaged 7 Ϯ 1.8 pA/pF (n ϭ 6). The current was reversibly blocked by To determine whether the endogenous hemichannels in myocytes could be activated by metabolic inhibition with normal extracellular Ca, calcein-labeled myocytes were exposed to rotenone (5 M) in dextrose-free Tyrode's solution containing 1.8 mM Ca. After 8 -10 min, a linear nonrectifying current developed that was fully blocked by 2 mM La (Fig. 5, D-F). The reversal potential of the La-sensitive component was 5.6 Ϯ 5.2 mV compared with 3.9 Ϯ 2.5 mV for the current induced by Ca removal. Myocytes underwent a rapid and progressive contracture as the current activated, indicating a severe disturbance of ionic homeostasis and intracellular Ca overload. We have previously shown (4) that this current is nonselective to molecules as large as N-methyl-D-glucamine (M r 195), which is consistent with hemichannel properties in HEK293 cells (Fig. 2). DISCUSSION The first major finding of this study is that both wild-type Cx43 and the Cx43-GFP assemble into functional hemichannels when heterologously overexpressed in HEK293 cells. The hemichannels open upon the removal of extracellular Ca, are nonselective and permeable to large molecules such as Lucifer Yellow (M r 450), and are reversibly blocked by halothane and irreversibly blocked by La. To our knowledge, this study is the first detailed electrophysiological characterization of hemichannels formed from heterologously expressed Cx43 and defines a convenient, useful assay system for future structurefunction studies. The findings are generally consistent with previously described studies of heterologously expressed Cx46 (13,15,16) and Cx56 (12) and studies of native hemichannels in skate (20) and catfish retina (11). Whether other connexins are activated by metabolic inhibition, however, is unknown. Our study also documents the presence of functional [Ca] osensitive hemichannels in normal ventricular myocytes, suggesting a novel potential explanation for the "Ca paradox" (21).
The most significant finding is the demonstration that endogenous and heterologously expressed Cx43 or Cx43-GFP hemichannels open when subjected to metabolic inhibition, even when extracellular Ca remains normal. In a beating cardiac myocyte, 10 open hemichannels producing a 100 pA (0.7 pA/pF) current at Ϫ80 mV would increase the Na influx by 75% (22), and the fully developed current (5 pA/pF or about 70 open hemichannels) would increase the Na influx 5-fold. This represents a tiny fraction of the estimated 2.6 ϫ 10 6 junctional connexins in a typical cardiac myocyte (23). To the extent that the Na-K pump is unable to compensate for this additional Na load, intracellular Na accumulation and K loss result (22), causing membrane depolarization, intracellular Ca overload by reverse Na-Ca exchange, and possibly direct Ca entry through hemichannels. We cannot absolutely exclude that hemichannel density in isolated myocytes is artifactually increased by the isolation procedure. However, existing evidence suggests that gap junction plaques are pulled off intact with either one myocyte or its partner, rather than separating into plaques of hemichannels (24).
The gap junction blocker halothane has been shown to have cardioprotective effects during myocardial ischemia (25), which could be mediated in part by inhibition of the Cx43 hemichannel. Also, activators of protein kinase C have been shown to prevent Cx43 hemichannels from opening during the removal of extracellular Ca (5), raising the possibility that the role of protein kinase C in ischemic preconditioning may be linked in part to the suppression of hemichannel activation.