Connexin 43 Hemichannels Contribute to Cytoplasmic Ca2+ Oscillations by Providing a Bimodal Ca2+-dependent Ca2+ Entry Pathway*

Background: Connexin hemichannels are Ca2+-permeable plasma membrane channels that are controlled by [Ca2+]i; therefore, they may contribute to Ca2+ oscillations. Results: Ca2+ oscillations triggered by bradykinin in connexin-expressing cells were inhibited by blocking hemichannel opening or by preventing their closure at high [Ca2+]i; ATP-triggered oscillations were unaffected. Conclusion: Hemichannels contribute to oscillations by controlling Ca2+ entry. Significance: Hemichannels together with InsP3 receptors help shape agonist-induced Ca2+ oscillations. Many cellular functions are driven by changes in the intracellular Ca2+ concentration ([Ca2+]i) that are highly organized in time and space. Ca2+ oscillations are particularly important in this respect and are based on positive and negative [Ca2+]i feedback on inositol 1,4,5-trisphosphate receptors (InsP3Rs). Connexin hemichannels are Ca2+-permeable plasma membrane channels that are also controlled by [Ca2+]i. We aimed to investigate how hemichannels may contribute to Ca2+ oscillations. Madin-Darby canine kidney cells expressing connexin-32 (Cx32) and Cx43 were exposed to bradykinin (BK) or ATP to induce Ca2+ oscillations. BK-induced oscillations were rapidly (minutes) and reversibly inhibited by the connexin-mimetic peptides 32Gap27/43Gap26, whereas ATP-induced oscillations were unaffected. Furthermore, these peptides inhibited the BK-triggered release of calcein, a hemichannel-permeable dye. BK-induced oscillations, but not those induced by ATP, were dependent on extracellular Ca2+. Alleviating the negative feedback of [Ca2+]i on InsP3Rs using cytochrome c inhibited BK- and ATP-induced oscillations. Cx32 and Cx43 hemichannels are activated by <500 nm [Ca2+]i but inhibited by higher concentrations and CT9 peptide (last 9 amino acids of the Cx43 C terminus) removes this high [Ca2+]i inhibition. Unlike interfering with the bell-shaped dependence of InsP3Rs to [Ca2+]i, CT9 peptide prevented BK-induced oscillations but not those triggered by ATP. Collectively, these data indicate that connexin hemichannels contribute to BK-induced oscillations by allowing Ca2+ entry during the rising phase of the Ca2+ spikes and by providing an OFF mechanism during the falling phase of the spikes. Hemichannels were not sufficient to ignite oscillations by themselves; however, their contribution was crucial as hemichannel inhibition stopped the oscillations.

effects of Ca 2ϩ on the open probability of InsP 3 R channels, which display a typical bell-shaped dependence (7,8). When [Ca 2ϩ ] i is below a certain threshold (ϳ300 nM), Ca 2ϩ potentiates InsP 3 -triggered Ca 2ϩ release (7), resulting in Ca 2ϩ -induced Ca 2ϩ release (9). A further rise in [Ca 2ϩ ] i above this level results in negative feedback, marking the start of the decaying phase of the Ca 2ϩ spike (7). After restoration of the ER Ca 2ϩ content (by sarcoplasmic/endoplasmic Ca 2ϩ ATPase pumps and SOCE) the cycle repeats to induce the next Ca 2ϩ spike. An important condition for Ca 2ϩ oscillations to occur is the necessity for kinetic differences between positive and negative Ca 2ϩ feedback, which means the positive feedback action should be faster than the negative one (8, 10 -12), a condition that is fulfilled for InsP 3 R channels (13,14). Continued Ca 2ϩ oscillations necessitate a slightly elevated intracellular InsP 3 concentration that sets a certain degree of Ca 2ϩ excitability of InsP 3 R channels, making them sensitive to small local Ca 2ϩ increases that can fire the next Ca 2ϩ spike through Ca 2ϩ -induced Ca 2ϩ release (10). When the InsP 3 concentration is elevated consequent to stronger GPCR stimulation, the oscillation frequency generally increases (10,15). In addition, the intracellular InsP 3 concentration is not solely determined by the level of GPCR stimulation but may also be influenced by direct or indirect feedback actions of [Ca 2ϩ ] i on Ca 2ϩ -or PKC-sensitive PLC isoforms (␦, , or ␤, ␥ isoforms, respectively) (16,17), thereby generating oscillations in the InsP 3 concentration. These InsP 3 oscillations may modulate the Ca 2ϩ oscillations (by augmenting Ca 2ϩ -induced Ca 2ϩ release (11)) but may also take the lead and provide the primary driving force for the Ca 2ϩ oscillations (18), depending on the GPCRs involved. Additionally, other feedback actions on the InsP 3 metabolism (11,19) and on Ca 2ϩ entry (20 -24) provide supplementary tools to modulate and shape the oscillatory signal (25).
Evidence is accruing that connexin channels play a role in Ca 2ϩ oscillations. Connexins form two kinds of channels, hemichannels and gap junction channels, the latter resulting from the head-to-head interaction of two hemichannels. Gap junction channels connect the cytoplasm of adjacent cells, whereas unapposed hemichannels, when open, link the cytoplasm with the extracellular fluid. Both types of channels are permeable to substances with a molecular mass below 1-1.5 kDa (26,27). Kawano et al. (28) reported that octanol, a nonspecific connexin channel blocker, inhibited spontaneous Ca 2ϩ oscillations in human mesenchymal stem cells. This work suggested the opening of hemichannels followed by ATP diffusing out of the cell and acting in an autocrine way on plasma membrane P2Y 1 receptors, thereby activating PLC ␤ and generating InsP 3 . Verma et al. (29) reported that 43 Gap26 and 37/43 Gap27, two synthetic peptides that mimic short sequences in, respectively, the first extracellular loop of connexin 43 (Cx43) and the second extracellular loop of Cx37/Cx43, inhibited Ca 2ϩ oscillations in connexin-expressing HeLa cells and in cardiac myocytes. 43 Gap26 and 37/43 Gap27 peptides are inhibitors of Cx43 gap junctions and have been reported to inhibit Cx43 hemichannels with faster kinetics (30 -33). Verma et al. (29) proposed that Gap inhibition of Ca 2ϩ oscillations was mediated by reducing Ca 2ϩ entry via hemichannels thereby affecting ER Ca 2ϩ release. We recently reported that 37/43 Gap27 inhibits bradykinin (BK)-triggered Ca 2ϩ oscillations in blood-brain barrier endothelial cells and thereby prevents a subsequent increase in barrier permeability (34). In addition to the fact that hemichannel-mediated ATP release and Ca 2ϩ entry may play a role in Ca 2ϩ oscillations, hemichannel opening is controlled by [Ca 2ϩ ] i (35)(36)(37). Because hemichannels both influence and are influenced by [Ca 2ϩ ] i , we examined the mechanisms by which hemichannels contribute to Ca 2ϩ oscillations. We specifically aimed to determine whether hemichannel-[Ca 2ϩ ] i interactions constitute a mechanism supporting oscillatory activity in a manner analogous to the InsP 3 R-[Ca 2ϩ ] i link by using tools that selectively target either InsP 3 Rs or connexin channels. We found that hemichannels contribute to InsP 3 R-based oscillations by providing a Ca 2ϩ entry pathway and by shutting down this Ca 2ϩ entry pathway by inhibiting hemichannel activity when [Ca 2ϩ ] i increases above ϳ500 nM. This contribution of hemichannels was essential as the Gap peptides blocked the BK-induced oscillations. Hemichannels were not involved in ATP-induced Ca 2ϩ oscillations, indicating that they may help in shaping distinct Ca 2ϩ response patterns to different agonists.
Ca 2ϩ Imaging-MDCK cells were seeded onto 18-mm-diameter glass coverslips (Knittel Glaser, Novolab, Geraardsbergen, Belgium), and experiments were performed at subconfluency the next day. The cells were loaded with 10 M fluo3-AM in HBSS-Hepes (1 mM CaCl 2 , 0.81 mM MgSO 4 , 13 mM NaCl, 0.18 mM Na 2 HPO 4 , 5.36 mM KCl, 0.44 mM KH 2 PO 4 , 5.55 mM D-glucose, and 25 mM Hepes) for 1 h at room temperature (RT). After 1 h, coverslips were washed and then left for an additional 30 min at RT in HBSS-Hepes to allow for de-esterification. Cells were thereafter transferred to an inverted epifluorescence microscope (Eclipse TE 300, Nikon Belux, Brussels, Belgium) equipped with a superfusion system that allowed changing the bath solution within ϳ1 min (bath volume ϳ300 l). Superfusion was switched off during the registration of oscillatory activity. Images were taken every second with a ϫ40 oil-immersion objective and an electron-multiplying CCD camera (Quantem 512SC, Photometrics, Tucson, AZ). We used a Lambda DG-4 filterswitch (Sutter Instrument Company, Novato, CA) to deliver excitation at 482 nm and captured emitted light via a 505-nm long-pass dichroic mirror and a 535-nm bandpass filter (35-nm bandwidth). Cells were loaded with fura2-AM (5 M) in a similar manner. Excitation was alternated between 340 and 380 nm at a rate of one image pair every second. Emitted light was captured using a 430-nm long-pass dichroic mirror and a 510-nm bandpass-filter (40-nm bandwidth). Fura2 in situ calibration was performed in zero Ca 2ϩ medium (10 mM EGTA) containing 10 M A23187 (R min ) and a saturating Ca 2ϩ solution (10 mM CaCl 2 ) containing 40 M digitonin (R max ). [Ca 2ϩ ] i was calculated from K d ⅐Q⅐[(R Ϫ R min )/ (R max Ϫ R)], where R is the F 340 /F 380 ratio, Q is F min /F max at 380 nm, and K d 224 nM (39). Mitochondrial Ca 2ϩ measurements were performed with the mitochondrial Ca 2ϩ indicator RhodFF as we described previously (40). MDCK cells were loaded with RhodFF-AM (5 M) for 1 h at RT followed by 30 min of de-esterification. Imaging was performed in a similar manner as fluo3 imaging but with excitation at 556 nm and long-pass filtering at 590 nm. A punctate distribution that matches the distribution of Mitotracker Green (100 nM, 1 h, RT) confirmed the mitochondrial localization of the dye (supplemental Figs. S1). Recordings and analysis were done with custom-developed QuantEMframes and Fluoframes software written in Microsoft Visual C ϩϩ 6.0. Oscillatory activity was recorded in a 10-min observation window, and only Ca 2ϩ spikes minimally 10% above baseline were considered in the analysis. To calculate the percentage of oscillating cells, an oscillating cell was defined as a cell displaying at least two Ca 2ϩ spikes subsequent to the initial spike. The percentage of cells oscillating was calculated relative to the total number of cells in view. The oscillation frequency is the average frequency of all cells in view including non-oscillating cells.
Electroporation Loading-MDCK cells were loaded with CytC or CT peptides by electroporation. Cells, seeded the day before the experiment, were rinsed with a low conductivity electroporation buffer and placed on the microscope stage.
Thereafter, a small volume (10 l) of CytC (3 M)/DTR (100 M) or CT9 peptide (300 M)/PI (12 M) dissolved in electroporation buffer was added to a parallel wire Pt-Ir electrode positioned 400 m above the cells. DTR (10 kDa) or PI (668 Da) have molecular masses approaching that of CytC (12 kDa) or CT9 (1 kDa), respectively, and were added to visualize the electroporated cells. In experiments using the Ca 2ϩ dye RhodFF, we visualized the CytC-loaded zone with 10-kDa dextran fluorescein. The IP 3 RCYT peptide (15 M) was added to the CytC (3 M) solution (see above) 30 min before electroporation to allow for interaction. Electroporation was performed after fluo3-AM loading with 50-kHz bipolar pulses at a field strength of 1000 V/cm, applied as 15 trains of 10 pulses of 2-ms duration each (41,42). Electroporation did not result in loss of fluo3/fura2/ RhodFF from the cells.
Caged Compound Loading and Photoliberation-Cells seeded on coverslips the day before the experiment were electroporated with cell-impermeant, caged InsP 3 (30 M), and DTR (100 M) as described above. Caged Ca 2ϩ was loaded into the cells by ester-loading, similar to the Ca 2ϩ -sensitive dyes. Thereafter, the coverslips were transferred to an inverted epifluorescence microscope. Photoliberation of InsP 3 was done by spot (20 m diameter) illumination with 1-kHz pulsed UV light (349 nm UV laser Explorer, Spectra-Physics, Newport, Utrecht, The Netherlands) applied during 20 ms (20 pulses of 90 J energy measured at the entrance of the microscope epifluorescence tube). For uncaging of Ca 2ϩ we applied UV illumination with different flash durations.
Hemichannel Assays-Hemichannel opening was investigated by calcein (623 Da) release, which is based on the efflux of the preloaded dye via open hemichannels (43). Subconfluent cultures of MDCK cells grown on glass coverslips (18 mm diameter), were preloaded with 50 M calcein-AM in HBSS-HEPES for 1 h at RT. Subsequently, the remaining calcein-AM was removed; cells were left for an additional 30 min at RT in HBSS-Hepes to allow the AM ester to de-esterify and were then transferred to the stage of an inverted epifluorescence microscope. For analysis, we measured the decrease in calcein fluorescence in function of time. The first 5 min, baseline leakage in HBSS-Hepes was measured (control); thereafter, the trigger solution was added, and efflux of calcein was further evaluated. The slope of the curve, calculated by linear regression, was used as a parameter describing the loss of dye in time. Calcein efflux in the presence of trigger is presented as % of control.
Gap Junction Dye Coupling Studies-Dye coupling via gap junctions was determined using fluorescence recovery after photobleaching. MDCK cell cultures were grown to confluence on 9.2 cm 2 Petri dishes (Techno Plastic Products, Trasadingen, Switzerland) and loaded with the gap junction-permeable fluorescent dye 5-carboxyfluorescein diacetate-AM (532 Da, 10 M) in HBSS-Hepes for 1 h at RT. After de-esterification, cells were transferred to a custom-made video-rate confocal laser scanning microscope with a ϫ40 water immersion objective (CFI Plan Fluor) and a 488-nm laser excitation source (Cyan CW Laser, 488 nm, 100 milliwatt; Newport Spectra-Physics, Utrecht, The Netherlands). After 1 min of recording, the cell in the middle of the field was photobleached by spot exposure (1 s) to increased power of the 488 nm laser, and fluorescence recov-ery, caused by dye influx from neighboring non-bleached cells, was recorded during an additional 5-min period. The fluorescence recovery trace was then analyzed for the recovery of the signal expressed relative to the starting level before photobleaching.
Ectonucleotidase Activity-Ectonucleotidase activity was assessed by measuring the breakdown of ATP added to the cells via the decline of luciferin/luciferase luminescence in the medium above the cells. MDCK and RBE4 cells were seeded at a density of 40,000 cells/cm 2 in 24-well plates. ATP (100 M) together with an ATP bioluminescent assay mix (luciferin/luciferase, 625-fold dilution, Sigma) was prepared in ATP assay mix dilution buffer (Sigma). Photon flux was counted using a multilabel counter (Victor-3, type 1420, PerkinElmer Life Sciences). The time constant () of the exponential luminescence decay was used as a parameter to express ectonucleotidase activity.
Apoptosis Assay-Annexin V staining detects the flip-flop of phosphatidylserine toward the outer plasma membrane leaflet that occurs during apoptosis. After electroporation with 10-kDa DTR (100 M) and CytC (3 M), cells were rinsed with PBS and incubated for 15 min (RT) with annexin V-FITC (1:50; Roche Diagnostics) and Hoechst 33342 (2 g/ml) in annexin V buffer (140 mM NaCl, 5 mM CaCl 2 , 10 mM HEPES, pH 7.4). The cultures were subsequently washed with PBS and transferred to a Nikon TE300 epifluorescence microscope, equipped with a ϫ10 objective (Plan Apo, NA 0.4, Nikon). Ten images inside the electroporation zone were taken, and annexin V-positive cells were counted. The number of apoptotic cells was expressed as the percentage of annexin V-positive cells relative to the total number of cells (determined by Hoechst 33342 staining).
Electrophoresis and Western Blot Analysis-For Western blots, cells were seeded in 75-cm 2 flasks. Total MDCK cell lysates were extracted with radioimmune precipitation assay buffer (25 mM Tris, 50 mM NaCl, 0.5% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 1 mM DTT, 0.055 g/ml ␤-glycerol phosphate, 30 l/ml phosphatase inhibitor mixture, and 20 l/ml mini EDTA-free protease inhibitor mixture). For separation of Triton X-100 soluble (cytosol) and insoluble (membrane) fractions, cells were harvested in 1% Triton X-100 sup-plemented with 50 mM NaF and 1 mM Na 3 VO 4 , and centrifuged at 16,000 g for 10 min. The Triton X-100 insoluble pellets were resuspended in 1ϫ Laemmli sample buffer. Protein concentration was determined using a Bio-Rad DC protein assay (Bio-Rad), and absorbance was measured with a 590-nm long-pass filter. Lysates were separated by electrophoresis over a 10% SDS-polyacrylamide gel and transferred to a nitrocellulose membrane (Amersham Biosciences). Membranes were subsequently blocked with TBS containing 5% nonfat milk and 0.1% Tween20. After blocking, blots were probed with rabbit anti-Cx43 antibody (Sigma), rabbit anti-Cx32 antibody (Sigma), rabbit anti-Cx26 antibody (Zymed Laboratories Inc., Invitrogen), rabbit anti-phospho-Cx43 (Ser(P)-368) (Cell Signaling Technology, Inc., Danvers, MA), rabbit anti-P2X 7 antibody (Alomone Labs, Jerusalem, Israel), anti-Panx1 antibody (a kind gift of Dr. Dale W. Laird, University of Western Ontario), and rabbit anti-␤-tubulin antibody (Abcam, Cambridge, UK) as a loading control. Membranes were subsequently incubated with an alkaline phosphatase-conjugated goat anti-rabbit IgG antibody (Sigma), and detection was done using the nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl-phosphate reagent (Zymed Laboratories Inc., Invitrogen). Quantification was done by drawing a rectangular window around the concerned band and determining the signal intensity using ImageJ software. Background correction was done by the same procedure applied to nitrocellulose membranes where protein was absent.
Statistical Analysis-Data are expressed as the mean Ϯ S.E. with n giving the number of independent experiments. Multiple groups were compared by one-way analysis of variance and Bonferroni post-test, making use of Graphpad Instat software. Two groups were compared with an unpaired Student's t test and two-tail p value. Results were considered statistically significant when p Ͻ 0.05 (one symbol for p Ͻ 0.05, two for p Ͻ 0.01, and three for p Ͻ 0.001).

Concentration and InsP 3 Dependence of BK-and ATP-induced Ca 2ϩ
Oscillations-We first characterized BK-and ATP-induced Ca 2ϩ oscillations in MDCK cells and determined the concentration dependence of the percentage of oscillating cells and of the oscillation frequency. We used non-confluent MDCK cell cultures to this purpose to limit the degree of gap junctional coupling. BK and ATP triggered an initial [Ca 2ϩ ] i transient followed by repetitive Ca 2ϩ spikes (recorded over a 10-min period) with quite a different profile (Fig. 1, A and B). BK concentrations ranging from 0.05 to 100 M all triggered Ca 2ϩ oscillations in an invariable percentage of cells (ϳ72%) and a similar oscillation frequency (ϳ5 spikes/10 min, measured over all cells in view) (Fig. 1C). By contrast, oscillations triggered by ATP concentrations between 0.5 M and 2 mM were characterized by a bell-shaped concentration-response curve for the percentage of oscillating cells and oscillation frequency (Fig. 1D). The maximum number of oscillating cells (89 Ϯ 3.4%; n ϭ 5) and the maximal oscillation frequency (13 Ϯ 1.1 spikes/10 min; n ϭ 5) were observed with 10 M ATP. These markedly different patterns of concentration dependence indicate distinct oscillation mechanisms for ATP and BK. We chose 10 M ATP (a concentration located between the two peaks depicted in Fig. 1D) and 0.5 M BK (relative location within the range of concentrations tested comparable with ATP) for further analysis of the differences between the oscillations triggered by these two agonists. We first determined whether the amplitude of the initial [Ca 2ϩ ] i transient triggered by BK (0.5 M) or ATP (10 M) was different but found they were very similar (650 Ϯ 59 and 783 Ϯ 110 nM, respectively, n ϭ 8, p Ͼ 0.05) which suggests that the intracellular InsP 3 elevation is comparable with the two stimuli. Yet the area under the curve (AUC) for both triggers differed markedly with BK exhibiting a much larger AUC compared with ATP (Fig. 1, E and F). Because the amplitudes were not different, it follows that the BK-triggered [Ca 2ϩ ] i transient is longer than the one triggered by ATP (see example traces in Fig. 1A). The longer duration of the BKinduced [Ca 2ϩ ] i transient is likely to be related to Ca 2ϩ entry from the extracellular space. In line with this, withdrawal of extracellular Ca 2ϩ (zero extracellular Ca 2ϩ ) did not much affect the initial [Ca 2ϩ ] i transient elicited by ATP, whereas it largely reduced the one triggered by BK (Fig. 1, E and F). Further probing of SOCE by reintroducing extracellular Ca 2ϩ after the [Ca 2ϩ ] i transient in zero extracellular Ca 2ϩ conditions showed that SOCE was much larger for BK than for ATP (Fig. 1G). Sneyd et al. (45) described a method to distinguish between oscillations characterized by a constant level of intracellular InsP 3 and those associated with oscillatory InsP 3 fluctuations by recording the response to an applied InsP 3 -concentration step. If InsP 3 fluctuates during the Ca 2ϩ oscillations, induction of a sudden InsP 3 increase will introduce a delay to the next Ca 2ϩ spike. The delay is caused by the fact that InsP 3 has to recover to a level that is again compatible with Ca 2ϩ oscillations; thereafter, the oscillation frequency stabilizes again. If InsP 3 is constant during the Ca 2ϩ oscillations, an induced InsP 3 elevation will temporarily increase the oscillation frequency, giving accelerated oscillations (45,46). We performed InsP 3 elevation experiments with flash photolysis of caged InsP 3 and found that the majority of the cells (ϳ48%) showed a delayed response for both BK-and ATP-induced oscillations (Fig. 1, H-K). A smaller fraction of the cells displayed an accelerated response: ϳ10% for BK and ϳ22% for ATP oscillations. In the remainder of the cells InsP 3 elevation had no effect. These data indicate that oscillations triggered by BK and ATP are similar, at least with respect to the occurrence of InsP 3 oscillations.
Connexin Channel Blockers Inhibit BK-induced Ca 2ϩ Oscillations but Not Those Triggered by ATP-SDS-PAGE and Western blot analysis revealed the presence of Cx32 and Cx43 in MDCK cells, with a small background expression of Cx26. Both Cx32 and Cx43 were present in the plasma membrane, whereas their presence in the cytosolic fraction was limited ( Fig. 2A). MDCK cells did not express Panx1 or P2X 7 , which is linked to Panx1 channels (47) (Fig. 2, B and C). When BKtriggered oscillations were elicited in the presence of the general connexin channel blocker Cbx (25 M, 30-min preincuba-tion and present during the 10-min observation window), the initial Ca 2ϩ spike remained, but the subsequent oscillations disappeared. We further applied two peptide connexin channel inhibitors, 32 Gap27 and 43 Gap26, that target Cx32 and Cx43 channels, respectively (36,48). Application of 32 Gap27/ 43 Gap26 ("Gap," 200 M, 30-min preincubation and present during the recording) also inhibited the Ca 2ϩ oscillations without perturbing the initial peak (Fig. 3A). 32 Gap27 and 43 Gap26, either separately or in an equimolar mix, reduced the percentage of oscillating cells to ϳ1/3, and Cbx reduced it to ϳ1/7 (Fig.  3B); a similar degree of inhibition was observed for the oscillation frequency (Fig. 3C). Superfusion experiments showed that inhibition by 32 Gap27/ 43 Gap26 was rapid, within ϳ1 min, and that oscillations reappeared upon wash-out of the peptides (Fig.  3D). We next tested the effect of Gap peptides and Cbx on Ca 2ϩ oscillations triggered by ATP (10 M) and found they had no effect on the percentage of oscillating cells or the oscillation frequency (Fig. 3, E-G). Thus, only BK-triggered Ca 2ϩ oscillations are influenced by Cbx or Gap peptides. The BK-triggered Ca 2ϩ oscillations were not synchronized in neighboring cells, pointing to absence of gap junctional or other synchronizing mechanisms (49 -51). The rapid block of oscillations by Gap peptides suggests an effect at the level of hemichannels, as generally longer incubations are needed to also influence gap junctions (30,32,34). In line with this, 32 Gap27/ 43 Gap26 (200 M, 60 min) had no effect on gap junction dye coupling studied with fluorescence recovery after photobleaching (Fig. 3H). Knockdown of Cx43 RNA showed that suppressing Cx43 expression prevented BK-triggered Ca 2ϩ oscillations, in line with inhibition of oscillations by 43 Gap26 added alone (without 32 Gap27). Western blot analysis indicated a ϳ50% reduction of Cx43 expression in cells transfected with SiCx43-1 (125 nM, 48 h) (Fig. 4A), and the number of oscillating cells in the presence of BK was reduced to a similar extent (Fig. 4, B and C). SiCx43-2 was less efficient and gave proportionally less inhibition (Fig. 4,  A and C). Importantly, neither SiCx43-1 nor SiCx43-2 reduced Cx32 expression (data not shown); therefore, the oscillations that remain after Cx43 silencing may result from incomplete Cx43 knockdown or from the contribution of Cx32 hemichannels. Cx43-gene silencing did not influence ATP-triggered Ca 2ϩ oscillations (Fig. 4, D and E).
Lowering Extracellular Ca 2ϩ Differentially Affects BK-and ATP-induced Oscillations-Hemichannels may contribute to the oscillations via ATP release acting in an autocrine manner (28,34) or via Ca 2ϩ entry (29). We first set out to find evidence for hemichannel opening in response to BK. Exposure of cells preloaded with calcein (a hemichannel-permeable fluorescent dye with a molecular mass of 623 Da (43)) to BK (0.5 M) increased the calcein efflux rate, and this effect was counteracted by 32 Gap27/ 43 Gap26 (Fig. 5A). In contrast, ATP (10 M) did not accelerate dye efflux (Fig. 5A). BK-triggered calcein release could be furthermore reduced by buffering increases in [Ca 2ϩ ] i using the Ca 2ϩ chelator BAPTA-AM (Fig. 5B). Phospholipase A 2 is reported to be activated by BK (52,53) and is involved in connexin hemichannel opening (36); however, phospholipase A 2 inhibition by arachidonyl trifluoromethyl ketone had no effect on the percentage of cells displaying oscillations in response to BK (63 Ϯ 7% oscillating cells in control versus 65 Ϯ 7% oscillating cells in arachidonyl trifluoromethyl ketone-treated cells, n ϭ 4), further emphasizing that [Ca 2ϩ ] i changes are necessary for hemichannel opening. We next tested whether ATP release through open hemichannels played a role in BK-induced Ca 2ϩ oscillations. We applied apyrase VI/VII (to degrade extracellular ATP), PPADS, or suramin (to inhibit purinergic P2 receptors), and 8-(p-sulfophenyl)theophylline (to inhibit adenosine A1/A2 B receptors), but these agents did not significantly influence the percentage of oscillating cells or the oscillation frequency (Fig. 5, C and D). We recently reported that ATP release via hemichannels was involved in BK-induced oscillations in RBE4 brain endothelial cells (34), and we speculated that MDCK cells display a stronger ectonucleotidase activity than RBE4 cells. In line with this, we found that ATP added to the incubation solution above the cells was more rapidly degraded in MDCK cells as compared with RBE4 cell cultures (Fig. 5, E and F), indicating stronger ectonucleotidase activity in MDCK cells. Thus, ATP released via BK-induced hemichannel opening likely has no downstream effects on purinergic receptors or Ca 2ϩ oscillations in MDCK cells due to its rapid degradation.
Because ATP release does not contribute to BK-triggered Ca 2ϩ oscillations in MDCK cells, we evaluated the role of Ca 2ϩ entry through hemichannels. To this purpose we decreased the driving force for Ca 2ϩ entry by lowering the extracellular Ca 2ϩ concentration. The latter may result in hemichannel opening (54 -56), compromising the interpretation of the intended experiments. However, concentrations below ϳ0.2 mM are reported to open hemichannels (57); thus, we applied 0.5 mM instead of the normal extracellular Ca 2ϩ concentration of 1 mM. Calcein release experiments confirmed that exposure of MDCK cells to 0.5 mM extracellular Ca 2ϩ did not trigger hemichannel opening, whereas a further reduction of extracellular Ca 2ϩ to 0.2 mM or the subnanomolar range indeed provoked the opening of hemichannels (Fig. 6, A and B). Application of 0.5 mM extracellular Ca 2ϩ during BK exposure interrupted the Ca 2ϩ oscillations, and re-addition of normal extracellular Ca 2ϩ restored the oscillations (Fig. 6C). The number of oscillating cells was strongly reduced by 0.5 mM extracellular Ca 2ϩ (Fig.  6D). Interestingly, Ca 2ϩ oscillations induced by ATP decreased in amplitude but were not suppressed by switching to 0.5 mM extracellular Ca 2ϩ , and the percentage of oscillating cells was not significantly altered (Fig. 6, E and F). Thus, BK-induced oscillations are dependent on extracellular Ca 2ϩ , indicating a contribution of SOCE or Ca 2ϩ entry via hemichannels to the oscillation mechanism. Importantly, hemichannels did not contribute to SOCE, as 32 Gap27/ 43 Gap26, added during reintroduction of extracellular Ca 2ϩ in an experiment as shown in Fig. 1G, did not influence SOCE after BK stimulation (data not shown).

ATP-induced Ca 2ϩ Oscillations Are Inhibited by CytC, whereas BK-induced Oscillations Are Inhibited by Both CytC
and Cx43-targeting CT9 Peptide-A critical and essential factor in the generation of Ca 2ϩ oscillations is the presence of positive and negative feedback actions (11,12,15). In the classical scheme of InsP 3 -triggered Ca 2ϩ oscillations, this feedback acts at InsP 3 Rs, with low [Ca 2ϩ ] i stimulating ER Ca 2ϩ release and higher concentrations being inhibitory (7) (Fig. 7B). Based on dye uptake and ATP release studies, hemichannels composed of Cx32 and Cx43 have also been demonstrated to display a bell-shaped [Ca 2ϩ ] i dependence for opening (35,36) (Fig. 7D). There is an interesting set of tools available to influence the bell-shaped Ca 2ϩ dependence of InsP 3 Rs and hemichannels.
Negative feedback of Ca 2ϩ on InsP 3 Rs can be alleviated by CytC (58) (Fig. 7, A and B). In fact, this is an important mechanism contributing to apoptotic cell death because CytC binding to the InsP 3 receptor removes the brake on ER Ca 2ϩ release, resulting in Ca 2ϩ accumulation in mitochondria that amplifies CytC release in a vicious circle (59 -61). Inhibiting the declining phase of InsP 3 R activity is expected to disrupt oscillations because of the essential role of negative feedback as an OFF signal in the oscillation cycle. We recently reported that a synthetic peptide composed of the last 10 amino acids of the C-terminal (CT) tail of Cx43 prevented the inhibitory phase of the bell-shaped [Ca 2ϩ ] i dependence of hemichannel opening (37) (Fig. 7, C and D). If this bell-shaped [Ca 2ϩ ] i dependence contributes as a hemichannel-related mechanism in the oscillations, it is expected (similar as for the InsP 3 R) that such CT peptide would inhibit the oscillations by removing the OFF signal. We used CytC and CT9 peptide (last nine amino acids of the Cx43 CT) to selectively interfere with the negative feedback of Ca 2ϩ on InsP 3 Rs and connexin hemichannels, respectively, and examined their effect on the BK-induced Ca 2ϩ oscillations. CytC binds to InsP 3 R type 1 and 3 (58), which are both present in MDCK cells (62). CT9 and CytC are plasma membrane-impermeable, and we used in situ electroporation to load these substances into the cells without disturbing cell function or viability (41,42). To identify the cells loaded with these agents, we included the fluorescent markers DTR/dextran fluorescein and PI that have molecular masses in the range of CytC (ϳ12 kDa) and CT9 (ϳ1 kDa), respectively. Electroporation loading allows analysis of the Ca 2ϩ oscillations in loaded as well as nonloaded (control) cells (Fig. 8, A and F). After applying BK (0.5 M), cells loaded with CytC (ϳ1 M intracellular concentration) displayed significantly less oscillatory activity as compared with control cells in the same culture and as compared with cells loaded with vehicle-only (Fig. 8, B and C). The decreased oscillations were not caused by apoptosis (triggered by CytC), as CytC exposure was short (10 min), and annexin V staining to detect early apoptotic cells was negative (data not shown). We further applied the IP 3 RCYT peptide that corresponds to the CytC binding residues of the InsP 3 R1 (located on the C terminus), thereby preventing the binding of CytC to InsP 3 R (38,63). Inclusion of the IP 3 RCYT peptide (ϳ5 M intracellular concentration) prevented the CytC-mediated decrease in percentage of cells displaying Ca 2ϩ oscillations (Fig.  8C). To further document the involvement of InsP 3 signaling in BK-triggered oscillations, we tested inhibition of PLC with U73122, inhibition of InsP 3 Rs with xestospongin C (XeC), and pre-emptying of thapsigargin-sensitive Ca 2ϩ stores; all these conditions suppressed BK-triggered oscillations, as expected (Fig. 8E). Loading cells with CytC did not affect the AUC (Fig.  8D) or peak amplitude (not shown) of the initial [Ca 2ϩ ] i transient triggered by BK under zero extracellular Ca 2ϩ conditions, indicating no effect of CytC on the Ca 2ϩ dynamics associated with the initial [Ca 2ϩ ] i transient triggered by BK. However, the addition of thapsigargin after BK stimulation (still under zero extracellular Ca 2ϩ conditions) released less Ca 2ϩ from CytCloaded cells than from control cells (Fig. 8D). Hence, Ca 2ϩ store emptying was more complete with CytC, and this substance thus potentiates BK-triggered ER Ca 2ϩ release. This effect of CytC is in line with its expected action as a stimulator of InsP 3 Rs, which is the result of a decreased InsP 3 R inhibition by high [Ca 2ϩ ] i (38). Because CytC did not influence the amplitude of the [Ca 2ϩ ] i transient, we anticipated that the larger ER Ca 2ϩ release flux was more effectively taken up by mitochondria. Accordingly, we found that the mitochondrial Ca 2ϩ response (measured with RhodFF) after BK stimulation was larger in CytC-loaded cells as compared with control (Fig. 8D).
When cells were loaded with the Cx43-targeting CT9 peptide (last 9 amino acids of the Cx43 CT; ϳ100 M intracellular concentration), BK-induced oscillations were significantly inhibited compared with non-loaded control cells in the same culture and to cells loaded with vehicle-only (Fig. 8, G and H). By contrast, CT9 Rev peptide (reversed sequence) did not affect the number of oscillating cells (Fig. 8H). The last isoleucine residue of Cx43 is essential for interaction with the scaffolding FIGURE 6. Depletion of extracellular Ca 2؉ inhibits BK-induced Ca 2؉ oscillations. A, applying 0.5 mM Ca 2ϩ solution did not trigger calcein release in MDCK cells, whereas 0.2 mM or ϳ1 nM free extracellular Ca 2ϩ solutions did. B, the bar chart summarizes the effect of low extracellular Ca 2ϩ on calcein release. The asterisk indicates a significant difference from the corresponding control. C, switching to 0.5 mM extracellular Ca 2ϩ during BK-induced Ca 2ϩ oscillations immediately interrupted the oscillations, and the effect was reversible upon switching to normal extracellular Ca 2ϩ . D, low extracellular Ca 2ϩ (0.5 mM) strongly reduced the number of oscillating cells. E and F, low extracellular Ca 2ϩ (0.5 mM) had no effect on ATP-induced oscillations. *, significantly different from 0.5 M BK before Ca 2ϩ depletion; # significantly different from oscillations after restoration of extracellular Ca 2ϩ .
protein ZO-1 (64), and CT9 peptide is thus expected to bind to ZO-1 and prevent Cx43/ZO-1 binding (37,65). To investigate whether Ca 2ϩ oscillations are suppressed by dissociation of the Cx43/ZO-1 complex, we used a peptide similar to CT9 that is lacking the last isoleucine residue (CT9⌬I). This peptide does not disrupt Cx43/ZO-1 interaction (37) but was equally potent in inhibiting BK-triggered Ca 2ϩ oscillations (Fig. 8H). Therefore, the CT9-induced block of Ca 2ϩ oscillations is not likely caused by altering Cx43/ZO-1 interactions. As observed for CytC, the AUC or peak amplitude of the first [Ca 2ϩ ] i increase was not influenced by CT9 or CT9⌬I, indicating that the InsP 3 production and initial InsP 3 R responses were not affected by these treatments. In addition, baseline [Ca 2ϩ ] i was not affected by CytC or CT9: 58 Ϯ 5 nM for CytC and 56 Ϯ 4 nM for CT9 compared with 57 Ϯ 5 nM for control (n ϭ 4). Calcein release was not different in CT9-or CytC-loaded cells from their controls outside the loading zone (88 Ϯ 8 and 107 Ϯ 31%, respectively, versus 100% in control, n ϭ 3), indicating that CT9 and CytC by themselves do not trigger hemichannel opening.
Recently, O'Quinn et al. (66) showed that the CT9 peptide triggers a PKC-mediated phosphorylation of the Cx43 Ser-368 consensus site in cardiomyocytes, but we could not observe such an effect when MDCK cells were treated with cell-permeable Tat-CT9 peptide (Fig. 8I).
We next tested CytC and CT9 peptide on Ca 2ϩ oscillations elicited by ATP (Fig. 9) and found that CytC was inhibitory and inclusion of the IP 3 RCYT peptide removed this inhibition. Again, Ca 2ϩ store emptying was more complete, and mito-chondrial Ca 2ϩ response was larger with CytC (Fig. 9C). Also, inhibition of PLC and InsP 3 R as well as pre-emptying of thapsigargin-sensitive Ca 2ϩ stores suppressed oscillatory activity (Fig. 9F). Importantly, CT9 peptide had no effect on ATP-induced Ca 2ϩ oscillations. These experiments indicate that ATPinduced oscillations rely on InsP 3 Rs, whereas those induced by BK rely on both InsP 3 Rs and hemichannels.
We further tested whether hemichannels were sufficient as a mechanism to obtain oscillations without a contribution of InsP 3 Rs. Hemichannel opening can be triggered by [Ca 2ϩ ] i elevation without increasing InsP 3 and thus without activating InsP 3 Rs and InsP 3 R-based oscillations. [Ca 2ϩ ] i elevation triggered by photolytic release of Ca 2ϩ did not trigger Ca 2ϩ oscillations (n ϭ 4), indicating that hemichannels are not sufficient as an oscillatory mechanism and InsP 3 Rs are the dominant mechanism leading to oscillations.

DISCUSSION
The present data demonstrate that BK-and ATP-induced Ca 2ϩ oscillations display distinct properties in MDCK cells. Oscillations induced by BK (i) showed a flat concentration dependence of frequency and percentage of responding cells, (ii) were inhibited by Cbx, Gap peptides, and Cx43 gene silencing, (iii) disappeared when extracellular Ca 2ϩ was lowered, and (iv) were inhibited by CytC and CT9 peptide. By contrast, ATPinduced oscillations (i) showed a bell-shaped ATP concentration dependence, (ii) were not influenced by Cbx, Gap peptides, or Cx43 silencing, (iii) were not abolished by lowering extracellular Ca 2ϩ , and (iv) were inhibited by CytC but not by CT9 peptide. Some data point to the involvement of hemichannels: (i) BK potentiated calcein dye release, whereas ATP did not, (ii) BK-triggered calcein release was rapidly (ϳ1 min) inhibited by Gap peptides, whereas 60 min of incubation with these peptides had no effect on dye coupling, and (iii) BK-induced Ca 2ϩ oscillations were inhibited by CT9 peptide that removes the Ca 2ϩ inhibition of Cx43 hemichannels but not by CT9 rev peptide. The fact that the oscillations were not synchronized between ATP-induced Ca 2ϩ oscillations displayed a bell-shaped concentration dependence. An S-shaped concentration dependence is more typical (13,(67)(68)(69), but it is well known that oscillations can only occur in a limited range of InsP 3 concentrations so the disappearance of the oscillations at high ATP concentration may well result from too high an intracellular InsP 3 level (45,70,71). BK-induced oscillations had a flat concentration dependence, a finding also reported by others (72)(73)(74). Despite the importance of encoding the strength of agonist stimulation in the frequency of oscillations, it is not uncommon that spiking frequencies are independent of agonist concentration. According to the dynamic desensitization mechanism that occurs upon stimulation of mGluR5 (75) or the M 3 muscarinic receptor (76), elevated levels of [Ca 2ϩ ] i and diacylglycerol activate PKC that phosphorylates and desensitizes the receptor, thereby limiting InsP 3 production and stabilizing the oscillation frequency. BK does activate PKC (77), and PKC phosphorylation consensus sites in BK receptors have been identified (78), but further studies are required to determine whether dynamic desensitization is involved here. Ca 2ϩ oscillations were previously found to be inhibited by octanol, palmitoleic acid, and 18␣GA (28) by antibodies directed against connexins and by Gap peptides (29,34), all known blockers of connexin channels. Conversely, stimulating hemichannel responses with the anti-arrhythmic peptide AAP (79) has been reported to induce Ca 2ϩ oscillations (80). Ca 2ϩ oscillations triggered by BK were inhibited by 32 Gap27/ 43 Gap26 peptides within 1 min, as reported by others (81), and the Gap peptides did not influence the initial [Ca 2ϩ ] i transient triggered by BK. The latter indicates that the InsP 3 -triggered Ca 2ϩ release was not influenced by the Gap peptides, which is in accordance with our observation that these peptides do not influence [Ca 2ϩ ] i transients triggered by photolytic release of InsP 3 (34). BK potentiated calcein dye efflux from MDCK cells, indicating hemichannel opening. This opening is likely caused by the elevation of [Ca 2ϩ ] i to values in the 500 nM range that are appropriate for activation of Cx32 and Cx43 hemichannels (35,36). BK-triggered calcein efflux was rapidly inhibited by the Gap peptides, in line with previous experimental work (34) and complemented by recent single-channel electrophysiological studies that indicate Gap peptide inhibition of unitary currents through Cx43 hemichannels with a time constant in the order FIGURE 9. ATP-induced Ca 2؉ oscillations are inhibited by CytC but not by CT9 peptide. Experiments as in Fig. 8 but now performed with ATP as the oscillation-inducing stimulus. A, example traces of vehicle/CytC-loaded cells and their controls. B, CytC inhibited the oscillatory activity, and this effect was suppressed by co-loading with IP 3 RCYT peptide. C, less Ca 2ϩ was liberated from intracellular stores in the presence of CytC, indicating more complete store emptying; accordingly, mitochondrial Ca 2ϩ uptake was larger. D, representative traces of vehicle/CT9-loaded cells. E, CT9 peptide had no effect on ATPtriggered oscillations. F, Xestospongin-C, U73122, and thapsigargin inhibited ATP-triggered oscillations. Significance symbols are as defined in Fig. 8.
of minutes. 4 The contribution of hemichannel opening to the Ca 2ϩ oscillations appeared to be related to Ca 2ϩ entry, as indicated by the low extracellular Ca 2ϩ experiments, and was not related to hemichannel ATP release. We cannot exclude hemichannel ATP release, but our results indicate rapid ATP degradation at the surface of MDCK cells, lowering its concentration below the threshold for purinergic receptor activation. A role for hemichannels as a non-selective Ca 2ϩ -entry route has been postulated under varying circumstances including alkalinization (Cx43 (82)) and metabolic inhibition (Cx32 (83) and Cx43 (84)). Ca 2ϩ entry contributes to reloading of ER Ca 2ϩ stores, and the entry rate can influence the interspike interval and thus the oscillation frequency (85). At low agonist concentrations, hardly any Ca 2ϩ is lost from the ER (86) and Ca 2ϩ entry may, under those circumstances, act to sensitize the InsP 3 R to produce a regenerative Ca 2ϩ spike (87) or exert positive feedback on PLC-mediated InsP 3 production (19,88). The present study shows that Ca 2ϩ entry, related to SOCE or as a consequence of hemichannel opening (or a combination of both), is more prominent with BK than with ATP as a stimulus.
Ca 2ϩ -mediated feedback actions residing at different levels of the Ca 2ϩ signaling cascade (GPCR, InsP 3 R, or PLC) lie at the basis of Ca 2ϩ oscillations (10, 18 -23). Ca 2ϩ activation of hemichannel-dye uptake and ATP release is, like the InsP 3 R open probability, characterized by a bell-shaped [Ca 2ϩ ] i response curve (35,36) (Fig. 7D) (37). Ca 2ϩ -dependent activation of hemichannels has been proposed to involve binding of the C-terminal tail to the intracellular loop of Cx43. Conversely, hemichannel inhibition at high [Ca 2ϩ ] i is mediated by Ca 2ϩ activation of actomyosin that disrupts the loop/tail interaction. The addition of CT9 peptide will substitute for the disrupted C-terminal tail binding by interacting with the cytoplasmic loop and thereby preventing high [Ca 2ϩ ] i inhibition of hemichannels (37). Here, we show that CT9 but not the reversed sequence inhibits Ca 2ϩ oscillations induced by BK. CT9 inhibition of BK-induced oscillations is mediated by an effect on Cx43 but not on Cx32. Unfortunately, a CT9 analog for Cx32 is not yet available. Both Cx43 and Cx32 appear to be essential in the BK-triggered Ca 2ϩ oscillations, as interfering with each of these connexins individually prevented the oscillations. This is in line with the fact that hemichannels composed of Cx32 also have, like those composed of Cx43, a bell-shaped [Ca 2ϩ ] i dependence.
Mathematical modeling has demonstrated that positive and negative feedback on voltage-gated Ca 2ϩ entry is by itself sufficient to generate Ca 2ϩ oscillations (89). In contrast, the present experiments show that a forced [Ca 2ϩ ] i change does not induce Ca 2ϩ oscillations, indicating that the Ca 2ϩ -sensitive and Ca 2ϩ entry-mediating hemichannels are not sufficient as an oscilla-tion mechanism. Consistent with this is the observation that BK-triggered Ca 2ϩ oscillations largely rely on InsP 3 Rs (Fig. 8). CytC removes InsP 3 R inactivation at high [Ca 2ϩ ] i , giving increased ER Ca 2ϩ release leading to a gradually increasing [Ca 2ϩ ] i , mitochondrial Ca 2ϩ overload, and apoptosis (58). We did not observe [Ca 2ϩ ] i elevation after CytC application, but this is probably related to the relatively short (10 min) time window used to record Ca 2ϩ oscillations. Additionally, high [Ca 2ϩ ] i inhibition of InsP 3 Rs is not the only OFF mechanism mediating [Ca 2ϩ ] i restoration to the resting level; Ca 2ϩ pumps, like sarcoplasmic/endoplasmic and plasma membrane Ca 2ϩ ATPases, which are not affected by CytC (58), will ensure proper maintenance of normal [Ca 2ϩ ] i within the 10-min time frame of our recordings. Importantly, our data are the first demonstration in intact cells that CytC promotes ER Ca 2ϩ release, supporting its proposed action as an antagonist of high [Ca 2ϩ ] i inhibition of InsP 3 Rs (38). CytC also stimulated mitochondrial Ca 2ϩ uptake, but this may be a direct consequence of the increased ER Ca 2ϩ release.
Interestingly, ATP-induced oscillations were also inhibited by CytC but not by CT9 peptide, indicating that these oscillations thrive exclusively on InsP 3 -based mechanisms. Thus, BKtriggered oscillations are based on InsP 3 Rs with an additional hemichannel component. Because hemichannels can be activated by moderate (Ͻ500 nM) [Ca 2ϩ ] i elevation, we hypothesize that these channels open with each Ca 2ϩ spike and contribute with Ca 2ϩ entry during the rising phase. When the spike amplitude rises above 500 nM, hemichannels close again and contribute with an OFF mechanism that adds to the OFF mechanism of InsP 3 R channels. The fact that hemichannels are by themselves not sufficient to mediate oscillations probably results from a slower kinetics for opening than for closing. Previous work has indeed suggested that hemichannel activation by Ca 2ϩ is characterized by several intermediate signaling steps pointing to slow activation kinetics (36). A point of notice is that the duration of the primary Ca 2ϩ peak was different for both triggers: ϳ40 s for 0.5 M BK and 15 s for 10 M ATP (see example traces in Fig. 1). Preliminary modeling using the activation kinetics of hemichannel opening presented in Refs. 35 and 36 indeed indicates that a [Ca 2ϩ ] i rise of 40 s (BK) can trigger the opening of twice as much hemichannels compared with a peak that lasts only 15 s (ATP). Additionally, the Ca 2ϩ spikes of BK-triggered oscillations were generally of longer duration than those triggered by ATP (see example traces in Fig. 1) making it possible that these Ca 2ϩ transients were more efficient in inducing hemichannel opening. We speculate that the more prominent contribution of SOCE with BK stimulation is more effective in triggering hemichannel opening because it occurs more localized and in close proximity of the hemichannels in the plasma membrane.
Our experiments indicate that hemichannels actively contribute to BK-induced Ca 2ϩ oscillations by providing a Ca 2ϩ entry pathway characterized by a bimodal [Ca 2ϩ ] i dependence. This contribution is crucial as inhibition of this pathway blocks the oscillations. Several compounds like the InsP 3 R inhibitor 2-APB used to explore the mechanism of Ca 2ϩ oscillations (90) are known to inhibit connexin hemichannels (91). Additionally, the non-selective Ca 2ϩ channel blocker La 3ϩ has frequently been used to investigate Ca 2ϩ entry during Ca 2ϩ oscillations, also in cells that express connexins (92)(93)(94), but these trivalent ions also inhibit hemichannels (31,95,96). In conclusion, connexin hemichannels may contribute to Ca 2ϩ oscillations in connexin-expressing cells. Interestingly, connexin-expressing cells may also display hemichannel-independent Ca 2ϩ oscillations, as exemplified here with ATP. This differential contribution of hemichannels to Ca 2ϩ oscillations may result in distinct downstream response patterns to this fundamental cell signal, as recently observed in brain endothelial cells (34).