Intercellular Calcium Signaling in Astrocytes via ATP Release through Connexin Hemichannels*

Astrocytes are capable of widespread intercellular communication via propagated increases in intracellular Ca2+ concentration. We have used patch clamp, dye flux, ATP assay, and Ca2+ imaging techniques to show that one mechanism for this intercellular Ca2+ signaling in astrocytes is the release of ATP through connexin channels (“hemichannels”) in individual cells. Astrocytes showed low Ca2+-activated whole-cell currents consistent with connexin hemichannel currents that were inhibited by the connexin channel inhibitor flufenamic acid (FFA). Astrocytes also showed molecular weight-specific influx and release of dyes, consistent with flux through connexin hemichannels. Transmembrane dye flux evoked by mechanical stimulation was potentiated by low Ca2+ and was inhibited by FFA and Gd3+. Mechanical stimulation also evoked release of ATP that was potentiated by low Ca2+ and inhibited by FFA and Gd3+. Similar whole-cell currents, transmembrane dye flux, and ATP release were observed in C6 glioma cells expressing connexin43 but were not observed in parent C6 cells. The connexin hemichannel activator quinine evoked ATP release and Ca2+ signaling in astrocytes and in C6 cells expressing connexin43. The propagation of intercellular Ca2+ waves in astrocytes was also potentiated by quinine and inhibited by FFA and Gd3+. Release of ATP through connexin hemichannels represents a novel signaling pathway for intercellular communication in astrocytes and other non-excitable cells.

the expression of connexins (the proteins that comprise gap junctions) in multiple cell types. For example, C6 glioma cells express low levels of connexins and do not exhibit intercellular Ca 2ϩ wave propagation. Expression of connexin43 (Cx43, the predominant connexin in astrocytes) in C6 cells is correlated with both intercellular dye transfer (3,4) and propagated Ca 2ϩ waves (5,6). Studies with other cell types have shown similar correlation of Ca 2ϩ wave propagation with connexin expression (7). In addition, pharmacological inhibitors of connexin channels have been shown to inhibit intercellular Ca 2ϩ wave propagation in astrocytes (1,8). These data show that connexin channel expression and function play a central role in Ca 2ϩ wave propagation.
Recent studies (9,10) have also clearly shown that the intercellular propagation of Ca 2ϩ signals in astrocytes involves the diffusion of an extracellular messenger, namely ATP. Ca 2ϩ waves in astrocytes induced by mechanical or electrical stimuli are associated with ATP release, and intercellular Ca 2ϩ waves are blocked by purinergic receptor antagonists (9,11). Multiple other stimuli that evoke astrocyte signaling, including glutamate and UTP, also evoke release of ATP (5,12).
The requirement for connexins and the involvement of ATP as a messenger in astrocyte Ca 2ϩ wave propagation seem to indicate paradoxically both gap junctional and extracellular messenger-mediated intercellular communication. However, in addition to docking with connexins in neighboring cells, connexins form "hemichannels" or "connexons" that exist independently within an individual cell (13)(14)(15)(16)(17). Similar to other membrane channels, connexin hemichannels are transmembrane channels that connect the cytoplasm and the extracellular space. Astrocytes in culture have been reported to express connexin hemichannels that allow transmembrane dye flux (13). These channels may provide a pathway for release of small signaling molecules, such as ATP, to the extracellular space. Roles for both an extracellular messenger and connexins in intercellular Ca 2ϩ signaling could therefore be reconciled by the hypothesis that the ATP that mediates Ca 2ϩ waves is released through connexin hemichannels.
Consistent with this hypothesis, Cotrina et al. (5) reported that ATP release from C6 glioma cells is increased in direct correlation with the level of forced connexin expression. Although these studies raised the possibility of a role for connexin channels in ATP release, they also found that ATP release was not inhibited by concentrations of octanol that were sufficient to inhibit gap junction channels as evidenced by dye coupling. Although this result would seem to contradict a direct role for connexin hemichannels in ATP release, it is possible that inhibitors of gap junctional coupling may not have the same effects or potency as inhibitors of connexin hemichannels. Bruzzone et al. (29) reported that transmembrane flux of NAD ϩ occurs via connexin43 hemichannels, indicating that the hemichannels composed of the primary astrocyte connexin are permeable to nucleotides. In the present study, we have used dye transfer and electrophysiological techniques to confirm the presence of connexin hemichannels in astrocytes. We have then used Ca 2ϩ imaging and ATP measurements to show that the extent of intercellular Ca 2ϩ waves and ATP release in astrocytes is correlated with the function of connexin hemichannels.

EXPERIMENTAL PROCEDURES
Cell Culture-Mixed glial cultures from postnatal day 1-5 mouse pups were prepared using the methods described previously (18). To prepare purified astrocyte cultures, flasks were shaken daily for 3 days for 1 h on a shaker at 37°C, and displaced cells were removed. Astrocytes were passaged onto poly-L-lysine-coated glass coverslips, 35-mm diameter culture dishes, or culture flasks at 50,000 cells/cm 2 and were maintained in growth medium (Dulbecco's modified Eagle's medium with 10% fetal calf serum, penicillin, and streptomycin).
A line of C6 glioma cells stably expressing connexin43 (Cx43-13 clone) that has been described in previous studies (4,6) was used along with non-transfected C6 cells to examine the functional roles of con-nexin43. C6 cells were passaged from flasks onto glass coverslips or 35-mm culture dishes and maintained in the same medium as described for astrocytes. Cells were grown for 1-3 days to a confluence of 70 -90% prior to experimentation.

Measurement of [Ca 2ϩ ] i -[Ca 2ϩ
] i was measured using a fluorescence imaging system that has been described previously in detail (18). In brief, cells on glass coverslips were loaded with fura2 by incubation in 5 M fura2-AM for 30 min. Coverslips were then placed on a Nikon inverted microscope and excited with a mercury lamp through 340 and 380 nm bandpass filters, and fluorescence at 510 nm was recorded through a ϫ20 objective with a SIT camera to a video tape recorder. Video images were digitized using an Axon Image Lightning board and Image WorkBench software, and ⌬F was calculated on a pixel-by-pixel basis as described previously.
Mass Mechanical Stimulation and Collection of Extracellular Samples-Mouse cortical astrocytes plated on 35-mm plastic Petri dishes (grown for 7-10 days) were rinsed with control Hanks' balanced salt solution (HBSS) and allowed to equilibrate (25 min). Then each plate was rinsed (3 times) with control HBSS (1 mM EGTA, 1 mM free Mg 2ϩ , 1.3 mM free Ca 2ϩ ) and bathed (1.5 ml) for 5 min, and at which time a sample was collected (500 l; control spontaneous release sample) and rapidly frozen. At this point the medium was changed to treatment HBSS. A sample was taken at 4.5 min, after which mass mechanical stimulation was achieved by dropping ϳ110 g of glass microBeads (30 -50 M diameter) through the ϳ3 mm of medium onto the cells. This type of stimulation has been shown to initiate multifocal calcium waves by mechanical stimulation of individual cells (9). All direct comparisons were made in sister cultures plated to the same density.
ATP Measurement and Analysis-ATP was measured using a luciferin/luciferase bioluminescence assay (Molecular Probes) and a fluorescence plate reader (Wallac VictorV Dye and LDH Release Measurement-For dye release assays, cells were gently rinsed (3 times; 1.5 ml) with control HBSS and then incubated with control HBSS supplemented with Calcein Blue AM (10 M) for 30 min. Some control plates were also loaded with Oregon Green BAPTA 1 by supplementing the dye loading solution with Oregon Green BAPTA 1 AM (5 M, Molecular Probes, Eugene OR). Cells were then rinsed (3 times; 1.5 ml) and incubated for an additional 30 min to allow for deesterification, after which they were placed in treatment HBSS (1.5 ml). After 4.5 min a sample (500 l for this and all subsequent samples) was collected as a base line. To control for fluorescence of components of treatment HBSS (e.g. Ca 2ϩ or Mg 2ϩ ), 250 l of treatment HBSS was added to base-line samples. Mass mechanical stimulation was then performed by adding medium containing microBeads (110 g in 500 l of HBSS) to the cells. After 40 s, medium samples were collected. The extracellular solution was then replenished, and the cells were lysed using a cell scraper. This was followed by the final sample collection. All collections were immediately put on ice and rapidly frozen.
Samples where I stim is measured intensity of the sample collected after mechanical stimulation. I spr is the measured intensity of the spontaneous release sample (i.e. first collected sample). I lysate is the measured intensity of the sample collected after the cells are lysed, and I bckg is the measured intensity of the treatment solution. Samples collected as described above were also analyzed for lactate deyhdrogenase (LDH) using a coupled enzymatic colorimetric assay in 96-well plates (Cytotox 96, Promega, Madison, WI). Cell lysate samples were used as positive controls for LDH.
Electrophysiology-Membrane currents were measured using the whole-cell configuration. The bath solution consisted of Hanks' balanced salt solution (Ca 2ϩ -and Mg 2ϩ -free) supplemented with HEPES (10 mM), EGTA (1 mM), Ca 2ϩ and Mg 2ϩ (pH 7.3 with NaOH). The pipette solution contained (in mM) the following: 140 KCl, 10 EGTA, and 10 HEPES (pH 7.2 with KOH). Pipette tips were fire-polished using a heated platinum filament to a tip resistance of 3-5 megohms. Patch clamp recordings were performed and analyzed using an Axopatch 200B amplifier and pClamp (Axon Instruments). All data were collected at 10 kHz and analog-filtered at 5 kHz.

RESULTS
Low Ca 2ϩ -activated Currents-We used the whole-cell configuration of the patch clamp technique to determine whether astrocytes display low [Ca 2ϩ ] e -activated currents consistent with connexin hemichannels. In extracellular medium containing normal (1.3 mM) [Ca 2ϩ ], astrocytes displayed variable inward and outward voltage-activated currents. The inhibition of the inward and outward currents by Cs 2ϩ and Ba 2ϩ was consistent with these currents being primarily due to inward rectifier and voltage-activated K ϩ channels (data not shown) as has been described previously in astrocytes (19,20). Low [Ca 2ϩ ] e -activated currents were more easily quantified in cells with low levels of base-line currents (Fig. 1).
Perfusion of astrocytes with medium containing no added Ca 2ϩ (with 1 mM EGTA) resulted in a base-line current at a holding potential of Ϫ50 mV, as well as a significant increase in the amplitude of both inward and outward currents evoked by voltage steps from this holding potential (Fig. 1). The average ratio of the maximum steady-state inward current in low [Ca 2ϩ ] e versus normal [Ca 2ϩ ] e was 1.56 Ϯ 0.17 (n ϭ 5). Currents returned to base line when the extracellular solution was exchanged for control solution. Approximately 60% of isolated astrocytes showed low [Ca 2ϩ ] e -activated currents. Flufenamic acid (FFA) has been reported to block currents through connexin hemichannels expressed in oocytes with high potency (21) and to inhibit Cx43-mediated intercellular communication (22). Application of 50 M FFA immediately abolished low [Ca 2ϩ ] e -activated currents (average ratio of maximum current in low [Ca 2ϩ ] e with FFA versus normal [Ca 2ϩ ] e was 0.97 Ϯ 0.02, n ϭ 3). The significant inhibition of low [Ca 2ϩ ] e -activated currents (p Ͻ 0.02) by FFA was partially reversible (ϳ70%; data not shown) with washout. C6 cells expressing connexin43 (C6 Cx43 cells, n ϭ 6), but not untransfected C6 cells (n ϭ 10), showed FFA-sensitive whole-cell currents similar to those observed in primary astrocytes (data not shown).
Dye Uptake-Transmembrane flux of low molecular weight dyes has been used to indicate the presence of connexin hemichannels in multiple cell types including astrocytes (13,14). We examined the pattern of Lucifer Yellow (LY) and rhodamine dextran (RD) (10 kDa) flux into astrocytes under conditions of normal and low [Ca 2ϩ ] e . Cells maintained in HBSS with normal [Ca 2ϩ ] containing LY (1 mg/ml) and RD (0.5%) showed little uptake of either dye. When cells were exposed to nominally Ca 2ϩ -free medium containing the same concentrations of LY and RD, they showed small but significant uptake of LY, but not of RD ( Fig. 2A). Similar uptake of LY in astro-cytes has been reported previously (13). Subsequent exposure of cells to Ca 2ϩ -free medium containing no dye caused release of the LY (Fig. 2B).
To determine whether uptake of LY was activated in association with intercellular Ca 2ϩ waves, we observed the pattern of LY uptake in the region of a mechanically stimulated cell. Mechanical stimulation of a single cell reproducibly elicits an intercellular Ca 2ϩ wave that is propagated from the stimulated cell to neighboring cells. Coverslips of mouse astrocytes were bathed in HBSS containing LY and RD, and single cells were mechanically stimulated using a glass micropipette. Mechanical stimulation resulted in the uptake of Lucifer Yellow in the stimulated cell as well as in immediately adjacent cells, suggesting that this uptake occurred in association with an intercellular Ca 2ϩ wave ( Fig. 2C; n ϭ 12). The uptake of LY by both the stimulated cell and the surrounding cells was greatly potentiated by removal of extracellular Ca 2ϩ . Although it is possible that some cell-cell spread of the LY occurred due to diffusion through gap junctions, we did not observe cell-cell spread of dye after rinsing of LY from the medium, suggesting that such diffusion was minimal. There was no uptake of RD in either the stimulated cell or surrounding cells, indicating that there was not nonspecific disruption of the cell membrane.
Dye Release-To characterize dye flux further, plates of cells were loaded with a connexin channel-permeable dye (Calcein Blue, approximate M r 400) and a connexin channel-impermeable dye (Oregon Green 1 BAPTA 488, approximate M r 1100) by incubating cells with the respective AM ester for 20 min. Dye-loaded cells were mechanically stimulated by dropping glass microbeads (30 -50 M in diameter) through the medium onto the cells. This mass mechanical stimulation has been found to initiate multifocal calcium waves without causing nonspecific disruption of the cell membrane (9).
Mass mechanical stimulation of astrocytes initiated significant release of Calcein Blue but not Oregon Green or LDH. Mechanical stimulation evoked release of 7.8% of intracellular Calcein Blue (Ϯ1.6% S.E., n ϭ 16) and 0.13% of Oregon Green (Ϯ0.25% S.E., n ϭ 8). There was no detectable release of LDH in response to mechanical stimulation under any condition (n ϭ 32, cell lysates were used as positive control). Calcein release was modulated by connexin channel modulators (Fig. 3). Low [Ca 2ϩ ] e medium evoked dye release in un-stimulated cells (not shown). Dye release evoked by mass mechanical stimulation was potentiated by low [Ca 2ϩ ] e medium and was inhibited by FFA and by Gd 3ϩ , which has also been shown to inhibit connexin hemichannels with high potency (21) (Fig. 3).
Identical stimulation of C6 glioma cells, which express very low levels of connexins, evoked a significantly reduced level of Calcein Blue release (2.86 Ϯ 1.3%, n ϭ 8), whereas C6 glioma cells with exogenously expressed connexin43 (C6 Cx43 cells) showed levels of Calcein Blue (9.6 Ϯ 1.3%, n ϭ 16) release that were comparable with or higher than that shown by astrocytes. As with primary astrocytes, Calcein Blue release from C6 Cx43 cells was potentiated by low [Ca 2ϩ ] e medium and was inhibited by FFA and Gd 3ϩ (Fig. 3). C6 Cx43 cells did not release Oregon Green in response to mechanical stimulation (not shown).
ATP Release-Mechanical stimulation initiates rapid release of ATP from astrocytes (9,11). We found that ATP release evoked by mechanical stimulation with glass microbeads could be detected in as little as 10 s following stimulation.
[ATP] e increased for the first ϳ40 s after mechanical stimulation and remained elevated throughout the time frame of the measurement (5 min; data not shown). We took samples at 30 s to reflect ATP release associated with intercellular Ca 2ϩ waves.
Treatment with low [Ca 2ϩ ] e medium significantly potenti-ated the ATP release induced by mechanical stimulation (Fig.  4). Conversely, the connexin channel inhibitors FFA and Gd 3ϩ significantly inhibited mechanically induced ATP release (Fig.  4). Because FFA is a chloride channel blocker, we also examined the effects of another chloride channel blocker DIDS, which does not inhibit connexin channels with high potency. DIDS did not inhibit ATP release by astrocytes (Fig. 4).
Cotrina et al. (5) reported previously that ATP release evoked by UTP was correlated with connexin expression in C6 cells. We found that mechanically induced ATP release was also correlated with connexin43 expression. Non-transfected C6 cells showed very low levels of ATP release in response to mass mechanical stimulation, whereas C6 Cx43 cells showed ATP release that was potentiated by low Ca 2ϩ , and inhibited by FFA but not DIDS, and Gd 3ϩ (Fig. 4). The inhibitory effect of FFA was concentration-dependent. In the presence of 25 M FFA, ATP release was 69 Ϯ 9% of control release, whereas in the presence of 50 or 100 M FFA ATP release was 10 Ϯ 6% of control release (no significant difference between the effect of 50 and 100 M FFA).
Calcium Signaling-We have reported previously that intercellular Ca 2ϩ waves are correlated with the level of exogenous connexin43 expression in C6 cells (6), and that reduction of [Ca 2ϩ ] e elicits intracellular and intercellular Ca 2ϩ signaling in astrocytes (26). Based upon the results described above, this effect could be explained by "unblocking" of connexin hemichannels with subsequent ATP release. Bath application of quinine evoked multifocal intercellular Ca 2ϩ waves in astrocytes (Fig. 5, n ϭ 5 coverslips). These intercellular Ca 2ϩ waves occurred within 3-5 min following application of quinine and immediately ceased upon washout of quinine. Quinine also evoked Ca 2ϩ oscillations and limited intercellular Ca 2ϩ waves in C6 Cx43 cells but did not evoke Ca 2ϩ signaling in parent C6 cells (Fig. 5, n ϭ 5 coverslips each). The Ca 2ϩ signaling responses of C6 Cx43 cells to quinine were inhibited (but not abolished) by apyrase or by the purinergic receptor antagonist pyridoxal phosphate-6-azophenyl-2Ј,4Ј-disulfonic acid (PPA-DS), indicating a role for ATP release in this response. In the presence of 50 units/liter apyrase, 100 M quinine evoked a significant (Ͼ100 nM) increase in [Ca 2ϩ ] i in an average of 14 Ϯ 4% of C6 Cx43 cells compared with an average response of 81 Ϯ 8% of cells after washout of apyrase (n ϭ 320 cells in four experiments). In the presence of 10 M PPADS, 100 mM quinine evoked a significant increase in [Ca 2ϩ ] i in an average of 30 Ϯ 8% of C6 Cx43 cells compared with 78 Ϯ 6% of controls in the absence of PPADS (n ϭ 300 cells in three experiments). As reported by others previously (5,27), we found that both C6 cells and C6Cx43 cells showed increases in [Ca 2ϩ ] i in response to bath application of ATP with similar sensitivity. In agreement with previous reports, we found that the ATP concentration at which 50% of cells responded was 2 M for both C6 and C6Cx43 cells (n ϭ 3 coverslips each).
Brief deformation of the membrane of a single cell initiates an increase in [Ca 2ϩ ] i in the stimulated cell that is propagated in a wave-like manner to surrounding cells (18) and Fig. 6. The extent of Ca 2ϩ wave propagation was significantly decreased by 50 M FFA and 50 M Gd 3ϩ but not DIDS. Conversely, the extent of wave propagation was significantly increased in medium containing quinine (Fig. 6C). DISCUSSION Astrocytes, like multiple other cell types, are capable of widespread communication via intercellular Ca 2ϩ waves. Although initial studies (1, 6) identified gap junctions as a pathway for communication of Ca 2ϩ waves, more recent studies (9, 10) have demonstrated an extracellular signaling pathway involving release of ATP and activation of purinergic receptors. Our results provide evidence that connexin hemichannels provide a pathway for this ATP release.
First, we found that astrocytes express functional connexin hemichannels as evidenced by whole-cell patch clamp studies. Activation by low [Ca 2ϩ ] e is characteristic of currents through connexin hemichannels (17,21). The low [Ca 2ϩ ] e -activated currents we observed in astrocytes are similar to those observed through connexin hemichannels expressed in Xenopus oocytes (17). Like connexin channel currents in oocytes, the low Ca 2ϩactivated currents in astrocytes were reversibly inhibited by similar concentrations of flufenamic acid (21,28). Although currents through connexin43 hemichannels have not been reported in oocyte models, our results indicate that expression of connexin43 in C6 cells results in the appearance of currents in single cells that are consistent with currents observed through gap junctions composed of connexin43.
Transmembrane flux of low molecular weight dyes is a commonly used method for assessing the presence and function of connexin hemichannels. Consistent with previous reports (13), we found that astrocytes show uptake of Calcein Blue or Lucifer Yellow, both low molecular weight dyes, but not the high molecular weight dye rhodamine dextran. In addition, we observed release of AM ester-loaded dyes in response to removal of extracellular Ca 2ϩ . Mechanical stimulation of astrocytes also evoked both uptake and release of low molecular weight dyes but not release of higher molecular weight dyes (Oregon Green BAPTA 1) or LDH. Mechanical stimulation evoked release of Calcein Blue from C6 Cx43 cells but not from non-transfected C6 cells. Dye release was inhibited by flufenamic acid and Gd 3ϩ , both of which are potent inhibitors of connexin hemichannels. In summary, the size selectivity, activation by removal of extracellular Ca 2ϩ , correlation with connexin43 ex- pression in C6 cells, and inhibition by connexin channel inhibitors are all consistent with transmembrane dye flux via connexin hemichannels.
Because intercellular Ca 2ϩ signaling induced by mechanical stimulation and other stimuli has been associated with ATP release (5,9,11), we examined the role of connexin hemichannels in mechanically induced ATP release. Cotrina et al. (5) reported previously a correlation of UTP-induced ATP release with connexin expression in glial cells. We found a similar correlation of mechanically induced ATP release with connexin expression in C6 glioma cells. We also found that low extracellular Ca 2ϩ and quinine evoke spontaneous ATP release and potentiate mechanically induced ATP release from astrocytes and C6 glioma cells expressing connexin43, whereas inhibitors of connexin hemichannels including FFA and Gd 3ϩ inhibited mechanically induced ATP release. Based upon these results, we conclude that mechanically induced release of ATP involves connexin hemichannels.
Modulators of connexin hemichannels had consistent effects on intercellular Ca 2ϩ signaling in astrocytes. We have reported previously that reduction of extracellular Ca 2ϩ evokes intercellular Ca 2ϩ signaling in astrocytes (26). This finding can now be explained by the unblocking of connexin hemichannels and subsequent ATP release to mediate intercellular signaling. Consistent with this hypothesis, the connexin hemichannel activator quinine evoked ATP release and intercellular Ca 2ϩ signaling in astrocytes and C6 cells expressing connexin43, and mechanically induced intercellular Ca 2ϩ waves in astrocytes were potentiated by quinine. Conversely, the extent of intercellular Ca 2ϩ wave propagation was significantly reduced by the same concentrations of connexin hemichannel inhibitors that inhibited wholecell currents, transmembrane dye flux, and release of ATP.
Cotrina et al. (5) raised the possibility of connexin hemichannels as a pathway for release of ATP and intercellular Ca 2ϩ signaling. However, their observations that octanol did not block ATP release at concentrations sufficient to block dye coupling seemed to contradict this possibility. Our results indicate that there is a consistent correlation between modulation of connexin hemichannel currents, low molecular weight dye release, and ATP release. It is possible that inhibitors of gap junctions such as octanol act with different potency and specificity on intercellular gap junction channels versus individual connexin hemichannels. Quinine represents an example of an agent with distinct effects on connexin hemichannel function versus gap junctional coupling; although quinine activates connexin hemichannels, it does not have a parallel effect on gap junctional conductance (23)(24)(25).
We used flufenamic acid and Gd 3ϩ as inhibitors of connexin hemichannels, because both have been shown to inhibit connexin hemichannels in oocytes with high potency (21), and neither evoked increases in [Ca 2ϩ ] i . Flufenamic acid has also been reported to be a potent and reversible blocker of gap junctions (22). By contrast, we found that other traditional gap junction inhibitors such as octanol, glycyrrhetinic acid, or oleamide all evoked Ca 2ϩ signaling that included elevations of base-line Ca 2ϩ , Ca 2ϩ oscillations, and intercellular Ca 2ϩ waves (not shown). The mechanism for the activation of Ca 2ϩ signaling by these agents is not known. Regardless of the mechanism, however, these increases in [Ca 2ϩ ] i confounded interpretation of their effects regarding a role for connexin channels.
Although both FFA and Gd 3ϩ also inhibit other channels as well as connexin channels, their effects with multiple different assays of connexin channel function support the interpretation that their effects in the present study are primarily due to Conversely, the extent of wave propagation was significantly increased in medium containing quinine (n ϭ 10, p Ͻ 0.0001). Error bars represent maximum variation in standard error of ratio ((condition ϩ S.E.)/(control Ϫ S.E.) Ϫ condition/control). connexin channel inhibition. In addition, we found that other Cl Ϫ channel blockers such as DIDS that do not inhibit connexin hemichannels (25,29) did not inhibit intercellular Ca 2ϩ signaling or ATP release. Quinine is also well known to have effects on other ion channels. However, the activation of Ca 2ϩ signaling only in C6 cells expressing connexin43, and not in nontransfected C6 cells, is consistent with the activation of connexin hemichannels, as has been reported previously (23)(24)(25) for multiple types of connexins.
The correlation of results of electrophysiological studies, dye flux measurements, ATP release measurements, and imaging of intercellular Ca 2ϩ signaling provides strong evidence for the hypothesis that intercellular Ca 2ϩ waves in astrocytes can occur by ATP release through connexin hemichannels. The parallel results obtained only with C6 cells expressing con-nexin43, and not with non-transfected C6 cells, corroborates this hypothesis. A primary role for connexin hemichannels in intercellular Ca 2ϩ signaling does not exclude other potential intercellular pathways, such as gap junctions. It is also possible that connexin expression is associated with up-regulation of some other pathway for ATP release. However, ATP release via connexin hemichannels is the most direct explanation for our results and provides a direct mechanism that reconciles both a role for connexins and a role for extracellular ATP in intercellular Ca 2ϩ waves.
Astrocyte Ca 2ϩ waves have been implicated in a variety of physiological and pathological processes. They have been shown to modulate synaptic signaling between neurons, suggesting a role in synaptic plasticity (30 -32). Haydon and colleagues (30,31,33) have provided evidence that this modulation of synaptic function is mediated by astrocytic release of glutamate, which they have shown to occur in association with Ca 2ϩ waves. Astrocyte Ca 2ϩ waves have also been shown to modulate the response of the retinal neurons to light stimulation (34). A role for astrocyte Ca 2ϩ waves has also been proposed in migraine headache and the spread of seizures (2). The identification of connexin hemichannels as a pathway for stimulus-evoked release of small molecules by our studies and others (29) raises the possibility that in addition to ATP, other intercellular messengers may also be released via this pathway. Further investigation of this mechanism of intercellular signaling may provide an opportunity for greater understanding of the functional significance of Ca 2ϩ waves in astrocytes and other cell types.