Colocalization of ATP Release Sites and Ecto-ATPase Activity at the Extracellular Surface of Human Astrocytes*

Extracellular ATP and other nucleotides function as autocrine and paracrine signaling factors in many tissues. Recent studies suggest that P2 nucleotide receptors and ecto-nucleotidases compete for a limited pool of endogenously released nucleotides within cell surface microenvironments that are functionally segregated from the bulk extracellular compartment. To test this hypothesis, we have used luciferase-based methods to continuously record extracellular ATP levels in monolayers of human 1321N1 astrocytoma cells under resting conditions, during stimulation of Ca2+-mobilizing receptors for thrombin or acetylcholine, and during mechanical stimulation by hypotonic stress. Soluble luciferase was utilized as an indicator of ATP levels within the bulk extracellular compartment, whereas a chimeric protein A-luciferase, adsorbed to antibodies against a glycosylphosphatidylinositol-anchored plasma membrane protein, was used as a spatially localized probe of ATP levels at the immediate extracellular surface. Significant accumulation of ATP in the bulk extracellular compartment, under either resting (1–2 nm ATP) or stimulated (10–80 nm ATP) conditions, was observed only when endogenous ecto-ATPase activity was pharmacologically inhibited by the poorly metabolizable analog, βγ-methylene ATP. In contrast, accumulation of submicromolar ATP in the cell surface microenvironment was readily measured even in the absence of ecto-ATPase inhibition suggesting that the spatially colocalized luciferase could effectively compete with endogenous ecto-ATPases for released ATP. Other experiments revealed a critical role for elevated cytosolic [Ca2+] in the ATP release mechanism triggered by thrombin or muscarinic receptors but not in basal ATP release or release stimulated by hypotonic stress. These observations suggest that ATP release sites are colocalized with ecto-ATPases at the astrocyte cell surface. This colocalization may act to spatially restrict the actions of released ATP as a paracrine or autocrine mediator of cell-to-cell signaling.

ATP and other nucleotides function as intercellular signaling molecules when released to extracellular compartments. Genes encoding seven ionotropic P2X nucleotide receptor subtypes (1), eight G protein-coupled P2Y nucleotide receptor subtypes (2,5), and at least nine different ecto-nucleotidases (3,4) have been identified in human and other vertebrate genomes. Most mammalian cell types express one or more subtypes of nucleotide receptor together with various combinations of the ectonucleotidases used for degrading and/or interconverting extracellular nucleotides (5). Recent studies using knockout mice that lack expression of particular nucleotide receptors or ectonucleotidases have suggested important in vivo roles for extracellular ATP or other nucleotides in a variety of inflammatory, nociceptive, hemostatic, and motility responses (6 -12).
The physiological sources of the extracellular nucleotides that elicit these complex tissue responses remain largely uncharacterized. A major exception is the well established ability of neurons, neuroendocrine cells, and platelets to release ATP via Ca 2ϩ -regulated exocytosis of nucleotides compartmentalized within synaptic vesicles or dense core granules (13). However, many cell types that express nucleotide receptors lack direct physical proximity to neurons or degranulating platelets. Thus, identification of alternative sources of extracellular nucleotides, as well as elucidation of additional cellular mechanisms that underlie the release of nucleotides, are significant areas of current investigation (14). Because all cells contain ATP, all cell types are potential sources of extracellular ATP given appropriate physiological or pathological stimuli. Two recurring themes have emerged in recent studies of ATP release from cell types other than neurons or platelets. First, intact cells constitutively release ATP at low rates by mechanisms that do not involve obvious lysis or loss of plasma membrane integrity (15)(16)(17). Second, the rate of non-lytic ATP release can be greatly increased by various extrinsic stimuli that result in perturbation of cell volume, shape, cytoskeletal organization, or intracellular Ca 2ϩ homeostasis.
Several factors complicate or limit studies aimed at the direct quantitative evaluation of physiologically relevant ATP release. Cells express multiple ecto-nucleotidases that can rapidly hydrolyze ATP released at the extracellular surface (3,4). Recent studies have indicated that ATP may also be synthesized from extracellular ADP via transphosphorylation reactions catalyzed by extracellular forms of nucleotide disphosphokinase (NDPK) 1 or adenylate kinase (17)(18)(19)(20). Although released or synthesized ATP will initially be spatially confined within unstirred layers at the cell surface, it will subsequently be diluted by diffusion and/or convection into the bulk extra-cellular volume. For these reasons, ATP concentrations at the extracellular surface will change rapidly and reflect a balance between the rates of release, hydrolysis, synthesis, and dilution into interstitial compartments (in vivo) or bathing media (in vitro). Most studies of ATP release have involved removal of samples from the bulk extracellular media bathing cultured cells or isolated tissues at selected time points following stimulation; this is followed by subsequent analysis of ATP or ATP-derived metabolites within these media samples (14 -20). The ATP levels measured in such bulk extracellular media samples are likely to significantly underestimate the amount of ATP actually released at the cell surface, particularly in cells/ tissues with substantial unstirred surface layers or high ecto-ATPase activities. For example, assays based on second messenger accumulation in astrocytes and epithelial cells indicate that non-lytic mechanical stimuli trigger the release of ATP and other nucleotides in amounts sufficient for robust autocrine (or paracrine) activation of various G protein-coupled P2Y nucleotide receptors (15)(16)(17). However, the concentrations of nucleotide directly measured in the media bathing these stimulated cells are orders of magnitude lower than the concentrations required for threshold activation of known P2Y receptors. This suggests that nucleotide receptors and ecto-nucleotidases compete for a limited pool of endogenously released nucleotides within cell surface microenvironments that are functionally segregated from the bulk extracellular compartment.
To address this possibility, we have used luciferase-based methods to continuously record ATP levels in extracellular compartments of human 1321N1 astrocytoma cell monolayers under resting conditions, during activation of Ca 2ϩ -mobilizing receptors, or during mechanical stimulation by hypotonic stress. Soluble luciferase was used as an indicator of ATP levels within the bulk extracellular compartment whereas a surfaceadsorbed protein A-luciferase was employed as a spatially localized probe of ATP levels at the immediate cell surface. Significant accumulation of ATP in the bulk extracellular compartment, under either resting or stimulated conditions, was observed only when endogenous ecto-ATPase activity was pharmacologically inhibited. In contrast, accumulation of submicromolar ATP in the cell surface microenvironment was readily measured even in the absence of ecto-ATPase inhibition. This indicated that the spatially colocalized luciferase could effectively compete with endogenous ecto-ATPases for released ATP. Other experiments revealed a critical role for elevated cytosolic [Ca 2ϩ ] in the ATP release mechanism triggered by thrombin or muscarinic receptors but not in ATP release stimulated by hypotonic stress. Our observations indicate that ATP release sites are functionally colocalized with ecto-ATPases at the astrocyte cell surface. This colocalization may act to spatially restrict the actions of released ATP as a paracine or autocrine mediator of cell-to-cell signaling.
Assay of ATP Levels in the Bulk Extracellular Media Compartment by Soluble Luciferase-An online luciferase-based protocol for measuring extracellular ATP levels in cultures of adherent cell monolayers was adapted from methods first developed by Schwiebert and co-workers (21,22) and recently modified by our laboratory (23). All extracellular ATP measurements were performed using a Turner Designs (TD 20/20) luminometer capable of accommodating 35-mm culture dishes. Unless otherwise stated, 35-mm dishes with adherent cells were washed twice with 2 ml of basal saline solution (BSS) containing 130 mM NaCl, 5 mM KCl, 1.5 mM CaCl 2 , 1 mM MgCl 2 , 25 mM Na-HEPES (pH 7.5), 5 mM glucose, and 0.1% bovine serum albumin (BSA). The washed monolayers were then bathed in 1 ml of BSS and incubated for ϳ45 min at room temperature (22-24°C), prior to experimental manipulation. Experiments were performed at room temperature using cell-free or cellcontaining 35-mm dishes treated under identical assay conditions. Lyophilized firefly luciferase ATP assay mix (FL-AAM; Sigma) containing luciferase, luciferin, MgSO 4 , dithiothreitol, EDTA, BSA, and Tricine buffer was reconstituted with 5 ml of sterile filtered water and stored in frozen 500-l aliquots. For experiments, aliquots were thawed at room temperature and diluted 1:25 (40 l) into the 35-mm dishes (Ϯ adherent 1321N1 cells) containing 1 ml of BSS immediately prior to start of luminescence recordings. ATP-dependent changes in extracellular luciferase activity were measured as relative light unit (RLU) values integrated over 5-s photon counting periods with dishes being gently stirred between readings. For most experiments, the luciferase activity was recorded every 1 or 2 min for up to 150 min. ATP contamination and possible effects of various test reagents on luciferase activity were assessed using cell-free 35-mm dishes under identical assay conditions. Calibration curves were generated for each experiment by addition of ATP standards (FL-AAS; Sigma) to cell-free dishes. Under these assay conditions, the limit of ATP detection was 100 pM (100 fmol/1 ml assay volume), and luminescence was linear with increasing ATP concentration to 1000 nM.
Ecto-ATPase Assay-To assess the ecto-ATPase activity of 1321N1 cell monolayers under conditions used for the ATP release experiments, a pulse of 100 nM exogenous ATP was added to cell-free dishes or 1321N1 monolayers containing luciferase assay medium as described above. Decreases in ATP-dependent luminescence were recorded every 2 min for 20 min. Ecto-ATPase inhibition studies were performed using resting 1321N1 monolayers that were incubated with various concentrations of the poorly metabolizable analog ␤␥MeATP 10 to 15 min prior to addition of the 100 nM ATP pulse.
ATP Release Stimuli-After extracellular luciferase activity reached steady state, 1321N1 cells were stimulated for up to 20 min with 2 units/ml thrombin (1 units/ml ϭ 24 nM), 3 M of the SFLLRD-TRAP, 100 M carbachol, or hypotonic stress (a 30% decrease in tonicity) in the absence or presence of ␤␥MeATP (3-300 M). In some experiments, thrombin was tested at concentrations ranging from 0.1 milliunits/ml (2.4 pM) up to 3 units/ml (72 nM) whereas SFLLRD-TRAP was varied from 30 nM through 10 M. Luciferase activity was recorded every 1-2 min during the stimulation period. All additions to the 1-ml ATP assay volumes were made from 100-to 1000-fold concentrated stocks of the various test reagents. To stimulate cells by hypotonic stress, 405 l of BSS was carefully replaced with 405 l of an otherwise identical, luciferase-supplemented assay solution lacking only NaCl; this reduced the NaCl concentration of the extracellular assay medium from 130 to 77 mM. Control experiments involving replacement of 405 l of assay solution with an equivalent volume of fresh, isotonic assay medium verified that this medium replacement protocol produced negligible mechanically induced ATP release. Separate ATP calibrations were performed for the reduced NaCl assay medium to control for the known effects of Cl Ϫ concentration on luciferase activity (51,66). At the end of each assay, the cells were permeabilized with digitonin (50 g/ml) to release cytosolic ATP as a measure of relative cell mass. Potato apyrase (5 units/ml) was used to scavenge released extracellular ATP and provide an index of background light production.
Assay of ATP Levels in the Extracellular Surface Microenvironment by Cell-attached Luciferase-Measurements of ATP in the cell surface microenvironment were performed by antibody-dependent adsorption of a protein A-luciferase chimeric protein (proA-luc) to the extracellular surface of intact 1321N1 cells. proA-luc was expressed in bacteria as described previously (24). Briefly, pMALU5-transformed cultures of JM109 E. coli were lysed by sonication in 150 mM NaCl, 100 mM Tris⅐HCl (pH 7.0), 5 mM EDTA, and 5 mM dithiothreitol supplemented with 200 g/ml lysozyme and protease inhibitors. The soluble lysate fraction containing recombinant proA-luc chimeric protein was filtered through 0.2-m Acrodisks (Gelman Sciences), applied to a HiTrap desalting column (Amersham Biosciences), and buffer-exchanged into a minimal BSS lacking calcium, magnesium, glucose, and BSA using fast protein liquid chromatography protocols. The buffer-exchanged bacterial lysates were pooled, tested for luciferase activity, and stored as 1-ml aliquots at Ϫ80°C. 35-mm culture dishes containing either 1321N1-CD14 or wild-type 1321N1 cell monolayers were washed once with 2 ml of BSS and incubated with gentle stirring for 1 h at room temperature in 1 ml of BSS containing 1 g of anti-CD14 antibody. Following removal of the primary antibody solution, the cell monolayers were washed once with 2 ml of BSS and then incubated with 1 ml of bufferexchanged bacterial lysate containing undiluted proA-luc for 1 h at room temperature with gentle stirring. The proA-luc solution was then aspirated, and the cell monolayers were washed with 2 ml of BSS. The cells were then re-equilibrated for ϳ30 min in 1 ml of fresh BSS at room temperature before being supplemented with 150 M luciferin (2.5 l of 60 mM stock/ml). ATP release was stimulated using 2 units/ml thrombin with and without ␤␥MeATP (300 M) to decrease ATP scavenging at the cell surface. proA-luc activity was recorded every 12 s for up to 18 min. Control experiments on wild-type 1321N1 versus 1321N1-CD14 cells incubated with and without anti-CD14 antibody verified that maximal proA-luc attachment was dependent on both CD14 expression and pre-incubation with anti-CD14 antibody. Calibration curves were generated by addition of exogenous ATP standards to anti-CD14 and proAluc-coated 1321N1-CD14 cells.
Manipulation and Measurement of Intracellular Ca 2ϩ Concentration-The role of cytosolic [Ca 2ϩ ] in basal and stimulated ATP release was studied using either 1321N1 cell monolayers or suspensions as described above. 1321N1 cells were incubated with serum-free Dulbec-FIG. 1. Constitutive ATP release and extracellular ATP metabolism by human 1321N1 astrocytes at steady state. A, extracellular ATP levels were recorded as described under "Experimental Procedures" at 1-2 min intervals during the initial 100 min following transfer to the luciferase-containing assay medium. At the 100-min time point, a bolus of 100 nM exogenous ATP was added, and ATP levels were recorded every 2 min for another 30 min. Digitonin (50 g/ml) and apyrase (5 units/ml) were added at the end of the assay. A representative single experiment with a 1321N1 monolayer (35-mm dish) assayed at 5 days post-plating is shown. B, extracellular ATP was assayed as described in panel A. 1321N1 monolayers were allowed to equilibrate for 45 min prior to treatment with (‚) or without (q) 300 M ␤␥MeATP. A bolus of 100 nM ATP was then added, and ATP levels were recorded every 2 min for another 20 min. Control cell-free dishes (Ⅺ) containing only BSS were also prepared. Data represents the mean Ϯ S.E. from four experiments using cell monolayers at 5-7 days post-plating. The averaged digitonin-releasable [ATP] was ϳ4.2 Ϯ 0.4 M. co's minimal essential medium containing 0.1% BSA plus or minus the cell-permeable calcium chelator BAPTA-AM at 10 M for 30 min at 37°C. BAPTA-loaded or mock-loaded 1321N1 monolayers were then used for luciferase-based assays of extracellular ATP exactly as described above. Suspended cells were prepared for measurements of intracellular Ca 2ϩ concentration by incubation in BSS containing 1 M fura2-AM at room temperature for 1 h. The suspensions were washed twice with BSS before fura2 fluorescence (339-nm excitation and 500-nm emission) was measured and calibrated at 37°C in a stirred cuvette as described previously (25). The effects of thrombin, SFLLRD-TRAP, carbachol, or ␤␥MeATP on intracellular Ca 2ϩ were measured in parallel samples of BAPTA-loaded or non-BAPTA-loaded cells.
Data Evaluation-Individual experimental manipulations were performed using duplicate or triplicate dishes of 1321N1 monolayers from the same cell passage. Integrated RLU recordings of luciferase activity were immediately downloaded into Microsoft Excel using the Turner Designs spreadsheet interface software (version 2.0.1; Sunnyvale, CA). RLU values were converted to ATP concentrations using calibration curves generated on the same day using identical assay solutions and conditions. Prism 3.0 TM software (GraphPad) was used to compute the S.E. of the calculated ATP levels from identical, independent experi-ments that were repeated two to seven times. For ecto-ATPase inhibition studies, Excel (Microsoft) was used to fit the decay of the exogenously added ATP bolus to an exponential curve. Rate constants (k) were then calculated for each curve fit. All experiments were repeated two to seven times. Figures were generated using Prism 3.0 TM (Graph-Pad) and Illustrator 7.0 TM (Adobe) software.

Constitutive ATP Release and Extracellular Metabolism by
Human 1321N1 Astrocytes at Steady State-Immediately after 35-mm dishes of 1321N1 monolayers were washed and transferred from tissue culture medium to the luciferase-supplemented assay medium, 50 -70 nM ATP was present in the bulk 1-ml extracellular volume (Fig. 1A). This elevated extracellular ATP reflects release of endogenous ATP from the cells because of mechanical stimulation of poorly characterized ATP release pathways (15,17). The ATP then rapidly decreased to ϳ2 nM within 20 min. Within 45-60 min after medium transfer, the extracellular ATP stabilized at a non-zero steady state of ϳ1 nM ATP that was 10-fold higher than the limit of ATP detection (100 pM) under these assay conditions. Addition of potato apyrase to the assay medium caused a rapid decrease in extracellular [ATP] to the limit of assay detection (not shown). The steady-state content of 1 pmol of ATP measured in the 1-ml extracellular volume was equivalent to only 0.01-0.02% of the total digitonin-releasable ATP (5-10 nmol) within the ϳ10 6 1321N1 cells that comprised each monolayer. Longer term incubation of the monolayers for up to 100 min failed to decrease extracellular ATP levels to the assay detection limit. Addition of digitonin (as a plasma membrane permeabilizing agent) to cell-containing dishes, but not cell-free dishes, resulted in a rapid increase in bulk extracellular [ATP] to ϳ5 M. This provided a measure of total cytosolic ATP content and was consistent with a cytosolic concentration of 5 mM ATP being diluted 1000-fold into the 1-ml assay volume from the ϳ1-l total intracellular volume of the 10 6 cells that comprise the 1321N1 monolayer.
When the monolayers were pulsed with 100 nM ATP after 100 min of equilibration, this exogenous ATP was degraded at a rate similar to that which characterized the scavenging of the ATP released from endogenous stores upon medium exchange (Fig. 1B). The ATP level decayed with a single exponential rate consistent with a t1 ⁄2 of ϳ12-14 min. In contrast, when 100 nM ATP was added to cell-free plates containing identical luciferase assay medium, the ATP was degraded at a much slower rate reflecting the intrinsic pyrophosphatase activity of the luciferase used as the ATP sensor. In another study, we have developed the use of ␤␥MeATP as an inhibitor of the ecto-ATPase(s) expressed by 1321N1 astrocytes. 2 Those experiments indicated that ␤␥MeATP could be used at millimolar concentrations with no confounding side effects because of ATP contamination, inhibition of luciferase as an ATP sensor, or action as a nucleotide triphosphate substrate for ecto-NDPKmediated phosphorylation of extracellular ADP (data not shown). Fig. 1B shows that 300 M ␤␥MeATP markedly inhibited (by 95%) the ability of 1321N1 monolayers to rapidly scavenge a 100 nM pulse of exogenous ATP.
Effects of Ecto-ATPase Inhibition on Steady-state ATP Levels in the Bulk Extracellular Medium-If the basal [ATP] within the bulk extracellular medium reflects a steady-state balance between constitutive ATP release and hydrolysis by ecto-ATPases, then acute inhibition of the ecto-ATPases should perturb this steady-state and result in an elevated level of extracellular ATP. Fig. 2A illustrates how acute addition of 300 M ␤␥MeATP causes extracellular ATP to gradually increase to ϳ3 nM ATP over 20 min. In the absence of ␤␥MeATP, extracellular ATP remained at the 1 nM steady-state level over this same time period. This slow accumulation of extracellular ATP induced by ␤␥MeATP corresponded to an ATP release rate of ϳ0.1 pmol/min/10 6 cells. Because similar results were observed using monolayers that were not periodically agitated after addition of the ␤␥MeATP (not shown), cell lysis was an unlikely cause for this slow accumulation of ATP in the bulk extracellular compartment. Identical experiments using cell-free dishes verified that the ␤␥MeATP-mediated increase in ATP concentrations was dependent on the presence of 1321N1 monolayers and was not because of de novo synthesis of extracellular ATP as a consequence of contaminating adenylate kinase or nucleotide diphosphokinase activities in the luciferase assay medium. These results are consistent with previous observations by Lazarowski et al. (17) who used isotopic equilibrium methods to demonstrate a steady-state balance between constitutive release of endogenous ATP from resting 1321N1 astrocytes and the degradation of that ATP by ecto-ATPases.
Thrombin-stimulated ATP Release from 1321N1 Astrocytes in the Absence or Presence of Ecto-ATPase Inhibition-Previous studies have established that 1321N1 astrocytes express G protein-coupled protease-activated receptors (PAR1) for thrombin (26,27). In the absence of ecto-ATPase inhibition, stimulation of 1321N1 monolayers with thrombin resulted in only a transient 2-fold rise in bulk extracellular ATP to 2 nM within 2 min followed by a return to the baseline level over the next 10 min (Fig. 2A). This very modest and transient increase contrasted with the sustained 30-to 80-fold increase in bulk extracellular ATP observed when 1321N1 monolayers were simultaneously exposed to both thrombin and 300 M ␤␥MeATP as costimuli (Fig. 2B). The strongly synergistic action of thrombin and ␤␥MeATP on accumulation of extracellular ATP indicates that ecto-ATPases can rapidly scavenge increased amounts of endogenous ATP released onto the cell surface in response to thrombin receptor activation. Moreover, the steadily maintained plateau in extracellular [ATP] throughout the 20-min period after costimulation with ␤␥MeATP and thrombin suggests that thrombin mobilizes a limited and slowly replenished pool of releasable ATP, such as exocytotic vesicles that sequester nucleotides. Increasing the ␤␥MeATP in the costimulus from 3 to 300 M at constant thrombin produced  a concentration-dependent increase in the peak magnitude of the extracellular ATP detected by luciferase (Fig. 2B). Consistent with the partial inhibition of ecto-ATPase activity at ␤␥MeATP Ͻ 300 M, 3 the increases in extracellular ATP were not steadily maintained when thrombin was applied as a costimulus with 3 or 30 M ␤␥MeATP.
Varying the thrombin concentration (0.0001-3000 milliunits/ml) at constant ␤␥MeATP (300 M) indicated that the threshold for thrombin-induced ATP release was in the 1 milliunits/ml (ϳ20 pM) range (Fig. 3, A and C). However, the concentration-response relationship describing thrombin-induced ATP accumulation was unusually steep with peak ATP accumulation at 10 milliunits/ml (240 pM) followed by modest decrease to a plateau level of ATP release at Ն30 milliunits/ml thrombin. Although most acute actions of thrombin on cellular function can be ascribed to its proteolytic activation of PAR family receptors, it was important to test whether thrombin might additionally potentiate extracellular ATP accumulation by proteolytic modification of other proteins directly involved in ATP release or extracellular ATP metabolism (28) . Fig. 3, B and D illustrates the rates and magnitudes of ATP accumulation observed in 1321N1 monolayers costimulated with ␤␥MeATP plus various concentrations of the hexapeptide, SFLLRD, which acts as a reversible PAR1 agonist or TRAP reagent. The maximal rates and peak magnitudes of SFLLRD-TRAP-induced ATP accumulation were similar to those observed when thrombin per se was used to activate PAR1. This suggested that the effects of thrombin on ATP release and accumulation are solely because of its proteolytic actions on PAR family receptors. The EC 50  induced only a small (3-4 nM peak) and transient accumulation of ATP within the bulk extracellular medium (data not shown).
Thrombin-stimulated Release of ATP into the Extracellular Surface Microenvironment-Figs. 2 and 3 indicate that, in the absence of ecto-ATPase inhibition, most of the ATP released in response to thrombin or SFLLRD-TRAP is hydrolyzed before it can diffuse into the bulk extracellular compartment containing the soluble luciferase sensor. This suggests that ATP release and hydrolysis occurs in a cell surface microcompartment that is functionally segregated from the bulk extracellular compartment. In a previous study, we demonstrated that proA-luc could be localized to the extracellular surface of intact cells via high affinity association between the protein A moiety and antibodies bound to extracellular epitopes of plasma membrane proteins (24). This effectively localizes and restricts the luciferase to the immediate extracellular surface. As schematically illustrated in Fig. 4, this may also facilitate targeting of the luciferase to plasma membrane subdomains containing the endogenous ecto-ATPases and thereby allow the luciferase-ATP sensor to effectively compete with ecto-nucleotidases for endogenous ATP as it is released onto the cell surface. To localize proA-luc to the surface of 1321N1 cells, we generated sublines (1321N1-CD14) that were stably transfected with an expression plasmid encoding human CD14. CD14 is a glycosylphosphatidylinositol-anchored plasma membrane protein, normally expressed only in myeloid leukocytes, that acts as a cell surface binding site for endotoxin/lipopolysaccharide (29). Immunofluorescence microscopy using a monoclonal anti-human CD14 antibody verified cell surface expression of CD14 in the 1321N1-CD14 lines but not in the wild-type parental 1321N1 cells (data not shown). This same monoclonal antibody was used to specifically adsorb the proA-luc chimeric protein to the extracellular surface of 1321N1-CD14 monolayers. proAluc attachment was strongly CD14-dependent, because calibrations with ATP standards revealed significantly higher RLU values from proA-luc-coated 1321N1-CD14 cells compared with wild-type 1321N1 monolayers that were subjected to identical incubations with anti-CD14 and proA-luc (Table I). In the absence of the anti-CD14 precoating, similar low levels of nonspecifically adsorbed proA-luc were observed with wild-type 1321N1 and 1321N1-CD14 monolayers (not shown). The limit of detection for ATP (added to the bulk extracellular compartment) was in the 10 -30 nM range (Table I). Fig. 5 compares the time courses of thrombin-induced ATP release from 1321N1-CD14 monolayers using either the previously described soluble luciferase protocol or the cell-attached proA-luc method; panels A and B illustrate the results obtained using two different subclones of CD14-transfected 1321N1 cells. To assess the early kinetics of the ATP release process, luminescence was recorded at 12-s intervals in the absence or and presence of 300 M ␤␥MeATP. When assayed by the soluble luciferase protocol, the thrombin-induced ATP release responses in 1321N1-CD14 cells were similar to those observed with wild-type cells, i.e. transient accumulation of extracellular ATP in the absence of ecto-ATPase inhibition versus a strongly synergistic response to thrombin and ␤␥MeATP added as costimuli. Thrombin activation of these particular 1321N1 sublines in the absence of ␤␥MeATP caused bulk ATP levels to transiently increase by ϳ15 nM within 3 min (Fig. 5, A and B), corresponding to an initial release rate of ϳ5 pmol/min/10 6 cells. This response was higher than was observed with the wild-type 1321N1 monolayers illustrated in Fig. 2A and may indicate either increased expression of the ATP release machinery in the clonally selected sublines or an elevated rate of ATP diffusion into the bulk phase medium because of increased mixing that accompanied the 10-fold higher frequency of luminescence readings (every 12 s versus every 2 min).
Significantly, when cell-attached luciferase was used as the ATP sensor, thrombin stimulation alone was sufficient to trigger rapid accumulation of ATP at the extracellular surface to peak concentrations in the 100 -300 nM range (Fig. 5, A and B). This corresponded to an initial release rate of Ͼ100 pmol ATP/ min/10 6 cells, which is at least 20-fold higher than was observed using soluble luciferase as the ATP sensor. The cell surface ATP concentration reached a peak within 1 min after thrombin addition and was then maintained at a plateau for another 60 -90 s before decaying toward baseline levels over the next 5 min. When thrombin was delivered as a costimulus with ␤␥MeATP, the increasing phase of ATP accumulation was prolonged to 2 min resulting in peak cell surface levels 1.5-to 1.8-fold higher (150 -550 nM) than was observed in the absence of ␤␥MeATP. The presence of ␤␥MeATP also affected decay kinetics of the cell surface ATP signal. The plateau was maintained for only 20 s, and ATP levels did not return to basal levels but instead remained elevated at the 50 -150 nM range. Although much of the ATP released onto the cell surface during thrombin stimulation may diffuse into and be diluted within the bulk medium compartment, ecto-ATPase inhibition clearly facilitates the further accumulation of ATP at the cell surface.
Carbachol-stimulated ATP Release from 1321N1 Astrocytes in the Absence or Presence of Ecto-ATPase Inhibition-To test whether agonists for other G protein-coupled receptors could mimic the effects of thrombin or SFLLRD-TRAP on ATP release, similar experiments were performed using 100 M carbachol to maximally activate the M1-muscarinic receptors that have been characterized previously in 1321N1 astrocytes (26). Fig. 6 shows that carbachol alone triggered only minor and transient accumulation of extracellular ATP (2 nM peak at 4 min) in the bulk medium compartment of 1321N1 monolayers. As observed with thrombin, carbachol acted synergistically with ␤␥MeATP (added as a costimulus) to trigger accumulation of ϳ12 nM extracellular ATP within 5 min. However, both the initial rate of ATP release and the maximal plateau level of accumulated ATP in these carbachol-stimulated cells were invariably 3-to 5-fold lower than measured in parallel 1321N1 monolayers (same passage assayed on the same day) stimulated with thrombin or SFLLRD-TRAP (see also Fig. 9).
Hypotonic Stress-stimulated ATP Release from 1321N1 Astrocytes in the Absence or Presence of Ecto-ATPase Inhibition-Multiple cell types have been shown to release nucleotides when subjected to hypotonic stress as part of a regulatory volume decrease response to osmotic swelling (22, 30 -35). 1321N1 monolayers challenged with a 30% decrease in tonicity accumulated 3 nM ATP in the bulk extracellular compartment over a 10-min period. However, induction of hypotonic stress as a costimulus with ␤␥MeATP resulted in a sustained rise in extracellular ATP to ϳ60 nM over a 20-min test period (Fig. 7). This result indicates that ATP released in response to either activated G protein-coupled receptors or mechanical stimulation can be catabolized with similar efficiency in extracellular compartments at the 1321N1 cell surface before dilution into the bulk extracellular fluid phase. Although the magnitude of the hypotonicity-induced ATP accumulation was similar to that triggered by thrombin receptor activation, the rate of the ATP accumulation was at least 5-fold slower with the hypotonic stimulus. This suggests that the two stimuli may target mechanistically distinct ATP release pathways.
Role of Cytosolic Ca 2ϩ in Basal and Stimulated ATP Release from 1321N1 Astrocytes-In platelets, thrombin activates Gqcoupled PAR1 receptors to trigger the Ca 2ϩ -dependent exocytosis of ATP-containing dense granules (28). Given that the PAR1 and M1 receptors expressed in 1321N1 astrocytes also activate phosphatidylinositol phospholipase-C-beta and mobilization of inositol 1,4,5-trisphosphate-sensitive Ca 2ϩ stores, we tested the possible role of cytosolic Ca 2ϩ as a regulator of ATP release. The basal cytosolic [Ca 2ϩ ] in 1321N1 cells was ϳ200 nM (Fig. 8). Thrombin and carbachol both rapidly triggered transient increases in [Ca 2ϩ ] that peaked in the 2 M range (Fig. 8A). However, the response to carbachol was sustained for several min in contrast to the rapid decay within 60 s of the thrombin-induced Ca 2ϩ transient. When ␤␥MeATP was tested at the 300 M concentration routinely used to inhibit ecto-ATPase activity, a transient increase in [Ca 2ϩ ] to ϳ300 nM was also observed in some (e.g. Fig. 8A) but not all preparations of 1321N1 astrocytes. 1321N1 cells lack significant expression of the known Gq-coupled P2Y nucleotide receptor subtypes (P2Y1, P2Y2, P2Y4, P2Y6, and P2Y11), and ␤␥MeATP is a poor agonist for those receptor subtypes (1,5,15). Although ␤␥MeATP can also act as an agonist for certain P2X receptors (1), 1321N1 cells lack expression of any of the seven known P2X subtypes (67). Thus, the underlying mechanism for this modest and variable ␤␥MeATP-induced Ca 2ϩ transient is unclear. As expected, the ability of carbachol, thrombin, or ␤␥MeATP to trigger transient increases in cytosolic [Ca 2ϩ ] was completely ablated in parallel samples of 1321N1 cells that were loaded with the intracellular Ca 2ϩ -chelator BAPTA (Fig. 8B).
Basal and stimulated ATP release were then compared in control versus BAPTA-AM-loaded 1321N1 monolayers assayed in the presence of 300 M ␤␥MeATP to inhibit ecto-ATPase activity. Fig. 9A shows that the slow rate of basal ATP release observed upon acute addition of ␤␥MeATP was similar in the control and BAPTA-loaded cells. However, BAPTA loading significantly (ϳ75% inhibition) attenuated the rate and extent of ATP release triggered by thrombin and completely abrogated the carbachol-stimulated ATP release (Fig. 9, B and C). These experiments also highlight the 4-fold larger magnitude of the ATP release triggered by thrombin versus carbachol despite the more sustained Ca 2ϩ transients elicited by the latter agonist. The results suggest that increased cytosolic [Ca 2ϩ ] is an important signal for activation of the ATP release machinery by Gq-coupled receptors but that thrombin additionally elicits ATP release via signaling pathways that do not require elevated Ca 2ϩ . An involvement of Ca 2ϩ -independent signaling pathways in ATP release was further supported by the inability of BAPTA loading to substantially attenuate the accumulation of extracellular ATP in 1321N1 cells subjected to hypotonic stress (Fig. 9D).
Additive Effects of Thrombin Receptor Activation and Hypotonic Stress on ATP Release from 1321N1 Astrocytes-The markedly different consequences of BAPTA loading on thrombin-induced ATP release versus hypotonicity-induced ATP release suggested that these stimuli trigger mechanistically distinct pathways of nucleotide export. This was tested by comparing the ATP release responses (in the presence of ␤␥MeATP) to costimulation by simultaneous hypotonic stress and thrombin treatment versus stimulation by hypotonicity alone or thrombin alone (Fig. 10). Significantly, the costimulation protocol induced strictly additive effects on net ATP accumulation with regard to both time course and magnitude. Such additive actions are consistent with the possibility that 1321N1 astrocytes may release ATP (and perhaps other nucleotides) by parallel, non-interacting mechanisms.

DISCUSSION
Accumulation of ATP within a particular extracellular compartment will reflect both the rate of ATP release into that compartment from intracellular pools and the rate of ATP hydrolysis by locally expressed ectonucleotidases. This study provides several novel insights regarding the dynamics and regulation of extracellular ATP accumulation in a non-excitatory cell type, the 1321N1 human astrocytoma line. Stimulation of these cells with Ca 2ϩ -mobilizing agonists for two different Gq-coupled receptors (PAR1 and M1-muscarinic) resulted in only minor increases in ATP levels within the bulk extracellular medium even at early time points following receptor activation. This contrasted with the rapid elevation of extracellular ATP to submicromolar levels at the immediate cell surface during PAR1 activation of 1321N1 cells in which the luciferase ATP sensor was tethered to a glycosylphosphatidylinositol-anchored plasma membrane protein. This result suggests that the PAR1-elicited ATP release is initially directed into an extracellular subcompartment that lacks rapid diffusional exchange with the extracellular medium bathing these astrocyte monolayers. As a result, much of the released ATP can be efficiently metabolized by ecto-ATPases that are present within, and possibly localized to, that same extracellular subcompartment prior to diffusion into the bulk extracellular fluid phase. This interpretation is further supported by pharmacological experiments that used high concentrations of an exogenous, poorly hydrolyzable ATP analog (␤␥MeATP) to competitively inhibit hydrolysis of the released ATP by the ecto-ATPases within the putative cell surface subcompartment. This repression of localized ATP scavenging readily facilitated diffusional equilibration of the released ATP with the bulk extracellular fluid volume where it could be registered by the soluble luciferase ATP sensor. Taken together, the results of the tethered luciferase assays and the pharmacological experiments suggest that the sites of initial ATP release are focused within plasma membrane subdomains characterized by high levels of local ecto-ATPase activity.
This apparent functional segregation of the immediate ATP release sites from the bulk extracellular compartment may involve spatial colocalization of the ATP release sites and ectonucleotidases within plasma membrane subdomains, such as caveolae or lipid raft-based invaginations, that enclose physically constrained extracellular spaces and thereby retard rapid exchange with the bulk extracellular space. The ability of the CD14-tethered luciferase to effectively monitor significant thrombin-induced ATP release even in the absence of ␤␥MeATP indicates that CD14 acts to concentrate the cell surface luciferase ATP sensor to levels sufficient for effective competition with local ecto-ATPases. glycosylphosphatidylinositol-anchored proteins such as CD14 are often localized to lipid raft domains of the plasma membrane (36). Significantly, various purinergic signaling elements, including receptors (37,38) and ecto-nucleotidases (39 -41), have also been localized to raft domains and/or caveolae. Most germane to our observations is the finding by Robson and co-workers (40) that the CD39 ecto-apyrase/eNTDPase-1, the prototype of the ecto-NT-DPase family, is localized to the caveolae of human endothelial cells. This caveolar association of CD39 was found to be dependent on post-translational modification by N-terminal palmitoylation (41).
It remains to be determined whether CD39 or another ectonucleotidase is the predominant ␤␥MeATP-sensitive ecto-ATPase expressed in 1321N1 human astrocytes. The mammalian CD39 family of eNTDPases comprises six unique gene products that encode intrinsic membrane proteins with two membrane-spanning domains, intracellular amino-and carboxyl termini, and a large connecting loop that contains the catalytic activity (3,4). Three of these family members (CD39/ eNTDPase-1, CD39-L1/eNTDPase-2, and CD39-L3/eNTD-Pase-4) function as plasma membrane-localized ecto-ATD-Pases that can hydrolyze extracellular ATP and/or ADP to AMP. These enzymes also metabolize other nucleotide tri-and diphosphates, including UTP and UDP, which are agonists for several P2Y receptor subtypes. Lazarowski and co-workers (15,17) have reported previously that 1321N1 cells express a significant ecto-NTPase activity that can metabolize both ATP and UTP. These investigators also observed that 1321N1 cells additionally express robust ecto-nucleotide pyrophosphatase and ecto-nucleotide diphosphokinase activities. The mammalian ecto-nucleotide pyrophosphatase family includes three unique gene products that are class 2-type plasma membrane proteins (42). In rat C6 glioma cells (43,44) and Xenopus oocytes (45), ecto-nucleotide pyrophosphatase enzymes can hydrolyze ␤␥MeATP to directly generate AMP, which is subsequently metabolized to adenosine via the CD73 ecto-5Ј-nucleotidase. Extracellular NDPK (17)(18)(19) or ecto-adenylate kinase activities (20) add yet another complication to analyses of extracellular ATP accumulation, because they can catalyze the extracellular synthesis ATP from ambient ADP in the presence of released or added nucleotide triphosphates. However, because of the methylene bridge between its ␤and ␥-phosphates, ␤␥MeATP cannot act as a phosphate donor for NDPK-catalyzed phosphorylation of ADP. 4 Depending on the cell type and extrinsic stimulus, ATP can be released to extracellular compartments either by exocytosis of ATP-containing vesicles/granules or by the facilitated efflux of cytosolic ATP through channels or transporters that can accommodate the size and charge of nucleotides. Our comparative analyses of ATP release from 1321N1 astrocytes stimulated by Gq-coupled receptor agonists versus hypotonic stress suggest that these cells express two (or more) parallel mechanisms for nucleotide export that are regulated by distinct signaling pathways. Given the strong inhibitory effect of BAPTA loading on ATP release induced by the Ca 2ϩ -mobilizing PAR1 or M1 agonists, but not on ATP release elicited by hypotonicity, it is plausible to speculate that thrombin and carbachol primarily activate Ca 2ϩ -dependent exocytosis of ATP-containing vesicles, whereas osmotic or mechanical stress may gate the Ca 2ϩ -independent opening of ATP-conductive channels. However, it is also possible that increased Ca 2ϩ may act to stimulate the gating of other ATP-permeable channels.
The best characterized analyses of exocytotic ATP secretion have utilized specialized secretory cells, such as chromaffin cells (46), PC12 cells (47), and platelets (24,48), that compartmentalize ATP to submolar concentrations within dense core granules that can be easily identified by morphological criteria and biochemically isolated by standard cell fractionation (13). However, recent studies have suggested that exocytosis may be partially or fully responsible for the stimulus-induced release of ATP observed in multiple cell types, including astrocytes, that lack large numbers of specialized granules or vesicles known to contain ATP (49 -51). It should be noted that even vesicles involved in the constitutive release of secreted proteins contain low levels of nucleotides because of important roles for intravesicular, nucleotide-dependent enzymes that catalyze covalent modification and maturation of secreted proteins (52). The studies of Maroto and Hamill (50) on ATP externalization from single Xenopus oocytes suggested that brefeldin A-sensitive exocytotic pathways were involved in both basal and mechanically stimulated nucleotide release in that cell type. Some recent studies suggest that astrocytes and related glial-type cells may compartmentalize ATP in vesicles that can be released via regulated exocytosis. Using subcellular fractionation methods, Maienschein et al. (53) observed that primary rat astroglial cells contain ATP-enriched vesicles that colocalize with multiple proteins (synaptotagmin-I, synaptobrevin-II, SNAP25) implicated in regulated exocytosis. Coco et al. (50) have reported that mechanical stimulation or pharmacological agonists (phorbol esters) induced a release of ATP from primary neonatal rat astrocytes that could be partially blocked by bafilomycin, an inhibitor of the vesicular H ϩ -ATPases that set up the proton electrochemical gradients required for active accumulation of neurotransmitters, or tetanus toxin, which proteolytically inactivates the V-SNARE, synaptobrevin-II.
In addition to regulated exocytosis, there is compelling evidence for the involvement of nucleotide-permeable channels in the release of ATP from various cell types, including astrocytes (54 -56). Because cells contain millimolar levels of negatively charged MgATP 2Ϫ within the cytoplasm and also maintain negative membrane potentials, there is a large electrochemical gradient for ATP efflux across the plasma membrane. Given this favorable electrochemical driving force, even brief activation of a nucleotide-permeable transporter could rapidly increase the rate of ATP delivery to local extracellular compartments. An involvement of channels as conduits for facilitated ATP efflux has been supported by two distinct but complementary experimental approaches. One is the direct electrophysiological measurement of membrane currents or channels for which ATP can act as an apparent charge carrier (32,34,57,58). The other analysis is based on the inhibition of ATP release by pharmacological agents known to target particular ion channels or classes of ion channels (30,31,33,35,57,68). Based on these approaches, three classes of channels have been implicated in ATP release: 1) ATP-binding cassette transporters (58), 2) volume-regulated anion channels or VRAC (32,34), and 3) connexin hemichannels (54 -56, 68). Whether one or more of these channels contributes to the ATP release observed in our 1321N1 astrocyte model system remains an open question.
Connexin hemichannels, which represent hexameric complexes of connexin subunits at non-junctional plasma membrane sites (59), have been particularly associated with ATP release from primary rat astrocytes subjected to metabolic inhibition (60) or direct mechanical stress (55,56). In addition, Braet et al. (68) have observed recently that photoliberation of caged inositol 1,4,5-trisphosphate in endothelial cells triggered a Ca 2ϩ -dependent release of ATP that was markedly repressed by peptide inhibitors of connexin hemichannels. Such hemichannels are appealing candidates as ATP efflux proteins given the known ability of some connexin-based gap junctions, which represent dodecameric complexes of connexin subunits at junctional plasma membrane loci, to directly mediate the cell-to-cell movement of nucleotides between junctionally coupled cells. It is important to note that Nedergaard and coworkers (54,55) have observed that BAPTA loading represses the mechanically induced ATP release that is correlated with gating of connexin hemichannels in rat astrocytes. This contrasts with the BAPTA insensitivity of the hypotonicity-stimulated ATP release we observed in the 1321N1 astrocytes. Alternatively, Hisadome et al. (32) have noted that the hypotonicity-induced ATP release associated with activated VRAC currents in bovine aortic endothelial cells can occur in the absence of elevated cytosolic Ca 2ϩ . Using the same bovine aortic endothelial cells model system and hypotonic stimuli, Koyama et al. (31) reported that ATP release was significantly inhibited by a pharmacological inhibitor (Y-27632) of the ROCK-I or ROCK-II Rho-activated kinases, as well as by botulinum C3 exotoxin, which acts to inhibit activation of Rhofamily GTPases. In preliminary experiments, we have failed to observe similar inhibitory effects of Y-27632 on the ATP release triggered by hypotonic stress (or PAR1 activation) in 1321N1 astrocytes. Thus, the nature or identity of candidate ATPpermeable channels that may contribute to nucleotide release from this latter cell model remains unclear at present. The possibility that a single cell type, such as a 1321N1 astrocyte, may release ATP via parallel, non-cytolytic mechanisms involving both exocytosis and facilitated efflux through a nucleotidepermeable channel transporter is supported by a recent analysis of hypertonicity-induced ATP release from single Xenopus oocytes (61).
Regardless of the mechanism(s) that underlie ATP release from astrocytes, a growing body of data indicates that nucleotide release plays important autocrine and paracrine signaling roles in several known functions of astrocytes as modulators of neuronal function and survival in the central nervous system (62)(63)(64)(65). The released ATP can act directly on ionotropic P2X receptors or certain G protein-coupled P2Y receptor subtypes that are expressed on astrocytes per se and adjacent neurons (both pre-and post-synaptic). In conjunction with ecto-nucleotidase cascades, the released ATP also serves as an extracellular reservoir for the generation of the ADP or adenosine agonists that subsequently target yet other G protein-coupled receptors. One of the more intensively studied roles of local ATP release and P2 receptor activation in astrocyte function is the intercellular propagation of Ca 2ϩ waves among astrocytes or between astrocytes and neurons (54 -56, 65). This intercellular propagation of signals that control a critical intracellular second messenger provides a rapid mechanism for coordinating the metabolic and electrical responses of local CNS regions to physiological changes in activity or pathological stresses, such as ischemia, mechanical trauma, or excitotoxicity. Although direct communication of Ca 2ϩ waves via gap junction channels among physically adjacent astrocytes provides one mechanism, the release of ATP followed by the activation of Ca 2ϩ -mobilizing P2Y receptors comprises a second pathway that does not require direct physical contact. In turn, the increase in astrocyte Ca 2ϩ initiated by the activated P2Y receptors can induce the release of additional paracrine factors such as glutamate, prostaglandins, or other nucleotides, which act to further reinforce and coordinate the local response to the stress that initiated release of the ATP.