Prostaglandin E2 stimulates a Ca2+-dependent K+ channel in human erythrocytes and alters cell volume and filterability.

To understand the mechanism by which human red blood cells (RBCs) contribute to hemostasis and thrombosis, we have examined the effects of metabolites released by activated platelets on intact RBCs. Prostaglandin E2 (PGE2), a signal molecule produced by activated platelets, was observed to lower the filterability of human erythrocytes by ~30% at 10−10 M. PGE2 also caused a reduction in mean cell volume of ~10%. The shrinkage of red cells after PGE2 treatment was confirmed by documenting a decrease in osmotic fragility and an increase in cell density following exposure to the hormone. Careful analysis, however, revealed that only ~15% of the erythrocytes responded to stimulation with PGE2. Examination of the cause of cell shrinkage showed that induction of a PGE2-stimulated K+ efflux pathway leading to rapid loss of cellular K+ was responsible. The PGE2-stimulated K+ loss was also observed to be Ca2+-dependent, suggesting the possible involvement of the Gardos channel. Gardos channel participation was supported by the observation that two Gardos channel inhibitors, charybdotoxin and clotrimazole, independently blocked the PGE2-stimulated K+ efflux. Further evidence for Gardos channel activation came from experiments aimed at characterizing the efflux pathway followed by the obligatory counterion. Thus, K+ efflux was readily stimulated even when NO−3 was substituted for Cl−, suggesting that neither KCl cotransport nor Na/K/2Cl cotransport plays a prominent role in the PGE2-induced cell shrinkage. Further, the anion transporter band 3 was implicated as the counterion efflux route, since DIDS inhibited the PGE2-stimulated cell volume change without blocking the change in membrane potential. Taken together, we propose that release of PGE2 by activated platelets constitutes part of a mechanism by which activated platelets may recruit adjacent erythrocytes to assist in clot formation.

To understand the mechanism by which human red blood cells (RBCs) contribute to hemostasis and thrombosis, we have examined the effects of metabolites released by activated platelets on intact RBCs. Prostaglandin E 2 (PGE 2 ), a signal molecule produced by activated platelets, was observed to lower the filterability of human erythrocytes by ϳ30% at 10 ؊10 M. PGE 2 also caused a reduction in mean cell volume of ϳ10%. The shrinkage of red cells after PGE 2 treatment was confirmed by documenting a decrease in osmotic fragility and an increase in cell density following exposure to the hormone. Careful analysis, however, revealed that only ϳ15% of the erythrocytes responded to stimulation with PGE 2 . Examination of the cause of cell shrinkage showed that induction of a PGE 2 -stimulated K ؉ efflux pathway leading to rapid loss of cellular K ؉ was responsible. The PGE 2 -stimulated K ؉ loss was also observed to be Ca 2؉ -dependent, suggesting the possible involvement of the Gardos channel. Gardos channel participation was supported by the observation that two Gardos channel inhibitors, charybdotoxin and clotrimazole, independently blocked the PGE 2 -stimulated K ؉ efflux. Further evidence for Gardos channel activation came from experiments aimed at characterizing the efflux pathway followed by the obligatory counterion. Thus, K ؉ efflux was readily stimulated even when NO 3 Ϫ was substituted for Cl ؊ , suggesting that neither KCl cotransport nor Na/K/2Cl cotransport plays a prominent role in the PGE 2 -induced cell shrinkage. Further, the anion transporter band 3 was implicated as the counterion efflux route, since DIDS inhibited the PGE 2 -stimulated cell volume change without blocking the change in membrane potential. Taken together, we propose that release of PGE 2 by activated platelets constitutes part of a mechanism by which activated platelets may recruit adjacent erythrocytes to assist in clot formation.
Red blood cells (RBCs) 1 comprise 99% of the blood cells in circulation. RBCs are typically thought to function only in gas transport, facilitating movement of O 2 from the lungs to the tissues, and CO 2 (as free HCO 3 Ϫ or as carbamates of hemoglobin) from the tissues to the lungs. Although generally considered hemostatically inert, isolated observations have suggested that erythrocytes might also participate in clot formation. Thus, clinicians have periodically reported a correlation be-tween low erythrocyte count and prolongation of bleeding time (1)(2)(3). Hellem et al. (3,4) have also documented a dependence of platelet adhesion to glass beads on the presence of red cells, increasing as hematocrit is elevated. More recently, Saniabdi and Lowe (5) reported that RBCs markedly enhance spontaneous platelet aggregation in vitro, and Santos and colleagues (6) have demonstrated that erythrocytes enhance platelet serotonin release as well as arachidonate/eicosapentaenoate production and eicosanoid formation (6). The same group (7) has further demonstrated that collagen-stimulated platelets aggregate 3 times more effectively and discharge 7 times more ADP in the presence of RBCs than in their absence. Taken together, these observations argue for some type of communication between red cells and activated platelets, although at present the diffusable second messengers remain unknown.
To evaluate the mechanism by which red cells recognize an activated platelet and respond molecularly to platelet-generated signals, we have begun to examine the effects of metabolites released by activated platelets on intact RBCs. Because prostaglandin E 2 (PGE 2 ) is a potent signaling molecule produced by activated platelets, PGE 2 was selected for initial examination. PGE 2 is a metabolic product of arachidonic acid which is released by the action of phospholipase A 2 during platelet activation (8). PGE 2 mediates a broad range of biological activities in diverse tissues through specific plasma membrane receptors (9 -11). In normal resting platelet-rich plasma, PGE 2 is essentially absent (8). However, during platelet activation, the level of PGE 2 in plasma increases to about 10 Ϫ9 M (8). Allen and Rasmussen (12,13) have reported that PGE 2 decreases the filtration rate of both human and rat RBCs, and they have consequently suggested that erythrocytes must contain some type of PGE 2 receptor. We report here that PGE 2 activates the Gardos channel in human erythrocytes and thereby induces K ϩ efflux, cell shrinkage, and reduced cell filterability. This reduced cell filterability could conceivably help retain the stimulated RBCs in the newly formed hemostatic plug.
Preparation of Erythrocytes-Human blood was withdrawn from healthy volunteers in acid citrate-dextrose solution on the day of the experiment. After centrifugation and removal of the plasma, "buffy coat," and upper 20% of the red cells, the remaining red cells were washed three times in phosphate-buffered saline (150 mM NaCl, 5 mM glucose 5 mM sodium phosphate, pH 7.4), and twice again in buffer A (125 mM NaCl, 3 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 1.2 mM sodium phosphate, 50 mM Tris-buffer, pH 7.4 (37°C), 10 mM glucose) in order to equilibrate the cells in a Ca 2ϩ -containing medium. During these washing steps, essentially all of the white cells were aspirated away. The washed erythrocytes at 10% hematocrit in buffer A were then treated with 0.001 volumes of PGE 2 dissolved in ethanol and incubated for the desired times.
Cell Filtration Studies-Cell filtration studies were performed as described by Ataullakhanov (14,15). Briefly, a filtration device was used consisting of an upper plastic reservoir equipped with two electronic fluid level detectors separated vertically by 9 mm. At the bottom of the reservoir was a 1-cm diameter membrane filter with 3-m cylindrical pores prepared by nuclear beam technology. The reservoir was filled either with an RBC suspension at 2% hematocrit in buffer A containing 5 mg/ml bovine serum albumin or with the serum albumincontaining buffer alone. As the erythrocytes passed through the membrane filter under force of gravity, the two level detectors first activated and then deactivated an electronic chronometer that recorded the time of movement of the meniscus from the first detector to the second. For each membrane filter, the filtration time of the erythrocyte suspension was normalized to the filtration time of the bovine serum albumin solution. The relative filterability of the erythrocyte suspension was then calculated according to the equation: relative filterability ϭ (T RBC / T BSA )/(TЈ RBC /TЈ BSA ), where T RBC and TЈ RBC represent the filtration times of erythrocyte suspensions without and with PGE 2 treatment, respectively; T BSA and TЈ BSA are the filtration times of the serum albumin buffer alone measured with the same filter used for the corresponding erythrocyte suspension.
The filtration device and nuclear membrane filters were gifts of Dr. F. I. Ataullakhanov, National Center of Hematology, Moscow, Russia.
Cell Volume Evaluation-Flow cytometric forward light scattering was used to estimate the relative sizes of the cells (16 -18). Briefly, cell suspensions (10% hematocrit) were aspirated directly into an EPICS Elite flow cytometer (Coulter Corp., Hialeah, FL) equipped with a 488 nm laser beam. Forward light scattering was recorded at the desired time points following PGE 2 treatment, and mean cell diameter was determined using the computer software provided by the manufacturer. The flow cytometer was calibrated prior to each experiment using the standard beads (9.99 m) provided by the manufacturer.
Osmotic Fragility of RBCs-Cell suspensions at 10% hematocrit were diluted 10-fold in media consisting of buffer A with decreasing NaCl concentrations. After incubating at 4°C for 4 min, cell suspensions were centrifuged (1,000 ϫ g, 5 min) and 100 l of supernatant was mixed with 10 volumes of Drabkins solution. The hemoglobin concentrations in the supernatants were determined by measuring the absorbance at 540 nm. Percent hemolysis was defined as C h /CЈ h ϫ 100, where C h and CЈ h are the hemoglobin concentrations obtained following incubation of the RBCs in the hypotonic media and deionized water, respectively.
Separation of Erythrocytes on Stractan Density Gradients-Erythrocytes were separated on a stractan density gradient as described by Corash et al. (19) with minor modifications. Briefly, stractan was supplemented to 3% (w/v) with bovine serum albumin, and the pH of the solution was adjusted to 7.4 with NaOH. Then, to 9 parts of the stractan solution was added 1 part of 0.15 M potassium phosphate, pH 7.4. Available water content was calculated according to Ref. 19, and 116 mg of MgCl 2 ⅐H 2 O and 200 mg of glucose were added to each 100 ml of available water. Finally, the osmolarity of the stock solution was brought to 291 mosm by addition of dry NaCl, and the desired density fractions were prepared by mixing the stractan stock solution with buffer A (adjusted to 291 mosm), as described by Dowson et al. (20).
For the separation of erythrocytes, 1 ml of cell suspension (50% hematocrit) was applied to the top of the stractan density gradient in a 1.58 by 10.16-cm cellulose nitrate centrifuge tube, and the tubes were centrifuged for 1 h at 50,000 ϫ g. The separated erythrocyte bands were collected and analyzed for hemoglobin content at 540 nm to estimate the number of erythrocytes at each density.
K ϩ Efflux Measurement-Packed RBCs were resuspended in buffer A at 10% hematocrit and prewarmed to 37°C for 10 min. The external K ϩ concentration was then continuously measured under gentle stirring using a K ϩ sensitive electrode and chart recorder. The decrease in cellular K ϩ content at any time t (⌬K i(t) ) was determined as follows: In the nitrate substitution experiments, nitrate salts were substituted for chloride salts in all components of buffer A. In the Gardos channel inhibition studies, the RBC suspension (10% hematocrit) was preincubated with charybdotoxin (ChTX) at room temperature for 100 min or in clotrimazole for 20 min prior to initiation of the K ϩ efflux assays. After raising the temperature to 37°C, efflux experiments were performed as described above. When ChTX was employed as an inhibitor, the Ca 2ϩ concentration of buffer A was reduced to 50 M, since elevated Ca 2ϩ has been shown to block ChTX binding (22).
Measurement of RBC Membrane Potential-The membrane potential of red cells was estimated using the cationic cyanine dye, diS-C 3 -(5), whose membrane partition coefficient is strongly sensitive to transmembrane potential. Briefly, 10 l of packed erythrocytes were incubated in 90 l of buffer A with or without PGE 2 for 40 min at 37°C. The cell suspension was then added to 3 ml of a 2 M diS-C 3 -(5) solution in buffer A, and after 5 min of equilibration, the suspension was centrifuged at 1,000 ϫ g for 5 min. The fluorescence of the supernatant was measured and used to calculate the cells' membrane potential according to the method of Hladky and Rink (23), assuming a resting potential in fresh cells of Ϫ9 mV (23).

Effects of PGE 2 on Erythrocyte Filterability-It has been
shown previously that PGE 2 decreases the filterability of human and rat erythrocytes with maximum potency at 10 Ϫ10 M PGE 2 (12,13). We have confirmed these results (data not shown) and extended them by evaluating the time course of the change in erythrocyte filterability. Red cells were treated with 10 Ϫ10 M PGE 2 for various times at 37°C, and their rates of passage through a 3-m membrane filter were measured (Fig.  1). Maximal decrease in filterability occurred 40 min after exposure to PGE 2 , by which time the relative rate of flow through the filter had diminished ϳ30%. Longer incubation times resulted in a gradual return to normal filterability. Unstimulated cells incubated at 37°C for up to 60 min showed no significant change in filterability (data not show), indicating the observed changes were not simply a consequence of prolonged incubation.
Effect of PGE 2 on Cell Volume, Osmotic Fragility, and Density-Reductions in cell filterability can be caused by a decrease in cell surface to volume ratio, an increase in viscosity of the cytoplasm, or an elevation of membrane rigidity. Since the former two sources can arise from a change in cell volume, we evaluated the effect of PGE 2 on erythrocyte volume. As shown in Fig. 2, red cells treated with 10 Ϫ10 M PGE 2 exhibited a gradual decrease in size that reached a maximum of ϳ10% decrease by 40 min of PGE 2 treatment.
To confirm the observed PGE 2 -stimulated shrinkage, we also evaluated the effect of PGE 2 on cell osmotic fragility. Osmotic fragility is a measure of the ability of red cells to resist osmotic lysis and is generally accepted as a sensitive assay of an eryth- rocyte's surface to volume ratio. When untreated red cells were exposed to hypotonic buffer, they began to hemolyze at ϳ175 mosm and were almost completely lysed by 70 mosm (Fig. 3a). However, when red cells treated with 10 Ϫ10 M PGE 2 were similarly examined, onset of hemolysis occurred normally but completion of osmotic lysis was shifted to much lower osmolarities. At 100 mosm, for example, the number of lysed PGE 2treated cells was reduced by ϳ15% relative to controls (Fig. 3a). These data confirm that PGE 2 can promote erythrocyte shrinkage, and they also suggest that only a fraction of the cells may participate in the PGE 2 -triggered response.
To further evaluate the fractional participation of erythrocytes in the PGE 2 -induced cell shrinkage, the density distribution of erythrocytes was measured before and after treatment with 10 Ϫ10 M PGE 2 . As seen in Fig. 3b, hormone treatment promotes a shift in ϳ15% of the population to higher density. Curiously, the sensitive fraction of cells seems to reside in the middle of the gradient, suggesting the responsive cells are not primarily the youngest cells of the population, since such cells are commonly enriched in lower density fractions.
Effect of PGE 2 on K ϩ Efflux and Membrane Potential-A decrease in red cell volume usually results from cell dehydration that accompanies loss of cellular electrolytes. Since K ϩ is the main osmotic regulatory ion in human red cells (24), we undertook to look for a change in K ϩ efflux following addition of PGE 2 . As shown in Fig. 4a, all three PGE 2 concentrations tested stimulated a loss in cellular K ϩ . Also, in agreement with previous observations, 10 Ϫ10 M PGE 2 elicited the greatest efflux among the concentrations examined. In fact, when the rate of loss of cellular K ϩ was quantitatively evaluated over the first 5 min of stimulation, 10 Ϫ10 M PGE 2 was found to promote an average decrease of 36 mM/h in intracellular K ϩ content (Fig.  4b). In general, however, initial efflux rates were sustained only for 10 -20 min, after which they gradually declined to control levels (data not shown).
If K ϩ were to exit stimulated erythrocytes faster than its counterion, a change in membrane potential to a more negative value might be anticipated. To explore this possibility, a membrane potential-sensitive dye frequently employed in red cell studies (diS-C 3 -(5)) was used (23). DiS-C 3 -(5) is a hydrophobic cation that partitions into erythrocytes in proportion to their negative membrane potential, thereby allowing the use of the Nernst equation and the equilibrium ratio of dye activities inside and outside the cell to estimate changes in membrane potential (23). In our studies, changes in membrane potential were calculated based on the assumption that fresh control erythrocytes have a membrane potential of Ϫ9 mV (23). Pretreatment of cells with PGE 2 (10 Ϫ10 M) for 40 min at 37°C changed the membrane potential from Ϫ9 mV to Ϫ16 mV (Table I), reflecting a partially uncompensated loss of cations.
Identification of the PGE 2 -stimulated K ϩ Efflux Pathway-The partially uncompensated loss of cations from PGE 2 -stimulated erythrocytes suggests that the cation efflux is not mediated by either KCl or Na/K/2Cl cotransport, since these efflux pathways are electrically silent (25,26). 2 In contrast, RBCs are known to be equipped with a Ca 2ϩ -activated K ϩ channel (Gardos channel), whose stimulation leads to cellular dehydration and a decrease in membrane potential (24,27,28). To examine the possible involvement of the Gardos channel in the PGE 2 signaling pathway, we looked to see whether the stimulated K ϩ efflux might be dependent on extracellular Ca 2ϩ . For this purpose, 8 mM EGTA was added to the efflux buffer and PGE 2stimulated K ϩ efflux was again monitored. As revealed in After pelleting the intact cells, the supernatant was assayed for hemoglobin content to determine percent hemolysis, as described under "Experimental Procedures." Panel b, red cell suspensions of 10% hematocrit were incubated in the presence (f) or absence of (Ⅺ) 10 Ϫ10 M PGE 2 at 37°C for 40 min. After centrifugation, the cells were resuspended at 50% hematocrit in Buffer A, layered onto a stractan density gradient, and separated according to density as described under "Experimental Procedures." Each point represents the average value (Ϯ S.E. of the mean) obtained from duplicate samples. Where error bars are smaller than the size of the symbol, they do not appear on the graph. Similar results were observed from three independent experiments.
To further explore the possible involvement of the Gardos channel in the PGE 2 -mediated volume changes, the impact of two specific Gardos channel inhibitors, ChTX and clotrimazole, on the PGE 2 -activated K ϩ efflux was examined (22, 29 -31). Importantly, both inhibitors have been previously shown to block the K ϩ loss and dehydration induced in RBCs upon addition of the Ca 2ϩ ionophore, A23187 (30,31). In our studies, we adopted the inhibitory conditions for ChTX of Brugnara et al. (29,30), who found it useful to lower the extracellular Ca 2ϩ concentration to 50 M to avoid Ca 2ϩ inhibition of ChTX binding. In our hands, this decrease in extracellular Ca 2ϩ content simultaneously reduced the magnitude of the PGE 2 -stimulated K ϩ efflux, consistent with the observed Ca 2ϩ dependence of the efflux pathway. Further, pretreatment of red cells with 100 nmol/ liter ChTX was found to largely abolish the effect of PGE 2 on K ϩ loss (Fig. 6). This inhibitory effect of ChTX on the Gardos channel was also observed in related studies of erythrocyte membrane potential. As shown in Table I, ChTX also blocked the hyperpolarization of erythrocytes treated with either PGE 2 or A23187, i.e. confirming that the majority of the membrane potential change derives from K ϩ efflux through the Gardos channel.
In related studies, we employed a structurally distinct type of Gardos channel inhibitor, clotrimazole, and observed a similar blockade of PGE 2 -induced K ϩ efflux (Fig. 6).
To further confirm that the PGE 2 -induced cell volume change is dependent on the Ca 2ϩ -activated K ϩ channel, we have taken advantage of the prediction that a measurable volume decrease cannot occur unless a counterion, i.e. Cl Ϫ , can accompany K ϩ in exiting the cell. Since Gardos channels do not cotransport Cl Ϫ , any accompanying anion must move through an alternate pathway. The anion transport protein, band 3, represents an obvious candidate for this counterion movement; hence, we examined the impact of the anion transport inhibitor, DIDS, on the PGE 2 -induced cell volume and membrane potential changes. Pretreatment of red cells with 100 M DIDS at 37°C for 1 h completely blocked anion transport activity (data not shown) as well as PGE 2 -induced cell shrinkage (Fig.  7). In contrast, the hyperpolarization of the membrane potential induced by PGE 2 was not significantly affected by DIDS treatment (Table I), changing as usual to ϳϪ17 mV. We interpret these data to suggest that the PGE 2 -induced volume change can occur only if a functional anion transporter is available to carry counterions across the membrane. Otherwise, the decrease in membrane potential will effectively prohibit osmotically relevant quantities of K ϩ from leaving the cell.
Evaluation of the Involvement of other K ϩ Efflux Pathways-The data presented above argue strongly for a PGE 2 -activated Gardos channel, but they do not exclude the possible minor participation of a Cl Ϫ -dependent K ϩ transport pathway, e.g.

FIG. 4. Stimulation of K ؉ efflux by PGE 2 in human erythrocytes.
Red cell suspensions were incubated with or without PGE 2 at 37°C. The external K ϩ concentration was monitored continuously using a K ϩ -sensitive electrode, and these data were used to calculate the loss of intracellular K ϩ . Panel a, loss of cellular K ϩ from erythrocytes treated with 0 mM (Ⅺ), 10 Ϫ11 M (å), 10 Ϫ10 M (E), or 10 Ϫ9 M (ࡗ) PGE 2 as a function of time. Panel b, rate of loss of cellular K ϩ over the first 5 min. Each point represents the average value (Ϯ S.E. of the mean) obtained from two different experiments. Variations among the data are sometimes smaller than the size of the symbol and therefore do not appear on the graph. Rates of K ϩ loss are calculated assuming all erythrocytes participate in the efflux.

TABLE I Effect of PGE 2 on membrane potential of human erythrocytes
Red cell suspensions were incubated with or without 10 Ϫ10 M PGE 2 at 37°C for 40 min, and membrane potential was measured as described under "Experimental Procedures." In the inhibition studies, cells were pretreated with or without ChTX (100 nmol/liter) or DIDS (100 mol/ liter), as described in Figs. 6 and 7. Inhibitor-treated cells were also treated with or without 19 M A23187 for 5 min, following which membrane potential measurements were also obtained. The membrane potential for the control samples was assumed to be Ϫ9 mV. These data represent the average of two experiments Ϯ S.E. of the mean.

Treatment
Estimated membrane potential a mV Control Ϫ9 a Membrane potential measurements assume all erythrocytes in the suspension participate in the stimulated responses. KCl or Na/K/2Cl cotransport. To evaluate this possibility, we substituted NO 3 Ϫ for Cl Ϫ in the incubation buffer and re-examined PGE 2 activation of K ϩ efflux. As shown in Fig. 8, removal of external Cl Ϫ did not inhibit the PGE 2 -stimulated K ϩ efflux. Thus, with reasonable confidence it can be concluded that the Gardos channel is the only volume regulatory channel activated by PGE 2 . DISCUSSION We have demonstrated that PGE 2 reduces human erythrocyte filterability by stimulating K ϩ efflux leading to a loss of osmotic water and cell shrinkage. We have further shown that the discharged K ϩ exits the cell through the Gardos channel. Evidence for this latter contention derives from (i) the dependence of channel activation on external Ca 2ϩ , (ii) the inhibition of K ϩ efflux by two specific Gardos channel inhibitors (i.e. charybdotoxin and clotrimazole), (iii) the retention of channel function upon substitution of NO 3 Ϫ for Cl Ϫ , and (iv) the absence of electrical neutrality that is characteristic of the more prominent KCl and Na/K/2Cl cotransport pathways. To the best of our knowledge, this is the first documentation of ligand activation of a Gardos channel in any cell type. Nevertheless, several papers have reported PGE 2 activation of K ϩ efflux from nonerythroid cells without identifying the channel involved (32,33), raising the possibility that our observations on red cells may not be an isolated phenomenon.
With the limited information available, it is difficult to evaluate the physiological relevance of the PGE 2 -stimulated volume changes. Arguments suggesting that erythrocyte shrinkage might contribute to hemostasis include (i) cell shrinkage significantly impacts cell filterability, (ii) the observed volume changes are maximal at PGE 2 concentrations released by activated platelets (8), and (iii) some type of communication between erythrocytes and activated platelets has already been well documented (5-7). The major argument against a physiological role for the PGE 2 -induced volume decrease is that it appears to proceed too slowly to significantly impact clot formation, i.e. clotting times for whole blood treated with procoagulant ranges from 1 to 5 min. However, many physiological regulators are missing from the simplified red cell suspension employed in these studies, and it is conceivable that co-stimulation by other signals surrounding activated platelets could greatly synergize the effect of PGE 2 . Indeed, we have recently observed that thrombin elicits a similar group of signaling Red cell suspensions at 10% hematocrit were pretreated with or without 100 nM charybdotoxin for 100 min at room temperature, or with or without 100 nM clotrimazole for 20 min at 37°C. The RBC suspensions were then incubated in the presence or absence of 10 Ϫ10 M PGE 2 at 37°C, and the rate of decrease in cellular K ϩ content was determined over the first 10 min, as described in Fig. 4. Results shown are the mean of two experiments. Because of the lower Ca 2ϩ concentration permitted in the charybdotoxin study (see "Experimental Procedures"), the control rate and rate of PGE 2 -stimulated K ϩ efflux in these cells is lower than normal. Column a, control for charybdotoxin study; column b, PGE 2 treatment; column c, PGE 2 plus charybdotoxin treatment; column d, control for clotrimazole study; column e, PGE 2 treatment; column f, PGE 2 plus clotrimazole treatment.
intermediates in red cells as PGE 2 . 3 Whether synergy between platelet-derived signals actually occurs in vivo must obviously await further experimentation, but it would indeed be surprising if a cell that constitutes Ͼ99% of the total blood cell mass did not participate in some manner in hemostasis.
The data we have presented obviously require PGE 2 activation of a regulated Ca 2ϩ influx channel in human erythrocytes. At present, Ca 2ϩ is believed to enter red cells primarily via passive leaks, especially in sickle cells where membrane deformation is thought to permit transient ion fluxes in and out of the cell (34). However, the normal impermeability of erythrocytes to Ca 2ϩ plus the absolute requirement of the PGE 2stimulated Gardos channel for extracellular Ca 2ϩ argues that PGE 2 somehow induces a regulated Ca 2ϩ uptake. In nonerythroid cells, PGE 2 is thought to bind a classical G protein-coupled receptor that functions via a pertussis toxin-sensitive GTPbinding protein to activate phospholipase C (11,33,35). In this pathway, the subsequent release of inositol trisphosphate leads to Ca 2ϩ channel opening and a transient elevation of cytoplasmic Ca 2ϩ . It will be important to learn whether a similar signal transduction pathway operates in erythrocytes, and whether red cells express an inositol trisphosphate-activated Ca 2ϩ channel.
One major question raised by our data concerns the nature of the erythrocyte subpopulation that participates in the PGE 2stimulated cell shrinkage. Thus, both the osmotic lysis and density gradient analyses suggest that up to 85% of erythrocytes may remain refractory to PGE 2 (Fig. 3a). Unfortunately, the flow cytometric analyses are too insensitive to allow further confirmation or rebuttal of this estimate (Figs. 2 and 7). Furthermore, the cell filterability assays (Fig. 1), the K ϩ efflux experiments (Figs. 4 -6 and 8), and the membrane potential measurements (Table I) report only average values for the entire population of cells examined. Indeed, if it were assumed that only 15% of the erythrocytes were stimulated by PGE 2 , then the PGE 2 -induced changes in membrane potential and K ϩ efflux rates of participating red cells would have to be increased by a factor of ϳ6. Furthermore, the ability of the Ca 2ϩ ionophore A23187 to protect more red cells against osmotic lysis than PGE 2 (Fig. 3a) and to lower average erythrocyte membrane potential more dramatically than PGE 2 ( Table I) also suggests that a larger fraction of cells can respond to direct stimulation with Ca 2ϩ than to PGE 2 . Where the signaling pathway between PGE 2 binding and Gardos channel activation fails in nonparticipating erythrocytes is obviously an important question, but without more information it will have to await further scrutiny.
Finally, we undertook these studies to explore how erythrocytes might facilitate both platelet secretion and aggregation, as previously reported by others (5-7). We have observed that a hormone released by activated platelets can indeed stimulate erythrocytes, but we have learned nothing regarding how stim-ulated erythrocytes can enhance activation of platelets, i.e. as required by the fact that collagen or thrombin can activate platelets more rapidly and extensively in the presence than absence of erythrocytes. It is, therefore, conceivable that erythrocyte stimulation results in more elaborate biochemical changes than simply erythrocyte shrinkage, and that chemical messages of some type are returned from the stimulated erythrocyte to the platelet to augment the latter's activation process. Studies are currently under way to identify these messages.