Paclitaxel Affects Cytosolic Calcium Signals by Opening the Mitochondrial Permeability Transition Pore*

We have characterized the effects of the antimitotic drug paclitaxel (Taxol TM ) on the Ca 2 (cid:1) signaling cascade of terminally differentiated mouse pancreatic acinar cells. Using single cell fluorescence techniques and whole-cell patch clamping to record cytosolic Ca 2 (cid:1) and plasma membrane Ca 2 (cid:1) -dependent Cl (cid:2) currents, we find that paclitaxel abolishes cytosolic Ca 2 (cid:1) oscillations and in more than half of the cells it also induces a rapid, transient cytosolic Ca 2 (cid:1) response. This response is not affected by removal of extracellular Ca 2 (cid:1) indicating that paclitaxel releases Ca 2 (cid:1) from an intracellular Ca 2 (cid:1) store. Using saponin-permeabilized cells, we show that paclitaxel does not affect Ca 2 (cid:1) release from an inositol trisphosphate-sensitive store. Furthermore, up to 15 min after paclitaxel application, there is no significant effect on either microtubule organization or on endoplasmic reticulum organization. The data suggest a non-endoplasmic reticulum source for the intracellular Ca 2 (cid:1) response. Using the mitochondrial fluorescent dyes, JC-1 and Rhod-2, we show that paclitaxel evoked a rapid decline in the mitochondrial membrane potential and a loss of mitochondrial Ca 2 (cid:1) . Cyclosporin A, a blocker of the mitochondrial permeability transition pore, blocked both the paclitaxel-induced loss of mitochondrial Ca 2 (cid:1) and the effect on Ca 2 (cid:1) spikes. We conclude that paclitaxel exerts rapid effects on the cytosolic Ca 2 (cid:1) signal via the opening of the mitochondrial permeability transition pore. This work indicates that some of the more rapidly developing side effects of chemotherapy might be due to an action of antimitotic drugs on mitochondrial function and an interference with the Ca 2 (cid:1) signal cascade. an (cid:1) on (cid:1) a of paclitaxel and fluorescence as For two independent experiments we obtained a value of value an of

We have characterized the effects of the antimitotic drug paclitaxel (Taxol TM ) on the Ca 2؉ signaling cascade of terminally differentiated mouse pancreatic acinar cells. Using single cell fluorescence techniques and whole-cell patch clamping to record cytosolic Ca 2؉ and plasma membrane Ca 2؉ -dependent Cl ؊ currents, we find that paclitaxel abolishes cytosolic Ca 2؉ oscillations and in more than half of the cells it also induces a rapid, transient cytosolic Ca 2؉ response. This response is not affected by removal of extracellular Ca 2؉ indicating that paclitaxel releases Ca 2؉ from an intracellular Ca 2؉ store. Using saponin-permeabilized cells, we show that paclitaxel does not affect Ca 2؉ release from an inositol trisphosphate-sensitive store. Furthermore, up to 15 min after paclitaxel application, there is no significant effect on either microtubule organization or on endoplasmic reticulum organization. The data suggest a nonendoplasmic reticulum source for the intracellular Ca 2؉ response. Using the mitochondrial fluorescent dyes, JC-1 and Rhod-2, we show that paclitaxel evoked a rapid decline in the mitochondrial membrane potential and a loss of mitochondrial Ca 2؉ . Cyclosporin A, a blocker of the mitochondrial permeability transition pore, blocked both the paclitaxel-induced loss of mitochondrial Ca 2؉ and the effect on Ca 2؉ spikes. We conclude that paclitaxel exerts rapid effects on the cytosolic Ca 2؉ signal via the opening of the mitochondrial permeability transition pore. This work indicates that some of the more rapidly developing side effects of chemotherapy might be due to an action of antimitotic drugs on mitochondrial function and an interference with the Ca 2؉ signal cascade.
Antimitotic drugs are used extensively for the treatment of cancer. For example, paclitaxel (Taxol) is used in the treatment of breast and ovarian cancers and for AIDS 1 -related Kaposi's sarcoma, and vinblastine is used in the treatment of Hodgkin's disease (1). The mechanism of action of antimitotic drugs, that leads to cancer cell death, is not clear. It is known that paclitaxel stabilizes microtubule dynamics thereby preventing the proper formation of the mitotic spindle apparatus and arresting cancer cells at the G 2 -M phase of the cell cycle (2,3). While it is thought that this action of paclitaxel on the cell cycle machinery precedes an apoptotic response of cells (4,5), some recent work has suggested that paclitaxel-induced apoptosis results from more direct effects of the drug on the mitochondria. In this context paclitaxel has been shown to bind to Bcl-2 (6) and this binding may regulate Bcl-2 effects on the mitochondrial permeability transition pore (PTP) (7,8). Furthermore, deletion of the loop region of Bcl-2 (which prevents Bcl-2 phosphorylation) blocks the apoptotic action of paclitaxel on cancer cells (9). Other proteins may also be involved in the paclitaxel effects on mitochondria, such as APAF-1 (10). In isolated mitochondria paclitaxel acts to release cytochrome c (11). This effect is blocked by cyclosporin A providing further evidence that paclitaxel directly targets mitochondria, independent of actions on microtubules.
The effect of antimitotic drugs on microtubule dynamics would confer drug specificity on actively dividing cancer cells. However, the actions of these drugs on Bcl-2 and on the mitochondria might be expected to be non-selective and affect all cells. Indeed paclitaxel treatment is associated with serious side effects, including neuropathy (12) and low white blood cell counts (13). These side effects occur rapidly, appear to be due to drug action on terminally differentiated cells, and are slowly reversible. Thus it is unlikely that these side effects are mediated by either mitotic block or apoptosis. Given that the clinical use of antimitotic drugs is limited by these side effects, understanding the mechanisms by which these drugs act is an important step toward optimizing the therapeutic benefits.
In our studies, on terminally differentiated epithelial cells, we now show rapid actions of paclitaxel on the cytosolic Ca 2ϩ signal that can be accounted for by effects of paclitaxel on the PTP of the mitochondria. Given the universality of Ca 2ϩ signaling, it is likely that this action of paclitaxel accounts for some of the side effects of antimitotic drugs.
Patch Clamp-Standard whole cell patch clamp techniques (15) were employed and all experiments were carried out at room temperature (ϳ21°C). Pipettes had a resistance of 3-6 M⍀ (pipette puller, Brown and Flaming). After breaking through to the whole cell configuration pipettes had a measured, but uncompensated, series resistance of 10 -20 M⍀. The pipette solution contained (mM): 140 KCl, 1 MgCl 2 , 2 Na 2 ATP, 0.01 EGTA, (0.05 calcium green (Molecular Probes, Eugene, OR), 10 HEPES KOH, pH 7.2. The extracellular solution contained (mM) NaCl 135, KCl 5, MgCl 2 1, CaCl 2 1, glucose 10, NaOH-HEPES 10, pH 7.4. In the Ca 2ϩ -free solution, used in the experiments of Fig. 2, no Ca 2ϩ was added. Cells were held at a membrane potential of Ϫ30 mV and currents were sampled by an A/D converter (EPC-9, HEKA) at 2 KHz. In most experiments Ins(2,4,5)P 3 (gift from Professor R. Irvine, Cambridge, UK) was added to the pipette solution (10 M) to establish trains of Ca 2ϩ spikes. These spikes have previously been shown to result from InsP 3 -dependent Ca 2ϩ release in the secretory pole (16).
Fluorescence Imaging-Ca 2ϩ imaging experiments were performed as previously described (17). Briefly, inclusion of 40 -50 M Calcium Green in the pipette solution allowed imaging of the cytosolic Ca 2ϩ signal. Cells were illuminated at 488 nm (Coherent Innova 70) and imaged through a Nikon ϫ40 UV, 1.3NA, oil immersion objective through a 510-nm long pass filter. Full frame images (128 ϫ 128 pixels) were captured on a cooled CCD camera (70% quantum efficiency, 5 electrons readout noise; MIT, Lincoln Laboratories) with a pixel size of 200 nm at the specimen and at rates of up to 500 Hz. After recording to computer, the data were analyzed with custom software with bleach correction routines and appropriate smoothing. Images were displayed where F is the recorded fluorescence and F o was obtained from the mean of 20 sequential frames where no activity was apparent.
Single-cell Fluorescence Measurement-Cells were loaded with actetoxymethyl esters of various fluorescent probes (Molecular Probes) and the fluorescent signal measured from single cells with a Cairn Dual Emission Fluoresence System (Cairn Research, Faversham, UK). Measurement of mitochondrial membrane potential (⌬ m ) was carried out using the dye JC-1 loaded at a concentration of 1 M for 20 min at room temperature (22°C). The dye was excited at a wavelength of 488 nm and the emitted light collected at 590 and 530 nm. The ratio of red (590 nm) to green (530 nm) gives an index of the mitochondrial membrane potential: the higher the mitochondrial potential the greater proportion of JC-1 aggregates in the mitochondria, the greater the intensity of the red light signal and so with active mitochondria we see a relatively high ratio of 590/530 nm (18). To measure mitochondrial Ca 2ϩ signals we used Rhod-2AM at a concentration of 10 M loaded at 37°C for 40 min. This protocol specifically loaded the dye into the mitochondria as judged by the overlapping fluorescent signal from the mitochondrial dye Mitofluor (Fig. 6). In these experiments data are presented as F/F o , where F is the recorded fluorescence and F o was measured before drug application.
Estimates of Intracellular Paclitaxel Concentrations-In our experiments we used 10 -20 M paclitaxel applied to the bathing solution of the cells. To estimate the intracellular concentration reached over a period of 10 min we used Oregon Green-conjugated paclitaxel (Molecular Probes) and imaged intracellular fluorescence using a Zeiss LSM 510 confocal microscope. The intracellular fluorescence was measured over time within an 8-m spot centered on the cell with an optical section of 1 m. A calibration curve was made using a range of Oregon Green paclitaxel concentrations and recording the fluorescence signal as above. For two independent experiments we obtained a value of 322.5 nM as the peak intracellular concentration reached over 10 min. This value represents an upper limit for the cytosolic concentration of the drug, as it would be expected to bind to microtubules.
Ca 2ϩ Flux Experiments-Hepatocytes were isolated from livers of male Wistar rats and then permeabilized by incubation with saponin (10 g/ml) in cytosol-like medium containing 140 mM KCl, 20 mM NaCl, 2 mM MgCl 2 , 1 mM EGTA, 20 mM Pipes (pH 7 at 37°C). The cells were washed and resuspended (10 million cells/ml) in cytosol-like medium supplemented with CaCl 2 (free [Ca 2ϩ ] ϳ 200 nM), ATP (7.5 mM), carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP, 10 M), and 45 Ca 2ϩ (7.5 Ci/ml) (19). FCCP was used in these experiments to exclude Ca 2ϩ flux from mitochondria. After 5 min at 37°C, during which the intracellular stores loaded to steady-state with 45 Ca 2ϩ , thapsigargin (1.25 M) was added to inhibit further Ca 2ϩ uptake together with paclitaxel (20 M) or its solvent (Me 2 SO) and either a maximal (5 M) or submaximal (100 nM) concentration of Ins(1,4,5)P 3 . After a further 5 min, the incubations were terminated by rapid filtration and the 45 Ca 2ϩ contents of the stores determined. After correction for the 45 Ca 2ϩ nonspecifically bound to the cells (determined by addition of 10 M iono-mycin), the 45 Ca 2ϩ contents of the stores were expressed as percentages of the control 45 Ca 2ϩ content.
Immunocytochemistry-Cells were washed quickly in PBS (including calcium and magnesium) and then in K-Pipes buffer (K-Pipes 80 mM, EGTA 5 mM, MgCl 2 2 mM, pH 6.5). Cells were then fixed in 4% paraformaldehdye in K-Pipes buffer for 30 min and permeabilized in 0.1% Triton X-100 in PBS for 5 min. After a 1-h block in 2% donkey serum plus 2% fish skin gelatin in PBS, cells were incubated in primary antibody (␤-tubulin mouse monoclonal, clone TUB 2.1, Sigma, UK) for 1 h. After washing, a secondary antibody (donkey anti-mouse Fab fragment, Jackson Immunoresearch, West Grove, PA) conjugated to a fluorescein isothiocyanate fluorphore was applied for 30 min. The images of Fig. 6 were obtained with a Zeiss LSM510 confocal microscope with a ϫ63 planacromat oil immersion lens and excitation light provided by an argon and a HeNe laser. The emitted light collected at the indicated wavelengths. Confocal sections (1 m in depth) were obtained in the mid-section of the cells and image overlays were produced with proprietary Zeiss software.

RESULTS
We have previously used the whole cell patch clamp method to record the activation of Ca 2ϩ -dependent chloride currents (Cl (Ca) ) in single mouse pancreatic acinar cells (16). The Cl (Ca) current is commonly used as an indirect measure of cytosolic Ca 2ϩ and is a particularly good index of the small subplasmalemal Ca 2ϩ signals that occur in the apical pole of polarized epithelia (17). Fig. 1 shows the typical effects of paclitaxel on the whole cell Cl (Ca) , currents in a single pancreatic acinar cell. In a cell stimulated with Ins(2,4,5)P 3 to produce trains of Ca 2ϩ spikes (16), paclitaxel induced a rapid increase in the Cl (Ca) current followed by an abolition of the spike response (n ϭ 14/26, Fig. 1A). In the other 12 cells paclitaxel led to a gradual loss of the Cl (Ca) current spikes (Fig. 1B). Applied alone paclitaxel had no effect (n ϭ 3 cells, Fig. 1C) and Me 2 SO, the solvent in which paclitaxel was dissolved, at 1% (v/v) (a ϫ5-10 greater concentration than that used in the paclitaxel experiments) had no significant effect on the current spikes (Fig. 1D).
These data indicate that paclitaxel can, in many cells, promote a rapid activation of the Cl (Ca) current. Since the Cl (Ca) channels are primarily under the control of cytosolic Ca 2ϩ (20) this suggests that paclitaxel elicits a Ca 2ϩ signal. To test this directly we carried out Ca 2ϩ imaging experiments using the Ca 2ϩ -sensitive fluorescent dye Calcium Green introduced into the cell cytosol via infusion through the solution of the patch pipette (21). We further probed the possible source of the Ca 2ϩ signal induced by paclitaxel by removing extracellular Ca 2ϩ in these experiments. Fig. 2 shows a typical response, the infusion of Ins(2,4,5)P 3 resulting in the induction of small, short-lasting Ca 2ϩ spikes in the apical region of the cell (indicated secretory pole, SP, in Fig. 2). Application of 10 M paclitaxel induced a rapid Ca 2ϩ response that spread from the apical region to all regions of the cell (n ϭ 5/6 cells). In these experiments we were limited to capturing images over an 80-s period. However, a second series of images, captured later in the experiment, showed no evidence of a Ca 2ϩ signal, indicating that the paclitaxel-evoked Ca 2ϩ rise was transient and the local Ca 2ϩ spikes were abolished (data not shown). This is entirely consistent with the data shown in Fig. 1A. Since there is no extracellular Ca 2ϩ we conclude that the paclitaxel induced Ca 2ϩ response must originate from an intracellular Ca 2ϩ store.
The major Ca 2ϩ store in pancreatic acinar cells is in the endoplasmic reticulum (ER) where Ca 2ϩ is released in an IP 3dependent manner (22,23). To test if paclitaxel might be acting directly on IP 3 -dependent stores we utilized the permeabilized hepatocyte preparation where the Ca 2ϩ stores were preloaded with 45 Ca 2ϩ and the loss of radiolabeled Ca 2ϩ in response to Ins(1,4,5)P 3 was measured (19). The data in Table I show that a saturating concentration of 5 M Ins(1,4,5)P 3 releases about 30% of the Ca 2ϩ from the store and that this response is unaffected by the presence of paclitaxel. At an intermediate concentration of Ins(1,4,5)P 3 (100 nM) about 20% of the store Ca 2ϩ content is released and again paclitaxel has no significant effect on this. We conclude that the intracellular Ca 2ϩ store affected by paclitaxel is not likely to be IP 3 -dependent. However, our data do not exclude the possibility that, while paclitaxel may not act directly on IP 3 -dependent release of Ca 2ϩ , it may act indirectly by physically changing the store morphology and this in turn may affect store behavior. In this context we have previously shown that microtubule depolymerizing agents have dramatic effects on Ca 2ϩ signaling by locally and rapidly reorganizing the ER (24). We therefore went on to study the possible effects of paclitaxel on microtubular and ER organization.
To measure any effects of paclitaxel on microtubule and ER organization we carried out experiments to immunolocalize TABLE I IP 3 -dependent Ca 2ϩ release is unaffected by paclitaxel 45 Ca 2ϩ was preloaded into the Ca 2ϩ stores of permeabilized hepatocytes. The figures show the 45 Ca 2ϩ remaining after the various conditions, expressed as % of the total content. 5 M Ins(1,4,5)P 3 is a saturating concentration and the entire Ins(1,4,5)P 3 -dependent Ca 2ϩ store is emptied (19). In our experiments this represents about 30% of the total loaded 45 Ca 2ϩ . With 100 nM Ins(1,4,5)P 3 about 20% of the total store is emptied. In all conditions paclitaxel had no significant effect on Ins(1,4,5)P 3 -dependent release. The 45 Ca 2ϩ remaining after application of 5 M Ins(1,4,5)P 3 is presumed to represent Ca 2ϩ uptake into Ins(1,4,5)P 3 -independent pools such as secretory vesicles and endosomes. Data are expressed as the mean Ϯ S.E.  figure. Images b, d, and f were obtained at the peak of the Ca 2ϩ spikes and illustrate that the signal does not spread beyond the secretory pole region. In contrast, image h was obtained at the peak of the paclitaxel-induced response and shows a large Ca 2ϩ rise (pseudocolor red) globally across the whole of the cell.
␤-tubulin and to image the fluorescent probe, ER tracker, respectively. Standard microtubule-preserving techniques were used to fix the acinar cells (K-Pipes buffer) and we probed with a monoclonal antibody against ␤-tubulin. We then imaged clusters of acinar cells taking confocal sections through the cluster of cells at various depths. For clarity, Fig. 3 shows confocal sections only through the lower, middle, and the upper parts of the cell clusters. In the phase images shown the apical domain is seen in the central part of the cell clusters and is delineated by the presence of the phase-dark secretory granules. In control cells an extensive network of microtubules was observed with predominance in the apical domain, the region presumed to contain the microtubule organizing centers (25) (Fig. 3). Application of paclitaxel (20 M) did not dramatically alter this pattern of staining although the impression was that after 5 min treatment there was a structural reorganization that led to a criss-cross pattern of microtubules in the basal pole (two independent preparations). To test if this possible rearrangement of the microtubular system has any effect on the ER distribution we used ER tracker that has the key advantage of allowing visualization of the ER in single live cells. Fig. 4 shows a typical experiment with the ER tracker fluorescence excluded from the granular region of the secretory pole and from the nucleus. This distribution of fluorescence is consistent with the distribution of the ER as visualized with immunolocalization of the ER-resident protein calreticulin (24,26). To determine any changes in the ER distribution we recorded the average fluorescent signal in a 5-m diameter region in the apical (secretory pole) and divided this by the average signal in a 5-m diameter region in the basal pole. The ratio of secretory pole to basal pole fluorescence (SP/BP fluorescence ratio) did not change in control cells (circles, n ϭ 7 independent cells) and did not change in the presence of 10 M paclitaxel (triangles, n ϭ 10 independent cells). We conclude that paclitaxel does promote changes in the microtubular system but, at least over this time frame, there is no reorganization of the ER. These observations are consistent with early observations, in other cell types, of paclitaxel effects on microtubules and ER organization (27).
Another intracellular Ca 2ϩ store, significant in many cell types, is the mitochondria. In pancreatic acinar cells mitochondria have been shown to contain low levels of Ca 2ϩ at rest, but to rapidly take up Ca 2ϩ during IP 3 -evoked Ca 2ϩ responses (28). If paclitaxel acted to release Ca 2ϩ from mitochondria this would be consistent with the cytosolic Ca 2ϩ responses we observed. We therefore set out to determine possible effects of paclitaxel on mitochondria. If paclitaxel were stimulating a route of Ca 2ϩ exit from the mitochondria there are three known mechanisms where it might act. These are: (a) the PTP; (b) Na ϩ -dependent efflux; (c) Na ϩ -independent efflux (29). Ca 2ϩ efflux through any of these pathways would be expected to affect the mitochondrial membrane potential (⌬ m ). To record this we utilized the potential-sensitive probe JC-l. We observed a rapid decrease in the ratio of emitted light collected at 590 and 530 nm on addition of the mitochondrial uncoupler FCCP (1 M, Fig. 5A) that represents a drop in ⌬ m . In separate experiments, 10 M paclitaxel (n ϭ 8/11 cells, Fig. 5B) induced a decrease in ⌬ m . To determine if these effects of paclitaxel on the ⌬ m might be associated with a loss of mitochondrial Ca 2ϩ we carried out experiments to directly measure mitochondrial responses. We used Rhod-2/AM, a Ca 2ϩ -sensitive dye that can be specifically loaded into the mitochondria. To verify that our dye loading protocol led to a specific loading of the mitochondria we carried out control experiments where we also loaded the cells with Mitofluor Green/AM, a mitochondrial specific dye (30). The excitation/emission wavelengths of these dyes are widely separated and we ensured, using experiments with only a single dye, that there was no "bleed" from one fluorescence channel to the other. We then took confocal images of Rhod-2 and Mitofluor Green fluorescence in the same cell. The results indicated that Rhod-2 and Mitofluor Green were loaded into the same cellular compartment with a characteristic distribution (Fig. 6) similar to that previously described for mitochondria in these cells (31). This result indicates that the Rhod-2 signal we record from single cells is predominantly from the mitochondria.
First we measured the Rhod-2 response following stimulation of the cells with the agonist acetylcholine. In 3/4 cells, 100 nM acetylcholine induced an increase in the mitochondrial Ca 2ϩ as measured by an increase in the Rhod-2 fluorescence (Fig.  7A). Subsequent addition of 10 M paclitaxel induced a drop in mitochondrial Ca 2ϩ (n ϭ 4/4 cells) and a further decline was seen with the addition of 1 M FCCP (n ϭ 3/3 cells). In cells not stimulated with agonist the measured Rhod-2 response from a single cell showed a rapid decline in response to paclitaxel (n ϭ 5/8 cells) and a further decline in response to 1 M FCCP (n ϭ 6/7 cells, Fig. 7B). From these results we conclude that paclitaxel does induce a loss of Ca 2ϩ from the mitochondria. To test for the possible route of exit of Ca 2ϩ we pretreated the cells with cyclosporin A (10 M), a drug that binds to cyclophilin D and leads to the block of the mitochondrial PTP. The subsequent addition of paclitaxel either failed to induce any decrease in the Ca 2ϩ signal (n ϭ 5/8) or slowed the rate of decrease of the Ca 2ϩ signal (n ϭ 3/8, Fig. 7C). To quantify these effects we measured the paclitaxel-induced decrease in Rhod-2 fluorescence 30 s after drug application in the absence and presence of cyclosporin A (Fig. 7D). An FCCP induced decrease in Ca 2ϩ was still observed after pretreatment with cyclosporin A (n ϭ 6/6), presumably because Ca 2ϩ can still exit via other efflux mechanisms.
If the paclitaxel action on the cytosolic Ca 2ϩ signal was mediated by an effect on mitochondria then cyclosporin A might be expected to protect the cell from paclitaxel action. We tested this first with 10 M cyclosporin applied during a train of spikes induced by Ins(2,4,5)P 3 . However, the drug had effects alone and acted to increase spike activity (n ϭ 3, data not shown). A lower concentration of cyclosporin A (5 M) was therefore used. This had no effect on spike activity. After treating the cells for 5 min in cyclosporin A, we then applied paclitaxel. In contrast to control cells paclitaxel did not abolish spiking and had only a small transient effect on the signal (n ϭ 3/4 cells, Fig. 8). This data supports our hypothesis that the primary action of paclitaxel on the Ca 2ϩ signal cascade is mediated via an effect on the PTP. DISCUSSION The major finding of this study is that the anticancer drug paclitaxel rapidly releases Ca 2ϩ from the mitochondria of pancreatic acinar cells. Although the mechanism underlying this effect is not completely clear, the ability of cyclosporin A to block the Ca 2ϩ release points to an effect of paclitaxel on the PTP. We suggest that paclitaxel acts by partially opening the PTP. Ion flow through the PTP then causes a drop in mitochondrial membrane potential and a release of mitochondrial Ca 2ϩ . These effects may underlie some of the side effects of antimitotic drug treatments.
Effects of paclitaxel on mitochondria have been noted before. Paclitaxel phosphorylates Bcl-2 in cancer cells and induces apoptosis, possibly by acting on the PTP (4 -6). Andre et al. (11) have shown that paclitaxel induces cytochrome c release from a preparation of isolated mitochondria. The same study also showed that this cytochrome c release was blocked by pretreatment with cyclosporin A. These direct or indirect actions on mitochondria (32,33) may in turn generate the cytosolic Ca 2ϩ responses we observe.
Our measurements of the effects of paclitaxel on the Ins(2,4,5)P 3 -induced spikes indicate that in more than 50% of cells the drug induces a release of Ca 2ϩ from an intracellular store followed by abolition of the spikes. In the remaining cells paclitaxel led to a rapid abolition of spike activity. Why does the drug have these different actions? The same action on mitochondria may underlie both effects, with the difference being heterogeneity of mitochondrial Ca 2ϩ loading. In cells where the mitochondria were replete with Ca 2ϩ , paclitaxelinduced opening of the PTP would elicit a cytosolic Ca 2ϩ signal. However, in cells where the mitochondria were less well loaded the expected cytosolic Ca 2ϩ signal may be too small to detect. Pharmacological block of mitochondrial function has been shown to enhance cytosolic Ca 2ϩ responses in astrocytes (30) and acinar cells (31). These effects may be equivalent to the transient cytosolic Ca 2ϩ response to paclitaxel we see in the acinar cells. In all of our experiments we observed an eventual abolition of the Ca 2ϩ spikes. This appears to conflict with the enhancement of responses described above. However, block of mitochondrial function has been shown to inhibit Ca 2ϩ reuptake into Ca 2ϩ stores (34) and since the Ca 2ϩ spikes in acinar cells are dependent on cycles of Ca 2ϩ release and Ca 2ϩ uptake this effect would lead to a loss of spiking. Mitochondria are physically held in close proximity to sites of Ca 2ϩ release (35) and therefore are in a good position to modulate both Ca 2ϩ release and Ca 2ϩ reuptake (36).
The actual concentration of intramitochondrial Ca 2ϩ results from a balance of Ca 2ϩ uptake by the uniporter and a variety of Ca 2ϩ efflux mechanisms (29). During periods of cell stimulation where the cytosolic Ca 2ϩ is high, the mitochondrial Ca 2ϩ rises (35). After stimulation, the intramitochondria Ca 2ϩ returns back to lower levels. It is not clear if the PTP contributes to the endogenous mechanisms of Ca 2ϩ efflux from mitochondria but it may be important in processes of charge balance that would be required for Ca 2ϩ exit. What is clear is that a drug-induced opening of the PTP, as we postulate for the action of paclitaxel, would both reduce the mitochondrial potential and lead to Ca 2ϩ efflux. This would modify the cytosolic Ca 2ϩ responses, as we show in this paper. In addition, since the intramitochondrial Ca 2ϩ signal has been shown to actively regulate mitochondrial functions such as Ca 2ϩ -dependent dehydrogenases (37), these would be expected to be affected by paclitaxel, and would lead to a loss of ATP production and the subsequent compromise of many other cell functions.
Our experiments indicate that the PTP plays a role in the paclitaxel-mediated responses but exactly how the PTP is affected is unclear. One possibility, put forward by Andre et al. (11) to explain paclitaxel effects on isolated mitochondria, was that the mitochondria might contain tubulin within their structure. In this model the binding of paclitaxel to this tubulin might influence PTP function either through a tubulin interaction with the voltage-dependent anion channel present on the outer mitochondrial membrane (7) or through the adenine nucleotide transporter at the inner membrane, both of which are thought to form the PTP. While this is an attractive hypothesis, other experiments clearly indicate Bcl-2 as a possible target for paclitaxel (6,9). Bcl-2 is resident in the mitochondrial membrane and it is thought that drug-induced phosphorylation may lead to PTP FIG. 7. Mitochondrial Ca 2؉ is decreased by paclitaxel and by FCCP.
In these experiments cells were loaded with Rhod-2 under exactly the same conditions as in experiments of Fig. 6. A, single cell fluorescence measurements showed an increase in mitochondrial Ca 2ϩ on addition of 100 nM acetylcholine and a drop in Ca 2ϩ on addition of paclitaxel to the bathing solution and a further drop when FCCP was added. B, in unstimulated cells the same paclitaxel and FCCP induced drop in Ca 2ϩ was observed. C, pretreatment with cyclosporin A inhibited the paclitaxel effect but did not block FCCP-induced reductions in Ca 2ϩ . D, graph showing the mean (Ϯ S.E., n ϭ 8) drop in Rhod-2 fluorescence induced by paclitaxel alone, or in the presence of cyclosporin A. The difference in the two points was significant at p Ͻ 0.05 (Student's t test). opening and the triggering of cyctochrome c release from mitochondria (9). There are further complications in understanding the effects of paclitaxel on the PTP since it has been shown that at least two possible states of PTP opening are possible. In both states cytochrome c release is observed but in only one of the states is a change in ⌬⌿ m seen (39,40). In one study the change in ⌬⌿ m was associated with mitochondrial swelling and was specifically induced by high, but not low, concentrations of Bax, the proapoptic agent (39). In another study different states were differentially induced by A23187 or arachadonic acid and cell death was only associated with the changes in ⌬⌿ m (40). We show that paclitaxel induces a rapid and dramatic change in ⌬⌿ m , Andre et al. (11) showed paclitaxel-induced mitochondrial swelling, and combined, these data suggest paclitaxel acts to induce a more sustained opening of the PTP. This would lead to paclitaxel-induced apoptosis as shown in cancer cells (10). However, in treating cancers paclitaxel is applied as a short infusion at a high concentration (41) and, in terminally differentiated cells, this transient exposure to the drug may disrupt mitochondrial function but not trigger cell death.
Although our experiments indicate that the effects of paclitaxel on the Ca 2ϩ signal are due to a rapid action on mitochondria, we cannot completely rule out other effects, on either the microtubules or the ER, that may contribute to the Ca 2ϩ response. The microtubule distribution appears little effected by paclitaxel over the time course we studied, but there may be more subtle effects on signaling cascades that might lead to modification of the Ca 2ϩ response. For example, it has been shown that microtubule stabilization leads to a rapid activation of Rac1 (38) and this could conceivably influence Ca 2ϩ release in our system. Although we do not see any effects of paclitaxel on ER distribution, ER integrity (42) or positioning (27) may be disrupted in a way not resolved by our techniques.
We used 10 -20 M paclitaxel, which is a similar concentration to that used in other studies (e.g. Ref. 11). It is also in the range of the peak plasma concentrations reached in the clinical situation during treatment regimes. For example, a typical treatment of a 6-h infusion of Taxol at 275 mg/m 2 leads to a peak plasma concentration of 8.11 M (42). The concentrations we use are therefore in this clinically relevant range. However, they are higher than the concentrations of paclitaxel (10 -100 pM) used to show effects of mitotic block in cultured cells (2). Paclitaxel accumulates in cells and tissues. For example during long-term incubation with 100 pM, intracellular concentrations of 40.5 M are reached (2). We have directly measured intracellular paclitaxel concentrations, using a fluorescent conjugate and show that the acute exposure of the cells to paclitaxel results in a very low initial uptake into the cells (see "Experimental Procedures"). Although it remains to be determined if the pharmacokinetic properties of the fluorescent paclitaxel are the same as the native drug. We conclude that the concentrations of paclitaxel we used in our study are relevant to the clinical use of the drug and therefore the acute effects we observe may underlie some of the drug side effects.