Comparison of Paclitaxel-, 5-Fluoro-2′-deoxyuridine-, and Epidermal Growth Factor (EGF)-induced Apoptosis

Epidermal growth factor (EGF), a hormone that stimulates proliferation of many cell types, induces apoptosis in some cell lines that overexpress the EGF receptor. To evaluate the mechanism of EGF-induced apoptosis, MDA-MB-468 breast cancer cells were examined by microscopy, flow cytometry, immunoblotting, enzyme assays, and affinity labeling after treatment with EGF, paclitaxel, or 5-fluoro-2′-deoxyuridine (5FUdR). Apoptosis induced by all three agents was accompanied by activation of caspases-3, -6, and -7, as indicated by disappearance of the corresponding zymogens from immunoblots, cleavage of substrate polypeptides in situ, and detection of active forms of these caspases in cytosol and nuclei using fluorogenic assays and affinity labeling. Further analysis indicated involvement of the cytochrome c/Apaf-1/caspase-9 pathway of caspase activation, but not the Fas/Fas ligand pathway. Interestingly, caspase activation was consistently lower after EGF treatment than after paclitaxel or 5FUdR treatment. Additional experiments revealed that the majority of cells detaching from the substratum after EGF (but not paclitaxel or 5FUdR) were morphologically normal and retained the capacity to readhere, suggesting that EGF-induced apoptosis involves cell detachment followed by anoikis. These observations not only indicate that EGF- and chemotherapy-induced apoptosis in this cell line involve the same downstream pathways but also suggest that detachment-induced apoptosis is responsible for the paradoxical antiproliferative effects of EGF.

EGF 1 is a 6-kDa polypeptide that binds to a 170-kDa cell surface receptor (the EGFR) expressed on a wide variety of normal and neoplastic cells (reviewed in Refs. [1][2][3][4]. The interaction of these two molecules results in activation of the EGFR tyrosine kinase, which in turn activates the mitogen-activated protein kinase and Janus kinase signaling pathways (reviewed in Refs. 1-4; see also Refs. 5 and 6). In most target cells, these events result in proliferation (1)(2)(3)(4). In a number of cell lines, however, EGF paradoxically inhibits proliferation (7)(8)(9)(10)(11)(12). Studies of A431 epidermoid carcinoma cells and MDA-MB-468 breast cancer cells suggest that this inhibition of proliferation results from EGF-induced apoptosis (5,13,14). Because the mechanism by which EGF induces apoptosis is not known, the physiological significance of these findings is unclear.
Apoptosis is a morphologically and biochemically distinct form of PCD that occurs in many cell types after withdrawal of trophic stimuli (15)(16)(17)(18) or treatment with a wide variety of cytotoxic agents (19,20). Current models separate the apoptotic process into at least two distinct phases, initiation and execution. The initiation phase involves biochemical changes that might be unique to each apoptotic stimulus (21). In contrast, the execution phase involves a series of stereotypic morphological and biochemical changes (19) that appear to result from the action of cysteine-dependent aspartate-directed proteases called caspases (22).
The mechanism by which caspases are activated after treatment with chemotherapeutic agents has been the subject of considerable investigation. The Fas/FasL pathway (reviewed in Refs. 27-30, 32, and 37) has been implicated in the initiation of apoptosis in methotrexate-or doxorubicin-treated CEM T cell leukemia cells (38,39), etoposide-or teniposide-treated Jurkat T cells (40), bleomycin-treated HepG2 hepatoma cells (41), and 5-fluorouracil-treated Gc3/cl colon cancer cells (42), but other studies have found no evidence for Fas/FasL interactions in drug-induced apoptosis (43)(44)(45). Likewise, release of cytochrome c to the cytosol has been observed early in the process of drug-induced apoptosis in some model systems (46 -49) but not others (50,51).
Once caspase activation is initiated, the work of disassembling the cell then falls to caspases-3, -6, and -7 and the downstream activities that they activate (24,31,32,52). These enzymes cleave a number of cellular polypeptides, inactivating some and producing enzymatically active forms of others (29,32). The net result of these cleavages is the disassembly of key structural components of the nucleus and cytoskeleton; inhibition of the processes of DNA repair, replication, and transcription; and activation of one or more endonucleases that irreversibly damage the genome (reviewed in Refs. 31 and 32). Collectively, these events contribute to the biochemical and morphological changes that make up the apoptotic phenotype.
Previous studies from our laboratories have demonstrated that MDA-MB-468 breast cancer cells undergo apoptosis after treatment with 5FUdR, paclitaxel, or EGF (13,53,54). The caspase substrates PARP and lamin B 1 were cleaved in all cases, although the cleavage was less extensive in EGF-treated cultures. The identity of the activated caspases and the mechanism of their activation were not addressed in these earlier studies. More recently, Chin et al. (5) reported that EGF treatment of MDA-MB-468 cells resulted in enhanced expression and activation of caspase-1 without any activation of caspase-3. In view of previous results indicating that caspase-1 plays a limited role in apoptosis under many circumstances (55,56), it was unclear whether the EGF-induced expression of caspase-1 was sufficient to explain the apoptosis observed in these cells. Accordingly, the present studies were undertaken to 1) characterize the complement of caspases activated during apoptosis induced by paclitaxel, 5FUdR, and EGF; and 2) compare the mechanisms involved in caspase activation after these treatments.
Cell Culture-MDA-MB-468 cells (American Type Culture Collec-tion, Manassas, VA) were cultured in improved minimal essential medium (Biofluids, Rockville, MD) supplemented with 5% heat-inactivated fetal bovine serum, 100 units/ml penicillin G, 100 g/ml streptomycin, and 2 mM glutamine (medium A) in a humidified atmosphere containing 5% (v/v) CO 2 . When cells were 40 -60% confluent, medium A was replaced by medium A containing 100 M 5FUdR, 100 ng/ml EGF, or diluent; and incubation was continued for 48 or 72 h, respectively. Alternatively, after medium A containing 100 nM paclitaxel was added, cells were incubated for 24 h, washed, and incubated in drug-free medium A for 24 h. These treatments were previously shown to induce apoptosis in this cell line (13,53,54). At the completion of the treatment, nonadherent cells were removed with the tissue culture medium, and adherent cells were released by trypsinization. Cells were sedimented at 200 ϫ g for 10 min and processed for the various biochemical assays described below. Alternatively, in some experiments, nonadherent cells were sedimented at 100 ϫ g, resuspended in fresh medium A, and seeded on new tissue culture plates.
After an additional 72 h, the nonadherent and adherent fractions were collected, counted on a hemacytometer in the presence of 0.2% trypan blue, and fixed for morphological examination as described below.
To assess the effect of growth under conditions in which cells could not adhere, semiconfluent MDA-MB-468 cells were released by trypinization and suspended in drug-free medium A at a concentration of 3-5 ϫ 10 5 cells/ml. Aliquots (1 ml) in medium A were incubated for up to 72 h in 17 ϫ 100-mm test tubes. Under these conditions, Ͼ95% of HL-60 human leukemia cells survive (58). In other experiments, tissue culture plates were treated with poly(HEMA) exactly as described by Folkman and Moscona (59). After the diluent was evaporated, cells were plated, allowed to adhere for 14 -16 h, and treated with 100 ng/ml EGF for 72 h. Adherent and nonadherent cells were then separately harvested.
Jurkat cells (kindly provided by Drs. C. M. Eischen and P. Leibson, Mayo Clinic) were grown in RPMI 1640 medium containing 5% heatinactivated fetal bovine serum, 100 units/ml penicillin G, 100 g/ml streptomycin, and 2 mM glutamine at concentrations below 1 ϫ 10 6 /ml. To examine the effects of the ZB4 blocking antibody, Jurkat cells were treated for 24 h with 20 ng/ml CH-11 in the absence or presence of 1 g/ml ZB4.
Immunoblotting-Sedimented cells were washed once with ice-cold RPMI medium containing 10 mM HEPES (pH 7.4 at 4°C) and lysed by vigorous vortexing in 6 M guanidine hydrochloride containing 250 mM Tris-HCl (pH 8.5 at 4°C), 10 mM EDTA, 150 mM ␤-mercaptoethanol, and 1 mM phenylmethylsulfonyl fluoride. After sonication, samples were treated with iodoacetamide to block free sulfhydryl groups and then dialyzed sequentially into 4 M urea and 0.1% (w/v) SDS as described previously (57). After an aliquot was removed for determination of protein (60), the sample was lyophilized to dryness; resuspended at a concentration of 5 mg protein/ml in SDS sample buffer consisting of 4 M deionized urea, 2% (w/v) SDS, 62.5 mM Tris-HCl (pH 6.8 at 21°C), and 1 mM EDTA; and heated to 65°C for 20 min. Aliquots containing 50 g of total cellular protein were subjected to SDS-PAGE on gels with 5-15% (w/v) acrylamide gradients, transferred to nitrocellulose or polyvinylidene difluoride, and probed with antibodies as described (24).
Cell Fractionation, Caspase Assays, and Affinity Labeling-Cytosol and nuclei were prepared at 4°C as recently reported (24) except that the previously described sedimentation of nuclei through 2.1 M sucrose was omitted because tenaciously adherent cytoskeletal components in MDA-MB-468 cells precluded purification by this technique. Instead, cells homogenized in Buffer A (25 mM HEPES (pH 7.5 at 4°C), 5 mM MgCl 2 , 1 mM EGTA supplemented immediately before use with 1 mM phenylmethylsulfonyl fluoride, 10 g/ml pepstatin A, and 10 g/ml leupeptin) were sedimented at 800 ϫ g. The supernatant was removed and sedimented at 280,000 ϫ g max to prepare cytosol (24). After the crude nuclei in the 800 ϫ g pellet were resuspended in Buffer A and resedimented at 800 ϫ g, they were stored in small aliquots at Ϫ70°C in Buffer A containing 5 mM EDTA and 2 mM dithiothreitol.
Cleavage of DEVD-AFC, VEID-AFC, and YVAD-AFC by caspases present in cytosol and nuclei was assayed as described previously (24). All reactions were run as end point assays using 100 M substrate concentrations, 2-h reaction times at 37°C, and 50-g aliquots of cytosolic or nuclear protein. Standards containing 0 -1500 pmol of AFC were utilized to determine the amount of fluorochrome released. Control experiments indicated that the assays were linear with respect to incubation time and enzyme content under these conditions. Cytosol from etoposide-treated HL-60 leukemia cells served as a positive control for DEVD-AFC and VEID-AFC cleavage, and cytosol from THP.1 monocytic leukemia cells was a positive control for YVAD-AFC cleavage (24).
For affinity labeling of active caspases, aliquots containing the indi-cated amounts of nuclear or cytosolic protein were incubated for 1 h at room temperature with 1 M Z-EK(bio)D-aomk (24), diluted with 1 ⁄2 volume of 3ϫ concentrated SDS sample buffer, heated to 95°C for 3 min, subjected to one-dimensional SDS-PAGE on 16% (w/v) acrylamide gels, transferred to nitrocellulose, probed with peroxidase-labeled streptavidin, and visualized using ECL reagents. Two-dimensional analysis was performed using isoelectric focusing for the first dimension and SDS-PAGE for the second dimension as described (24), except that precast pH 4 -7 Immobiline gels (Pharmacia, Uppsala, Sweden) were utilized for isoelectric focusing. Labeled polypeptides were visualized using peroxidase-coupled streptavidin followed by SuperSignal ULTRA chemiluminescent substrate. As a control, recombinant caspases expressed in Sf9 cells (24) were subjected to the same analysis.
Fluorescence Microscopy and Quantitation of Apoptotic Cells-Adherent and nonadherent cells were collected separately, sedimented at 200 ϫ g for 10 min, washed with ice-cold calcium-and magnesium-free phosphate-buffered saline, fixed in 3:1 (v/v) methanol:acetic acid, stained with 1 g/ml Hoechst 33258, and examined by fluorescence microscopy as described previously (13,58). A minimum of 300 cells/ sample were scored for apoptotic changes (fragmentation of the nucleus into multiple discrete fragments). Samples were photographed using Eastman Kodak Co. Elite II ASA 400 film and a 1 ⁄4 s exposure time.
Flow Cytometry-Adherent and nonadherent cells were collected separately, sedimented at 200 ϫ g for 10 min, washed with ice-cold phosphate-buffered saline, fixed in 50% ethanol, treated with 1 mg/ml RNase A, stained with 100 g/ml propidium iodide, and subjected to flow cytometry as described previously (61).
Statistical Analysis-Experiments were replicated as indicated in the figure legends. The statistical significance of differences between treated and untreated cells was assessed using two-sided t tests and a post hoc Bonferroni correction to take into account the effect of multiple comparisons (62).

EGF-induced Apoptosis Is Mediated by the EGFR Tyrosine
Kinase-Although EGF stimulates the proliferation of most cells, it paradoxically inhibits the growth of a number of cell lines (5,(7)(8)(9)(10)(11)(12)(13)(14). To begin to elucidate the mechanism of this latter effect, MDA-MB-468 cells were treated with 100 ng/ml EGF, a concentration previously shown to result in apoptosis in this cell line (13). To provide a basis for comparison, MDA-MB-468 cells were also treated with the chemotherapeutic agents paclitaxel (100 nM) or 5FUdR (100 M) under conditions previously shown to induce PCD in these cells (53,54). Initial examination revealed that treatment with paclitaxel for 24 h resulted in accumulation of cells in mitosis. The paclitaxeltreated cells had a tetraploid DNA content (Fig. 1B) and contained condensed chromosomes when examined after staining with Hoechst 33258 (Fig. 1B, inset). In contrast, 5FUdR caused a marked decrease of cells in G 2 , with a concomitant accumulation of cells in G 1 and early S phase (Fig. 1C). Finally, treatment with EGF for 24 h caused a decrease in the S phase population (cf. Fig. 1, A and D) with accumulation of cells in G 1 and G 2 phases of the cell cycle. Further EGF treatment, however, was associated with recovery of the S phase population (Fig. 1D, inset) followed by detachment of a portion of the cells from the tissue culture plate (Fig. 1E) (13).
To confirm that the detachment process initiated by EGF was mediated by the EGFR, cells were treated with EGF in the presence or absence of PD153035 at a concentration (5 M) that has been observed to inhibit the EGFR tyrosine kinase (63,64) without itself inducing apoptosis (64). PD153035 inhibited the EGF-induced detachment of MDA-MB-468 cells (Fig. 1E), suggesting that signaling through the EGFR is required for EGFinduced apoptosis. In contrast, PD153035 did not have any effect on paclitaxel-or 5FUdR-induced detachment (data not shown).
Comparison of Polypeptide Cleavages after Treatment with EGF, Paclitaxel, or 5FUdR-In further experiments, nonadherent and adherent cells were separated and examined for evidence of caspase activation. Nonadherent cells harvested after treatment with paclitaxel or 5FUdR (Fig. 2, lanes 4 and 6, respectively) contained diminished levels of full-length PARP and the lamins. Fragments of PARP, lamin A, and lamin B 1 were detected (Fig. 2, arrowheads). The same polypeptides were also cleaved in nonadherent cells resulting from EGF treatment (Fig. 2, lane 8), although cleavage was less complete than in paclitaxel-or 5FUdR-treated cells. In all cases, the corresponding polypeptides remained intact in cells that remained adherent (Fig. 2, lanes 3, 5, and 7), suggesting that proteolysis was occurring concomitant with or subsequent to detachment of cells from the substratum. FAK, another putative caspase substrate (32), was also diminished in the nonadherent cells, although discrete fragments of FAK were not detectable using this antibody (Fig. 2).
The cleavage of PARP, lamin A, and lamin B 1 to their signature fragments (65)(66)(67)(68) suggested that caspases had been activated by the three treatments. Based on the observation that PARP is cleaved efficiently by caspases-3 and -7 (69), whereas lamin A is cleaved preferentially by caspase-6 (67), we initially focused on these three caspases. Immunoblotting of the same cell lysates revealed that the precursors for caspases-3, -6, and -7 were extensively or quantitatively decreased in nonadherent cells after treatment with paclitaxel or 5FUdR (Fig. 3, A-C, lanes 4 and 6). In the case of caspases-6 and -7, fragments corresponding to the large subunits of the active enzymes were detectable in some apoptotic cell fractions (Fig. 3, B and C, arrow). Cleavage of the procaspases was less complete in the nonadherent cells resulting from EGF treatment but was nonetheless evident (Fig. 3, A-C, lane 8). Two closely spaced isoforms of procaspase-1 were also detectable in MDA-MB-468 cells (Fig. 3, D, lanes 1 and 2). In cells that remained adherent after treatment with paclitaxel, 5FUdR, or EGF, levels of these isoforms remained constant (Fig. 3D, lanes 5 and 7) or decreased slightly (Fig. 3D, lane 3). Although these procaspase-1 species decreased markedly in nonadherent cells (Fig. 3D, lanes 4, 6, and 8), bands corresponding to the large subunit of active caspase-1 were not detectable even upon prolonged exposure of the blot. Procaspase-2, which did not appear to be cleaved after any of the treatments (Fig. 3E, lanes  4, 6, and 8), served as a loading control.
Comparison of Active Caspases after Treatment with EGF, Paclitaxel, or 5FUdR-To determine whether the disappearance of procaspases reflected the generation of enzymatically active proteases, cleavage of fluorogenic substrates that correspond to preferred cleavage sites of caspase-3 (DEVD-AFC) and caspase-6 (VEID-AFC) was examined (Fig. 4, A and B). Activities that cleaved both of these substrates were markedly increased in cytosol of cells that became nonadherent after treatment with paclitaxel or 5FUdR (Fig. 4, A and B, solid bars). Smaller, but statistically significant, increases in activity were also observed in cells that remained adherent after these two treatments, suggesting that caspase activation precedes detachment after exposure to these drugs. Increased DEVD-AFC and VEID-AFC cleavage activities were also detected after EGF treatment. Two differences, however, were noted. First, the activities observed in cytosol of nonadherent cells after EGF treatment were consistently ϳ2-fold lower than activities observed after drug treatment (Fig. 4, A and B, solid bars). Second, the activities were not significantly increased in cells that remained adherent after EGF treatment.
Because of the suggestion that caspase-1 might play a role in triggering EGF-induced apoptosis, activity that cleaved YVAD-AFC, a fluorogenic substrate of caspase-1, was also measured after paclitaxel, 5FUdR, or EGF treatment. This activity increased less than 2-fold after treatment (Fig. 4C) and was very low compared with the other activities (compare y axes in Figs. 4, A-C), suggesting that caspase-1 activation is not a major feature of apoptosis in this cell line.
Recent results (24,49) indicate that active caspases can also be detected in nuclei of apoptotic leukemia cells. Consistent with these results, activities that cleave DEVD-AFC and VEID-AFC (but not YVAD-AFC) were easily detected in nuclei from nonadherent MDA-MB-468 cells harvested after treatment with paclitaxel, 5FUdR or EGF (Fig. 4, A and B, hatched bars). These activities were also elevated in nuclei from adherent cells after paclitaxel or 5FUdR.
To provide more complete identification of the caspases that are activated in MDA-MB-468 cells, cytosol and nuclei were reacted with the affinity label Z-EK(bio)D-aomk (24,49), subjected to one-and two-dimensional gel electrophoresis, and reacted with peroxidase-coupled streptavidin. Results of the one-dimensional analysis (Fig. 5A) revealed that multiple active caspase species were detectable in cytosol and nuclei after all three treatments. Interestingly, none of these species migrated with the large subunit of active caspase-1 (Fig. 5A, cf.  lane 0 with lanes 4 -6 and 9 -11). Although samples from EGF-  Fig. 3 are from the same experiment. Arrowheads, previously described proteolytic fragments of PARP, lamin A, and lamin B 1 (66 -68). The lower intensity of the signal for fragments of lamins A and B 1 as compared with the intact polypeptides (cf. lanes 5 and 6) suggests that multiple epitopes recognized by the polyclonal sera have been lost from these carboxyl-terminal fragments. Results are representative of three to six experiments with each treatment. Fig. 2 were reprobed with antibodies to procaspases. MDA-MB-468 cells that became nonadherent (NA) or remained adherent (A) after treatment with paclitaxel, 5FUdR, or EGF were subjected to immunoblotting with reagents that recognized procaspase-3 (panel A), procaspase-7 (panel B), procaspase-6 (panel C), or procaspase-1 (panel D). All lanes were loaded with 50 g of protein.

FIG. 3. Effect of paclitaxel, 5FUdR, and EGF on caspases. A-E, blots shown in
To confirm equal loading, blots were reprobed with antibodies to procaspase-2 (panel E), a polypeptide that does not appear to be activated in chemotherapyinduced PCD in other model systems (24,44,96). Arrowheads, active large subunits resulting from proteolytic cleavages of procaspase-7 and procaspase-6, respectively. Results are representative of three separate experiments. treated cells usually contained less signal than samples from paclitaxel-or 5FUdR-treated cells, qualitative differences among the treatments were not discernible at this level of resolution.
In further experiments, the active caspase species present in nuclei and cytosol were compared by two-dimensional gel electrophoresis (Fig. 5B). Previous results in leukemia cells have demonstrated that the major caspase species detected by this technique are active forms of caspases-3 and -6 (24, 52). Two major and several minor caspase species were detected in cytosol of paclitaxel-and EGF-treated cells by this method. Interestingly, several additional species (Fig. 5B, numbered arrowheads) were detected in the cytosol from 5FUdR-treated cells. With one exception (Fig. 5B, arrowhead 1, 5FUdR), all of these species were detectable in nuclei as well. In general, nuclei contained higher amounts of many of the more weakly labeled species, raising the possibility that certain species have been specifically targeted or activated in this organelle. In addition, it appeared that one caspase species (Fig. 5B, indicated by arrowhead 2 in cytosol and circles in nuclei) was preferentially present in cytosol of 5FUdR-treated cells and nuclei of paclitaxel and EGF-treated cells, raising the possibility that nuclear accumulation of some caspase species might vary with the apoptotic stimulus. Despite these differences, the overall impression was that caspase species activated by the three treatments were generally similar.
Evaluation  Fig. 6D). This treatment failed to induce PCD in MDA-MB-468 cells (Fig. 6, B and E), suggesting that these cells are resistant to cytotoxic effects of Fas ligation. To further evaluate the potential role of the Fas/FasL system in MDA-MB-468 PCD, the effects of the blocking anti-Fas antibody ZB4 (44, 71) were evaluated. Although ZB4 blocked induction of apoptosis in thymidine-deprived human colon cancer cells (42) and in Jurkat cells subjected to T cell receptor cross-linking (44) or Fas ligation (Fig.  6F), it did not alter the number of MDA-MB-468 cells that detached from the plates (Fig. 6E) or developed apoptotic morphological changes (Fig. 6F)

FIG. 5. Affinity labeling of active caspases after treatment of MDA-MB-468 cells with paclitaxel, 5FUdR, or EGF.
A, nonadherent and adherent cells were collected after treatment with 100 nM paclitaxel for 24 h followed by a 24-h incubation in drug-free medium, 100 M 5FUdR for 48 h, or 100 ng/ml EGF for 72 h. In each case, cytosol and crude nuclei prepared as described under "Experimental Procedures" were incubated with 1 M zEK(bio)D-aomk, subjected to onedimensional SDS-PAGE followed by blotting with peroxidase-coupled streptavidin. Mobilities of the large subunits of recombinant caspases (determined as described previously in Ref. 24) are indicated in lane 0 for comparison. B, after affinity labeling as described for A, cytosol and nuclei were subjected to isoelectric focusing followed by SDS-PAGE as indicated. Numbered arrowheads indicate species present in cytosol from 5FUdR-treated cells but not other cells. The circled species in paclitaxel-and EGF-treated nuclei appears to be diminished in 5FUdRtreated nuclei but enhanced in 5FUdR-treated cytosol.
of EGF-induced apoptosis, we examined the possibility that activation of downstream caspases might involve the release of cytochrome c from mitochondria to cytosol (see the Introduction). Cytochrome c was readily detectable in cytosol from nonadherent MDA-MB-468 cells after treatment with paclitaxel, 5FUdR, or EGF (Fig. 7). Thus, paclitaxel, 5FUdR, and EGF all belong on the growing list of agents that are capable of inducing cytochrome c release from mitochondria.
Effects of Paclitaxel 5FUdR and EGF on Bcl-2 Family Members-Recent studies have suggested that the release of cytochrome c from mitochondria can be induced by Bax, a proapoptotic member of the Bcl-2 family (72,73). Previous studies also demonstrated that paclitaxel treatment is associated with the appearance of a slow-migrating hyperphosphorylated species of Bcl-2 (74) that has a lower affinity for Bax (75). One model suggests that the resulting increase in free Bax mediates the induction of apoptosis by paclitaxel, although this concept remains controversial (76,77). An alternative mechanism of triggering apoptosis involves induction of bax expression, e.g. after antibody ligation of the EGFR or inhibition of EGFR kinase activity in colon cancer cells (78). To assess the potential role of these types of alterations, MDA-MB-468 cells treated with paclitaxel, 5FUdR, or EGF were harvested at various times and reacted with anti-Bcl-2 antiserum. Results of these studies (Fig. 8A) readily confirmed the existence of a paclitaxelinduced alteration in Bcl-2 migration but failed to provide any evidence that 5FUdR or EGF induced a similar change. Additional experiments failed to provide any evidence that EGF caused alterations in levels of the antiapoptotic Bcl-2 family members Bcl-x L and Mcl-1 or the proapoptotic homologs Bad, Bax, or Bak (Fig. 8B). Accordingly, it does not appear that EGF-induced alterations in the expression of these polypeptides contribute to activation of the apoptotic process.
EGF Treatment Induces Anoikis-In a final series of experiments, we attempted to further evaluate the observation that EGF-treated cells displayed lower levels of active caspases (Fig.  4), less cleavage of caspase precursors (Fig. 3) and less degradation of caspase substrates (Fig. 2). Morphological examination revealed that 80 -90% of the cells detaching from the tissue culture plates after treatment with paclitaxel or 5FUdR had fragmented nuclei (Fig. 9, A and B). Even at paclitaxel concentrations as low as 5 nM, which caused Ͻ10% of the cells to detach from the plate, ϳ80% of the nonadherent cells were frankly apoptotic (data not shown). In contrast, at all time points up to 96 h after addition of EGF, 50 -70% of the nonadherent MDA-MB-468 cells appeared morphologically normal (Figs. 6F and 9C), although the number of floating cells progressively increased. When cell viability was analyzed by trypan blue exclusion, a similar distinction between drug-and EGF-induced apoptosis was observed. Forty-eight hours after initiation of paclitaxel or 5FUdR treatment, 47 Ϯ 8% (mean Ϯ S.D.; n ϭ 3) and 41 Ϯ 9% of the nonadherent cells were nonviable, respectively. In contrast, only 23 Ϯ 4% of the nonadherent cells stained with trypan blue 48 h after addition of EGF. Upon further incubation, however, the EGF-treated cells also died, as evidenced by the fact that 46 Ϯ 10% and 62 Ϯ 1% of the nonadherent cells took up trypan blue at 96 and 144 h, respectively. Coupled with the results of the caspase assays in Fig. 4, these observations raised the possibility that EGF might initially cause detachment of viable cells, which then become apoptotic due to the lack of interactions with the substratum. This model suggested three testable predictions: 1) MDA-MB-468 cells would become apoptotic if adhesion to a substratum were prevented by other means; 2) EGF might have a larger effect if the strength of the adhesion between MDA-MB-468 cells and the substratum were diminished; and 3) EGF-treated cells might survive if they could reattach. Experiments were performed to test each of these predictions.
First, semiconfluent MDA-MB-468 cells were trypsinized and seeded in fresh tissue culture plates or in test tubes to which they could not adhere. Over the course of 3 days, a substantial fraction of the cells cultured in test tubes became apoptotic (Fig. 9D), establishing that this cell line requires contact with its substratum for survival.
Next, MDA-MB-468 cells were cultured on plates coated with various concentrations of poly(HEMA), a hydrophobic polymer that diminishes cell-substrate adhesion (59). As the amount of poly(HEMA) coating was increased, EGF treatment caused more cell detachment (Fig. 9E), 2 establishing that cell-substratum interactions play an important role in the effect of EGF on these cells. Morphological examination of cells that detached from the poly(HEMA)-treated plates revealed that the percentage of apoptotic cells was the same in the absence and presence of EGF, suggesting that it was detachment rather than EGF treatment per se that was toxic.
Finally, MDA-MB-468 cells that detached after treatment with paclitaxel, 5FUdR, or EGF were sedimented, resuspended in drug-free medium, seeded on fresh tissue culture plates, and subsequently examined for viability (Fig. 9F). Fewer than 5% of the paclitaxel-or 5FUdR-treated cells attached and remained viable. In contrast, ϳ40% of the EGF-treated cells reattached and excluded trypan blue 72 h after EGF removal. DISCUSSION In the present study, we have compared the apoptotic pathways activated by paclitaxel, 5FUdR, or EGF in MDA-MB-468 breast cancer cells. Our results indicate that the vast majority of cells detaching from the tissue culture plates after treatment with paclitaxel or 5FUdR are frankly apoptotic, whereas 50 -70% of the cells becoming nonadherent after EGF treatment are morphologically normal (Fig. 9, A-D). Further experiments have demonstrated that the cells becoming nonadherent after EGF treatment can reattach and survive (Fig. 9F). In contrast, cells that detach after treatment with paclitaxel or 5FUdR cannot. Despite these differences, all of the treatments appear to activate the cytochrome c/Apaf-1 f caspase-3 pathway, as indicated by the release of cytochrome c to the cytosol (Fig. 7) and the appearance of active caspases-3, -6, and -7 (Figs. 2-5).
These observations have potentially important implications for the understanding of drug-and hormone-induced PCD in human breast cancer cells.
Recent studies (reviewed in Refs. 79 -82) have demonstrated that cell-substratum interactions can provide survival signals for many cell types. Conversely, removal of certain cell lines from their substrata can result in apoptosis. These recent descriptions of the process termed "anoikis" lead us to propose the following sequence of events following EGF treatment: 1) EGFtreated cells adhere less tightly to the tissue culture plates; 2) a fraction of cells detach from the substratum (Figs. 1E, 6E, and 9C) and lose integrin-mediated signaling; 3) those cells that do not reattach in a timely fashion initiate the process of anoikis, which includes release of cytochrome c to the cytosol (Fig. 7) and subsequent activation of downstream caspases (Figs. [2][3][4][5]. A number of observations support this model. First, MDA-MB-468 cells clearly undergo apoptosis when deliberately deprived of the opportunity to attach to a substratum (Fig. 9D). Second, treatment of the tissue culture plates with poly-(HEMA), which decreases the adhesion of the MDA-MB-468 Adherent cells were then harvested and subjected to SDS-PAGE followed by immunoblotting with anti-Bcl-2 antiserum. In addition to the usual species (a), a slower migrating species (b) appeared in paclitaxel-treated cells as described previously (74,75). In contrast, this species was not observed after 5FUdR or EGF treatment. B, after MDA-MB-468 cells were incubated with 100 ng/ml EGF for the indicated period of time, adherent cells were harvested, subjected to SDS-PAGE, and blotted with antisera that recognize the indicated Bcl-2 family member. To confirm equal loading, blots were reprobed with antiserum that recognizes the nucleolar protein B23 (not shown). Blots were derived from two experiments and were representative of three to five experiments examining each polypeptide. cells, 3 enhances the effect of EGF (Fig. 9E). Third, MDA-MB-468 cells that detach during EGF treatment can be rescued by replating in EGF-free medium (Fig. 9F). Fourth, EGF-treated cells show little increase in caspase activity until they have detached, whereas paclitaxel-or 5FUdR-treated cells contain elevated levels of DEVD-AFC and VEID-AFC cleavage activity before detachment (Fig. 4). Finally, when MDA-MB-468 cells are cultured in suspension as illustrated in Fig. 9D, EGF fails to exert any pro-or anti-apoptotic effect that can be distinguished from the effect of detachment alone. 3 Collectively, all of these observations support the view that EGF-induced apoptosis results from the loss of cell-substratum attachment. Although this process of anoikis has been implicated in a variety of physiological processes (79,80,82), the present report appears to be the first implicating anoikis in the antiproliferative effects of any hormone. Further studies are required to determine whether detachment-induced apoptosis plays a similar role in the antiproliferative effects of other hormones.
The present studies do not address the question of how EGF treatment causes detachment. EGF is known to enhance motility of various epithelial cells (83)(84)(85)(86), providing one potential explanation for the decreased adherence. A change in the affinity of integrins for their ligands (87), either in association with altered motility or independent of this process, might also contribute to the decreased adhesion. Further studies are required to examine these possibilities.
While the present studies were in progress, Chin et al. (5) reported that EGF treatment of serum-deprived MDA-MB-468 cells resulted in enhanced procaspase-1 expression. Based in part on these results, these investigators proposed that EGFinduced apoptosis was triggered by procaspase-1 autoactivation. Our results are not compatible with this model. First, when MDA-MB-468 cells were cultured in serum-containing medium, immunoblotting failed to provide evidence that EGF treatment enhanced procaspase-1 polypeptide levels (Fig. 3D,  lanes 1, 2, and 7) even though this treatment induced apoptosis (Figs. 6F and 9C). Second, assays for cleavage of the caspase-1 substrate YVAD-AFC failed to demonstrate an EGF-induced increase in the low level basal activity of this enzyme (Fig. 4C). Finally, an affinity labeling technique with nanogram sensitivity (24) failed to provide evidence for active caspase-1 in cytosol or nuclei of EGF-treated cells (Fig. 5A). Although these observations argue against a role for caspase-1 in EGF-induced apoptosis of the MDA-MB-468 cells, we nonetheless observed decreased levels of procaspase-1 concomitant with PCD in this cell line (Fig. 3D). Several potential explanations could account for these observations, including the possibility that caspase-1 is exported out of the MDA-MB-468 cells upon activation (88) or the possibility that procaspase-1 has been proteolytically cleaved at one or more sites that do not result in its catalytic activation.
In contrast to procaspase-1, cleavage of procaspases-3, -6, and -7 (Fig. 3, A-C) is clearly accompanied by their activation, as evidenced by the generation of protease species of appropriate molecular weight that react with the affinity label zEK-(bio)D-aomk (Fig. 5), appearance of activities that cleave DEVD-AFC and VEID-AFC (Fig. 4, A and B), and degradation of caspase substrates to fragments that are known to reflect activity of caspases-3, -6, and/or -7 in situ (Fig. 2). In short, four separate pieces of evidence (Figs. 2-5) support the view that EGF treatment is accompanied by activation of these downstream caspases. The cohort of effector caspases activated during EGF treatment is very similar to the cohort activated by paclitaxel or 5FUdR (Fig. 5), although the specific activities of these caspases in vitro (Fig. 4, A and B) and in situ (Fig. 2) are lower, presumably reflecting the lower percentage of cells that are undergoing apoptosis after EGF treatment (Figs. 6F and 9C). These observations not only provide additional support for previous claims that different stimuli activate the same downstream effectors of PCD (44,52,89,90) but also provide the first characterization of the caspases that are activated as cells undergo anoikis.
Additional experiments were designed to examine, in a preliminary fashion, the potential role of various pathways in the activation of these downstream caspases. Although a number of studies have provided evidence that the Fas/FasL pathway can participate in the induction of apoptosis by chemotherapeutic agents in some Fas-expressing cells (see the Introduction), this pathway does not appear to contribute to PCD in MDA-MB-468 cells. Treatment with a blocking anti-Fas antibody failed to attenuate the effects of 5FUdR, paclitaxel, and EGF in this cell line (Fig. 6, E and F). Furthermore, ligation of Fas by exogenous antibodies did not induce PCD in MDA-MB-468 cells (Fig. 6, B and E). Although these observations are difficult to reconcile with a model in which drug treatment induces PCD by activating the Fas/FasL pathway in this cell line, they do not rule out the possibility that another death receptor (e.g. death receptor 5 (91)) and its ligand might be involved in MDA-MB-468 cell death, nor do they rule out the possibility that the Fas/FasL pathway might play a role in chemotherapy-induced apoptosis in other cell lines. These issues require further investigation.
Recent studies have provided evidence that a pathway involving cytochrome c, Apaf-1, and procaspase-9 can also lead to activation of downstream caspases (32)(33)(34)(35)(36). To evaluate the possible activation of this pathway, the distribution of cytochrome c was examined by immunoblotting. Results of these experiments (Fig. 7) revealed that cytochrome c was readily detectable in cytosol of cells that became apoptotic after treatment with EGF as well as paclitaxel or 5FUdR (Fig. 7). Because of the lack of sensitivity of currently available anti-procaspase-9 antibodies, we were unable to directly confirm that procaspase-9 was cleaved in these cells. Nonetheless, the observations in Fig. 7 are highly suggestive that the cytochrome c/Apaf-1/procaspase-9 pathway contributes to activation of the downstream caspases after treatment with any of these agents.
Current evidence suggests that this pathway can be triggered by alterations in the expression or phosphorylation of certain Bcl-2 family members. In particular, increased levels of Bax (78,92,93), decreased levels of Bcl-2 (94), and altered phosphorylation of Bcl-2 (74,75,95) have been observed in various model systems prior to the onset of apoptosis. To search for similar changes in MDA-MB-468 cells, whole cell lysates were probed with antibodies that recognize six Bcl-2 family members. Consistent with previous reports (74,75,77,95), paclitaxel treatment caused a transient reduction in the mobility of part of the Bcl-2 molecules present in MDA-MB-468 cells (Fig. 8A). In contrast, changes in Bcl-2 phosphorylation were not evident after treatment with 5FUdR or EGF (Fig. 8A). Moreover, changes in phosphorylation state or levels of Bcl-x L , Mcl-1, Bak, Bax, or Bad were not discernible during treatment with EGF (Fig. 8B), suggesting that other alterations must trigger the apoptotic cascade.
In summary, the present studies have demonstrated that EGF treatment induces the process of anoikis in MDA-MB-468 cells. As is the case with drug-induced PCD, anoikis in these cells involves release of cytochrome c from mitochondria and activation of downstream caspases, including caspases-3, -6, and -7. The changes occurring between treatment with EGF and activation of downstream caspases do not appear to involve activation of the Fas/FasL pathway or alterations in levels of commonly studied Bcl-2 family members. Although the intervening steps remain unidentified at present, EGF-treated MDA-MB-468 cells might provide a useful model for studying the process by which loss of integrin-mediated signaling results in caspase activation and cell death.