Modulation of Caspase-Independent Cell Death Leads to Resensitization of Imatinib Mesylate–Resistant Cells

Chronic myelogenous leukemia (CML) is a clonal proliferative malignancy originating from a pluripotent hematopoietic stem cell. CML accounts for ∼20% of newly diagnosed cases of leukemia in adults. About 20% to 30% of patients with CML will die within 2 years of the diagnosis, and about 25% die each year after. All CMLs are triggered by the fusion gene product of the Philadelphia chromosome translocation t(9;22). Such a translocation leads to the generation of a p210 Bcr-Abl protein harboring constitutively active tyrosine kinase. Bcr-Abl protein is essential for the induction of in vitro cellular transformation (1) and in vivo leukemogenesis (2, 3). Imatinib mesylate (Gleevec, STI-571) functions through competitive inhibition of the ATP-binding site of the Bcr-Abl enzyme, leading to the inhibition of tyrosine phosphorylation of proteins involved in Bcr-Abl signaling (4). Molecular targeting of signal transduction molecules by imatinib has greatly improved the treatment of chronic phase and blast crisis in CML. Although imatinib is unquestionably effective in CML treatment, some patients ultimately relapse with resistant disease (5, 6). It is estimated that 20% to 25% of treated patients will develop some resistance (primary or secondary). Resistance may be through several mechanisms: about half of them are due to reactivation of Bcr-Abl kinase activity within the leukemic cells by either point mutation or gene amplification and the second half are resistant for reasons not directly related to Bcr-Abl (several mechanisms have been suggested but the nature of intrinsic resistance still needs clarification; refs. 7, 8).

It has been extensively reported that imatinib, by counteracting Bcr-Abl activity, will induce cell death in a caspase-dependent manner. However, it has recently been suggested that other death mechanisms could be engaged in response to imatinib treatment (9–12). Cell death has been classified in several categories. To date, at least three are known: apoptosis, necrosis, and the so-called caspase-independent cell death (CID). Apoptosis is by far the best-characterized type of cell death. It is defined by morphologic modifications (chromatin condensation, loss of mitochondrial membrane potential, plasma membrane asymmetry, overall cell shrinkage, blebbing of the plasma membrane, and detachment from the cellular matrix), all occurring before loss of plasma membrane integrity. Generally, those modifications are considered as hallmarks for executioner caspase activation (13). Apoptosis is characterized by the very rapid removal of the dying cell in vivo (14). In contrast, necrosis has long been considered as a passive mode of death without established regulatory mechanism. Classically, necrosis is defined by an early rupture of the plasma membrane and dilation of organelles (mainly mitochondria). Nevertheless, recent findings suggest that its occurrence and course might also be tightly regulated (15, 16).

A large number of studies have proposed the existence of a type of cell death that diverges from apoptosis or necrosis (17–19). In conditions where caspase activation is completely prevented, the process of cell death is delayed but rarely inhibited, leading to the occurrence of CID. This CID occurs physiologically when a signal that normally engages apoptosis fails to activate caspases. Different CID characteristics have been reviewed in refs. 20, 21. A CID-specific signature has yet to be found, but several morphologic modifications are observed: ragged plasma membrane, nucleus shrinkage and absence of apoptotic body, DNA fragmentation, and cellular blebbing.

We have recently identified the glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a specific inhibitor of CID but not of apoptosis (22). GAPDH overexpression was shown to rescue cells from CID stresses, allowing them to proliferate. Importantly, we established that cell survival and proliferation could only be observed in conditions where both apoptosis and CID (through GAPDH expression) were blocked. Until now, very little is known about the implication of this new type of cell death in CML. The need for additional cancer therapy due to treatment resistance prompted us to investigate the potential role of CID in the context of CML.

In this study, we established that apoptosis inhibition was not sufficient to prevent imatinib-induced cell death in Bcr-Abl–positive cell lines. We showed that CID was as efficient as apoptosis in killing CML cells and that GAPDH was able to protect cells from CID in this context. Interestingly, we also found that some imatinib-resistant CML cell lines spontaneously overexpressed GAPDH. Finally, we could establish that GAPDH knockdown was sufficient to resensitize those resistant cells to imatinib-induced death. Therefore, this study represents the first demonstration that CID modulation can sensitize resistant cancer cells to death and also that drugs able to modulate GAPDH, when used together with imatinib, should find some therapeutic benefit in CML patients.

Reagents. Imatinib was kindly provided by Novartis Pharma. Reagents were purchased from Sigma unless stated otherwise. Caspase inhibition was achieved by including 20 μmol/L of qVD-oph (MPbio). The caspase inhibitors were added 30 min before the apoptotic stimuli and replaced at 48-h intervals. Ac-DEVD-AMC, Ac-DEVD-CHO, and zVAD-fmk were from Alexis Biochemical. Mouse anti-V5 was from Invitrogen. Goat anti–heat shock protein 60 (Hsp60) was purchased from Santa Cruz Biotechnology. Rabbit anti-GAPDH was from Abcam. Antimouse-horseradish peroxidase (HRP), antigoat-HRP, and antirabbit-HRP were purchased from Dakopatts.

Cell cultures. Two human CML cell lines (Bcr-Abl–positive cells), K562 and JURL-MK1, were grown in RPMI 1640 (Life Technologies, Inc.) containing 5% FCS (Perbio), 50 units/mL penicillin, 50 μg/mL streptomycin, and 1 mmol/L sodium pyruvate. Imatinib-resistant cells were isolated from K562 cell lines (previously described in ref. 23). Mock or GAPDH-expressing K562 and JURL-MK1 cells were isolated after nucleofection (Amaxa) with either an empty pcDNA3.1 vector or a pcDNA3.1 encoding GAPDH-V5.

Cell death measurement. After the indicated treatment, cells were harvested and percentage of viability was measured by propidium iodide (PI) staining (0.5 μg/mL) and flow cytometry analysis in FL-3.

DEVDase activity measurement. After the indicated treatment, cells were lysed in buffer A (see “Western blot”). Lysates were standardized for protein content and loaded on a black 96-well plate (Cellstar) in the presence of 0.2 mmol/L caspase-3 substrate Ac-DEVD-AMC diluted in the following buffer: 50 mmol/L HEPES (pH 7.5), 150 mmol/L NaCl, 20 mmol/L EDTA, and 10 mmol/L DTT. Caspase activity was determined on a fluoroscan at 460 nm with or without 1 μmol/L Ac-DEVD-CHO, and specific activities were expressed as the change in absorbance per minute per milligram of protein.

Clonogenicity assay. Cells (3,000) were loaded in each well of the first column of a 96-well plate and then diluted by a factor of 2 by serial dilution. Cells were treated in the presence or absence of imatinib as indicated. When needed, 20 μmol/L of fresh qVD-oph were added to the cell culture every 48 h. Ten days later, the number of clones present in the last six wells of each lane was quantified.

The results represent the number of clones in the last six wells of one lane. Each treatment is done in duplicate in one experiment and results represent an average of three to four independent experiments.

Measurement of GAPDH activity. Cells were lysed in buffer A (see Western blot). Lysates were standardized for protein content and incubated with 0.25 mmol/L NAD, 3.3 mmol/L DTT, 13 mmol/L Na₄P₂O₇ (pH 8.5), 26 mmol/L sodium arsenate, and 25 mmol/L D-glyceraldehyde-3-phosphate on black 96-well plate (Cellstar). GAPDH activity is measured on a fluoroscan at 445 nm as the increase in fluorescence relating to NADH accumulation. Activity is expressed as the change in absorbance per milligram of protein.

Western blot. Briefly, after treatment, cells were collected, washed in PBS, and lysed for 10 min at 4°C in buffer A [10 mmol/L HEPES (pH 7.4), 150 mmol/L NaCl, 5 mmol/L EDTA, 1% NP40, 10 μg/mL aprotinin, 1 mmol/L phenylmethylsulfonyl fluoride (PMSF), 10 μmol/L leupeptin]. Lysates were cleared at 10,000 × g for 10 min, and proteins (50 μg) were separated on polyacrylamide gels (10%) and blotted on polyvinylidene difluoride membranes. After blocking nonspecific binding sites, the membrane was incubated for 2 h at room temperature with the appropriate primary antibody. The membranes were washed thrice with 50 mmol/L Tris, 150 mmol/L NaCl (pH 7.5), and 1% NP40 and further incubated with HRP-conjugated antibody for 1 h at room temperature. Immunoblots were revealed by autoradiography by using enhanced chemiluminescence detection kit (Pierce).

ATP measurement. On a 96-well plate, each condition was present in triplicate. Cells (5,000 per well) were incubated in 100 μL of medium. One hundred microliters of substrate solution (ATPLite 1 Step kit, Perkin-Elmer) were added and ATP content was analyzed using a luminometer following the manufacturer's instructions.

Vector construction. Three different short hairpin RNAs (shRNA), targeting GAPDH, were generated: sh546 (5′-GATCCCCGGTCATCCATGACAACTTTTTCAAGAGAAAAGTTGTCATGGATGACCTTTTTGGAAA-3′), sh675 (5′-GATCCCCCATCATCCCTGCCTCTACTTTCAAGAGAAGTAGAGGCAGGGATGATGTTTTTGGAAA-3′), and sh813 (5′-GATCCCCCCTGCCAAATATGATGACATTCAAGAGATGTCATCATATTTGGCAGGTTTTTGGAAA-3′). Complementary sense and antisense oligonucleotides (Eurogentec) were annealed and cloned into BglII/HindIII–cut pTER, generating sh546, sh675, and sh813 constructs. All constructs were confirmed by DNA sequencing analysis.

K562 cells and imatinib-resistant K562 cells (ImaR) stably expressing one of the shRNAs mentioned above or an empty pTER as control were obtained after Amaxa nucleofection and zeocin selection (200 μg/mL).

Apoptosome formation assay. The intrinsic potentials of K562 and ImaR to form apoptosome were examined in a cell-free system as previously described (24). Briefly, 50 × 10⁶ cells were collected and washed with 50 mL of ice-cold PBS, washed in 10 mL of hypotonic extraction buffer [50 mmol/L PIPES (pH 7.4), 50 mmol/L KCl, 5 mmol/L EGTA, 2 mmol/L MgCl₂, 1 mmol/L PMSF, 1 mmol/L DTT], and then lysed in 100 μL of hypotonic extraction buffer in a Dounce homogenizer. The lysate was cleared at 16,000 × g. One hundred micrograms of protein were incubated with 1 mmol/L dATP and 10 μmol/L cytochrome c for 30 min at 37°C. Then, DEVDase activity was measured as previously described.

Statistics. Statistics were done using the percentage comparison test or the Student's t test. All values shown in the text and figures are means ± SD.

Imatinib induces both apoptosis and CID of Bcr-Abl–positive cells. To confirm the importance of caspase dependency in imatinib-induced death of Bcr-Abl–positive cells, K562 and JURL-MK1 cells were incubated in the presence of imatinib for several days. Cell death was measured as PI⁺ cells by flow cytometry (Fig. 1A). In the same experiment, cells were pretreated or not with a large-spectrum caspase inhibitor (qVD-oph) and then treated as indicated. To minimize any cell type specificity, experiments were systematically done in the two different Bcr-Abl–positive cell lines. As expected, imatinib alone induced a massive (80–90% PI⁺ cells) and time-dependent death of both cell populations. We observed that JURL-MK1 cells were more susceptible than K562 cells to this stimulus. Interestingly, in the presence of imatinib and qVD-oph, a massive cell death was also observed. The death kinetic was slower, but after 5 days of imatinib treatment, the percentage of dead cells was found to be independent of the presence of a caspase inhibitor. The same results were observed in the presence of another caspase inhibitor, zVAD-fmk (data not shown). To verify that the cell death observed in the presence of imatinib and qVD-oph was indeed caspase independent and not due to some residual caspase activation, we measured caspase-3 activity in both cell lines (Fig. 1B). In the absence of qVD-oph, caspase activity was readily observed in those conditions but was undetectable in the presence of qVD-oph at any time point. Next, we determined whether CID was as efficient as apoptosis in preventing colony formation. To this end, cells were treated in the presence of different doses of imatinib ± qVD-oph. The ability of those cells to form colonies was analyzed 10 days after stimulation. As illustrated in Fig. 2, no clone could be obtained when cells were treated with imatinib alone. Interestingly, the addition of qVD-oph had no significant effect on imatinib efficacy. Therefore, we could conclude that imatinib-induced CID in CML cells represents a central mechanism controlling cell viability.

GAPDH can protect Bcr-Abl–positive cells from imatinib-induced CID but not from apoptosis. We recently described the ability of the glycolytic enzyme GAPDH to rescue cells from CID (22). Because imatinib was able to induce both apoptosis and CID, we wondered whether GAPDH could prevent imatinib-induced CID in Bcr-Abl–positive cells. For that purpose, we established two Bcr-Abl–positive cell lines stably overexpressing a V5-tagged form of GAPDH. Figure 3A (left) showed the expression of the V5-tagged GAPDH in K562 and JURL-MK1 cells. The right image illustrated the overexpression of the total amount of GAPDH present in those two cell lines compared with the mock-transfected K562 and JURL-MK1 cells. Using a colony formation assay (Fig. 3B), we evaluated the ability of those cells to be protected from CID. As expected, neither parental nor GAPDH-overexpressing cells can be protected from imatinib-induced apoptosis. Strikingly, GAPDH-overexpressing cells can form colonies at higher rates in the presence of qVD-oph, indicating that GAPDH can significantly protect against imatinib-induced CID in CML cell lines.

Imatinib-resistant cells are protected from CID via endogenous GAPDH overexpression. To further analyze the importance of CID in CML, we took advantage of ImaR cells isolated in our laboratory (23). Unfortunately, despite several attempts, no JURL-MK1–resistant cells could be isolated. ImaR cells present a resistance to high doses of imatinib (up to 10 μmol/L imatinib) as shown in Fig. 4A. We also verified that these cells can form colonies in the presence of 5 μmol/L imatinib and that the addition of qVD-oph did not significantly affect the number of clones that are able to grow (Fig. 4B). To understand why those cells are resistant to imatinib, we looked for described Bcr-Abl mutations or overexpression (7). We could not find any of those in our ImaR cells (data not shown; ref. 25). We reasoned that because those cells failed to die in the presence of imatinib, they should have acquired protection from both apoptosis and CID. Therefore, we first analyzed apoptosis. For that matter, we measured apoptosome-mediated caspase-3 activation in K562 and ImaR cells. As presented in Fig. 4C, we observed that ImaR cells present a 60% reduction of caspase activity as compared with sensitive K562 cells, suggesting that those cells are resistant to the intrinsic apoptotic pathway. CID resistance in ImaR cells was monitored by investigating GAPDH levels. ImaR cells exhibited a significant GAPDH overexpression as visualized by Western blot (Fig. 4D,, top) and quantified by measuring the GAPDH-specific activity (Fig. 4D , bottom). We found that resistant cells displayed ∼30% more active GAPDH than control cells.

Down-regulation of GAPDH expression levels resensitized ImaR cells to imatinib. Because ImaR cells overexpressed GAPDH, we hypothesized that reducing GAPDH level could be sufficient for resensitizing those cells to imatinib-induced cell death. For that purpose, we generated three different shRNAs specifically designed to stably knock down GAPDH. Figure 5A shows the level of GAPDH reduction obtained in ImaR cells expressing each of the three shRNAs. sh813 and sh546 gave a mild reduction (40%), whereas sh675 led to a larger GAPDH decrease (60%). Modulation of GAPDH levels was carefully quantified by measuring specific activity of this enzyme (Fig. 5B). This quantification correlated very closely with the level of protein observed in Fig. 5A. Because GAPDH is a key enzyme of the glycolytic pathway, we verified that its reduction had no significant effect on the overall energetic status of those cells. As presented in Fig. 5C, no major differences in ATP level could be measured comparing K562, ImaR-pTER, and ImaR cells expressing one of the GAPDH shRNAs.

To further characterize GAPDH knockdown effect on cell viability and response to imatinib treatment, we generated sensitive K562 cells expressing different GAPDH shRNAs as controls. The observed GAPDH extinction was equivalent to the one obtained for ImaR cells (Supplementary Fig. S1). Figure 6A indicated that sensitive K562 cells expressing shRNA 546 (K562-sh546) or shRNA 675 (K562-sh675) were not distinguishable from K562-pTER cells in the presence and absence of treatment, altogether indicating that a partial knockdown of GAPDH in sensitive K562 has no significant effect on their viability and their response to treatment.

As previously noted, reducing GAPDH levels had no significant effect on ImaR cells in the absence of treatment (Fig. 6B). Because sh813 and sh546 gave a similar GAPDH reduction in ImaR cells, we chose to further characterize only two of the three knockdown cell lines generated (see Fig. 5A and B). Strikingly, the viability of ImaR cells expressing GAPDH shRNA 546 (ImaR-sh546) or shRNA 675 (ImaR-sh675) was profoundly affected by imatinib treatment as compared with ImaR-pTER cells. A net increase in cell death could be measured by flow cytometry as PI⁺ cells in both cell lines (Fig. 6B,, left). Then, we verified the effect of GAPDH knockdown on the ability of those cells to form colonies (Fig. 6B,, right). As expected, no difference was observed among cells in the absence of treatment. Importantly, as opposed to ImaR-pTER cells that could easily grow as colonies, neither K562 nor ImaR cells expressing sh546 or sh675 could form a significant number of colonies in the presence of imatinib independently of the presence of qVD-oph (Fig. 6A  and B, right).

Altogether, we could establish that reduction of GAPDH level was sufficient to resensitize ImaR cells to imatinib-induced death.

In the present study, we established the importance of CID in the context of CML. We showed that apoptosis inhibition (using broad caspase inhibitors) failed to prevent imatinib-induced cell death of Bcr-Abl–positive cells (Fig. 1). In fact, on apoptosis inhibition, CID was acting as a death backup mechanism to kill the leukemic cells. Therefore, we showed that apoptosis and CID occurred in parallel in response to imatinib, albeit with different kinetics (apoptosis occurring faster than CID). We used PI staining as a robust way to measure cell death. One could argue that a PI⁻ cell observed was in fact a dying cell that did not lose its integrity at the time of the analysis. Therefore, we addressed this question using a very strict test: the capacity of those cells to grow as clones. As presented in Fig. 2, we established for the first time that the ability of imatinib to prevent colony formation was mediated by both apoptosis and CID. Indeed, we could show that even if CID is slower than apoptosis, it is as efficient in preventing any colony formation.

It seems that CID can be observed only in conditions where caspase-mediated apoptosis is inhibited. Apoptosis inhibition has been shown to occur in cancer cells by various mechanisms. These include several tumor lines that lose Apaf-1 expression (26, 27) and can have mutation in executioner caspases. Similarly, many tumors overexpressed inhibitor of apoptosis (IAP) proteins such as XIAP (28). In this line, expression of one member of this IAP family (survivin) has been linked to CML progression and resistance to apoptosis (29), although it remains unclear whether survivin is very effective as a caspase inhibitor. It has also been reported that Bcr-Abl is able to prevent apoptosis downstream of mitochondrial cytochrome c release by perturbing caspase-9 recruitment to Apaf-1 (30) mainly through modulation of Hsp90β (31). Finally, apoptosome-induced caspase-3 activation was found to be deficient in several patient samples presenting a resistance to imatinib treatment (12).

CID occurrence in CML cells has previously been suggested (9–12), but the exact role of this type of cell death and the importance of its modulation were unknown. To elucidate the implication of imatinib-induced CID in Bcr-Abl–positive cells, we took advantage of the ability of GAPDH to specifically inhibit CID without affecting apoptosis (22). For that matter, we generated GAPDH-overexpressing K562 and JURL-MK1 cells and showed that those cells became protected from CID (imatinib + qVD-oph) but not from apoptosis (imatinib alone; Fig. 3). This is of particular interest because glycolytic enzymes and particularly GAPDH were shown to be overexpressed in a wide variety of cancers (32). It is known since Warburg's work in 1929 that cancer cells very frequently up-regulated glucose metabolism leading to a high uptake and use of glucose but moderate rates of mitochondrial respiration under aerobic conditions. Frequently, up-regulation of glycolytic metabolism has been shown to correlate with increased tumor aggressiveness and poor patient prognosis in several cancers (33, 34). The exact reason of glycolytic enzyme overexpression remains unclear but we could postulate that, among other reasons, GAPDH was overexpressed to protect cells from CID.

Imatinib discovery represents an important advance in the management of CML and has led the way for targeted therapy of cancer as a general “proof of concept.” Despite the achievable remission rates, resistance to imatinib is an important issue for the therapy because a minority of CML patient in chronic phase and a substantial proportion in advanced disease phases either display refractoriness to imatinib treatment or lose imatinib sensitivity over time and experience relapse (7). To investigate CID importance in the context of resistance to imatinib, we took advantage of ImaR cells (23, 25). We showed that those cells were resistant not only to imatinib-induced apoptosis as expected but also to imatinib-induced CID. We could establish that those cells are presenting a resistance to imatinib-induced apoptosis through an inhibition of the apoptosome-induced caspase-3 activation (Fig. 4C). The exact reason of this resistance would need further studies but did not seem to be related to the decrease of apoptosome component factors (data not shown).

Because imatinib is inducing apoptosis and CID, we therefore investigated how those cells can escape CID. For that reason, we monitored GAPDH levels in those ImaR cells and established that they are expressing 25% to 30% more GAPDH enzyme than control cells (Fig. 4D).

To further investigate the role of GAPDH in controlling imatinib-induced CID in ImaR cells, we specifically knocked down GAPDH in those cells using several shRNAs (Fig. 5A and B). We carefully verified that GAPDH reduction failed to affect the energy status (i.e., ATP content) of the cell (Fig. 5C). Because GAPDH is not a limiting enzyme of the glycolytic pathway, its partial reduction in cells did not result in a major ATP depletion (Fig. 5C) and did not modify the ability of those cells to form colony in the absence of treatment (Fig. 6). We also verified that knocking down GAPDH to the same extent in sensitive K562 cells did not alter their ability to grow and to react to imatinib-induced death. Finally, we showed that lowering GAPDH below a certain level was enough to resensitize ImaR cells to imatinib-induced death (Fig. 6B).

Interestingly, recent clinical reports have indicated that changes in glucose metabolism might be related to and/or predictive for the development of imatinib resistance (35). Additionally, clinical studies of patients with imatinib-resistant c-KIT–positive tumors have shown avid glucose uptake on positron emission tomography scans (36). Finally, others also observed GAPDH overexpression at the protein level in imatinib-resistant cells (37).

In conclusion, our work describes the first evidence asserting the importance of CID in Bcr-Abl–positive cells. It seems clear that combination, rather than sequential therapy, is an attractive option to prevent the expansion of mutant subclones that confer resistance to therapy, which may allow more effective elimination of residual leukemic cells. There is a growing consensus that inhibition of individual targets is unlikely to succeed as a therapeutic strategy. Indeed, even in CML, it has been argued that the efficacy of imatinib treatment requires inhibition of targets other than Bcr-Abl (38). Therefore, regulating CID (via GAPDH knockdown or inhibition) in addition to classic treatment could constitute a new approach for sensitizing imatinib-resistant cells to death.

No potential conflicts of interest were disclosed.

Grant support: Association pour la Recherche sur le Cancer grant. L.A. Pradelli received a fellowship from the Région Provence-Alpes-Cote-d'Azur and Institut National de la Sante et de la Recherche Medicale (INSERM). M. Bénéteau received a fellowship from La Fondation de France. J-E. Ricci is a recipient of a contrat d'interface INSERM-CHU de Nice.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Sandrine Marchetti, Dr. Pascal Colosetti, Sebastien Grosso, and Dr. Bernard Mari for invaluable help and discussion; Dr. Hans Clever for the kind gift of the pTER vector; and Novartis for the gift of imatinib mesylate.