Autophagy induction by Bcr-Abl-expressing cells facilitates their recovery from a targeted or nontargeted treatment


  • Lisa C. Crowley,

    1. Leslie C. Quick Laboratory, Cork Cancer Research Centre, BioSciences Institute, University College Cork and Mercy University Hospital, Grenville Place, Cork, Ireland
    Search for more papers by this author
  • Baukje M. Elzinga,

    1. Leslie C. Quick Laboratory, Cork Cancer Research Centre, BioSciences Institute, University College Cork and Mercy University Hospital, Grenville Place, Cork, Ireland
    Search for more papers by this author
  • Gerald C. O'Sullivan,

    1. Leslie C. Quick Laboratory, Cork Cancer Research Centre, BioSciences Institute, University College Cork and Mercy University Hospital, Grenville Place, Cork, Ireland
    Search for more papers by this author
  • Sharon L. McKenna

    Corresponding author
    1. Leslie C. Quick Laboratory, Cork Cancer Research Centre, BioSciences Institute, University College Cork and Mercy University Hospital, Grenville Place, Cork, Ireland
    • Leslie C. Quick Laboratory, Cork Cancer Research Centre, BioSciences Institute, University College Cork, Cork, Ireland
    Search for more papers by this author

  • Conflict of interest: Nothing to report


Although Imatinib has transformed the treatment of chronic myeloid leukemia (CML), it is not curative due to the persistence of resistant cells that can regenerate the disease. We have examined how Bcr-Abl-expressing cells respond to two mechanistically different therapeutic agents, etoposide and Imatinib. We also examined Bcr-Abl expression at low and high levels as elevated expression has been associated with treatment failure. Cells expressing low levels of Bcr-Abl undergo apoptosis in response to the DNA-targeting agent (etoposide), whereas high-Bcr-Abl-expressing cells primarily induce autophagy. Autophagic populations engage a delayed nonapoptotic death; however, sufficient cells evade this and repopulate following the withdrawal of the drug. Non-Bcr-Abl-expressing 32D or Ba/F3 cells induce both apoptosis and autophagy in response to etoposide and can recover. Imatinib treatment induces both apoptosis and autophagy in all Bcr-Abl-expressing cells and populations rapidly recover. Inhibition of autophagy with ATG7 and Beclin1 siRNA significantly reduced the recovery of Imatinib-treated K562 cells, indicating the importance of autophagy for the recovery of treated cells. Combination regimes incorporating agents that disrupt Imatinib-induced autophagy would remain primarily targeted and may improve response to the treatment in CML. Am. J. Hematol., 2011. © 2010 Wiley-Liss, Inc.


Chronic myeloid leukemia (CML) is a progressive disease requiring lifelong treatment and is rarely cured. CML patients usually present in the chronic phase, which is a relatively manageable disease. However, without effective treatment, chronic phase progresses to a more aggressive and drug-resistant phase known as blast crisis [1].

Imatinib is now a first-line treatment for CML and specifically targets the tyrosine kinase activity of Bcr-Abl. This fusion protein is produced by the Philadelphia chromosome and is regarded as the hallmark of CML. Imatinib has dramatically improved the prognosis for chronic-phase patients [2]. However, as with any single-agent therapy, chronic-phase patients can develop resistance, which is primarily due to the amplification or mutation of Bcr-Abl [3]. For patients with disease in advanced stage (blast crisis), responses to Imatinib are short-lived, and drug resistance and relapse occur in almost all patients [4]. This has been attributed to both amplification of Bcr-Abl and the acquisition of Bcr-Abl-independent mechanisms [5, 6].

Other challenges in treatment include the need to give inhibitory drugs continuously. Following Imatinib withdrawal, the majority of cases will relapse. Even patients who achieve molecular remission have been reported to relapse when the drug is withdrawn [7]. Imatinib can therefore control the disease, but sufficient cells remain that are capable of regenerating the leukemia. Recent studies have suggested that these are Imatinib-resistant CML stem cells [8], and their resistance has been attributed to elevated expression of Bcr-Abl [9], their quiescent state [10], and the protective effects of the stem-cell niche [11]. Studies are on-going into alternative inhibitors and potential new combinatory regimes to more effectively target these cells [12]. Undoubtedly, an improved understanding of survival mechanisms operating in Bcr-Abl-expressing cells could improve regime design.

A number of studies have now reported that treatment with cytotoxic agents can induce an autophagic response [13–17]. Imatinib-treated CML cells have been reported to undergo apoptosis [18, 19] and more recently—autophagy [20, 21]. Autophagy is a highly conserved catabolic process in which cells self-digest organelles and other macromolecules by forming a double-membraned autophagosome around sequestered cytoplasmic material. Fusion with a lysosome generates an autolysosome, which degrades and recycles cytoplasmic material (reviewed in Ref. [22]). One paradox in the literature is that autophagy is a survival mechanism induced by growth limiting conditions but may also lead to cell death [23], which has recently been reported to occur in developmental processes [24–26].

The role of autophagy in the response of CML cells to therapy requires further clarification in order to advance the design of curative regimes. It remains unclear whether autophagy is solely a consequence of Imatinib treatment or if CML cells will respond to other agents in this way. Several studies have demonstrated the antiapoptotic properties of the Bcr-Abl protein, particularly, at high-expression levels, and the impact of this on resistance to various chemotherapeutic drugs [27–30]. Autophagy is an alternative response that may be engaged when cells are apoptosis incompetent or resistant [31].

In this study, we investigated the response of Bcr-Abl-expressing cells to two classes of agent; a DNA-damaging agent (etoposide/VP16) and a tyrosine kinase inhibitor (TKI; Imatinib). We have also examined models with Bcr-Abl expression at low and high levels as elevated expression has been associated with treatment failure. We have found with both classes of drug that autophagy can be engaged in a subset of viable cells following chemotherapy. With VP16, this will happen in cells expressing high levels of Bcr-Abl [or in low-Bcr-Abl-expressing cells, supplied with interleukin-3 (IL-3)] but also “normal/untransformed” non-Bcr-Abl-expressing cells. Where autophagy is observed, it is associated with a gradual recovery of the leukemic population. This type of response ultimately limits treatment success. The contribution of autophagy to recovery following Imatinib treatment has been demonstrated by siRNA knockdown of autophagy regulators (Beclin1 and ATG7). This reduced the recovery of treated populations and improved the cytotoxic effect.

Materials and Methods


Mouse myeloid progenitor 32D cells were maintained in RPMI 1640 medium, HEPES modification, with 10% FCS, 1% penicillin/streptomycin, 2 mmol/l L-glutamine (Sigma-Aldrich, Ireland), and 10% WEHI conditioned media as a source of IL-3. C2 (clone2) and C4 (clone4) BCR-ABL-transformed clones of 32D cells [5] were grown under the same conditions minus WEHI media. TonB.210 cell line (a generous gift from Prof. Daly) is a stably transfected Ba/F3 hematopoietic cell line containing an inducible 210 kDa BCR-ABL controlled by doxocyclin. Uninduced TonB.210 cells were maintained in the same media as 32D cells. The following Bcr-Abl-expressing cells were maintained under the same conditions minus WEHI media; K562, a human erytholeukemia cell line, and TonB.210 cells induced for 48 hr (TonB0.2μgDox and TonB2.0μgDox). All cells were cultured at 37°C under humidified atmosphere containing 5% CO2 (C2 and TonB0.2μgDox—have low-Bcr-Abl expression and C4 and TonB2.0μgDox have higher Bcr-Abl expression).

Western blotting

Lysates were prepared using RIPA buffer (50 mM Tris–HCL pH 7.4, 1% NP-40, 150 mM NaCl, 1 mM EDTA, 0.25% Na-deoxycholate, 1 mM PMSF, 1 mM Na3VO4, 1 mM NaF, and Protease Inhibitor Cocktail (Sigma-Aldrich, Ireland). Protein concentrations were determined by Bradford reagent (BioRad Laboratories, UK), and lysates were resolved on SDS–PAGE gels or precast 4–12% gradient gels (Invitrogen, Ireland). Proteins were transferred to nitrocellulose or polyvinylidene difluoride membrane by wet transfer or iBlot system (Invitrogen, Ireland). The anti-cAbl antibody (Santa Cruz), antiphosphtyrosine (Calbiochem), anti-LC3 (MBL), anti-Beclin1, and anti-ATG7 (Cell Signaling) were used at 1:1,000 dilution. Anti-β-actin (Millipore) was used at 1:5,000. Membranes incubated with appropriate HRP-conjugated secondary antibodies (DakoCytomation, UK) were detected via enhanced chemiluminescence (Amersham Biosciences) detection system. Membranes incubated with appropriate IRDye (700DX/800CW) conjugated secondary were detected via LI-COR Odyssey system (LI-COR Biosciences).

Cell-viability analysis and long-term liquid culture assay

Cells were seeded at 2 × 105 cells/ml in a T25 flask 24 hr before treatment. TonB.210 cell lines were treated with 1.0 μg/ml of VP16 (Etoposide; Sigma-Aldrich, Ireland) for 16 hr. All 32D and K562 cells were treated with 10 μg/ml of VP16 for 24 hr. Imatinib (a gift from Novartis, Ireland) treatment was as follows; C2/C4 cells were treated with 1 μM, and TonB0.2μgDox/TonB.2.0μgDox and K562 were treated with 5 μM Imatinib for 24 hr. Cells were collected in PBS and 50 μg/ml of propidium iodide (PI) solution (Sigma-Aldrich, Ireland) and added immediately before testing. Viability was measured using the dye exclusion method using FACSCalibur® flow cytometer (Becton Dickinson, Mountain View, CA). Data were analyzed using Graphpad 4.0 software. Prolonged Viablity and recovery analysis: following initial 24-hr treatment, drug was removed and cells were resuspended in fresh media. Media were removed and replaced when appropriate until population had fully recovered.

Cell morphology and scoring of apoptosis

Samples were transferred to glass slides using the Cytospin System (Shandon Technology-Thermo Scientific) and stained using Pro-Diff Stain (Braidwoods laboratories, UK). Apoptosis was defined as any cell, which appears shrunken with condensed chromatin or has clear nuclear fragmentation. A minimum of a hundred cells from three independent experiments were scored and graphed using Graphpad 4.0 analysis software.


Cytospins were fixed in 4% PFA for 20 min and washed with PBS. Permeabilization and blocking were carried out with 0.005% saponoin/PBS and 0.2% BSA/PBS, respectively. Antibody staining; anti-Atg8b (LC3) (Abgent, Cambridge Biosciences, UK), Beclin1, and ATG7 (Cell Signaling Technology, UK) were incubated (1:200) for 1 hr at room temperature. After incubation with the appropriate secondary, Alexa Fluor 488 or Alexa Fluor 594 samples were cover-sliped with “Gold plus with Dapi Anti-fade” solution (Molecular Probes, Invitrogen, Ireland).

siRNA transfection

K562 cells were transfected using Hiperfect (Qiagen, Ireland) with 0.85 μg Beclin1 siRNA (BECN1 ON TARGETplus SMARTpool siRNA) and 0.85 μg ATG7 siRNA (ATG7 ON TARGETplus SMARTpool siRNA) (Dharmcaon,Thermo Scientific, UK) 24 hr before treatment. Both immunofluorescence and Western blot were used to analyze knock-down efficiency.

Vesicular visualization and imaging

LC3-GFP plasmid was transfected into the appropriate cells using Turbofect transfection reagent (Fermentas, Ireland). Cytospun samples were fixed using 4% PFA for 20 min. Cells were incubated with monodansylcadaverine (MDC) 50 μM for 15–30 min with or without 50 μg/ml PI and washed with PBS. All slides were viewed using a DP70 digital microscope camera; images were captured with Olympus DP-Soft823 version 3.2 acquisition software. Images were processed using Photoshop CS2 software.


Cell-line models with low and high expression of Bcr-Abl

We and others have previously reported that cellular responses to chemotherapeutic agents can differ in CML cells expressing low and high levels of Bcr-Abl [5, 32]. We therefore used cell-line models that have low and high-Bcr-Abl expression. The first model is an IL-3-dependent 32D mouse myeloid progenitor cell transfected with Bcr-Abl. Cells expressing Bcr-Abl are IL-3 independent and cause a CML like disease in mice [33]. Clones were selected, with high (C4 clone) and low (C2 clone) expression of Bcr-Abl. C4 cells are more resistant to DNA-damaging cytotoxic drugs, whereas C2 cells have similar drug sensitivity to the parental 32D cells [5].

The second model is the TonB.210 cell line, which expresses Bcr-Abl in a doxocyclin-dependent manner (Fig. 1A) [34] and can survive without IL-3 at minimum of 0.1 μg doxocyclin (Fig. 1B). Subsequent experiments on drug sensitivity were carried out at 0.2 μg (to ensure viability was not compromised by absence of IL-3). In TonB.210 cells, a higher level of Bcr-Abl was induced by the addition of 2.0 μg doxocyclin, which results in elevated levels of phosphotyrosine (Supporting Information Fig. S1A). In subsequent experiments with TonB.210 cells, Bcr-Abl is induced 48 hr before the removal of IL-3 and addition of treatment agent.

Figure 1.

Analysis of Bcr-Abl expression in cell-line models and impact of a DNA damaging agent (Etoposide/VP16) on cell death morphology. A(i): Increasing concentrations of doxocyclin (0.1–2.0 μg) resulted in increased expression of the Bcr-Abl protein. A(ii): Western blot of Bcr-Abl expression in the 32D parental cells and the C2 and C4 subclones. Bcr-Abl expression is absent in 32D, low in C2, and high in C4. A(i),(ii): Forty micrograms of whole-cell lysates were loaded. Bcr-Abl was detected using the c-Abl antibody, HRP-conjugated secondary antibody, and enhanced chemiluminescence (ECL). β-Actin was used as a loading control. UID, uninduced. B: Viability of TonB.210 by FACS analysis of PI exclusion (percentage PI negative). TonB.210 cells were cultured for 48 hr in the presence of doxocyclin (0.05–2.0 μg) before IL-3 withdrawal. IL-3 independence is induced by expression of Bcr-Abl. C: Morphology of cell lines 32D [C(i)] TonB.210 [C(ii)] and K562 [C(iii)] following 24 hr/16 hr VP16 treatment; [C(i)] 32D, C2 and C4 cells (32D cell line) treated with 10 μg/ml VP16 for 24 hr. C(ii): TonB uninduced, TonB0.2μgDox and TonB2.0μgDox (TonB cell line) treated with 1 μg/ml VP16 for 16 hr. C(iii): K562 cells treated with 10 μg/ml VP16 for 24 hr. C(i–iii): Parental (32D and TonB.210) and low expressors (C2 and TonB0.2μgDox) show apoptotic morphology (black arrows). In high Bcr-Abl-expressing cells (C4 cells and TonB2.0μgDox), apoptotic morphology is absent, and cells show cytoplasmic vacuolization (red arrows) and intact nuclei. The K562 cells C(iii), which have endogenous Bcr-Abl expression, show no evidence of apoptosis at 24 hr. Images; detailed view shows an enlarged section indicated on the “VP16 treated” 40× picture. Images were taken with a 40× optical lens of a DP70 digital microscope camera, and Olympus DP-Soft823 version 3.2 acquisition software was used to capture images, and these were processed using Adobe Photoshop CS2.

Impact of a DNA-damaging agent (etoposide/VP16) on cell death morphologies

Optimal drug concentrations for evaluation of death responses were determined in each cell line. Cellular responses were evident in parental 32D cells and K562 cells treated with 10 μg/ml VP16 for 24 hr. A lower concentration of drug was required to induce a similar level of response in parental TonB.210 cells—1 μg/ml for 16 hr was sufficient to cause a reduction in viability to 59.0% ± 2.9%. Viability and apoptosis assays for all cell lines have been described before at limited time points (18–24 hr) and are not repeated here [35]. Morphology following VP16 treatment of 32D, TonB.210, and K562 cell lines is shown in Fig. 1C(i),(ii),(iii), respectively. These results clearly demonstrate the presence of apoptosis in non-Bcr-Abl-expressing cells [top panels (1C(i,ii)] and low-Bcr-Abl expressers (middle panel). In contrast, in higher Bcr-Abl-expressing cells [C4 and TonB2.0μg; bottom panels C(i,ii)], apoptosis is either absent (C4) or present at much lower levels (TonB2.0μg). Both cell lines do however exhibit increased cytoplasmic vacuolization characteristic of autophagy. K562 cells, as previously reported, do not exhibit any apoptosis nor do we see prominent cytoplasmic changes at this time point [Fig. 1C(iii)].

Cell recovery after VP16/etoposide treatment

Although the consequences of apoptosis for cell viability are clear, other cellular responses require long-term growth and viability assays to assess delayed responses and establish if recovery of the leukemic population is possible. Cell viability was assessed for up to 18 days following an initial 24-hr drug treatment. This was measured using the dye exclusion method and therefore encompasses all cell death mechanisms [Fig. 2A(i),(ii),(iii)].

Figure 2.

Analysis of cell recovery after treatment with VP16/etoposide. A: Graphical representation of percentage PI negative (viable) cells. A(i): 32D, C2 and C4 (32D cell line) treated with 10 μg/ml VP16 for 24 hr. A(ii): TonB uninduced, TonB0.2μgDox and TonB2.0μgDox (TonB cell line) treated with 1 μg/ml VP16 for 16 hr. A(iii): K562 cells treated with 10 μg/ml VP16 for 24 hr. Note: The parental 32D and TonB.uninduced were grown with IL-3 [graphs A(i) and A(ii)]. B: 40× morphology images at 48 hr (recovery) following VP16 removal; (i) 32D and TonB (non-Bcr-Abl-expressing cells). Very low numbers of intact TonB cells are present at 48-hr post-treatment, and these are surrounded by cell debris. B(ii): C4, TonB2.0μgDox and K562 cells (high-Bcr-Abl-expressing cells). Detailed view shows an enlarged section indicated by the small box on the 40× view. Examples of apoptosis are represented by black arrows, red arrows indicate vacuoles. Note: This highlights the cytoplasmic vacuoles and intact nuclei of cells during recovery. C: IL-3 rescue of low-Bcr-Abl-expressing cells; (i) graphical representation of percentage PI negative (viable) of C2 and TonB0.2μgDox treated with and recovered from 10 μg/ml and 1 μg/ml VP16 treatment, respectively, for 24 hr in the presence of IL-3. C(ii): 40× images and detailed view of C2 and TonB0.2μgDox 48 hr (recovery) after removal of VP16. Note: With the addition of IL-3, low-Bcr-Abl-expressing cells now show recovery 9 days post-treatment. This is delayed when compared with C4 or TonB2.0μgDox cells but cultures will repopulate. Graphs were prepared using Graphpad Software4.0. Bars indicate mean values ± standard deviation. Morphology; cells were viewed under the 40× optical lens of a DP70 digital microscope camera; pictures were captured with Olympus DP-Soft823 version 3.2 acquisition software. Pictures were then processed using Adobe Photoshop CS2.

For non-Bcr-Abl expressers (32D and TonB uninduced) and low-Bcr-Abl expressers (C2 and TonB0.2μgDox), PI positivity/uptake at 24 hr VP16/etoposide treatment was consistent with apoptotic morphology [Fig. 1C(i),(ii)]. In high-Bcr-Abl-expressing cells (C4 and TonB2.0μgDox), viability was 55.1% ± 2.6% and 43.7% ± 6.3%, respectively, following a 24/16-hr treatment. However, the absence of apoptotic morphology was not consistent with the levels of death indicated by PI uptake. The human cell line K562, previously reported as resistant, had almost no loss in viability after 24 hr of treatment (Fig. 2A).

Following 16/24-hr treatment, all cells were placed in fresh media in the absence of the drug. Viability was assessed after a recovery period of 48 hr, the media was then replaced to ensure minimal toxicity due a dying population and subsequently reassessed after a further 96 hr in the case of K562 cells (referred to as 5 days post-treatment) and 7 days post-treatment in 32D and TonB cell lines. Following this time point, cultures were assessed up to 18 days post-treatment (Fig. 2A).

Low-level Bcr-Abl expressers [C2 and TonB0.2μgDox (which exhibit apoptosis)] did not recover; 48 hr after drug removal, no viable cells were detected in C2 cultures, while in TonB0.2μgDox, no viable cells remained at 7 days post-treatment. The higher Bcr-Abl expressers (C4 and TonB2.0μgDox) recover within 1 week of drug removal. Surprisingly, non-Bcr-Abl expressers (“normal/untransformed”) cells also recover, repopulating the culture and regaining a normal morphology [Fig. 2A(i),(ii)]. Evidently, these populations have subset of cells that are not committed to apoptosis. These cells are IL-3 dependent and unlike Bcr-Abl-expressing cells had IL-3 present. The viability of K562 cells decreased to 55.7% ± 0.2% at 48 hr following drug removal indicating a delayed but significant response. Loss in viability continued for up to 9 days post-treatment. Following 18 days post-treatment, K562 cells show remarkable recovery, returning from viability lows of just 3% [Fig. 2A(iii)].

The morphology of all cells was examined during recovery [Fig. 2B(i),(ii)]. Cells appeared misshapen and swollen with numerous cytoplasmic vacuoles of differing sizes and sometimes had intact but distorted nuclei.

We were intrigued by the fact that non-Bcr-Abl-expressing cells, provided with IL-3, can recover from cytotoxic drugs, whereas, low-level Bcr-Abl-expressing cells, without IL-3, did not. Although IL-3 independence is conferred by low-Bcr-Abl expression, this is insufficient to rescue cells from drug-induced damage. Low and high-Bcr-Abl expressers were subsequently treated with VP16 in the presence of IL-3. High-Bcr-Abl expressers treated in the presence of IL-3 showed no additive effect (data not shown). In contrast, cells expressing low levels of Bcr-Abl, treated for 24 hr in the presence of IL-3 with either 10 μg/ml (C2) or 1 μg/ml VP16 (TonB0.2μgDox), are able to recover [Fig. 2C(i)]. Morphology during recovery was very similar to the recovering populations described earlier [Fig. 2C(ii)].

Recently, the involvement of autophagy in response to DNA damaging therapeutics has been reported in cell lines derived from solid tumors. Its role in the response of hematopoietic cells to DNA damaging drugs has not been evaluated. We therefore examined whether the vesicles apparent in drug-treated cells were autophagosomes. Upon induction of autophagy, the LC3 protein becomes lipidated and sequestered onto autophagic vesicles [36]. The presence of this autophagic marker was assessed by Western blot (to examine both isoforms) and immunofluorescence (to assess up regulation and punctate staining in treated cells).

Western blot analysis (Fig. 3A) indicated that all cells possess a basal level of endogenous LC3I and LC3II. In C4 high-Bcr-Abl expressers, this basal level was slightly higher compared to 32D and C2 cells. K562 cells also had high-basal levels of LC3II. The amount of LC3II relative to LC3I increases upon treatment of high-Bcr-Abl-expressing cells with VP16. Low-Bcr-Abl expressers (C2 and TonB0.2μgDox) have lower levels of expression of both LC3 isoforms. Parental 32D cells also show LC3I depletion and slight elevation of LC3II in treated cells. This isoform is also present in treated TonB cells although expression levels are lower.

Figure 3.

Evaluation of autophagic markers following VP16 treatment. A: Endogenous LC3 levels in the 32D cell line (32D,C2,C4) treated with 10 μg/ml VP16 for 24 hr and TonB cell line (TonB uninduced, TonB0.2μgDox and TonB2.0μgDox) treated with 1 μg/ml VP16 for 16 hr. K562 cells were treated with 10 μg/ml VP16 for 24 hr. LC3 levels in K562 cells were analyzed at both 24-hr treatment and 24 hr following removal of drug. Eighty micrograms of total cell lysates were analyzed by Western blot. Blots were incubated with anti-LC3 and detected using IRDye secondary antibody and viewed with the LI-COR Odyssey system. B: Immunofluorescence staining of endogenous LC3 was performed on treated and untreated 32D, C4, and TonB2.0μgDox cells. C: Immunofluorescence staining of endogenous LC3 and Beclin1 in K562 cells following 24-hr treatment of VP16 and 24-hr postdrug removal (recovery).

Immunofluorescence analysis (Fig. 3B,C) of parental 32D cells showed LC3 accumulation and punctate staining following VP16 treatment. This suggests that, even though apoptosis is detected in these treated populations, the co-existence of other cells undergoing autophagy will contribute to the recovery of non-Bcr-Abl-expressing cell populations. In addition, the comparison of untreated 32D and C4 cells indicates a higher basal level of autophagy in C4 cells—with levels of punctate LC3 staining increasing further following VP16 treatment (Fig. 3B). TonB2.0μgDox cells have lower basal expression of LC3 (as indicated by Western analysis), but show increased punctate staining following treatment. VP16-treated K562 cells also show an increase in punctate LC3II staining (Fig. 3C) particularly during recovery (i.e., drug withdrawn for 24 hr). In K562 cells, immunofluorescence of Beclin1 (another autophagy marker) showed increased expression during recovery. The increased expression of both LC3II and Beclin1 strongly suggests elevation of autophagy during recovery.

Imatinib treatment and recovery

We then evaluated the response of Bcr-Abl-expressing cell lines to Imatinib and examined longer term recovery. Previously, we reported that C4 cells are resistant to Imatinib at low concentrations of drug (0.5 μM) compared to C2 cells, but at 1 μM, both cell lines respond to treatment in a similar way—suggesting that saturation of the active site is not achieved in higher expressing C4 cells until 1 μM [5]. Cells were treated with 1 μM (C2 and C4 cells) or 5 μM Imatinib (TonB cells and K562 cells) for a period of 24 hr as the latter cell lines were more resistant. PI negativity/viability was 26.9% ± 1.7% for C2 cells, 31.0% ± 1.1% for C4 cells, 14.2% ± 0.2% for TonB0.2μgDox, and 27.2% ± 3.1% for TonB2.0μgDox [Fig. 4A(i),(ii)]. Similar to VP16 treatment, K562 cells were initially resistant to Imatinib (24-hr treatment) with viability remaining at 91.7% ± 0.3% [Fig. 4A(iii)]. Drug was withdrawn from all cells at 24 hr and fresh media added. Viability of K562 cells reduced to 51.1% ± 1.7% in the first 24-hr recovery period and to 21.5% ± 0.71% in the subsequent 24 hr [Fig. 4A(iii)]. At 9 days, post-treatment cells were 58.2 ± 1.0%.

Figure 4.

Cellular response to Imatinib treatment and recovery. A: Viability assessment following treatment and removal of Imatinib; bar charts represent percentage viability based on the exclusion of PI at 24-hr treatment and subsequent recovery following removal of drug. A(i): C2 and C4 cells treated with 1 μM Imatinib for 24 hr. A(ii): TonB0.2μgDox and TonB2.0μgDox cell line treated with 5 μM Imatinib for 24 hr. A(iii): K562 cells treated with 5 μM Imatinib for 24 hr. B: Morphology 48 hr following Imatinib removal from C2, C4, TonB2.0μgDox, and K562 cells. Images show 40× view and detailed enlargement of boxed section. Black arrow = apoptosis, red arrow = vacuoles. C(i,ii): Bar charts represent percentage nonviable cells (PI positive = clear bar) and percentage apoptotic morphology (=black bar). A minimum of hundred cells from three separate experiments were scored (any small shrunken or condensed chromatin stained cells or cells with clear nuclear fragmentation). C(i): Bar graph representing C2/C4 and TonB0.2μgDox following 24-hr treatment with 1 and 5 μM Imatinib, respectively. C(ii): Bar graph representing K562 at 24-hr treatment with 5 μM Imatinib and 24 hr following drug removal. D: Evaluation of the autophagy marker LC3. D(i): Western blot of equal amounts (80 μg) of total protein lysate from C2/C4 and TonB2.0μgDox showing endogenous LC3I and LC3II expression in cells treated with or without Imatinib (1 and 5 μM Imatinib, respectively) for 24 hr. K562 cell lysates were from untreated, 24-hr treatment, and 24 hr following drug removal (recovery). D(ii): Immunofluorescence of LC3 was performed on C2, C4, TonB2.0μgDox and K562 treated with Imatinib (concentrations stated above) for 24 hr. E: 40× view of LC3 immunofluorescence analysis of TonB2.0μgDox following the removal of doxocyclin for a 48-hr period in the absence of IL-3 (red arrow = fragmented nuclei/apoptosis, yellow arrow = intact nucleus). Graphs were prepared using Graphpad Software4.0. Bars indicate mean values ± standard deviation. Morphology; cells were viewed under the 40× optical lens of a DP70 digital microscope camera; pictures were captured with Olympus DP-Soft823 version 3.2 acquisition software. Pictures were then processed using Adobe Photoshop CS2.

Consistent with the previous reports, apoptotic morphology was observed following treatment with Imatinib (black arrows Fig. 4B). This occurred within the first 24-hr treatment period in C2 and C4 and TonB2.0μgDox cells (morphologies/treatment effects were similar for TonB0.2μgDox—Data not shown) or after 24 hr without Imatinib in K562 cells. However, the levels of apoptosis were not consistent with levels of PI uptake [Fig. 4C(i,ii)]. Apoptotic morphology was scored as 25.1% ± 5.6% for C2 cells, 18.4% ± 2.7% for C4 cells, and 25.7% ± 2.1% for TonB2.0μgDox cells [Fig. 4C(i)]. For K562 cells, apoptotic morphology was 31.0% ± 3.4% following a 24-hr recovery period, which is ∼20% lower than PI positivity [Fig. 4C(ii)]. When apoptosis was scored in K562 cells following 48 hr recovery, this remained relatively unchanged at 34.0% ± 4.1%. PI positivity in this period increased by a further 30%, suggesting the presence of an alternative cell death process. Cells not undergoing apoptosis demonstrate swelling and vacuolization of the cytoplasm with an intact nucleus (Fig. 4B, red arrows). In all cases, regardless of Bcr-Abl levels, cells recover rapidly, repopulating the culture, and regaining normal morphology.

We then assessed whether the increased vesicular content within these cells was due to elevation of autophagy. Western blot analysis indicated that the ratio of LC3II to LC3I increases upon treatment with Imatinib [Fig. 4D(i)]. In K562 cells, this further increases during recovery. Immunofluorescence also demonstrated punctate LC3 staining following Imatinib treatment [Fig. 4D(ii)].

The TonB.210 cell line depends upon the addition of doxocyclin for Bcr-Abl expression. Removal of doxocyclin from TonB2.0μgDox induces both apoptotic morphology (fragmented nuclei) and intact cells with elevated punctate LC3II expression (Fig. 4E), resembling Imatinib treatment.

To further validate the induction of autophagy, we transfected an expression plasmid containing LC3-GFP into C4 cells. Punctate green fluorescence was evident, and this overlapped with MDC another marker of autophagy. The number of cells with elevated GFP increased following Imatinib treatment [Supporting Information Fig. S2(i),(ii)].

Role of autophagy in recovery of Imatinib-treated cells

As autophagy is clearly present in treated and recovering populations, it remains unclear whether it is a feature of cells about to die (by an undefined/nonapoptotic mechanism/Type II cell death) or whether autophagy plays a role in the recovery of cells. We therefore investigated whether inhibition of autophagy would reduce the recovery of treated cells. K562 cells were evaluated as these were the most chemoresistant cell line and are derived from human blast crisis CML.

Autophagy was initially inhibited with 3-methyladenine (3MA), a known PI3K(III) inhibitor. At 10 μM, there was minimal effect on viability of the culture alone. The combination treatment (3MA + Imatinib) did not have any significant effect on culture viability at 24 hr. However, as this treatment was extended to evaluate recovery, it was clear that 3MA, when combined with Imatinib, significantly impeded recovery at 5 days post-treatment (P = *0.0316) (Fig. 5A).

Figure 5.

Inhibition of autophagy in K562 cells with 3MA and ATG7&Beclin1 siRNA knockdown. A: Graphical representation of viability determined by PI negativity in K562 cells treated with 5 μM Imatinib alone or combined with 10 mM 3MA for 24-hr treatment and up to 5 days postwithdrawal of both agents. B(i,ii): Confirmation of siRNA knock-down of autophagy genes (i) ATG7 and (ii) Beclin1. B(i): ATG7 knock-down confirmation by immunofluorescence. Upper panels of the first image quartet represent untreated K562 WT and K562 24-hr post-siRNA-transfected cells. Two lower panels represent these cells treated with 5 μM Imatinib for 24 hr. The second image quartet shows K562 WT and K562 72-hr post-siRNA transfection untreated (upper panels) and 48-hr post-Imatinib withdrawal (lower panels). B(ii): Immunofluorescence confirmation of Beclin1 knock-down [Image layout similar to B(i)]. First image quartet shows K562 WT and K562 24-hr post-siRNA transfection, second image quartet shows K562 WT and K562 72-hr post-siRNA transfection. Upper panels in both image quartets show untreated controls while lower panels show 24-hr treatment or 48-hr post-Imatinib withdrawal. Note: B(i,ii): For siRNA-transfected samples 48-hr post-Imatinib treatment, cell numbers are very low as shown in Fig. 5D below. B(iii): Total protein lysates (20 μg) analysis of Beclin1 in K562 WT versus K562 siRNA-transfected cells. Graphical representation of densitometric analysis of percentage Beclin1 expression. C: Percentage viability bar chart (PI negativity) in K562 cells siRNA transfected and untransfected at initial 24-hr Imatinib treatment and up to 48-hr postdrug removal. Statistical analysis was performed using Graph-pad software 4.0. P values generated using an unpaired Student's t-test *P value = 0.001/**P value = 0.0072. D: 10× view of K562 untransfected and K562 siRNA-transfected 48 hr following drug withdrawal. Black arrowheads—potentially viable cells in siRNA-transfected culture. Several of these are evident in the untransfected cultures (right hand box), whereas they are rare in siRNA-transfected cultures (left hand box). E(i): 40× View, K562 untransfected (upper panels) and K562 siRNA transfected (lower panels); first column = untreated, second column = 24-hr post-treatment, third column = 48-hr post-treatment. E(ii): Bar graph representing K562 untransfected and siRNA transfected at 24-hr post-treatment, morphology score versus PI positivity. Think verses may be spelled wrong in Figure 5Eii.

We more specifically inhibited autophagy by knocking down two essential autophagy genes, ATG7 and Beclin1. To ensure knockdown of either gene alone would not impact on cell viability, genes were initially transfected separately and subsequently together (Data not shown). Over 72 hr, the “knock-down” treatments alone did not affect viability. To ensure autophagy was effectively inhibited, the double knockout strategy was employed. Levels of knock-down were confirmed by Western blot and/or immunofluorescence for Beclin1 and ATG7, respectively (Fig. 5B(i),(ii), respectively). Western blot analysis [Fig. 5B(ii)] indicates a knock down of Beclin1 by ∼25% at 24 hr. At 48 hr, this was reduced to 40% of the original level of expression of Beclin1. Levels began to increase after 72 hr. A reduction in autophagy was also indicated by reduced LC3II immunofluorescence in the presence of siRNA (Supporting Information Fig. S2A). Viability at the initial 24-hr treatment period was not affected, whereas the effects of autophagy knock-down were evident during recovery (Fig. 5C). The reduction in recovery from 51.1% ± 1.73% (untransfected cells) to 22.5% ± 0.5% (siRNA-transfected cells) was significant at 24-hr post-treatment (P = 0.0010) and was further reduced from 21.5% ± 0.7% to 4.2% ± 0.4% 48-hr post-treatment (P = 0.0017).

Fewer intact cells were also visible in siRNA-treated cultures [Fig. 5D,E]. The increased levels of cell death can also be seen in more detail in Fig. 5E. These cultures contained a mixture of morphologies. siRNA-transfected cells had increased levels of very small-shrunken cells, which appear to have very condensed and irregular chromatin [Supporting Information Fig. 2B(i),(ii)]; therefore, we cannot rule out apoptosis involvement (NB inhibition of autophagy may also affect morphology of cells undergoing apoptosis). Accurate determination of the cell death involved would require further assays of additional cell death mechanisms and is beyond the scope of this article. This data collectively suggests that autophagy has an important role in the recovery of drug-treated CML populations and that overall cell killing by Imatinib can be improved by inhibition of autophagy.


Despite the success of Imatinib, residual disease exists in responsive patients, which necessitates long-term treatment. This failure to eradicate CML provides an opportunity for resistant cells to emerge. More effective targeting of residual cells is therefore required for curative regimes in CML. To achieve improved regime design, a greater understanding of how leukemic cells respond to drug treatment is needed. In this study, we have investigated the response of Bcr-Abl-expressing cells to two different types of drug. In addition, we have assessed the impact of elevated Bcr-Abl expression. We have found that the induction of autophagy in response to drug treatment is associated with persistence and recovery of a subpopulation of leukemia cells. The incorporation of autophagy inhibitors therefore has the potential to improve treatment regimes.

Following treatment with VP16/etoposide, we found that cells expressing low levels of Bcr-Abl undergo apoptosis and cannot recover. In contrast, leukemic cells expressing high levels of Bcr-Abl are resistant to apoptosis, but undergo autophagy accompanied by an alternative delayed cell death process. This is a p53-independent response, as K562, a well-characterized cell line, lacks p53 expression [37]. The induction of autophagy may present an opportunity for survival as a subset of cells that do not progress to a death process, can recover, and repopulate. Non-Bcr-Abl-expressing parental cells exhibit both autophagy and apoptosis following treatment and recover following drug withdrawal. This implies that, following chemotherapy, the normal hematopoietic progenitor population, despite sustaining some damage and death, is likely to recover. Undoubtedly, the bone marrow microenvironment will provide much more survival signaling through colony-stimulating factors (CSFs) and adhesion molecules than a limited culture system supplemented with IL-3. It was also apparent from our studies that (CSF) IL-3 is particularly protective and could help cells expressing low levels of Bcr-Abl to inhibit apoptosis, induce autophagy, and recover. The consequences of these findings for treatment with a DNA damaging agent could therefore be envisaged as follows: (1) normal progenitor cells in a CSF rich niche may sustain some apoptosis—but also induce autophagy and have good potential for recovery; (2) low Bcr-Abl-expressing cells will undergo extensive apoptosis in CSF-depleted environments and are unlikely to recover. However, in a high-CSF niche, they will have good potential for recovery; (3) for high-Bcr-Abl-expressing cells, despite extensive nonapoptotic/delayed cell death, populations will recover regardless of the presence/absence of CSFs. Although we have focused on one DNA-damaging agent, we expect that these findings could be transferable to other DNA-damaging drugs. Autophagy is a response to damage/stress, and it may be the primary response in apoptosis defective cancer cells [38]. Autophagy has been reported in solid tumors in response to cisplatin [39], 5-flurouracil [40, 41], and alkylating agents [41], and several of these studies suggest that treatment could be improved by inhibiting autophagy [16, 42].

We also evaluated cellular responses to Imatinib. Cells that express higher levels of Bcr-Abl require escalation of dose to achieve kinase inhibition (as expected from previous work and clinical studies), and when inhibition is achieved, these cells will die in the same way. We found two (or more) cell death pathways (apoptosis and nonapoptotic/possibly Type II death) present in Imatinib-treated cultures. Importantly, all treated cells recovered and rapidly repopulated the culture following Imatinib withdrawal. Autophagy was important for this recovery as it could be significantly reduced by inhibition with 3MA or knock-down of the autophagy regulators ATG7 and Beclin1.

We have referred to the alternative death mechanism induced in these cells as Type II cell death, based on the presence of autophagy. Type II death describes cells that retain an intact nucleus, accompanied by extensive cytoplasmic vacuolization. However, this death may be accompanied by, but not directly caused by autophagy [43]. In this study, we propose that autophagy plays a significant role in the survival and recovery of leukemic populations following drug treatment. However, we cannot say conclusively whether it plays a proactive or passive role in the alternative death mechanism observed, as when autophagy was inhibited with siRNA an alternative and undefined death process was present. Other studies have also reported an undefined death mechanism (nonapoptotic, caspase-independent necrotic-like cell death) in Imatinib-treated cells [44] and in INNO-406 (TKI)-treated cells [17].

A recent study has reported the induction of autophagy in response to Imatinib and its importance for the survival of CML cells [42]. This study evaluated continuous long-term treatment in colony assays and found a greater cytotoxic effect when autophagy was inhibited. Our long-term cultures enabled us to remove the drug after a single 24-hr treatment and monitor cellular responses over time in the absence of drug. This clearly demonstrated delayed and prolonged autophagic responses—particularly in the human K562 cell line. Collectively, both studies indicate that the induction of autophagy is a major response to Imatimib and can significantly impede its efficacy.

Other studies have indicated that the induction of autophagy in response to Imatinib is not exclusive to Bcr-Abl-expressing cells. Targeting c-Kit with Imatinib in GISTs has been reported to induce autophagy leading to enhanced cell survival [15, 45]. Other non-Bcr-Abl-expressing cell lines have also been reported to undergo autophagy in response to Imatinib [20], suggesting that the drug can inhibit other tyrosine kinases important for survival. New TKIs have also been developed for Imatinib-resistant CML. These may have a higher affinity for Bcr-Abl (Nilotinib) or target other kinases in addition to Bcr-Abl—such as src family kinases (Dasatinib, INNO-406) or Aurora kinases (PHA-739 358) [reviewed in ref. [46]). The Bcr-Abl/Lyn kinase inhibitor INNO-406 has also been reported to induce a protective autophagic response in CML cell lines [17]. Autophagy has been reported in response to other targeted therapies such as Tamoxifen [14, 47], Sorafenib [13], and EGF inhibitors [48]. Collectively, these studies suggest that inhibition of key survival/oncogenic signaling by targeted therapeutics may induce a default autophagic response, potentially negating treatment.

Overcoming the autophagy barrier may be key to improving drug treatments, but the clinical translation of these observations is limited by the deficit of selective autophagy inhibitors. One agent that has however shown that potential is the inhibitor of lysosomal acidification—Chloroquine, which is an FDA approved antimalarial agent. This has been reported to improve therapeutic response to a number of agents [16, 17] and to enhance Imatinib activity in CML cell lines and primitive CML cells (CD34+ CD38) [42]. Patients are being recruited to clinical trials with Chloroquine and Imatinib (CHOICES—Chloroquine and Imatinib Combination to Eliminate Stem cells—Hollyoakes; unpublished). This strategy is designed to limit the protective effects of autophagy in the stem/progenitor cells that remain following Imatinib treatment.

For Imatinib-resistant patients, such as those harboring the Bcr-Abl T315I mutation (15–20% of resistant patients), alternative strategies are likely to be required. The histone deacetylase inhibitor—suberoylanilide hydroxamic acid—has been reported to have activity against T315I cells, but its efficacy was also limited by autophagy and improved with chloroquine [21]. Our study has further indicated that a less selective agent such as a DNA-damaging agent (etoposide/VP16) will induce apoptosis in cells with low levels of Bcr-Abl, but, predominantly, autophagy in cells with high levels of oncogenic/CSF-survival signaling. However, the manipulation of autophagy should be broached with some caution as our data indicated the induction of autophagy by non-Bcr-Abl-expressing cells during recovery from nonselective drug treatment. Therefore, if nonselective drugs are combined with autophagy inhibitors, it could impact negatively on the recovery of normal cells.

Imatinib resistance may be augmented by a new emerging problem. As with any long-term self-administration drug, Imatinib will have compliance issues [49]. It was recently reported that, of the 14% of patients who took less than 80% of their recommended Imatinib dose, none achieved a complete cytogenetic response [50]. With rising numbers requiring continuous Imatinib treatment, more patients will eventually require alternative therapies.

Current studies indicate the potential for strategically designed regimes to circumvent autophagic survival and promote more extensive cell death in leukemic populations. The use of a selective TKI combined with an autophagy inhibitor may provide a unique opportunity to target autophagy that is induced only in the persisting (possibly progenitor) leukemic compartment. Early implementation of such regimes may limit the need for dose escalation or nonselective agents as patients develop resistance to TKIs. As CML progenitor cells appear to be compromised but not killed by TKI inhibition alone, selective inhibition of their survival mechanisms is critical in the development of curative treatments for CML.


We thank Dr. Michelle Nyhan and Dr. Tracey O'Donovan, Cork Cancer Research Centre, for advice and proofreading of final manuscript.