Introduction

Chronic myelogenous leukemia (CML) is a clonal myeloproliferative disorder characterized by Philadelphia (Ph) chromosome, which is formed by a reciprocal translocation between chromosome 9 and 22. This chromosomal rearrangement fuses a truncated BCR gene to sequences upstream of the second exon of ABL gene, resulting in a fusion oncoprotein BCR/ABL with constitutive tyrosine kinase activity.1 Since multiple hematopoietic lineages contain the translocation, it is generally believed that CML arises in a pluripotent hematopoietic stem cell. Imatinib mesylate, a small compound that inhibits the tyrosine kinase activity of BCR/ABL has been successfully developed for CML therapy with impressive clinical responses.2, 3 Currently, imatinib offers an effective treatment for patients with chronic phase CML; however, many patients with advanced disease experience only a short-term response, with subsequent emergence of resistant leukemic cells and clinical relapse.4, 5 Even in patients with chronic phase CML, it has been shown the malignant progenitors persist upon complete cytogenetic remission with imatinib treatment,6 indicating that imatinib predominantly targets more mature CML cells and that leukemic stem cells are relatively resistant to imatinib.

Diverse mechanisms of resistance to imatinib have been identified, including amplification of BCR/ABL gene, point mutations inside or outside BCR/ABL kinase domain and increased plasma level of α1 acid glycoprotein.7, 8, 9, 10, 11 Among them, the pre-existence or acquisition of point mutations in BCR/ABL oncogene has been identified as the most common mechanism of relapse for CML patients treated with imatinib.10 However, in some imatinib-resistant CML patients, neither point mutation nor amplification of BCR/ABL gene was detected, indicating that other factors likely contribute to imatinib resistance.8, 10

Drug efflux by major multidrug transporters, such as P-glycoprotein (P-gp), ABCG2 (also known as BCRP) and multidrug-resistant proteins (MRP), has been well characterized in pharmacokinetic resistance to a variety of antineoplastic agents. Recently, it has been reported that overexpression of the MDR1 gene in leukemia cell lines was associated with resistance to imatinib.12, 13, 14, 15 Moreover, it has been shown that imatinib can interact with P-gp, but not with MRP1,16, 17 and distribution of imatinib in the central nervous system was found to be limited by P-gp-mediated efflux in the blood–brain barrier, suggesting that imatinib is a substrate of P-gp.18 Given that P-gp is highly expressed in normal hematopoietic stem cells (HSCs),19, 20 it is possible that this transporter pumps imatinib out of stem cells and contributes to resistance in CML. However, Ferrao PT et al21 reported that overexpression of P-gp in K562 cells did not confer resistance to imatinib in vitro. Therefore, the role of P-gp in resistance to imatinib remains unclear. To test the in vivo contribution of P-gp to imatinib resistance, we established an Mdr1a/1b-null CML animal model via BCR/ABL retroviral transduction and bone marrow transplantation (BMT), and compared the sensitivity of diseased mice to imatinib. Our studies showed no changes in hematopoietic responses to imatinib in the absence of P-gp, indicating that P-gp expression in HSCs does not confer resistance to imatinib in CML.

Materials and methods

Animals and cell lines

Mdr1a/1b-null mice22 on the FVB background were originally obtained from Taconic Farms (Germantown, NY, USA) and raised in our animal research center. This mouse strain was chosen because mice have two Mdr1-type P-glycoproteins (Mdr1a and Mdr1b) that fulfill together the same functions as the single human P-gp (MDR1).22 Female FVB/NJ mice were purchased from the Jackson Laboratories (Bar Harbor, MA, USA). All animals were housed in ventilated cages, provided with sterilized food and acidified water, and used between 6 and 14 weeks of age. All mouse studies were conducted according to protocols approved by the Institutional Animal Care and Use Committee of St Jude Children's Research Hospital.

K562 human chronic myeloid leukemia cells, 293T and ecotropic retroviral packaging cell line GP+E86 were grown in DMEM (Mediatech, Herndon, VA, USA) supplemented with 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine. Cells were cultured and maintained in a log-phase growth at 37°C in a humidified atmosphere with 5% CO2.

Retroviral constructs and reagents

The HaMDR1sc retroviral vector and MSCV-BCR-ABL-IRES-EGFP vector was described previously.23, 24 Imatinib mesylate was extracted from capsules of Gleevec (Novartis, Basel, Switzerland) by dissolving in pH 5 sterile distilled water. The stock solution was prepared at 10 mM, filtered and stored at −20°C. For animal experiments, fresh preparations at 4 and 8 mg/ml were made immediately prior to use.

Preparation of MDR1-transduced clonal K562 cells

The HaMDR1sc retroviral stocks were obtained by transient transfection of 293T cells,25 using GenePORTER™ transfection reagent (Gene Therapy Systems, San Diego, CA, USA). In all, 2 × 105 K562 cells were transduced in a cocktail containing DMEM, 10% FBS, 50% retroviral supernatant and 6 μg/ml polybrene. After incubation at 37°C for 16 h, the cells were grown in fresh medium. Thereafter, transduced K562 cells underwent three rounds of fluorescent activated cell sorting (FACS) for Rhodamine 123 efflux, as previously described.26 Briefly, cells were resuspended in DMEM with 10% FBS at 2 × 106 cells/ml, and Rhodamine 123 was added to a final concentration of 0.5 μg/ml. After incubation in a 37°C water bath for 15 min, cells were washed with PBS, then resuspended in fresh medium, and incubated for 90 min at 37°C. Then, cells were spun down and resuspended in DMEM with 10% FBS, and kept on ice until FACS analysis. The expression levels of P-gp on the cell surface were analyzed on flow cytometry (Becton Dickenson, Franklin Lakes, NJ, USA) by staining cells with monoclonal antibody 4E3 (DAKO, Carpinteria, CA, USA), as previously described.26 Individual clones of K562/pHaMDR1 cells were isolated by single cell sorting, expanded and screened for high level of P-gp expression using 4E3 staining.

Cell viability assay

The parental K562 cells and MDR1-transduced K562 (K562/pHaMDR1) cells were seeded at 2 × 105 cells/ml in 12-well plates. The stock solution of imatinib was thawed and diluted before use, then added at various concentrations as indicated. After incubation at 37°C for 96 h, viable cells were counted, as assessed by trypan blue exclusion. Results are expressed as the mean from three independent experiments of the percentage of viable cells relative to the untreated controls.

Retroviral transduction of murine bone marrow (BM) cells

The BCR-ABL-IRES-EGFP ecotropic producer cell line was generated in GP+E86 packaging cells as described previously.25 The 50% of cells with the highest GFP expression were sterilely sorted by flow cytometry, subsequently expanded and cryopreserved. Retroviral transduction of BM cells was performed as described.26 Briefly, BM cells were harvested by standard methods from Mdr1a/1b-null mice and FVB/NJ mice 3 days after treatment with 150 mg/kg 5-fluorouracil. Cells were prestimulated in DMEM with 15% FBS, 20 ng/ml murine IL-3, 50 ng/ml human IL-6 and 50 ng/ml murine stem cell factor for 48 h, and then replated onto a confluent monolayer of irradiated (3000 cGy) producer cells. After 48 h coculture, the loosely adherent and nonadherent cells were recovered. The transduction efficiencies of BM cells with GFP expression were determined, by flow cytometry, to be at between 17 and 32%.

BMT and treatment of animals

BCR-ABL-IRES-EGFP-transduced BM cells (1–9 × 105 GFP-positive cells/mouse) were injected into lethally irradiated (two doses of 450 cGy irradiation separated by 4 h) syngeneic FVB/NJ mice via the lateral tail veins. Equal numbers of transduced Mdr1a/1b-null or wild-type BM cells were transplanted into recipients in each independent experiment. The administration of imatinib or placebo (sterile distilled water) was performed as described previously,27 except that treatment started from day 1 or day 7 after BMT and continued until the animals were dead. Briefly, in the first two experiments, primary recipients were given imatinib, at 50 mg/kg every morning and 100 mg/kg every evening starting on day 7, by gavage using 20G feeding needles (Popper & Sons, New Hyde Park, NY, USA). A higher dose of imatinib (200 mg/kg per day) was used starting on day 1 in another two independent experiments. The mice were monitored for clinical signs of disease twice daily. For complete blood counts (CBC) with differential, peripheral blood was collected at various intervals from the retro-orbital cavity using a heparinized glass capillary. GFP expression and lineage of peripheral white blood cells (WBCs) were analyzed by flow cytometry after lysis of the red blood cells with ammonium chloride and staining cells with an antibody cocktail consisting allophycocyanin (APC)-conjugated anti-Gr-1/Mac-1 and phycoerythrin (PE)-conjugated anti-B220/Thy1.2 (PharMingen, San Diego, CA, USA). The weight of spleen and liver were obtained at necropsy. For secondary BMT experiments, FVB/NJ mice were lethally irradiated (900 cGy) and received 1–6 × 106 BM cells per mouse from primary recipients that had received 4- or 5-week treatment of imatinib. Survival and CBC were measured until 6 months after BMT.

Statistical analyses

The statistical differences between experimental groups were compared by two-tailed unpaired Student's t-test or nonparametric test using InStat 2.03 software for Macintosh (Apple, Cupertino, CA, USA). P-value of 0.05 is chosen as the limit of statistical significance.

Results

Overexpression of P-gp in K562 cells conferred minimal resistance to imatinib in vitro

Ferrao et al21 recently described that K562 cells overexpressing P-gp were not resistant to imatinib, contradicting several reports showing that overexpression of P-gp was associated with imatinib resistance in leukemia cell lines.12, 13, 14, 15 To address this discrepancy, we first prepared K562/pHaMDR1 cells by pHaMDR1sc retroviral transduction. After three rounds of FACS selection, 96% of cells exhibited high Rhodamine 123 efflux activity (Figure 1a). The expression of P-gp on the cell surface was confirmed by flow cytometry after staining with an anti-P-gp monoclonal antibody 4E3 (Figure 1b). The bimodal distribution of P-gp expression in K562/pHaMDR1 cells may reflect selection of an oligoclonal population due to the relative low transduction efficiency.

Figure 1
figure 1

Expression of P-gp in the K562/pHaMDR1 cell line and its sensitivity to imatinib. (a) Rhodamine 123 efflux assay of K562 cells (shaded area) and K562/pHaMDR1 cells (solid curve) by flow cytometry; (b) expression of P-gp on the K562 (shaded area) and K562/pHaMDR1 (solid curve) cell surface after staining with monoclonal antibody 4E3. (c) Cell viability assessed by trypan blue exclusion of K562/pHaMDR1 and its parental cells after continuous exposure to imatinib at the indicated concentration for 96 h. The differences in viability with statistical significance are indicated by asterisks (P<0.05). Results are expressed as the percentage of viable cells relative to the untreated controls. Error bars represent the s.d. from three separate experiments.

Full size image

The imatinib sensitivity of K562 cells engineered to overexpress P-gp was compared to parental K562 cells by measuring cell viability following 96 h of continuous incubation with imatinib. We found there was a modest difference in cell viability between these two cell lines, which reached statistical significance (P<0.05) only at 0.01 and 0.05 μ M concentration of imatinib (Figure 1c).The IC50, the dose of imatinib that resulted in a 50% decrease in cell survival, was 0.135 μ M for K562 cells and 0.178 μ M for K562/pHaMDR1 cells. At a 0.5–1 μ M dose of imatinib, which is the steady-state plasma concentration of imatinib in patients given 400 mg Gleevec/day,9 K562/pHaMDR1 cells remained sensitive to imatinib. Moreover, a similar dose-dependent cell killing curve was found in an individual clone of K562/pHaMDR1 cells with highest expression of P-gp isolated by single cell sorting (data not shown), suggesting that P-gp expression level was not associated with imatinib resistance in this system. Thus, our results indicated that overexpression of P-gp in K562 cells only conferred minimal resistance to imatinib in vitro.

Mice transplanted with BCR/ABL-transduced Mdr1a/1b-null BM cells did not have improved responses to imatinib

To examine the in vivo effect of P-gp expression in HSCs on imatinib sensitivity, we induced CML-like myeloproliferative disease (MPD) in recipient mice by using a well-established BCR/ABL retroviral transduction/transplantation model.27, 28, 29, 30 In comparison with mice that received wild-type BM expressing BCR/ABL, mice reconstituted with transduced Mdr1a/1b-null BM exhibited similar clinical features of MPD, including an abrupt increase in peripheral WBC counts (Figure 2a) and marked hepatosplenomegaly (Figure 2b, c). The median spleen and liver weights were about 380 mg and 2.81 g (wild type), and 457 mg and 3.04 g (Mdr1a/1b null), respectively, with no statistically significant difference (P>0.05). More than 75% of peripheral WBCs were GFP positive (data not shown). The median survival of two groups of placebo-treated mice was about 20 days after BMT (Figure 3). Consistent with previous studies,27, 28, 29, 30 extensive pulmonary hemorrhage was observed, which may be the direct cause of death in most sick animals.

Figure 2
figure 2

Efficacy of imatinib for treatment of myeloproliferative disease derived from BCR/ABL-transduced Mdr1a/1b-null or wild-type BM. (a) Effects of imatinib treatment on the peripheral WBC counts in mice reconstituted with BM expressing BCR/ABL. Animals transplanted with BCR/ABL-transduced Mdr1a/1b-null (KO) or wild-type (WT) BM were started on treatment with imatinib or placebo from day 7 after BMT. Complete blood counts were obtained at day 16. The results were generated from two independent experiments. Spleen (b) and liver (c) weights of mice in the different groups were determined at necropsy. Each point represents an individual mouse and the median values are indicated by bars. The results summarized four independent experiments.

Full size image
Figure 3
figure 3

Mice that received Mdr1a/1b-null BM expressing BCR/ABL had similar survival after imatinib treatment as compared with those reconstituted with transduced wild-type BM. Kaplan–Meier plot showing survival of recipients of BCR/ABL retroviral transduced BM from wild-type (placebo: n=22; imatinib: n=23) or Mdr1a/1b-null donors (placebo: n=26; imatinib: n=32).

Full size image

Treatment of diseased mice with imatinib (150–200 mg/kg/day in two divided doses) significantly lowered the peripheral WBC counts and improved hepatosplenomegaly (Figure 2). Imatinib-treated mice had an average WBCs of 9.4 × 103/μl (wild type) and 18.5 × 103/μl (Mdr1a/1b null) at day 16 after BMT, compared with 80.6 × 103/μl (wild type) and 84.2 × 103/μl (Mdr1a/1b null) for placebo-treated mice. There were very significant differences in spleen and liver weight between placebo and imatinib-treated mice (P<0.01). Imatinib treatment prolonged the median survival of diseased mice to 34 days after BMT (Figure 3). However, most of the treated mice still died within 5 weeks; and the majority of WBCs in peripheral blood were GFP-positive myeloid cells (Figure 4), suggesting that many BCR/ABL-expressing cells survived imatinib treatment. To exclude the toxicity of high dosage of imatinib, the nontransplanted FVB/NJ mice were treated with imatinib at 200 mg/kg per day by gavage. We found that continuous administration of imatinib for 7 weeks did not cause myelosuppression or death in these animals (data not shown).

Figure 4
figure 4

Imatinib treatment did not eradicate GFP-positive leukemic cells. (a) Representative lineage analysis of peripheral WBCs from imatinib-treated mice (top: wild type; bottom: Mdr1a/1b null). (b) The percentage of GFP-positive cells in peripheral WBCs from diseased mice that received imatinib for 30 days. (Wild type: n=4; Mdr1a/1b null: n=4. P>0.05.)

Full size image

Importantly, mice reconstituted with Mdr1a/1b-null BM cells expressing BCR/ABL did not show better responses to imatinib with comparison to mice transplanted with transduced wild-type BM, as evidenced by similar leukemic burden indicators in these two imatinib-treated groups. The peripheral WBC counts in imatinib-treated Mdr1a/1b-null and wild-type groups are shown in Figure 2a. There was no significant difference in spleen weight (168±76 mg for wild type, 207±100 mg for Mdr1a/1b null, P>0.05) and liver weight (1.33±0.27 g for wild type, 1.34±0.70 g for Mdr1a/1b null, P>0.05) at necropsy between these two groups (Figure 2b and c). Furthermore, a similar survival time was observed between imatinib-treated mice reconstituted with Mdr1a/1b-null BM and wild-type BM expressing BCR/ABL (Figure 3). These two groups of mice had the same median survival of 34 days after BMT. There was no statistically significant difference in the percentage of GFP-positive cells in peripheral blood between two imatinib-treated groups (Figure 4b). All these results indicated that BCR/ABL-transduced hematopoietic cells were not sensitized to imatinib in the absence of P-gp.

Imatinib-treated MPD was transplantable to secondary recipients

BM cells from diseased mice with 4 or 5 weeks imatinib treatment were transplanted into secondary recipients. Two independent experiments showed that MPD was present in most lethally irradiated secondary recipients (Table 1). The disease was characterized by hepatosplenomegaly and gross pulmonary hemorrhage, but with low WBC counts in some mice. In the first experiment, 71% (five out of seven) of secondary recipients of imatinib-treated mice transplanted with transduced Mdr1a/1b-null BM died from MPD within 4 months after BMT, while 67% (two out of three) of secondary recipients of imatinib-treated wild-type group had a similar disease. In the other transplantation experiment using 4–5 × 106 BM cells per mouse, all the secondary recipients of two imatinib-treated groups developed acute MPD with shorter latency. We found in these two experiments that the transplantability and latency of disease were not changed by the absence of P-gp in BM cells, indicating that BCR/ABL-transduced Mdr1a/1b-null leukemic initiating cells survived imatinib treatment.

Table 1 Manifestation of MPD in secondary recipients of imatinib-treated mice
Full size table

Discussion

In this study, we found that overexpression of P-gp in K562 cells only conferred minimal resistance to imatinib, which is consistent with the data of Ferrao et al.21 Our results of cell viability assay demonstrated that at low doses of imatinib, P-gp could have a minor protective effect on imatinib-induced growth inhibition and cell death. However, at 0.5–1 μ M level, which is the physiological plasma concentration of imatinib in patients given 400 mg Gleevec per day,9 K562 cells engineered to overexpress P-gp did not show resistance to imatinib. Several previous reports suggested that P-gp is involved in imatinib resistance.12, 13, 14, 15 However, it is important to note that drug (daunorubicin, adriamycin, vincristine or etoposide)-selected resistant cells were used in most of these previous studies, which would exhibit changes other than overexpression of P-gp. In contrast, we engineered to overexpress P-gp by retroviral transduction in isogenic cells where the only difference was P-gp expression on the cell surface.

A previous study showed that there was resistance to imatinib at the HSC or progenitor level in a murine BM retroviral transduction and transplantation model of CML-like disorder.27 The underlying mechanisms are unknown. Here, we demonstrated equivalent hematopoietic responses to imatinib in the presence or absence of P-gp, indicating that P-gp expression in leukemic cells was not responsible for resistance to imatinib in these CML mice. Wolff NC et al27 described that continuous imatinib treatment could prolong the survival of CML mice up to 101 days after BMT. However, in our study, despite the obvious hematopoietic responses to imatinib, most of the imatinib-treated mice still died within 5 weeks after BMT. One explanation for this discrepancy could be due to a different mouse strain (Balb/c) used in their studies.27 We found that more than 50% imatinib-treated FVB/NJ mice with MPD had gross pulmonary hemorrhage at necropsy, which may result in the lower survival of our imatinib-treated mice.

Increasing evidence shows that only a small subset of cancer cells are tumorigenic – referred to as cancer stem cells – which have been found in human AML, CML, ALL, breast and brain tumors.31 From these studies, it seems that cancer stem cells share many similar characteristics with normal stem cells, including quiescence of cell cycling, side population phenotype, common cell surface marker and signal transduction pathway. Since high expression of P-gp in normal HSCs was previously described,19, 20 it is possible that CML stem cells will also express this ABC transporter, which may be associated with resistance to imatinib.12, 13, 14, 15, 16, 17, 18 However, our results of in vivo imatinib sensitivity experiments showed that BCR/ABL-transduced hematopoietic cells were not sensitized to imatinib in the absence of P-gp, Mdr1a/1b-null leukemic initiating cells still remained intact after 4–5 weeks treatment of imatinib, indicating that P-gp expression has no significant contribution to imatinib resistance at the HSC/progenitor level.

One potential explanation of our results is that ABCG2, another drug-resistance-associated ABC transporter, could provide a redundant mechanism for imatinib resistance in the absence of P-gp, since ABCG2 has been shown to be highly expressed in HSCs and provide protection against mitoxantrone toxicity.32, 33 Moreover, two recent studies showed that imatinib was a substrate for ABCG2,34, 35 suggesting that simultaneous expression of ABCG2 and P-gp at HSCs might have redundant effects on resistance to imatinib in CML, especially when combined with the quiescent properties of HSCs and other mechanisms.36, 37 However, Houghton et al38 reported that overexpression of ABCG2 in the human osteosarcoma cell line Saos2 did not confer resistance to imatinib and accumulation and efflux of imatinib were similar in the presence or absence of ABCG2 function, indicating that imatinib was not a substrate for ABCG2 transporter. Therefore, further studies are necessary to define the interaction of imatinib with ABCG2, and Abcg2/Mdr1a/1b triple knockout mice39 could be a valuable model for determining the potential redundancy between P-gp and ABCG2 in imatinib resistance.