MRP3, BCRP, and P-Glycoprotein Activities are Prognostic Factors in Adult Acute Myeloid Leukemia

Despite improvements accomplished in the last 30 years on the use of combination cytarabine (Ara-C) and intercalating agents, the overall prognosis for adult acute myeloid leukemia (AML) remains poor (1). One of the best-characterized resistance mechanisms in AML is the drug extrusion mediated by ABCB1 [P-glycoprotein (Pgp)], which has been shown to be associated with a poor outcome (2–8). Other ATP-binding cassette (ABC) proteins, such as ABCC1 (MRP1), also seem to contribute to the resistance in AML, especially in association with Pgp (6, 9).

A new ABC protein, ABCG2 (BCRP), was simultaneously characterized by different groups (10–12) from the cell lines resistant to anthracycline and mitoxantrone. The drug-resistance profile of BCRP-positive cells was done in vitro by Litman et al. (13). They were cross-resistant to mitoxantrone, daunorubicin, doxorubicin, bisantrene, and topotecan. Daunorubicin was also shown to be a BCRP substrate in AML cells (14), and Ara-C was not a BCRP substrate in infant acute lymphoblastic leukemia (15). The specificity of BCRP binding to its substrate could be modified by changes in its amino acid sequence (16). Amino acid 482 mutation in human cancers could affect the clinical application of ABCG2 antagonists. In a recent study by Ross et al., high expression of BCRP mRNA was sufficiently frequent to warrant more extensive investigations to determine the relation of disease and treatment outcome to BCRP expression (17). However, in another study, only 7% of the 40 AML samples tested contained BCRP mRNA levels within the range of their drug-resistant clone, although another 78% were higher than normal blood and bone marrow (18). In a clinical study, 59 childhood AML patients who expressed high levels of BCRP (the authors used the median of BCRP expression as the cutoff for high expression) had a worse prognosis (19). In a recent study, using mRNA expression, we showed that BCRP could be implicated in resistance to chemotherapies in adult AML (20). van den Heuvel-Eibrink et al. have shown that BCRP mRNA expression was significantly higher in refractory/relapsed AML than in de novo AML, suggesting that BCRP is associated with clinically resistant disease in AML (21), in contrast to the results of van der Kolk' study (22). Therefore, BCRP protein expression and functionality for prognosis in AML should be investigated.

MRP1 homologues, such as ABCC2 (MRP2), ABCC3 (MRP3), and ABCC5 (MRP5), have been shown to be expressed at variable levels in AML patient cells. These proteins have been described to confer a resistance to chemotherapeutic agents, such as daunorubicin, mitoxantrone, and etoposide, which are used in the treatment of AML patients (23). However, the role of these proteins in drug resistance in adult AML has not been clarified. In recent studies, MRP3 and possibly MRP2, but not MRP4 and MRP5, have been shown to be involved in drug resistance in childhood AML and childhood T acute lymphoblastic leukemia (24, 25). Specific assays and protein expression studies are necessary to confirm these results. The different functional assays for these new ABC proteins are currently being analyzed, compared, and applied to the prognostic analysis.

Here, we report the protein expression of Pgp, MRP1, MRP2, MRP3, MRP5, and BCRP in six cell lines using flow cytometry. We compared three different probes and modulators to measure the functionality of those proteins. Finally, we have studied the prognostic effect of these six ABC proteins (expression and activity) in 85 adult AML samples.

Cell lines. The present study was carried out with six cell lines at different levels of six ABC proteins, Pgp, MRP1, MRP2, MRP3, MRP5, and BCRP (Table 1): K562, a human erythroleukemia (a gift from B Sickic, Pharmacology Department, Stanford University, Stanford, CA.) and its derivation, K562/HHT30 cell line, which was developed in our laboratory (26). HL60, human myeloid leukemia cells, and its derivation HL60/MRP, which is resistant to daunorubicin. PC13, a lung large cell carcinoma, and PC13 2-2, which are PC13 cells transfected with BCRP (a gift from Dr. H. Komatani, Banyu Tsukuba Research Institute, Ibaraki, Japan; ref. 27).

Patients. Between January 1998 and December 2002, 85 samples from de novo adult AML patients were successfully tested for the six ABC protein expressions. The diagnosis was based on French-American-British (FAB) criteria [11 M1 (13%), 40 M2 (47%), 17 M4 (20%), 16 M5 (19%), and 1 M6 (1%); ref. 28]. Immunophenotyping was done by flow cytometry. Acute promyelocytic leukemia patients were excluded from the study (due to retinoic acid treatment). Patients with t(9;22) were also excluded from the study. For each patient, several clinical and biological characteristics were analyzed (age, WHO performance status, WBC count at diagnosis, CD34 expression, FAB subtypes, and cytogenetics). Karyotypes were defined as previously reported (29).

All patient in this study had a history of prior therapy with anticancer drugs. All patients were treated with uniform chemotherapy and transplantation; having received a combination of Ara-C (100 mg/m²/d) for 10 days, with daunorubicin (45 mg/m²/d) for 3 days, and etoposide (100 or 50 mg/m²/d) for 5 days. Those patients who achieved complete remission (CR) after one or two cycles of therapy received one cycle of consolidation therapy (Ara-C, 500 mg/m²/12 hours for 6 days with daunorubicin, 45 mg/m²/j for 3 days). Patients achieving CR were subsequently scheduled to proceed to allogeneic bone marrow transplantation if a matched sibling donor was available (10 patients); patients >50 years old or lacking a suitable donor received an autograft or chemotherapy (30). Patients receiving one allograft were censored at the time of transplant.

Level of ABC protein expression in cell lines and acute myeloid leukemia samples. Pgp, MRP1, MRP2, MRP3, MRP5, and BCRP protein expression was measured by labeling fresh viable cells with the UIC2, QCRL3, M2I4, M3II9, M5I1, and BXP-34 monoclonal antibodies, respectively (concentration, 5 × 10³ μg/mL), and phycoerythroin-labeled second antibody as described before (31). If necessary, cells were permeabilized in 15% (v/v) lysing solution G (Becton Dickinson, Le Pont de Claix, France) in water and incubated for 15 minutes in PBS/bovine serum albumin containing 1% (v/v) normal goat serum. The expression of ABC proteins was established only in blast cells selected by CD34 antibody (HPCA₂ clone, Becton Dickinson; two-color assays) or other markers (for example, CD33/CD7, CD33/CD2, CD33/CD19, or CD33/CD22 by three-color assays), or selected by physical characteristics when blast cells did not express characteristic markers. Fluorescence was analyzed on a EPICS Altra (Beckman Coulter, Brea, CA) flow cytometer. ABC-Protein expression was determined by the ratio of mean fluorescence intensity (MFI) of each specific antibody/control antibody (Pgp, MRP1, or MRP5/IgG_2a; MRP2, MRP3, or BCRP/IgG₁). For each sample, 5,000 events were collected.

Functional assays of ABC protein in cell lines and acute myeloid leukemia samples. Firstly, six cell lines with different levels of these five ABC proteins were used to assess the ability of three probes [calcein-AM (CAM; 10⁻⁶ mol/L), JC1 (7.5 μmol/L), and mitoxantrone (20 × 10⁻⁶ mol/L)] ± three modulators [MK571 (MK; 20 × 10⁻⁶ mol/L), GG918 (10⁻⁶ mol/L), and fumitremorgin C (10⁻⁵ mol/L)] to measure the functionality of these proteins by flow cytometry. Cell fluorescence was recorded using a Beckman Coulter (EPICS Altra) flow cytometer. The function of ABC proteins was established with blast cells selected by CD34 antibody (FL3 channel, HPCA₂ clone; Becton Dickinson), or by physical characteristics when blast cells did not express characteristic markers.

We used JC1 ± CsA (2 × 10⁻⁶ mol/L) to assess Pgp activity as previously described (8). CAM and JC1 functional assays were assessed as previously described (8, 31). For mitoxantrone staining, cells were washed twice and resuspended in PBS containing 20 × 10⁻⁶ mol/L of mitoxantrone at a concentration of 5 × 10⁵ cells/mL and was incubated at 37°C for 20 minutes with or without modulator. Cells were washed twice in cold PBS, and then were analyzed. Mitoxantrone fluorescence was logarithmically emitted at 675 nm wavelength when excited by laser at 633 nm wavelength. The effect of the inhibitors was expressed as a shift of MFI of the dye accumulation. For each sample, 5,000 events were collected.

Statistical analysis. Correlations between multidrug resistance protein expression and multidrug resistance protein activity were estimated using Spearman's rank order correlation coefficient. The associations between variables were analyzed by Fisher's exact test for categorical variables and by Mann-Whitney U test or Kruskal-Wallis test for continuous variables. Clinical and biological factors were investigated for their influence on remission rate by Fisher's exact test for binary variables and by Mann-Whitney U test or Kruskal-Wallis test for continuous variables. The rates of (a) disease-free survival (DFS) were measured from the establishment of CR until relapse or death from any cause, with observation censored for patients last known alive without report of relapse; (b) overall survival (OS) was measured from diagnosis until death from any cause, with observations censored for time when patients were last reported to be alive. They were estimated with the method of Kaplan and Meier (31) and compared with a log-rank test. Analyses of prognostic factors for treatment outcomes were based on proportional hazards regression models for DFS and OS (32). Significance was defined as two-tailed P ≤ 0.05. The Cox proportional model was used for the multivariate analyses (33). The median follow-up time for censored patients was 1,098 days. The time point used for the proportion of DFS and OS was December 31, 2003. We used StatView software (version 5.0) for statistical analysis (SAS Institute Inc., CA).

We have studied six cell lines: K562, K562/HHT30, HL60, HL60/MRP, PC13, and PC132.2. We found a good correlation between MRP3 expression and MRP3 functionality using mitoxantrone incorporation with or without modulator MK (mitoxantrone ± MK, r = 0.93; P = 0.001 using Spearman's rank correlation test), and also between BCRP expression and its functionality (mitoxantrone ± fumitremorgin C, r = 0.98; P < 0.0001; see data in Table 1). There was no correlation between MRP1 expression and its functionality (mitoxantrone ± MK). As previously published, Pgp functional assays using JC1 ± CsA were correlated with its expression. Any other functional assay was correlated with one protein expression.

In this study, we have also investigated 85 samples from AML patients. Five AML samples are shown in Fig. 1. We found good correlations between MRP3, BCRP, and Pgp expression with their respective functionalities (mitoxantrone ± MK, r = 0.78; P < 0.001; mitoxantrone ± fumitremorgin C, r = 0.83; P < 0.001; JC1 ± CsA, r = 0.69; P < 0.001; Fig. 2). Interestingly, nine M5 AML patients expressed a nonfunctional Pgp (Fig. 2). This dissociation between expression and function was not apparent for the other transporters. Other functional assays were not correlated with expression of ABC proteins. We failed to find a specific functional assay for MRP1. MRP1 expression correlated neither with its functionality (mitoxantrone ± MK; Fig. 2) nor with its functionality (CAM ± MK). CAM ± MK could detect both MRP1 and MRP3 activity at the same time. There was a good correlation between MRP1 expression and functionality only when the patient cells did not express MRP3 (CAM ± MK, r = 0.65; P = 0.001). Conversely, there was a good correlation between MRP3 expression and functionality when the patient cells did not express MRP1 (CAM ± MK, r = 0.60; P = 0.002; data not shown). Therefore, CAM ± MK might be able to assess MRP1 or MRP3 activity when one of two these proteins were absent. In Fig. 2, patients were considered to have high MRP3 activity when their mitoxantrone ± MK MFI levels were ≥200, and patients were considered to have high BCRP and Pgp activities when their mitoxantrone ± fumitremorgin C and JC1 ± CsA MFI levels were ≥150, respectively. Using this cutoff (Fig. 2), 50% and 40% of patients presented Pgp expression and activity, 39% and 35% for BCRP, and 32% and 31% for MRP3, respectively. In AML samples, MRP2 and MRP5 expression was very poor. Only two patients had an expression of MRP2 >1.2, and one patient expressed MRP5 at >1.2. According to the variables in Table 1 and Fig. 1, any assay could be used to evaluate MRP1 activity alone. CAM ± MK was available for MRP3 and MRP1 activities together (not to be used for one specific assay). Therefore, we have studied the correlation between MRP3, BCRP, and Pgp activities with clinical and biological variables.

MRP3 activity (mitoxantrone ± MK) was higher in M5 AML than other FAB subtype AMLs (289 ± 90 versus 134 ± 45; P = 0.04), as well as in patients with a hyperleucocytosis (275 ± 145 versus 89 ± 34; P = 0.01), and in patients with poor or intermediate cytogenetics than in patients with good cytogenetics (189 ± 103 versus 156 ± 45 versus 87 ± 39, respectively; P = 0.05; Table 2). BCRP activity (mitoxantrone ± fumitremorgin C) was not correlated with the clinical and biological variables.

Pgp activity (JC1 ± CsA) was higher in elderly patients than in younger patients (256 ± 60 versus 107 ± 96; P = 0.03), higher in CD34+ patients than in CD34− patients (254 ± 123 versus 58 ± 37; P = 0.002), higher in M1, M2, M4, and M6 FAB subtypes than in the M5 FAB subtype (287 ± 115 versus 48 ± 10; P = 0.002), higher in the poor or intermediate cytogenetic subgroups than in the good cytogenetic subgroup (231 ± 125 versus 182 ± 89 versus 68 ± 59, respectively; P = 0.02).

Fifty six of 85 patients achieved CR (66%). The percentage of DFS was 31 ± 5% at 4 years (median DFS, 371 days) and the percentage of OS was 27 ± 4% at 4 years (median OS, 341 days).

Univariate analysis. In this analysis, Pgp (JC1 ± CsA), MRP3 (mitoxantrone ± MK), and BCRP (mitoxantrone ± fumitremorgin C) activities, as continuous variables, were shown to correlate with prognosis. The Pgp mean value of MFI was 90 ± 45 in patients achieving CR versus 290 ± 78 in other patients (P = 0.04). This high activity was found in patients with poor DFS [P = 0.04; relative risk (RR), 1.90] and poor OS (P = 0.01; RR, 3.10). Elsewhere, MRP3 activity was very high in nonresponders with a MFI of 245 ± 89 versus 79 ± 34 in responders (P = 0.009). This high activity was also found in patients with poor DFS (P = 0.01; RR, 2.7) and poor OS (P = 0.02; RR, 2.01). BCRP activity was also higher in nonresponders with a MFI of 210 ± 78 than in responders with a MFI of 94 ± 34 (P = 0.05). This high activity was also found in patients with a poor DFS (P = 0.03; RR, 2), and poor OS (P = 0.03; RR, 2.12).

Similarly, the expressions of Pgp, MRP3, and BCRP were also correlated with treatment outcome (data not shown). Other significant variables in univariate analysis (CR, age, leukocytes, WHO performance status, and cytogenetics) are shown in Table 3.

Multivariate analysis. We evaluated several prognostic variables for CR, DFS, and OS in a multivariate Cox regression model (Table 3). ABC protein activity was analyzed as a continuous variable. We included all significant variables in univariate analysis in the models. Cytogenetics (P = 0.01; RR, 3.1), high Pgp activity (P = 0.02; RR, 2.89), high BCRP activity (P = 0.04; RR, 1.90), and high MRP3 activity (P = 0.04; RR, 2.1) had an independent adverse prognostic significance for CR. Cytogenetics (P = 0.02; RR, 3.1), high Pgp activity (P = 0.02; RR, 2.70), and high MRP3 activity (P = 0.03; RR, 2.12) had an independent adverse prognostic significance for duration of DFS. For duration of OS, the absence of CR (P = 0.001; RR, 5.1), cytogenetics (P = 0.001; RR, 4.01), high Pgp activity (P = 0.04; RR, 2.60), high MRP3 activity (P = 0.04; RR, 2.10), and high BCRP activity (P = 0.05; RR, 1.95) had an independent adverse prognostic significance.

Treatment outcome of patients with high activity of MRP3, BCRP, and P-Glycoprotein. In this analysis, ABC protein activity was analyzed as a dichotomized variable. Among 85 patients, 30 patients (35%) had high BCRP activities with a MFI of ≥150, 26 patients (31%) had high MRP3 activity with a MFI of ≥200, and 34 patients (40%) had high Pgp activity with a MFI of ≥150 (Fig. 2).

Patients with high BCRP activity had a poorer percentage of CR than other patients (50% versus 75%, respectively; P = 0.03), as well as for DFS (21 ± 11% versus 39 ± 11% at 4 years; median DFS, 375 versus 600 days; P = 0.12; Fig. 3A); and for OS (11 ± 7% versus 41 ± 9% at 4 years; median OS, 230 versus 760 days; P = 0.003; Fig. 3B).

Independently, patients with a high MRP3 activity also had a poorer percentage of CR than other patients (46% versus 75%; P = 0.01) and for DFS (21 ± 9% versus 35 ± 7% at 4 years; median DFS, 156 versus 622 days, respectively; P = 0.04; Fig. 3 C), as well as for OS (15 ± 6% versus 38 ± 9% at 4 years; median OS, 230 versus 790 days, respectively; P = 0.02; Fig. 3D).

Patients with a high Pgp activity have also a poorer percentage of CR than other patients (52% versus 74%; P = 0.04), and for DFS (21 ± 5% versus 42 ± 8% at 4 years, median 350 versus 704 days, respectively; P = 0.02; Fig. 3E), as well as for OS (20 ± 5% versus 39 ± 3% at 4 years, median 198 versus 452 days, respectively; P = 0.008; Fig. 3F).

The patients who had only one or no (12 and 27 patients, respectively) high-activity Pgp, MRP3, or BCRP functional proteins had a better prognosis than the patients who had two or three (32 and 14 patients, respectively) high-activity Pgp, MRP3, or BCRP functional proteins. CR levels were 79% (31 patients) versus 54% (25 patients), respectively (P = 0.02); DFS was 43% versus 21% at 4 years, median 800 versus 370 days, respectively (P = 0.01; Fig. 3G); OS was 57 ± 11% versus 11 ± 5% at 4 years, median not reached versus 220 days, respectively (P < 0.0001; Fig. 3H).

After discordant results during the past few years, several large studies confirmed the importance of functional Pgp in the clinical resistance of AML (2–8). However, the results of randomized trials, with the addition of potent Pgp inhibitors such as cyclosporine A, quinine, or PSC833 showed contradictory results. This gave revealed alternative or additional mechanisms of resistance (34–38). Other new ABC proteins, MRP2, MRP3, MRP5 BCRP, were described as potential chemotherapy resistance mechanisms in acute leukemia (17, 19, 20, 23, 24). However, the prognostic effect of these new ABC proteins is not well known.

In cell lines with different levels of resistance, we have analyzed several functional assays of these new ABC proteins. Some of those functional assays correlated with the protein expression. Thereafter, we selected these functional assays to assess the functionality of ABC proteins in AML samples.

In a cell line study, both mitoxantrone ± fumitremorgin C and mitoxantrone ± GG918 were correlated with BCRP expression. However, in AML samples, only the mitoxantrone ± fumitremorgin C assay could be correlated with BCRP expression as in some other studies (22, 39, 40), except for one study (41), because GG918 is also a good modulator of Pgp (39, 42). Therefore, we have selected the mitoxantrone ± fumitremorgin C assay rather than the mitoxantrone ± GG918 assay to assess BCRP function. In addition, BCRP substrate specificity could be modified by changes in its amino acid sequence. However, mitoxantrone is a substrate for both wild-type and mutant BCRP (16, 43, 44). Therefore, the modulation of mitoxantrone retention by the “BCRP-specific” modulator fumitremorgin C may serve as an assay for BCRP function for the cells expressing either wild-type or mutant protein. Despite numerous reports showing BCRP expression in AML, there is little evidence about the correlation between BCRP expression and any adverse clinical outcomes (17–19, 45). In our study, BCRP functionality was very strong in 35% of patients. BCRP function was not associated with clinical or biological characteristics. In this study, BCRP function was an independent prognostic factor, concordant with the mRNA expression results in childhood and adult AML (19, 20).

MRP3 activity was able to correlate with MRP3 expression only in mitoxantrone ± MK assay in patient samples. This activity was a prognostic factor as in other MRP3 mRNA expression studies in childhood AML and acute lymphoblastic leukemia (24, 25). MRP3 function was higher in patients with M5 AML (24), which expressed a nonfunctional Pgp (shown recently by us) but had a poor prognosis compared to other AMLs, except for AML with good cytogenetics (46). In another study, M5 and M4 AML also expressed low Pgp activity (5). Our study highlights the potential role of MRP3 in drug resistance in AML, especially in M5 AML.

The functions of BCRP, MRP3, and Pgp are independent prognosis factors. Previously, we have shown that the coexpression of Pgp and MRP1 was a poorer prognostic factor than the expression of only one of these proteins (9). In a recent study, childhood AML patients who expressed high levels of both MRP2 and MRP3 genes had a lower rate of survival than the patients who expressed only one of these two genes (24). In the same way, we have shown in this study that patients who coexpressed a high activity of two or three ABC proteins among Pgp, BCRP, and MRP3 proteins had a poorer prognosis than patients who expressed a high activity of only one or none of these proteins. Therefore, these studies emphasized the importance of the simultaneous activity of ABC proteins in drug resistance. Hence, the modulation of only one of these proteins in clinical trials may not be enough. Some modulators could inhibit two ABC proteins. For example, GG918 could inhibit Pgp and BCRP activity in patients who coexpressed these two proteins (41, 42). Apart from anthracycline, Ara-C is a drug administered in the treatment of AML. The transfected cells with the cDNA of Pgp are not more resistant to Ara-C than nontransfected cells. Nevertheless, the sensitivity of blast cells to Ara-C is not known in cells transfected with MRP3 or BCRP. However, in one study, in infant acute lymphoblastic leukemia, despite the observed correlation between BCRP mRNA expression and Ara-C resistance, in a subline which expresses a high level of BCRP, Ko143, a specific BCRP inhibitor, did not sensitize the leukemic cell to Ara-C (14).

MRP2 protein expression was found only in a few patients, which is contrary to other studies. It is possible that the different results are due to the different methods used. We have studied MRP2 protein expression by flow cytometry, but others have studied MRP2 mRNA expression (24, 47). MRP2 is expressed in normal lymphocytes (48), and even <10% of lymphocytes could increase the level of MRP2 in the blast samples (49). It is also possible that the affinity of our MRP2 antibody was not strong enough.

In conclusion, the ABC proteins, BCRP and MRP3, could be implicated in the resistance to chemotherapies in AML. Therefore, the modulation of BCRP and MRP3, as well as Pgp, could be essential.

We thank Dr. H. Komatani for generously providing the vector and BCRP PC13 cells.