High multidrug resistance protein activity in acute myeloid leukaemias is associated with poor response to chemotherapy and reduced patient survival


O. Fardel, INSERM U456, Faculté de Pharmacie, 2 Avenue du Pr L. Bernard, 35043, Rennes, France. E-mail: olivier.fardel@univ-rennes1.fr


Summary.  Multidrug resistance protein (MRP) activity was investigated in 44 newly diagnosed acute myeloid leukaemia (AML) patients using a functional assay based on efflux of carboxy-2′,7′-dichlorofluorescein, an anionic dye handled by both MRP1 and MRP2. Elevated MRP transport was detected in 29% of cases, but was not significantly correlated with sex, age, white blood cell count at diagnosis or karyotype. In contrast, it was associated with secondary AML (P = 0·002), CD34 positivity (P = 0·041) and P-glycoprotein activity (P = 0·01). There was a lower rate of complete remission in MRP-positive patients versus MRP-negative patients (23% versus 81%; P = 0·001); overall survival was also better for MRP-negative patients (P = 0·004). These data indicate a probable role for MRP activity in the clinical outcome of AML.

Low response to chemotherapeutic agents is a major cause of failure in the treatment of malignancies. Several mechanisms of cellular resistance have been incriminated. A major one results from reduced intracellular accumulation of structurally unrelated anticancer drugs and usually involves overexpression of ATP-binding cassette membrane proteins acting as drug efflux pumps (Ross, 2000). The first identified of these transporters is P-glycoprotein (P-gp) handling anthracyclines, vinca alkaloids and epipodophyllotoxins. More recently, export pumps belonging to the multidrug resistance protein (MRP) subfamily, such as MRP1 and MRP2, have also been demonstrated to confer multidrug resistance on tumoral cells (Borst et al, 2000). In contrast, other members of the MRP subfamily, such as MRP4, MRP5 and MRP6, are not thought to be involved in multidrug resistance, whereas MRP3 only confers a limited resistance phenotype restricted to epipodophyllotoxins and methotrexate (Borst et al, 2000).

Elevated expression of P-gp has been reported in acute myeloid leukaemia (AML) and has been shown to be associated with poor treatment response in terms of obtaining complete remission, indicating that P-gp is probably a predictive factor for the outcome of AML (Ross, 2000). In contrast, the contribution of MRP transporters to clinical resistance in AML is still under discussion. Indeed, MRP1 expression has been reported to have no impact on treatment outcome in some studies (Filipits et al, 1997; Leith et al, 1999), whereas others have demonstrated that MRP activity may be a prognostic factor (Legrand et al, 1998). In addition, although MRP2 gene expression has been described in AML cells (Van der Kolk et al, 1998), the putative contribution of this transporter sharing numerous substrates with MRP1, including anticancer drugs, to clinical drug resistance has not yet been investigated. Additional studies are therefore required to determine the exact role of MRPs in AML. In the present work, we have used a functional cytometric assay allowing concomitant detection of both MRP1- and MRP2-mediated transport, in order to investigate MRP activity in 44 cases of AML. After correlation with clinical outcome, our data indicate that high MRP activity may predict a poor response to chemotherapy and a reduced overall survival of patients in AML.

Patients and methods

Patients.  Bone marrow aspirates or peripheral blood from 44 AML patients were tested at diagnosis; 30 cases were de novo AML, whereas 14 cases were secondary AML (10 arose from myelodysplastic syndromes and four followed chemotherapy for solid tumours). Twenty-four patients were female, and 20 were male. The diagnosis was made using French–American–British (FAB) criteria; one AML was classified as M0, 20 as M1, 12 as M2, three as M3, four as M4, three as M5 and one remained FAB unclassified. The median age was 47 years (18–70 years), and the mean of white blood cell count at diagnosis was 52·97 × 109/l (0·5–240 × 109/l). Seven patients displayed favourable karyotypes defined as inversion of chromosome 16 (inv 16) or translocation t(8; 21) or t(15; 17), whereas 35 had normal or unfavourable karyotypes; cytogenetic analysis was not available for two patients. Twenty-nine cases were considered CD34 positive, as revealed by immunophenotyping, i.e. > 20% of the leukaemic cells were stained with CD34 antibody. Twelve patients were found to be P-gp positive according to the rhodamine 123 efflux assay performed as described previously (Laupeze et al, 2001), whereas the others were P-gp negative. All patients were treated with conventional chemotherapy including cytosine arabinoside and anthracycline and were evaluable for complete remission defined as disappearance of clinical signs, normocellular blood and percentage of blastic cells < 5% of total nuclear cell count in bone marrow smears. Patients entering complete remission received two cures of consolidation according to the French GOELAM (Groupe Ouest d'Etudes des Leucemies Aigues Myeloblastiques) protocol. Eleven patients underwent allogeneic bone marrow transplantation as the final post-remission treatment. Patients were included in the study until December 1999.

MRP-related transport activity.  MRP activity was investigated using the carboxy-2′,7′-dichlorofluorescein (CF) efflux assay (Laupeze et al, 2001). Cells were first loaded at 37°C with 2 µmol/l CF diacetate for 30 min in the presence of the MRP1 and MRP2 blocker probenecid used at 2·5 mmol/l. CF diacetate, a non-polar non-fluorescent esterified form of CF, diffuses freely into cells, where it is cleaved by intracellular esterases, resulting in fluorescent CF, handled by both MRP1 and MRP2 but not by P-gp (Payen et al, 2000; Laupeze et al, 2001). After washing, cells were reincubated at 37°C in phosphate-buffered saline (PBS) for 120 min in the absence or presence of 2·5 mmol/l probenecid. Intracellularly retained CF fluorescence, depending on MRP-mediated efflux of the dye, was then determined using a FACSCalibur flow cytometer (Beckton Dickinson, San Jose, CA, USA). This cytometric analysis was performed routinely in the blast region according to granularity and size parameters; in some cases, leukaemic cells were also identified in dual fluorescence experiments according to their expression of a specific marker, usually CD34. Results were expressed as the ratio of dye retention value in the presence of probenecid versus CF staining in the absence of probenecid as reported previously (Laupeze et al, 2001); this ratio was called probenecid modulating factor (pmf).

Statistical analysis.  The association between categorical variables was analysed using the chi-square test or the Yates corrected chi-square test. The link between categorical variables and continuous values was investigated by the Mann–Whitney U-test. Multivariate analysis of prognostic factors for the achievement of complete remission was performed using the logistic regression model. Overall survival was measured from diagnosis until death from any cause, with observations censored for patients last known alive; rates were estimated by the Kaplan–Meier method and compared by the log-rank test. The Cox proportional model was used for multivariate analysis of overall survival. Significance was defined as P < 0·05.


Determination of MRP activity in blast cells from 44 AML patients using the CF efflux assay indicated that the mean pmf value was 1·51 + 0·44 (range 0·89–2·96). An example of an AML exhibiting high MRP activity, i.e. elevated pmf, is given in Fig 1. For six cases (14% of the total number of AML cases studied), pmf was lower than 1·1, indicating an absence of detectable MRP activity. Twenty-five patients (57%) had a pmf between 1·1 and 1·65; this pmf range has also been reported previously in normal peripheral blood mature cells (Laupeze et al, 2001) and most probably corresponds to basal MRP activity. Only 13 patients (29%) were found to display a pmf > 1·65, unequivocally reflecting an elevated MRP activity. To establish a clear threshold for MRP positivity, in order to search for correlations between MRP activity status and clinical and biological parameters and treatment outcome, the pmf value of 1·65 was thereafter retained for distinguishing MRP-positive patients (pmf > 1·65) and patients with no or only basal MRP activity (pmf < 1·65) that we considered as MRP-negative patients. The pmf value means for MRP-positive patients (n = 13) and MRP-negative patients (n = 31) were 2·03 + 0·42 and 1·29 + 0·19 respectively.

Figure 1.

Flow cytometric graph of carboxy-2′,7′-dichlorofluorescein (CF) efflux assay in an AML sample exhibiting high MRP activity. CD34+ leukaemic cells were stained with 2 µmol/l CF diacetate, washed and reincubated in dye-free medium in the absence or presence of 2·5 mmol/l probenecid for 120 min. Flow cytometric analysis was then performed to detect blast cells displaying MRP activity, i.e. exhibiting probenecid-inhibitable loss of dye fluorescence. Results are expressed as probenecid modulating factor (pmf), i.e. the ratio of dye retention value in the presence of probenecid versus CF retention in the absence of probenecid.

MRP activity was not found to be associated with sex, age or white blood cell count. The proportion of MRP-positive patients was higher for CD34+ patients (41%) than for CD34 patients (7%) (P = 0·041). P-gp activity was also found to be associated with CD34 expression; indeed, 62% of CD34+ patients were P-gp positive versus 7% of CD34 patients (P = 0·001). Patients with favourable karyotypes were all MRP negative, whereas 32% of patients with normal or unfavourable karyotypes were MRP positive; this difference was not significant (P = 0·21). MRP positivity was significantly linked to the diagnosis of secondary AML, as the proportion of MRP-positive cases was 13% and 64% for de novo and secondary AML respectively (P = 0·002); similarly, P-gp positivity was associated with secondary AML (P = 0·024). MRP activity was found to be linked to P-gp activity; indeed, 53% of P-gp-positive patients were also positive for MRP, whereas only 14% of P-gp-negative patients were MRP positive (P = 0·01).

Twenty-eight patients (64%) entered complete remission. Values of pmf according to the achievement of complete remission are summarized in Fig 2A. Means of pmf values in patients who achieved or did not achieve complete remission were 1·35 + 0·28 and 1·79 + 0·54 respectively (P = 0·003). In addition, MRP-positive patients defined as described above (pmf > 1·65) displayed a lower rate of complete remission than MRP-negative patients (pmf < 1·65) (23% versus 81%, P = 0·001). CD34 positivity (P = 0·009) and P-gp activity (P = 0·023) were also found to be significantly associated with a low rate of complete remission in univariate analysis; in contrast, the diagnosis of secondary AML, the presence of normal or unfavourable cytogenetic abnormalities, age or white blood count at diagnosis did not retain significance, probably because of the relatively small number of patients entering our study. In multivariate analysis, MRP status, unlike CD34 status or P-gp positivity, was still associated with the achievement of complete remission (P = 0·016). Overall survival was finally investigated in MRP-positive and MRP-negative patients using Kaplan–Meier analysis (Fig 2B). MRP-negative patients were found to display a better overall survival than MRP-positive patients (P = 0·004). Besides MRP status, the diagnosis of secondary AML (P = 0·009), a favourable karyotype (P = 0·018) and the achievement of complete remission (P < 0·001), but not CD34 status or P-gp positivity, were identified as parameters influencing overall survival in univariate analysis; however, only the achievement of complete remission retained significance (P = 0·015) in multivariate analysis.

Figure 2.

(A) MRP activity in patients according to the achievement of complete remission (CR). MRP activity is expressed as probenecid modulating factor (pmf), i.e. the fold increase in dye retention in the presence of the MRP blocker probenecid. (B) Overall survival of MRP-positive and MRP-negative patients estimated according to the Kaplan–Meier method.


The exact contribution of MRPs to clinical drug resistance in AML remains controversial. This may result from the variability in the analytical methods used for MRP detection in clinical samples and discrepancies in the thresholds adopted for positivity. In addition, studies have been aimed primarily at investigating correlation between MRP1 expression and treatment outcome, and MRP2 relevance has not been evaluated, although this MRP pump shares numerous substrates with MRP1, including anticancer drugs (Borst et al, 2000), and has recently been shown to be present in myeloid leukaemia cells (Van der Kolk et al, 1998). In the present study, we have used a functional assay based on cellular retention of the fluorescent anionic dye CF in order to analyse MRP relevance in 44 cases of AML. This efflux assay was used because: (i) it belongs to the functional methods that have previously been shown to allow sensitive detection of multidrug resistance phenomenon in leukaemic patients (Pall et al, 1997); (ii) it does not detect inactive drug transporters that may be expressed by some leukaemic cells, whereas other methods, including immunolabelling assays, do; (iii) the anionic dye retained, CF, is handled by both MRP1 and MRP2 (Payen et al, 2000), thereby enabling the activity of these two structurally and functionally related transporters to be investigated concomitantly; (iv) CF, even under its esterified form CFDA, is not transported by P-gp (Laupeze et al, 2001), enabling discrimination between MRP and P-gp activity; and (v) fluorescent substrates specifically recognized by each member of the MRP subfamily, especially MRP1 and MRP2, have not yet been described, precluding the use of a flow cytometric functional assay that may discriminate between the activities of these MRP export pumps. We have found that 29% of AML patients displayed elevated MRP activity, whereas the other patients (71%) had no or only basal MRP activity; this percentage of MRP-positive patients was similar to those reported previously in various studies. Indeed, Van der Kolk et al (1998) and Filipits et al (1997) showed that 33% and 30% of AML cases exhibited MRP activity or MRP1 expression respectively; similarly, Legrand et al, 1998) showed that 34% and 27% of AML patients were positive for MRP1 expression and MRP1 function, whereas Borg et al (1998) demonstrated that 21% of AML patients displayed increased MRP levels. Only Leith et al (1999) have reported a lower rate (10%) of AML cases exhibiting MRP1 expression.

MRP activity status was not found to be associated with sex, age, white blood cell count or unfavourable karyotypes; such data agree with previous reports (Borg et al, 1998). In contrast, MRP positivity, like P-gp activity, was shown to correlate with CD34 expression; this probably reflects the elevated activity of P-gp and MRP occurring in normal CD34+ haematopoietic stem cells (Legrand et al, 1998; Laupeze et al, 2001). Interestingly, we also found a significant association between MRP activity and P-gp positivity, thus suggesting that co-expression of these resistance efflux pumps probably occurs in AML cases. MRP was also found to be associated with secondary AML; in a similar manner, P-gp correlated with this type of AML, in keeping with previous results (Ross, 2000). Such increased expression of multidrug transporters may account for the relatively poor clinical outcome of secondary AML compared with that of de novo AML.

The assessment of our data in respect of MRP activity with clinical outcome highlighted the strong correlation between MRP positivity, a low achievement of complete remission and poor overall survival. Patients who did not achieve complete remission were shown to have high pmf values, reflecting enhanced MRP-mediated efflux. Similarly, Legrand et al (1998) reported that AML patients exhibiting MRP activity displayed a lower rate of complete remission; overall survival was also reduced. However, this correlation between MRP status and response to treatment and survival was not found by others (Filipits et al, 1997; Borg et al, 1998; Leith et al, 1999). This discrepancy may result from differences in experimental protocols designed for MRP analysis in clinical samples as described above. In this context, it is notable that detection of MRP1 by immunolabelling or reverse transcription–polymerase chain reaction (RT–PCR) assays failed to show a correlation with treatment outcome, whereas concomitant analysis of MRP activity in the same leukaemic cell samples did (Legrand et al, 1998), thus demonstrating the need to retain functional tests for exploring multidrug resistance transporters. A similar conclusion also arose from studies investigating P-gp expression and activity (Pall et al, 1997). The fact that we have investigated the functional expression of both MRP1 and MRP2, whereas other studies have been primarily aimed at analysing MRP1 alone may also explain the discrepancy with respect to correlations with treatment outcome.

Beside MRP and P-gp activity, cellular efflux resulting from the breast cancer resistance protein (BCRP), also known as mitoxantrone resistance protein (MXR), may influence the clinical response to chemotherapy in AML. Indeed, this drug resistance protein has recently been shown to be expressed in about one-third of AML patients (Ross et al, 2000).

In conclusion, we have used a functional efflux assay to demonstrate that high MRP activity in AML correlated with poor achievement of complete remission and reduced overall survival. Such data outline the relevance of detecting MRP activity in AML and suggest that the clinical use of MRP inhibitors may improve the efficacy of chemotherapeutic drugs in such acute leukaemias.


Beatrice Laupeze is a recipient of a fellowship from the Association pour la Recherche sur le Cancer.