Dr Liu Yin Department of Clinical Haematology, Manchester Royal Infirmary, Oxford Road, Manchester M13 9WL, UK.
The in vitro intracellular daunorubicin accumulation (IDA) of blast cells from 69 patients with newly diagnosed acute myeloid leukaemia (AML) was correlated with the expression and functional activity of the multidrug resistance (MDR) proteins, P-glycoprotein (Pgp), multidrug resistance-associated protein (MRP) and lung-resistance protein (LRP). An inverse and significant association was found between IDA and Pgp-related efflux activity (r = −0.31, P = 0.01) and also MRP (r = −0.25, P = 0.04) but not with LRP (r = −0.13, P = 0.28). Coexpression of the MDR proteins had an additive effect in further lowering of IDA levels, suggesting that the clinical MDR phenotype is dependent on the sum of multiple MDR factors available to the leukaemic cell. Thus, the median IDA of leukaemic cells without any MDR proteins was significantly higher than that of blasts carrying two MDR proteins (0.466 vs. 0.296, P = 0.046). Seven patients with no expression of Pgp, MRP and LRP still had low IDA levels, suggesting the presence of efflux MDR mechanisms other than those studied. The relation of IDA to clinical parameters known to be associated with poor prognosis, such as age, secondary AML, karyotype, peripheral blood blast and CD34 counts, was also studied, but no significance was found on multifactorial analysis. There was a non-significant trend for earlier relapse in patients with low IDA levels (leukaemia-free survival of 16.3 months compared with 21.1 months in patients with high IDA levels). Our data suggest that, while the IDA assay is a quick and relatively easy test for the combined efflux MDR phenotype, it is unable to detect other MDR mechanisms, such as LRP, which may be important to the clinical outcome of patients with AML.
Most induction of remission chemotherapy regimens for acute myeloid leukaemia (AML) include the anthracycline antibiotic, daunorubicin. The response to multiagent treatment will partly depend on a lethal level of daunorubicin being achieved within the leukaemic cells. While pharmacological factors influencing plasma daunorubicin levels have been largely taken into account by drug dosage and schedules, biological factors intrinsic to the leukaemic blast cell may be more important in determining intracellular daunorubicin levels.
Cancer cell lines that are chemoresistant often have much lower intracellular cytotoxic concentrations when exposed to daunorubicin than their chemosensitive parent cell lines ( Ramu et al, 1989 ; Sehested et al, 1989 ). These resistant lines usually exhibit a multidrug resistance (MDR) phenotype in which there is cross-resistance to structurally unrelated cytotoxic agents, probably as a result of high levels of transmembrane transporters, such as P-glycoprotein (Pgp) and multidrug-resistance protein (MRP), in the cytoplasmic membrane ( Pastan et al, 1988 ; Zaman et al, 1994 ). Lung resistance protein (LRP) is a major vault protein often overexpressed in MDR cell lines and may also be involved in cellular drug transport processes ( Scheffer et al, 1995 ).
There is increasing evidence that Pgp is a poor prognostic marker in AML and, to a lesser extent, LRP and MRP have also been reported to correlate with a poor treatment outcome ( Del Poeta et al, 1996 ; Borg et al, 1998 ; Legrand et al, 1998a ). These drug resistance proteins often coexist in the same cell and may all contribute to lower the intracellular daunorubicin concentration ( Slapak et al, 1994 ; Zhou et al, 1996). Daunorubicin levels can be measured conveniently by direct cellular fluorescence, and this method could usefully be used as a surrogate test for the multifactorial clinical resistance phenotype ( Pallis & Russell, 1998).
To determine the relationship between the intracellular daunorubicin concentration and degree of expression of the drug resistance proteins, Pgp, MRP and LRP, in acute myeloid leukaemic cells, we used a flow cytometric technique to measure intracellular daunorubicin fluorescence and immunofluorescent and/or immunocytochemical methods for the drug-resistant proteins. We found that increased expression of Pgp and MRP but not LRP was associated with lower intracellular daunorubicin levels.
MATERIALS AND METHODS
Sixty-nine patients with newly diagnosed AML (median age 53 years; range 16.4–73.5 years; 38 males and 31 females) were treated according to established trial chemotherapy protocols. Forty-three patients received MRC (UK) AML 10 induction therapy DAT 3+10 (daunorubicin 50 mg/m2 × three doses, cytosine arabinoside 100 mg/m2 × 20 doses, 6-thioguanine 100 mg/m2 × 20 doses); 14 patients had Amgen CSF-91134 DAV 3+7+5 (daunorubicin 45 mg/m2 × three doses, cytosine arabinoside 100 mg/m2 × 14 doses, etoposide 100 mg/m2 × five doses) and 12 patients were treated with an elderly AML protocol MAC 4+5 (mitozantrone 10 mg/m2 × four doses and cytosine arabinoside 100 mg/m2 × 10 doses). Consolidation therapy consisted of DAT 3+8 or DAV 2+5+5 followed by MACE (m-amsacrine 100 mg/m2 × five doses, cytosine arabinoside 200 mg/m2 × five doses and VP-16 100 mg/m2 × five doses) and MidAC (mitozantrone 10 mg/m2 × five doses and cytosine arabinoside 1 g/m2 × six doses). Elderly patients received two pulses of MAC 2+4 as consolidation therapy. Eight patients underwent bone marrow transplantation (seven allogeneic, one autologous) as the final post-remission treatment.
Patients with fewer than 5% blasts in a bone marrow regenerating normal haemopoietic elements after one or two chemotherapy courses were considered to have achieved complete remission. All other patients were classed as failures of remission induction therapy.
AML was diagnosed on bone marrow smears routinely stained and evaluated according to the revised French–American–British (FAB) criteria. Bone marrow aspirates were collected before chemotherapy into medium containing preservative-free heparin. Mononuclear cells were isolated by Ficoll–Paque density gradient sedimentation. Examination of cytospin preparations showed that the samples contained more than 95% blast cells. The samples were resuspended in medium with 60% fetal calf serum (FCS) and 10% dimethyl sulphoxide (DMSO) and stored in liquid nitrogen after programmed cell freezing. For analysis, the frozen samples were thawed in a waterbath at 37°C and immediately diluted in 10 ml of RPMI 1640 with 40% FCS, prewarmed to 37°C. After 60 min of incuation, the cells were centrifuged and resuspended in RPMI with 10% FCS and then gassed with 5% CO2. These cells had a median viability of 80%, as shown by fluorescein diacetate/propidium iodide tests. Previous work had shown frozen samples to be comparable to fresh ones.
Cell lines were prepared in a similar manner to the clinical samples and used as negative and positive controls of the LRP, MRP and Pgp assays. The human lung cancer cell lines COR-L23/P and COR-L23/R were used as the negative and positive MRP controls respectively. The Pgp-negative cell line K562 cl.6 and the Pgp-positive K/DAU100 line were used for control of Pgp assays. The K/DAU100 cell line was chosen because it expressed Pgp and efflux activity at a level similar to strongly Pgp-positive AML samples. The small lung cancer cell line SW1573 and its drug-resistant subline SW1573/2R120 were used as negative and positive controls for LRP. SW1573 was weakly positive for LRP and positive for MRP, but cross-reactivity was not observed between MRP- and LRP-specific antibodies. Besides cell line controls, irrelevant antibodies were also used to evaluate the endogenous staining and fluorescence of the blast cells.
Flow cytometric detection of intracellular daunorubicin accumulation and rhodamine/PSC 833 efflux
Cells (0.5 × 106) were incubated for 90 min at 37°C with 2 m M daunorubicin (Sigma) in RPMI 1640 and 10% FCS. Immediately afterwards, the cells were washed in ice-cold Dulbecco's phosphate-buffered saline (PBS), and then cell fluorescence was measured. A Coulter Epics XL flow cytometer was used to analyse 5000 events of gated viable cells for each sample. The mean log fluorescence of intracellular daunorubicin accumulation was recorded.
In order to measure Pgp function, cells were loaded for 60 min at 37°C with 400 ng/ml Rhodamine 123 (Sigma) and, after being washed twice in PBS, were incubated in prewarmed drug-free RPMI with and without 3 m M PSC833 (Sandoz) for 90 min at 37°C. Centrifuging the cells and adding ice-cold PBS stopped further efflux. The relative decrease in the effluxing cells after adding the modulator was then calculated.
Flow cytometric detection of MRP and Pgp expression
Pgp expression of viable leukaemic cells was analysed by incubating them for 1 h at 4°C with the mouse monoclonal antibody MRK-16 (20 μg/ml; Immunotech) or an irrelevant mouse IgG2a antibody (5 μg/ml; Coulter). Cells were washed twice and incubated for 30 min in the dark with goat anti-mouse fluorescein isothiocyanate (FITC; Coulter).
In order to detect MRP, cells were permeabilized in Permaefix (Ortho) for 40 min, washed twice, and then incubated for 1 h at room temperature with the mouse monoclonal antibody MRPm6 (5 μg/ml; R.J. Scheper) or an isotypic mouse IgG1 control (5 μg/ml; Coulter). After two washes, antibody binding was detected by incubating the cells for 1 h with 5 μg/ml goat anti-mouse FITC (Coulter). The mean fluorescence ratio (MFR), defined as the mean fluorescence of MRPm6- or MRK-16-labelled cells divided by the mean fluorescence of the isotypic control antibody-labelled cells, was calculated. The percentage of MRK-16-positive cells relative to the control was also measured.
Immunocytochemical detection of LRP
Cytospins were fixed in ice-cold acetone and air dried before application of the LRP-specific monoclonal antibody, LRP56 (a kind gift from R. J. Scheper, Amsterdam, The Netherlands). A modified streptavidin–biotin complex technique was then used to produce a red-coloured precipitate as an end-point ( Borg et al, 1998 ). By adjusting incubation times and recycling steps, the sensitivity level for the assay was maximized to produce a full range of staining intensities on random AML patient samples before test sample processing. Besides the cell line controls, SW-1573 (LRP negative) and SW-1573/2R120 (LRP positive), irrelevant antibodies were also used to evaluate endogenous avidin activity. No staining was detectable with these antibodies. Slides were analysed by two independent observers blinded to clinical outcome, and the percentage of stained blast cells was measured.
Flow cytometric immunofluorescent detection of LRP was attempted using LRP56 monoclonal antibody and a method similar to that of MRP (see above). Sensitivity was low, in that the range of expression between LRP-positive and -negative controls was small. This technique was therefore abandoned in favour of immunocytochemistry.
The consensus recommendations were adhered to when reporting and analysing data ( Beck et al, 1996 ). Pearson correlation coefficient (r) and least-squares regression were used to test the relationship between continuous variables. Comparison of the distribution of a continuous variable between the different subgroups of a categorical variable was tested by a one-way analysis of variance (ANOVA) test. Samples were also dichotomized by arbitrarily considering an IDA of 0.4 and above as a low daunorubicin efflux sample and below 0.4 as a high efflux sample. Pgp expression, Pgp function and LRP were divided into positive and negative groups with 20% positivity as cut-off. MRP expression was considered positive if the MFR was 2.5 or more. Frequencies were then tested by chi-squared analysis. IDA, MRP, LRP and Pgp expression and function were treated as continuous variables and not as dichotomized categories to test their relationship to chemotherapy response and survival times. Response to chemotherapy was analysed using multivariate logistic regression models. A multivariate Cox survival function model was fitted to test the predictive value of these drug resistance factors for leukaemia-free survival (LFS) and overall survival (OS). A comparison of Kaplan–Meier survival curves was also performed with the log-rank test for IDA.
The median IDA of the study group was 0.395 ± 0.12 with a range of 0.107–1.41. The cut-off between high and low efflux groups was chosen at an IDA of 0.4, as 34 (49.3%) patients had a high efflux with an IDA of less than 0.399, whereas 35 (50.7%) had low efflux with an IDA of 0.4 or more (Fig 1).
LRP, MRP and Pgp
The median MRP fluorescence ratio of the study population was 1.23 with a range of 1–4.84. The median LRP was 19% positive cells with a range of 0–100%. There was a very strong correlation between MRK-16 MFR and the percentage of MRK-positive cells (P 0.0001) and, therefore, only the results of MRK-16 percentage-positive cells are reported in this study. The median number of cells expressing Pgp when using the MRK16 antibody was 10.1% positive cells (range 0–69.5%), while the median number detected by the rhodamine/PSC 833 Pgp efflux assay was 8.5% (range 0–39.6%).
Correlation of IDA with multidrug resistance proteins
Table 1. Table I. Pearson correlation coefficient (r) for intracellular daunorubicin accumulation (IDA) and multidrug resistance (MDR) proteins and other clinical parameters.
Table 2. Table II . Relationship between the multidrug resistance (MDR) proteins, LRP, MRP and Pgp and intracellular daunorubicin accumulation (IDA). F-ratio = 3.218, P = 0.046. As there were only two cases in which all the MDR proteins were positive, they were included with the group positive for two MDR proteins for statistical analysis.
Correlation of IDA and drug resistance proteeins to treatment outcome
Of the 69 patients who received remission induction chemotherapy, 49 (71%) achieved complete remission (CR). Patients entering remission had a mean IDA of 0.42 compared with an IDA of 0.35 of patients in failure (P = 0.31). The CR rate of the low daunorubicin efflux group was 77% (27/35) compared with 65% (22/34) for the high efflux group (P = 0.254). The LFS of the low efflux group was 21.1 months compared with 16.3 months for the high efflux group (P = 0.669). The low daunorubicin efflux group had an OS of 17.8 months compared with 13.1 months for the high-efflux group (P = 0.167; Fig 2).
Multivariate analysis showed that both LRP and Pgp function, but not MRP or IDA, were predictive of response to chemotherapy and survival ( 3 Table III).
Table 3. Table III . Relationship of drug resistance proteins and intracellular daunorubicin accumulation to response to induction chemotherapy and leukaemia-free survival (LFS) and overall survival (OS). The drug resistance factors were all entered as covariates in a Cox regression model.
We investigated the in vitro intracellular daunorubicin accumulation (IDA) of AML blasts by flow cytometry to study its relationship to the MDR proteins, MRP, Pgp and LRP and to clinical drug resistance. We found that IDA is influenced by Pgp and, to a lesser extent, by MRP but not LRP. Other clinical markers, such as CD34, peripheral blood blast count, age and poor risk karyotype, had no effect on IDA levels. Despite a significant correlation with Pgp function and MRP, IDA was not predictive of response to chemotherapy.
There was a wide range of IDA levels in the leukaemic samples, suggesting large differences in the activity and/or number of cellular drug efflux mechanisms between patients. Pgp functional performance correlated better than Pgp expression with IDA. This discordance arose because of patients with no or very low Pgp efflux, despite overexpressing the protein, and also others with high efflux activity but low or no Pgp expression. Similar findings have been reported previously ( Lamy et al, 1995 ; Leith et al, 1995 ).
MRP staining in AML cells has been reported to be primarily intracytoplasmic ( Nooter et al, 1996 ). In some cell lines, MRP has also been found to be predominantly in the endoplasmic reticulum ( Krishnamachary & Center, 1993), while in others, including SW1573 and MRP-transfected HeLa cells, the protein was predominantly expressed on the plasma membrane ( Zaman et al, 1994 ). Our study showed that the presence of MRP in the leukaemic cell depresses IDA levels, suggesting that a significant proportion of MRP is associated with the plasma membrane. These results are not necessarily contradictory, as the antibody used by Nooter et al (1996 ) is specific to the MRP1 homologue, while the MRPm6 antibody used in our study detects all five homologues of MRP ( Legrand et al, 1999 ). Also, no significant correlation was found between the fluorescence indices of MRPm6 and MRPr1 ( den Boer et al, 1998 ). Only five of the 69 patients studied overexpressed the protein, and MRP would therefore play a role in drug resistance only in a minority of patients with AML. This observation was supported by previous work, which showed that patients with high expression of MRP tended to be refractory to chemotherapy but, statistically, this relationship was lost on multivariate analysis ( Borg et al, 1998 ).
LRP has been shown to correlate with clinical drug resistance and a poorer outcome in AML in some ( List et al, 1996 ; Hart et al, 1997 ; Filipitis et al, 1998; Borg et al, 1998 ; Legrand et al, 1998b ), although not all ( Leith et al, 1999 ) studies. A non-significant trend for lower IDA levels with increasing LRP expression was observed in our study. This may possibly reflect an LRP-related process that does not pump out daunorubicin fast enough to be detected within the duration of the test in use. Alternatively, it may be caused by nucleo-cytoplasmic transport and cytoplasmic sequestration of cytotoxics, as in the current hypothesis for LRP function based on the cellular distribution of LRP ( Scheffer et al, 1995 ). Broxterman et al (1999 ) have also reported that there was no correlation between daunorubicin levels and LRP, unlike Michieli et al (1996 ), who found a significant association in their study. These discrepant findings between the various studies regarding IDA and LRP may well be the result of differences in techniques for LRP detection, as immunocytochemistry was used in Broxterman's studies and ours, whereas flow cytometric immunofluorescence was utilized by Michieli. There is, however, no consensus at present as to the most appropriate method for detecting LRP, as no functional assay has yet been established.
Co-expression of drug resistance proteins had an additive effect of further lowering IDA levels consistent with a clinical MDR phenotype dependent on the interaction of multiple mechanisms available to the leukaemic cell. MDR mechanisms, other than the ones studied, are also certainly contributing to daunorubicin efflux. Indeed, 20% of patients with low daunorubicin levels had no expression of Pgp, MRP and LRP. Furthermore, the coefficients of correlation between IDA and Pgp and MRP were only −0.31 and −0.25, respectively, albeit still significant. In this respect, a novel energy-dependent mechanism reducing IDA has been described recently ( Hedley et al, 1997 ).
Paradoxically, we did not find an association between IDA and treatment outcome, although there was a trend for earlier relapse and shorter survival in patients with low IDA, but this could be caused by the relatively small number of patients in our study. Other studies have also not found a correlation between IDA and outcome ( Campos et al, 1992 ; Kessel et al, 1984 ; Kokenberg et al, 1988 ; Pallis et al, 1999 ), although Guerci et al (1995 ) did show a positive association. In this study, both LRP and Pgp function, but not MRP, were predictive of response to chemotherapy and overall survival. Moreover, it is clear from work on MDR cell lines that, besides plasma membrane efflux, processes involving nuclear transport and cytoplasmic sequestration may also play a significant role in drug resistance ( Keizer et al, 1989 ; Lautier et al, 1997 ). While tests for IDA can be used to measure the combined functional activity of MRP and Pgp, they will miss these other processes, such as LRP, as IDA represents the sum of nuclear and cytoplasmic daunorubicin. Functional assays of nuclear drug accumulation may therefore reflect clinical MDR better than IDA, and there is a need for their development in the future.
Dr A. G. Borg was a Leukaemia Research Fund Training Fellow.