Quantitative determination of mdrl gene expression in leukaemic cells from patients with acute leukaemia

By using a quantitative RNA-RNA solution hybridisation method, the average number of mdrl RNA transcripts per cell was measured in total nucleic acid extracts of leukaemic cells from patients with acute leukaemia. The results in different types of leukaemia were (number of patients with detectable mdrl RNA/total number of patients; median number of transcripts per cell in samples with detectable mdrl RNA); de novo untreated acute myelocytic leukaemia (AML): 20/44; 0.7, secondary acute myelocytic leukaemia: 8/13; 1.1, acute lymphocytic (ALL) and undifferentiated leukaemia: 5/14; 0.6, relapsed leukaemia: 7/15; 0.7. Forty-six patients with de novo untreated acute leukaemia (AML: n = 34, ALL: n = 12) were evaluable for response to induction chemotherapy. Twelve of 18 patients (67%) with detectable mdrl RNA levels achieved complete remission compared to 23 of 28 (82%) with undetectable levels (P = 0.40). The remission duration tended to be longer among patients with undetectable mdrl RNA (P = 0.08). Leukaemic cells were analysed

Concomitant incubation with agents that prevent drug efflux, e.g. verapamil, cyclosporin A, and quinidine can restore intracellular drug concentration and overcome resistance in vitro (Tsuruo et al., 1983;Twentyman et al., 1990). Cell lines expressing the MDR phenotype may also have decreased sensibility to drugs that do not induce MDR like mitoxantrone and amsacrine (Taylor et al., 1991). cDNA clones encoding the gene for P-glycoprotein (mdrl) have been isolated (Gros et al., 1986;Van der Bliek et al., 1988), and transfection studies have confirmed that expression of the mdrl gene is sufficient to create the MDR phenotype (Ueda et al., 1987a). The mdrl gene is frequently expressed in certain tissues such as liver, large and small intestine, kidney, adrenal cortex and pancreas (Fojo et al., 1987;Thiebaut et al., 1987).
Analyses of human tumours have shown that malignant cells from organs that normally express mdrl RNA often express mdrl. Untreated neoplasms that occasionally express mdrl include acute leukaemia in adults, non-Hodgkin's lymphoma and neuroblastoma (Goldstein et al., 1989). An association between mdrl RNA expression in leukaemic cells and response to chemotherapy in acute myelocytic leukaemia (AML) has earlier been suggested (Sato et al., 1990;Pirker et al., 1991;Marie et al., 1991). In sarcoma of childhood and neuroblastoma, expression of P-glycoprotein determined with the monoclonal antibody C219 was found to be an important prognostic factor (Chan et al., 1991;. Another group using the same monoclonal antibody found that P-glycoprotein expression was restricted to normal cells in neuroblastoma biopsies (Favrot et al., 1991). In P-glycoprotein positive resistant multiple myeloma verapamil has been shown to increase the effect of vincristine, doxorubicin and prednisone chemotherapy in some but not all patients (Salmon et al., 1991).
To further delineate the clinical relevance of mdrl expression in leukaemic cells, we have quantified the mdrl RNA expression in 92 samples of peripheral blood leukaemic cells from 76 patients with untreated de novo or secondary acute leukaemia and relapsed acute leukaemia. Leukocytes from seven blood donors, cell line K562 and two vincristine resistant sublines, three human liver specimens and the hepatoma cell line HepG2 were also investigated.

Patients
Ninety-two samples of leukaemic cells from 76 patients were analysed. The study included 58 patients with de novo untreated acute leukaemia-44 with AML, 12 with acute lympocytic leukaemia (ALL), two with acute undifferentiated leukaemia (AUL), 15 patients with relapsed leukaemia (samples from ten had also been analysed at first presentation) and 13 with secondary AML (12 patients with AML evolving from a myelodysplastic syndrome and one patient who had previously received cytostatic treatment). Median ages of the different patient subgroups were: de novo AML 64 years (15-87), ALL and AUL 36 years (17-80), secondary AML 69 years (46-86). The median white blood cell count was 38 (4-544) x 109 1-1. In patients studied > 70% of the mononuclear cells in peripheral blood were leukaemic cells.
Mononuclear cells were isolated from peripheral blood by centrifugation on Lymphoprep® (Nycomed A/S, Oslo, Norway), frozen to -90°C in a programmed freezer in RPMI 1640 medium (Gibco, Life Technologies Ltd, Paisley, Scotland) supplemented with 1% L-glutamine, 50% human serum and 10% dimethylsulfoxide, and stored in liquid nitrogen.
The patients were treated between 1982 and 1991 and received combination chemotherapy according to different protocols. The majority of the patients with AML, in which an attempt to achieve complete remission (CR) was made, received daunorubicin-vincristine-ara-C, mitoxantrone-etoposide-ara-C, or daunorubicin-ara-C-thioguanine (Table I). Patients with ALL received daunorubicin-cyclophosphamidevincristine-prednisone-L-asparaginase. CR was defined as normal cellularity and less than 5% blasts in the bone marrow in combination with normal peripheral blood values.
Patients who did not enter remission following at least two induction courses were defined as having resistant disease. Patients who died during induction treatment or received palliative low dose cytostatic treatment were not considered evaluable for response to chemotherapy.
The study was approved by the local ethics committee.
Normal leukocytes, liver specimens and cell lines Mononuclear cells from seven healthy blood donors were separated from buffy coats by centrifugation on Lympho-prep®. Monocytes were separated from lymphocytes by adhesion to a plastic culture flask. Granulocytes were isolated from the buffy coats after removal of mononuclear cells. The remaining pellets (approximately 15 ml, containing red blood cells, platelets and granulocytes) were mixed with 10 ml of human plasma, 4 ml of 4.5% dextran (T250, Pharmacia, Uppsala, Sweden) containing 1000 IU of heparin, and allowed to sediment for 1.5 h at 4°C. The supernatant containing the granulocytes was collected. The isolated cells were frozen as pellets, containing approximately 60 x 106 cells, in liquid nitrogen and kept at -90°C.
The cell line HepG2 and three specimens of human normal liver, obtained during surgery (liver resection because of metastatic cancer) were also analysed. The specimens were immediately after removal frozen in liquid nitrogen and then kept at -90°C.
As controls, cell line K562 and the vincristine resistant sublines K562/Vcr3O and K562/Vcrl 50 were analysed. The resistant cell lines were developed by growing K562 cells in medium with increasing concentrations of vincristine. K562/ Vcr3O and K562/Vcrl 50 were maintained in medium with vincristine concentrations 30 and 150 nM, respectively, and expressed MDR characteristics (not shown).

RNA analysis
Nucleic acids extracts were prepared from approximately 60 x 106 cells or in case of liver specimens approximately 100mg of tissue (Durnam & Palmiter, 1983). The samples were lysed in 4 ml of 1 x SET (1% SDS, 20 mM Tris-HCI, pH 7.5, and 10 mM EDTA), homogenised in a Polytron® (Kinematica, Kriens, Switzerland), and then treated with proteinase K 200 iLg ml-l (Merck, Darmstadt, Germany) for 45 min at 45°C. After extraction with phenol/chloroform and precipitation with ethanol the precipitate was dissolved in Of the five patients where no remission duration is stated, one refused further treatment, three died from other causes than leukaemia and one underwent an allogeneic bone marrow transplantation. For explanation of abbreviations of cytostatic drugs, see Table V. 0.2 x SET and aliquots were taken for determination of DNA concentration by Hoechst fluorometry (Labarca & Paigen, 1980). Plasmid pGEM4 (Promega Corporation, Madison, WI, USA) carrying the 1383 basepair mdrl cDNA sequence 5A, was kindly provided by M.M. Gottesman and I. Pastan, NCI, USA (Ueda et al., 1987b). A 403 nucleotides long antisense probe (with 393 nucleotides transcribed from mdrl cDNA, position 2561-2168) was generated by in vitro transcription of Stul (New England Biolabs, Inc, Beverly, USA) cleaved 5A with SP6 RNA polymerase (Promega Corporation) in the presence of 35S-UTP (>37 x 106 MBq mmol-', Amersham, England). Non incorporated nucleotides were separated from the transcript on a Sephadex G-50 column (Nickcolumn, Pharmacia). Polyacrylamide gel electrophoresis showed that >90% of the transcripts were of the expected size. Two types of sense complementary to the labelled antisense were used. One 1732 nucleotides long unlabelled sense RNA was transcribed by T7 RNA polymerase (Promega Corporation) from 5A after linearisation with NaeI (New England Biolabs, Inc). A shorter sense, 439 nucleotides long, was transcribed by SP6 RNA polymerase from EcoR I (New England Biolabs, Inc) cleaved plasmid pGem-3Zf( + ) (Promega Corporation), into which the 393 basepair sequence, which constitutes the labelled antisense probe had been subcloned. After Sephadex G-50 chromatography, the RNA containing fraction was ethanol precipitated, dissolved in 0.2 x SET and the RNA concentration determined spectrophotometrically at 260 nm.
Aliquots of the nucleic acid extracts, or unlabelled sense RNA, were adjusted to 20 p1l with 0.2 x SET and mixed with 30 x 103 counts per min (c.p.m.) of antisense probe, dissolved in 20 fIl hybridisation solution (0.6 M NaCl, 4 mM EDTA, 20 mM Tris-HCI, pH 7.5, 7.5 mM DTT, and 25% deionised recrystallised formamide), and incubated for 18 h at 68°C. Subsequently, the samples were treated with 1 ml of 40 tLg ml-RNase A (Sigma Chemical Company, St Louis, MO, U.S.A.), 2 fg ml' RNase TI (Sigma), 100 jig ml-' salmon sperm DNA (Sigma), 0.3 M NaCl, 2 mM EDTA, and 10 mM Tris-HCI, (pH 7.5) and incubated for 45 min at 37°C. After addition of 100 glI trichloroacetic acid, the samples were kept on ice for 30 min, and the RNase resistant precipitates were collected on Whatman GF/C filters (Whatman International Ltd, Maidstone, England). After addition of 4 ml scintillation liquid (Insta-gel; Packard Instrument Company, Downers Grove, IL, USA), the radioactivity was determined in a liquid scintillation counter (Packard). The background radioactivity, measured in RNase treated samples containing the same amount of probe but no sense RNA or extract, was below 0.5% of the input value. Samples classified as positive for mdrl RNA showed at least twice the background radioactivity and a proportional increase in radioactivity with increasing amounts of added extract (Figures la and b). The quantities of mdrl RNA in the extracts were determined by comparison with a standard curve, generated by hybridisations with increasing amounts of sense RNA (Steen et al., 1990) (Figure lc). The average number of mdrl RNA transcripts per cell could be calculated based on the standard curve, the molecular weight of sense RNA, c.p.m. per jig of DNA in the positive extracts, Avogadros number (6 x 1023 molecules in a mole) and the assumption of a DNA content of 6 pg/cell. The limit of detection in extracts containing 30 iLg of DNA was 7.5 x I05 transcripts of mdrl RNA, corresponding to 0.15 RNA copies per cell. Cell extract corresponding to at least 30 gig of DNA was analysed in all samples classified as having undetectable levels of mdrl RNA.
Recovery of RNA and DNA The recovery of RNA was measured by the addition of 135 x 103 c.p.m. of labelled RNA transcribed with SP6 RNA polymerase from plasmid pGEM 5Zf (+) (Promega Corporation) to cell samples lysed in 1 x SET. Thereafter the samples were treated with proteinase K, extracted with phenol/chloroform, precipitated with ethanol dissolved in 0.2 x SET, treated with trichloroacetic acid, the remaining radioactivity collected on a GF/C filter and counted in a liquid scintillation counter. The recovery of DNA was calculated by comparison of DNA concentration in cell samples before and after the extraction and precipitation procedure. The recovery of RNA was between 70-80% and of DNA 80-90%, respectively. n I v Specificity and reproducibility of the method The specificity of the method was tested by performing an RNase protection assay with 50 pg of 439 nucleotides long sense, RNA extracts from K562, its vincristine resistant subline K562/Vcrl 50, HepG2 and leukaemic cells from two patients with AML (one with 3.9 mdrl RNA transcripts per cell and one with undetectable levels of mdrl RNA). RNA was extracted from approximately 20 x 106 cells (Andersson et al., 1992) and hybridised with 50 x 103 c.p.m. of labelled sense under conditions identical to those described above. The intactness of the extracted RNA was confirmed by agarose gel electrophoresis. After RNase treatment, extraction with phenol/chloroform and ethanol precipitation, RNA was size separated by electrophoresis through a denaturating 5% polyacrylamide gel. Subsequently the gel was dried and the hybridised probe detected by autoradiography for 12 h. An RNA ladder was used as marker of length in nucleotides. The assay revealed one major protected fragment of the expected size in the lanes of K562/Vcrl50, HepG2, the leukaemic cells with detectable mdrl RNA and sense RNA.
No probe could be seen in the lanes of K562 and the leukaemic cells with undetectable mdrl RNA (Figure 2). The human mdr2 gene (also called mdr3), which is expressed in liver and in some B cell lymphatic leukaemias but not in myelocytic leukaemias (Herweijer et al., 1990), shows a high DNA sequence homology with the mdrl gene (Chin et al., 1989). However, in the region encompassed by the labelled antisense probe, the homology is only 69% and the longest homologous sequence consists of 14 nucleotides, preventing cross-hybridisation under the used stringent conditions. This is also shown by the single protected fragment in the RNase protection assay of HepG2 cells, which expresses both the mdrl and the mdr3 gene ( Van der Bliek et al., 1987).
To check the reproducibility of the method 19 samples with detectable mdrl RNA and nine samples with undetectable levels were extracted and hybridised on a second occasion. All samples with undetectable levels at first analysis also had undetectable mdrl RNA levels at the second analysis. The coefficient of variation of calculated mdrl RNA transcripts per cell among the 19 samples with detectable mdrl RNA levels was 25% (Figure 3).
The coefficient of variation using the two different sense RNAs for quantification was 4%.

Statistical analyses
The relationship between mdrl expression and age was analysed using the Mann Whitney U test. 2 x 2 tables were analysed with Fisher's exact test. Remission duration was analysed using the method of Kaplan and Meier and comparisons made by the log rank test (Peto et al., 1977).

Results
Leukocytes from blood donors The expression of mdrl RNA in lymphocytes, monocytes, and granulocytes from seven blood donors is shown in for response to induction treatment. The relationship between mdrl RNA expression and achievement of CR is shown in Table IV. If patients with AML and ALL are analysed together 23 of 28 patients with leukaemic cells without detectable mdrl RNA entered CR compared to 12 of 18 with leukaemic cells where mdrl RNA could be detected (P= 0.40). Most patients had a short remission duration (less than one year). The remission duration tended to be longer among patients whose leukaemic cells had undetectable levels of mdrl RNA but the difference was not significant, (P = 0.08) when patients with AML and ALL are analysed together, P = 0.15 when only patients with AML are included) (Figure 4).
Among the 24 patients with AML who achieved a CR there was no difference in the number of chemotherapy courses needed to achieve CR between patients with detectable mdrl RNA levels and patients with undetectable levels.
The 20 patients, with de novo AML, with detectable mdrl RNA levels in their leukaemic cells had a median age of 74 years (37-87) as compared to 54 years (15-78) for the 24 AML patients without detactable mdrl RNA levels (P <0.02). Considering all 76 patients (two patients who changed their expression are analysed twice) the median age of the 36 patients with detectable mdrl RNA levels was 66 years (17-87) compared to 54 years (15-81) for the 42 patients with undetectable levels (P<0.02). The median age of the 24 AML patients who achieved CR was 53 years, as  compared to 64 years for the 10 patients with resistant disease (P = 0.48). No association was observed in AML patients between FAB subtype (Bennett et al., 1976) or peripheral white blood cell count and expression of the mdrl gene.
Two or more consecutive samples were analysed in 12 patients, ten at diagnosis and relapse, one during a resistant relapse and one primary resistant patient ( Table V). One of six patients (pt 6) who initially had undetectable levels of mdrl RNA increased the expression to 0.7 transcripts per cell after multiple chemotherapy. In three of six patients (pts 2, 7, 8) who initially had detectable mdrl RNA in their cells there was a small increase, and in two (pts 1 and 4) there was a decrease of mdrl RNA expression.

Discussion
An accurate method for mdrl RNA quantification is important to clarify the relationship between mdrl expression in leukaemic cells and response to chemotherapy and also to evaluate the effect of chemotherapy on mdrl expression. We have employed a solution hybridisation technique, which allows reproducible quantification of less than one transcript of mdrl RNA per cell in a few million cells. Like other investigators we found that expression of the mdrl gene is quite common in untreated acute leukaemia (Sato et al., 1990;Noonan et al., 1990;Pirker et al., 1991;Marie et al., 1991). However, the level of expression is low, generally below one transcript per cell compared to 6-9 transcripts per cell in human liver. The level of expression in leukaemic cells and liver is in the same order of magnitude as that observed by Noonan et al. (1990) who used polymerase chain reaction to quantify mdrl RNA.
The interpretation of the clinical significance of mean levels of mdrl RNA is complicated by the difficulty in distinguishing between situations where many cells express low levels, or a small clone of leukaemic cells express high levels. These alternative situations may differently infl4ence the antileukaemic effect of chemotherapy. The impact of mdrl RNA levels on response to chemotherapy also depends on the association between mdrl RNA levels and amount of P-glycoprotein on the cell surface which needs to be clarified.
The biologicial significance of the low mdrl RNA expression as well as the demonstrated variation in mdrl RNA concentrations in the different leukocyte fractions from healthy individuals needs to be further investigated. Studies of drug efflux have indicated that the mdrl gene expression may vary in subpopulations of lymphocytes (Bines et al., 1991).
We found that a low mdrl RNA expression is common in tumour cells from patients with acute leukaemia but could not find any association between mdrl expression and liability to achieve a CR. The majority of patients with detectable and with undetectable levels of mdrl RNA had a remission duration of less than one year. However, all patients with a remission duration of more than 18 months had undetectable mdrl RNA in their leukaemic cells at diagnosis. Most likely expression of the mdrl gene influences response to chemotherapy in combination with several other mechanisms of resistance, for example non P-glycoprotein mediated multiple drug resistance, altered topoisomerase II activity or increased levels of glutathion-transferases (Danks et al., 1987;Harker et al., 1989;Beck, 1989;Holmes et al., 1990;Schuurhuis et al., 1991). The fact that our patients with AML received different combinations of cytostatic drugs may also hamper the interpretation of the assocation between mdrl expression and response to chemotherapy. Furthermore comparison of remission frequency may also be too blunt a tool to differentiate between resistant and chemosensitive disease.
The findings in 12 patients analysed on consecutive occasions and in patients with relapsed leukaemia support the notion that treatment with cytostatic drugs seldom seems to select clones of leukaemic cells that express mdrl RNA.
In the present series mdrl expression was more common in elderly patients, which might be one explanation for the more dismal prognosis in the elderly with AML. In secondary AML that generally is chemotherapy resistant the median level of expression among the samples with detectable mdrl RNA was higher than in de novo AML and the two patients with the highest expression were found in this group.
A negative association between mdrl RNA expression and remission frequency and/or duration of first CR has been reported in other studies where more uniform treatment protocols have been used (Sato et al., 1990;Pirker et al., 1991;Marie et al., 1991). In the study by Sato et al., only in one of seven patients analysed on consecutive occasions there was an increase of mdrl expression and in the study by Marie et al., two of four cases analysed on consecutive occasions increased their mdrl expression, one was unchanged and one changed from positive to negative. A negative relationship between response to chemotherapy (CR or partial remission) and P-glycoprotein expression, detected with the antibody C219, in leukaemic cells from patients with AML and ALL has also been demonstrated (Kuwarzuru et al., 1990). However, in another study where mdrl expression was determined with the monoclonal antibody MRK16 P-glycoprotein was not detected in any sample from 12 AML patients who were investigated at diagnosis and at relapse (Ito et al., 1989). In contrast, in another study where the antibody C219 was used 61-100% of the leukaemic cells from six of eight AML patients studied at relapse were positive (Musto et al., 1991).
If mdrl expression is a main cause for resistance to chemotherapy one would expect improvement of treatment results by the use of resistance modifiers. Until now there is only one AML patient reported in whom a resistance modifier has been used to overcome drug resistance. In the reported case no mdrl RNA could be detected in the leukaemic cells at diagnosis. In contrast, the cells at relapse 4 months after CR were positive. The patient was retreated with the original induction chemotherapy with the addition of cyclosporin A and attained a second CR of short duration . From the results of the present study and those by other investigators we conclude that expression of the mdrl gene could be one of several mechanisms of primary resistance to chemotherapy in acute leukaemia. But that selection of leukaemic cells that express mdrl RNA is a rare event following conventional antileukaemic treatment.
We thank Dr P Gunven for liver specimens. The study was supported by grants form the Swedish Cancer Society, the Karolin-ska Institute Foundations and U. Lundahl's and R. Lundberg's memorial funds.