MDR1 Gene Polymorphisms Affect Therapy Outcome in Acute Myeloid Leukemia Patients Free

Susceptibility to cancer represents a complex interplay between exposure to potential toxins and carcinogens and the genes involved in the detoxification pathways. Accordingly, genetic polymorphisms in genes associated with carcinogen metabolism may result in increased exposure to potential carcinogens. Even drug effects may be different in the presence of alterations of genes involved in drug metabolism (1). In addition to pathways involved in drug metabolism, efflux mechanisms such as the MDR1 gene product P-gp,² have been shown to be importantly involved in mediating the multidrug resistance phenotype during cancer chemotherapy. P-gp functions as an energy dependent efflux pump, transporting a wide array of structurally divergent compounds from the intracellular to the extracellular compartment (2). Whereas the MDR1 knockout mice model suggests a dispensable role for P-gp in normal development, the main physiological property of P-gp has been demonstrated as a protective mechanism against toxic substances (3, 4). This protective function may be especially important for CD34-positive stem cells, expressing P-gp during conditions of cytotoxic stress (5). Moreover, P-gp expression in liver, intestinal, and renal cells contributes to an increased clearance of cytotoxic substances (6) and is protecting the central nervous system against such toxins (7, 8).

P-gp drug substrates include anthracyclines and epipodophyllotoxins that have been shown to be efficacious in the treatment of AML (9). A reduced intracellular concentration of cytotoxic drugs attributable to the action of P-gp in AML blasts may therefore be related to resistant disease and failure of AML therapy. Indeed, ∼30% of younger patients (<60 years; Ref. 10) and ∼50% of older patients (>60 years; Ref. 11) presenting with AML have been shown to express readily detectable levels of P-gp on their blast cells. A number of studies have indicated that in relapsed patients, the incidence of P-gp expression increases (12, 13), indicating either drug selection of P-gp-positive cells and/or a positive regulatory effect of drugs used in the initial treatment regimens. The mechanisms by which P-gp is up-regulated are still a matter of debate. Altered activity of certain transcription factors have been implicated in MDR-1 regulation (14, 15, 16, 17). Another possibility to account for the up-regulation of P-gp is the hypomethylation of the MDR1 promoter (18). MDR1 up-regulation has also been reported after gene rearrangements, resulting in partial deletion of the MDR1 gene and expression of a deregulated fusion gene (19). Recently, it has been assumed that the extent of P-gp expression may be influenced by structural variations of the MDR1 gene (20). Furthermore, Hoffmeyer et al. (21) demonstrated multiple sequence variations in the MDR1 gene in a Caucasian population. They identified 15 polymorphisms of the MDR1 gene with various frequencies. The authors described a noncoding sequence change in exon 26 of the MDR1 gene at a wobble position as being significantly correlated with altered enterocyte P-gp expression and function. Accordingly, functionally relevant polymorphisms in the MDR1 gene are likely to be present among AML patients and may have an impact on response to antileukemic drug treatment. If MDR1 polymorphisms have an impact on P-gp expression, drug efficacy, in terms of eradicating leukemic blasts may also be altered. Furthermore, the related alterations of SNPs in P-gp expression pattern may change pharmacokinetic profiles of the drugs used to treat this disease, with a potential for enhanced toxicity or possibly suboptimal drug levels.

As determined by Cascorbi et al. (22), the three most frequent SNPs in the Caucasian population are located in exons 12, 21, and 26 (positions 1236, 2677, and 3435). Moreover, it was shown in healthy volunteers that these changes are in linkage disequilibrium and may therefore be associated with transcriptional regulation of MDR1 mRNA (23). We have therefore investigated these loci in a large AML patient population and determined frequencies of MDR1 SNPs in correlation with a quantitative MDR1 mRNA expression assay. In addition, the presence of MDR1 genetic variants and their expressed levels were compared with the clinical course of AML treatment.

Four hundred and five previously untreated AML patients were studied. Patient characteristics are summarized in Table 1. The patients were uniformly treated according to the protocol of the SHG-AML-96 study as published previously (24). Patients with the diagnosis AML FAB-subtype M3 were excluded and treated in a separate trial.

CR was defined as the presence of <5% of blast cells in a standardized bone marrow puncture after the second course of induction therapy. The study was approved by the ethics committee of the University of Dresden (review number EK210396). Each patient gave written informed consent.

Cytogenetic studies were performed at bone marrow samples taken before the beginning of therapy after standard G-banding techniques. The cytogenetic risk assessment was done as described previously (24). In brief, cytogenetic risk groups were defined as follows: poor risk, −5/del(5q), −7/del(7q), hypodiploid karyotypes (besides 45, X, ±Y or ±X), inv(3q), abn 12p, abn 11q, +11, +13, +21, +22, t(6;9); t(9;22); t(9;11); t(3;3), multiple aberrations; intermediate risk, patients without low-risk or high-risk constellation; low risk, t(8;21) and t(8;21) combined with other aberrations.

Bone marrow samples were taken at the time point of diagnosis. Mononuclear blood cells were isolated using a standardized Ficoll centrifugation procedure. Samples were subsequently frozen in liquid nitrogen. At the time of RNA extraction, samples were thawed according to routine protocols. Samples containing <80% of myeloblasts were referred to CD-3 depletion. We performed depletion with CD-3-coated Dynabeads (Dynal, Hamburg, Germany) according to the manufacturer’s recommendations. CD3-positive cells were eliminated with a sensitivity of 98% (data not shown). RNA extraction and cDNA synthesis were done as described previously with minor modifications (25). In brief, 5 × 10⁶ cells were pelleted and washed twice in PBS. Cells were dissolved in 0.2 ml of PBS. RNAzol was added to a total volume of 1 ml. Thereafter, 0.1 ml chloroform/isoamyl alcohol was added. RNA was recovered using standard ethanol precipitation protocols. Thereafter, 1 μg of RNA was used for first-strand cDNA synthesis with random hexamers.

We used probes and primers for the detection of MDR1 cDNA as described by Schiedlmeier et al. (26). The following primers were used: MDR1 forward, 5′-TGG TCC GAC CTT TTC TGG CCT TAT CCA-3′, and MDR1 reverse, 5′-CGA ACT GTA GAC AAA CGA TGA GCT A-3′. The PCR fragment was detected by the 6-carboxyfluorescein-labeled probe 5′-TGG TCC GAC CTT TTC TGG CCT TAT CCA-3′.

Primers and probe were purchased from TibMolBiol (Berlin, Germany). Amplification conditions were as follows: 2 min preincubation time at 50°C, 10 min at −95°C enzyme activation, and 50 cycles at 95°C for 15-s denaturation and 60°C for 1min annealing and extension.

Reference gene determination was carried out using commercially available GAPDH primers and probes from Perkin-Elmer-Applied Biosystems (Weiterstadt, Germany) with the same conditions as mentioned above. PCR reactions were run in triplicate. The critical threshold cycle (C_t) was defined as the cycle at which the fluorescence becomes detectable above background. A standard curve for MDR1 determination was obtained for C_t values in each Taqman reaction by plotting known quantities of genomic DNA of the retroviral SF-MDR vector (27). These standard curves were used to transform C_t values to the relative number of DNA molecules. MDR1 copy numbers of individual probes where then normalized to the respective GAPDH value. Homogeneity of the assays was controlled by comparing individual MDR1 copy number values with the respective GAPDH expression value. A standard curve of GAPDH values was obtained by cDNA dilutions of the human T-lymphoblastoid cell line CCRF-CEM.

Bone marrow or peripheral blood samples were taken at the time point of diagnosis. Mononuclear blood cells were isolated using a standard Ficoll centrifugation procedure. Genomic DNA was extracted from 5 × 10⁶ cells using either phenol/chloroform extraction after proteinase K digestion or a silica-based procedure (Qiagen DNA Blood kit; Qiagen, Hilden, Germany) according to the manufacturer’s protocols.

The most frequent polymorphic sequence strings of the MDR1 gene were amplified by PCR. PCR was performed on genomic DNA using the previously described primer molecules: exon 12 forward, 5′-TCC TGT GTC TGT GAA TTG CCT TG; exon 12 reverse, 5′-GCT GAT CAC CGC AGT CTA GCT CGC; exon 21 forward, 5′-GTT TTG CAG GCT ATA GGT TCC; exon 21 reverse, 5′-TTT AGT TTG ACT CAC CTT (23); exon 26 forward, 5′-GAT CTG TGA ACT CTT GTT TTC A; and exon 26 reverse, GAA GAG AGA CTT ACA TTA GGC (21). PCR conditions were as follows. Fifty ng of genomic DNA were amplified in a volume of 50 μl containing 50 mm KCl, 10 mm Tris-HCl (pH 8.3), 2.5 mm MgCl₂ (exon 26), or 3.0 MgCl₂ (exon 12 of 21) and 0.001% (w/v) gelatin, 200 μm deoxynucleotide triphosphates and oligonucleotides (0.5 μm each). One unit of AmpliTaq Gold DNA-polymerase (Perkin-Elmer, Norwalk, CT) was used per reaction. Cycling conditions were as follows: exon 12 + 21: annealing 53°C for 30 s, extension at 72°C for 30 s, denaturation 95°C for 15 s for 32 cycles; exon 26: annealing 60°C for 30 s, extension at 72°C for 30 s, and denaturation 95°C for 15 s for 32 cycles. PCR products were analyzed on standard 1.5% agarose gels stained with ethidium bromide (0.5 μg/ml).

For SSCP analysis, the PhastGel system (Amersham Pharmacia Biotech, Freiburg, Germany) was used. Five μl of the PCR product were heat denatured in an equal volume of deionized formamide and 0.01% xylene cyanol for 5 min. The probes were placed on ice and immediately loaded on a SSCP gel, essentially as recommended by the manufacturer. Gels were run with continuous cooling at 4°C. SSCP bands were visualized using a commercially available silver staining kit (Amersham Pharmacia Biotech) as recommended by the manufacturer.

The variants of exon 12 are further denoted as follows. The homozygous presence of the published MDR1 sequence (wild type or reference sequence) from chromosome 7 (GenBank accession no. AC 002457) is described as variant A of the respective exon, variant B is identical with heterozygosity at the described gene locus, and variant C resembles the homozygous “mutated” allele or better “alternative allele (aa).” The alternative alleles were for exon 12 C1236T, for exon 21 G2677T, and for exon 26 C3435T. Primer pairs covering the described sequence strings of exons 21 and 26 may amplify additional polymorphic loci. However, the frequencies of these polymorphisms are low (observed frequency for allele G2677A + T2677A, 3.8%, Ref. 22; for allele C3396T, 0.5%, Ref. 21). We therefore did not consider those changes in our survey and excluded patients with SSCP “extrabands.”

We assumed that the resulting SSCP pattern is reflecting the MDR genotype of the patient that is not different from the genotype of the patients blasts. However, because there are AML patients with loss of one part of the chromosome 7 (where the MDR1 gene is located), these patients may in our analysis reflect a status of artificial homozygosity. Thirty-six patients showed loss of chromosome 7 [including −7, del(7q)]. Ten of those 36 patients presented a state of heterozygosity of either the polymorphisms in exon 12, 21, or 26. To reduce the probability of false homozygous cases in our analysis, 12 marrow or peripheral blood samples with proven homozygosity and loss of chromosome 7 were referred to CD3 depletion. SSCP analysis was carried out both in the CD3-positive (normal T-cells) and in the CD-3 negative (mainly reflecting blast cells) fraction of samples. None of the investigated cases showed divergent SSCP pattern between T-cells and blast samples as it may be expected in the case of an artificial homozygosity by monosomy 7 (data not shown).

To further rule out misclassification of patients with loss of chromosome 7, the sensitivity of the SSCP assay to detect a status of heterozygosity was evaluated. Therefore, HeLa cells with loss of chromosome 7 were mixed with the HL 60 cell line (which was previously shown to be heterozygous at the investigated MDR1 gene loci). PCR was conducted exactly as it was described for the study population. As exemplified for the analysis of exon 26, the SSCP assay is able to detect 20% of contaminating HL 60 cells. Because only 4 patients with both homozygosity of the MDR1 polymorphisms and loss of chromosome 7 show a blast percentage of >80%, the likelihood of misclassification is very low. Therefore, we assume that the SSCP analysis described here is detecting the somatic genotype of the patient rather than a specific genotype of the patients blasts.

For sequence analysis of representative alleles of the MDR1 gene, each variant from exons 12, 21, and 26 was sequenced. The PCR product was cloned into pCR 2.1 TOPO vectors (Invitrogen, Leek, the Netherlands). After transformation of competent Escherichia coli and plating on selective agar plates (50 μg/ml kanamycin and 40 μg/ml 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside), 4–8 single white colonies were picked per sample and grown in 5 ml of LB medium cultures overnight. Plasmid DNA was prepared (Qiagen Plasmid Mini kit; Qiagen) and sequenced using Big Dye Terminator cycle sequencing chemistry as recommended by the manufacturer using T7 and M13 primers. Sequences were compared with the wild-type sequence using the Lasergene software package (DNASTAR, Madison, WI).

Expression analysis of different cDNA polymorphic variants of exon 21 was done using previously published primer pairs covering adjacent exon/exon sequence strings (20). PCR conditions were exactly as described previously. Forty cycles of amplification were run. PCR products were subsequently referred to SSCP analysis as described above.

Differences in age, leukocyte, and CD34 distributions between different polymorphic and MDR1 expression groups were calculated using the Mann-Whitney test. Pearson χ² test was used to identify differences in the distribution of cytogenetic risk groups and MDR1 expression frequencies in the respective patient category. Distribution inhomogeneities in FAB groups were tested with a logistic regression analysis. Furthermore, logistic regression analysis using multiple parameters was performed to identify the impact of single variables on CR. Cox proportional hazard regression was performed for the determination of variables affecting OS. Estimates of OS and DFS were calculated by the method of Kaplan and Meier. Differences between OS rates were compared using the log-rank test. All Ps are 2-sided, and a significance level of 0.05 was used. All analyses were performed using SPSS software.

We used the PCR-SSCP method to determine MDR1 SNPs in 405 AML patients. Different SNPs were identified by gel shifts within the SSCP gels (Fig. 1). Prototypes of SNPs as detected by SSCP were sequenced (Fig. 1). Thus, variants of MDR1 polymorphisms for exons 12, 21, and 26 could be determined in AML patients as it is depicted in Table 1.

There are 27 potential combinations of allelic variants when investigating the three MDR1 gene loci (three loci with three possible polymorphisms each). However, only 18 combinations were detected in our study (Table 2). The most frequent combination of polymorphic variants was a complete heterozygous state of the loci with 32.3% of all patients investigated (B genotype at each investigated position). The combinations of homozygous A variants and homozygous C variants were detected in 22.7 and 14.1% of the cases investigated, respectively. Thus, these three combinations cover 69.1% of all investigated cases and indicate linkage disequilibrium.

Patients characteristics are shown in Table 1. Median age was lower in the patients with the C genotype at any given position. Patients with the homozygous Ex-12-C and Ex-21-C showed a tendency to be less often diagnosed as AML FAB type M1. In contrast, an increased rate of patients with FAB type M4 was diagnosed in those patients. There were no significant differences between the polymorphism subgroups according to bone marrow blast cells, leukocyte count, lactate dehydrogenase, and CD 34 percentage in bone marrow (data not shown).

However, we observed differences in the distribution of cytogenetic risk groups between different polymorphism subgroups. Patients with the homozygous C variant are more likely to have a cytogenetic poor risk constellation as compared with all other cases. This difference is statistically significant when analyzed in subgroups of exons 12 and 26.

MDR1 cDNA expression was assessed in blast samples of 136 patients using the Real-Time PCR assay. Dilution curve analysis of the SF-MDR plasmid indicated that we were able to detect 10¹-10⁸ copies of MDR1 transcripts (data not shown). There was only one blast sample of AML patients with 8 copies ranging below that threshold. All other cases ranged between 22 and 847,705 copies. Median expression of MDR1 was 4,794.5 copies (25th percentile, 410; 75th percentile, 13,794). We therefore considered all cases expressing >4,794.5 copies as high MDR1-expressing patients, whereas patients with a median ≤4,794.5 copies were furthermore considered as low MDR1-expressing patients.

Older age was more likely to be associated with high MDR1-expressing blast samples (low-MDR1 patients median age: 52 years versus high MDR1 patients: 58 years; P = 0.06). Leukocyte count showed an inverse correlation with MDR1 expression (low MDR1 patients: 57 × 10⁹/l versus high MDR1 patients: 20 × 10⁹/l; P < 0.001), whereas the CD34 percentage in the bone marrow was directly associated with MDR1 expression (low MDR1-expressing patients: 11% CD34-positive cells versus high MDR1 patients: 49%; P < 0.001).

Interestingly, MDR1 blast expression significantly correlated with the presence of MDR1 polymorphisms in exons 21 and 26 (Table 3). Thus, patients with the A genotype in exons 21 and 26 demonstrate a lower MDR1 expression levels as compared with all other patients. Furthermore, heterozygous patients showed the highest percentage of MDR1-expressing cases at any gene position.

Patients who were homozygous for the A variants in all three positions (exons 12, 21, and 26) appeared to have a tendency for lower levels of MDR1 expression in their blast cells (Table 4). In contrast, the highest frequency of MDR1 expression could be shown for patients with the combination of three B variants. However, the observed differences were not statistically significant.

Representative samples of patients who were shown to be heterozygous for the exon 21 polymorphism were subjected to Real-time PCR and subsequent SSCP analysis. Because the primers used for genomic DNA typing cover both adjacent intronic-extronic sequence strings, previously published primer pairs with specificity for cDNA amplification of the exon 21 were used. Twelve cDNA samples with the Ex-21-B constellation were investigated. The results are consistent with the presence of the heterozygous genotype in that all samples showed expression of both alleles (Figure 2). Thus, it is unlikely that clonal inactivation of one of the MDR1 alleles may have occurred, and both alleles seem to be expressed in the heterozygous cases.

We performed a logistic regression analysis in the 136 AML patients for whom MDR1 expression data were available to identify variables that could predict achieving a CR after induction therapy. Age (>/< 60 years), status of the disease (secondary or de novo AML), cytogenetic risk profile according to the stratification of the AML-SHG96 study, MDR1 expression groups, and variants of the MDR1 gene were included into the investigation. As expected, age <60 years (P < 0.001) and the cytogenetic risk profile (P = 0.01) proved to be significantly correlated with CR. Moreover, in the multivariate analysis, high MDR1 expression as defined by a blast expression rate above the median of all patients was significantly associated with induction treatment failure (P = 0.02). Univariate analysis demonstrated that only 33.8% of patients with high MDR1 blast expression reached CR, whereas 60.3% of low MDR1-expressing patients were shown to be in CR after the second course of induction therapy (P = 0.002). In contrast, MDR1 gene variants did not show a significant influence on CR rate in the regression analysis (see also Table 3).

We determined the influence of MDR1 gene variants on survival parameters in the AML population with a Cox regression model. Again, age (>/<60 years), status of the disease (secondary or de novo AML), cytogenetic risk profile, MDR1 blast expression groups, and MDR1 gene variants were investigated for their impact on OS and DFS. Investigating all patients revealed age <60 years as the most powerful variable that correlated with survival (P < 0.01). Moreover, the cytogenetic good risk constellation of AML patients was associated with increased survival rates (P = 0.01). Neither MDR1 expression nor MDR1 gene variants showed an independent significant impact on survival when analyzing the whole cohort of patients.

Because older age is a well known prognostic factor of both low therapy response rate and OS, we wanted to determine the impact of MDR1 gene variants in patients <60 years. Multivariate Cox regression analysis including status of the disease, the cytogenetic risk profile, and the MDR-SNP groups revealed MDR1 gene variant Ex-26-A as an independent variable (P = 0.02) for decreased OS, the cytogenetic good risk constellation was furthermore associated with increased OS rate (P = 0.01), whereas the poor risk classification predicted for lower survival rates (P < 0.001). When included in the analysis, there was no statistically significant impact of MDR1 expression on OS (data not shown).

Therefore, Kaplan-Meier distributions were calculated for polymorphism of the exon 26. As it is shown in Fig. 3,A, patients with the Ex-26-A constellation show the worst survival in the study population (P < 0.01). The same Kaplan-Meier distribution could be detected when analyzing the impact of polymorphism of exon 26 on DFS, implying an increased rate of relapses in Ex-26-A as major reason for the observed decreased OS (data not shown). Indeed, patients with the Ex-26-A genotype display significantly higher probability of relapse, as it is shown in Fig. 3,B. Because allogeneic transplantation may substantially influence survival analysis in the study population, the impact of this procedure was analyzed. Analyzing the patient groups according to their postremission therapy revealed that 16, 21, and 39% of patients received allogeneic transplant therapy in the Ex-26-A, Ex-26-B, and Ex-26-C groups, respectively. Despite this imbalance in transplantation frequencies, censored survival analysis still revealed the negative impact of Ex-26-A on OS rate (Table 3), which is caused by an increased rate of relapses in the Ex-26-A group.

Similarly, when subjects with linked polymorphisms in exons 12, 21, and 26 were analyzed, only patients with the combination of the A variants showed a significantly decreased cumulative survival rate as compared with all other patients (Table 4 and Fig. 4,A). In contrast, patients with heterozygosity at any investigated genetic locus show increased survival rates (Tables 3 and 4). Favorable survival rates of heterozygous patients are mainly attributable to a low probability of relapse (Fig. 4 B).

A major problem of cytotoxic drug treatment is intrinsic or acquired drug resistance. One potential mechanism of drug resistance is mediated through expression of the P-gp efflux pump, enabling AML blasts to decrease intracellular toxic drug levels and thereby lower the rate of apoptosis.

Recently, it has been reported that genetic polymorphisms of the MDR1 gene may affect the expression and function of the P-gp efflux pump in healthy volunteers (21, 22, 23). Thus, another component of altered efficacy to drug therapy may be variable P-gp function attributable to the presence of variant MDR1 genotypes, among AML patients. These polymorphisms could not only influence sensitivity or resistance of the leukemic blast but could also impact on therapy outcome by altered drug clearance. In the current study, we found that AML patients who were homozygous for the wild-type polymorphism of the MDR1 gene (A variant) at position 3435 were likely to have decreased survival when treated according to the current SHG-AML 96 trial protocol. In contrast, patients with heterozygous (B variants) for the polymorphisms at the sites studied in our investigation (in exons 12, 21, and 26) had survival rates above average.

The genotype-phenotype association in our study shows a clear correlation between the homozygote A variant(s) and a lower MDR1 expression in blast samples. Similar results on genotype-dependent gene expression were obtained in one previous study that showed a tendency to higher MDR1 expression rates in heterozygotes (28). In contrast, there are two other studies that link the mutant or C-type polymorphism with lower MDR1 expression (21, 29). It is of note that the former study used bone marrow for MDR1 analysis as it was done in our survey, whereas the latter studies used biopsy specimens from the duodenum and placenta. These differences raise the possibility of differential gene regulation in different body tissues. There are currently no data about differential MDR1 gene regulation in normal tissues. However, there are some reports about specific mechanisms of up-regulation of the MDR1 transcript by leukemic blasts. Recently, Nakayama et al. (18) reported hypomethylation of the MDR1 promoter as a predictive factor for MDR1 up-regulation in AML patients, and Mickley et al. (19) showed gene rearrangements as causative events for MDR1 expression. In this light, it seems important that patients with the homozygous C variant harbor significantly more cytogenetic poor risk aberrations, which could have a positive regulatory effect on MDR1 expression (30) and could thus explain the higher expression rate of MDR1 in patients with the C variant(s).

Currently, it is not known in which way MDR1 polymorphisms influence gene expression and protein function with one exception. In exon 21, the C variant of the MDR1 gene is responsible for an altered MDR1 gene product. This polymorphism accounts for an alanine to serine substitution at codon 893 of P-gp and is responsible for an altered functional activity of the protein (23). In agreement with recent findings in healthy volunteers (23), we found that changes at position 2677 of the MDR1 gene are in linkage disequilibrium with two other loci in close proximity to exon 21 (polymorphisms at position 1236, exon 12; and polymorphisms at position 3435 at position 26, exon 26). Therefore, in addition to protein changes at position 893 of the P-gp protein, it is possible that polymorphisms in exons 12, 21, or 26 are functionally linked to polymorphic positions at regulatory sites of the MDR1 promoter and may account for different regulatory kinetics. One potential candidate for a functional linkage is a recently identified polymorphism at position 129 of the MDR1 gene that was shown to be associated with lower Pgp expression (29). Because this region of the MDR1 promoter does not map to the known regulatory elements such as the G-box, CAAT box, and heat shock-responsive element, the precise functional mechanism of this polymorphism remains to be determined. However, linkage of the investigated polymorphisms with regulatory elements of MDR1 is likely, because all homozygous A variants associate with lower median MDR1 expression in AML patients. Moreover, patients with a combination of these loci show the lowest median of MDR1 mRNA expression. Similarly, combined heterozygous B variants of the MDR1 genotype represent a phenotype that is associated with higher median MDR1 expression. Therefore, determination of linked polymorphisms is of value because polymorphic regulatory regions may comprise different MDR1 expression kinetics after exposure to the cytotoxic drugs used in the trial. In this light, it seems important that heterozygous cases express both MDR1 alleles as it could be shown for the B variant in exon 21. Thus, chromosomal inactivation does not occur in AML patients with functional expression and regulation of both alleles in heterozygote patients.

The results about MDR1 polymorphisms presented here raise questions about a crucial mechanism in the pathogenesis of AML. In this study, patients with the homozygous C variant in exon 12, 21, or 26 were ∼10 years younger than patients with the heterozygous B variant or the homozygous A variant. C-variant patients show significantly more cytogenetic poor risk aberrations. Moreover, these patients have a tendency to a lower overall survival rate despite their younger age, and one may speculate whether the homozygous C variant is associated with the acquisition of mutations in genes coding for DNA repair mechanisms or tumor suppressors. An example for such acquired alterations could be mutations of the tumor suppressor gene p53, which may create genetic instability (31). Additionally, mutated p53 may positively regulate MDR1 gene expression (15).

Taken together, the study presented here shows an independent impact of the homozygous wild-type polymorphism in exon 26 on survival of AML patients by a clearly increased relapse risk. Given the presence of linkage disequilibrium between the different genetic loci of the MDR1 gene, it is questionable whether the presence of Ex-26-A in AML patients or other as yet unidentified regulatory regions of the MDR1 gene are causative for the lower survival rate observed in those patients. Moreover, we cannot determine whether this phenomenon is attributable to characteristics of the AML blast population or attributable to altered P-gp-mediated drug pharmacokinetics. However, the suspected differential gene regulation between normal tissues and leukemic blasts is highlighting the role of increased drug clearance in patients with the A variant. This phenomenon may than override the initially observed lower MDR1 expression rate and lead subsequently to an increased rate of relapse. This would be one explanation of observations made by us and others (10) that MDR1 blast expression is predictive for achievement of CR but fails to predict influence on long-term survival. Therefore, investigations including pharmacokinetic studies are warranted to finally assess the impact of MDR1 polymorphisms on AML therapy.

We thank Dr. K. Kuehlcke (Idar, Oberstein, Germany) for providing us with the previously described SF-MDR vector. The technical assistance of Anja Liebkopf is greatly appreciated. We thank Silke Soucek for help with the database and for statistical advice and acknowledge their support in preparing the manuscript.