MDR1 causes resistance to the antitumour drug miltefosine

Miltefosine (hexadecylphosphocholine) is used for topical treatment of breast cancers. It has been shown previously that a high percentage of breast carcinomas express MDR1 or MRP. We investigated the sensitivity of MDR1 -expressing cells to treatment with miltefosine. We show that cells overexpressing MDR1 (NCI/ADR-RES, KB-8-5, KB-C1, CCRF/VCR1000, CCRF/ADR5000) were less sensitive to miltefosine treatment when compared to the sensitive parental cell lines. HeLa cells transfected with MDR1 exhibited resistance to the compound, indicating that expression of this gene is sufficient to reduce the sensitivity to miltefosine. The resistance of MDR1 -expressing cells to miltefosine was less pronounced than that to adriamycin or vinblastine. Expression of MDR2 did not correlate with the resistance to miltefosine. As shown by a fluorescence quenching assay using MIANS-labelled P-glycoprotein (PGP), miltefosine bound to PGP with a K d of approximately 7 μM and inhibited PGP-ATPase activity with an IC 50 of approximately 35 μM. Verapamil was not able to reverse the resistance to miltefosine. Concentrations of miltefosine up to approximately 60 μM stimulated, whereas higher concentrations inhibited the transport of [3H]-colchicine with an IC 50 of approximately 297 μM. Binding studies indicated that miltefosine seems to interact with the transmembrane domain and not the cytosolic nucleotide-binding domain of PGP. These data indicate that expression of MDR1 may reduce the response to miltefosine in patients and that this compound interacts with PGP in a manner different from a number of other substrates. © 2001 Cancer Research Campaign www.bjcancer.com

The two multidrug-resistant MDR1-overexpressing HeLa cell lines were obtained by transfection of human HeLa S3 (HeLa-WT) cervix carcinoma cells with a MDR1 wild-type gene construct (HeLa-MDR1-G185) and with a mutation in codon 185 (Gly-Val, kindly provided by Dr M M Gottesman, HeLa-MDR1-V185), respectively (Kane et al, 1989). Following transfection, HeLa-MDR1-G185 cells were grown in the presence of vinblastine (100 nM) and HeLa-MDR1-V185 in the presence of colchicine (240 ng ml -1 ). One clone of each cell line was taken for further cultivation. MDR1-mRNA expression was controlled by reverse transcriptase PCR (Hofmann et al, 1997). Wild-type and mutant genes were controlled by sequencing (Spitaler et al, 1998).
Dose-response curves for calculations of the IC 50 values (Table  1) were obtained by plating the cells in 96-well plates. Following an incubation period of 4 hours, the drugs were added and the cells were exposed to the drugs continuously for 72 hours. Subsequently, cell proliferation was detected by the MTT assay (Mosman, 1983). The IC 50 values were calculated using CalcuSyn software from Biosoft, Cambridge, UK.

Detection of MDR1 and MDR2 mRNA levels
For detection of the mRNA levels, total RNA was isolated using RNAzol (Biotexs Laboratories Inc, Houston, TX, USA). Synthesis of cDNA and amplification of the MDR1-mRNA by polymerase chain reaction was performed as described (Noonan et al, 1990). Primers for the amplification of the MDR2-mRNA were: 2061-2083 (5′-TGT CAG AAG AGC CTT GAT GTG G-3′) and 2193-2215 (5′-TGG CAA TGG CAC ATA CTG TTC C-3′). β-Microglobulin was used to control the correct amount of RNA in the experiments (Noonan et al, 1990). Amplifications (30 cycles) were performed with a denaturation temperature of 94˚C (35 seconds), an annealing temperature of 57˚C (30 seconds), and an extension temperature of 73˚C (1 minute). Starting with cycle 16, the time for synthesis was extended (5 seconds per cycle). The reaction products were separated on a 10% polyacrylamide gel and stained with ethidium bromide.

MIANS-PGP quenching assay
Binding of miltefosine to P-glycoprotein (PGP) was carried out using fluorescence quenching, as described previously for drugs, chemosensitizers and hydrophobic peptides (Liu and Sharom, 1996;Sharom et al, 1998aSharom et al, , 1998b. Highly purified PGP, labelled with MIANS (Liu and Sharom, 1996) was titrated with miltefosine and quenching of the fluorescence emission at 420 nm was monitored. The dissociation constant K d for binding was estimated by fitting the data to an equation describing interaction with a single class of binding site.

PGP ATPase activity
The ATPase activity of P-glycoprotein in CH R B30 plasma membrane was measured as described previously  by detection of the release of inorganic phosphate from ATP, using a colorimetric method. Membrane vesicles (1-2 µg of protein) in buffer containing 2 mM ATP and 5 mM Mg 2+ were preincubated with miltefosine for 5 minutes before initiation of the assay by addition of ATP.

PGP-mediated [ 3 H]-colchicine transport
ATP-dependent uptake of [ 3 H]-colchicine into CH R B30 plasma membrane vesicles was determined by rapid filtration as outlined earlier , in the presence of increasing concentrations of miltefosine. Colchicine uptake was calculated as percent relative to a control in the absence of drug.

Effects of miltefosine on MDR1-expressing cells
It has been demonstrated that the resistant sublines shown in Table  1 overexpress MDR1 ( Figure 1; Akiyama et al, 1985;Kimmig et al, 1990;Hofmann et al, 1997). The resistance of the cell lines employed in this study to vinblastine and adriamycin (as attested by IC 50 values and factors of resistance) is shown in Table 1. All MDR1-expressing sublines exhibited cross-resistance to miltefosine. The resistance to the compound was less pronounced than that to vinblastine or adriamycin (Table 1). Resistance to miltefosine was also observed in a cell line transfected with wild-type or mutant PGP. This supports the notion that expression of MDR1 is sufficient to elicit resistance to miltefosine, and that it is not due to additional resistance mechanisms possibly induced during selection with adriamycin, vincristine or colchicine. Compared to the degree of resistance to vinblastine or adriamycin, the KB-C1 cell line exhibits low resistance to miltefosine ( Figure 1). PGP expressed in KB-C1 cells harbours a mutation in position 185 (glycine to valine) (Choi et al, 1988;Safa et al, 1990). Compared to wild-type PGP, the mutated PGP exhibited different substrate specificity for drugs (Choi et al, 1988;Safa et al, 1990) and differences in the sensitivity to reversing agents (Cardarelli et al, 1995). We investigated whether this mutation might influence the resistance to miltefosine. In a HeLa subline transfected with mutant PGP (HeLa-MDR1-V185) the profile of resistance to vinblastine and adriamycin was altered. Despite slightly higher expression of MDR1 in the HeLa-MDR1-V185 compared to the HeLa-MDR1-G185 cell line (Figure 1), the resistance to vinblastine was decreased (Table 1). This is in accordance with results published previously (Choi et al, 1988;Safa et al, 1990). However, this mutation did not alter the resistance to miltefosine significantly (Table 1; HeLa-MDR1-G185 = 8.3-fold, HeLa-MDR1-V185 = 9.6-fold).

MDR2-expression
It has been reported that the MDR2-encoded PGP transports phospholipids out of the cell (Smit et al, 1993;Ruetz and Gros, 1994). Thorgeirsson et al (1991) proposed a possible mechanism for co-induction of the MDR1 and MDR2 genes. If both genes are co-expressed, MDR2 might be responsible for the resistance to miltefosine. In order to investigate whether MDR2 might be involved in the resistance to miltefosine, we detected the expression of the MDR2 gene. As shown in Figure 1, there was no association between the resistance to miltefosine and the expression of the MDR2 gene. For example, the sensitive MCF7 and the resistant NCI/AD-RRES cell lines express MDR2 to a similar extent. On the other hand, neither sensitive HeLa-WT, nor the multidrugresistant MDR1-transfected HeLa cell lines express MDR2. Drugsensitive CCRF/CEM cells express low levels of MDR2, whereas the miltefosine-resistant CCRF/ADR5000 show no MDR2-expression ( Figure 1). Thus, MDR2-expression does not correspond to the resistance to miltefosine. In contrast, there is a clear association between the expression of MDR1 and the resistance to miltefosine (Table 1, Figure 1).

Interaction of miltefosine with PGP
In order to investigate whether miltefosine binds to PGP, we carried out a fluorescence quenching assay using highly purified protein isolated from CH R B30 cells. Binding of a wide variety of substrates to PGP, including drugs, chemosensitizers, and hydrophobic peptides, leads to saturable fluorescence quenching of highly purified PGP labelled with the fluorophore MIANS at two conserved cysteine residues within the Walker A motifs of the nucleotide-binding domains (Liu and Sharom, 1996;Sharom et al, 1998aSharom et al, , 1998b. The quenching curve can be used to estimate the dissociation constant, K d , which is a measure of the affinity of binding (Liu and Sharom, 1996;Sharom et al, 1998aSharom et al, , 1998b. To date, K d values have been measured covering the range 250 µM to 25 nM (Sharom et al, 1998a(Sharom et al, , 1999. As shown in Figure 2, miltefosine bound to purified MIANS-PGP with a K d of approximately 7 µM, indicating that the compound is a relatively high-affinity substrate for PGP. Drugs and chemosensitizers that interact with PGP can either stimulate or inhibit the basal ATPase activity of PGP (Thorgeirsson et al, 1991;Gottesman and Pastan, 1993;Sharom et al, 1995;Sharom, 1997). In plasma membrane vesicles from the highly drug-resistant CHO cell line CH R B30, miltefosine inhibited the ATPase activity with an IC 50 of 35 µM (Figure 3). In addition to the data shown in Figure 2, this is a further indication of an interaction between miltefosine and PGP. Miltefosine was also found to bind to the Bacillus subtilis ABC transporter YvcC with a K d of 2-3 µM (data not shown).
In order to investigate whether miltefosine interacts directly with the cytosolic NBD of PGP, murine NBD2 was purified and   (Conseil et al, 1998). A wide range of miltefosine concentrations (2 µM-2 mM) did not show any significant quenching. The binding site of the antiprogestin RU486 is close to the ATP-binding site (Conseil et al, 1998). The K d for RU486 in the absence of miltefosine, 17.9 ± 2.6 µM, was not markedly changed upon addition of miltefosine up to 320 µM (not shown here). This is an indication that the cytosolic NBDs do not contain the main binding site for miltefosine. The interaction site is therefore likely to be located within the transmembrane domain. The main binding site for steroids and other PGP transport substrates has also been found to be located in the transmembrane regions (Sharom, 1997;Vo and Gruol, 1999).

Combination of verapamil and miltefosine
If the resistance of MDR1-expressing cells to miltefosine is due to PGP-mediated efflux, verapamil should block the efflux and thereby reverse the resistance to miltefosine. Usually 5 µM verapamil is sufficient to reverse the resistance to antitumor drugs. However, in multidrug-resistant HeLa cells (HeLa-MDR1-G185), 10 µM verapamil did not reverse resistance ( Figure 4). Verapamil also did not enhance the antiproliferative activity of miltefosine in drug-sensitive HeLa-WT cells. In HeLa-MDR1-V185 harbouring a mutant PGP, the combination verapamil/miltefosine is slightly more effective than in drug-sensitive HeLa-WT and HeLa-MDR1-G185. In HeLa-MDR1-V185 this effect is not synergistic, but is at best additive. For comparison, the synergistic effect of vinblastine in combination with verapamil in HeLa-MDR1-G185 cells is shown ( Figure 5). Thus, although miltefosine seems to be transported by PGP, the efflux cannot be blocked by verapamil ( Figure  4). This is an indication that miltefosine seems to show anomalous behaviour compared to other substrates.

Modulation of the colchicine transport by miltefosine
If miltefosine interacts directly with PGP, it should modulate the transport of other drugs by the protein. Many PGP substrates and chemosensitizers inhibit colchicine transport into plasma membrane vesicles from MDR cells, or into reconstituted proteoliposomes containing PGP Sharom et al, 1993Sharom et al, , 1995Sharom, 1997). A good correlation has been found between the K d value for many drugs, chemosensitizers and peptides (as determined by MIANS-PGP quenching), and the IC 50 for inhibition of PGP-mediated [ 3 H]colchicine uptake into CH R B30 plasma membrane vesicles (Sharom et al, 1998a(Sharom et al, , 1998b. Some peptides and other compounds have been found to stimulate the transport of drugs, likely via positive allosteric effects Shapiro and Ling, 1997;Shapiro et al, 1999). Miltefosine at concentrations up to 60 µM led to a stimulation of colchicine transport into CH R B30 plasma membrane vesicles ( Figure 6). The observed activation in the range of 27% was highly reproducible in each experiment. This activation of colchicine transport can be explained by a positive allosteric interaction between two distinct, but possibly overlapping, substrate-binding sites for miltefosine and colchicine. Much higher concentrations of miltefosine led to inhibition of colchicine transport, with an IC 50 of approximately 300 µM. The inhibition of transport at high concentrations of miltefosine is probably due to non-specific detergent-like effects, as were observed previously for certain membrane-active peptides (Sharom et al, 1995).

DISCUSSION
We have investigated the susceptibility of human tumour cell lines to treatment with the phospholipid analogue miltefosine. We found that the expression of MDR1 elicits resistance to this compound, although it was less pronounced than that to adriamycin or vinblastine. In view of the fact that breast cancers frequently  (Trock et al, 1997), the observed resistance may suffice to reduce the efficacy of miltefosine, or to cause treatment failures. Breast cancers are, among others, treated with drugs transported by PGP, such as adriamycin, vinblastine, vincristine, etoposide or taxol. Our results show that tumours resistant to these compounds, due to the expression of MDR1, also exhibit reduced sensitivity to miltefosine. Cross-resistance of multidrug-resistant cells to miltefosine was observed previously, when multidrug resistance was induced with adriamycin, but not with colchicine (Himmelmann et al, 1990). Our results are in agreement with these data. The resistance to miltefosine of KB-8-5 and KB-C1 cells in which the resistance was induced by colchicine is very modest (Table 1). Resistance to miltefosine is more pronounced in adriamycin-induced cells (NCI/ADR-RES, CCRF/ADR5000) compared to the level of resistance to adriamycin. It is conceivable that crossresistance to miltefosine may develop not only via induction of MDR1 expression by adriamycin, but also by other mechanisms. However, we show here that two HeLa cell lines in which the resistance was obtained by transfection with MDR1 also exhibited resistance to miltefosine (Table 1). These results demonstrate that expression of MDR1 is sufficient to elicit resistance to the compound. A mutation at position 185 that alters the substrate specificity for several compounds does not influence the miltefosine resistance significantly.
One possible explanation for the cross-resistance of multidrugresistant cells to miltefosine would be that the compound is transported by PGP. This assumption is confirmed here by the fact that the expression of MDR1 correlates with the resistance to miltefosine. mRNA levels of MDR2 are not related to the miltefosine resistance. One approach to investigate whether a particular compound is a substrate of PGP is the determination of the quenching of the PGP-bound fluorescence probe MIANS (Liu and Sharom, 1996;Sharom et al, 1998aSharom et al, , 1998b. Using this assay, interaction of miltefosine was observed with PGP with a K d value of approximately 7 µM (Figure 2). Interaction with comparable affinity was observed with the bacterial ABC transporter YvcC. As indicated by experiments with NBD2 of murine PGP, miltefosine does not interact with the cytosolic ATP-binding site. A likely binding site may therefore be a drug/modulator-binding site within the transmembrane domain. Miltefosine also inhibited the ATPase activity of PGP ( Figure 3). These data illustrate that miltefosine does interact with PGP and the resistance to miltefosine seems to be due to efflux of the compound. Abulrob and Gumbletan (1999) reported that in MDR2-negative KB and MCF cells in which the expression of MDR1 was induced, the intracellular accumulation of a fluorescently-labelled phosphatidylcholine analogue was reduced as compared to the sensitive cells. These data confirm that the MDR1-encoded PGP is able to transport phospholipids. This proposal is supported by the recent report from one of our laboratories that PGP reconstituted into proteoliposomes acts as a flippase for a number of fluorescent phospholipids (Romsicki and Sharom, 2001).
As shown in Figure 4, the multidrug resistance modulator verapamil was not able to reverse the resistance to miltefosine. Recently, it has been shown that cells selected for resistance by the addition of miltefosine also express MDR1, indicating a connection between PGP and resistance to miltefosine. This resistance also could not be reversed by verapamil (Fu et al, 1999), which indicates that miltefosine does not interact with the verapamilbinding site. In agreement with this hypothesis is the observation that miltefosine increased the transport of [ 3 H]-colchicine ( Figure  6). Most PGP substrates show a good correlation between the K d value and the IC 50 for inhibition of the ATP-dependent [ 3 H]-colchicine uptake. Miltefosine exhibits unusual behaviour in that it activates transport at concentrations up to 60 µM. Although higher concentrations inhibit the colchicine uptake, such high concentrations of miltefosine may lead to detergent-like effects as observed previously for membrane-active peptides (Sharom et al, 1993). An explanation for these unusual results might be that miltefosine does not interact, as do many other substrates, with a site of PGP that leads to transport inhibition, but rather with a site that leads to allosteric activation of drug transport. To date, over 60 PGP substrates and modulators have been found to inhibit colchicine transport into membrane vesicles (Sharom et al, 1999). A few compounds have been found to stimulate colchicine transport (Lu and Sharom, unpublished data). Results similar to those with miltefosine were obtained with the synthetic peptide NAc-LLYamide which is a substrate of PGP, but activates colchicine transport (Sharom et al, 1998b). NAc-LLY-amide also did not compete with azidopine photolabelling. More recently, Shapiro and coworkers have shown that transport of the fluorescent PGP substrates rhodamine 123 and Hoechst 3342 in CH R B30 plasma membrane vesicles was stimulated by several other compounds in a positive allosteric fashion (Shapiro and Ling, 1997;Shapiro et al, 1999). MDR1-expressing cells have also been shown to be resistant to the phospholipid analogue ilmofosine. This compound slightly increased the photolabelling of PGP by azidopine (Hofmann et al, 1997). This illustrates that ilmofosine, like miltefosine, increases the interactions of PGP with certain substrates. From the facts that i) ilmofosine enhanced azidopine photolabelling (Hofmann et al, 1997), ii) the PGP-modulator dexniguldipine-HC1 did not reverse the resistance to ilmofosine (Hofmann et al, 1997), iii) verapamil did not reverse the resistance to miltefosine (Fu et al, 1999), and iv) cells resistant to miltefosine, in addition to MDR1, also expressed elevated levels of bcl-2, it was concluded that the phospholipid analogues ilmofosine (Hofmann et al, 1997) and miltefosine (Fu et al, 1999) are not substrates of PGP. However, these conclusions were not based on direct experiments designed to answer this question. The results obtained by the experiments shown contradict these conclusions, and show direct interaction of miltefosine with PGP.