3.1. P-Glycoprotein

PGP-mediated or classic MDR, which was identified in the 1970s, is a well-characterized experimental phenomenon (reviewed extensively by Gottesman et al., 1995; Ueda, Yoshida, and Amachi, 1999; Ambudkar et al., 1999; Hrycyna, 2001; Borst and Oude Elferink, 2002). Classic MDR is characterized by (1) cross-resistance between a series of chemically unrelated drugs, (2) decreased drug accumulation (Biedler and Riehm, 1970; Danø, 1971), (3) increased expression of PGP (Juliano and Ling, 1976), and (4) reversal of the phenotype by a variety of different compounds [e.g., verapamil (VER)] (Tsuruo et al., 1981). The drugs most often involved in PGP-mediated MDR are of fungal or plant origin, including the anthracyclines [primarily daunorubicin (DNR) and doxorubicin (DOX)] and vinca alkaloids. Apart from drugs within these groups, a number of other, nonrelated compounds are able to induce PGP-mediated MDR [e.g., epipodophyllotoxins, actinomycin D, colchicine (COL), the taxanes, and the anthracenedione derivatives]. All these drugs are hydrophobic, and most are weak bases.

The primary sequence of PGP has been determined from sequence data obtained from cDNA. The protein consists of 1276 to 1280 amino acids with a molecular mass of 170 kDa. The commonly accepted model for the topologic structure of PGP is given in
Figure 1 .

According to this model, the protein has a tandemly duplicated structure. Each half of the molecule contains a nucleotide-binding domain and reveals six predicted transmembrane regions. The N- and C-termini, as well as the nucleotide-binding domains, are located intracellularly, and the first extracellular loop is glycosylated. Both nucleotide-binding domains are essential for proper functioning of the protein. Each consists of two core consensus motifs referred to as the Walker A and B motifs (Walker et al., 1982). These motifs generally are found in a wide range of ATPases, and they are involved directly in the binding and hydrolysis of nucleotides. Different topologic orientations of PGP have been reported, and several studies have indicated that conformational changes in the structure of PGP are involved in the mechanism of substrate efflux (Zhang, 2001; reviewed in Litman et al., 2001).

A small gene family with two members in humans and three in rodents encodes PGP. Despite extensive homology (amino acid identity > 70 percent) among all PGPs, they are divided into two different classes. Class 1 consists of the drug-transporting PGPs, which include the human MDR1 and the mouse mdr1a ( mdr3 ) and mdr1b ( mdr1 ) gene products. Class 2 (also referred to as class 3) includes the non-drug-transporting PGPs, such as the human MDR2 ( MDR3 ) and the mouse mdr2 gene products (reviewed in Germann, 1996). This thesis refers to the MDR-conferring isoforms (class 1) as PGP.

PGP is a member of a superfamily of ATP-dependent membrane transporters called ABC transporters (ATP-binding cassette transporters). Members of this transport superfamily display high amino acid similarity of the 200 amino acids surrounding the ATP-binding folds. Approximately 1100 ABC transporters are known at this time. The family includes bacterial transporters, the cystic fibrosis transmembrane conductance regulator, the Plasmodium falciparum drug-resistance gene, and genes apparently involved in peptide transport during antigen presentation. The human family of ABC transporters includes at least 48 members. Classification of known human ABC genes into families based on homology of the ATP-binding domain generates 7 families (Klein, Sarkadi, and Varádi, 1999; Dean and Allikmets, 2001; Dean, 2002). MDR1 belongs to the ABCB subfamily with 11 members (www.nutrigene.4t.com/humanabc.htm; www.gene.ucl.ac.uk/nomenclature/genefamily/abc).

Highest expression of PGP in humans has been detected in the liver bile canaliculi, the small and large intestine, the kidney, the pancreas, and the adrenal gland. PGP is also seen in endothelial cells of capillaries in the brain and testis. The tissue specificity of PGP has led to a number of proposals about the normal function of the protein, including transport of endogenous toxic compounds and transport of steroid hormones.

The best evidence as to the normal functions of PGP has been derived from mouse knockout models. Loss of either or both mdr1a or mdr1b had no effect on viability, fertility, or life span. Mice lacking functional mdr1a developed paralytic symptoms and died after administration of the antihelmintic ivermectin. Characterization of ivermectin biodistribution indicated that the brains of mice deficient in mdr1a accumulated almost 100-fold more ivermectin than control mice. This observation is consistent with the observation that PGP plays a role in the maintenance or generation of the blood-brain barrier. The phenotype of mdr1a knockout mouse appeared to be the same as that of mdr1a and mdr1b double-knockout mice, which, under laboratory conditions, suggests that mdr1b contributes little to the pharmacokinetics of drugs and xenobiotics (this result does not exclude the possibility that mdr1b may have some other, as yet undetermined, function) (Schinkel et al., 1994).

Although the true clinical relevance of PGP-mediated MDR is still heavily debated, PGP is believed to be one of the key molecules that confers MDR in cancer. Thus the presence of PGP in clinical tumours has been documented as well as its relevance for prognosis in several different neoplastic diseases such as acute myeloid leukaemia (Noonan et al., 1990; van den Heuvel-Eibrink et al., 1997; Dhooge et al., 1999).

3.2. Multidrug Resistance-Associated Protein

Several MDR cell lines with decreased drug accumulation but without expression of PGP have been described (Nielsen and Skovsgaard, 1992). One the most extensively characterized has been H69AR, a DOX-selected small cell lung cancer cell line. Using a differential hybridization approach, Cole et al. (1992, 1993) identified a 6.5-kb mRNA in this cell line. Sequencing of DNA clones derived from this mRNA has revealed a 1531-amino-acid (190-kDa) protein named multidrug resistance-associated protein 1 (MRP1) (reviewed by Cole and Deeley, 1998; Hipfner, Deeley, and Cole, 1999).

Like PGP, MRP1 belongs to the ABC superfamily of transport proteins. MRP1 belongs to the ABCC subfamily with at least 12 members (Klein, Sarkadi, and Varádi, 1999; Dean and Allikmets, 2001; Dean, 2002), including the cystic fibrosis transmembrane conductance regulator (CFTR) (www.nutrigene.4t.com/human abc.htm; www.gene.ucl.ac.uk/nomenclature/genefamily/abc). The genes encoding MRP1 and PGP are evolutionarily very distant, and the primary structure of the two proteins is quite dissimilar, sharing only 15 percent amino acid identity (Cole et al., 1992). Most of the sequence similarity between MRP1 and PGP is found within the nucleotide-binding domains that generally are conserved among members of the ABC superfamily. MRP1 is larger than other full-length ABC proteins, containing approximately 250 additional amino acids in its NH2 terminal. Thus, in addition to the 12 transmembrane segments characterizing PGP, MRP1 has five transmembrane domains (Cole and Deeley, 1998; Hipfner et al., 1997)
( Figure 2 ). Stable transfections of HeLa cells (Grant et al., 1994) and non-small cell lung cancer SW-1573 cells (Zaman et al., 1994) have established that MRP1 increases resistance to several anticancer drugs, such as DOX, DNR, vincristine (VCR), vinblastine (VBL), and VP16, with the specific exceptions of paclitaxel and MITOX (Cole et al., 1994). Unlike PGP, however, MRP1 appears to cause resistance to some heavy metal ions, including arsenite and antimonials (Cole et al., 1994), which is consistent with the extensive homology of MRP1 with the Leishmania arsenite transporter-encoding gene ( ltpgpA) and the yeast cadmium factor gene ( ycf1 ).

MRP1 is classified as an organic anion transporter; it transports anionic drugs and neutral drugs conjugated to acidic ligands, such as glutathione, glucuronide, or sulphate (Jedlitschky et al., 1994; Leier et al., 1996; Loe et al., 1996). One of the endogenous glutathione-S-conjugates transported by MRP1 is leukotriene C 4 (LTC 4 ) (Leier et al., 1994). LTC 4 is an arachidonic acid derivative involved in several receptor-mediated signal-transduction pathways. Together with other leukotrienes, LTC 4 plays an important role in the pathogenesis of human bronchial asthma and the inflammatory response. Since MRP1 has a very high affinity for LTC 4 , it has been postulated that LTC 4 may be a physiologically relevant MRP1 substrate.

Knockout mice without mrp1 have a decreased response to inflammatory stimuli, increased levels of glutathione, and increased sensitivity to VP16 but are otherwise healthy and fertile (Lorico et al., 1997; Wijnholds et al., 1997).

The human MRP family contains nine members (Borst and Oude Elferink, 2002). An important member is cMOAT (Oude Elferink and Jansen, 1994). This homologue of MRP1 (49.0 percent identity with human MRP1 at the protein level), also known as MRP2, is mainly expressed in the canalicular membrane of hepatocytes. It has been suggested that this protein could contribute to resistance (Cui et al., 1999; Kawabe et al., 1999; reviewed by Kuwano et al., 1999). Recently, seven new homologues of MRP1 have been identified [MRP3, MRP4, MRP5, MRP6, MRP7, MRP8, and MRP9 (MRP7-MRP9 have not yet been biochemically characterized)]. Among those, MRP3 also seems to confer resistance to VP16 and VCR (Zeng et al., 1999). Furthermore, available results suggest that at least some of the methotrexate-transporting GS-X pumps previously identified in cultured cells are actually MRP1, MRP2, and MRP3. Whether increased levels of MRPs could contribute to methotrexate resistance in vivo has not yet been studied (reviewed in Borst et al., 1999, 2000).

MRP1 has been detected in almost every tumour type examined, both in solid tumours and in haematologic malignancies. Further, it has been demonstrated that expression of MRP1 has implications for prognosis in some solid tumours (Norris et al., 1996; Nooter et al., 1997). However, data concerning the contribution of MRP1 to clinical drug resistance are still sparse, and the issue is complicated by the widespread expression of MRP1 in normal tissue. In general, it is still unclear whether MRP1 expression significantly affects drug sensitivity of most types of tumours.

3.3. Lung Resistance-Related Protein

In 1993, another resistance-related protein was described (Scheper et al., 1993; reviewed by Scheffer et al., 2000). It was first identified in a human MDR lung cancer cell line and, because of that, named lung resistance-related protein (LRP). The LRP gene is located on chromosome 16p13.1-16p11.2, close to the gene encoding for MRP1 (Scheffer et al., 1995). However, the two genes are not normally located in the same amplicon. LRP consists of 896 amino acids and has a molecular mass of 110 kDa. The deduced amino acid sequence of LRP shows 87.7 percent identity with the rat major vault protein (Scheffer et al., 1995). This protein is one of the structural proteins of the vault, a large, abundant ribonucleoprotein particle. Vaults are distributed widely throughout eukaryotes, and their morphology is highly conserved among various species. The remarkable structural conservation and broad distribution suggest that their function is essential to eukaryotic organisms. However, the function of vault proteins is still unclear. Most of these proteins have a cytoplasmic localization (Kedersha and Rome, 1986), and it has been suggested that vaults may regulate nucleocytoplasmic as well as vesicular transport of different substrates, including cytostatic drugs. This suggestion is supported by a decreased nuclear-cytoplasmic ratio of drugs and increased sequestration of drugs in exocytotic vesicles in LRP-expressing MDR cell lines (Schuurhuis et al., 1991). Increased expression of LRP has been associated with decreased DNR accumulation in leukaemic blast cells (Michieli et al., 1997). Further, accumulation of DNR seems to be inversely correlated with expression of LRP in childhood acute lymphoblastic leukaemia (den Boer et al., 1999). Recently, Kitazono et al. (1999) demonstrated that LRP is involved in resistance to DOX, VCR, VP16, paclitaxel, and gramicidine-D and has an important role in the transport of DOX between the nucleus and the cytoplasm.

Scheffer et al. (1995) have established stable LRP gene transfectants. Drug sensitivity results indicate that the LRP gene itself is insufficient to confer the MDR phenotype. More recently, a vault protein with a molecular mass of 193,000 kDa has been identified. It is possible that other vault components are essential for their role in MDR (Schroeijers et al., 2000).

LRP is frequently increased in acute myelogenous leukaemia and multiple myeloma and may have a potential relevance for prediction of response (Rimsza et al., 1999; Xu et al., 1999). Thus results from several, but not all, clinicopathologic studies have shown that expression of LRP is a strong independent prognostic factor (Izquierdo et al., 1998).

3.4. MXR/BCRP/ABCP1/ABCG2

This drug transporter gene was discovered in 1998 (reviewed in Litman et al., 2001; Borst and Oude Elferink, 2002). It was discovered and cloned independently by three different laboratories, resulting in three names: MXR (the mitoxantrone-resistance gene; Miyake et al., 1999) (this name will be used herein), BCRP (the breast cancer-resistance gene; Doyle et al., 1998), and ABCP (the ABC transporter gene expressed in placenta; Allikmets et al., 1998). The Human Genome Nomenclature Committee (HUGO) recently recommended that the gene should be renamed ABCG2 (Klein, Sarkadi, and Varádi, 1999; Dean, 2002). MXR encodes a 655-amino-acid (72.1-kDa) protein. This protein has six transmembrane domains and one ATP-binding domain. Thus it is thought to be a half-transporter and is believed to homodimerize to produce an active transport complex (Ozvegy et al., 2001). The mouse ortholoque of the human MXR gene, mxr/bcrp/abcp , has been cloned and shown to encode a 657-amino-acid protein with 81 percent identity with MXR (Allen et al., 1999). Recently, a murine putative partner protein, abcp2, has been identified (Bates et al., 2001).

The phenotype conferred by MXR is characterized by high levels of resistance to MITOX and topoisomerase I inhibitors [topotecan and SN-38 (the active metabolite of irinotecan)], moderate resistance to anthracyclines, and lack of resistance to vinca alkaloids, paclitaxel, and cisplatin (Litman et al., 2000).

The protein is highly expressed in the placenta, and preliminary results suggest that the protein may eliminate foetal waste products or may protect the foetus from potential maternal toxins. Preliminary studies also suggest localization in the small intestine, indicating a role in elimination of xenobiotics. Other localizations are the liver, colon, lung, kidney, adrenals, and the endothelia of veins and capillaries (Litman et al., 2001). The clinical relevance has not yet been revealed. However, a relatively high expression of MXR has been observed in approximately 30 percent of patients with acute myeloid leukaemia, suggesting a potential role for this transporter in drug resistance related to leukaemia (Ross, 2000).

3.5. at-MDR

Cell lines that are drug resistant due to altered TOPO II are called at-MDR ( a ltered t opoisomerase II activity or at ypical MDR). This type of MDR is characterized by cross-resistance between anthracyclines and epipodophyllotoxins but not vinca alkaloids. In cell lines with at-MDR only, the accumulation is unchanged.

Most at-MDR cell lines have been selected with epipodophyllotoxins. However, at-MDR is also a common mechanism of resistance in anthracycline-selected cell lines (Friche et al., 1991; Prost, 1995). Many cell lines resistant to complex-forming TOPO II inhibitory drugs have more than one mechanism of resistance. Thus many have reduced uptake due to the presence of PGP or other membrane-associated transport proteins (Sullivan and Ross, 1991).

3.5.1. Topoisomerase II

Topoisomerases are nuclear enzymes that modulate the topologic state of DNA and are involved in virtually every aspect of DNA metabolism. These enzymes are essential in survival of eukaryotic cells and play important roles in DNA replication and transcription, chromosome condensation and segregation, and repair (Burden and Osheroff, 1998; Bakshi, Galande, and Muniyappa, 2001). Mammalian topoisomerases were classified originally according to the number of strands they cleave. Thus TOPO II acts by cleaving a double-strand DNA helix, whereas TOPO I cleaves only single-strand DNA. Recently, mammalian TOPO III-cleaving single-strand DNA was identified (Hanai, Caron, and Wang, 1996), making this classification obsolete. This thesis will concentrate on TOPO II. This enzyme alters the topology of DNA by passing an intact double strand of DNA through a transient double-strand break. During this process, an intermediate DNA-enzyme complex (the cleavable complex) is formed. By covalent binding to and stabilizing this complex, TOPO II-interacting drugs (complex-forming TOPO II inhibitors) such as anthracyclines, epipodophyllotoxins, and anthracenediones inhibit the rejoining action of the enzyme, resulting in DNA double-strand breaks (Osheroff, Zechiedrich, and Gale, 1991).

Two closely related isoforms of TOPO II have been demonstrated. These have different biochemical and pharmacologic profiles. The smaller, 170-kDa form is termed TOPO II α , and the larger, 180-kDa form is termed TOPO II β . TOPO II α gradually increases through S-phase, reaching a peak in G 2 M, followed by a decrease as mitosis is completed. Conversely, TOPO II β expression is constant once cells have entered the cell cycle (Drake et al., 1989; Woessner et al., 1991). The cDNAs for mammalian TOPO II α and II β have been cloned and sequenced (Zechiedrich et al., 1989; Tsai-Pflugfelder et al., 1988). Although TOPO II α and II β share extensive amino acid sequence identity (~70 percent), they are products of different genes located on chromosomes 17q21-22 and 3p24, respectively.

Alterations in TOPO II drug activity can be divided into two groups: (1) quantitative changes, e.g., decreased levels of TOPO II protein through downregulation of transcription, increased degradation, or deletion of one allele, and (2) qualitative changes, e.g., mutations resulting in an altered drug-DNA-protein interaction or ATP binding, alterations in the ratio of isoenzymes, or altered enzymatic function by posttranslational modification. TOPO II has phosphorylation sites available on its serine residues, and posttranslational modification of the phosphorylation status of the protein may play a significant role in regulating the activity (Nitiss and Beck, 1996; Valkov and Sullivan, 1997).

4.1. Tumour Cell Lines

Sensitive cell line. Wild-type Ehrlich ascites tumour (EHR2) cells were used in the experiments. The EHR2 cells have been investigated extensively in context of accumulation and efflux of anthracyclines and vinca alkaloids (Danø 1971, 1973; Skovsgaard 1977, 1978a, 1978b, 1978c). The transport experiments performed in this study are a further development of these studies. Additionally, Sehested et al. (1989a) and Demant, Sehested, and Jensen (1990) used the EHR2 cells for investigation of anthracycline and vinca alkaloid binding and transport.

The use of EHR2 cells has several advantages. First, as the cells grow intraperitonally in mice, it is easy to obtain the large number of cells necessary for transport assays. Second, the viability of EHR2 cells is greater than 95 percent, as judged by the exclusion of trypane blue. Finally, most cells are clonogenic; thus the plating efficiency of EHR2 cells is very high ( > 60 percent) compared with that of human tumour cell lines (10-20 percent) (Roed et al., 1987). On the other hand, the rodent cells are selected in very high levels of drugs, and EHR2 cells do not invade or metastasize. The resistance mechanisms in rodent tumour cell lines may be different from the resistance mechanisms in human tumour cells, making the study less relevant. Such criticism seems impossible to refute. However, all resistance mechanisms reported in human tumour cell lines also have been demonstrated in rodent cell lines and vice versa (Beck and Danks, 1991; Sugimoto and Tsuruo, 1991).

Subline exposed to irradiation. EHR2/irr was derived by sequential exposure of logarithmically growing, in vitro -established EHR2 cells to 12 fractions of 5 Gy of irradiation. The EHR2/irr cell line was twofold resistant to VCR and sixfold resistant to VP16 but remained sensitive to DNR (Nielsen et al., 2001).

Cell lines selected with combinations of DNR and chemosensitizers. Three sublines were developed from EHR2 and six sublines from the PGP-positive EHR2/ DNR cell line [developed by treatment with DNR (0.1 mg/kg four times weekly)]. The sublines were developed by treatment in vivo with DNR, a combination of DNR and VER, or a combination of DNR and cyclosporin A (CsA) (Nielsen et al., 2002). In the resistant cell lines, the postscripts stated the doses of DNR (in mg/kg four times weekly) used for development of the sublines; postscript V indicated that the subline had been developed in a combination of DNR and VER, and postscript C indicated that the subline had been developed in a combination of DNR and CsA. Compared with EHR2, the DNR-selected sublines EHR2/ 1.6, EHR2/DNR/0.1, and EHR2/DNR/0.4 were 26.1-, 29.8-, and 68.4-fold resistant to DNR. The DNR + VER-selected sublines (EHR2/1.6V, EHR2/DNR/0.1V, and EHR2/DNR/0.4V) were 44.1-, 47.4-, and 71.1-fold resistant to DNR, respectively, whereas the DNR + CsA-selected sublines (EHR2/1.6C, EHR2/DNR/0.1C, and EHR2/DNR/0.4C) were significantly less resistant (1.8-, 18.6-, and 22.6-fold).

Anthracycline-selected sublines. A series of DNR-resistant sublines were established in vivo in mice by intraperitoneal treatment with DNR 0.1, 0.2, 0.4, 0.8 and 1.6 mg/kg four times weekly, corresponding to 6.25, 12.5, 25, 50, and 100 percent of the LD 10 dose, respectively (Nielsen et al., 1994; Nielsen, Maare, and Skovsgaard, 1994, 1995). After 36 passages, the cell lines were 2.9-, 10.4-, 38.6-, 25.4-, and 31.6-fold resistant to DNR as compared with EHR2, respectively.

Etoposide-selected subline. This cell line (EHR2/ VP16) was developed and maintained in vivo in mice by intraperitoneal treatment with VP16 20 mg/kg four times weekly, corresponding to the LD 10 dose (Nielsen et al., 2000a). Compared with the parent cell line (EHR2), EHR2/VP16 was 114-fold resistant to VP16, 6-fold resistant to DNR, and 4-fold resistant to VCR.

MITOX-selected subline. The MITOX-resistant cell line (EHR2/ MITOX) was developed in vivo in mice by intraperitoneal treatment with MITOX 2.5 mg/kg five times weekly, corresponding to the LD 10 dose (Nielsen et al., 2000b). The EHR2/MITOX cell line was highly resistant to MITOX (6123-fold), cross-resistant to DNR (32.6-fold) and VP16 (29.8-fold), but remained sensitive to VCR.

To reduce or avoid sensitivity drifting, all experiments were performed within a few passages, and cell lines were reestablished from frozen cell cultures at regular intervals.

4.2. Sensitivity Assays

A comparison of the different sensitivity assays has shown that the clonogenic assay is the best measure of the proliferative capacity of the cells, provided that the plating efficiency is high enough (Roper and Drewinko, 1976). The clonogenic assay employed in this study was developed and improved by Roed et al. (1987).

Previously, Roed et al. (1987) reported a large interexperimental variation in DOX sensitivity when comparing dose-response curves obtained with the clonogenic assay. In addition, Jensen et al. (1989a) found a fivefold variation in DOX sensitivity of small cell lung cancer cells after 1 h exposure to the drug. On the other hand, these authors were able to reduce the variation significantly by culturing in a rigorously standardized manner. Furthermore, they demonstrated much less interexperimental variation when using a continuous drug exposure (Jensen et al., 1989a, 1989b). With a continuous drug exposure, the drug concentration is poorly defined because many drugs are unstable and/or metabolized by the cells. Theoretically, this is a major disadvantage. However, Jensen et al. (1992, 1993) found similar results when comparing a 72-h incubation procedure with a continuous incubation procedure. Accordingly, a continuous drug exposure was used in this study.

The tetrazolium dye colourimetric assay [MTT; 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide] (Tada et al., 1986) was used to evaluate sensitivity in a large group of cell clones (Nielsen et al., 1998). Previously, this assay was compared with the clonogenic assay and the dye exclusion assay (Carmichael et al., 1987). The assay was found to offer a valid and simple method of assessing chemosensitivity in established cell lines.

4.3. Drug Transport Assays

4.3.1. Steady-state accumulation

Since DNR is one of the anthracyclines, which has a very high cellular accumulation rate, and since PGP and MRP1 transport it, this drug was used primarily. Since our studies and studies of Stride et al. (1996, 1997) showed that murine mrp1 had little affinity for DNR, [ 3 H]VP16 was used in studies concerning mrp1. The method used to determine drug accumulation was essentially that of Skovsgaard (1978c). The steady-state accumulation typically was measured by incubation of the cell lines for 60 to 120 min at 37°C in the presence of the drug. The intracellular DNR concentration was assessed spectrofluoroscopically, the intracellular concentration of MITOX was measured by spectrophotometry (611 nm), and cellular content of [ 3 H]VP16 and [ 3 H]VCR was measured by scintillation counting.

4.3.2. Influx of daunorubicin

Influx of DNR was measured as previously described (Skovsgaard, 1977). The medium was supplemented with DNR (2.5-25.0 mM) and either glucose (10 mM) or Na + -azide (10 mM). Measurements of influx are complicated by factors such as self-association of drug molecules, adsorption of drug to the cell surface membrane, difficulties in quantifying low intracellular drug concentrations, and the need for measurements to be performed rapidly. Difficulties in measuring influx in this study were reflected by a considerable interexperimental variation (26.2 percent for EHR2 and up to 24.5 percent for the DNR-resistant cell lines when investigated in glucose-enriched medium).

4.3.3. Efflux (washout kinetics) of daunorubicin

Studies of DNR transport in intact cells are simple in principle. However, several limitations exist regarding the kinetics analysis. The anthracyclines bind not only to nuclear DNA but also to membrane lipids. Intracellularly, the drug distributes inside a system of compartments. Consequently, only a small fraction of the total DNR that accumulates in cells is in solution in the cytoplasm and free to diffuse. Thus the rate of efflux generally is limited by the rate of debinding of drug from its intracellular binding sites. This greatly complicates the analysis.

In this study, the efflux of DNR was investigated in EHR2 and a series of DNR-resistant sublines using the method described by Skovsgaard (1977). The effect of the chemosensitizer VER was investigated by adding different concentrations of VER to the efflux medium. The optimal concentration of VER was selected according to previously performed dose-response studies.

This method also was used to investigate the relation between efflux and intracellular concentration of DNR in PGP-positive cells (EHR2/0.8, passage 72) preloaded with DNR (5-150 μ M).

4.4. Determination of Resistance Proteins

4.4.1. Determination of P-glycoprotein

Various techniques are available to assess expression of MDR1 mRNA or PGP (Herzog et al., 1992). The majority of published studies reported determination of MDR1 mRNA (Noonan et al., 1990). In general, the level of mRNA expressed correlates with the level of protein, although exceptions have been reported (Brophy et al., 1994), indicating that PGP may be regulated at the translational or posttranslational level. Thus, when the function of PGP is investigated, measurements of PGP content should be preferred to MDR1 mRNA measurements. Several techniques are available for the measurement of PGP, including immunocytochemical techniques (Chan et al., 1988), flow cytometric techniques (Herzog et al., 1992), and Western blot. All these techniques rely on the use of antibodies to detect PGP, and numerous antibodies are available. Naturally, differences in the affinity of various antibodies for PGP and the varying sensitivities of these immunoassays can lead to a great deal of variability in the detection/quantification of PGP.

Western blot analysis is essentially a qualitative technique. However, semiquantitative measurements can be achieved (Garfin and Bers, 1989). To improve the ability to semiquantify PGP, we optimized a Western blot analysis (Nielsen, 1994; Nielsen et al., 1994). This method was described in detail previously (Nielsen, 1994; Nielsen et al., 1994). Briefly, the method described previously by Hitchins et al. (1988) was used to prepare a plasma membrane-enriched microsomal fraction. This preparation was resolved by the method of Laemmli (1970), and PGP was identified by immunoblot analysis. The antibody applied was C219, a murine monoclonal IgG2 subclass antibody developed by Kartner et al. (1985). C219 recognizes the mdr1a and mdr1b gene products as well as the mdr2 gene product.

In the absence of purified PGP, the concentrations of the protein in test cells could be related only to an arbitrary standard. We decided to use membrane preparations from sensitive EHR2 and multidrug-resistant EHR2/DNR+ cells (Danø, 1971) for assay calibration. The membrane preparations were prepared in bulk and stored frozen in order to produce an internal reference. A calibration curve was achieved by mixing membrane preparations of EHR2 and EHR2/DNR+. As discussed by Nielsen (1994) and Nielsen et al. (1994), the method has several limitations. Anyhow, the standardized procedure allowed reproducible measurements of PGP because the overall coefficient of variation for the assay was acceptable (10.4 and 7.2 percent for cells with high and low content of PGP, respectively).

4.4.2. Determination of multidrug resistance-associated protein 1

Preparations of membrane fractions and Western blot analysis were performed as described for PGP. The monoclonal antibody used was anti-MRP (MRPr1) (catalog no. 475726, Calbiochem-Novabiochem Corporation, San Diego, CA, USA). mrp1 was visualized by chemiluminescence using the sensitive cell line EHR2 as a reference.

4.4.3. Determination of mdr1a and mdr1b mRNA

Total RNA was purified by the method of Chomczynski and Sacchi (1987). RNA concentrations and purity were determined spectrophotometrically. Reverse-transcriptase polymerase chain reaction (RT-PCR) amplification of both mdr1a and mdr1b mRNA was performed qualitatively, whereas a semiquantitative determination was performed for mdr1a . Amplification of mRNA for the household gene b -actin was used as a control. In brief, RNA aliquots were treated with DNase in the presence of RNAguard RNase inhibitor and thereafter extracted with phenol-chloroform and precipitated with ethanol. Samples containing DNase-treated total RNA were reverse transcribed in buffer with random hexadeoxynucleotides, deoxynucleotides, RNAguard RNase inhibitor, and reverse transcriptase. These cDNA preparations were diluted and used for RT-PCR amplification. PCR reactions were run on a Perkin Elmer 9600 Thermo Cycler.

The following primers were used [Genbank accession numbers (www.ncbi.nlm.nih.gov/Genbank) and nucleotide positions in brackets]: mdr1a sense [M33581 (366-388)], mdr1b sense [M14757 (351-373)], mdr1a + mdr1b [common downstream primer, 1a: M33581 (620-598), 1b: M14757 (602-580)].

For semiquantitative determination of mdr1a mRNA, mdr1a and β -actin mRNA were subjected to five amplifications using different numbers of cycles, and the PCR products were run in agarose gels and visualized by ethidium bromide fluorescence. Both mdr1a and β -actin were in the linear part of the exponential amplification between 25 and 30 cycles. The β -actin curves overlapped, and therefore, mdr1a differences were used as semiquantitative estimates.

4.4.4. Determination of multidrug resistance-associated protein 1 (mrp1) mRNA

RT-PCR amplification of mrp1 mRNA was performed semiquantitatively. The primers used were sense [AF022908 (2552-2574)] and antisense [AF022908 (2755-2733)]. Semiquantitative determination of mrp1 mRNA was performed as described for mdr1a .

4.4.5. Determination of lung resistance-related protein

Immunocytochemical analysis was performed to determine the expression of LRP (the monoclonal antibody was unsuitable for immunoblotting analysis; Scheper et al., 1993). The APAAP (immune complexes of alkaline phosphatase and monoclonal antialkaline complexes) technique described by Cordell et al. (1984) was used. The primary monoclonal antibody used was LRP-56 (Scheper et al., 1993). This monoclonal antibody was developed by immunization of mice with 2R120 cells (PGP-negative human non-small cell lung cancer cells). Sequence analysis indicated 87.7 percent amino acid identity between LRP and the major vault protein from Rattus norvigecus (Kickhoefer and Rome, 1994). It is possible, however, that the antibody LRP-56 did not recognize murine major vault protein. Yet, at this time, the LRP-56 antibody was the only antibody available for detection of LRP. Recently, several monoclonal antibodies have been isolated (Schroeijers et al., 2001).

4.4.6. Determination of topoisomerase II

Nuclear extracts of cells in exponential growth were used for the experiments. The expression of TOPO II α and II β was measured by Western blot analysis (Friche et al., 1991). This assay only allowed determination of the relative amount of TOPO II in the respective cell lines. The TOPO II antiserum used has been described previously by Danks et al. (1988). Drug resistance due to altered TOPO II activity was not investigated.

4.5. Fluctuation Analysis

The Luria-Delbrück fluctuation analysis is a combined experimental and statistical method. The analysis allows one to distinguish between variant cells arising by spontaneous mutations and variant cells arising through adaptation to an environmental selection (Luria and Delbrück, 1943). It is based on the variation observed in colonies from parallel cultures. When a small number of cells are expanded, the prevalence of variant cells in the resulting population can show considerable fluctuation. This is a result of the variation rate and the time of appearance of any variation. When similar cell populations from the same parental culture are expanded, the number of variant cells within each population shows a considerable variation due to the random appearance in time of the variant clones ( Figure 3a ). However, when samples from the parental culture are seeded directly into selective media, the number of variant colonies in the samples follows a Poisson distribution and expresses less variability ( Figure 3b ). The mean number of variant clones from a direct sampling of the parental culture represents the prevalence of variant cells within the parental population. The wide fluctuation in the number of variant colonies in parallel cultures reflects variations. The fluctuation analysis is a quantitative analysis of such fluctuations.

Originally designed for bacterial populations, the fluctuation analysis has been applied widely in somatic cell genetics (Law, 1952). There are fundamental genetic differences between bacteria and somatic cells. Moreover, a considerable statistical error is associated with the fluctuation analysis, and the method has several limitations. Kendal and Frost (1988) have reviewed the use and limitations of the analysis and have concluded that the method can be applied to somatic cell genetics. However, reliable results can be obtained only when the purpose is to demonstrate the qualitative consequences of variation.

In this study, Luria-Delbrück fluctuation analysis was used to distinguishing between DNR-resistant cells arising by spontaneous mutations and DNR-resistant cells arising through adaptation. Two thousand cells were seeded in 13 parallel cultures and allowed to expand (4.0 × 10 6 cells/flask). Cells from each population were then plated in a selective medium (DNR 7.5 × 10 -9 and 10 -8 , respectively) for 14 days and then replaced in drug-free medium. Surviving colonies were allowed to grow for 14 days and then were harvested and propagated individually for further studies. The probability of preexisting resistant clones in the original seeded population was excluded because no surviving colonies were observed when 13 populations of 2 × 10 3 cells were treated directly with DNR 7.5 × 10 -9 and 10 -8 M without expansion of the populations before drug exposure.

4.6. Determination of ATPase Activity

Microsomes were prepared as described by Hitchins et al. (1988). The total and the vanadate-sensitive membrane-associated ATPase activities were measured according to Borgnia, Eytan, and Assaraf (1996). The release of inorganic phosphate from ATP was quantified using a colourimetric method described by Chifflet et al. (1988). This method was chosen because it allowed determination of ATPase activity of the respective proteins in their natural environment. This could be important because the ATPase activity of PGP has been found to depend on the lipid composition and the presence of detergents (Senior, al-Shawi, and Urbatsch, 1995).

5.1. Regulation of Resistance Proteins

As concerns resistance proteins, only regulation of the MDR1 genes has been investigated in detail. Analysis of the expression of MDR1 suggests that regulation may be both species- and tissue-specific. Three complementary approaches have been used to investigate the mechanisms of regulation of the MDR1 genes. The first defines cellular conditions under which the levels of expression of the various MDR1 genes are modulated. The second involves use of the fluctuation analysis described by Luria and Delbrück (1943) to distinguish between variant cells arising due to mutations and variant cells arising due to adaptation. The third defines the cis-acting DNA sequences and trans-acting factors that influence MDR1 gene expression. Among those listed, the third approach was not applied in this study and accordingly will not be discussed.

5.1.1. P-glycoprotein selection/induction

A possible mechanism of drug resistance is gene amplification, which is defined as a mechanism whereby cells can generate multiple copies of discrete regions of their genome. The amplified genes are usually located either within expanded chromosomal regions (homogeneously staining regions or abnormally banded regions) or in extrachromosomal elements (double minutes). Studies have confirmed that the MDR1 genes are amplified in many MDR cell lines (Borst, 1991). On the other hand, selection for gene amplification is only possible when the gene is active. Silent genes with low or no transcriptional activity require transcriptional activation before selection for amplification. The MDR1 genes belong to this group of genes. Accordingly, a number of MDR cell lines have shown increased expression of MDR1 mRNA with little, if any, gene amplification (Roninson, 1992).

Gene amplification always arises by mutation, whereas increased transcription may arise by both mutation (e.g., in genes coding for regulatory proteins) and adaptation (Borst, 1991). Adaptation depends on a continuous presence of drug. Mutations can occur at random in both the presence and absence of drug. Drug treatment subsequently selects the resistant cells, and their progeny will dominate in recurring tumours. Understanding the manner in which MDR1 gene expression occurs may provide crucial information for the design of more rational chemotherapy for human cancers.

Expression of PGP was investigated in EHR2 after selection in vitro with DNR. Fluctuation analysis experiments were performed to assess whether selection or adaptation determined resistance and expression of PGP. Thirteen expanded populations of EHR2 cells were exposed to DNR 7.5 × 10 -9 M or 10 -8 M for 2 weeks. Surviving clones were scored and propagated. Only clones exposed to DNR 7.5 × 10 -9 M could be expanded for investigation (the clones exposed to DNR 10 -8 M were not viable probably due to cell cycle arrest; the dose applied was presumably too high). Compared with EHR2, the variant cells developed 2.5- to 5.2-fold resistance to DNR (mean 3.6-fold). PGP was significantly increased in only 44 percent of the clones. Thus possibly other resistance mechanisms were present in the variant clones. It also is possible that the variant cells could employ multiple, concurrent mechanisms of resistance. The nature of these mechanisms was not investigated. If resistance was acquired by adaptation via drug exposure, the number of surviving colonies would be expected to have a Poisson distribution, with the variance equal to the mean (Luria and Delbrück, 1943). The number of surviving EHR2 clones arising from the different populations showed a substantial fluctuation. The variance in the number of colonies was 14-fold greater than the mean, supporting the hypothesis that under these experimental conditions (5 log cell killing) spontaneous mutations conferred drug resistance.

Our result confirmed previous results reported by Sikic and coworkers investigating human sarcoma (MES-SA) cells exposed to DOX, paclitaxel, and VP16, respectively (Chen et al., 1994; Jaffrézou et al., 1994; Dumontet et al., 1996). In contrast, adaptation seemed to determine resistance in EHR2 cells exposed to DNR 10 -8 M. This latter finding was compatible with the result of Jaffrézou et al. (1994), exposing MES-SA cells to a very high dose of VP16 (5 μ M). These results suggest that at highly stringent conditions of selection (6 log cell killing), the resistance could be a result of adaptation.

In order to investigate whether PGP could be induced by acute treatment with DNR, we treated EHR2 cells and the revertant EHR2/0.8/R cells in vivo with DNR 6 mg/kg for 24 h. After treatment, PGP remained unchanged in EHR2 cells, whereas the protein increased significantly in EHR2/0.8/R cells (7-fold), and the cell line developed significant resistance to DNR (12-fold). Immunocytochemistry using the monoclonal antibody C219 showed a uniform weak immunoreactivity in treated EHR2/0.8/R cells, suggesting that in this cell line the mdr1 gene was activated by induction. This finding could be in accordance with either a direct activation of the mdr1 promoter as described by Kohno et al. (1989) or an indirect activation of the promoter by interaction with a regulatory protein (Gant et al., 1992). Previously, several authors have investigated the inducibility of PGP after treatment with cytostatics (Chevillard et al., 1992; Fardel et al., 1997). In contrast with our findings, these authors demonstrated substantially increased PGP in sensitive cells after acute treatment with anthracyclines (treatment with DOX for 24 and 4 h, respectively). The discrepancy probably could be explained by differences in the experimental design. Fardel et al. (1997) have shown the inducibility of PGP by anthracyclines to be dose-dependent. Both Chevillard et al. (1992) and Fardel et al. (1997) investigated the inducibility of the protein in vitro using very high doses of DOX [1.8 μ M for 24 h and 0.5 μ g/ml (approximately 0.8 μ M) for 4 h, respectively], doses that are not clinically relevant (Frost et al., 2002).

The literature is generally unclear on the exact events regulating the appearance of PGP-mediated resistance. Both selection of spontaneous mutations and adaptation have been described. Although mutation seems to be the most frequent mechanism, no convincing evidence has been presented to support the idea that one or the other mechanism occurs in all cellular/tumour models. Prior studies have been performed in many different cell lines using various drugs, exposures, and analytical approaches. It is possible that the sometimes contradictionally conclusions apparent in the literature could reflect such differences of experimental conditions. In addition, the MDR1/mdr1 genes have been shown to be tightly regulated in a cell-specific fashion, and expression of MDR1/mdr1 may relate to the growth or differentiation state of the tumour cells (Kane, 1996). Studies regarding the MDR1/mdr1 promoters have shown that the DNA sequences of human and rodent promoters differ greatly (Chin, Pastan, and Gottesman, 1993). Although some of these differences may not be functionally significant, the regulation of the resistance proteins may be distinct in human and murine cells.

In 1978-1979, the studies of somatic mutations by Law (1952) found clinical application through ideas predominantly developed by Skipper, Schabel, and Lloyd (1978) and Goldie and Coldman (1979). Goldie and Coldman (1979) suggested a mathematical model based on the following postulate: The number of resistant tumour cells is a function of the frequency of spontaneous mutations and the time at which treatment is initiated. Thus a greater tumour size will produce a greater number of resistant cells. Further, the chance of having two independent mutations in one cell is low. This model suggested the success of the simultaneous administration of drugs affected by different resistance mechanisms and helped inspire the development of multiple-drug regimens. In this study we provided an example of selection of variant cells that supports this model. Resistance is often multifactorial, and a tumour does not consist of completely sensitive or completely resistant cells but a continuous spectrum of cells with different levels of sensitivity (Borst, 1991). However, this finding does not detract from the basic soundness of this mutation-selection theory for understanding some aspects of tumour cell resistance.

5.1.2. Expression of resistance proteins in EHR2 cells exposed to fractionated irradiation

Drug resistance has been reported in patients after treatment not only with anticancer drugs but also with radiotherapy (Hill, 1991). Studies have shown that in vitro exposure of mammalian cells to irradiation may result in increased expression of PGP (Hill et al., 1990a, 1990b; McClean, Hosking, and Hill, 1993). This expression seemed to occur without any concomitant increase in expression of MDR1 mRNA (Hill et al., 1994). Further, the resistance profile seemed to differ from the profile seen after exposure to MDR drugs, suggesting that irradiation could induce a distinctive PGP-mediated phenotype (Hill et al., 1990a; McClean et al., 1993). However, in these studies, very high single doses of irradiation were applied (McClean, Hosking, and Hill, 1993: 30 Gy × 1; Hill et al., 1990a: 45-180 Gy in fractions of 9 Gy), making the results less clinically relevant. At present, little is known about the processes regulating MRP1. Further, the involvement of this protein in the phenotype of irradiated cells is unknown. Thus we designed this study to investigate the phenotype of irradiated cells using fractionated irradiation in clinically relevant doses.

The EHR2 cell line became resistant to VCR and VP16 but remained sensitive to DNR after exposure to 5 Gy × 12. The level of PGP increased approximately threefold. Previously, Hill et al. (1990a, 1990b) reported essentially identical results. The increased expression of PGP occurred without any increase in mdr1a and mdr1b mRNA. This observation confirmed results by McClean and Hill (1993). These authors demonstrated an increased half-life of PGP in irradiated cells, suggesting the increased PGP levels could be attributed to stabilization of the protein.

Drug-induced PGP possesses significant basal ATPase activity, and VER has constantly been reported to stimulate the activity. The ATPase activity of drug-induced PGP depends on Mg 2+ and is inhibited by vanadate and N-ethylmaleide (Senior, al-Shawi, and Urbatsch, 1995). In this study, the vanadate-sensitive ATP-ase activity of microsomes from irradiated cells was very low and comparable with the activity in sensitive cells. In addition, the activity could not be stimulated by VER. Thus our findings contrast with previous findings in drug-selected MDR cell lines. One possible explanation could be that the amount of PGP in EHR2/irr was low compared with the amount of the protein in drug-selected cells; thus its activity may account for only a small part of the ATPases in the membrane preparations. Previously, Litman et al. (1997) showed the drug-stimulatable ATPase activity to be directly proportional to the amount of PGP in a series of DNR-selected EHR2 cell lines. On the other hand, the result might indicate that PGP induced by irradiation differs from drug-induced PGP. Several possible explanations exist. A suggestion could be that the protein was changed by a given mutation. Specific site-directed mutations in certain transmembrane regions and cytoplasmic loops of PGP have been shown to alter both drug resistance and drug-stimulated ATP hydrolysis profiles (Gros et al., 1991; Germann, 1996). Alternatively, cellular factors could either directly or indirectly interact with the protein, stabilizing it and obscuring some of the drug-binding sites. Although the significance of phosphorylation in drug resistance has been a matter of much debate (see Sec. 5.3.1), it is possible that the phosphorylation of PGP could be changed in irradiated cells.

In this study, both the protein and mRNA levels of mrp1 increased approximately eightfold after irradiation. The increased expression of mrp1 could be independent of the damaging agent because mutation or gene amplification during the proliferation of cells after irradiation could explain the findings. However, in accordance with our results, Harvie, Davey, and Davey (1997) reported increased expression of MRP1 mRNA in human T-cell leukaemia cells (CEMRR) after exposure to irradiation. Comparing EHR2/irr with either MDR cells expressing murine mrp1 or cells transfected with mrp1 (Stride et al., 1996; Stride et al., 1997; Stride, Cole, and Deeley, 1999), it appeared that the resistance profiles and drug-transporting properties were similar. Furthermore, the ATPase activity of microsomes from irradiated cells was similar to that of mrp1-positive EHR2/VP16 microsomes (Nielsen et al., 2000a).

The processes regulating MRP1/mrp1 expression are not well defined. The human MRP1 promoter is extremely GC-rich and similar to many housekeeping genes. This promoter differs significantly from the MDR1/mdr1 promoters (Zhu and Center, 1994). Clearly, our results demonstrate that the mechanisms of PGP and mrp1 expression differ in irradiated murine cells. Thus the increased mrp1 mRNA found in this study could be explained by either increased transcription or posttranscriptional regulation, i.e., increased half-life of RNA in irradiated cells.

In order to determine the relevance of the two resistance mechanisms, the effect of the PGP inhibitor VER was investigated. It has been a general finding that the effect of VER on MRP1-positive cells is moderate or absent (Doyle et al., 1995). Thus VER should be expected preferentially to reverse PGP-mediated resistance. VER increased the cytotoxicity of VP16 in both EHR2 and EHR2/irr cells. The effect of VER was approximately the same in the two cell lines. Several authors have investigated the relationship of VP16 to the classic MDR phenotype (Watanabe et al., 1991; Sehested et al., 1992; Maare, 1997). Our results agree with the findings of Sehested et al. (1992), whereas Watanabe et al. (1991) found the chemosensitizing effect of VER to be greater in resistant than in sensitive cells. The results suggest that VER retains its chemosensitizing effect and that the resistant phenotype at least partly is mediated by increased PGP. Supporting this view, McClean et al. (1993) reported increased accumulation of [ 3 H]VCR after addition of VER to irradiated cells.

Several reports have suggested an important role of glutathione in MRP1-mediated drug efflux (O'Brien and Tew, 1996; Renes et al., 1999). Thus depletion of intracellular glutathione by treatment with buthionine sulphoximine (BSO), an inhibitor of γ -glutamylcysteine synthetase, increased the sensitivity of MRP1-positive cells (Schneider et al., 1995). In this study we observed only a minor but significant sensitization (approximately twofold) to VP16 after treatment with BSO. This sensitization ratio was significantly lower than the ratio reported by Schneider et al. (1995). One possible explanation could be that mrp1 only played a minor role in determining the resistant phenotype. Incomplete depletion of glutathione could not be excluded. Nevertheless, BSO treatment (50 μ M during 24 h) has been shown to decrease the glutathione level by 80 to 90 percent (probably a near-total depletion of cytosolic glutathione) in DOX-resistant MCF-7 cells with significantly increased glutathione peroxidase (Dusre et al., 1989).

In conclusion, the results show that both PGP and mrp1 increased after irradiation. Although a substantial increase in mrp1 was demonstrated, functional analyses suggest that PGP-mediated resistance was the dominating resistance mechanism in irradiated EHR2 cells.

5.1.3. Expression of resistance proteins in cell lines selected in combinations of daunorubicin and chemosensitizers

Several in vitro and in vivo studies have shown that various agents are capable of circumventing PGP-mediated MDR in tumour cells (reviewed by Ford, 1996; Krishna and Mayer, 2000). The compounds that modulate or reverse MDR represent a wide range of chemical structures, including calcium channel blockers, calmodulin inhibitors, glucocorticoids, immunosuppressives, and tranquilizers. Generally, these agents by themselves have no anticancer activity but are able to potentate the activity of the MDR drugs. The exact nature of the mechanisms of reversal is unclear. There is considerable evidence; however, that some of the chemosensitizers may interact directly with PGP by competitive inhibition.

Studies of many chemosensitizers have demonstrated that most appear to function in a manner similar to that of VER, although with differing potencies and perhaps at different sites on the PGP molecule (Ford, 1996). Therefore, the effects of VER on evolution of the drug-resistant phenotype may serve as a paradigm for most chemosensitizers. On the other hand, several hydrophobic cyclic peptides have been found to possess potent activities for modulating MDR. These peptides have distinctly different pharmacologic and structural properties as compared with VER. Representative of this class of chemosensitizers is CsA, an immunosuppressive agent used widely in human organ transplantation (Ford, 1996). This study was initiated to examine the influence of different chemosensitizers on the drug-resistant phenotype.

In this study, three sublines were developed from EHR2 and six sublines from the PGP-positive EHR2/ DNR cell line. The sublines were developed by treatment with DNR, a combination of DNR and VER, or a combination of DNR and CsA. The mdr1a mRNA levels increased in all but one drug-selected subline, whereas mdr1b mRNA remained unchanged. The results suggest either a selective growth advance of cells that overexpress mdr1a as compared with cells that overexpress mdr1b (Lothstein et al., 1989) or, alternatively, that anthracyclines might select for increased mdr1a . Previously, Raymond et al. (1990) demonstrated increased expression of mdr1a mRNA in 7 of 12 independently derived MDR cell lines, of mdr1b mRNA in 3, and of both in 2. Two sublines selected in DOX had increased mdr1a mRNA levels only.

The chemosensitizers seemed to suppress PGP in six individually selected sublines. Studies focusing on expression of MDR1 /PGP after the addition of VER or CsA to the chemotherapeutic treatment have shown conflicting results. Several results are in accordance with ours. Thus Beketic-Oreskovic et al. (1995), who used Luria and Delbrück fluctuation analysis to determine the mutation rate and resistance mechanisms after selection with DOX in the presence of the cyclosporine PSC 833, concluded that coselection with this cyclosporine reduced the mutation rate for DOX-selected resistance by 10-fold and suppressed the emergence of MDR1 mutants. Muller et al. (1994) reported a decreased PGP level in a human MDR leukaemia cell line selected with VER, and Futscher et al. (1996) found that inclusion of VER during drug selection with DOX prevented the emergence of PGP. These results apparently contradict the findings of others (Herzog et al., 1993; Dietel et al., 1994), who reported an increase in the expression of MDR1 /PGP in cell lines treated with different calcium channel blockers or CsA. It should be noted, however, that Herzog et al. (1993) demonstrated that increased differentiation was associated with the increases in MDR1 . Furthermore, Dietel et al. (1994) used an in vitro -selected Friend leukaemia cell line (F4-6RADR-CsA) that was developed by continuous incubation for more than 6 months with increasing doses of CsA (the dose was increased 90-fold). At any rate, the discrepancies emphasize the fact that the effect of various drugs on expression of MDR1 / PGP could be highly dependent on the cell type (i.e., capability of differentiation), the model used, and probably the method of measuring MDR1 /PGP.

In this study, the mdr1a mRNA level was significantly decreased in two cell lines (EHR2/1.6V and EHR2/1.6C) selected from the sensitive cell line EHR2, whereas the mdr1a mRNA levels in four VER- or CsA-selected sublines were similar to the levels in the parent EHR2/DNR cell line and the DNR-selected counterparts. The dynamic range of the RT-PCR could explain these findings because this technique is not able to show differences in cells with high levels of mRNA. Our results suggest that selection in the presence of chemosensitizer decreases the expression of mdr1a . The mechanism by which the decreased mdr1a occurred is unknown because we did not investigate mutation or transcription rate of mdr1a . At any rate, previous studies have shown that both mechanisms could account for the results (Beketic-Oreskovic et al., 1995; Muller et al., 1995).

Compared with EHR2, the mRNA and protein levels of mrp1 remained stationary in all selected cell lines. In accordance, neither Futscher et al. (1996) nor Abbaszadegan et al. (1996) found changes in the expression of MRP1 in VER-coselected cell lines.

One subline (EHR2/1.6V) selected in DNR + VER developed changes in TOPO II. Furthermore, Futscher et al. (1996) demonstrated decreased TOPO II level in myeloma cells selected in a combination of VER and DOX. VER may select for TOPO II-mediated resistance. It should be emphasized, however, that this resistance mechanism frequently has been reported in MDR cell lines selected in cytostatic alone (Friche et al., 1991; Skovsgaard et al., 1994). Since only one of six DNR + chemosensitizer-selected sublines developed changes in TOPO II, we suggest that neither VER nor CsA influenced the expression of TOPO II.

Comparing the DNR- and DNR + VER-selected cell lines; it appeared that the resistance was increased in two of three VER-selected cell lines. In accordance, Formelli et al. (1988) reported increased resistance of melanoma cells selected in DOX and VER. A decreased TOPO II α level could explain the findings in EHR2/ 1.6V, and we did not investigate for changes in TOPO II activity. Furthermore, in all sublines a reasonable correlation between expression of PGP and accumulation of DNR was demonstrated, suggesting that PGP was the predominant drug-transporting resistance mechanism. At any rate, increased resistance following selection with DNR in the presence of VER could be a consequence of induction of other resistance mechanisms.

It is noteworthy that the resistance was significantly lower in all DNR + CsA-selected sublines as compared with the DNR- and DNR + VER-selected counterparts ( Figure 4 ).

In a murine model system, CsA seemed to prevent both PGP expression and development of resistance. Additionally, the effect of VER and CsA on cytotoxicity was retained in all cell lines treated with a chemosensitizer. The duration of treatment in the present protocols was rather short. Thus resistance may develop after additional treatment. However, addition of CsA to the cytostatic treatment seemed at least to delay development of resistance. Essentially similar results have been reported by Beketic-Oreskovic et al. (1995), who used fluctuation analysis to investigate MDR1 expression and resistance in human tumour cells. Clinical trials combining a chemosensitizer with cytostatics are in progress (Ferry, Traunecker, and Kerr, 1996). Basically, two strategies have been adopted. Most often patients with drug-refractory disease were retreated with the primary treatment combined with a chemosensitizer. A second option was to perform trials in previously untreated patients with tumour types in which it was a reasonable assumption that the tumours contained a proportion of PGP-positive cells. However, apart from a few isolated case reports, these trials have failed to demonstrate conclusively that modulation of the MDR phenotype is of any benefit (Robert, 1999). One should be cautious when interpreting the relevance of results obtained in model systems. Nonetheless, the present and other preclinical studies allow speculation that the use of cyclosporines (CsA or its analogue PSC 833) in initial treatment regimens could have the advantage of delaying/preventing the emergence of PGP and resistance in tumour types known commonly to develop increased PGP. Examples include breast cancer, lymphoma, leukaemia, neuroblastoma, and multiple myeloma. Thus clinical trials including cyclosporine early in the course of disease should be considered in patients with potentially PGP-negative, drug-sensitive tumours. The major clinical end point of such trials would be duration of remission and time to progression (Gottesman, Fojo, and Bates, 2002).

5.2. Resistance Mechanisms in Cell Lines
Selected in Topoisomerase II Inhibitors

5.2.1. Resistance mechanisms in an etoposide-selected cell line

The epipodophyllotoxins VP16 and teniposide (VM26) are derivatives of podophyllotoxin. These drugs belong to the TOPO II inhibitors, and their chemotherapeutic action correlates with their ability to stabilize the cleaved TOPO II-DNA complex (Jensen et al., 1990). VP16 has an established activity against a wide range of cancers and is one of the prevailing cytostatics in the clinic today. However, there are only a limited number of reports in which a drug-resistant tumour cell line has been developed following specific exposure to an epipodophyllotoxin (reviewed in Prost, 1995; Yamada and Ando, 1996). In most cell lines, only one resistance mechanism has been investigated. Therefore, it seemed interesting to dissect the importance of the different resistance mechanisms in a VP16-selected EHR2 subline. Generally, VP16- or VM26-selected cell lines show alterations in TOPO II content and/or activity. Likewise, we demonstrated a 30 to 40 percent reduction in both TOPO II α and TOPO II β . TOPO II β was identified only recently. Accordingly, most studies have either measured the total amount of TOPO II or focused on TOPO II α (Prost, 1995; Yamada and Ando, 1996). The inability to detect TOPO II β probably has been related to the enhanced susceptibility of the enzyme to proteolysis during sample preparation or the use of antibodies that did not recognize TOPO II β . Recently, Brown et al. (1995) investigated the correlation between expression of TOPO II α and II β and resistance to DOX and VP16, respectively. Their findings suggested that the cytocidal activity of the two drugs could be mediated, at least in part, by TOPO II β .

Although tumour cell lines with alterations in TOPO II as the only resistance mechanism have been described, most cell lines resistant to TOPO II-inhibiting agents have more than one mechanism of resistance (Sullivan and Ross, 1991). Previously, a few epipodophyllotoxin-selected cell lines with increased expression of PGP were described (Brock et al., 1995). However, this is not a general finding, and it has been suggested that the epipodophyllotoxins rarely select for PGP expression (Koike et al., 1996). On the contrary, there are several indications that epipodophyllotoxins may select for MRP1 expression (Schneider et al., 1994; Doyle et al., 1995; Koike et al., 1996). The summarized data are in accordance with the result of this study. Thus EHR2/VP16 displayed 20-fold increased expression of mrp1 mRNA, whereas PGP was unchanged as compared with EHR2. VER produced a moderate degree of sensitization of EHR2/VP16 cells to DNR and VP16, whereas CsA produced only a modest sensitization to DNR and no sensitization to VP16. Our results are in concordance with those of others because it has been a general finding that the effect of chemosensitizers in MRP1-positive cells is moderate or absent (Doyle et al., 1995).

5.2.2. Resistance mechanisms in a mitoxantrone-selected cell line

The dihydroxyanthracenedione derivative MITOX is an intercalating agent that inhibits TOPO II and induces double-strand breaks. The drug is used widely in haematologic diseases, i.e., chronic lymphocytic leukaemia and chronic myelogeneous leukaemia (Jacobs and Wood, 2002; Axdorph et al., 2002), whereas its use in solid tumours has been limited.

Only a limited number of MITOX-selected tumour cell lines have been described (Miyake et al., 1999). The data available suggest that resistance to MITOX in tumour cells is multifactorial (Hazlehurst et al., 1999). Most MITOX-resistant tumour cell lines have an atypical resistance profile (Litman et al., 2000), suggesting that the resistance mechanism may well be different from previous known mechanisms. Thus this study was performed to elucidate the resistance mechanisms in a MITOX-resistant cell line.

In accordance with Litman et al. (2000), the EHR2/ MITOX subline displayed an atypical resistance profile showing a very high degree of resistance to the selecting agent, MITOX, moderate cross-resistance to DNR and VP16, and retained sensitivity to VCR.

The amount of immunoreactive TOPO II α in EHR2/MITOX was reduced to one-third relative to that in EHR2 cells, whereas TOPO II β could not be detected in EHR2/MITOX. Previously, Harker et al. (1991) demonstrated a marked reduction, if not absence, of immunodetectable TOPO II β protein in MITOX-resistant leukaemia cells (HL-60/MX2). An explanation could be that MITOX preferentially produces cytotoxic damage via TOPO II β rather than TOPO II α (Errington et al., 1999). Nevertheless, decreases in TOPO II β level or complete disappearance of the protein also have been reported in DOX-, amsacrine-, and 9-hydroxy ellepticine-resistant cell lines (Prost, 1995). Further, Kellner et al. (1997) were unable to detect decreases in TOPO β in a MITOX-resistant gastric carcinoma cell line.

Compared with EHR2, EHR2/MITOX showed increased expression of mrp1 mRNA, whereas PGP was decreased. In contrast, Futscher et al. (1994) found that none of four independently MITOX-selected cell lines exhibited increased expression of PGP or MRP1, and Yang et al. (1995) did not find the two proteins in a MITOX-resistant MCF7 cell line.

Previously, LRP was detected in two MITOX-resistant cell lines investigated by Futscher et al. (1994). However, immunocytochemical analysis using the monoclonal antibody LRP-56 (Scheper et al., 1993) did not show increased lrp in EHR2/MITOX as compared with EHR2. As stated previously, it could not be excluded that LRP-56 did not recognize murine major vault protein.

Recently, Eriksen (personal communication) demonstrated increased expression of mxr/bcrp/abcp ( abcg2 ) in EHR2/MITOX. Further, Allen et al. (1999) found murine mxr to be functionally comparable with human MXR, strongly arguing that murine models are highly informative in understanding the pharmacologic role of MXR.

The EHR2/MITOX subline clearly showed an ATP-dependent reduction of DNR and MITOX accumulation. Compared with EHR2, the EHR2/MITOX cell line had significantly increased efflux of DNR. A comparison of EHR2/MITOX with PGP-positive cell lines showed several similarities in the transport of DNR. However, addition of VER had only minor influence on DNR transport in EHR2/MITOX cells, whereas it has been a general finding that VER corrects accumulation defects in PGP-positive cells. A comparison of EHR2/MITOX with mrp1-positive cell lines also showed several similarities in the transport of DNR. However, EHR2/MITOX cells had a significant constitutive ATPase activity, which was stimulated by DNR but not by VBL. Furthermore, the apparent K i value for inhibition of the ATPase activity by vanadate did not differ from the K i value previously obtained for PGP-positive cells. In contrast, both the total and in particular the vanadate-sensitive ATPase activity of mrp1-positive cells were very low (see Sec. 5.4). Various groups have identified cell lines selected in MITOX that display phenotypes comparable with EHR2/MITOX (Dalton et al., 1988; Harker et al., 1989; Taylor et al., 1991; Nakagawa et al., 1992; Kellner et al., 1997; Lee et al., 1997). In all investigated sublines, the drug kinetics was identical. Recently, increased expression of MXR has been demonstrated in several of the cell lines (Doyle et al., 1998; Allikmets et al., 1998; Miyake et al., 1999; reviewed in Litman et al., 2001 and Bates et al., 2001).

Previously, Dietel et al. (1990) found an intensive formation of membrane-coated vesicles containing cytostatic drug to be associated with resistance in the MITOX-resistant EPG85-257 cell line. Redistribution of drug away from the target sites may be a significant factor in determining resistance. The nuclear/total cellular DNR fluorescence ratio was, however, similar in EHR2 and EHR2/MITOX. In accordance with our result, Lee et al. (1997) found no apparent alterations in the subcellular DNR distribution in MCF7 AdVp cells expressing 100,000-fold resistance to MITOX. Thus it is unlikely that cross-resistance between DNR and MITOX in EHR2/MITOX is due to changes in the subcellular distribution of DNR. This finding, however, does not exclude the possibility that the marked resistance to MITOX could be explained by trapping of drug within the Golgi complex. Thus Fox and Smith (1995) have suggested that cytoplasmic sequestration may be an important determinant of MITOX resistance. Furthermore, studies with the lysosomal probe LysoTracker Green have confirmed the localization of MITOX in acidic vesicles in both parental cells and resistant cell lines (Litman et al., 2000).

5.3. P-glycoprotein-Mediated Transport

5.3.1. Correlation between expression of P-glycoprotein
and drug transport

Identification of PGP as a member of a superfamily of transport proteins and the many identical reports of ATP-dependent drug efflux from MDR cells suggested a direct role of PGP in the efflux of chemotherapeutic agents out of the cell. When this project was initiated, no systematic studies of the relationship between PGP and the mechanism of drug kinetics in several MDR cell lines had been performed. Comparing one resistant cell line with its sensitive counterpart might yield important information about cellular changes accompanying development of resistance. However, only by comparing sublines with different degrees of resistance one can determine whether any property is invariably associated with resistance (and other cellular changes) in a predictive and quantitative way.

Five MDR cell lines were established by treatment of EHR2 cells with different doses of DNR (0.1, 0.2, 0.4, 0.8, and 1.6 mg/kg four times weekly). In all cell lines, expression of PGP, degree of resistance, net accumulation of DNR, and unidirectional influx and efflux were measured. We did not investigate for other resistance mechanisms, i.e., change in TOPO II or increased expression of mrp1 .

In each individual subline, an initially linear correlation between PGP and steady-state accumulation was demonstrated (Nielsen et al., 1994). However, continuous measurements in four of the five sublines showed significantly increased expression of PGP without concomitant changes in the drug accumulation. Moreover, a comparison of the individual cell lines showed a four- to fivefold variation in PGP with approximately similar steady-state accumulation. In contrast, transfection studies have shown that the density of PGP in the plasma membrane determines the level of MDR (Choi et al., 1991) and that PGP expression is inversely related to the accumulation of chemotherapeutic drugs (Chen and Simon, 2000). In order to further evaluate the kinetics of DNR transport, we measured DNR sensitivity, PGP expression, and efflux of DNR in EHR2 cells and five DNR-resistant sublines [one passage from EHR2/0.1 (passage 72), EHR2/0.2 (passage 24), and EHR2/0.4 (passage 54) and two passages from EHR2/0.8 (passage 12 and 72)] (Nielsen, Maare, and Skovsgaard, 1994). In four of the sublines, a reasonable correlation between resistance, PGP expression, and PGP-mediated efflux capacity was established. In one subline (passage 12 of EHR2/0.8), however, a high amount of PGP was found despite a low degree of resistance and a low efflux capacity. Our results support the hypothesis that PGP is a transport protein. However, the actual level of PGP does not, by itself, determine the efflux capacity. Other factors regulating the efficacy of PGP also seem to be of importance. First, the antibody C219 used for detection of PGP recognizes PGP encoded by mdr1a and mdr1b as well as mdr2 . Since only mdr1 encodes for functional PGP, it cannot be excluded that increased levels of PGP due to mdr2 gene expression could confound the results. Further, studies of murine MDR J774.2 cells have shown that the mdr1a and mdr1b are transcripted to two different precursors of PGP that possess different affinities for metabolites (Yang et al., 1990). All the present cell lines have not been investigated systematically for mdr1a and mdr1b expression. However, investigations of the mdr1a and mdr1b mRNA levels in EHR2/1.6, EHR2/0.1, and cell lines selected in a combination of DNR and chemosensitizers (Nielsen et al., 2002) have shown increased mdr1a mRNA but unchanged mdr1b mRNA, indicating that DNR selects primarily for increased mdr1a mRNA.

Theoretically, concomitant changes in mrp1 could confound the results. Thus investigations of EHR2 and EHR2/0.4 have shown expression of mrp1 in EHR2 and increased expression of mrp1 (approximately threefold) in early passages (passage 12) of EHR2/0.4 as compared with EHR2. In contrast, expression of mrp1 seemed to decrease (approximately 1/10 of the amount in EHR2) in later passages of EHR2/0.4 (passage 54). It should be noticed, however, that the affinity of murine mrp1 for DNR is very low (see Sec. 5.4).

The biologic activity of PGP could be changed. The role of posttranslational modifications in PGP activity is not well elucidated. PGP has three carbohydrates on the first extracellular loop. There is evidence that glycosylation is not required for drug transport activity (Beck and Cirtain, 1982); however, it could change the stability of the protein (Schinkel et al., 1993). PGP is phosphorylated as well. In a recent report, Goodfellow et al. (1996) demonstrated indistinguishable transport properties of mammalian cells expressing wild-type PGP and cells expressing PGP mutated at all potential phosphorylation sites. These authors concluded that phosphorylation unequivocally had no role in modulating drug transport activity. Nevertheless, it cannot be excluded that phosphorylation affects other aspects of PGP function, e.g., specificity, quaternary structure, or processing/turnover rate.

Finally, alterations in the plasma membrane lipids could modulate drug accumulation and resistance. Our group (Litman et al., 1995) has investigated the lipid composition of different passages of the DNR-selected EHR2 cell lines. Although significant variation was found during development of resistance, the composition of the lipids in the most resistant cell line (EHR2/0.8) was not significantly different from the composition in the EHR2 cell line. Thus alterations in the lipid composition may influence but do not determine the resistance in the present cell lines. At any rate, several studies have reported that changes in the lipid bilayer of the plasma membrane could alter drug and/or nucleotide binding (reviewed in Litman et al., 2001). Recently, Romsicki and Sharom (1999) found that alterations in the plasma membrane lipid environment altered the affinities of VBL, VER, and DNR to PGP and also affected ATP binding and hydrolysis, suggesting that the plasma membrane may participate directly in the interaction between nucleotide-binding domains and drug-binding sites of PGP.

5.3.2. Kinetics of P-glycoprotein-mediated drug transport

There is still controversy about the kinetics of PGP-mediated drug transport. Consequently, we sought a method to describe this process in detail. Siegfried, Burke, and Tritton (1985) reviewed the current literature and found drug efflux from sensitive cells to be compatible with simple diffusion. Theoretically, the solubility/diffusion model adequately describes drug efflux from wild-type EHR2 cells (Stein, 1997, 1998). In contrast, several authors have demonstrated that efflux from PGP-expressing cells (Spoelstra et al., 1992; Frezard and Garnier-Suillerot, 1991a, 1991b) is composed of two main components: Diffusion and active efflux. These observations prompted us to describe the drug kinetics in sensitive and MDR cells using the following assumptions: Efflux by diffusion could be described by first-order kinetics given by the equation C t = C 0 × e - kt (where C = intracellular concentration of DNR, t = time and k = a constant). Efflux from PGP-positive MDR cells could be described by biexponential kinetics using the equation Ct = A × e - kt + B × e - qt (where C = intracellular concentration of DNR, A + B = C 0 , t = time, and k and q = constants). A curve-fitting computer programme (Sigma Plot, version 5.0) was used to fit the efflux data to the functions and to calculate A, B, k , and q (A = B, k = q for first-order kinetics) in EHR2 and the five DNR-selected, PGP-positive EHR2 sublines [the calculations are given in detail in Nielsen (1994) and Nielsen, Maare, and Skovsgaard (1994)] ( Figure 5 and Figure 6 ).

Fitting efflux data for EHR2 to monoexponential kinetics, we obtained a correlation of 0.99, making it reasonable to conclude that diffusion was the predominant efflux component from sensitive cells (EHR2 showed a low level of PGP). Likewise, the experimental data for the DNR-resistant cell lines in glucose-enriched medium fitted well with the equation for biexponential kinetics. In the DNR-resistant cells, the efflux was composed of an initial fast component followed by a significantly slower component.

The rapid-efflux component was blocked completely by VER, and it seemed reasonable to assume that this component represented PGP-mediated efflux. However, Stein (1997) has reviewed our data concerning efflux of DNR from EHR2 and EHR2/0.8 (passage 72). He found by a systematic analysis of the variance between data and curve fit that the best fit for the efflux curve for sensitive cells was two exponential functions and a linear component (y = Ae -Bt + Ce - Dt + E , where A, C , and E had the values 6.975, 46.58, and 46.44 percent and B and D were 20.61 and 0.102 min -1 , respectively), and the best fit for the resistant cell line was three exponential functions (y = Ae-Bt + Ce-Dt + Ee-Ft , with A, C, and E having the values of 26.93, 36.72, and 36.35 percent and B , D , and F having the values of 36.72, 0.809, and 0.092 min -1 , respectively). It is correct that the best fit is obtained by these equations, and we completely agree that efflux is a complex process influenced by several factors, e.g., drug binding [for a detailed discussion see Stein (1997)]. In this study, we did not use the zero-points for computer calculations. All calculations were performed using the relative amount of drug because these values only showed little variation, whereas determinations of the exact amount of drug showed considerable variation (especially zero-points). Thus our model did not account for 14, 12, 27, 19, and 32 percent of drug efflux in EHR2/0.1, EHR2/0.2, EHR270.4, and EHR2/0.8 passages 12 and 72, respectively. Besides, we did not try to obtain the best-fitting parameters but used theoretical predictions to obtain approximate parameters, making it possible to compare resistant cell lines developed under similar conditions. Moreover, in analyzing the equation for resistant cells (y = Ae-Bt + Ce-Dt + Ee-Ft ) suggested by Stein (1997), we suggest that Ce-Dt describes the fast-efflux component (PGP-mediated efflux) and that Ee-Ft describes the slow-efflux component (diffusion). Thus the "efflux constants" obtained in our study could be compared with "efflux constants" suggested by Stein (1997). Knowing that such calculations should be evaluated with considerable reservation, all calculations concerning efflux from DNR-selected cell lines were repeated (Sigma Plot, version 7.0). The results are given in Table 1. A comparison of our results with the result using the equation suggested by Stein (1997) shows approximately similar values for the efflux constants.

Exocytosis could contribute significantly to the efflux. However, Demant, Sehested, and Jensen (1990), who used computer-generated simulations to analyze some possible mechanisms controlling DNR fluxes, found the exocytotic model to be physically impossible. Furthermore, Lankelma et al. (1990) demonstrated efflux to be independent of increased intravesicular pH.

In this study, efflux was found to follow Michaelis-Menten kinetics. The efflux from resistant cells was found to saturate [efflux half-saturated (K m ) at an external concentration of DNR of approximately 15 μ M]. Since we used very high DNR doses to achieve this saturation, however, the conclusion should be interpreted with caution because self-association or unspecific toxicity (damage to the membrane) of the anthracycline molecules could confound the results (Chaires, Dattagupta, and Crothers, 1982). However, the conclusion agrees with that of several investigators (Watanabe, Inaba, and Sugiyama, 1989; Cazenave and Robert, 1991; Spoelstra et al., 1991). Among these, Spoelstra et al. (1991) found PGP-mediated efflux to saturate with half-maximal rate at 0.8 to 3.3 μ M free intracellular DNR. In another study, Frezard and Garnier-Suillerot (1991a) reported that the efflux of DNR increased almost linearly with the free cytoplasmic concentration of drug in the 0 to 2.5 μ M range. A direct comparison of the studies is difficult due to the different experimental conditions and the indirect method used to calculate the free concentration of cytoplasmic DNR.

5.3.3. P-glycoprotein-drug interaction

Pivotal for the mechanism of MDR resistance is (1) where in the cell and (2) when in the course of drug transport PGP interacts with its substrate. Yet there is no published model for the action of PGP that fully explains the prevailing kinetic and physicochemical data. According to the classical model for MDR, drug molecules enter the cell by passive diffusion and bind to PGP on the cytoplasmic side of the plasma membrane. PGP then acts as a carrier that actively extrudes drugs through a pore or channel in the membrane formed by transmembrane domains of the molecule, leading to drug transport directly from the cytosol to the extracellular medium (conventional transport hypothesis) (Chen et al., 1986). This model, however, requires enormous flexibility of PGP's substrate-recognition site(s) to explain the exceptional broad specificity of the protein for chemically unrelated compounds.

Alternatively, PGP could recognize the lipophilic drugs by their physical property of partitioning into the phospholipid bilayer of the plasma membrane. Two models have been proposed. The "vacuum cleaner" model proposes that drug molecules are pumped by PGP from the lipid bilayer of the plasma membrane to the extracellular medium. Primarily, this model was based on studies demonstrating that PGP could reduce the apparent concentration of DOX in the plasma membrane through the use of energy transfer from DOX to a photoactivatable hydrophobic probe (Raviv et al., 1990). In 1992, Higgins and Gottesman (1992) suggested that PGP might function as a "flippase". In this model, drugs are transported from the inner to the outer leaflet of the lipid bilayer, from which they diffuse into the extracellular medium. Drug recognition within the plasma membrane is supported by a number of observations: First, the lipophilicity and partition coefficient of a drug seem to be key factors in determining the efficacy of PGP. Second, PGP reduces the concentration of DOX within the plasma membrane through energy transfer from DOX to a photoactivatable hydrophobic probe (Raviv et al., 1990). Third, MDR cells have been shown to pump out esters of fluorescent probes before they gained access to intracellular esterases (Homolya et al., 1993; Hollo et al., 1994). Fourth, Shapiro and Ling (1997) measured the kinetics of Hoechst 33342 using PGP-enriched plasma membrane vesicles from Chinese hamster ovary cells. Because Hoechst 33342 is fluorescent when bound to the membrane but not when in the aqueous medium, it was possible to determine the movement of the dye out of the membrane by monitoring the fluorescence intensity. The initial specific rate of transport was directly proportional to the amount of dye in the lipid phase and inversely proportional to the concentration in the aqueous phase, demonstrating that PGP extracted Hoechst 33342 from the lipid membrane. Lastly, kinetics studies by Stein et al. (1994) have supported the vacuum cleaner model. They suggested that the PGP molecule encounters all substrates within the plasma membrane but extracts these substrates differentially out of the two halves of the membrane bilayer.

This study was designed to further investigate the above-mentioned hypotheses. Both models predict that the initial influx should reflect PGP-mediated efflux and thus be significantly reduced in PGP-positive cells. Influx of DNR was measured in five DNR-resistant cell lines (EHR2/0.1, EHR2/0.2, EHR2/0.4, and passages 12 and 72 of EHR2/0.8) (Nielsen, Maare, and Skovsgaard, 1995). Compared with EHR2, the influx was significantly decreased in all but one of the resistant cell lines. Similar data have been reported by Sirotnak et al. (1986), who investigated VCR uptake in DC-3F Chinese hamster lung cells, and by Pereira et al. (1994), who investigated uptake of DOX in K562 leukaemia cells. In theory, several cellular changes could account for the decreased influx. Insertion of large quantities of PGP in the plasma membrane could have a pronounced effect on the structural order of the membrane and thereby on the diffusional permeability. Since the EHR2/0.8 subline (passage 12) had a minimal decrease in influx despite a high level of PGP, it is unlikely that this mechanism alone could explain the results. Alterations in the plasma membrane lipids could modulate drug influx. Previously, this topic has been discussed in relation to efflux (see Sec. 5.3.1), and the same arguments could be put forward with regard to influx (Litman et al., 1995). Thus alterations in the lipid composition may influence but do not determine the influx in the present cell lines. Changes in the fluidity of the plasma membrane may contribute to the decreased permeability. Supporting this hypothesis, VER has been shown to have an ability to interact with and perturb membranes (Cano-Gauci and Riordan, 1987) and was in this study found to increase influx significantly in energy-depleted cells. Clearly, these explanations cannot be excluded. However, significantly decreased influx has been found in MDR1 -transfected cells differing only from sensitive cells by the increased level of PGP (Shalinsky et al., 1993; Stein et al., 1994). Thus, using KB-GRC1 cells transfected with MDR1 , Shalinsky et al. (1993) found significantly reduced influx of VBL that equalized after addition of VER or depletion of metabolic energy as compared with sensitive cells. Stein et al. (1994) used NIH/3T3 cells transfected with the wild-type and mutant MDR1 gene, respectively. In both cell lines, a significantly reduced influx of VBL that was dependent on metabolic energy was demonstrated. Hence the decreased influx could represent increased PGP-mediated efflux. The following results favoured this hypothesis: First, in all cell lines a linear correlation between influx and efflux was demonstrated ( r = 0.97) ( Figure 7 ). Second, in PGP-expressing cells the influx was decreased even after 6 s ( Figure 8 ). This was well before sufficient drug had accumulated within the cells to make outward pumping substantial, whereas the concentration of drug in the plasma membrane probably was high due to the hydrophobic nature of the drug [see Stein (1998) for a detailed discussion]. Third, a comparison of the effect of VER on influx and efflux revealed that VER affected influx preferentially (VER 3.3 μ M equalized influx in highly resistant cells, whereas VER 5.5 μ M was needed to equalize efflux in the same cells when compared with sensitive cells).

On the other hand, Sirotnak et al. (1986) found indistinguishable uptake rates of DNR in sensitive and resistant CHO cells, and Stein et al. (1994) reported similar uptake rates for DNR in 3T3 mouse fibroblasts transfected with human MDR1 . The reason for the different results is by no means clear. The methods used by Sirotnak et al. (1986) were essentially identical to ours. These authors, however, did not correct DNR data for surface-bound drug. Stein et al. (1994) stopped uptake by rapid washing in ice-cold phosphate-buffered saline. Thus these authors did not use centrifugation for terminating the flux reaction. Further, differences in incubation conditions could be an explanation for this discrepancy. Finally, the ability of PGP to extract drugs probably depends on the cell type, the structure of PGP itself, and the nature of the drug.

The original vacuum cleaner model suggests that drugs could be removed from both leaflets of the plasma membrane (Stein, 1997). More recently, however, Chen, Pant, and Simon (2001) used MDR1 -transfected HeLa cells to demonstrate that PGP extracted the fluorescent lipid analogue TMA-DPH from the inner and not the outer leaflet of the plasma membrane. Eytan and Kuchel (1999) reviewed a survey of studies of the influx and efflux of chemotherapeutic drugs in both sensitive and PGP-positive cells. They combined the results with computer-generated simulations and demonstrated that MDR drugs (DOX was used as a model drug) probably may transverse the plasma membrane by a flip-flop mechanism and concluded that the present data indicated that PGP functions as a flippase and removed the drugs from either the inner leaflet of the membrane or the cytoplasm. An explanation of the different results could be that PGP handles substrates differently. Thus the protein probably could extract some substrate from the inner leaflet and others from the outer. Further elucidation of the hydrophobic vacuum cleaner or flippase models will require use of biophysical techniques and substrates conjugated with spin probes.

Other models for PGP-substrate interaction have been suggested. According to Demant, Sehested, and Jensen (1990), the MDR pump is a typical enzyme system showing saturation with increased concentrations of substrates. This model also predicts the reduced initial uptake and the reduced steady-state uptake - which is in accordance with the characteristics of the MDR system. Anyhow, it is possible to distinguish between the enzyme model and the vacuum cleaner model. The enzyme model predicts that the initial ratio between the concentration of intracellular and extracellular drug depends on the external drug concentration and rises when the external concentration increases, becoming equal to the ratio without pumping. In contrast, the vacuum cleaner model predicts that the ratio of intracellular/extracellular drug concentration as a function of time is independent of the extracellular substrate concentration [for arguments, see Stein (1997)]. Using our data to calculate the ratio of intracellular/extracellular DNR, we found that over a tenfold range of extracellular DNR (2.5-25 μ M), the initial intracellular/extracellular DNR concentration in EHR2/0.8 (passage 72) was still well below that seen in EHR2 (absence of pumping). Thus the pump (PGP) was not saturated at the inner face of the membrane, and DNR seemed to be extracted from the membrane before it reached the intracellular compartment.

5.3.4. Effect of verapamil on efflux/influx

In this study, the ratio between VER and DNR seems to determine the relative inhibition of efflux in all the resistant cell lines; i.e., the probability that PGP was occupied by either VER or DNR depended solely on the relative substrate concentrations. Thus it was suggested that VER acts as a competitive inhibitor on efflux (Stein, 1986). This assumption was supported by a Lineweaver-Burk plot showing unchanged V max (double reciprocal plot of initial rapid efflux as a function of total intracellular content of DNR). In contrast, our data (Lineweaver-Burk plot showing unchanged K m ) suggest that VER acted as a noncompetitive inhibitor of influx. This finding supports the idea that the binding surfaces for substrates are extended over a larger protein structure of PGP (Borgnia, Eytan, and Assaraf, 1996; Boer et al., 1996). Our results confirmed the statement of Skovsgaard (1978a). Thus a V max for influx was obtained, indicating that influx is saturable. The results must be interpreted with caution because artifacts probably could explain in part these findings. Intracellular binding and intercalation could interfere with the efflux results, whereas the influx results could be confounded by self-association of the DNR molecules (Chaires, Dattagupta, and Crothers, 1982). At any rate, Yusa and Tsuruo (1989) reported similar results concerning drug efflux.

5.4. Multidrug Resistance-Associated
Protein 1-Mediated Transport

The analysis of cell lines that express MRP1 has raised several questions concerning the pharmacology associated with MRP1. Some drug-selected MRP1-positive and MRP1 -transfected cell lines have been reported to have a decreased drug accumulation and/or increased efflux (Slovak et al., 1988; Krishnamachary et al., 1994; Cole et al., 1994; Zaman et al., 1994; Doyle et al., 1995; Breuninger et al., 1995), suggesting that MRP1 may function as a drug efflux pump similar to PGP. However, the highly resistant H69AR cell line from which MRP1 was isolated initially did not display altered drug accumulation or efflux (Cole et al., 1991), and this has led to speculations that MRP1 may confer resistance by a mechanism that does not involve altered drug kinetics. Thus the present study was performed to investigate the drug kinetics in a cell line expressing the murine mrp1 solely (we only investigated PGP; other, as yet unknown transport proteins could be present). The EHR2/VP16 cell line had significantly increased [ 3 H]VP16 efflux ( Figure 9 ), whereas DNR efflux from EHR2/VP16 cells was only slightly increased as compared with the wild-type cell line ( Figure 10 ), suggesting that mrp1 had only little affinity for DNR. Our results were consistent with those of Stride et al. (1997), who compared the pharmacologic properties of human MRP1 and murine mrp1. Although the amino acid sequences of murine mrp1 and human MRP1 are 88 percent identical (Stride et al., 1996), some biochemical differences were demonstrated. Both proteins were found to transport VP16 and VCR. However, a significant transport of DNR was detected for cells transfected with human MRP1 , whereas efflux of DNR was only slightly increased in cells expressing murine mrp1 . Recent studies with hybrid murine/human proteins indicate that the structural causes of these functional differences reside in the COOH-proximal third of the protein (Stride, Deeley, and Cole, 1999).

It seems unlikely that methodological differences alone account for the considerable variation in the transport properties of MRP1-positive cell lines. Many observations were made in cell lines obtained by stepwise selection in drug. Thus multiple collateral resistance mechanisms could have developed. None of the resistant sublines had increased PGP; however, other resistance proteins could be present (Slovak et al., 1988; Krishnamachary et al., 1994; Doyle et al., 1995). Indeed, several MRP1 -expressing cell lines have been shown to have decreased TOPO II (Zijlstra, de Vries, and Mulder, 1987; Cole et al., 1991). Another possible explanation is that drug transport by MRP1 may be influenced by cellular components that may be cell type-specific. An alternative explanation is that the activities of the proteins encoded by MRP1 are different. Thus it is possible that genetic polymorphism may explain the results (Hipfner, Deeley, and Cole, 1999).

MRP1 has been demonstrated to transport a wide range of substrates that are conjugated to glutathione, glucuronide, or sulphate (Jedlitschky et al., 1994; Leier et al., 1996; Loe, Deeley, and Cole, 1998). Further, the protein has been shown to transport unmodified anionic compounds as the antifolate methotrexate (Hooijberg et al., 1999). Previously, Paul et al. (1996) investigated drug transport of DNR, VP16, and VCR into HL60/ADR membrane vesicles. Their results supported the hypothesis of a direct transport of unaltered lipophilic drugs. On the other hand, experiments have demonstrated that glutathione is required for the transport of DNR and that glutathione depletion completely reversed resistance to DOX, DNR, VCR, and VP16 in MRP1 -transfected cells (Zaman et al., 1995). Versantvoort et al. (1995) reported increased accumulation of DNR and rhodamine-123 in MRP1-positive human lung cancer cells after glutathione depletion. It was shown that ATP-dependent uptake of unmodified VCR by membrane vesicles derived from MRP1-transfected HeLa was dependent on the presence of reduced glutathione (Loe et al., 1996; Loe, Deeley, and Cole, 1998). Lastly, an ATP- and glutathione-dependent uptake of unmodified DNR and VCR has been demonstrated into membrane vesicles from drug-selected MRP1-positive cells and MRP1 -transfected cells (Renes et al., 1999). Reducing agents such as 2-mercaptoethanol, dithiothreitol, and l-cystine could not increase uptake of VCR into membrane vesicles from MRP1-transfected cells (Loe et al., 1996), indicating that the reducing capacity of glutathione was not responsible for its effect. Glutathione itself seems to play an essential role in MRP1-mediated transport of hydrophobic agents. The mechanism by which glutathione facilitates transport has not yet been fully elucidated. For most drugs, negatively charged conjugates are not known to exist, and there is no evidence for conjugation of glutathione with drugs to which MRP1 confers resistance (O'Brien and Tew, 1996). There are indications that MRP1 may function as a cotransporter for glutathione and drug (Rappa et al., 1997; Loe, Deeley, and Cole, 1998). Glutathione may be a low-affinity substrate for MRP1. It is possible that glutathione as well as anticancer drugs interact directly with MRP1 and that this interaction is necessary for transport (Taguchi et al., 1997). Transport of anionic substrates such as glutathione and glucuronide conjugates is inhibited by vinca alkaloids and anthracyclines (cationic) (Müller et al., 1994; Loe et al., 1996). For explanation of these facts, it has been suggested that MRP1 may contain two binding sites: One for hydrophobic compounds and one for hydrophilic compounds. This would allow a similar binding of glutathione, hydrophobic drugs, and compounds conjugated to glutathione, glucuronate, or sulphate (Renes et al., 1999).

Little is known about the MRP1-drug interaction. Our measurements showed that DNR entry into mrp1-positive and sensitive EHR2 cells occurred at rates that were indistinguishable. In contrast, Mülder et al. (1993) demonstrated decreased influx in MRP1-positive cells. This finding prompted the authors to suggest the vacuum cleaner model as also being appropriate for MRP1-substrate interaction.

The ability of MRP1 to bind ATP has been demonstrated unequivocally (Almquist et al., 1995). Thus the current model for MRP1 function predicts that ATP hydrolysis is coupled directly with drug transport (Versantvoort et al., 1992). In our study, both the total and in particular the vanadate-sensitive ATPase activities were negligible in the EHR2/VP16 microsomes. Several authors have determined the ATPase activity of human MRP1 (Mao et al., 1999; Hooijberg et al., 2000; Chang, Hou, and Riordan, 1997). In agreement with our results, neither Mao et al. (1999) nor Hooijberg et al. (2000) could demonstrate any effect of VCR, VP16, or DNR on the ATPase activity of MRP1-containing plasma membranes from human lung cancer cells. In contrast, Chang, Hou, and Riordan (1997) found a maximal specific ATPase activity of MRP1 that was of the same order as the basal-specific ATPase activity of PGP being moderately stimulated by anthracyclines and vinca alkaloids. The different results probably could be explained by the small amount of mrp1 in EHR2/VP16. The mrp1 activity accounts for only a small part of the ATPases in the membrane preparations. The very high K i value for vanadate inhibition found possibly could be explained by an unspecific inhibition of the background activity of other ATPases. In this study, the composition of the incubation medium has been optimized for assaying PGP ATPase activity. It may well be that the putative mrp1 ATPase requires cofactors that are different from those of PGP. A possible explanation could be that mrp1 acts as a cotransporter requiring presence of glutathione (Leier et al., 1996; Loe, Deeley, and Cole, 1998; Manciu et al., 2001).