Expression of MRP4 Confers Resistance to Ganciclovir and Compromises Bystander Cell Killing*

The multidrug resistance protein MRP4, a member of the ATP-binding cassette superfamily, confers resistance to purine-based antiretroviral agents. However, the antiviral agent ganciclovir (GCV) has not been shown to be a substrate of MRP4. GCV is important not only in antiviral therapy, but also in the selective killing of tumor cells modified to express herpes simplex virus thymidine kinase (HSV-TK). We therefore tested the effect of MRP4 on the cytotoxicity of GCV, on the ability of GCV to kill cells genetically modified to express HSV-TK, and on the bystander effect in which unmodified target cells are killed by GCV. Cells overexpressing MRP4 had markedly increased resistance to the cytotoxicity of GCV. Although, expression of recombinant HSV-TK increased the intracellular concentration of GCV nucleotide, key the accumulation and cytotoxicity of GCV metabolites. These novel variable response in gene therapy based on HSV-TK expression.

The multidrug resistance proteins (MRPs) 1 are a family of ATP-binding cassette transporters (ABC transporters; for an overview, see //nutrigene.4t.com/humanabc.html) that mediates drug efflux and multidrug resistance (1). We previously demonstrated that MRP4 (also known as ABCC4) severely reduces the antiviral efficacy of several nucleoside reverse transcriptase inhibitors, such as zidovudine (3Ј-azido-3Ј-deoxythymidine) (AZT)) and 9-(2-(phosphonomethoxy)ethyl)-adenine (PMEA), in mammalian cells and that this effect corresponds with increased ATP-dependent efflux of their nucleotide derivatives (2). These findings were, in part, replicated by Lee et al. (3). A better understanding of the role of MRP4 as a nucleotide efflux transporter may lead to improved nucleoside-based therapies for HIV, herpes viruses, and cancer. Ganciclovir (9-(1,3-dihydroxy-2-propoxymethyl)guanine (GCV)) is widely used against cytomegalovirus infection in patients with AIDS (4 -6) but is poorly tolerated in combination with AZT (7). Although the molecular basis of this interaction is unknown, in vitro studies indicate that the combination is extremely cytotoxic (8,9). We therefore postulated that MRP4 might be involved in the cytotoxicity induced by the AZT-GCV combination. If so, such an interaction could have important therapeutic potential because of the widespread use of GCV as a prodrug in gene therapies for cancer (10,11). Transduction of tumor cells with the herpes simplex virus thymidine kinase (HSV-TK) gene and treatment with GCV is a common anticancer gene therapy (10 -12). HSV-TK specifically converts GCV to its monophosphate nucleotide form, GCVmonophosphate, which is then further anabolized by cellular kinases, and accumulation of these metabolites causes cell death (5,12). However, the cytotoxicity of this therapy varies dramatically. The mechanism of this variability is unknown, but it does not appear to reflect the level of HSV-TK expression (13,14).
The variable toxicity of GCV to HSV-TK gene-modified tumor cells also markedly affects the efficacy with which bystander (adjacent, non-transduced) tumor cells are killed (15). In vitro studies show that bystander cell death is associated with the accumulation of GCV metabolites (16); however, this effect is variable and idiosyncratic, and the mechanism accounting for variable accumulation of these metabolites is unclear (17,25). Some investigators believe that direct intercellular communication by gap junctions is a prerequisite for GCV metabolite transfer (18), while others do not (19,20). Other studies suggest that diffusion of GCV metabolites and the proximity of bystander cells are important variables (21)(22)(23). Although a combination of these mechanisms is likely, it is unknown if nucleotide efflux plays a role. We reasoned that because the efflux transporter MRP4 transports some natural and modified nucleotides (2,24) its expression could impact HSV-TK based gene therapy by hastening the removal of cytotoxic GCV metabolites. However, it is unknown if MRP4 plays a role in GCV metabolite accumulation and cellular egress.
We investigated whether MRP4 affects the cellular accumulation and cytotoxicity of GCV in cells specifically overexpressing MRP4 and in cells we engineered to overexpress MRP4. The cDNA encoding human MRP4 was isolated and stably introduced into MCF-7 cells that expressed negligible immunoreactive MRP4. These cells had enhanced resistance to GCV cytotoxicity and to bystander cell killing. We also investigated the effect of MRP4 on the cytotoxicity of GCV in cells engineered to express HSV-TK. Taken together, our results show that MRP4 plays a key role in the intracellular accumulation and cytotoxicity of GCV metabolites. These findings reveal ganciclovir egress by MRP4 as a novel mechanism contributing to the variable response in gene therapy based on HSV-TK expression. Cell Lines-The human T-lymphoid cell line CEMss and its PMEAresistant variant, CEMr1, were grown as described previously (2). The human breast tumor cell line, MCF7, and the HSV-TK-expressing cell line, SW620-HSV-TK, were cultured as previously described (26,27). Human osteosarcoma cell line, Saos-2, and 293T human embryonic kidney were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and 100 units/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine. CEMss and CEMr1 cells were transduced with a recombinant murine retrovirus (pLENTK) encoding HSV-TK (27).

Materials
Immunoblot Analysis-Immunoblot analyses of MRP4 and MRP1 were performed as described previously (2). N-glycosylation was assayed by treating the lysates of transfected MCF7 cells with PNGase F (New England Biolabs) and immunoblotting was as described (2).
Cytotoxicity Assay-Cells were plated in 6-cm dishes (2 ϫ 10 5 cells/ dish) and incubated for 24 h at 37°C. Medium was removed, drugs were added, and cells were incubated at 37°C for the indicated intervals. Viability and inhibition of cell growth were determined by direct counting of trypan blue-dyed cells on a hemocytometer. A 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (Sigma) was used according to the manufacturer's instructions to determine cellular toxicity. Each assay included duplicate samples for each drug concentration, and all experiments were performed at least twice.
Flow Cytometry-After exposure to GCV, cells were harvested, washed, and resuspended at ϳ1 ϫ 10 6 cells/ml in propidium iodide staining solution (0.05 mg/ml propidium iodide, 0.1% sodium citrate, 0.1% Triton X-100). Immediately before analysis, cells were treated for 30 min at room temperature with RNase (Calbiochem, San Diego, CA) at a final concentration of 14 g/ml and filtered through nylon mesh. Fluorescence was measured on a Becton Dickinson FACScan flow cytometer (BD Biosciences) using laser excitation at 488 nm. The percentages of cells in different phases of the cell cycle were determined by using the Modfit computer program (Verify Software House, Topsham, ME).
Sub-G 1 Determination-Analysis of the DNA content less than G 1 was used as an additional measurement of apoptosis as previously reported (31). Briefly, after drug treatment both floating and attached cells were harvested in a propidium iodide solution (50 g/ml propidium iodide in 0.1% sodium citrate and 0.1% Triton X-100), treated with 5 g/ml RNase (Calbiochem) for 30 min at room temperature, and then analyzed by flow cytometry on a Becton Dickinson FACscan (BD Biosciences) using laser excitation at 488 nM. The percentage of sub-G 1 cells was determined.
DAPI Staining of Apoptotic Cells-CEMss and CEMr1 cell lines expressing TK were plated (5 ϫ 10 4 cells/well) in 8-well plastic chamber slides (Lab-Tek) and treated with the indicated concentrations of GCV. The cells were then washed with phosphate-buffered saline, stained in 1 g/ml DAPI, and examined on a Zeiss fluorescent microscope at 40ϫ magnification (28).
Assays of Drug Accumulation-Cells (2 ϫ 10 5 cells/ml growth medium) were seeded in 6-cm dishes 24 h prior to the assay. They were then washed with Hanks' balanced salt solution and incubated for the desired interval with the radiolabeled drug dissolved in warmed medium. Aliquots of cells were removed, washed with chilled phosphatebuffered saline, solubilized in 1 ml of 0.5 N NaOH, and assayed for radioactivity.
Cloning of MRP4 cDNA-The human MRP4 cDNA (expressed sequence tag 38,091, described in Ref. 2) was used to screen a human lung ZAPII cDNA library (Stratagene) by plaque hybridization. A 4.5-kb cDNA that lacked the initiator methionine was identified, and a 5Ј portion of this cDNA was used to screen a human lung 5Ј-STRETCH PLUS cDNA library (Clontech). Ten positive clones that contained only the 5Ј portion of MRP4 were isolated. The 5Ј portion of MRP4, together with the overlapping 4.5-kb clone of the 3Ј portion, yielded a 5.8-kb MRP4 cDNA (GenBank TM accession number: AY081219). The 5Ј-and 3Ј-untranslated regions of this cDNA fragment were removed, and the complete 4.2-kb cDNA encoding the 1325 amino acids of MRP4 was used to construct the expression vector MSCV-MRP4-IRES-GFP (contains the murine stem cell virus (MSCV) long terminal repeat as well as a ribosomal entry site (IRES) to permit expression of green fluorescent protein (GFP) and was kindly provided by Dr. Robert Hawley, Holland Laboratory, American Red Cross, Rockville, MD). The complete MRP4 cDNA was also subcloned into the EcoRI site of pcDNA3 (Invitrogen) to generate pcDNA3-MRP4.
Metabolic Labeling with [ 3 H]GCV-CEMss HSV-TK and CEMr1-HSV-TK cells were treated in triplicate with 1 C of [ 3 H]GCV (plus 1 M unlabeled GCV), washed, and resuspended in fresh medium. Nucleotides were extracted from pelleted cells in 0.2 ml of 7% methanol at 4°C for 15 min (22). Radioactivity was measured in an aliquot of each methanol supernatant by scintillation counting. The proportion of phosphorylated GCV was measured in an aliquot of medium by anion exchange chromatography (22).

Effect of MRP4
Overexpression on the Cell Cycle, Sensitivity to GCV, and Apoptosis-Immunoblot analysis showed the PMEA-resistant cell line, CEMr1 (4), expressed 6-to 8-fold more MRP4 and slightly less MRP1 than did the CEMss parent cell line (Fig 1A). Treatment with GCV caused an indistinguishable cell cycle arrest in both CEMr1 and CEMss cells, i.e. the proportion of S-phase cells increased and the G 1 -phase decreased (Fig. 1B). However, the viability of GCV-treated CEMss cells decreased in a dose-dependent manner, whereas the viability of CEMr1 cells was only modestly affected by GCV (Fig. 1C); this finding was confirmed by measuring the population of cells with DNA content less than G 1 indicative of apoptotic cells (31) (Fig. 1D). CEMr1 and CEMss cells were equally sensitive to vinblastine and, as expected, retained their respective resistance and sensitivity to PMEA (Fig. 1E) (4). These results indicate that cells overexpressing MRP4 are sensitive to cell cycle arrest induced by GCV but are substantially more resistant to its cytotoxic effects.

Efflux of GCV Is Enhanced in HSV-TK-transduced Cells
That Overexpress MRP4 -Sensitivity to GCV can be enhanced by modifying targeted cells to express HSV-TK, but identically modified cells vary markedly in their sensitivity to GCV (13). To evaluate whether MRP4 modulates the GCV sensitivity, we modified CEMss and CEMr1 cells to stably express HSV-TK, isolated cells that expressed similar levels of HSV-TK (Fig 2B), and treated them with GCV. Although expression of HSV-TK increased the sensitivity of both sets of cells to GCV, their relative sensitivity was preserved (Fig 2A). The IC 50 of GCV was 60 M in the CEMr1-TK cells, a value five times that in the CEMss-TK cells (12 M). Thus, MRP4 overexpression strongly influences sensitivity to GCV in cells that are and are not modified to express HSV-TK.
We compared efflux of GCV and its metabolites from CEMss-HSV-TK and CEMr1-HSV-TK cells by pretreating the cells with GCV, suspending them in drug-free medium for 4 h, and quantifying GCV and phosphorylated GCV in the medium by anion exchange chromatography (27). Sixty-one percent of GCV metabolites in the CEMr1-HSV-TK cells and 44% in the CEMss-HSV-TK cells were released (Fig. 2C). Thirty percent more of the GCV metabolites exported by CEMr1-HSV-TK cells than by the CEMss-HSV-TK cells were phosphorylated. These findings indicate that cells that overexpress MRP4 readily remove GCV metabolites.
Accumulation of GCV Is Reduced in Cells That Overexpress MRP4 -The initial uptake of GCV (0 -5 min after exposure) was very rapid and was essentially identical in CEMss and CEMr1 cells (not shown). However, marked differences in GCV accumulation were observed after 60 min or more. CEMr1 cells accumulated substantially less drug than CEMss cells at every GCV concentration tested (Fig. 2D). Treatment with indomethacin, an inhibitor of MRP4 transport (see below), dramatically increased GCV accumulation in the CEMr1 cells (and to a lesser extent in the CEMss cells; Fig. 2D) and decreased their efflux of radiolabeled GCV nucleotides into the media to the level observed in CEMss cells (not shown).
Isolation and Expression of Human MRP4 -A cDNA encoding human MRP4 was isolated from two independent human lung cDNA libraries (see "Materials and Methods"). Its coding sequence was identical to the reported sequence (GenBank TM / EBI accession number AF071202) with the exception of seven single-base substitutions. Of these, only four caused amino acid changes. These differences may reflect polymorphisms or may be related to derivation of the reported MRP4 cDNA from a drug-selected cell line; drug selection has been noted to cause such alterations in other mammalian ABC transporters (e.g. MDR1 and BCRP). These differences clearly do not affect the transport of PMEA (see Fig. 3B).
Uptake of bis-POM-PMEA, Vinblastine, and Methotrexate by Cells Modified to Express MRP4 -The human breast carcinoma cell line, MCF-7, was shown by Western blotting to express very low levels of MRP4. We transduced MCF-7 cells with either the MSCV-MRP4-IRES-GFP or the control MSCV-IRES-GFP and isolated GFP ϩ cell populations. We then quantified MRP4 in crude membrane fractions from each cell population by Western blotting (Fig. 3A). As expected, MRP4 was highly expressed in cells transduced with MSCV-MRP4-IRES-GFP, whereas it was expressed at very low levels in cells transduced with MSCV-IRES-GFP. However, treatment of plasma membrane fractions with PNGase F to de-glycosylate MRP4 yielded a band of ϳ150 kDa (a value close to that predicted from the amino acid sequence) in both cell lines, which suggests that posttranslational modifications may affect detection of MRP4 with this antibody. Immunohistochemical and subcellular fractionation studies revealed that most of the MRP4 expression in MCF-7 cells was localized to the plasma membrane (not shown).
Uptake of PMEA, an acyclic nucleoside phosphonate that contains a nonhydrolyzable bond and resembles a nucleoside 5Ј-monophosphate (4), was decreased in cells transfected with MRP4. However, because PMEA uptake is slow, we also incubated cells with bis-POM-PMEA, which is taken up and is converted to PMEA intracellularly before efflux (4). Fig. 3B shows that uptake of bis-POM PMEA by MCF-7-MRP4 cells H]MTX concentration was measured as previously described (33). E, the osteosarcoma cell line Saos-2, which expresses minimal MRP4, was transfected with either pcDNA3 or pcDNA3-MRP4, and cells were selected in G418. Asterisks indicate the clones that were expanded for functional analysis. F, uptake of bis-POM-[ 3 H]PMEA was inversely proportional to the expression of MRP4. G, indomethacin inhibited MRP4 function, but probenecid did not. Uptake of bis-POM-[ 3 H]PMEA was measured in cells preincubated with no drug, with indomethacin, or with probenecid. was less than one-third of the uptake by control cells at every concentration tested. In contrast, vinblastine uptake was identical in test and control cells (Fig. 3C). Interestingly, whereas earlier studies suggested that MRP4 plays a role in MTX uptake and efflux (3), we found that the steady-state intracellular concentration of methotrexate was the same in test and control cells (Fig. 3D).
We next compared MRP4 function with the quantity of MRP4 protein expressed. Saos-2 human osteosarcoma cells, which express very little MRP4, were transfected with pcDNA3-MRP4 and selected in G418. Immunoblot analysis revealed a wide range of MRP4 expression (Fig. 3E). Although there appeared to be an inverse relationship between expression of MRP4 and expression of MRP1 in CEMss cells (Fig. 1A), MRP1 and MRP4 expression were completely unrelated in Saos-2 transfectants (Fig. 3E). Intracellular accumulation of PMEA was inversely proportional to the quantity of MRP4 expressed: accumulation of PMEA was decreased (ϳ41%) at low levels of MRP4 expression and substantially (more than 95%) reduced at very high levels of MRP4 expression (Fig. 3F). Accumulation of PMEA in the Saos-2 MRP4 transfectants was increased over 600% by treatment with 100 M indomethacin FIG. 4. Ectopic expression of MRP4 in MCF-7 cells reduces GCV uptake and cytotoxicity. A, MCF-7 and MCF-7-MRP4 cells were incubated for various intervals with radiolabeled GCV, and intracellular radioactivity was measured. B, MCF-7 and MCF-7-MRP4 cells were incubated with various concentrations of radiolabeled GCV, and intracellular radioactivity was measured. C, MCF-7 and MCF-7-MRP4 cells were pretreated with radiolabeled GCV, washed in ice-cold buffer, and resuspended in warmed drug-free medium. Intracellular radioactivity was measured at the indicated intervals. D, MCF-7 and MCF-7-MRP4 cells were exposed for 72 h to GCV at the indicated concentrations, and an 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay was performed to determine viability. E, MCF-7 and MCF-7-MRP4 cells were exposed for 4 h to various concentrations of MTX. No difference in MTX sensitivity was seen at either 4 h or after 72 h of exposure (not shown).  (27). The values are obtained from triplicate determinations in a single experiment that was repeated twice with similar results. (Fig. 3G). In contrast, 100 M probenecid treatment had no effect (Fig. 3G). Indomethacin-treated Saos-2 control cells also showed a modest increase (56%) in PMEA accumulation, possibly reflecting inhibition of an endogenous transporter.
Accumulation, Efflux, and Cytotoxicity of GCV in MCF7 Cells Modified to Express MRP4 -We observed no difference between control and MRP4-transfected MCF-7 cells in the initial rates of uptake (Ͻ5 min) of GCV, gemcitabine (a cytidine analog). This finding is consistent with the patterns of mRNA expression we observed for the major nucleoside transporters (hCNT1a, ENT2, and hENT1) (not shown). However, there were striking differences between test and control cells in GCV accumulation: the steady-state concentration of GCV in the MRP4-transfected cells was only ϳ50% that in the vector cells (Fig. 4A). This disparity was observed over a wide range of concentrations of GCV (Fig. 4B).
We investigated whether efflux was altered in the cells expressing MRP4. First, we treated these cells with ϳ1.5 times the amount radiolabeled GCV used to treat control cells to obtain a comparable intracellular GCV concentration. We then washed the cells, placed them in drug-free medium, and measured intracellular radioactive GCV (Fig. 4C). The decline of intracellular GCV was more rapid in the MRP4-transfected cells compared with vector cells (t 1/2 s of ϳ16 and 32 min, respectively). This increased efflux was accompanied by a striking decrease in GCV-induced cytotoxicity (Fig. 4D) but not in MTX-induced cytotoxicity (Fig. 4E). Similar results were obtained in parallel experiments with Saos-2 cells modified to express MRP4 (not shown).
Expression of MRP4 Affects Bystander Cell Killing-Because MRP4 transports nucleotide derivatives (1-3, 24, 26, 32) including GCV anabolites from cells, we reasoned that the killing of bystander cells (those that do not express TK) by GCV could be significantly influenced by MRP4 expression. To explore this possibility, we used an SW620-HSV-TK cell line that activates GCV and readily effluxes GCV nucleotides (27). We performed a clonal dilution assay (27) in which either MCF-7 or MCF-7-MRP4 cells were co-cultured with increasing percentages of the SW620 HSV-TK cells (see "Materials and Methods") (Fig. 5). As the proportion of HSV-TK-expressing cells increases the delivery of GCV metabolite increases (30) leading to greater cell death of bystander cells. The MCF-7-MRP4 cells are more resistant to bystander killing. However, with very high proportions of HSV-TK cells we see that MRP4 overexpression offers less protection. These studies indicate that MRP4 plays a strong role in bystander cell survival, but suggest its protection may be overcome by increasing the proportion of HSV-TK expressing cells. DISCUSSION We have shown that expression of MRP4 affects the sensitivity of cells to the therapeutically important antiviral guanine analog GCV and that it modulates the direct and bystander cytotoxicity in which cells modified to express HSV-TK are treated with GCV. Cells that overexpressed MRP4 accumulated smaller quantities of GCV metabolites than did controls. Although the sensitivity to GCV-induced cell cycle arrest was not reduced in these cells, they were protected from GCV cytotoxicity. This finding is consistent with a relationship between GCV cytotoxicity and the intracellular concentration of the drug (7,14). However, the effect did not extend to all cytotoxins: sensitivity to methotrexate and vinblastine was not affected.
Because we and others have demonstrated that GCV metabolites are readily transported from cells expressing HSV-TK (17, 27, 30), we evaluated the effect of MRP4 expression on GCV cytotoxicity of cells gene-modified to express HSV-TK.
HSV-TK cells that overexpressed MRP4 had significantly enhanced resistance to GCV (Fig. 2, A and B). This finding has strong implications for gene therapy with HSV-TK because varying MRP4 expression among cells gene-modified with HSV-TK would lead to decreased cytotoxicity.
The success of gene therapy requires not only that the HSV-TK-modified cells export GCV nucleotides but also that the adjacent bystander cells accumulate and retain sufficient nucleotides to induce apoptosis. An increase in the extracellular concentration of GCV can induce the killing of otherwise intractable bystander cells (13), suggesting the existence of a mechanism that excludes GCV nucleotides or decreases their intracellular concentration in these cells. Bystander killing by GCV can also be enhanced by prolonged exposure to the drug (13). Our results strongly suggest that expression of MRP4 by bystander cells protects them from GCV cytotoxicity when they are co-cultured with initiator cells expressing HSV-TK (Fig. 5). However, this protection decreases as the proportion of HSV-TK-expressing cells increases, implying that a threshold level of GCV nucleotide is exceeded and overrides the protective effect of MRP4. Nevertheless, this result suggests that cells overexpressing MRP4 would be less susceptible to the bystander effect because of their reduced accumulation of GCV nucleotides.
Our results demonstrate that cellular retention and accumulation of the guanine analog ganciclovir is strongly influenced by MRP4. Expression of this nucleotide transporter may reduce the antiviral efficacy of GCV therapy, which is often used to treat or prevent cytomegalovirus disease in transplant recipients and patients with AIDS. It can also attenuate both direct GCV cytotoxicity and bystander cell killing in cells modified to express HSV-TK to induce selective cell killing. We have also shown that cells that express MRP4 are much less susceptible to GCV despite modification with an HSV-TK expression vector. These studies provide the mechanistic basis for future studies evaluating the role of MRP4 expression in antiviral response to GCV and in HSV-TK and GCV gene therapy with solid malignancies (e.g. brain tumors) Finally, a challenge for the future will be to determine the degree to which MRP4 modulates the levels of toxic metabolites near bystander cells.