Introduction

Chemotherapy is one of the mainstays of current anti-tumour treatment. A number of anti-tumour drugs have been developed which, when used singly or in combination, may affect tumour regression. In the main, such agents are dependent upon the tumours having a relatively higher proliferative index when compared to surrounding, normal tissue. However, few drugs are specific and normal tissues with a high proliferative index, eg gut, lung and haemopoietic system, are often subjected to collateral toxicity during anti-tumour chemotherapy. Where such toxicity is severe, it may become dose limiting, leading to a reduction or cessation of therapy and to insufficient tumour kill.1

To overcome this problem of normal tissue toxicity, particularly myelotoxicity, a number of strategies have been formulated. These include the use of haemopoietic growth factors to support bone marrow recovery following treatment and the transplantation of autologous, peripheral blood progenitor cells (PBPCs) post treatment.2,3 Both of these strategies, whilst successful to some extent, have the disadvantage that the recovered haemopoiesis is still (at least as) sensitive to further exposure to chemotherapeutic agents as prior to initiation of therapy.

Therefore, a further strategy has been developed which utilises the gene transfer of drug resistance functions into bone marrow in order to confer resistance to either particular chemotherapeutic drugs or to classes of agents.1 Many such drug resistance genes have been characterised and are under investigation for this purpose and these include O6-alkylguanine-DNA-alkyltransferase (resistance to nitrosoureas and related methylating agents),4,5 mutated dihydrofolate reductase (resistance to methotrexate and derivatives)6 and the multi-drug resistance (MDR-1, p-gp) gene product.7,8 The latter has been extensively studied, and the utility of this in providing protection against a wide range of chemotherapeutic agents has been demonstrated in cell line and animal models, as well as in human clinical trials.7,8,9,10

However, some worries have been expressed regarding the potential long-term effects of exogenous MDR-1 expression in bone marrow stem cells and their progenitors. One worry is that expression of MDR-1 in cells may facilitate their survival by reducing intracellular levels of drug to a non-toxic level, yet allowing sufficient levels of drug to accumulate to exert a mutagenic effect. Recently, we have shown that expression of MDR-1 in murine bone marrow in vivo reduces the clastogenic effects of etoposide, suggesting that MDR-1 expression does not necessarily result in an increase in occult damage.11 However, in one recent study Sorrentino and colleagues have shown an apparent increase in myeloproliferation and leukaemia in murine bone marrow following MDR-1 gene transfer and in vivo expansion of transduced cells.12 Whilst this remains the only study to suggest a leukaemic effect of exogenous MDR-1 expression, the fact that such an effect was seen poses serious questions as to the potential long-term effects of MDR-1 gene transfer and expression in humans, as well as to the advisability of utilising MDR-1 as a chemoprotective agent.

In order to determine whether a biological basis might be found for the myeloproliferative syndrome seen by Sorrentino and colleagues, we have introduced the MDR-1 cDNA, via retroviral gene transfer, into the murine, haemopoietic multipotent progenitor cell line FDCP-mix A4.13 This cell line was derived from long-term bone marrow cultures, is karyotypically normal and non-leukaemic. In the presence of IL-3 the cells self-renew but will undergo multilineage differentiation in response to physiological stimuli; eg in the presence of GM-CSF and G-CSF the cells develop into granulocytes and macrophages;14 in the presence of erythropoietin and haemin mature erythroid cells are formed.15 As well as responding to growth stimulatory growth factors this cell line will respond to negative regulators of haemopoiesis in an appropriate manner.16,17,18 These characteristics make this cell line a suitable model for the investigation of the effects of expression of exogenous MDR-1 on the survival and myeloproliferation of haemopoietic progenitor cells.

Materials and methods

FDCP-mix cells

FDCP-mix is a karyotyopically normal, factor-dependent murine multipotent haemopoietic cell line derived from long-term bone marrow culture. FDCP-mix cells were cultured in Fischer's medium (Life Technologies, Paisley, UK) supplemented with 20% (v/v) horse serum (Sigma, Poole, UK) and 2% (v/v) mIL3 conditioned medium (maintenance conditions).19 For analysis of factor dependence, FDCP-mix cells were pelleted from maintenance medium, washed twice with Fischer's medium and resuspended in this supplemented with 20% (v/v) horse serum. For analysis of proliferation, FDCP-mix cells were pelleted and washed as above. They were then re-seeded in Fischer's/20% (v/v) horse serum supplemented with either recombinant murine IL-3 (rmIL-3, 10 ng/ml or 1 ng/ml; R&D Systems, Abingdon, UK), or in a combination of rmIL-3 (10 ng/ml), recombinant human IL-6 (25 units/ml, R&D Systems) and recombinant rat stem cell factor (100 ng/ml; a gift of Amgen, Thousand Oaks, CA, USA). Viable cell counts were then performed daily. Differentiation potential of FDCP-mix cells was assessed as previously described.14,15 Cell morphology was examined on cytospin preparations stained with O-dianisidine (3,3-dimethoxybenzidine; Sigma) for erythroid cells and May and Grunwald/ Giemsa (BDH, Poole, UK). For analysis of colony formation, FDCP-mix cells were plated at 5 × 103/ml in Iscove's medium (Life Technologies) supplemented with 20% (v/v) horse serum 1% (w/v) bovine serum albumin, 2% (v/v) mIL3 conditioned medium, 3.3% (w/v) agar and various concentrations of etoposide as appropriate. Three aliquots of 1 ml were plated per assay and following 7 days incubation at 37°C in an atmosphere of 5% CO2 in air, colonies of more than 50 cells were counted.

Retroviral constructs and producer cells

The SF-MDR-1 amphotropic retroviral producer cell line was a gift from Dr Christopher Baum (Heinrich Pette Institut, Hamburg, Germany).20 This was incubated in a ‘ping-pong’ co-culture with GP-E86 cells,21 then amphotropic cells were selected with 0.15 mg/ml Hygromycin B (Sigma), for use in transducing FDCP-mix cells. The tNGFR cDNA was inserted into the retroviral vector pSF, which contains the spleen focus-forming virus LTR. Retroviral producer cells were derived from this construct following Lipofectin-mediated (Life Technologies) co-transfection of pSFtNGFR (10 μg) and pcDNA3 (neo resistance, 1 μg; Clontech laboratories, Palo Alto, CA, USA) into GP-E86. After selection of G418-resistant cells, these were co-cultivated in a ‘ping-pong’ culture with GP + envAm1222 and selected with 0.15 mg/ml Hygromycin B (Sigma) to isolate amphotropic cells. Following FACS sorting of tNGFR-positive cells (anti-NGFR mouse IgG1; Chemicon, Harrow, UK) these were expanded for use in transducing FDCP-mix cells.

Transduction of FDCP-mix cells

Confluent 25-cm2 flasks of producer cells were irradiated (30 Gy, Cs137, 4.2 Gy/min), then seeded with 2 × 105 FDCP-mix cells in 10 ml of maintenance medium. Seventy-two hours later, FDCPmix cells were harvested from the producers for isolation either on the basis of rhodamine exclusion (see below) for MDR-1, or via FACS sorting for tNGFR, using the same anti-NGFR as described above.

Rhodamine exclusion assay

FDCPmix cells were incubated for 20 min at 37°C in the dark, in maintenance medium supplemented with rhodamine-123 (100 ng/ml; Sigma). Half of this cell suspension was then washed twice in cold PBS, then incubated for 30–60 min in maintenance medium without rhodamine-123. This efflux sample was compared to the non-efflux sample (other half of cells) by FACS analysis. In some experiments, verapamil (Sigma; 25 μM) was added during the efflux incubation.

Results

Following transduction of FDCP-mix cells with either the tNGFR or MDR-1 vectors, positive cells were selected by FACS either using the anti-NGFR antibody or, for MDR-1, via rhodamine exclusion. Cells were then plated in soft agar to allow development of discrete colonies, these were then picked, individually expanded and the subsequent clonal cell lines examined for their ability to exclude rhodamine or for the expression of tNGFR. Figure 1 shows the data obtained when parental FDCP-mix and clones of either tNGFR or MDR-1 transduced FDCP-mix were subjected to FACS analysis following incubation with rhodamine. In all cases, when FACS analysis was performed without allowing time for dye efflux (Figure 1a–h, solid histograms), all of the cells became rhodamine bright. The cells were washed free of rhodoamine and incubated for 30 min in fresh (rhodamine-deficient, efflux) medium to allow efflux of dye (open histograms). No significant difference in the fluorescent population was seen in any of the control (FDCP-mix parental and tNGFR transduced) cells following this procedure (Figure 1a–d). In contrast in all MDR-1-transduced clones analyzed (Figure 1e–h, open histograms), significant efflux of rhodamine was evidenced by a decrease in the mean fluorescence of the cells. To confirm that the efflux rhodamine observed in the MDR-1-transduced clones was indeed due to p-glycoprotein expression, we next tested the effects of the p-glycoprotein inhibitor, verapamil, on the ability of MDR-1-transduced FDCP-mix cells to exclude rhodamine. In all, four clones of FDCP-mix-MDR-1 were tested and Figure 2 shows a typical result from one of these. In the absence of verapamil, efflux of rhodamine was again evident (Figure 2a, open histogram). However, when verapamil was added to the incubation and efflux media, FDCP-mix-MDR-1 cells showed a diminished capacity to exclude rhodamine and retained a high mean fluorescence (Figure 2b). Thus the exclusion of rhodamine from FDCP-mix-MDR-1 cells is due to expression of p-glycoprotein.

Figure 1
figure 1

Rhodamine exclusion by control and MDR-1-transduced FDCP-mix clones. Data shown are derived from four control (parental FDCP-mix (a), and three tNGFR transduced clones (b–d)) and four MDR-1-transduced clones (e–h). Solid histograms indicate rhodamine fluorescence when no efflux time was allowed prior to FACS analysis. Open histograms indicate rhodamine fluorescence following efflux.

Full size image
Figure 2
figure 2

Effect of verapamil on rhodamine exclusion by MDR-1-transduced cells. Rhodamine exclusion was performed in (a) the absence of verapamil and (b) the presence of 25 μM verapamil. Solid histograms indicate rhodamine fluorescence when no efflux time was allowed prior to FACS analysis. Open histograms indicate rhodamine fluorescence following efflux.

Full size image

We next examined the sensitivity of tNGFR and MDR-1-transduced cells to the cytotoxic drug, etoposide. Cells were exposed to increasing doses of etoposide in a colony formation assay and after 7 days colonies of surviving cells counted. As shown in Figure 3, cells expressing tNGFR (in common with parental cells – data not shown) exhibited a dose response to etoposide, with colony formation decreasing steadily with dose to around 30% of control levels at the highest drug concentration (125 nM) used. In contrast, MDR-1 transduced cells showed increased resistance to killing, with 75–85% survival of colony-forming cells at 125 nM etoposide. These data, combined with the rhodamine exclusion data, thus demonstrate that the MDR-1 gene product was expressed and functional in the transduced FDCP-mix cells.

Figure 3
figure 3

Dose–response of two MDR-1 expressing (solid squares and circles) clones and one tNGFR-expressing (open triangles) lines to the cytotoxic effect of various doses of etoposide. Data show colony formation at day 7 expressed as percentage of control (untreated cells) and representing the mean of three experiments ± s.e.m.

Full size image

Having demonstrated functional expression of MDR-1, we proceeded to examine the effects of expression on other phenotypic characteristics of FDCP-mix. FDCP-mix cells are dependent on high levels of interleukin-3 (IL-3) for survival and proliferation. Thus, to examine the effects on factor dependence, we washed cells free of growth factor and then incubated them in maintenance medium minus IL-3. As a control we included a previously described FDCP-mix line, which had been engineered to express the anti-apoptotic factor, bcl-2.23 As shown in Figure 4, and in agreement with our previously published data, bcl-2-expressing cells continued to survive without proliferation in the absence of IL-3. All other cell lines (parental, tNGFR-expressing and p-glycoprotein-expressing) showed extensive cell death as a result of growth factor withdrawal, demonstrating that p-glycoprotein expression led neither to a general resistance to cell death nor to factor independence. We next examined the growth rate of control and p-glycoprotein-expressing cells, in the presence of maximal (10 ng/ml) amounts of IL-3 or reduced (0.1 ng/ml) levels of this growth factor. As evidenced by the data presented in Figure 5, expression of p-glycoprotein was not associated with an increase in the growth rate of FDCP-mix cells, either at high or low levels of IL-3. In the study by Sorrentino and colleagues, increased amplification of MDR-1 transduced vs control haemopoietic cells was observed in vitro in a combination of IL-3, IL-6 and stem cell factor (SCF).12 Therefore, we also examined the growth characteristics of tNGFR and MDR-1 transduced FDCP-mix under these growth factor conditions. As seen in Figure 6, there was again no difference in the expansion of cells, regardless of whether they expressed p-glycoprotein or the control gene product.

Figure 4
figure 4

Factor dependence of control and MDR-1 expressing FDCP-mix. Data show viable cell counts at different time points after factor withdrawal, for one bcl-2-expressing clone (open squares), parental FDCP-mix cells (open triangles), the mean of four tNGFR transduced clones (open circles) and the mean of four MDR-1-transduced clones (solid triangles). The data represent the mean of three experiments ± s.e.m.

Full size image
Figure 5
figure 5

Proliferation of tNGFR (open symbols) and MDR-1 (solid symbols) expressing cells with time in days after seeding in the presence of either 0.1 ng/ml (squares) or 10 ng/ml (triangles) IL-3. Data represent the mean of four clones in two separate experiments ± s.e.m.

Full size image
Figure 6
figure 6

Expansion with time of MDR-1 (solid bars) and tNGFR (hatched bars)-expressing clones incubated with IL-3, IL-6 and SCF. Data are mean of four clones ± s.e.m. and show fold expansion at days 4, 7 and 12.

Full size image

Finally, we determined the differentiation capacity of the different FDCP-mix clones that we had derived (Figure 7). In the absence of conditions that promote differentiation of parental FDCP-mix cells, tNGFR and MDR-1-transduced cells, in common with the parental line,13,14 exhibited a mainly (>80%) blast cell morphology, with a high nuclear to cytoplasmic ratio. When cells were placed in conditions that promote mainly granulocyte/macrophage differentiation, both control (tNGFR) and MDR-1 cells differentiated normally, producing mostly granulocytes, with some macrophages and few blast cells remaining after 7 days. A corresponding result was also seen when cells were placed in conditions that additionally promote erythroid differentiation. In this case, a significant (30–40%) proportion of erythroid cells were also produced. The patterns of differentiation of both tNGFR and MDR-1-transduced cells mimicked those seen with the parental cells. Thus expression of MDR-1 was not associated with a block in either granulocyte/macrophage or erythroid differentiation.

Figure 7
figure 7

Morphology of tNGFR and MDR-1-expressing cells incubated under maintenance (M, see Materials and methods) conditions or those that promote granulocyte/macrophage (G) or erythroid (E) differentiation. Solid bars indicate blast cells, hatched bars indicate granulocytes, crosshatched bars indicate macrophages and stippled bars indicate erythroid cells. The results are the mean of four clones from three separate experiments.

Full size image

Discussion

The concept of using drug resistance genes to protect human haemopoiesis from the cytotoxic side-effects of chemotherapy has been under development for some time. A number of genes have been used in pre-clinical models, including dihydrofolate reductase,6 O6-alkylguanine-DNA-alkyltransferase4,5 and MDR-1.7,8 Of all of these, MDR-1 was initiated first and has consequently progressed furthest, with some clinical trials of the approach already conducted. Whilst the earliest trials suffered from poor transduction of human repopulating stem cells,24,25 recent developments in HSC-transduction and in ex vivo manipulation of HSCs are likely to overcome this difficulty.26,27,28 Indeed, a recent report of MDR-1-transduced cells in human patients highlights improvements in this regard.10 Ironically, at the same time as MDR-1-mediated genetic chemoprotection in humans appears technically feasible, a new problem has arisen as a result of studies in mice. The description of a myeloproliferative disorder in animals that received an MDR-1 gene-modified graft is of obvious concern.12 Numerous previous murine studies of MDR-1 gene transfer have not documented such an effect. Furthermore, non-human primate29 and Nod/SCID repopulating cell30 studies have found no evidence for myeloproliferative disease following MDR-1 transduction and transplantation, and the available human clinical data do not support a myeloproliferative role for MDR-1.9,10 However, the possibility of an adverse long-term effect of MDR-1 gene transfer on haemopoiesis is an important impediment to the further development of this approach.

We reasoned that if MDR-1 plays a direct role in the induction of a myeloproliferative effect it should be possible to mimic this in a cell line model by effecting retroviral gene transfer of MDR-1. As a test system we chose the murine haemopoietic progenitor cell line, FDCP-mix. This is a non-leukaemic, karyotypically normal, factor-dependent (for both survival and proliferation) cell line that may be maintained in high levels of IL-3 but can undergo terminal differentiation in appropriate culture conditions. A number of studies have looked at the response of this cell line to various oncogenes and it has proved sensitive to the transforming effects of these. Even subtle effects on proliferation and survival are detectable as evidenced by the effects of v-abl or bcr/abl on FDCP-mix.31,32 These cell lines have also been used to study the influence of proteins that are known regulators of hematopoiesis (for example, growth factor receptors and transcription factors) and been shown to be sensitive to small changes in expression levels.33,34

Following retroviral gene transfer of either MDR-1 or tNGFR, cells expressing these were isolated and clonal cell lines derived. Those cell lines that had been derived from MDR-1-transduced FDCP-mix showed clear evidence of rhodamine exclusion which was reversed by the MDR-1 inhibitor, verapamil. Furthermore MDR-1, but not tNGFR, cells showed increased resistance to killing by etoposide. Thus, functional p-glycoprotein was expressed as a consequence of MDR-1 gene transfer and at a sufficiently high level to confer resistance to a clinically important anti-tumour agent.

Having established functional expression of p-glycoprotein, we next examined the effects of this expression on a variety of parameters, changes in which have previously been associated with oncogenesis in humans, and which are altered in FDCP-mix cells by various oncogenes. For example, transfer and expression of bcl-2 leads to resistance of FDCP-mix cells to apoptosis following growth factor removal or cytotoxic insult.31 Expression of activated ras leads to a block in differentiation of these cells,35 whilst v-abl and bcr-abl both lead to changes in the growth factor response of FDCP-mix.31,36,37 No effect of p-glycoprotein expression on any of these parameters was seen which could explain the previously reported myeloproliferative disease. Growth factor dependence, proliferation (even under the growth factor conditions used by Bunting et al12) and differentiation were all the same as in control cells. This despite the fact that MDR-1-transduced cells were maintained in culture over 3–6 months and were entered into the above assays at multiple points during that time.

Thus our data do not support a role for p-glycoprotein expression in dysregulated haemopoiesis. Notwithstanding this, the results obtained by Sorrentino and colleagues are still disturbing and require explanation. It may be that the precise transduction conditions utilised in the Sorrentino study contributed significantly to the emergence of a hyperproliferative population. Thus, multiple gene insertions, coupled with excessive proliferative stress, may have led to a selective pressure for cells with a hyperproliferative phenotype. This may have arisen as a result of insertional mutagenesis by provirus affecting key cell cycle control elements, and may even be vector backbone-specific. Southern blot analysis (not shown) indicated that the FDCP-mix-MDR1 clones we used in this study had only one proviral insertion, in contrast to up to 20 copies seen in studies by the Sorrentino group. This is a fundamental difference between our study and that of Sorrentino et al.

An alternative explanation may involve a gene dosage phenomenon whereby multiple copies of an MDR-1 gene may lead to hyperexpression of p-glycoprotein and thus to a proliferative effect. We did not see any evidence of a differential phenotype (other than in terms of rhodamine exclusion and drug resistance) in different clones with varying levels of p-glycoprotein expression. However. we cannot discount the possibility that even greater levels of expression could unmask an otherwise occult effect of p-glycoprotein. Given the efficiency of retroviral gene transfer to human haemopoietic stem cells, it is unlikely that a large number of proviral copies would be introduced into a patient's stem cells during a clinical trial. Thus, our model may more accurately effect the gene dosage that might be expected clinically.

In view of the possibility of a gene dosage effect, further careful evaluation of the phenotypic effects of p-glycoprotein over-expression, particularly in primary cells is warranted.