P-glycoprotein plays a drug-efflux–independent role in augmenting cell survival in acute myeloblastic leukemia and is associated with modulation of a sphingomyelin-ceramide apoptotic pathway

The following products were used in this study: MRK-16 anti-pgp (Kamiya, Seattle, WA); 15D3 anti-pgp (Becton Dickinson, Cowley, England); control and secondary antibodies for pgp analysis (Dako, High Wycombe, England); rhodamine 123 (R123), 7-amino-actinomycin D (7-AAD), the annexin V apoptosis detection kit, propidium iodide, FB1, L-glutamine, and antibiotics (Sigma Aldrich, Poole, England); daunorubicin (Cerubidin, Rhone-Poulenc Rorer, West Malling, UK); C₆-NBD-SM (Molecular Probes, Cambridge Bioscience, Cambridge, England); and granulocyte-macrophage–colony stimulating factor (GM-CSF) (Leucomax, Novartis, Basel, Switzerland).

Fumonisin B1 was dissolved in sterile water and stored at 4°C. PSC-833 (gift from Novartis) was stored at 1 mmol/L in ethanol. A 1 mmol/L stock solution of C₆-NBDSM in ethanol was stored at −20°C. UIC2 anti-pgp blocking antibody (Immunotech, Beckman Coulter, High Wycombe, England) was stored in 200 μg/mL aliquots at −20°C, as were preservative-free control IgG2a antibodies (Calbiochem, Nottingham, England). Fetal calf serum (FCS) (First Link, Wolverhampton, England) was heat inactivated at 56°C for 60 minutes.

U937 and TF1 leukemic cell lines were grown in RPMI 1640 with 10% FCS, 2 mmol/L L-glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin. The medium of the growth factor dependent cell line TF1 was supplemented with 200 units/mL GM-CSF. Experiments were performed on cells in log phase. Fresh or cryopreserved blood or bone marrow samples from untreated patients with AML were used in this study. As several assays were to be performed on the same samples, all samples selected were from patients with high blast counts; 9 samples were collected at presentation, and the remaining 2 were from relapsed patients (RF and GM in Table 1). The viability of thawed cells was assessed using trypan blue dye exclusion, and only samples with more than 90% thawed cell viability were used.

The MDR status of patient blasts was determined by 3 previously described methods.15 First, pgp was measured using an unconjugated MRK-16 antibody and a fluorescein isothiocyanate–conjugated (FITC-conjugated) second layer. In 1 case there was high nonspecific binding of the control antibody, and the phycoerythrin-conjugated (PE-conjugated) 15D3 antibody gave a clear result. This antibody was also used to confirm protein results in 2 protein-negative, R123 modulation–positive samples. Pgp substrate efflux modulation by PSC-833 was determined in an accumulation assay using R123. Thawed or freshly isolated patient blasts were rested in culture medium containing 20% FCS for 60-90 minutes. They were then pelleted; resuspended at 10⁶/mL; and incubated in medium containing 10% FCS at 37°C, in duplicate, with 200 ng/mL R123, with or without 2 μmol/L PSC-833. Control tubes with R123 incubated at 4°C were also set up. Following a 75-minute incubation, the cells were pelleted at 4°C, rinsed, and resuspended. They were analyzed using a FACScan or FACSCalibur flow cytometer (Becton Dickinson) as previously described.15 The mean fluorescence intensity (MFI) of each sample in the FL1 channel was recorded. Results were expressed as (MFI with PSC-833–cold control MFI)/(MFI with diluent control–cold control MFI). Third, results were validated using the standardized daunorubicin uptake assay. Rested leukemic blasts at 10⁶/mL and carboxylate microspheres (Polysciences) were incubated in duplicate for 75 minutes with 2 × 10⁻⁶mol/L daunorubicin at 37°C. Control cells were also set up with daunorubicin at 4°C. Cells and beads were rinsed and analyzed as described.15 Results were expressed as (cell MFI–control cell MFI)/bead peak fluorescence.

Thawed or fresh mononuclear cell preparations from patients with AML were rested for 60-90 minutes in culture medium consisting of RPMI 1640 with 20% FCS and 1% L-glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin. The cells were then counted, rinsed in 20 mL RPMI, and resuspended at 5 × 10⁵/mL in culture medium with only 1% FCS. Only samples with a post-rest viability greater than 90% were used. For the C₆-NBD-SM assay, aliquots were incubated for 30 minutes at 37°C at 2 × 10⁶/mL with 4 μmol/L C₆-NBD-SM in 1% FCS, pelleted, and resuspended in 1% FCS. Preliminary studies of pgp+ve cells had indicated that 30 minutes of preincubation with C₆-NBD-SM was sufficient for saturation, and that 30 minute fluorescence at a concentration of 4 μmol/L C₆-NBD-SM was approximately 1 log greater than autofluorescence in pgp+ve samples.

For chemosensitivity assays, cells were cultured for 48 hours in medium containing 10% FCS and dilutions of daunorubicin ranging from 0.01-1 μg/mL. The IC₃₀ (ie, the concentration of daunorubicin that caused apoptotic cell death in 30% of cells) was determined for each patient and used for a follow-up assay in which patient cells were cultured with daunorubicin at the IC₃₀ with and without 50 μmol/L FB1. Apoptotic cell death was quantified using 7-AAD as detailed below.

7-AAD was diluted in PBS as a 50 μg/mL stock solution, which was stored at 4°C. This was mixed with cultured cells to give a final concentration of 12.5 μg/mL. Cultures were incubated for 20 minutes in the dark and analyzed using flow cytometry (Cellquest software, Becton Dickinson). Analysis was performed as illustrated by Philpott27 and Schmid28 to select cells undergoing apoptotic cell death, which was represented by 7-AAD–high cells with low forward scatter. The apoptotic nature of the cell death induced by growth factor withdrawal was examined in additional studies using MGM-stained cytospins for cells displaying morphological characteristics of apoptosis, ie, condensed nuclear chromatin and membrane blebbing, and by quantitative analysis using the annexin V/propidium iodide kit (Sigma Aldrich) according to manufacturer's instructions.

Statistical analyses were performed (software; SPSS, Chicago, IL). The differences in response to anti-pgp and control antibodies were analyzed using the Wilcoxon signed rank test. Differences in responses between pgp+ve and pgp−ve groups were analyzed using the Mann-Whitney U test.

The D value used to measure pgp is calculated in the Kolmogorov-Smirnov test by the flow cytometer's Cellquest software (Becton Dickinson) and is a measure, on a scale of 0 to 1, of the difference in distribution between 2 fluorescence curves. The D value is used to calculate the difference between test and control fluorescence when that difference is likely to be small.15 29 Values of less than 0.1 were considered to be negative.

P-glycoprotein protein and function are measured routinely on AML samples in our laboratory. Out of 11 samples, 5 protein positive (pgp D values: 0.15-0.47; mean: 0.30) and 6 protein negative samples were selected for further study (Table 1). PSC-833 increased R123 uptake more than twofold in 4/5 pgp+ve samples, but its effect was less marked in pgp−ve samples. Daunorubicin uptake was also characterized in these patients (Table 1). A previous study has shown that a daunorubicin cell to bead ratio of less than 2.1 is associated with MDR mediated by pgp, lung resistance protein (lrp), or multidrug resistance-associated protein (mrp).15Daunorubicin uptake was found to be low in 4/5 pgp+ve samples and 2/6 pgp−ve samples. TF1 cells are strongly pgp+ve (D = 0.93 ± 0.03), with a high PSC-833 modulation (5.0 ± 0.47), whereas U937 cells are pgp−ve phenotypically and have a low modulation ratio (1.12 ± 0.01).

To examine a role for pgp in apoptosis that could be separated from its role in drug uptake, a growth and serum factor withdrawal system was chosen, as this induces considerable apoptosis in most AML patient samples as well as in the pgp+ve growth factor–dependent TF1 cell line within 48 hours. Apoptotic cell death was quantified by flow cytometry using uptake of the dye 7-AAD in addition to low forward scatter. The quantitative use of 7-AAD to measure apoptosis was validated by comparison with the measurement of apoptotic cells identified by exposure of phosphatidyl serine using annexin V (Figure1). A close association was found, as has also been noted by others.30 Cultured leukemic patient cells do not undergo cell death at the same rate, and moreover they undergo considerable secondary necrosis, such that many of the nonviable cells observed after overnight incubation are necrotic rather than apoptotic. Figure 1 illustrates that the propidium iodide–positive (necrotic) cells are also annexin V positive, strongly indicating that necrosis was secondary to apoptosis. The apoptotic nature of cell death induced by growth factor withdrawal was examined in additional studies through MGM-stained cytospins, which showed cells displaying morphological characteristics of apoptosis, ie, condensed nuclear chromatin and membrane blebbing (data not shown).

In pgp+ve primary AML samples cultured overnight in 1% FCS without growth factors, there was an increase in apoptotic cell death in all samples assayed in the presence of the pgp-specific antibody UIC2 compared with the isotype- and concentration-matched control antibody, with a mean increase of 58% (range: 11%-95%;P < .05 using the Wilcoxon signed rank test) (Table2 and Figure2). TF1 cells were also cultured in 1% FCS without growth factors in the presence of UIC2 and control IgG2a antibodies. At 30 hours, apoptosis had increased 2.5-fold compared with that of cells cultured with isotype- and concentration-matched control antibodies (Figure 3). In the presence of growth factors, UIC2-induced apoptosis was less than 1% in all experiments, demonstrating that the antibody was not toxic. As an additional negative control, apoptosis was also measured in U937 cells deprived of growth factors overnight in the presence of UIC2 or control IgG2a. We did not observe UIC2-dependent apoptosis.

The effect of the pgp inhibitor PSC-833 on growth factor withdrawal in TF1 and primary AML blasts was also determined. TF1 cells were cultured with 0.5- and 1-μmol/L concentrations of PSC-833, which were found to enhance the cell death induced by growth factor withdrawal in a dose-dependent manner (Figure 4). In the presence of growth factors, PSC-833–induced apoptosis was less than 2% in all experiments, which demonstrates minimal toxicity for this compound.

A pgp independent role for PSC-833 in interleukin-1 (IL-1) secretion and leukemic blast survival has recently been postulated in AML.31 To determine whether the apoptotic effect of PSC-833 on growth factor withdrawn cells was pgp specific, pgp+ve and pgp−ve patient cells were cultured in the presence of PSC-833, which was found to enhance cell death in all samples tested including 6 pgp−ve samples (Figures 1 and 5). At 1 μmol PSC-833, in pgp−ve samples, the increase was 6%-350%, and the median was 53%; in pgp+ve samples, the increase was 13%-204%, and the median was 52%. There was no significant difference between groups using the Mann-Whitney U test. These experiments indicated that the enhancement of cell death induced by growth and serum factor withdrawal by PSC-833 is not pgp specific.

The role of the SM-ceramide pathway in apoptosis is controversial. The controversy largely surrounds the lack of specificity of the lipid reagents used and the methodology for measuring ceramide up-regulation.32 Before investigating the possible role of pgp in SM distribution, it was thought advisable to establish a role for the SM-ceramide pathway in the apoptosis of primary AML cells. The natural fungal product FB1 is structurally similar to the ceramide precursor sphinganine and inhibits ceramide synthase activity, thus inhibiting de novo ceramide synthesis.20 To the best of our knowledge, the specific inhibitory role of FB1 in ceramide synthesis has not been challenged, and this compound has previously been used to demonstrate a role for the ceramide pathway in the daunorubicin-induced apoptosis of U937 cells.20 We therefore cultured primary leukemic samples with and without daunorubicin and FB1. The presence of FB1 was found to reduce daunorubicin-specific apoptosis by a median of 40% (range: 18%-81%) in the 7 samples studied (Table 3). This implicates a direct role for ceramide in daunorubicin-induced apoptosis in AML cells.

The low bioavailability of plasma membrane inner leaflet SM has been postulated as a limiting factor in TNF-induced apoptosis in a pgp+ve leukemic cell model.22 To expand this hypothesis we constructed models using a fluorescent SM, C₆-NBD-SM, to label pgp+ve and pgp−ve leukemic cell membranes. If the hypothesis proposed by Bezombes et al22 were applicable, we would expect to observe that C₆-NBD-SM, transported to the inner leaflet of stressed pgp−ve cells, would augment apoptosis; whereas in pgp+ve cells, the exogenous SM would be largely confined to the outer leaflet, and apoptosis would not be enhanced. We measured the effects of exogenous C₆-NBD-SM on apoptosis. Preliminary experiments using flow cytometry to measure cell fluorescence enabled us to observe that C₆-NBD-SM uptake reached a plateau within 30 minutes at 4 μmol/L in pgp+ve cells (data not shown), and this loading schedule was adopted in subsequent experiments. An experiment was performed on pgp−ve U937 cells to determine whether C₆-NBD-SM induces apoptosis or whether it augments a preinduced apoptotic pathway. U937 cells survived for several days in 1% FCS; we therefore used daunorubicin (rather than serum and growth factor withdrawal) to induce apoptosis in these cells. Minimal apoptosis was induced by C₆-NBD-SM alone in the absence of daunorubicin (Figure 6A). However, in combination with daunorubicin, C₆-NBD-SM enhanced apoptosis. In contrast, C₆-NBD-SM was ineffective at augmenting apoptosis induced by growth factor withdrawal in pgp+ve TF1 cells (Figure 6B).

The effect of C₆-NBD-SM on the in vitro apoptosis of primary leukemic cells was then studied. Following a 30-minute incubation with C₆-NBD-SM, blast samples were pelleted and incubated for a further 18 hours at 37°C in 1% FCS. In 6 pgp−ve cases, increases of 19%-88% in cell death, median 33%, were observed, whereas in the 5 pgp+ve cases, increases of 0%-27%, median 7%, were noted (Table 4 and Figure7). The Mann-Whitney U test demonstrated the difference between groups to be significant atP = .028.

When exogenous phospholipids are incubated with cells in the presence of protein, the lipid retained in the membrane is affected by both uptake into the cell and back-exchange from the outer leaflet into the protein-containing medium.33 We therefore postulated that 18 hours after C₆-NBD-SM loading, fluorescence in leukemic cells would be related to the amount of SM internalized. The studies by Bettaib et al34 documented low inner leaflet SM content in pgp+ve KG1a cells compared with pgp−ve U937 cells and subsequently determined that PSC-833 can enhance inner leaflet SM content in the presence of an apoptotic stimulus.22Therefore, we were interested in C₆-NBD-SM fluorescence distribution in apoptotic and nonapoptotic cells. We examined cytofluorographs from samples that had been incubated overnight with C₆-NBD-SM in 1% FCS without growth factors (Figure8). From these, we determined that high C₆-NBD-SM fluorescence is largely associated with apoptotic cells. (Cells with reduced forward scatter and increased side scatter have been shown to be apoptotic in several models.35 36)

Patient cells were also treated with both C₆-NBD-SM and 1 μmol/L PSC-833. C₆-NBD-SM fluorescence was increased by PSC-833 in 10/11 patient samples (all except JD). Increased C₆-NBD-SM may reflect an alteration in SM distribution preceding apoptosis or alternatively may be nonspecifically associated with a loss of cellular integrity in these samples. To distinguish between these possibilities, we examined cultures from pgp+ve samples that underwent only minimal apoptosis at 18 hours. An increase in C₆-NBD-SM fluorescence was clearly distinguishable in the viable cells of samples treated with PSC-833 (Figure 8), supporting the finding that PSC-833 alters SM distribution.22 However, the pgp blocking antibody UIC2 was also used in conjunction with C₆-NBD-SM in these assays and was found to cause a slight fall in the C₆-NBD-SM fluorescence of viable pgp+ve cells (Figure 8). This suggests that PSC-833 is altering SM distribution, at least in part, by a pgp independent mechanism.

Using the pgp-specific blocking antibody UIC2, we demonstrated a drug efflux independent role for pgp in augmenting the survival of primary leukemic blasts and TF1 cells in vitro. A similar role in apoptosis augmented by growth factor withdrawal was previously demonstrated in Chinese hamster ovary fibroblasts.18 When added to the findings of Smyth16 and Johnstone,17 who have demonstrated a role for pgp in the fas- and UV-induced caspase-mediated apoptosis of T cell lines, it is clear that a fundamental role for pgp in leukemic cell survival, which is much broader than its drug efflux role, has now been established. It is important to note in this context that only blocking antibodies can be considered as totally pgp specific. UIC2 is a well-documented antibody that recognizes an extracellular portion of pgp but not of the MDR 2/3 gene product.26,37,38 UIC2 reactivity is increased in the presence of pgp substrate drugs or in experiments in which adenosine 5′-triphosphate (ATP) is directly depleted, and it has therefore been suggested that the inhibitory function of the antibody may involve the trapping of a transient conformation of pgp associated with ATP hydrolysis.39 Many pharmaceutical agents that inhibit pgp function, including PSC-833, have broader roles,31 and in this study we have shown that PSC-833 enhances cell death when growth and survival factors are withdrawn from pgp−ve as well as pgp+ve patient samples. A direct cytotoxic effect of PSC-833, independent of initial blast pgp status, has also been noted in an in vitro AML blast chemosensitivity assay.40 

Apart from pgp's function as a drug efflux pump, it has been suggested that pgp can transport phospholipids across cell membranes.21 In view of this, a potential role for pgp in reducing the bioavailability of SM has been postulated.22To determine whether reduced SM is relevant to the induction of apoptosis in AML blasts exposed to cytotoxic drugs, we initially used the ceramide synthase inhibitor FB1, which blocks the de novosynthesis of ceramide. In this system FB1 was found to reduce daunorubicin-specific apoptosis. In this respect our findings agree with those of Bose et al,20 who, like ourselves, cultured cells continuously with daunorubicin. In contrast, Jaffrezou et al19 found evidence for SM hydrolysis but not for de novo ceramide synthesis in the daunorubicin-induced apoptosis of U937 cells after a 1-hour pulse of daunorubicin. As ceramide is both a precursor and a breakdown product of SM,41,42the precise role of de novo ceramide synthesis vis-à-vis SM hydrolysis as contributors to apoptosis will be hard to establish and may partly depend on an experimental schedule. Furthermore, a direct drug-independent role for PSC-833 in enhancing ceramide synthesis has been shown in pgp+ve and pgp−ve epidermoid carcinoma cells.43 It also has been recently demonstrated that PSC-833 interacts with daunorubicin to induce SM hydrolysis and ceramide synthesis in drug-resistant HL60 leukemic cells.44 

As further evidence of a role for the SM-ceramide pathway in the apoptosis of AML cells, we have shown that exogenous fluorescence-labeled SM enhances loss of viability induced by daunorubicin in the pgp−ve U937 cell line. Although SM was not toxic when used without drugs, the availability of exogenous SM enhanced daunorubicin-induced apoptosis. The availability of SM is therefore shown to be a factor limiting the degree of apoptosis in AML cells in response to cytotoxic drug exposure. The evidence for a proapoptotic role for SM is further supported by the demonstration that SM augments apoptosis in pgp−ve patient blasts. In contrast, SM failed to enhance apoptosis in the pgp+ve TF1 cell line and had a significantly decreased effect on pgp+ve patient blasts compared with that seen in the pgp−ve cases.

Many of the models used to study pgp have involved the selection of drug-resistant sublines in vitro, which tend to express pgp at much higher levels than those found in unselected leukemic cell lines.45 The unselected TF1 cell line has low pgp expression compared to other pgp+ve AML cell lines.45However, even this cell line expresses more pgp than any of the 100-plus clinical AML samples studied in our laboratory (data not shown). For this reason we consider that it is important to use patient data to help understand whether an in vitro pgp model is likely to be pathophysiologically relevant in AML. However, it is also important, when evaluating the data which we have presented, to realize that a pgp+ve patient sample may contain a mixture of pgp+ve and pgp−ve blasts. Hence the differences in SM metabolism between pgp+ve and pgp−ve cells may be even more clear-cut than our patient data reveals.

The choice of a fluorescence-labeled SM for this study enabled us to observe the uptake and retention of this compound on a single-cell basis. We found that the addition of PSC-833 enhanced the SM fluorescence in viable cells, compatible with the finding of Bezombes et al22 that PSC-833 enhanced the internalization of SM, potentially providing an internal pool of hydrolyzable SM to up-regulate ceramide-mediated apoptosis. However, we were unable to establish that PSC-833 was acting specifically on pgp in this context because the pgp-specific UIC2 antibody, while inducing a similar degree of apoptosis at 18 hours, did not increase C₆-NBD-SM fluorescence. Cyclosporin A, which is closely related to PSC-833, is known to have a membrane fluidizing effect,46 which may account for an altered, pgp-independent distribution of phospholipids in the plasma membrane. In this context, cyclosporin A was also found to enhance C₆-NBD-SM fluorescence in our system (data not shown).

The central finding of this study has been the demonstration that when pgp is blocked, apoptosis induced by growth factor withdrawal in acute myeloid leukemic cells is enhanced. We have also demonstrated for the first time that the SM-ceramide pathway is involved in cytotoxic drug-induced apoptosis in primary AML blasts. Furthermore we have shown that exogenous SM enhances apoptosis induced by growth factor withdrawal in pgp−ve AML blasts, but that pgp+ve cells are protected from SM enhancement of apoptosis. PSC-833 was found not only to enhance apoptosis but also to increase cellular SM accumulation. The fact that PSC-833 enhances cell death and SM internalization in both pgp+ve and pgp−ve cases is of particular clinical relevance, as this agent is currently undergoing clinical evaluation, and these findings suggest that it may be of benefit to a broader range of patients than just those with pgp+ve blasts.

The question of how pgp exerts its antiapoptotic effects remains open. Pgp-mediated alterations of the electrochemical gradient across the plasma membrane are well documented and may be relevant in this context.47,48 Additionally, a role for pgp has been documented in the transport of several membrane lipids.21,49,50 There is also a reported role for pgp in cytokine export.51 Which, if any, of these mechanisms is relevant to the survival of leukemic clones is currently a major focus of research in our laboratory.