Abstract
Aim/Background: Various interactions between Caveolae membrane domains, multidrug resistance transporter P-glycoprotein (P-gp) and cholesterol have been suggested. We tested the assumption that anthracycline-induced P-gp and Caveolin-1 have correlated effects on cholesterol distribution in plasma membrane. Materials and Methods: The present study was performed in four lymphoblastic K562 cell lines expressing none (KS), one (Cav and KR cells) or both P-gp and caveolin-1 proteins (CavKR cells). Results: The CavKR cell line exhibits a significantly higher free cholesterol content than the other cell lines. Cholesterol distribution at the outer leaflet was distinct from the total cellular cholesterol by its accessibility to cholesterol oxidase (COase). When cells were ATP-deprived, cholesterol accessibility to oxidation was significantly delayed in CavKR cells. Caveolin-1 or P-gp expression did not induce detectable changes in membrane cholesterol accessibility to COase. Conclusion: Combination of functional P-gp, caveolae presence and lasting effect of anthracycline treatment appear determinant in free membrane cholesterol homeostasis and likely modulate cholesterol membrane order.
P-glycoprotein (P-gp or ABCB1) is a member of the ATP-binding cassette (ABC) transporters family, with broad substrate specificity. Acquired multidrug resistance (MDR) involving P-glycoprotein eventually leads to treatment failure in T-cell acute lymphoblastic childhood leukemia in young patients (1) and still places P-gp as one of the key prognostic factors for adult acute myeloid leukemia (AML) (2, 3). Recently, Kornblau et al. presented the blockade of adaptive defensive changes in cholesterol uptake and synthesis in AML by the addition of pravastatin in a phase 1 study (4). The authors observed that following exposure to cytotoxic agents, AML blasts elevate cellular cholesterol in a defensive adaptation that increases chemoresistance.
The multidrug resistance gene, MDR1/ABCB1, is coordinately regulated with the caveolin-1 gene in normal and leukemic bone marrow (5). Caveolin-1 coordinated expression with Caveolin-2 at the membrane level form hetero-oligomers and membrane vesicles. The so-called caveolae membrane vesicles, as well as the raft membrane domains, are cholesterol-rich membrane microdomains. Caveolin-1 protein can be phosphorylated on several residues and has been reported to interact with a long and diverse list of proteins (6, 7). The nature of the physical interaction between P-gp and the caveolae marker caveolin-1 (Cav-1) is unclear (8-10). Co-localisation of P-gp in membrane microdomains seems to depend on the cell type investigated, if not on the experimental procedures employed (11-13). Nevertheless, Cav-1 expression levels are often increased in MDR cancer cell lines compared to sensitive cell lines (14).
On the contrary, coupling at the cellular level between cholesterol and Cav-1 or between cholesterol and P-gp functionality, is extensively documented. Caveolae structures work as platforms for membrane cholesterol and cellular cholesterol homeostasis in several processes. Cav-1 is a cholesterol-binding protein (15). In addition to the favoured localisation of the LDL receptor in caveolae (16), a direct molecular interaction between Cav-1 and the cholesterol transporter ABCA1 has been reported (17). Non-esterified cholesterol also regulates dynamic caveolin trafficking from the Golgi complex to the cell surface and to lipid bodies (18). As for P-gp, cholesterol controls both its ATPase and drug-binding and efflux activities in many human drug selected cell lines, among which CEM acute lymphoblastic leukemia cells (19) and K562 erythrolymphoblastic cells (12), as well as in reconstituted systems (20). Eckford and Sharom proposed that cholesterol modulates P-gp function via drug membrane partitioning and changes to the local environment of the protein (21). Conversely, cholesterol has been suggested as a substrate of P-gp (22).
The aim of this study was to test the coupling between caveolae and P-gp function in living cancer cells on cholesterol distribution at the plasma membrane level. Different human erythrolymphoblastic K562 cell lines were used, expressing either or both P-gp and Cav-1 proteins. The wild type K562 cell line (KS) and the MDR P-gp-overexpressing K562/ADR (KR), derived from KS by doxorubicin (DOX) selection (23), do not express Cav-1. The caveolae structure is induced by Cav-1 gene transfection in KS (24) leading to the K562/Cav cell line (Cav). A K562 cell line expressing both P-gp and Cav-1 (CavKR) is obtained with DOX treatment of the Cav cells. Knowing that Cav-1 transfection is sufficient for induction de novo of caveolae in K562 (9), the CavKR cell line is an interesting model for investigation of putative P-gp and Cav-1 coupled effects onto cholesterol content and distribution within the plasma membrane. The stimulation of P-gp expression in caveolae-containing K562 by DOX significantly affects the cholesterol content and the mechanism maintaining the distribution of the membrane cholesterol.
Materials and methods
Cell lines. Cav-1 and green fluorescent protein (GFP)-expressing K562 cell line (Cav) as well as the GFP-expressing K562 cell line (Pinca) were a gift from Professor Sargiacomo (Istitute Superiore de Sanita, Rome, Italy). Both cell lines were obtained by a retroviral vector-based gene transfer procedure (24). K562/S cells (KS) and the DOX-selected, resistant counterpart cell line, KD56/ADR (KR) were a gift from Professor Marie (Hôpital Hotel-Dieu, Paris, France). K562 cell line expressing both P-gp and Cav-1 (CavKR) was selected from the Cav-1-transfected K562 by DOX treatment at increasing concentrations. All K562 leukemia cells were cultured in RPMI 1640 (Sigma-Aldrich Chemie, Saint-Quentin Fallavier, France) medium supplemented with 10% foetal calf serum (Biomedia, Boussens, France) at 37°C in a humidified incubator with 5% CO2. Cell cultures were split 1:2, one day before the experiment.
Every month, resistant KR and CavKR cells were cultured for three days with 400 nM of DOX. The cell lines were used one week later, for three weeks. The stability of P-gp expression was checked before each experiment by measuring the P-gp transport activity, i.e. the rate of the P-gp-mediated efflux of pirarubicin (PIRA, see below). All experiments were carried out with cells cultured without DOX for at least two passages.
Drugs and chemicals. Purified DOX and pirarubicin (PIRA) were purchased from Sigma-Aldrich and Pharmacia-Upjohn (Saint Quentin en Yveline, France), respectively. Concentrations were determined through measurements at 480 nm optical absorption of stock solutions diluted to approximately 10−5 M−1 with ε480=11500 l mol−1 cm−1. Stock solutions were prepared immediately before use. Cholesterol oxidase (COase) and Triton X100 were purchased from Sigma-Aldrich and were dissolved in water. The fluorescent probe 10-acetyl-3,7-dihydroxyphenoxazine (Amplex Red) was obtained from Molecular Probes (Eugene, USA) or Sigma-Aldrich. Anti-P-gp mouse serum antibody (C219) was suppplied by DAKO (Carpinteria, USA) and anti-Cav-1 (N-20 sc 894) rabbit serum was supplied by Santa Cruz Biotechnology (Santa Cruz, USA). Polyvinylidene difluoride membrane (Hybond-P) and the chemioluminescent-based antibody detection kit (ECL plus kit with mouse IgG, HRP-linked whole antibody) were purchased from Amersham Pharmacia Biotech (Orsay, France). The P-gp and MRP1 inhibitor, PAK-104P, was a gift of Drs Shudo, Iisaki and Akiyama, (Nissan, Chemical Industries, Tokyo, Japan). Other chemicals were of the highest available grade. Before the experiments the cells were counted, centrifuged and suspended in HEPES buffer solution containing 20 mM Hepes plus 132 mM NaCl, 3.5 mM KCl, 1 mM CaCl2, 0.5 mM MgCl2 at pH 7.3, with or without 5 mM glucose. Deionised double-distilled water was used throughout the experiments.
Western blotting measurement of P-gp and Cav-1 expressions. P-gp (12) and Cav-1 (25) detection by Western blotting was performed as previously described by the respective authors. Typically, equal volumes of cell suspension (0.12 ×105 cells in 12 μl PBS) were mixed with concentrated SDS reducing buffer (final concentrations were 0.75% SDS, 45 mm Tris, pH 6.8, 75 mM dithiothreitol). The samples were then incubated for 1 h at 50°C for P-gp detection, or for 5 min at 95°C for Cav-1 detection. Protein samples were separated on 7.5% (P-gp) or 12% (Cav-1) SDS/PAGE, and then transferred to a poly(vinylidene difluoride) membrane for 2 h and 20 min for P-gp in transfer buffer (25 mm Tris-base, 192 mm glycine, 0.1% SDS) and for 1 h and 45 min for Cav-1 in transfer buffer (25 mm Tris-base, 192 mm glycine, 0.1% SDS, 10% methanol). The membrane was blocked with 5% non-fat dry milk in 0.1% Tween/NaCl/Pi overnight at 4°C and treated with 0.5 μg/ml C219 anti-P-glycoprotein mouse serum antibody or with 1 μg/ml anti-Cav-1 for 2 h at room temperature. Detection by HRP-linked was performed according to the manufacturer's protocol.
Cellular anthracycline accumulation and P-gp-mediated efflux of PIRA. The rationale and validation of our experimental set-up to measure the anthracyclines accumulation and ATP-dependent anthracycline efflux rate in cancer cells have been extensively described and discussed previously (12, 26-29). It was based on the continuous spectrofluorometric monitoring (Perkin Elmer, LS50B spectrofluorometer, Courtaboeuf, France) of the decrease in fluorescence signal of PIRA at 590 nm (λex=580 nm) added to cell suspension (106 cells per ml).The cells were maintained with intact ATP levels with 10 mM glucose or were ATP depleted with 30 min pre-incubation in 10 mM sodium azide (Figure 2).
The decrease in fluorescence observed during the incubation of cells with PIRA resulted from the fluorescence quenching after PIRA intercalation between the DNA base pairs. This methodology allowed for the accurate measurement of the free cytosolic PIRA concentration under steady-state conditions and the kinetics of its active efflux. Levels of cellular anthracycline accumulation at steady state were related to the levels of the resistance phenotype after treatment of cells with DOX. The P-gp inhibitor PAK-104 (30) and Triton X100 (31) were used to demonstrate the involvement of P-gp in resistant cells.
The P-gp-mediated PIRA efflux was assayed in the four cell lines. The cells (1×106 /ml; 2 ml per cuvette) were pre-incubated for 30 min in HEPES buffer with 10 mM sodium azide, but without glucose (energy-deprived cells). The cells remained viable throughout the experiment, as checked with Trypan blue and calcein vital stain (not shown). After addition of PIRA, the signal decrease was monitored until the steady state was reached. Glucose was added and ATP levels recovered within 2 min. Consequently, PIRA fluorescence signal increased, due to its ATP dependent efflux by P-gp. The efflux rate, Va, expressed in mole cell−1 s−1, was calculated from the slope of the tangent to the F=f (t) curve (F is the fluorescence intensity at time t) after glucose addition. PIRA efflux rate with P-gp inhibitors, PAK-104 and Triton X100, was performed at non-permeabilizing inhibitor concentrations. PAK and Triton X100 concentrations ranged from 0.1 to 1.2 μM and 4 to 20 μM, respectively.
Determination of non-oxidised cholesterol in cholesterol oxidase-treated and untreated cells. Free cellular oxidised cholesterol was titrated by an Amplex Red-based fluorescent assay as previously described (12, 25), following a method provided in the Molecular Probe assay kit (32). Standard measurements were carried out in the presence of sonicated cell suspensions to which cholesterol (0 to 1.5 μM) was added. The standard curve was linear within this concentration range. Cholesterol titration was not affected by the presence of P-gp inhibitors or sodium azide. This method was used to determine the free cellular non-oxidised cholesterol content throughout the experiments.
Enzymatic oxidation of plasma membrane cholesterol. The accessible membrane cholesterol oxidation was measured as a function of incubation time with COase in the four cell lines CavKR, KR, Cav and KS. Intact ATP cells were incubated with 10 mM glucose and ATP-depleted cells were incubated with 10 mM sodium azide for 30 min prior to the experiments. Oxidation obtained at various incubation times was stopped by two washings with cold PBS and followed by titration of the remaining non-oxidised and non-esterified cholesterol. Incubation times with cholesterol oxidase ranged from 1 to 5 minutes. Throughout all the experiments, the cholesterol oxidase concentration was 0.01 U/ml. Short incubation times and low COase concentrations present the advantage to minimize the effect of variation of Cav-1 content in the plasma membrane (33).
Data were expressed as the ratio of non-oxidised cholesterol after COase treatment to total cellular cholesterol in untreated cells, [chol]+COase / [chol]‒COase. The variation of this ratio with time was an apparent normalized oxidation rate. The initial normalized apparent oxidation rate (Vinitial) was estimated from the slope of cholesterol oxidation rate between 0 and 1.5 minutes. From 2 minutes to 10 minutes, the non-oxidised normalized cholesterol ratio was constant and corresponded to a pseudo steady state. Treatment of cells with membrane-impermeant COase leads to specific oxidation of the membrane cholesterol from the outer leaflet (34), assayed here by the ratio [chol]+COase/[chol]–COase. The apparent normalized cholesterol oxidation rate was the result of cholesterol movements between the plasma membrane and intracellular pools, within the plasma membrane, and cholesterol efflux or uptake between extra- and intracellular medium.
Data analysis. Analysis of variance (ANOVA) with the Student-Newman-Keul test was performed in order to compare membrane cholesterol content between the cell lines. Throughout the analyses, statistical significance was accepted for p<0.05. The cholesterol content and cholesterol oxidation ratio are expressed as mean±standard deviation.
Results
MDR K562 cells expressing Cav-1 and P-gp. Exposure of increasing concentration of DOX for two years induced a stable MDR phenotype with a 400 nM DOX concentration in Cav cells. The K562 cell line obtained, called CavKR, expressed P-gp protein as shown in Figure 1A. Furthermore, Cav-1 protein was detected as an intense band in the CavKR cell line, suggesting an over-expression of Cav-1, compared to the Cav cell line (Figure 1B).
KR and CavKR cells presented a reduced PIRA intracellular accumulation, when compared to the KS and Pinca cells (Figure 2A). They exhibited a similar ATP-dependent efflux of PIRA (Figure 2B) that was reversed by the P-gp inhibitor PAK-104P and Triton X100 (data not shown). Va was 5.5±0.2×10−10 and 5.6±0.6×10−10 mole cell−1 s−1 in KR and CavKR cells, respetively. Concentrations inhibiting 50 % of PIRA efflux rate were 0.5 and 6 μM for PAK-104P and Triton X100, respectively, in both cell lines. Pinca and KS cells exhibit identical phenotypes, consequently, further experiments were carried out using the KS cell line as the unique control cell line concerning P-gp and Cav-1 expression.
Cellular cholesterol content. The cellular non-esterified cholesterol content was similar in the sensitive cell lines KS and Cav and resistant KR cell line but was significantly increased in CavKR cell line (5.6×10−15 mol l−1 cell−1) when compared to Cav and KR cells (4.8 and 4.4×10−15 mol l−1 cell−1, respectively). Cellular ATP-depletion by sodium azide incubation did not change the cholesterol content.
Cholesterol accessibility to enzymatic oxidation in cells with intact intracellular ATP levels. The ratio of non-oxidised cholesterol to total cholesterol was plotted as a function of time, in cells not deprived of ATP (Figure 3) for the four cell lines. The oxidation profiles were similar in sensitive and resistant K562 cells expressing and not expressing Cav-1 (KS,KR, Cav and CavKR cells). This ratio decreased below 20% at the first measurement at 1 min. Vinitial was equal or below 5×10−18 mol l−1 cell−1 s−1. The experimental conditions allow for a ratio measurement after 1 minute. An apparent steady state was reached at 2 min up to 10 min (data showed to 6 minutes) where the non-oxidised cholesterol represented 20% of the total free cellular cholesterol. This value does not include the esterified cholesterol content.
Cholesterol accessibility to enzymatic oxidation in ATP-depleted K562 cells. The ATP depletion caused different oxidation profiles for CavKR cells (Figure 4). Indeed, in the ATP-deprived KS, KR and Cav cells, as in previous experiments, non-oxidised cholesterol decreased up to 20% within 1 minute whereas in CavKR cells, it remained at its initial value during the 1.5 min COase treatment (Figure 4). Therefore, CavKR cells were not affected up to 1.5 minutes leading to a null Vinitial. Despite this delayed oxidation, at 2 minutes, the oxidised cholesterol ratio was similar to those of the other cell lines resulting in an apparent oxidation rate equal to 15×10−18 mol l−1 cell−1 s−1. The non-oxidised free cholesterol ratio at 2 and 6 minutes ranged between 20% and 40% for the four cell lines and were not significantly different considering the SD values.
Discussion
The free cellular cholesterol (unesterified) content and distribution within the plasma membrane were studied in leukemia cell lines expressing or not expressing P-gp and Cav-1. With this approach, the possible coupling between the MDR phenotype and caveolae structures with regard to plasma membrane cholesterol was investigated. The CavKR cell line was obtained by transfection with Cav-1-containing plasmid, followed by treatment with DOX for induction of P-gp expression. This approach completes previous studies that employed Cav-1-containing plasmid transfection in P-gp-expressing resistant cells or naturally P-gp and Cav-1-expressing leukemic bone marrow (5, 35). The CavKR cells, expressing both P-gp and Cav-1, are different from the three other cell lines used in this study as their free cholesterol content is higher and their membrane cholesterol accessibility depends on the presence of energy.
The higher free cholesterol content found in CavKR when compared to other cell lines is not, per se, a proof of coupled function between P-gp and caveolae. First, Cav-1 expression is described as an early membrane response to a stress caused by cytotoxic agents, taking place before drug resistance (35). Belanger et al. show that under drug pressure, up-regulation of Cav-1 expression by cytotoxic agents can be observed in still drug-sensitive cancer cells (35). Second, caveolae microdomains are reported as being membrane platforms for cholesterol traffic exchange (15, 17, 18). This suggests that the higher cellular non-esterified cholesterol content observed in CavKR may originate from a more important recruitment of cholesterol at the DOX-induced caveolae platforms, without link to P-gp expression. The effect of drug treatment does not involve P-gp as caveolae-free KR cells, also selected by drug pressure, do not display increased cellular non-esterified cholesterol when compared to KS. In a Cav-1 containing cell line transfected with mdr1 gene, cholesterol perturbation of cholesterol homeostasis was found only when cells were treated with vincristine leading Le Goff et al. to a similar conclusion (36). The apparent discrepancy of not observing an increase in cholesterol levels in Cav cells could originate from other cholesterol homeostasis-linked protagonists, besides Cav-1, that are down-regulated in DOX-treated CavKR cells.
The CavKR cell line is also remarkable in its membrane cholesterol accessibility. A number of different pathways have been identified by which free cholesterol is removed from cell membranes. There are three known mechanisms for free cholesterol efflux: aqueous diffusion, SRB1-mediated free cholesterol flux and ABCA1-mediated active efflux (37). Aqueous diffusion is a passive mechanism described for free cholesterol efflux occurring in all cell types. This diffusion takes place in the presence of cholesterol acceptor or donor particles, such as LDL lipoproteins, and is driven by the cholesterol concentration gradient (37). Nevertheless, the size of the acceptor particle will affect diffusion mediated collision and for this reason, large particles such as cells themselves, are inefficient acceptors. SRB1-mediated free cholesterol flux is a facilitated diffusion mechanism and involves binding to acceptors including LDL, HDL and SUV (37). ATP-dependant ABCA1-mediated pathway is an unidirectional net efflux that also involves acceptors (37).
These three mechanisms are negligible in our experimental conditions whatever the ATP level is. In fact, the two successive serum free washings and the use of a serum free incubation buffer throughout the experimental procedure, lead to an acceptor- and donor-free extracellular medium. Intracellular cholesterol trafficking and intra-membrane cholesterol movements such as flip-flop or lateral diffusion are then the major cholesterol movements affecting cholesterol membrane homeostasis in both leaflets in our experimental conditions.
As shown with the COase assay, the membrane cholesterol accessibility exhibits two different behaviours with time: up to 2 minutes, a fast membrane cholesterol oxidation and after 2 minutes, a partial recovery to an apparent steady state value. The COase assay provided estimates of the cellular cholesterol fraction at the plasma membrane. The ratios ranging between 20% and 40% for the four cell lines demonstrated that free cholesterol was mainly located at the plasma membrane level (between 80% and 60% of total cholesterol), which is in agreementwith other authors (38). As reviewed by Lange and Steck, COase activity depends on membrane lipid structure (39). The enzyme has been shown to interact at the surface of the bilayer rather than penetrating the membrane (40, 41). In the membrane model, the higher the order of lipid and cholesterol structure is the lower is the COase catalytic rate (42). The COase rate was proposed as an indicator of cholesterol membrane organisation imposed by membrane lipids on cholesterol (43, 44).
Considering these COase characteristics, Vinitial is a snapshot of the cholesterol membrane content, accessibility and organisation at the outer leaflet. Oxidation profile reflects cholesterol movements between the two leaflets and eventual intra-membrane cholesterol movements between ordered domains (low COase accessibility) and cholesterol disordered domains (high COase accessibility). It relates subsequent cell adaptation to the enzymatic cholesterol oxidation. High Vinitial is in agreement with cholesterol undergoing flip-flop between the two leaflets at a submicrosecond to second time scale (45, 46). Intracellular cholesterol trafficking takes place at longer time scales (several minutes) (47). These two mechanisms are not directly ATP-dependent.
For the cell lines investigated with intact ATP levels, no difference in Vinitial was observed, whatever the type of membrane microdomains was present (caveolae in Cav and CavKR cells or rafts in KS and KR cells). However, when experiments were carried out in ATP deprivation conditions, CavKR cells exhibited null Vinitial. For longer COase incubation times, oxidation profiles remain identical for the four cell lines. This all-or-none action of COase would originate from the ability of the reaction product, cholestenone, to displace cholesterol from its association with phospholipids, thereby promoting its enzyme susceptibility (37, 41, 42, 48, 49). The null Vinitial was observed only in CavKR cells, P-gp and Cav-1 expressing, anthracycline-selected cells and only when depleted of ATP. This result suggests a higher organized lipid bilayer or a lower movement of cholesterol from inner to outer leaflet in CavKR. Higher membrane organisation in CavKR may result from the combined presence of Pgp and of Cav-1 considering that the latter is reported to be critical for ordered plasma membrane domains (50). Indeed, in CavKR cells, an ATP-dependent mechanism may compensate the effect of caveolae on membrane cholesterol organisation by increasing outer leaflet cholesterol accessibility to COase. We suggest that this disorganisation is induced by P-gp and is suppressed in ATP-depleted cells. The phenomenon induced by P-gp is related to its ATP-dependent transporter activity and is a limiting rate within the scale time of the study. Nevertheless, the involvement of P-gp is not sufficient as in KR cells similar cholesterol accessibility profiles are observed with or without ATP.
From the data collected here, we conclude that anthracycline treatment and caveolae presence are responsible for an increase in the total free cholesterol content. The presence of a functional membrane P-gp and membrane caveolae induce change in the mechanism of free cholesterol accessibility at the plasma membrane level. The comparison between the different leukemic cell lines strongly suggests an energy-dependent phenomenon on membrane free cholesterol distribution in the presence of P-gp activity, and caveolae, linked to anthracycline cell line selection. As a consequence, anthracycline-based chemotherapy may have a role on cellular cholesterol homeostasis in cells with caveolae, underestimated for patient with AML. This provides a further argument of interest for cholesterol lowering-therapy in AML as proposed by Kornblau et al (4).
Acknowledgements
The Authors would like to acknowledge Professor Sargiacomo for providing the Cav cell lines. We thank Dr. S. Orlowski for helpful discussions.
- Received May 24, 2010.
- Revision received June 21, 2010.
- Accepted June 28, 2010.
- Copyright© 2010 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved