Clinical evidence shows that following initial response to treatment, drug-resistant cancer cells frequently evolve and, eventually, most tumors become resistant to all available therapies. We compiled a focused library consisting of >500 commercially available or newly synthetized 8-hydroxyquinoline (8OHQ) derivatives whose toxicity is paradoxically increased rather than decreased by the activity of P-glycoprotein (Pgp), a transporter conferring multidrug resistance (MDR). Here, we deciphered the mechanism of action of NSC297366 that shows exceptionally strong Pgp-potentiated toxicity. Treatment of cells with NSC297366 resulted in changes associated with the activity of potent anticancer iron chelators. Strikingly, iron depletion was more pronounced in MDR cells due to the Pgp-mediated efflux of NSC297366–iron complexes. Our results indicate that iron homeostasis can be targeted by MDR-selective compounds for the selective elimination of multidrug resistant cancer cells, setting the stage for a therapeutic approach to fight transporter-mediated drug resistance.
Modulation of the MDR phenotype has the potential to increase the efficacy of anticancer therapies. These findings show that the MDR transporter is a “double-edged sword” that can be turned against resistant cancer.
Clinical evidence shows that, following initial response to treatment, drug-resistant cancer cells frequently evolve, and eventually most tumors become resistant to all available therapies (1). Cancer cells can escape treatment by various mechanisms. The most straightforward cause of therapy resistance is linked to cellular alterations that prevent drugs to act on their target. For example, therapies hitting cancer-specific pathways are often blunted by mutations altering the target or by the activation of compensatory pathways. Alternatively, cell-intrinsic resistance mechanisms can change the “cellular pharmacology” of the cancer cell, influencing the uptake, metabolism or the efflux of the drugs to reduce their concentrations below a cell-killing threshold (1, 2). Upregulation of cell membrane efflux transporters of the ATP-binding cassette (ABC) superfamily leads to simultaneous resistance against structurally and functionally unrelated chemotherapeutic agents. In particular, P-glycoprotein (Pgp, MDR1), the product of ABCB1 gene, was shown to be expressed in several drug resistant malignancies, including ovarian carcinoma (3), acute myeloid leukemia (4), chronic myeloid leukemia (5), small cell lung cancer (6), and more (1, 2).
On the basis of the correlation of P-glycoprotein expression and function (7) with unfavorable treatment response, it is universally accepted that pharmacologic modulation of the multidrug resistant (MDR) phenotype has the potential to significantly increase the efficacy of currently available anticancer therapies. Unfortunately, despite a few early successes, clinical trials conducted with Pgp inhibitors did not fulfill this expectation, failing to confirm clinical benefit (8). Failure of the trials led to a setback in research, and the shutdown of the pharmaceutical development of transporter inhibitors for the improvement of anticancer therapy. Yet the “transporter problem” has not vanished, as evidenced by new studies supporting the relevance and benefit of research on the role of ABC transporters in clinical drug resistance (2). For example, studies conducted with genetically engineered mouse models (GEMM) mirroring many aspects of the human disease have shown that acquired resistance to docetaxel, doxorubicin, topotecan or olaparib is caused by the increased expression of the Abcb1 or Abcg2 genes, and the contribution of drug efflux to drug resistance was also confirmed in ABC transporter-deficient tumors (9).
Failure of the inhibitors has boosted research in other directions, exploring the possibility to evade efflux, or to exploit the paradoxical sensitivity associated with efflux-based drug resistance mechanisms (1). Alterations that confer selective advantage during the evolution of cancer cells might also create vulnerabilities that can be exploited therapeutically (10, 11). From this perspective, resistance can be interpreted as a trait that could be targeted by new drugs. In recent years we have undertaken a systematic effort to identify “MDR-selective” compounds exhibiting increased, rather than decreased toxicity against MDR cells. Our method, based on the correlation of toxicity profiles with Pgp expression patterns in the NCI60 cell panel, has led to the discovery of several compounds with robust, Pgp dependent toxic activity across diverse cell lines (12–15).
To date, the mechanism of Pgp potentiated toxicity of the MDR-selective compounds is not known, limiting intelligent drug design. Strikingly, MDR-selective compounds show structural coherence, highlighting chemical features that may be responsible for a common mechanism of action. In particular, there is a significant enrichment of metal chelators, indicating that metal–ion interaction could be key to the cytotoxicity of MDR-selective compounds (16, 17).
Here, we describe the relation of iron chelation to the MDR-selective toxicity of NSC297366, a compound containing an 8-hydroxyquinoline substructure possessing exceptionally strong Pgp-potentiated toxicity. We show that the toxicity of NSC297366 is linked to cellular iron depletion, which is exacerbated by Pgp.
Materials and Methods
MES-SA, MES-SA/DX5 cells were obtained from ATCC where they were characterized by DNA fingerprinting. The ABCB1-overexpressing subline MES-SA/B1 was produced by lentiviral transduction of pRRL-EF1-ABCB1. MES-SA mCherry and MES-SA/B1 mOrange cell lines were engineered to overexpress the fluorescent protein with lentiviral transduction of pRRL-EF1 plasmids. P-glycoprotein function was characterized by the Calcein AM assay (Supplementary Fig. S2; ref. 7). Cells were periodically tested and resulted negative for Mycoplasma contamination with the MycoAlert Mycoplasma Detection Kit (Lonza). Cells were cultured for not more than 30 passages or 3 months after thawing. Cells were cultured in DMEM (Gibco) supplemented with 10% fetal bovine serum (Gibco), 5 mmol/L L-glutamine (Gibco), and 50 U/mL penicillin/streptomycin solution (Life Technologies).
Doxorubicin was purchased from TEVA. NSC297366 was obtained from DTP's drug repository (NCI, NIH). Tariquidar was a kind gift from Susan Bates (NCI, NIH, Bethesda, MD). Verapamil, iron chloride, deferiprone, cobalt chloride, human holotransferrin and deferasirox were purchased from Sigma Aldrich. Iron chloride, cobalt chloride and transferrin were used in aqueous solutions, all other chemicals were dissolved in DMSO.
Cell viability assays
MTT assay was used to quantify cell viability, as described previously (14). For the drug combination tests, 384-well tissue culture plates were used. In each well, 2,500 cells were seeded in 20 μL medium, 1 day prior to the addition of the 2 compounds to be combined in additional 20 + 20 μL volumes. Fully automated pipetting steps were performed by a Hamilton StarLet liquid handling workstation. Plates were incubated for 6 days. Growth inhibition (GI) of MES-SA mCherry and MES-SA/B1 mOrange lines was assessed based on the detection of the respective fluorescent intensities scanned from the wells (585ex/610em for mCherry and 545ex/567em for mOrange). Data were normalized to negative (live cells, maximal fluorescence) and positive (dead cells, minimal fluorescence) controls. GI50 values of “compound A” with the fixed concentrations of “compound B” (and vice versa) were paired, and plotted on an equipotent graph as GI50 isoboles. For each data point of the isobole, significance was calculated as the combination index (CI; ref. 18). Drug combinations were considered additive at 0.85 ≤ CI ≤ 1.2; moderate and strong synergism was defined as CI < 0.85 and CI < 0.7, respectively; moderate and strong antagonism was defined as CI > 1.2 and CI > 1.45, respectively.
RNA isolation and RT-PCR
Cells grown in 6-well plates were homogenized in 1 ml TRIzol Reagent (Thermo Fisher Scientific) at 80% confluency. Total RNA was isolated from cell lysates using Direct-zol MiniPrep Kit (Zymo Research) according to the manufacturer's guidelines. cDNA samples were prepared from 1,000 ng total RNA using the AMV Reverse Transcription System Kit (Promega). TaqMan probes for ABCB1, RPLP0, TFRC, NDRG1 and VEGFA were ordered from Thermo Fisher Scientific. Real-time PCR analyses were carried out using the StepOnePlus Real-Time PCR System (Thermo Fisher Scientific); mRNA fold changes were determined by the 2−ΔΔCt method. Gene expression levels were normalized to RPLP0 (60S acidic ribosomal protein P0) expression.
Western blot analysis
Cells were lysed with MLB (No. 20–168, Merck KGaA) cell lysis solution containing 10 μg/mL aprotinin and 10 μg/mL leupeptin. Following SDS-PAGE and blotting, HIF1α was detected using anti–HIF1α Rabbit polyclonal antibody (ab51608, Abcam UK) in 1:500 dilution. Cell cycle–related protein levels were revealed using the cell-cycle regulation sampler kits (#9932 & #9870, Cell Signaling Technology Inc.). Anti-p53 antibody (#2524, Cell Signaling Technology) was used in 1:500 dilution. Anti–β-actin (clone AC-15, Sigma) was used to reveal endogenous control protein levels. Signals were detected using the HRP-conjugated goat anti-mouse IgG secondary antibody (Jackson ImmunoResearch). Immunoblots were revealed by the enhanced chemiluminescence system (Advansta Inc.).
After treatment, cells were trypsinized, resuspended in DMEM, and distributed to 5 ml sample tubes (3 × 105 cells). Before labeling, the cells were washed in 1 ml washing solution (0.5% BSA in PBS), to be resuspended in 50 μL of BSA-PBS. For cell-cycle measurement, we followed the two-dimensional cell-cycle analysis protocol (19), labeling the cells with propidium iodide (PI) and bromodeoxyuridine (BrdU, Sigma), the latter further labeled by anti-BrdU (clone B44, BD Biosciences) and anti-mouse IgG-Alexa488–conjugated antibody (Life Technologies). Apoptosis was measured by Annexin V and PI staining, using the Annexin V, FITC Apoptosis Detection Kit (Dojindo Molecular Technologies). Samples were analyzed with an Attune NxT flow cytometer (Life Technologies).
Determination of cellular dNTP levels
Cultured cells were treated with DMSO, hydroxyurea (2 mmol/L), or NSC297366 (4 μmol/L) for 24 hours. Cells were rinsed with PBS to remove residual media. dNTPs were extracted based on ref. 20, with a few modifications. Briefly, adherent cells were detached by trypsin, resuspended gently in 10 mL of ice-cold PBS, and an 500 μL aliquot was removed to determine the cell number using a hemocytometer. Samples were centrifuged for 10 minutes at 3,000 × g at 4°C, cell pellets were resuspended in 500 μL of ice-cold 60% methanol, placed at 95°C for 3 minutes. The extracts were centrifuged (16,000 × g for 5 minutes at 4°C), and the supernatant was transferred into a new tube and evaporated under centrifugal vacuum at 45°C. The resultant pellet was resuspended in 50 μl nuclease-free water ready to assay or stored at −20°C until use. The dNTP pool was determined using the fluorescence polymerase-based assay developed by Wilson and colleagues (20), with modifications as detailed in the Supplementary Material.
ATPase activity of crude membranes isolated from Spodoptera frugiperda (Sf9) cells expressing human P-glycoprotein was measured by colorimetric detection of inorganic phosphate liberation as described in (21). Preformed iron complexes of NSC297366 or 8-hydroxyquinoline were prepared by mixing the ligand with 0.5 equivalents of iron.
Total-reflection X-ray fluorescence measurements
MES-SA and MES-SA/B1 cells were pretreated with 25 μmol/L holotransferrin in serum-free medium for 4 hours (37°C) to fill intracellular iron pools. Following incubation with the chelators (see Fig. 6 legend for details), trypsinized cells were resuspended in DMEM supplemented with 10% v/v FBS, and were washed three times in 1 mL of PBS. In the accumulation experiments, cells were treated for 4 hours with 5 μmol/L of the iron-chelator complexes. Intracellular iron content was determined by total-reflection X-ray fluorescence (TXRF) as described in ref. 22. Briefly, cell pellets were digested for 24 hours at room temperature in a solution containing 20 μL 30% H2O2, 80 μL 65% HNO3, and 10 μL of 15 μg/mL Ga (Merck; added as an internal standard). Two microliters of the resulting solutions was pipetted on the quartz reflectors of an Atomika 8030C TXRF spectrometer (Atomika Instruments GmbH). The K α line used for determination of Fe was set to 6.403 keV.
A total of 106 cells were seeded in 6-well plates. After 4h treatment with 5 μmol/L FeCl3-NSC297366 complex at 37°C, cells were precipitated in 50 μL acetonitrile (Sigma). Following centrifugation, supernatants were analyzed on an AB Sciex 3200QTrap hybrid tandem mass spectrometer coupled to a PerkinElmer series 200 HPLC system. NSC297366 levels were quantified in multiple reaction monitoring (MRM) mode using phenylalanine as an internal standard (see Supplementary Material for more details).
IC50 values were calculated from normalized dose-response curves using GraphPad Prism 6 (GraphPad Software). Student (two-sided) t test was used to evaluate statistical significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Deviation from the additive isobole was determined by the CI, described in ref. 18. All data represent at least 3 independent experiments.
Screening of a focused 8-hydroxyquinoline library identifies MDR-selective derivatives whose toxicity is potentiated by P-glycoprotein
Previously, we identified several 8-hydroxyquinoline (8OHQ) derivatives as MDR-selective compounds targeting the collateral sensitivity of Pgp-expressing MDR cancer cells (14, 15). To increase the scope of structurally related MDR-selective compounds, we compiled a focused library consisting of >500 commercially available or newly synthetized 8-hydroxyquinoline derivatives and screened the compounds using the parental MES-SA and the multidrug resistant MES-SA/DX5 cell lines. This screening effort has confirmed the relevance of the 8-hydroxyquinoline backbone in MDR-selective toxicity, and the requirement of functional Pgp to mediate sensitization of MDR cells (Fig. 1A). NSC297366, in which the 8OHQ substructure is embedded into a ring is one of the highly active derivatives, and thus was chosen to characterize the mechanism of action of MDR-selective toxicity (results obtained with 4 additional MDR-selective 8OHQ derivatives are shown in Supplementary Table S1 and Supplementary Fig. S1). To explore the specific contribution and the functional relevance of P-glycoprotein in collateral sensitivity to NSC297366, we generated MES-SA/B1 cells by lentiviral transduction of MES-SA cells with an expression vector encoding human MDR1 cDNA. Using this isogenic cell line pair, we confirmed that the moderate expression of P-glycoprotein in MES-SA/B1 cells (Supplementary Fig. S2A-S2C) is sufficient to confer resistance to doxorubicin as well as increased sensitivity to NSC297366 (Fig. 1B). Both phenotypes were fully reversed by the specific Pgp inhibitor tariquidar, showing the reliance of collateral sensitivity on P-glycoprotein function. Notably, MDR-selective toxicity was not restricted to MES-SA/B1 cells. NSC297366 proved to preferentially target several further MDR cell lines with variable Pgp expression and different tissue of origin (Supplementary Table S2), whereas overexpression of ABCG2 did not sensitize cells (Supplementary Fig. S3A-S3B). To expand the scope of the in vitro models to cells expressing low Pgp levels, NSC297366 was assayed in sorted MES-SA/B1 cells (Supplementary Fig. S4), an independent MES-SA/B1 cell line transfected with lower viral titers (MES-SA/B1 M50) and the KB epidermoid carcinoma series selected from KB-3–1 cells in increasing concentrations of colchicine (KB-8, KB-8-5, KB-8-5-11; ref. 23). These results confirmed that paradoxical hypersensitivity of MDR cells is proportional to Pgp levels, pertaining to cells expressing clinically relevant levels of Pgp (Supplementary Fig. S5). Selectivity was abrogated by functional inhibition of Pgp, indicating that Pgp is both necessary and sufficient to convey sensitivity to this compound.
Modulation of cellular iron levels influences selective toxicity of NSC297366
Our recent work indicates that MDR-selective 8-hydroxyquinolines are capable of binding iron(III) and copper(II) ions (24). To test the role of iron in the toxicity of NSC297366, we repeated the cytotoxicity experiments in the presence of established iron chelators or excess iron (FeCl3). Excess iron attenuated the toxicity of NSC297366. Strikingly, the protective effect of excess iron was more pronounced in MES-SA/B1 cells (Fig. 2A), leading to a significant reduction of selective toxicity in the presence of higher FeCl3 concentrations (Fig. 2B; Supplementary Table S1). Conversely, MES-SA/B1 cells proved to be more sensitive to the combined effect of NSC297366 and the established iron chelator deferiprone. Whereas the combination of NSC297366 and deferiprone resulted in pronounced synergistic toxicity in MES-SA/B1 cells, the same two compounds were only additive in MES-SA cells or in MES-SA/B1 cells treated with tariquidar (Fig. 2C and D). Taken together, these results demonstrated that the MDR-selective toxicity of NSC297366 is linked to intracellular iron levels and suggested that P-glycoprotein expressing cells may be particularly prone to iron depletion caused by this compound.
P-glycoprotein increases the NSC297366-mediated induction of TfR mRNA levels and the stabilization of HIF1α
In iron-starved cells, upregulation of TfR1 levels ensure increased acquisition of iron through the receptor-mediated endocytosis of the transferrin-iron complex. In the next set of experiments, we evaluated the effect of NSC297366 on transferrin receptor (TfR1) mRNA levels in MES-SA and MES-SA/B1 cells. Despite similar steady state TfR1 levels, induction of TfR1 mRNA expression by 1 μmol/L NSC297366 was exclusively observed in MES-SA/B1 cells (Fig. 3A). The P-glycoprotein inhibitor tariquidar had no effect on baseline TfR1 levels but prevented TfR1 induction by NSC293766 in MES-SA/B1 cells. Whereas excess iron downregulated TfR1 levels in both cells, exogenous iron was no longer able to downregulate TfR1 levels in MES-SA/B1 cells in the presence of NSC297366. Again, this difference was not observed when the transporter was inhibited by tariquidar (Fig. 3B). Thus, in line with data presented in Fig. 2, these results confirm that cellular iron depletion induced by NSC297366 is exacerbated by P-glycoprotein. A decrease of the labile iron pool results in impaired prolyl hydroxylase domain enzyme (PHD) activity, thereby inhibiting hypoxia-inducible factor (HIF)-1α hydroxylation and proteasomal degradation (25). Treatment with NSC297366 or cobalt chloride (which inhibits PHD activity through a different mechanism; ref. 26) increased HIF1α protein levels (Fig. 3C), while HIF1α mRNA expression remained unchanged (Supplementary Fig. S6A). However, the NSC297366-induced increase of HIF1α levels was detected at much lower NSC297366 concentrations in MES-SA/B1 cells as compared to the parental cell line. Again, differential sensitivity of the cells was abrogated by excess iron or the Pgp inhibitor tariquidar (Fig. 3D). Increased HIF1α levels resulted in the upregulation of target genes such as VEGFA (Fig. 3E) or NDRG1 (Supplementary Fig. S6B), but not that of Pgp (Supplementary Fig. S7A-S7B). Transporter inhibition and excess iron had no significant effect in MES-SA cells, but the same treatment could prevent the selective, NSC297366-induced VEGFA expression in MES-SA/B1 cells (Fig. 3F).
Chelation is not sufficient to convey MDR-selective toxicity
The results presented so far indicate that collateral sensitivity of MDR cells may be due to the increased susceptibility of P-glycoprotein expressing cells to NSC297366-induced iron depletion. Since similar baseline TfR or HIF1α levels in MES-SA and MES-SA/B1 cells did not suggest a direct effect of Pgp on iron homeostasis, we compared the susceptibility of the two cell lines to a series of established iron chelators. Deferasirox (DFS), which is widely used for the treatment of iron overload, exhibited similar toxicity in MES-SA and MES-SA/B1 cells (Fig. 4A), and similar results were obtained with deferiprone and ciclopirox. Intriguingly, the unsubstituted 8OHQ core structure also showed equivalent dose dependent toxicity in MES-SA and MES-SA/B1 cells, suggesting that iron chelation per se is not responsible for MDR-selective toxicity (Supplementary Table S3). DFS induced a remarkable, concentration dependent increase in the expression of the HIF1α protein and (VEGFA) mRNA, but in contrast to NSC297366, this induction was identical in the two cell lines (Fig. 4B and C). These results clearly indicate that whereas chelation may be a necessary prerequisite to the MDR-selective toxicity (14), it is certainly not sufficient to exploit the collateral sensitivity of P-glycoprotein expressing MDR cells.
Selective toxicity of NSC297366 is mediated by the inhibition of ribonucleotide reductase, resulting in cell-cycle arrest and selective apoptosis in MDR cells
Selective depletion of iron in P-glycoprotein expressing MDR cells is expected to result in the targeted inhibition of the iron-dependent enzyme ribonucleotide reductase (RNR). A commonly described target of anticancer chelators, RNR catalyzes the conversion of ribonucleoside diphosphates (NDP) to deoxyribonucleoside diphosphates (dNDP) that are subsequently converted into deoxyribonucleoside triphosphates (dNTP; ref. 27). Treatment with the established RNR inhibitor hydroxyurea lead to an imbalance of the dNTP pools, resulting in the significant decrease of dATP levels in both cell lines (28). In contrast, treatment with NSC297366 did not affect MES-SA cells, but induced a selective depletion of dATP levels in MES-SA/B1 cells (Fig. 5A; further dNTPs are sown in Supplementary Fig. S8A–S8C; baseline RNR levels are shown in Supplementary Fig. S9). Consistent with the inhibition of RNR activity, NSC297366 caused a P-glycoprotein dependent increase of cell cycle suppressing proteins p53, p21Cip/Waf1 and p27Kip1, along with a moderate decrease in cyclin D3 (CCND3) and cyclin-dependent kinase 4 (CDK4) levels 24 hours after the treatment (Fig. 5B), which lead to a G1–S arrest marked by the dramatic decrease of G2 phase (Fig. 5C; Supplementary Fig. S10) and eventually the selective induction of apoptosis in MES-SA/B1 cells (Fig. 5D; Supplementary Fig. S11A and S11B). Addition of tariquidar rescued MES-SA/B1 cells from cell cycle arrest and apoptosis, proving that the MDR-selective toxicity of NSC297366 is linked to the efflux activity of P-glycoprotein.
P-glycoprotein depletes cellular iron below critical levels by effluxing the NSC297366–iron complex from the cells
Finally, to directly test if selective toxicity of NSC297366 is indeed linked to excess iron depletion, intracellular iron levels were followed by TXRF (22). Resting iron levels were not different, and loading with transferrin-bound iron resulted in a comparable increase of iron levels in both cell lines (Fig. 6A). Following incubation at 37°C, cellular iron levels showed a gradual decrease indicative of the activation of regulatory mechanisms restoring iron levels after loading with holotransferrin (29). The presence of 8OHQ did not influence this pattern, and treatment with NSC297366 also did not change the kinetics of iron release in MES-SA cells during the time course (12 hours) of the experiment. In sharp contrast, NSC297366 induced a significant reduction of iron levels in MES-SA/B1 cells, essentially depleting P-glycoprotein expressing cells of iron. Pgp mediated depletion of iron levels was confirmed in the KB cell series (Supplementary Fig. S12A). In both models, NSC297366-induced depletion of iron levels could be prevented by tariquidar, suggesting that the selective reduction of iron levels is due to the P-glycoprotein–dependent efflux of the NSC297366-iron complex. To verify this proposition, we evaluated the cellular accumulation of preformed iron complexes. Following incubation with NSC297366 in the presence of iron, cellular iron levels increased significantly in MES-SA and KB-3-1 cells, indicating that iron was shuttled into the cells by in situ formed complexes. In sharp contrast, iron uptake of MES-SA/B1 and the MDR KB cells was only incremental (Fig 6B; Supplementary Fig. S12B; further MDR-selective 8OHQ derivatives are shown in Supplementary Fig. S1). In parallel, intracellular levels of NSC297366 were significantly lower in MES-SA/B1 cells unless Pgp-mediated efflux was inhibited by tariquidar, which restored both iron and NSC297366 levels to that observed in MES-SA cells (Fig. 6B). Collectively, these data indicate that MDR cells accumulate less iron and NSC297366 because Pgp expels iron chelates from the cells. Finally, the role of Pgp in the transport of preformed iron complexes was verified by ATPase activity measurements (21). As shown in Fig. 6C, the iron complex of NSC297366 significantly stimulated the Pgp ATPase, whereas the iron complex of 8OHQ did not have any effect. These results indicate that the iron complex of NSC297366 is a P-glycoprotein substrate, providing an explanation of the selective iron depletion and the ensuing elimination of NSC297366-treated MES-SA/B1 cells.
There exists an unmet medical need for developing targeted and efficient anticancer drugs acting against drug resistant tumors expressing ABCB1/Pgp. Our prior work has identified several compounds whose toxicity is increased, rather than decreased by the activity of P-glycoprotein. We have shown that the function of Pgp is both necessary and sufficient to induce collateral sensitivity in MDR cells. That Pgp is necessary was evidenced by the loss of collateral sensitivity in the presence of specific efflux inhibitors; that Pgp is sufficient was demonstrated by the persistent MDR-selective toxicity of the compounds across diverse MDR cell lines (Supplementary Table S2; Supplementary Fig. S5) (14). However, the mechanism of action of MDR-selective compounds targeting multidrug resistant cells has remained a mystery. The results presented in this study clearly delineate a series of cellular events originating from the Pgp-mediated efflux of iron complexes, leading to the selective depletion of intracellular iron and the eventual elimination of MDR cells.
Several theories have been put forward to explain the paradoxical sensitization effect of Pgp. It has been hypothesized that high affinity substrates engage Pgp in a futile transport cycle, leading to the selective depletion of ATP in MDR cells. While such a mechanism would be consistent with the effect of efflux inhibitors on cellular sensitivity, we have been unable to demonstrate significant ATP depletion of MDR cells (16). Another model suggested Pgp-mediated lysosomal uptake to be responsible for the collateral sensitivity of MDR cells towards compounds from the structurally unrelated chelator class of thiosemicarbazones (30), but our own results did not confirm this hypothesis (14). Interpretation of these and other models (reviewed in ref. 16) should take into account specific alterations linked to the long-term in vitro drug selection used to derive resistant subclones. As a result of the drug-selection, multidrug resistant cell lines often exhibit cellular phenotypes with questionable relevance, such as mislocalization of abundantly expressed Pgp (31). Here our aim was to discern the specific contribution of Pgp to the mechanism of action of MDR-selective toxicity – we therefore engineered an isogenic cell line pair only differing in the expression of P-glycoprotein. Expression of Pgp at moderate levels conferred resistance to doxorubicin, but also hypersensitivity to NSC297366, proving that Pgp is indeed a “double-edged sword” (32).
NSC297366 belongs to a focused library of approximately 1,000 MDR selective agents that were designed around the 8OHQ scaffold. Although the 8OHQ scaffold is equally toxic to parental and MDR cells, the library includes several derivatives possessing variable levels of Pgp-dependent MDR selective toxicity (Fig. 1A; Supplementary Table S2). The anticancer activity of 8OHQ derivatives is linked to the ability of the compounds to form chelate complexes with metal ions (24). To understand the mechanisms leading to the selective death of MDR cells, we wanted to identify cellular phenotypes that only occur in NSC297366-treated MES-SA/B1 cells with uninhibited Pgp and are not induced in MES-SA cells or MES-SA/B1 cells treated with 8OHQ. Based on the described ability of 8OHQ to mobilize cellular iron (33), we focused our attention on changes in iron homeostasis.
Intracellular iron levels are tightly controlled by proteins regulating iron uptake, storage and export. Iron is acquired from holotransferrin, which is internalized by receptor-mediated endocytosis after binding to transferrin receptor 1 (TfR1). Early signs of iron depletion include the upregulation of TfR1 through the activation of iron regulatory proteins that stabilize TfR1 mRNA, and the stabilization of HIF1α protein through the reduced activity of iron-dependent PHDs. 8OHQ derivatives bearing weak structural resemblance to our compound set have recently been identified as inhibitors of PHD; however, the mechanism of action of those analogues were shown to rely on tight binding to the active site of the enzyme (34). In our hands, NSC297366 induced all of the changes associated with the activity of potent iron chelators such as deferasirox, deferiprone or the unsubstituted 8OHQ, which tightly sequester iron from the labile iron pool, leading to depletion of intracellular iron levels and the slowdown of iron-dependent cellular processes (35, 36).
Additional experiments performed with HIF1α silenced cells indicated that HIF1α does not play a direct role in MDR-selective toxicity (Supplementary Fig. S13A–S13D). Similarly, collateral sensitivity to NSC297366 is not tied to p53, as selective toxicity of NSC293766 persists in A431-ABCB1 cells expressing a mutant p53 protein variant that cannot bind to DNA (Supplementary Fig. S14A and S14B; ref. 12). Cellular iron depletion may lead to the death of cancer cells through multiple mechanisms, including inhibition of ribonucleotide reductase or induction of cell cycle arrest (37, 38). It has been shown that RNR can be regulated irrespective of the HIF1α or p53 status (39), in line with the ability of iron chelators to directly inhibit RNR by depleting intracellular iron pools (40).
8OHQ derivatives that are unable to chelate metal ions are not active (24). However, iron chelation alone cannot be responsible for MDR-selective toxicity, since we were not able to detect any differences in the iron homeostasis of MDR cells, and the toxicity of several iron chelators, including the parental compound 8OHQ, was found to be identical in MES-SA and MES-SA/B1 cells (Supplementary Table S3). We provide several lines of evidence supporting the link between MDR-selective toxicity and the efflux of iron complexes by P-glycoprotein: first, NSC297366 treatment results in the selective depletion of iron in MDR cells (Fig. 6A; Supplementary Fig. S12A); second, Pgp significantly hinders the intracellular accumulation of preformed iron complexes (Fig. 6B; Supplementary Figs. S1 and S12B); third, selective toxicity and iron depletion in MES-SA/B1 cells is abolished by tariquidar; fourth, the iron complex of NSC297366 stimulates the Pgp ATPase (Fig. 6C). It would be possible to propose alternative models based on indirect mechanisms resulting in the simultaneous decrease of iron and NSC297366 levels in Pgp-expressing cells. However, the most straightforward interpretation needing the smallest number of assumptions is that iron depletion is exacerbated by the Pgp-mediated efflux of iron complexes formed with MDR-selective 8OHQ derivatives. Extrusion of iron shifts the equilibrium towards increased iron mobilization from intracellular iron pools, leading to more significant iron depletion and the selective apoptosis of MDR cells (Fig. 7A–C; refs. 27, 41–43).
It has to be noted that chelators can harm cancer cells through further mechanisms, some of which would also activate p53 and HIF1α. Interestingly, the toxicity of NSC73306, which was the first MDR-targeting compound identified in the DTP database (12), was not significantly affected by iron, and MES-SA/B1 cells did not show reduced accumulation of the NSC73306-iron complex (Supplementary Fig. S1). Thus, it seems that the thiosemicarbazone (TSC) NSC73306 acts through a different mechanism, in line with the broad metal affinity of TSCs, which allows interaction with several biologically relevant metal ions including iron, copper and zinc (41). It has been suggested that the paradoxical vulnerability of MDR cells is based on an imbalance of the redox homeostasis (44). The anticancer activity of 8-hydroxyquinolines was shown to be at least partially mediated through complex formation with redox active copper and iron ions (24), and several compounds were reported to selectively kill Pgp-expressing cells, which are not chelating agents (16). However, at present it is unclear how the function of Pgp would increase the formation of redox-active complexes that could explain MDR-selective toxicity. The results presented in this paper clearly indicate that the targeted toxicity of MDR-selective 8OHQ derivatives is linked to the increased loss of iron as a result of the efflux of iron-complexes. We cannot rule out further mechanisms that may contribute to the general cytotoxicity of the compounds, such as depletion of an iron pool that is inaccessible to other chelators. Future work, exploiting the structure-activity relationship hidden in the focused 8OHQ library will address these exciting possibilities.
Understanding the molecular mechanism of action of MDR selective toxicity opens the way for the synthesis of a new class of drugs that can overcome and even exploit drug resistance. MDR-selective compounds are different from both transporter inhibitors and typical transporter substrates. In contrast to NSC297366, whose toxicity is exacerbated by Pgp, the therapeutic benefit of Pgp inhibitors such as tariquidar or zosuquidar depends on the inhibition of efflux of the concomitantly administered cytotoxic drugs. In MES-SA/B1 cells, NSC297366 shows moderate synergism with doxorubicin (Supplementary Fig. S15A–S15D). Whereas cytotoxic Pgp substrates show decreased activity in MDR cells, Pgp-mediated transport of MDR selective compounds results in increased toxicity, due to the smuggling of iron molecules outside of the cells. Conventional wisdom would dictate against the design of anticancer agents that are Pgp substrates. Paradoxically, in the case of MDR-selective compounds, the aim is to design 8OHQ derivatives whose iron complexes are recognized by the transporter and are expelled from the cells. Our focused 8OHQ library offers a good starting point for the identification of the critical structural determinants needed for Pgp-mediated transport, and thus, MDR-selective toxicity. In parallel, development will have to overcome challenges related to the low solubility of the compounds, which have thus far prevented in vivo studies. Several iron chelators, including deferoxamine (Desferal) and Triapine showed promising activity against hematologic diseases in phase II trials (41, 45, 46). However, if artefacts linked to the in vitro propagation of cell lines contribute to the elevated toxicity of MDR-selective compounds, it is unlikely that the concept of MDR-targeted therapy will be transferred to the bedside. In patients, drug resistance is achieved by low Pgp levels that can be also found in pharmacologic barriers that regulate the passage of drugs and xenobiotics (47). To be of clinical relevance, MDR-selective compounds should show selective toxicity in tumors expressing Pgp, without threatening physiological sites such as the blood brain barrier. As shown in Supplementary Fig. S5, selective toxicity of NSC297366 extends to cells expressing Pgp at clinically relevant levels. At the same time, hCMEC/D3 cells, which are routinely used as an in vitro blood brain barrier model, were shown to be insensitive to MDR-selective treatment (14). These results suggest that normal tissues expressing Pgp are not differentially sensitive to MDR-selective agents, in line with the increased reliance of malignantly transformed cells on iron. By analyzing the mechanism of action of a lead 8OHQ molecule, we show that iron homeostasis can be targeted for the selective elimination of multidrug resistant cancer cells. Our results lay the ground for the ligand-based development of MDR-selective compounds, setting the stage for a therapeutic approach to tackle transporter-mediated drug resistance.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: M. Cserepes, D. Türk, N. Szoboszlai, G. Szakács
Development of methodology: M. Cserepes, D. Türk, S. Tóth, V.F.S. Pape, A. Gaál, M. Gera, N. Kucsma, N. Szoboszlai, G. Szakács
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Cserepes, D. Türk, V.F.S. Pape, A. Gaál, J.E. Szabó, N. Kucsma, G. Várady, P.T. Szabó, J. Tovari, N. Szoboszlai
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Cserepes, D. Türk, S. Tóth, V.F.S. Pape, J.E. Szabó, B.G. Vértessy, P.T. Szabó, G. Szakács
Writing, review, and/or revision of the manuscript: M. Cserepes, D. Türk, V.F.S. Pape, J.E. Szabó, B.G. Vértessy, J. Tovari, G. Szakács
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Cserepes, N. Kucsma, C. Streli
Study supervision: J. Tovari, G. Szakács
We thank Judit Sessler for her help in the assessment of drug synergisms, Éva Bakos for her valuable advice, and András Füredi for the critical reading of the manuscript. G. Szakács was supported by a Momentum grant of the Hungarian Academy of Sciences and an ERC Starting Grant (StG-260572). Funding from the Austrian Science Fund (SFB35, G. Szakács) is also acknowledged. We acknowledge support from the National Research, Development and Innovation Office (PD124330 to J.E. Szabó; K128011 to G. Várady; K116295 to J. Tovari); Research and Technology Innovation Fund of Hungary FIEK_16-1-2016-0005 to G. Szakács; and by the ÚNKP-18-4-BME-391 National Excellence Program (J.E. Szabó). Financial support from the 2019 Thematic Excellence Program (TUDFO/51757/2019-ITM) is also acknowledged. J.E. Szabó is the recipient of the János Bolyai Research Scholarship of the Hungarian Academy of Sciences.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.