Unprecedented collateral sensitivity for cisplatin-resistant lung cancer cells presented by new ruthenium organometallic compounds

Ru compounds exhibit collateral sensitivity in cisplatin-resistant NSCLC and increase cisplatin activity by inhibiting efflux pumps.


Introduction
Cancer is a leading cause of death worldwide, with an estimated 9.6 million deaths in 2018. Among all types of cancer, lung cancer is the most common and most lethal disease, due to late diagnosis and intrinsic and/or acquired resistance. In particular, non-small cell lung cancer (NSCLC) accounts for 85% of all lung cancers and its treatment relies on surgery, chemotherapy, radiation, immunotherapy or combinations thereof. 1 The first line treatment for locally advanced stage NSCLC is simultaneous chemotherapy with different che-motherapeutics and radiation. 1 A multi-chemotherapeutics approach is frequently used and includes a platinum-based agent (such as cisplatin or carboplatin) and other drugs with different mechanisms of action such as a taxane or a vinca alkaloid. 2 Yet, due to drug resistance these therapeutic options have only limited success.
During the last years several research groups have developed new metallodrugs/metallodrugs formulations aiming at overcoming several forms of multidrug resistance. [3][4][5][6][7][8][9][10][11][12] However, in what concerns lung cancer, promising examples are still scarce. One of the strategies used has been the codelivery of valproic acid (VPA), a histone deacetylase inhibitor, with a metallodrug. 13,14 For example, Mao and co-workers have developed VPA-functionalized cyclometalated iridium(III) complexes through a hydrolysable ester bond. 13 The results have shown a significant increased activity for the new conjugates (vs. VPA alone or a mixture of complexes + VPA) validating the strategy adopted. Also, the conjugates were able to overcome cisplatin resistance in human lung carcinoma cells (A549R). In another work, Du, Meng and co-workers, used L-cysteine decorated Zr-based metal-organic frameworks (MOFs) to co-deliver cisplatin and VPA. 14 Immunoblot and immunofluorescence analyses showed that the new MOFs were able to downregulate the expression of vascular endothelial growth factor (VEGF) ruthenium MDR modulators, i.e., compounds able to block the drug efflux, only a few examples are known. Among them, the compound [Ru(η 6 -p-cymene)Cl 2 (N-(Anthracen-9-yl)-imidazole)] (Fig. 1, left), developed by Juillerat-Jeanneret, Dyson and coworkers, showed good cytotoxicity against A549 lung, HT29 colon, and T47D breast carcinoma and inhibited P-gp close to the levels of the reference inhibitor verapamil. 23 In this regard, we have been also engaged in the last years in the development of ruthenium(II) compounds with ATP Binding Cassette (ABC) efflux pumps inhibitory properties. Notably, during structure-activity studies on compounds from the [Ru II (η 5 -C 5 H 4 R)(2,2′-bipyridine-4,4′-R′)(PPh 3 )] + family, one compound (LCR134 for R = H, R′ = biotin ester, Fig. 1, right) has shown remarkable ability to inhibit P-gp (even better than the reference inhibitor verapamil) 24 and another stood out as a possible MRP1 inhibitor (RT11 for R = CH 3 and R′ = CH 3 ) (Fig. 1,  bottom). 25 In addition, both compounds (LCR134 and RT11) were also cytotoxic for breast (MCF-7 and MDA-MB-231) and ovarian (A2780/A2780cisR) cancer cells, respectively, exhibiting a rare dual behavior as both cytotoxic agents and ABC pump inhibitors. It was further observed that the substituent at the bipyridine is key to provide inhibitory properties to this family of compounds. Thus, in the continuation of these studies other substituents were added at the cyclopentadienyl ring (-CHO or -CH 2 OH) allowing the synthesis of a new family of compounds. The results disclosed in this work will highlight the importance that specific chemical groups have on the inhibitory properties of ABC pumps and how this might contribute to improve the activity of the compounds and to sensitize NSCLC cells to cisplatin.

Synthesis and characterization
For the synthesis of the new compounds, we first optimized the method described by Bai Wei et al. 26 to prepare the precur- sor [Ru(η 5 -C 5 H 4 CHO)(PPh 3 ) 2 Cl], by slightly changing the purification protocol avoiding chromatography (cf. Experimental section). Using this complex as starting material, three new cationic compounds with general formula [Ru(η 5 -C 5 H 4 CHO) (bipy)(PPh 3 )][CF 3 SO 3 ], where bipy is 2,2′-bipyridine (1), 4,4′dimethyl-2,2′-bipyridine (2) and 4,4′-bis(hydroxymethyl)-2,2′bipyridine (3), were synthesized in moderate to good yield by chloride abstraction with silver trifluoromethanesulfonate (Fig. 2, left). However, the reaction with 4,4′-dibiotin ester-2,2′bipyridine (bipy-biotin) originated a secondary product very difficult to eliminate from the mixture even by column chromatography or repeated recrystallization. In order to overcome this difficulty, the 'chemistry-on-the-complex' concept 27 was used. Thus, complex 3 was used as a platform to directly conjugate the envisaged vitamin via carbodiimide-based activation method using N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (Scheme S1 †). 28 This proved to be an efficient alternative method to afford complex [Ru(η 5 -C 5 H 4 CHO)(bipy-biotin)(PPh 3 )][CF 3 SO 3 ] (4) with a high yield (88%). Finally, we reduced the formyl group in the cyclopentadienyl ring of the previous four compounds into their hydroxymethyl analogues of general formula [Ru(η 5 -C 5 H 4 CH 2 OH) (bipyR)(PPh 3 )][CF 3 SO 3 ] (where bipyR = 2,2′-bipyridine (5), 4,4′dimethyl-2,2′-bipyridine (6), 4,4′-di(hydroxymethyl)-2,2′-bipyridine (7) and 4,4′-dibiotin ester-2,2′-bipyridine (8)) ( Fig. 2, right). This reduction reaction was successfully performed for complexes 1-3 in THF in a one-step high yield procedure, using sodium borohydride as reduction agent (Scheme S2 †). Yet, the reduction reaction of complex 4 did not lead to the expected hydroxymethyl analogue and the main product of the reaction contained the methylated-bipyridyl counterpart (confirmed by NMR and single crystal X-ray diffraction). Thus, under the reduction reaction conditions, the bipy-biotin ester bonds were found to be very sensitive and ended by being cleaved. Therefore, another approach was to reduce the formyl group at the precursor prior to the coordination of the biotinylated ligand, since the presence of three equally reactive primary alcohol groups at compound 7 prevented the use of the previously proposed 'chemistry-on-the-complex' approach. Unfortunately, all purification attempts of complex 8 were unsuccessful and therefore the characterization and, consequently, biological activity of complex 8 was not evaluated in this study. Thus, the seven new compounds were fully analyzed by 1 H, 13 C and 31 P-NMR, FTIR and UV-vis spectroscopies, ESI-MS and their purity confirmed by elemental analysis. In addition, single crystals of six compounds (1, 2, 3, 5, 6 and 7) were successfully obtained and studied using single-crystal X-ray crystallography. 1  was reported for related ruthenium cationic compounds. 29 Additionally, other resonances in the aromatic region (6.95 ppm < δ < 7.42 ppm) were ascribed to the aromatic ortho, meta and para protons of the triphenylphosphane co-ligand. As expected, compounds 5, 6 and 7 exhibit shielded resonances for the hydroxymethylcyclopentadienyl protons when compared to their formylcyclopentadienyl analogues, which is in good agreement with the electron donating character of the newly introduced pendant group. Detailed spectroscopic data concerning APT-13 C{ 1 H} and 31 P{ 1 H}-NMR experiments are included in the Experimental section and are in accordance with the effects discussed in this 1 H NMR analysis. The solid state FTIR of the complexes (KBr pellets) of the novel organometallic ruthenium-cyclopentadienyl compounds 1-7 shows the bands for the νC-H stretching vibration of the phosphane, cyclopentadienyl and bipyridyl ligands in the range 3100-3050 cm −1 and bands for νCvC ranging from 1489 to 1390 cm −1 . In addition, the presence of the triflate counter-ion was observed at 1250-1260 cm −1 . Characteristic stretching vibrations of the formyl and hydroxymethyl appended groups were also found at the expected region at ∼1670 cm −1 and ∼3420 cm −1 , respectively.
Optical absorption spectra of complexes 1-7 were recorded at room temperature using ∼10 −4 to ∼10 −6 M solutions in dichloromethane and dimethylsulfoxide. Table S1 † presents molar absorptivity coefficient (ε) values and the correspondent wavelength (λ max ) for the bands observed. Fig. 3A shows the spectra of compounds 1 and 5 which are representative of the behavior of formyl-or hydroxymethylcyclopentadienyl series of compounds (which exhibit very similar spectral features), whereas Fig. 3B illustrates the solvatochromic phenomenon observed for compound 3. The same trend is observed in the electronic spectra of all complexes with two very intense absorption bands at high energy values (λ = 240-330 nm) that are ascribed to electronic transitions occurring in the organometallic fragment ({[Ru(η 5 -C 5 H 4 R)((PPh 3 )] + } R = CHO or CH 2 OH) and the coordinated ligands. In the visible range, one or two medium-strength absorption bands assigned to metalto-ligand charge transfer transitions (MLCT) from Ru 4d orbitals to the π* orbitals of the phosphane and bipyridyl ligands (λ = 330-550 nm), as previously reported for related compounds. 25,30 The charge transfer character of these bands is corroborated by solvatochromic studies in DMSO where a clear blue-shift is observed with the increase of polarity of the solvent for compound 3 (379 nm in CH 2 Cl 2 vs. 366 nm in DMSO).
The crystal structures of complexes 1-3 and 5-7 have been solved by X-ray diffraction analysis. Fig. 4 displays the ORTEP diagrams of their molecular structures whereas the main crystallographic data and selected bond distances and angles can be found in the ESI † section (Tables S2-S3 †). All compounds crystallize in the monoclinic system, space group C12/c1 (1), P12 1 /c1 (3,7) or P12 1 /n (5, 6) with exception of compound 2 which crystallizes in the triclinic system, space group P1. In all cases the corresponding unit cells of the complexes display two enantiomers in the racemic crystal.
All complexes adopt the classical pseudo-octahedral threelegged piano stool geometry and, hence, the cyclopentadienyl arene rings exhibit the usual π-bonded η 5 coordination mode, whereas the bipyridyl ligands are coordinated in a bidentate fashion. The sixth coordination site is occupied by the triphenylphosphane ligands. The distances between Ru and the centroids of the cyclopentadienyl moiety in all the complexes, are within a narrow interval (1.828-1.841 Å), similar to other cyclopentadienyl ruthenium complexes described in the literature. 29 The Ru-P bond distances in the complexes are all in the same range (2.3086(11)-2.3351(3) Å), similar to the Ru-N bipy bond distances, which are in the same order for all the compounds (2.0685(11)-2.1112(3) Å). The Ru-C distances are in general higher than Ru-N bipy , probably due to the σ-donor character of the cyclopentadienyl rings. Moreover, the distances between Ru and the monosubstituted carbon on the η 5cyclopentadienyl ring in complexes 1-3 (Ru-C21 = 2.1901(13) 1; Ru-C2 = 2.212(3) 2; Ru-C21 = 2.1820(17) 3 Å) are in general slightly shorter than in complexes 5-7 (Ru-C21 = 2.230(2) 5; Ru-C2 = 2.213(3) 6; Ru-C21 = 2.235(3) 7 Å). This behavior could be due to the presence of the -CH 2 OH substituent onto the arene ring in the latter, which presents an electron-withdrawing character a lower than that of the -CHO substituent. In the case of complex 1 (see Fig. S29 †) an intramolecular hydrogen bond is formed between one pyridyl H atom and the  The angles N-Ru-N and C-Ru-C in the complexes, show the geometrical restrictions imposed by the bipyridyl and cyclopentadienyl ligands, and the values found are similar to other compounds described in the literature. 25,29 It is worth mentioning the packing structure of the compounds, where additional intermolecular hydrogen bonds can be observed between oxygen atoms of the triflate anions and the hydrogen atoms of the triphenylphosphane (1) or the bipyridine ligands (2,3); in addition, intermolecular hydrogen bonds are displayed between the oxygen atom of the -CH 2 OH substituent and the hydrogen atoms of the bipyridine ligand

Inorganic Chemistry Frontiers
Research Article of a neighboring molecule (5,6,7). In the case of 7 additional intermolecular hydrogen bonds are observed between O atom of the triflate anions and H atoms of the bipyridine ligand ( Fig. S30 †).

Stability in organic and aqueous media
All compounds were evaluated over time by monitoring their UV-Vis profile in 100% DMSO and 2-3% DMSO/DMEM in order infer about their stability in conditions mimicking the biological assays. These studies were performed during 24 h at room temperature. UV-Vis absorption spectra of the complexes in cellular media exhibit one strong absorption band in the UV range and two broad absorption bands in the visible range, similarly to their correspondent spectra in organic solvents (

Biological evaluation of the compounds
In order to access the biological activity of the new compounds, four non-small cell lung cancer cell lines (A549, NCI-H228, Calu-3, NCI-H1975 cells) with different expression levels of P-gp and MRP1 transporters were used ( Fig. S33 †), i.e. the two main transporters involved in cisplatin efflux. 31 Cells were incubated with complexes 1-7 and cisplatin and cell viability was assessed after a 72 h period and their cytotoxicity (IC 50 values) was determined using the WST-1 assay.
Analyzing the different IC 50 obtained for cisplatin (Table 1), we observed that the higher the level of MRP1 in the cells analyzed (A549 and NCI-H228), the greater their resistance to cisplatin (i.e., a higher IC 50 value is obtained), suggesting that MRP1 plays a key role in the efflux of cisplatin in our cellular models.
The cytotoxicity of cisplatin, commonly evaluated with the MTT assay, often results in lower IC 50 values in NSCLC cells 32 than those measured in the present work with the WST-1 assay. The different methods as well as the incubation time of the dye in the MTT assay (ranging from 1 to 4 h in different experimental works) can explain such discrepancies.
In a first experimental set, we evaluated the cell viability of the cell lines in the presence of increasing concentrations of the new ruthenium compounds (Table 1).
According to the IC 50 values obtained for the Ru compounds it can be noticed that four compounds -1, 2, 4 and 6were more active. For the panel of compounds tested herein, the presence of the substituent -CH 2 OH seems to be detrimental for the activity of the overall compound, whether placed in the cyclopentadienyl ring (in 5 and 7) or in the bipyridine ligand (in 3). The formyl substituent in the cyclopentadienyl ring seems advantageous, with compounds 1, 2 and 4 being active. The methyl substituent in the bipyridine co-ligand (complexes 2 and 6) is, in this panel of complexes, highly beneficial for the activity, rendering active complexes. Indeed, compounds 1, 2, 4 and 6, were stronger killers of A549 and NCI-H228 cells, i.e. the wo cell lines showing the highest resistance to cisplatin among those analyzed, suggesting that they could be inducers of collateral sensitivity of drug resistant cells.
This is not the first time that metal-based compounds are reported to exert a selective cytotoxicity against drug resistant cells: for instance, cobalt-based complexes 15 and ferrocenebased compounds 33 overcome resistance to taxol-resistant tumors. Ruthenium-based compounds have been described as potent anti-cancer drugs, 34-37 but their potential as collateral sensitivity inducers and/or MDR-reversing agents remains poorly known.
To verify if the compounds can improve the sensitivity to cisplatin, in particular in the cell lines with the highest degree   Table 2, was considered an index of sensitization toward cisplatin 38 and indicates how many folds the compound enhances the efficacy of cisplatin. All the compounds strongly sensitized A549 and NCI-H228 cell lines, showing a moderate sensitization effect in the mild cisplatin-resistant Calu-3 cells, and losing their effect on the cisplatin-sensitive NCI-H1975 cells.
Of note, A549 and NCI-H228 cells were the cell lines with the highest expression of MRP1, whereas in Calu-3 the most abundant transporter was P-gp. Thus, we decided to investigate if the Ru compounds can directly modulate the activity of these transporters by measuring the ATPase activity of cell extracts after incubation with the compounds (Fig. 5). This assay indicates if the compounds, once entered within the cells, bind the ABC transporters and interfere with their catalytic activity that is based on the hydrolysis of two ATPs associated to the efflux of the drug. Therein, a reduction in the ATPase activity of MRP1 or P-gp may suggest that the compounds inhibit the efflux activity of these transporters, e.g. by acting as inhibitors, substrates or negative allosteric modulators, as previously reported. 39 MRP1 and P-gp activities differed between the cell lines, in agreement with the expression levels of the protein, as observed for the controls. Interestingly, all Ru compounds inhibited MRP1 ATPase activity (Fig. 5A). They also inhibited P-gp activity in Calu-3 cells, that expresses P-gp to a greater extent (Fig. 5B). We could not detect any significant decrease of P-gp ATPase activity in A549 and NCI-H228 cells, most likely as a consequence of the lower amount of P-gp expressed. From the four complexes tested, compounds 4 and 6 seem to be the most efficient in inhibiting P-gp activity in Calu-3 cell extracts.
These results provided evidence that the selected ruthenium compounds increased the sensitivity to cisplatin in the resistant cell lines by directly impairing the catalytic activity of the main drug efflux transporters expressed the panel of lung cancer cell lines chosen. We are now investigating if this inhibition is exerted on other ABC transporters and by which molecular mechanisms. Considering our previous results on anti-cancer ruthenium cyclopentadienyl compounds, it seems that a trend is gaining some consistency: concerning their cytotoxicity, hydroxylated [Ru(η 5 -C 5 H 4 R)(PPh 3 )(bipy)] + (R = H, CH 3 , CHO or CH 2 OH) based compounds are less active or inactive in any of the cell lines tested (lung, breast or ovarian). This could possibly be related to their behavior as substrates of ABC pumps as we have previously observed for [Ru(η 5 -C 5 H 5 ) (4,4′-di(hydroxymethyl)-2,2′-bipyridine)(PPh 3 )] + ( pmc79) and [Ru(η 5 -C 5 H 4 CH 3 )(4,4′-di(hydroxymethyl)-2,2′-bipyridine)(PPh 3 )] + (RT12) compounds that were possibly P-gp substrates. 25,28 RT12 was also a poor substrate for MRP1, MRP2 and BCRP. On the contrary, [Ru(η 5 -C 5 H 5 )(4,4′-dibiotin ester-2,2′-bipyridine) (PPh 3 )] + (LCR134) and [Ru(η 5 -C 5 H 4 CH 3 )(4,4′-dimethyl-2,2′bipyridine)(PPh 3 )] + (RT11) were found to be P-gp and MRP1/ MRP2 inhibitors, respectively. 25, 28 We will now focus on understanding how these structural changes influence the ability of the compounds to act as inhibitors at the molecular level.  SO 3 ], where bipyR = 2,2′-bipyridine (1, 5), 4,4′-dimethyl-2,2′-bipyridine (2, 6), 4,4′-di(hydroxymethyl)-2,2′-bipyridine (3, 7) and 4,4′-dibiotin ester-2,2′-bipyridine (4), have been synthesized and completely characterized by analytical and spectroscopic techniques. In addition, the molecular structure of compounds 1-3 and 5-7 was unveiled by X-ray diffraction studies. All compounds were evaluated as prospective chemotherapeutics for lung cancer, and their activity assessed in a panel of different human lung cells. From this set of compounds, four of them (1, 2, 4 and 6) have shown strong activity against cisplatin-resistant non-small cell lung cancer cell lines, disclosing them as highly promising agents for the treatment of lung cancer. These four complexes were further tested for their ability to sensitize cisplatin-resistant cells. Complexes 1, 2, 4 and 6 increased cisplatin cytotoxicity up to 1390-fold by inhibiting MRP1 and P-gp transporters. This is of particular interest since, besides exerting an important role in cisplatin resistance, 40 MRP1 has been proposed as a potential mediator of collateral sensitivity. [41][42][43] Few compounds targeting MRP1 are inducers of collateral sensitivity 44 but the research in this field is in great expansion. In lung cancer, where cisplatin is the first-line treatment and MRP1 is one of the main responsible for resistance to this drug, 45 finding MRP1-targeting agents, inducing collateral sensitivity is absolutely preeminent. In this context, complexes 1, 2, 4 and 6 emerge as valuable prospective agents for lung cancer chemotherapy, specially compounds 4 and 6, both quite active as cytotoxic agents and able to inhibit both MRP1 and P-gp, suggesting an important ability to overcome drug resistance mechanisms.

Materials
All chemicals were purchased from commercial sources and used without further purification (unless otherwise stated).

Instrumentation and methods (Experimental section)
General procedures. All reactions and purification of compounds were performed under nitrogen atmosphere using Schlenk techniques. All solvents were used as purchased. Only dichloromethane, n-hexane and tetrahydrofuran used for synthetic procedures and work-up were dried using an MBRAUN solvent purification system (MB SPS-800, M Braun Inertgas-Systeme GmbH, Garching, Germany). NMR spectra were recorded on a Bruker Advance 400 spectrometer at probe temperature using commercially available deuterated acetone. Chemical shifts (δ) are reported in parts per million ( ppm) referenced to tetramethylsilane (δ 0.00 ppm) using the residual proton solvent peaks as internal standards. 31 P{ 1 H} NMR chemical shifts were reported downfield from external standard 85% H 3 PO 4 . The multiplicity of the peaks is abbreviated as follows: s (singlet), d (doublet), t (triplet), m (multiplet), comp (complex). Coupling constants ( J) are reported in Hertz (Hz). All assignments were attributed using COSY, HMBC, and HMQC 2D-NMR techniques. Infrared spectra were recorded on KBr pellets using a Mattson Satellite FT-IR spectrophotometer. Only considered relevant bands were cited in the text. Electronic spectra were recorded at room temperature on a Jasco V-660 spectrometer from solutions of 10 −4 -10 −6 M in quartz cuvettes (1 cm optical path). Elemental analyses were performed at Laboratório de Análises, at Instituto Superior Técnico, using a Fisons Instruments EA1 108 system. ESI-MS data acquisition, integration and handling were performed using a PC with the software package EAGER-200 (Carlo Erba Instruments). 26 but in this work another purification method was applied. A mixture of RuCl 2 (PPh 3 ) 3 (307 mg, 0.32 mmol), 6-dimethylaminopentafulvene (77 mg, 0.63 mmol) in 10 mL of THF/water (v/v = 7 : 3) was refluxed for 3 h. After that period, the reaction mixture was cooled to room temperature and concentrated to dryness and the aqueous phase discarded. The brown residue was washed with water (5 mL × 2), light petroleum ether (5 mL × 3), a mixture of light petroleum ether/diethyl ether (v/v = 3/3 × 3) and hexane (5 mL × 3). The experimental data obtained is in agreement with the previous literature.   [Ru(η 5 -C 5 H 4 CHO)(4,4′-CH 2 Biotin-2,2′-bipy)(PPh 3 )][CF 3 SO 3 ] (4).

X-ray structure analysis
The X-ray intensity data were measured on a D8 QUEST ECO three-circle diffractometer system equipped with a Ceramic X-ray tube (Mo Kα, λ = 0.71076 Å) and a doubly curved silicon crystal monochromator, using APEX3 46 software package. The frames were integrated with the Bruker SAINT. 47 Data were corrected for absorption effects using the Multi-Scan method (SADABS). 48 The structures were solved and refined using the Bruker SHELXTL. 49 The crystallographic data as well as details of the structure solution and refinement procedures are reported in ESI. † CCDC 2042407-2042412 † contain the supplementary crystallographic data for this paper.

Stability studies
For the stability studies, all complexes were dissolved in 100% DMSO and a sample containing each compound in 2% DMSO/ DMEM at 90-180 μM was prepared. Their electronic spectra were recorded in the range allowed by the solvent mixture at set time intervals. Samples were stored at room temperature and protected from light between measurements. The vari-ation percentage between measurements were calculated by the following expression:

Cytotoxic activity
Cells were seeded in 96 well plates. In a first experimental set, cells were incubated for 72 h with DMSO as a solvent or cisplatin at the following concentrations: 1 nM, 10 nM, 100 nM, 1 μM, 10 μM, 100 μM. In a second experimental set, cells were incubated with DMSO or each compound at the following concentrations: 1 nM, 10 nM, 100 nM, 1 μM, 10 μM, 100 μM. In the third experimental set, cells were incubated for 72 h with 1 μM of the selected compound plus cisplatin at the following concentrations: 1 nM, 10 nM, 100 nM, 1 μM, 10 μM, 100 μM. Cell viability was evaluated using the WST-1 assay (Sigma-Merck), as per manufacturer's instructions, using a Packard EL340 microplate reader (Bio-Tek Instruments, Winooski, VT). The absorbance units of the untreated cells were considered 100%; the absorbance units of the other experimental conditions were expressed as percentage versus untreated cells. IC 50 and IC 25 were defined as the concentrations of cisplatin of compound that killed 50% and 25% cells, respectively. The Resistance Factor was the ratio between IC 50 of cisplatin alone/ IC 50 of cisplatin + compound.

ATPase activity
The P-gp ATPase activity was measured in membrane vesicles as described previously. 39 Cells were washed with Ringer's solution (148.7 mM NaCl, 2.55 mM K 2 HPO 4 , 0.45 mM KH 2 PO 4 , 1.2 mM MgSO 4 ; pH 7.4), lysed on crushed ice with lysis buffer (10 mM Hepes/Tris, 5 mM EDTA, 5 mM EGTA, 2 mM dithiothreitol; pH 7.4) supplemented with 2 mM phenylmethylsulfonyl fluoride, 1 mM aprotinin, 10 μg ml −1 pepstatin, 10 μg ml −1 leupeptin, and subjected to nitrogen cavitation at 1200 psi for 20 min. Samples were centrifuged at 300g for 10 min in the pre-centrifugation buffer (10 mM Tris/HCl, 25 mM sucrose; pH 7.5), overlaid on a sucrose cushion (10 mM Tris/HCl, 35% w/v sucrose, 1 mM EDTA; pH 7.5) and centrifuged at 14 000g for 10 min. The interface was collected, diluted in the centrifugation buffer (10 mM Tris/HCl, 250 mM sucrose; pH 7.5) and subjected to a third centrifugation at 100 000g for 45 min. The vesicle pellet was re-suspended in 0.5 ml centrifugation buffer and stored at −80°C until the use, after the quantification of the protein content. 1 mg of total protein were immunoprecipitated with the anti-P-gp or anti-MRP1 antibody. 100 μg of each immunpurified protein was incubated for 30 min at 37°C with 50 μl of the reaction mix (25 mM Tris/HCl, 3 mM ATP, 50 mM KCl, 2.5 mM MgSO 4 , 3 mM dithiothreitol, 0.5 mM EGTA, 2 mM ouabain, 3 mM NaN 3 ; pH 7.0). The reaction was stopped by adding 0.2 ml ice-cold stopping buffer (0.2% w/v ammonium molybdate, 1.3% v/v H 2 SO 4 , 0.9% w/v SDS, 2.3% w/v trichloroacetic acid, 1% w/v ascorbic acid). After 30 min incubation at room temperature, the absorbance of the phosphate hydrolyzed from ATP was measured at 620 nm, using a Packard EL340 microplate reader. The absorbance was converted into nmoles hydrolyzed phosphate (Pi) per min per mg proteins, according to the titration curve previously prepared.

Statistical analysis
All data in the text and figures are provided as means ± SD. The results were analyzed by a one-way analysis of variance (ANOVA) and Tukey's test. p < 0.05 was considered significant.

Conflicts of interest
There are no conflicts to declare.