Localization of MRP-1 to the outer mitochondrial membrane by the chaperone protein HSP90b

Overexpression of plasma membrane multidrug resistance-associated protein 1 (MRP-1) in Ewing’s sarcoma (ES) predicts poor outcome. MRP-1 is also expressed in mitochondria, and we have examinedthe submitochondrial localization of MRP-1 and in- vestigated the mechanism of MRP-1 transport and role of this organelle in the response to doxorubicin. The mito- chondrial localization of MRP-1 was examined in ES cell lines by differential centrifugation and membrane solu- bilization by digitonin. Whether MRP-1 is chaperoned by heat shock proteins (HSPs) was investigated by immuno- precipitation, immunofluorescence microscopy, and HSP knockout using small hairpin RNA and inhibitors (apop- tozole, 17-AAG, and NVPAUY). The effect of disruptingmitochondrial MRP-1–dependent efflux activity on the cytotoxic effect of doxorubicin was investigated by counting viable cell number. Mitochondrial MRP-1 isglycosylated and localized to the outer mitochondrial membrane, where it is coexpressed with HSP90. MRP-1 binds to both HSP90 and HSP70, although only inhibitionof HSP90b decreases expression of MRP-1 in the mito- chondria. Disruptionofmitochondrial MRP-1–dependent efflux significantly increases the cytotoxic effect ofdoxorubicin (combination index, <0.9). For the first time, we have demonstrated that mitochondrial MRP-1 is expressed in the outer mitochondrial membrane and is a client protein of HSP90b, where it may play a role in thedoxorubicin-induced resistance of ES.—Roundhill, E.,Turnbull, D., Burchill, S. Localization of MRP-1 to theouter mitochondrial membrane by the chaperone protein HSP90b. FASEB J. 30, 000–000 (2016). www.fasebj.orgBoth intrinsic resistance and acquired multidrug resistance (MDR) are major challenges for the successful treatment of many patients diagnosed with Ewing’s sarcoma (ES),with 60–80% of patients dying due to disease progressionand less than half responding to second-line treatment (1).MDR is frequently associated with overexpression of se- lected ATP-binding transporter proteins in the cellmembrane (2), where they may actively efflux cytotoxic agents to reduce intracellular levels and drug efficacy. We and others have recently demonstrated that MDR transporter proteins can be expressed in different sub-cellular compartments of normal and cancer cells (3–5), including the mitochondria of ES (4). Because multidrug resistance-associated protein 1 (MRP-1) is preferentially transported to the mitochondria follow- ing treatment of ES cells with the MRP-1 substratedoxorubicin (4), we have hypothesized that MRP-1 in this organelle may increase drug efflux from the mito- chondria to inhibit induction of the mitochondrial death cascade and contribute to the development of the drug-resistant phenotype. In this study, we have therefore investigated the submitochondrial localization of MRP-1, the mechanism of MRP-1 transport to the mitochondria,and its role in the cellular response of ES to the MRP-1 substrate doxorubicin and the MRP-1–independent sub- strates vincristine and fenretinide.The substrate-adherent ES (A673, RD-ES, SKES-1, SK-N-MC, TTC 466, and TC-32), neuroblastoma (SHEP-1), rhabdomyosarcoma (A204), colon carcinoma (HT-29), breast cancer (MCF-7), and glioblastoma (T98G; positive control for MRP-1 expres- sion) cell lines were cultured as previously described (4). All cell lines were purchased from the American Type Culture Collection (Manassas, VA, USA), except for the T98G and HT-29 cell lines, which were a gift from Professor M. Knowles (Leeds Institute of Cancer and Pathology, University of Leeds, Leeds, United Kingdom).Subcellular fractions were prepared and proteins deglycosylated using the Protein Deglycosylation Mix (New England Biolabs, Ipswich, MA, USA) for analysis as previously described (4).1 Correspondence: Children’s Cancer Research Group, Leeds Institute of Cancer and Pathology, St. James’s University Hospital, Beckett St., Leeds LS9 7TF, United Kingdom.E-mail: [email protected] doi: 10.1096/fj.15-283408This article includes supplemental data. Please visit http:// www.fasebj.org to obtain this information. Subcellular fractionation, SDS-PAGE, and Western blotting were performed as previously described (4). The purity of frac- tions and equal protein loading was confirmed by Western blot for the mitochondrial markers Grp75 (1:1000, ab2799), porin (0.2 mg/ml) for the plasma membrane marker NaKATPase (4 mg/ml, ab7671), the nuclear marker TATA-TBP (1.3 mg/ml, ab818), and the cytoplasmic and total cellular protein marker b-actin (1:1000, ab8227; all from Abcam, Cambridge, United Kingdom). Expression of heat shock protein (HSP)70 and HSP90 was analyzed usingthe mouse mAbs anti-HSP70 (1 mg/ml, ab47455), anti-HSP90 (1 mg/ml, ab13492; identifies epitopes from both HSP90a and HSP90b isoforms), anti-HSP90b (0.25 mg/ml, ab53497), and anti-HSP90a (0.25 mg/ml, ab59459; all from Abcam). TNF receptor-associated protein 1 (TRAP-1) expression was determined using a mouse anti-TRAP-1 mAb (1 mg/ml, ab2721; Abcam). Inhibition of HSP90 was confirmed by decreased expression of the client proteins RAF-1 [0.2 mg/ml, sc-133; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA (6)] and AKT [1:1000, #9272; Cell Signaling Technology, Danvers, MA, USA (7)], examined by Western blot. MRP-1 expression wasdetected using the internal pre-MSD2–binding MRP-1 rat mAb (1 mg/ml, ALX-801-007-C125, MRPr1; Enzo Life Sciences, Farmingdale, NY, USA). Following immunoprecipitation, MRP-1expression was confirmed using a second mouse mAb (1:50, ALX-801-013-C500, MRPm6; Enzo Life Sciences) thatbinds to the internal C-terminal. Western blots were incubated with primary antibodies overnight at 4°C and secondary (0.4 mg/ml, Alexa Fluor; Molecular Probes, Invitrogen–Life Technologies, Carls-bad, CA, USA) antibodies for 1 h at room temperature. Proteinswere visualized and quantified using the Odyssey Infrared Imag- ing System (Li-Cor Biosciences, Lincoln, NE, USA). Mitochondria were isolated from ES cells (12 3 107) using the Mitochondria Isolation Kit (Mitoiso2; Sigma-Aldrich, St. Louis, MO,USA) following the manufacturer’s instructions, and then in- cubated with 5–25 mg digitonin per microgram protein for 30 min at 30°C to solubilize the outer mitochondrial membrane. Mito-chondria were then centrifuged at 10,000 g for 10 min at 4°C, to pellet any nonsolubilized mitochondria and mitoplasts (mito- chondria with no outer membrane) and separate these from the supernatant containing the solubilized outer mitochondrial mem- brane proteins. Mitochondrial bound and solubilized proteins were analyzed by Western blot. Solubilization of mitochondria was monitored by thedetectionof thesubmitochondrial fraction- specific marker porin, which appears asa doublet onthe membrane when analyzed by Western blot and we predict represents the phosphorylated and dephosphorylated forms of the protein (8).MRP-1–dependent efflux activity of isolated mitochondria and whole cells was measured using the calcein-F efflux assay (4).Briefly, ES cells were incubated for 30 min with calcein AM (Sigma-Aldrich): 0.05 mM for analysis of whole cells, or 1 mM for isolated mitochondria. Calcein-F accumulation was mea- sured (0 h) by flow cytometry using the FACSCalibur (BD Biosciences, Oxford, United Kingdom) with an excitation la- ser of 488 nm and emission detection using a 530/30 nm filter. Unlabeled control samples were included to correct for autofluorescence (4). Valinomycin (Sigma-Aldrich), which causes uncoupling of the electron transport chain and collapse of mitochondrial mem- brane potential (9), was used to inhibit mitochondrial MRP-1 activity. Valinomycin-dependent uncoupling of the electron transport chain was confirmed by the depolarization of the mi- tochondrial membrane, detected by cellular fluorescence fol- lowing incubation with 3,39-dihexyloxacarbocyanine (40 nM for 15 min), as previously described (10), and inhibition of mito- chondrial MRP-1 efflux activity using the calcein-F assay in isolated mitochondria and whole cells. To determine the response of ES cells to chemotherapeutics following mitochondrial disruption, TC-32 cells (5 3 104 per well) were seeded into Primaria 24-well plates (Corning, Corning, NY, USA) and allowed toadhere overnight. Cells were then treated with increasing concentrations of valinomycin (0.025–0.4 nM; Sigma-Aldrich) and the MRP-1 substrate doxorubicin (3.5–56 nM; Sigma-Aldrich) or the non-MRP-1 substrates actinomycin D (1–16 nM; Sigma-Aldrich) and fenretinide (0.4–7 mM; National Institutes of Health/National Cancer Institute, Bethesda, MD, USA) for 24 h.Viable cell number was determined using the trypan blue exclusion assay (Vi-Cell; Beckman Coulter, Brea, CA, USA) (4).The MRP-1 amino acid sequence was initially interrogated for the presence of 8–18 uncharged consecutive residues (11) and for enrichment with the amino acids arginine, leucine, and serine (12), key components of a mitochondrial targeting sequence (MTS). The MitoMiner, v3.1-2015_04 database of the mitochon-drial proteome [MRC Mitochondrial Biology Unit, Cambridge, United Kingdom (13, 14)] was also searched for the Homo sapiens MRP-1 gene (ABCC1). Using existing online databases containing previously described MTSs (iPSort, MitoProt, and TargetP), MitoMiner was used to assign a probability value for the expression of MRP-1 in mitochondria based on known MTSs (13, 14).ES cell lysates were prepared in RIPA buffer containing protease inhibitors as previously described (4). Lysed cells were collected and centrifuged at 10,000 g at 4°C for 10 min. The cleared cell extract (200 mg) was incubated with primary antibody (25 mg cell lysate/mg of antibody; HSP70 ab47455 and HSP90 ab13492; HSP90 ab13492 binds both HSP90a and HSP90b isoforms, TRAP-1 ab2721) for 16 h at 4°C with rotation. After which, 20 ml Protein A/G Plus-Agarose Immunoprecipitation Reagent (Santa Cruz Biotechnology, Inc.) was added, and the extracts were incubated with rotation for a further 4 h at 4°C. The agarose beads were then isolated by centrifugation at 1000 g for 3 min at 4°C, supernatant was discarded, and beads were washed 4 times with 1 ml RIPA buffer, with centrifugation at 1000 g for 1 min at 4°C between each wash step. The washed beads were then resuspended in sodium dodecyl sulfate loading buffer (4) and immunoprecipitates examined for MRP-1 expression by SDS-PAGE and Western blot using 2 different MRP-1 antibodies, as described above: internal pre-MSD2–binding MRP-1 rat mAb (;1 mg/ml; ALX-801-007-C125, MRPr1), and internal C-terminal–binding mouse mAb (1:50, ALX-801-013-C500, MRPm6). Protein extracts were incubated with beads alone (inthe absence of antibody) to control for nonspecific binding of proteins to the beads, and cell lysate in the absence of beads or antibody to control for any nonspecific protein precipitation. The effect of HSP70 and HSP90 inhibition on the expression of MRP-1 andviability of TC-32 and SK-N-MC ES cells was examined. Briefly, ES cells were seeded (5 3 104 cells per well into Primaria 24-well plates for viable cell counts and 10 3 106 per 10 cm dish for isolation of mitochondria) and allowed to adhere overnight.Cells were then incubated with increasing concentrations of the HSP70 ATPase activity inhibitor apoptozole [100–6400 nM for 48 h (15, 16)], and HSP90 inhibitors 17-AAG [25–1600 nM for 48 h, inhibition of both HSP90a (17) and HSP90b (18)] and NVPAUY [20–1280 nM for 24 h, inhibition of both HSP90a (19) and HSP90b (20, 21)]. HSP90 inhibitors were used at concen-trations previously described to decrease the expression of HSP90 client proteins and decrease viable cell number in vitro [17-AAG (17, 18) and NVPAUY (19–21)]. For analysis of mito- chondrial protein expression, mitochondria were isolated at the appropriate time point following treatment with HSP inhibitors [MITOISO2 (4); 6.25–1600 nM 17-AAG for 48 h, and 5–1280 nMNVPAUY for 24 h], and the expression of MRP-1 and the HSP90client proteins, AKT and RAF1, was determined by SDS-PAGE and Western blot. Viable cell number following treatment with HSP inhibitors was determined by trypan blue exclusion, using the Vi-CELL.TC-32 cells were infected with HSP70 (sc-29355-v), HSP90a (sc- 29353-v), HSP90b (sc-35606-v), MRP-1 (sc-35962-v), and control (sc-108080; all from Santa Cruz Biotechnology, Inc.) small hair- pin RNA (shRNA). Infected cells were selected in puromycin (0.5 mg/ml; Sigma-Aldrich) for 10 d before placing cells in nor- mal growth media. Knockdown of proteins in total protein ex- tracts and isolated mitochondria was confirmed by SDS-PAGE and Western blot.TC-32 cell populations (5 3 104 per well) were seeded into Primaria 24-well plates and allowed to adhere overnight, andthen treated with the HSP90 inhibitors, 17-AAG (50–800 nM) or NVPAUY (5–80 nM), and doxorubicin (7–112 nM) for 48 h. TC-32.shHSP90b, TC-32.shHSP90a, and shControl cells were also treated with doxorubicin (7–448 nM for 48 h). Viable cell number was determined using the trypan blueexclusion assay.Cells (2 3 104) were fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100, and incubated with primary antibody (1 mg/ml HSP70 ab47455, 1 mg/ml HSP90a ab59459, and 0.5 mg/ml HSP90b ab53497; MRP-1, 1 mg/ml, ALX-801-007-C125, MRPr1)for 1 h at room temperature or MitoTracker CMXRos (100 nM, M7512; Molecular Probes, Invitrogen–Life Technologies) for 30 min at 37°C followed bya 30 min incubation with the secondary antibody (0.4 mg/ml; Molecular Probes, Invitrogen–Life Technol- ogies), as previously described (4). Cells were then visualized by immunofluorescence (IF) microscopy using a Zeiss 200 inverted microscope (Carl Zeiss Ltd., Welwyn Garden City, United King- dom) and a Nikon Eclipse TE2000-E confocal microscope (Nikon UK Ltd., Kingston upon Thames, United Kingdom) (4).Unless otherwise stated, experiments were performed 3 times with triplicates in each experiment. Statistical analyses were per- formed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). Any significant differences in viable cell number or calcein-F efflux were determined using ANOVA followed byBonferroni’s post hoc multiple comparison test. Regression anal- ysis was performed on viable cell number to calculate the differ-ence between curves. Evidence of synergistic, additive, or antagonistic activity was calculated using CalcuSyn Software for Dose Effect Analysis (Biosoft, Cambridge, United Kingdom); a combination index (CI) .1.1 is antagonism, 0.9–1.1 is addi-tive, and 0.85–0.1 is synergism. RESULTS MRP-1 in isolated mitochondria is glycosylated and was heterogeneously expressed in the cell lines examined (Fig. 1A). It is capable of calcein-F efflux in mitochondria isolated from ES cell lines (Fig. 1B), consistent with the hypothesis that functional MRP-1 is expressed in the mi- tochondrial membrane. Interestingly, the calcein-F efflux activity of MRP-1 in the mitochondria is greater than that of MRP-1 in the plasma membrane of whole cells (P , 0.001; Fig. 1B), possibly reflecting the level of ATP in these different subcellular localizations.Solubilization of the outer mitochondrial membrane by treatment with digitonin was monitored by measuring porin expression. Expression of MRP-1 decreased with the loss of porin in the mitoplast fraction, consistent with localization of MRP-1 to the outer mitochondrial membrane (Fig. 1C).Inhibition of mitochondrial MRP-1 activity enhances the response of ES cell lines to the MRP-1 substrate doxorubicin, but not the nonsubstrates actinomycin D or fenretinideBecause we identified MRP-1 in the outer mitochon- drial membrane, we sought to establish if this might alter the response of cells to cytotoxics by disrupting the mitochondrial membrane potential and efflux activity using valinomycin (0.025–0.4 nM). TC-32 viable cellnumber was significantly decreased after treatment for24 h with valinomycin and the MRP-1 substrate doxo- rubicin, when compared to the cell death induced by doxorubicin alone (0.025 nM valinomycin and 7 nM doxorubicin, CI = 0.518; and 0.05 nM valinomycin and 7 nM doxorubicin, CI = 0.803; Supplemental Fig. 1A). However, there was no such additive effect when cells were treated with valinomycin and the non-MRP-1 substrates, fenretinide (CI . 1; Supplemental Fig. 1B) or actinomycin D (CI . 1; Supplemental Fig. 1C). Valinomycin-induced uncoupling of the electron transport chain and collapse of mitochondrial mem- brane potential (9) were confirmed by a significant decrease in 3,39-dihexyloxacarbocyanine fluorescence (up to 10 nM valinomycin; P , 0.05) and disruption ofmitochondrial MRP-1–dependent efflux activity (up to 10 nM; decrease in calcein-F efflux by 9 6 2%; P , 0.05);the amino acids arginine, leucine, and serine (12). Fur- thermore, interrogation of the online database, MitoMiner, did not predict for the presence of an MTS (probability of an MTS: iPSort 0.0, MitoProt 0.0778, and TargetP 0.106). Because the chaperone proteins HSP70 (22) and HSP90(23) are reported to transport hydrophobic proteins to the mitochondrial outer membrane, through an associa- tion with a translocase of the mitochondrial outer mem- brane [TOM70 (24)], we went on to investigate the hypothesis that the hydrophobic MRP-1 may be an HSP70 or HSP90 client protein. plasma membrane MRP-1 activity remained unchanged (up to 10 nM; P . 0.05).Because we have established localization of MRP-1 to the mitochondria, we hypothesized that MRP-1 may contain an MTS. Interrogation of MRP-1 using the online database MitoMiner and literature searches failed to identify any components of a previously described MTS, including 8–18uncharged consecutive residues (11) and enrichment with MRP-1 coprecipitated with HSP70 and HSP90 in 6 out of 6 ES cell lines examined, demonstrated by Western blot of immunoprecipitates using 2 independent MRP-1 an- tibodies (Fig. 2). Interestingly, both HSP70 and HSP90 appeared to preferentially precipitate lower molecular mass MRP-1s [;100 kDa (4); Fig. 2]. This was most striking when immunoprecipitates were blotted with the MRP-1r1 anti- body (Fig. 2B), which binds MRP-1 in the first cytoplasmicloop at amino acids 1511–1520; the MRP-1m6 antibody binds MRP-1 in the intracellular C terminus at amino acids 238–247. The functional relevance of the 100 kDa variant is yet to be established. The specificity of HSP70 and HSP90 immunoprecipitations was confirmed by Western blot of TC-32 precipitates for HSP70 (Supplemental Fig. 2A) and HSP90 (Supplemental Fig. 2B). This was independent ofthe HSP90-like chaperone TRAP-1, which is expressed in the mitochondrial matrix (25) and did not precipitate with MRP-1 in ES family of tumor cells (Supplemental Fig. 2C).Because MRP-1 coprecipitated with HSP70 and HSP90 proteins from ES cells, we went on to examine whether knockdown of either HSP70 or HSP90 decreased MRP-1 expression in the mitochondria. Expression of HSP70 was decreased in cells treated with the HSP70 inhibitor apoptozole (15) or using shRNA to HSP70. MRP-1 expres- sion in mitochondria isolated from ES cell lines treated with apoptozole (Fig. 3A, B) was unchanged, suggesting that HSP70 does not chaperone MRP-1 to the mitochondria. Consistent with this observation, MRP-1 expression was the same in mitochondria isolated from TC-32.shHSP70 and shControl cells (Fig. 3C). The HSP70 doublet observed only in the total cell fraction (Fig. 3C) is likely to represent both the full-length HSP70 (70 kDa, HSPA1A) and also alternative splice variant (64 kDa) (26). The isolated mi-tochondria (Fig. 3A–C) only express the smaller (64 kDa) splice variant, although the significance of differential splice variant expression requires further investigation.Confirming the specificity of HSP70 shRNA knockdown, the expression of HSP90a and HSP90b was unchanged when examined by Western blot (Fig. 3C). Immunofluorescent confocal microscopy confirmed the knockdown of HSP70 in TC-32.shHSP70 and revealed that HSP70 and MRP-1 colo- calized to the nucleus (Fig. 3D), in agreement with previous studies (4) and the detection of HSP70 inthe nuclear fraction of ES cells by Western blot (Supplemental Fig. 2D). Inhibition of the stress-induced HSP90a and constitu- tive HSP90b using 17-AAG or NVPAUY decreased MRP-1 expression in mitochondria isolated from TC-32 (Fig. 4A) and SK-N-MC (Fig. 4B) cells, as demonstrated by a decrease in the expression of the HSP90 client proteins RAF-1 (6) and AKT (7). Expression of total cellular MRP-1 remained unchanged (data not shown). These observations suggest that HSP90 does not have a role in maintaining total cel- lular MRP-1 expression but regulates the intracellular transport of MRP-1 to the mitochondria. Expression of MRP-1 was decreased over the same time course and treatment conditions (Fig. 4) as RAF-1 and AKT, support- ing the hypothesis that MRP-1 is also an HSP90 client protein. Expression of the non-HSP90 client protein and loading control, GRP75, was unchanged in isolated mito-chondria following treatment with 17-AAG and NVPAUY (Fig. 4). In agreement with previous studies (18, 20, 27–29) and a hallmark of HSP90 inhibition, expression of HSP70 was increased following treatment with 17-AAG and NVPAUY (Fig. 4) because HSP70 is also reported to be acochaperone of HSP90 (30).The compounds 17-AAG and NVPAUY inhibit both HSP90a and HSP90b isoforms (17–20); therefore, we utilized shRNA to investigate which of these HSP90 proteins might be chaperones for MRP-1 to the mito- chondria. Suggesting that HSP90a is not a chaperoneprotein for MRP-1 to the mitochondria, there was no change in the expression of MRP-1 in both total (Fig. 5A) and mitochondrial (Fig. 5B) fractions isolated from TC-32.shHSP90a and TC-32.shControl cells. Consistent with this, MRP-1 expression in the mitochondria was retained when examined by IF microscopy in TC-32.shHSP90a cells (Fig. 5C–F and Supplemental Fig. 3A) compared to TC-32.shControl cells (Supplemental Fig. 3B). In agreement with the precipitation of MRP-1 usingthe HSP90 antibody (which binds both HSP90a and HSP90b isoforms), cytoplasmic colocalization was ob- served between MRP-1 and HSP90a in TC-32 cells (Fig. 5D and Supplemental Fig. 3C) and was decreased in TC- 32.shHSP90a cells (Fig. 5C and Supplemental Fig. 3A, red arrow) compared to TC-32.shControl cells (Sup- plemental Fig. 3B, red arrow). These observations suggest that HSP90a may have a role regulating thetransport of MRP-1 to cytoplasmic organelles or vesicles, which requires further investigation.In contrast, MRP-1 expression was decreased in the mi- tochondria of the TC-32.shHSP90b cells compared to that in the TC-32.shControl cells (Fig. 5B). There was no change in MRP-1 expression in the total cellular protein fractions isolated from the TC-32.shHSP90b cells (Fig. 5A), demonstrating that HSP90b does not control the expres- sion of total MRP-1 but regulates the transport of MRP-1 to the mitochondria.Colocalization of MRP-1 and HSP90b to the mitochon- dria was confirmed by IF microscopy in TC-32 cells (Fig. 5E and Supplemental Fig. 3D). Consistent with a role for HSP90b in the transport of MRP-1 to the mitochondria, expression of MRP-1 was decreased in the mitochondria of TC-32.shHSP90b cells (Fig. 5F and Supplemental Fig. 3E) compared to TC-32.shControl cells (Supplemental Fig. 3B). Some residual HSP90b expressionwas observed inthe TC-32.shHSP90b cells by Western blot (Fig. 5B), and consistent with the role of MRP-1 as an HSP90b client protein, residual MRP-1 expression was also observed in TC-32.shHSP90b cells by both IF microscopy (Fig. 5F and Supplemental Fig. 3E) and Western blot (Fig. 5B). Because we have demonstrated that MRP-1 is a client protein for HSP90 and increased expression of MRP-1 can decrease the cytotoxic activity of some chemotherapeutics (4), we went on to test whether inhibition of HSP90 de- creased ES cell number and whether this might sensitize cells to the effect of MRP-1 substrates. Compounds 17-AAG and NVPAUY decreased cell viability of both TC-32(Fig. 6A; EC50 17-AAG, 79 nM, and 95% confidence in- tervals 45–81 nM) and SK-N-MC (Fig. 6B; EC50 17-AAG,304 nM, and 95% confidence intervals 87–1023 nM) cells,demonstrating that these cells are sensitive to inhibition ofHSP90. Because NVPAUY at the concentrations shown to inhibit HSP90 induced .40% cell death in both cell lines, it was not possible to generate an EC50 for this compound (Fig. 6A, B). In contrast, TC-32 (Fig. 6C) and SK-N-MC (Fig. 6D) ES cells were resistant to the HSP70 inhibitor apoptozole; cell death was ,10%.Incubation of cells with 17-AAG (Supplemental Fig. 4A) or NVPAUY (Supplemental Fig. 4B) did not increase the sensitivity of SK-N-MC cells to the MRP-1 substrate doxo- rubicin (CI . 1). Furthermore, TC-32.shHSP90b and TC-32.shHSP90a cells were not sensitized to doxorubicin compared to shControl cells (P . 0.05; Fig. 6E). These results suggest that inhibition of HSP90 proteins and treatment with doxorubicin are unlikely to be effective as a combination treatment for ES using current small molecules. DISCUSSION For the first time, we have demonstrated expression of glycosylated MRP-1 in the outer mitochondrial membrane of cancer cells. We have been unable to identify an MRP-1 MTS, although we have demonstrated that mitochondrial MRP-1 is a client protein for the chaperone HSP90b. HSPs are known to transport some hydrophobic proteins to the mitochondrial outer membrane (22, 23), but this is the first evidence that they regulate the subcellular trans- portation of the ABC transporter protein MRP-1.Importantly, cells were more sensitive to the cytotoxic activity of the MRP-1 substrate doxorubicin when cotreated with valinomycin at a concentration that decreased efflux activity in the mitochondria, but not in the plasma mem-brane. Furthermore, cells were not sensitized to the effect of the MRP-1–independent substrates actinomycin D and fenretinide. At higher concentrations of valinomycin, the additive effect with doxorubicin was lost, which most likely reflects additional cellular effects of valinomycin-induceduncoupling of the mitochondrial electron chain (9, 31) and efflux–independent activity (32, 33).Treatment of ES cells with the HSP90 inhibitors 17-AAGor NVPAUY (20) decreased viable cell number, in agree- ment with previous observations that the HSP90 inhibitor PU-H71 induces death in ES cell lines (34, 35) and de- creases tumor growth in vivo (35). Furthermore, inhibition of HSP90 or the HSP90 cochaperone Sgt1 both decreased levels of the ES-driving fusion protein, EWS-FLI1 (35, 36), suggesting that these chaperones are involved in the sta- bility of such oncoproteins. Interestingly, expression of hTERT and IGF receptor 1 (37), targets often associated with the progression of ES (34), were also decreased fol- lowing HSP90 inhibition (35). Together, these observations suggest that HSP90 proteins have multiple roles in the initi- ation, maintenance, and progression of drug-resistant ES, which requires further investigation.Although treatment with 17-AAG induced greater cell kill, the NVPAUY-induced decrease in expression of the HSP90 client proteins AKT and RAF1 and mitochondrial MRP-1 was most marked, suggesting that 17-AAG may in- duce off-target effects independent of HSP90 inhibition (38). These observations are consistent with reports that NVPAUY is a more potent and specific inhibitor of HSP90 than 17-AAG (20). The decrease in mitochondrial MRP-1 expression following treatment with HSP90 inhibitors paralleled the altered profile of the HSP90 client pro-teins AKT and RAF1 (18, 19, 27, 28, 39–41), supporting the hypothesis that HSP90 is a chaperone for MRP-1 to the mitochondria. Although mitochondrial MRP-1 was de-creased when HSP90b was reduced, total MRP-1 expression was unchanged, confirming that MRP-1 is not protected from degradation by HSP90b, consistent with observations in colon cancer cells (18).We have shown that MRP-1 is a client protein of HSP90b, and its expression in the outer membrane of the mito- chondria has functional efflux activity, which when re- duced following treatment with valinomycin decreases the effect of the MRP-1 substrate doxorubicin. However, direct inhibition or knockdown of HSP90b and the concomitant reduced expression of MRP-1 in the mito- chondria did not sensitize ES cells to doxorubicin. This may reflect the incomplete inhibition or knockdown of HSP90b, consistent with previous attempts that have failed to completely knock down HSP90b (42). Alternatively, off- target effects of HSP90 inhibition and knockdown leading to the degradation of client proteins and modification of several signaling pathways (43, 44) may in part explain the lack of sensitization to MRP-1 substrates. For instance, HSP90 inhibitors, including 17-AAG, have been reported to specifically increase the protein levels of both mitochon- drial and nuclear-encoded genes by a posttranscriptional regulation (45) and induce mitochondrial swelling, a hall- mark of mitochondrial dysfunction (46). HSP90 inhibition has also been reported to induce a change in cell metabo- lism to glycolysis and poor incorporation of TOM40 into the mitochondrial membrane (47). Moreover, we cannot ex- clude the possibility that there is some redundancy and compensatory feedback pathways regulating intracellular MRP-1 expression that enable the cell to withstand chemo- therapeutic insult under such conditions.Like others (20), we have found HSP90 inhibitors to decrease ES viable cell number in vitro as a single agent. The inactivation of multiple cellular pathways (44), stability in vivo (20, 48), and promising efficacy in early-phase clinical trials (49) have made HSP90 an attractive therapeutic target. Because we have shown that mitochondrial MRP-1 is more active than the plasma membrane form and treatment with the mitochondrial-depolarizing agent modifies efflux to sensitize ES cells to MRP-1 substrates, exploiting the organelle-specific chaperone function of HSP90s in combination with inhibitors of MRP-1 for ther- apeutic benefit may be attractive. In particular, the devel- opment of organelle-specific HSP90 inhibitors or strategies to target HSP90 inhibitors to specific organelles may min- imize off-target effects and overcome current challenges of complete inhibition to improve efficacy of MRP-1 inhibi- tors while limiting toxicity (50). Furthermore, although TRAP-1, the mitochondrial HSP90, does not appear to have a role in MRP-1 transport, inhibition of this mito- chondrial chaperone has been reported to induce apo- ptosis and increase sensitivity to chemotherapeutics and overexpression results in sensitization to cisplatin (51, 52). Therefore, we hypothesize that TRAP-1 may also be an exciting target for combination therapy in ES, although this requires further investigation. The colocalization of HSP90a and MRP-1 in the cyto- plasm, and the secretion of extracellular proteins from the cytoplasm by HSP90 (53), suggests that HSP90a may chaperone the excretion of ABC transporter proteins such as MRP-1 from the cancer cell. This hypothesis is consistent with the extracellular secretion of ABC transporter MDR proteins (54, 55) and further supports a role for HSP90 in drug resistance. In agreement with our studies, HSP90a has been described in the cytoplasm of 3T3 cells (56, 57) and increased expression of HSP90 (58), and in particular, the isoform HSP90a (59) has been observed in cancer compared to that in normal tissue. Furthermore, recent studies have shown that HSP90 (60), including HSP90a, regulates the migration and invasion of cancer cells (57, 61) and can transport client proteins to lysosomes for se- cretion (62). Moreover, a variant of HSP90a has been shown to chaperone some proteins toexosomes ina variety of cancer cell lines (61). Cancer cell migration has also been associated with MRP-1 expression in vitro (63), sup- porting the hypothesis that the colocalization of HSP90a and MRP-1 in the cytoplasm may regulate several cellular processes. Previous studies have described MRP-1 in the nucleus (3, 4, 64), andinthis study, we demonstratecoprecipitation and colocalization of HSP70 with MRP-1 in this organelle. This observation suggests that MRP-1 may also be a client protein for HSP70, and supports the hypothesis that mul- tiple proteins of the HSP family may be organelle-specific chaperones for MRP-1. Whether they are also chaperones for other ABC transporter proteins remains to be seen. CONCLUSIONS In summary, we have demonstrated transport of MRP-1 to the outer mitochondrial membrane by the chaperone protein HSP90b, where it may decrease the response of ES cells to MRP-1 substrates. Although the combination of current HSP90b inhibitors with MRP-1 substrates appears unlikely to enhance the activity of MRP-1 substrates in ES, the development of organelle-specific HSP90 inhibitors may prove to be attractive to increase the efficacy of MRP-1 substrates while Apoptozole minimizing toxicity.