Mitochondrial Hsp90s suppress calcium-mediated stress signals propagating from mitochondria to the ER in cancer cells

To investigate whether mitochondrial Hsp90s modulate mitochondrial calcium stores, we used the mitochondria-targeted Hsp90 inhibitor gamitrinib, a conjugated of triphenylphosphonium (a mitochondria-targeting moiety) and geldanamycin (an Hsp90 inhibitor) [33, 34]. A cytotoxic dose (30 μM) of gamitrinib dramatically increased the intracellular calcium concentration within an hour in human cervical (HeLa), prostate (22Rv1), and breast (MDA-MB-231) cancer cell lines in calcium-free medium (Figure 1A and B). A non-targeted Hsp90 inhibitor, 17-allylamino-17-demethoxygeldanamycin (17AAG), did not increase cytosolic calcium (Additional file 1: Figure S1A), consistent with a previous report that gamitrinib is specific to mitochondrial Hsp90 without affecting cytosolic Hsp90 function [33]. After gamitrinib treatment, PTP opening and loss of mitochondrial membrane potential (ΔΨm) occurred within 30 minutes (Figure 1C, TMRM staining), whereas cytochrome c release, caspase activation, and cell death were not prominent until after 2 hours (Figure 1C, cytochrome c staining; Figure 1D), suggesting that calcium flux concurs with PTP opening, prior to mitochondrial outer membrane permeabilization (MOMP). Consistently, cytosolic calcium elevation was inhibited by cyclosporin A (CsA) (Figure 1E), a potent Cyp-D inhibitor, blocking PTP opening [19]. Thus, mitochondrial Hsp90 inhibition immediately induces PTP opening, loss of ΔΨm, and discharge of the calcium stored in the mitochondrial matrix. Thereafter, a cascade of MOMP, cytochrome c release, and caspase activation ensues (Figure 1F).

The PTP opening has been shown to immediately discharge calcium stored in the mitochondria [36]; however, after mitochondrial Hsp90 inhibition in this study, calcium release continued even after a significant drop in ΔΨm (Figure 1A and C), suggestive of additional sources of calcium flux. We postulated that the primary calcium-storing organelle, the ER, contributes to the cytosolic calcium increase after gamitrinib treatment. To prove this, we directly measured calcium depletion using the calcium sensor protein, Cameleon, targeted to mitochondria and the ER (mtCameleon and D1ER, respectively) [37]. Gamitrinib treatment resulted in FRET signal loss in both mtCameleon- and D1ER-transfected HeLa cells, comparable to that seen with FCCP or Thap treatment (Figure 2A and B), clearly indicating calcium depletion in the ER as well as in mitochondria. Consistent with previous reports [33], gamitrinib has no effect on the ΔΨm of a normal MCF10A breast cell (Additional file 1: Figure S1B and C), and the non-targeted Hsp90 inhibitor 17AAG did not affect the mtCameleon FRET signal (Additional file 1: Figure S1D).

Gamitrinib has been reported to trigger the unfolded protein response in mitochondria, and, through unknown mechanisms, to subsequently activate CHOP, the pro-apoptotic transcription factor often induced during unfolded protein responses in the ER (UPR^ER) [4, 38‐40]. siRNA knockdown of the mitochondrial Hsp90 homolog TRAP1 results in spliced XBP1 mRNA production and eukaryotic translation initiation factor 2α (eIF2α) phosphorylation (Additional file 1: Figure S2A and B), suggesting activation of UPR^ER sensor proteins such as inositol-requiring protein 1α (IRE1α) and PKR-like ER kinase [41, 42]. Consistently, pharmacological inactivation of mitochondrial Hsp90s by gamitrinib also triggered eIF2α phosphorylation and XBP1 mRNA splicing (Figure 2C; Additional file 1: Figure S2C). In addition to UPR^ER sensor protein activation, CHOP induction was clearly seen after both pharmacological and genetic inhibition of mitochondrial chaperones (Figure 2C; Additional file 1: Figure S2D). To investigate the critical involvement of mitochondrial calcium discharge through the PTP for the ER stress response, gamitrinib was administered in the presence or absence of the PTP inhibitor CsA and the calcium chelator BAPTA. Both substances compromised UPR^ER induction, resulting in a dramatic reduction in eIF2α phosphorylation and CHOP expression (Figure 2C).

IP₃Rs and RyRs are ER membrane channels responsible for calcium release from the organelle [26]. Silencing IP₃R1, the major isoform in HeLa cells [43] (Additional file 1: Figure S3A), did not affect the elevation of cytosolic calcium and the induction of CHOP after gamitrinib treatment (Figure 3A and B), but was enough to compromise lysophosphatidic acid-induced ER calcium release in calcium-free medium (Additional file 1: Figure S3B). By contrast, specific RyR inhibitors such as ryanodine (100 μM) and tetracaine (300 μM) [44] strongly inhibited gamitrinib-induced ER calcium release, similar to the PTP inhibitor CsA (Figure 3C). Genetic knockdown of RyR2, the dominant RyR isoform in HeLa cells [45‐47], also blocked gamitrinib-induced cytoplasmic calcium increase (Figure 3D). Consistently, ryanodine and RyR2-specific siRNAs inhibited eIF2α phosphorylation and the subsequent CHOP induction (Figure 3E and F). Collectively, our data suggest that RyR, not IP₃R, is the ER sensor that propagates the signal initiated by discharged calcium from mitochondria in cancer cells.

Impaired mitochondrial function [16] (Figure 4A, TMRM staining) and slightly elevated cytoplasmic calcium (Figure 4A, Fluo-4 staining) were frequently found in gamitrinib-treated cells, even at non-toxic dose of the drug. Therefore, we hypothesized that calcium-mediated stress propagation can render cells sensitive to additional stresses, i.e. lowering the cell death threshold. A representative UPR^ER inducer, Thap, was combined with gamitrinib to test this hypothesis. Gamitrinib sensitized cancer cells to Thap treatment at various concentrations, while the non-targeted Hsp90 inhibitor 17AAG did not (Figure 4B and C). Consistent with pharmacological data, TRAP1 knockdown also sensitized cancer cells to Thap treatment (Additional file 1: Figure S4). The combination of gamitrinib and Thap synergistically induced apoptotic cell death, causing a dramatic increase in caspase activity (Figure 4D).

Proapoptotic CHOP expression induced by combination treatment with gamitrinib and Thap was faster and higher compared to single-agent treatment (Figure 5A). A cell-based reporter assay also showed elevated CHOP transcription activity following combination treatment (Figure 5B). siRNA-mediated knockdown of either CHOP or RyR significantly suppressed this increased cytotoxic activity, but did not affect the toxicity seen with single-agent treatment (Figure 5C and D), suggesting important roles of RyR and CHOP in the drug combination effect. Silencing RyR compromised CHOP induction by the drug combination but not by single-agent treatment (Figure 5E), further confirming that RyR opening is an essential upstream event in the stress response elevating CHOP expression. CHOP-dependent death receptor 5 (DR5) expression [48] has been reported before, but was not involved in the drug combination, considering marginal elevation of DR5 expression and no activation of caspase-8 (Additional file 1: Figure S5A) [49]. Neither did reactive oxygen species (ROS) scavengers affect the increase in cytoplasmic calcium and CHOP induction (Additional file 1: Figure S5B and C). Collectively, our data argue that gamitrinib lowers the cellular threshold against ER stresses by increasing CHOP expression in an RyR-dependent manner.

The mitochondrial Hsp90 pool is dramatically elevated in many cancer cells to cope with various stresses, but expression is very low or undetectable in most normal tissues except brain and testis [7, 8, 10, 51, 52]. To test whether mitochondrial Hsp90-regulated interorganelle calcium signaling is functional in normal cells, we examined primary astrocytes from mouse brain, where Hsp90 expression in mitochondria is higher than in other tissues [7]. Gamitrinib did not affect CHOP induction and eIF2α phosphorylation (Figure 6A), whereas Thap increased CHOP expression in astrocytes (Additional file 1: Figure S6A). Gamitrinib treatment in combination with Thap did not sensitize astrocytes (Figure 6B), possibly due to very low expression of both TRAP1 and Cyp-D in astrocytes compared to cancer cells (Figure 6C). Collectively, gamitrinib does not affect the cell death threshold in astrocytes, probably due to the limited contribution of the chaperones to PTP opening in normal cells; this is in stark contrast with data from cancer cells (Figure 4B-D). Next, the gamitrinib and Thap combination was further examined using a xenograft of relapsed prostate cancer cells (22Rv1) [53], to test whether the cancer cell-specific lowering of the cell death threshold occurs in vivo. Because Thap has been reported to be highly toxic in vivo[54], we administered a very low dose of the drug. Suboptimal individual doses of Thap and gamitrinib did not result in significant inhibition of tumor growth, whereas combined treatment inhibited tumor growth (Figure 6D) without remarkable histological abnormalities and body weight changes (Additional file 1: Figure S6B and C). Individual treatment with either gamitrinib or Thap slightly elevated CHOP expression, whereas combined treatment further elevated CHOP expression synergistically in cancer cells, but not in the brain or liver (Figure 6E; Additional file 1: Figure S6D). Therefore, similar to the in vitro data, mitochondrial Hsp90 inhibition lowers the cell death threshold of cancer cells to Thap treatment in vivo.

HeLa, MDA-MB-231, and NCI-H460 cells were purchased from the Korean Cell Line Bank and 22Rv1 from the American Type Culture Collection. Cell lines were maintained as recommended by supplier. Cells were cultured in DMEM or RPMI medium (Lonza) containing 10% fetal bovine serum (FBS; GIBCO) and 1% penicillin/streptomycin (GIBCO) at 37°C in a 5% CO₂ humidified atmosphere.

Gamitrinib conjugated with triphenylphosphonium was prepared as described previously [33]. MitoTracker, Fura-2-AM, and tetramethylrhodamine methyl ester (TMRM) were purchased from Molecular Probes, Ryanodine was from Santa Cruz Biotechnology. Mn(III) tetrakis(1-methyl-4-pyridyl) porphyrin (MnTMPyP) was from Calbiochem. 1,2-bis(o-aminophenoxy) ethane-N,N,N’,N’-tetraacetic acid acetoxymethyl ester (BAPTA), cyclosporine A (CsA), carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP), tetracaine, and thapsigargin (Thap), and N-acetylcysteine (NAC) and all other chemicals, were from Sigma.

Anti-CEBP homologous protein (CHOP) antibodies were obtained from Cell Signaling; anti-RyR, anti-IP₃R, anti-eIF2α and anti-cytochrome c antibodies from Santa Cruz Biotechnology; anti-cyclopholin D from Calbiochem; anti-eIF2α[pS52] from Invitrogen; anti-β-actin from MP Biomedicals; and anti-TRAP1 from BD Biosciences.

Primary cultures of astrocytes were prepared as previously described [64]. Briefly, the mouse brain cortex, after removing the meninges, was dissected and dissociated with moderate pipetting. Cells were plated on 100-mm dishes coated with 10 μg/ml poly-D-lysine (Sigma) and grown to confluence in DMEM supplemented with 10% FBS, 10% horse serum (GIBCO), 100 units/ml penicillin, and 100 μg/ml streptomycin at 37°C in a 5% CO₂ humidified atmosphere. Afterward, astrocytes were trypsinized and plated on 6-well plates coated with poly-D-lysine to administer drugs.

Small interfering RNAs (siRNA) against TRAP1, RyR2, IP₃R, and CHOP were synthesized by Genolution (Korea) as follows:

RyR2-#1, 5′-AAGTGGTTCTGCAGTGCACCG; RyR2-#2, 5′-AAGTACGAGTTGGAGATGACC; TRAP1-#1, 5′- AAACATGAGTTCCAGGCCGAG; TRAP1-#2, 5′- CCCGGTCCCTGTACTCAGAAA; IP₃R1-#1, 5′-GAGAATTTCCTTGTAGACATCTGCA; IP₃R1-#2, 5′-GGCCTGAGAGTTACGTGGCAGAAAT; IP₃R2, 5′-GAGAAGGCTCGATGCTGAGACTTGA; IP₃R3, 5′-CCGAGATGACAAGAAGAACAAGTTT; CHOP-#1, 5′-AGAACCAGCAGAGGTCACAA; CHOP-#2, 5′-AAGAGAATGAACGGCTCAAGC; control, 5′-ACUCUAUCUGCACGCUGAC. Cells were cultured on 6-well plates at 50–75% confluence, transfected with 20 nM siRNA mixed with G-Fectin (Genolution) for 48 hours, and then analyzed or treated with drugs.

Cells (5 × 10³ cells/well) were cultured in 96-well plates overnight and treated with gamitrinib and Thap alone or in combination for 24 hours. To determine cell viability, cells were exposed to 3 (4,5-dimethyl-thyzoyl-2-yl)2,5 diphenyltetrazolium bromide (MTT), and crystalized formazan was quantified by measuring the absorbance at 595 nm with an Infinity M200 microplate reader (TECAN). Absorbance data were compared with that of vehicle control and expressed as percent viability. Alternatively, after treatment with drugs, DNA content (propidium iodide, red fluorescence) and caspase activation (DEVDase activity, green fluorescence) of the cells were analyzed using the CaspaTag in situ apoptosis detection kit (Millipore). Labeled cells were analyzed using the FACS Calibur™ system (BD Biosciences). Data were processed using FlowJo software (TreeStar).

To generate a CHOP reporter stable cell line, PC 3 cells were co-transfected with 8 μg of a promoter construct (CHOP::GFP) [50] obtained from Addgene (Addgene plasmid 21898) and 800 ng of puromycin linearized selection marker (Clontech) using Lipofectamin (Invitrogen) per manufacturer’s instructions. Transfected PC3 cells were cultured in RPMI (Lonza) with 1 μg/ml puromycin (Clontech) for 3 weeks and colonies were picked using cloning cylinders. GFP expression was monitored in the IncuCyte™ imaging system (Essen Bioscience) at an excitation wavelength of 450–490 nm and an emission of 500–530 nm, and analyzed by Image J software (National Institutes of Health).

HeLa cells were incubated with 5 μM Fura-2-AM for 30 min at 37°C and 5% CO₂. After washing with Hank’s Buffer, the cells were incubated with calcium-free Locke’s solution (154 mM NaCl, 5.6 mM KCl, 3.2 mM MgCl₂, 5 mM HEPES, 10 mM glucose, 0.2 mM EGTA; pH 7.4). Fluorescence changes were monitored every 5 minutes using an IX81 ZDC microscope (Olympus) at an emission wavelength of 510 nm with dual excitation at 340 nm and 380 nm. Images of the 340/380 fluorescence ratio were generated and analyzed by the Xcellence software package (Olympus).

Fluorescence resonance energy transfer (FRET) measurements were performed using an FV1000 laser confocal scanning microsope (Olympus) with a FRET module and a UPLSAPO 100× oil immersion objective with a 1.40 numerical aperture. HeLa cells were seeded on a Lab Tek II slide chamber at 40–80% confluency in DMEM (Lonza) supplemented with 10% FBS and 1% penicillin/streptomycin at 37°C and 5% CO₂. D1ER or mtCameleon constructs (kind gifts from Dr. R.Y. Tsien, University of San Diego) [37] were transfected into HeLa cells using the Lipofectamine transfection reagent (Invitrogen) per manufacturer’s instructions. Cells were imaged at 24 or 48 hours after transfection. All analyses were performed under the same conditions. D1ER and mtCameleon, containing FRET donor (CFP) and acceptor (citrine) components, were excited with a 440-nm diode laser source; the emitted fluorescence bands were separated by a grating and detected by photomultiplier tubes in the CFP channel (480 nm) and FRET channel (535 nm). The FRET ratio (R_FRET) was calculated as described previously [65] from confocal images using FV10-ASW 3.1 software (Olympus) by pixel-by-pixel quantification of fluorescence intensity: R_FRET = I_FRET/I_CFP , where I_FRET and I_CFP represent the fluorescence intensities from the FRET and CFP channels, respectively. The FRET ratio (relative units) was plotted after comparing R_FRET values.

Total RNA was prepared from cells suspended in cold PBS using the RNeasy mini kit (QIAGEN), and cDNA was synthesized using the ProtoScript® First Strand cDNA Synthesis Kit (New England Biolabs) using an oligo(dT) primer. The PCR reaction was performed in a Mastercycler PCR machine (Eppendorf) with the following sets of oligonucleotide primers: glyceraldehyde phosphate dehydrogenase (GAPDH), 5′-CGGGAAGCTTGTCATCAATGG-3′ and 5′-GGCAGTGATGGCATGGACTG-3′; CHOP, 5′- CTTTCTCCTTCGGGACACTG-3′ and 5′-AGCCGTTCATTCTCTTCAGC-3′; TRAP1, 5′- ATGGCGCGCGAGCTGCGG-3′ and 5′-CAGTCGTCCTGCCTGCAA-3′; X-box binding protein 1 (XBP1), 5′-CCTTGTAGTTGAGAACCAGG-3′ and 5′-GGGGCTTGGTATATATGTGG-3′.

All experiments involving animals were approved by UNIST (IACUC-12-003-A). 22Rv1 (7 × 10⁶) cells suspended in sterile PBS (200 μl) were injected subcutaneously into both flanks of 6-week-old BALB/c nu/nu male mice (Japan SLC Inc.) and allowed to grow to an average volume of approximately 100 mm³. Animals were randomly divided into four groups (two tumors/mouse, five mice/group). Gamitrinib or vehicle (DMSO) dissolved in 20% Cremophor EL (Sigma) in PBS was injected intraperitoneally, and Thap dissolved in 0.9% NaCl in PBS intravenously. The mice were administered 10 mg/kg gamitrinib and 0.2 mg/kg Thap twice a week. Tumors were measured daily with a caliper, and tumor volume was calculated using the formula: V = 1/2 × (width)² × length. At the end of experiment, animals were euthanized, and organs including brain, heart, kidney, liver, lung, spleen, and tumor were collected for histology or western blotting. For histological analysis, harvested organs were fixed in 10% formalin and embedded in paraffin. Sections (5 μm) were placed on high-adhesive slides, stained with H&E, and scanned using the Dotslide system (Olympus) with 10× magnification. For western blot analysis, tissue samples were lysed in RIPA buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% NP-40, and 0.25% N-deoxycholate) containing protease inhibitor and phosphatase inhibitor cocktails (Calbiochem) using a homogenizer (IKA).

All MTT experiments were duplicated and repeated independently at least three times. Statistical analyses were performed using the software program Prism 5.0 (GraphPad). In an unpaired t-test, p < 0.05 was considered significant.