Resveratrol Specifically Kills Cancer Cells by a Devastating Increase in the Ca²⁺ Coupling Between the Greatly Tethered Endoplasmic Reticulum and Mitochondria

Resveratrol is a natural phenol in several plants like grapes [1,2,3] and mulberries [4,5]. This compound is known as antioxidant and to exhibit counteracting efficiency against cancer [6], inflammation [7], and aging [8]. Remarkably, resveratrol and its derivative piceatannol [9] specifically cause cell death in cancer cells [10,11,12,13,14]. Different mechanisms including inhibition of respiratory chain complexes [15] and the F1 subunit of ATP synthase [16,17], changes in ROS levels [18,19,20,21], dissipation of mitochondrial membrane potential [13], and Ca²⁺ release from the endoplasmic reticulum (ER) [22] have been discussed as possible mechanisms for cancer cell death that is triggered by increased caspase activity, cytochrome c release and, finally, apoptosis [23].

One hallmark of cancer cells is the very high ATP demand of the ER due to the enormous protein folding activity in this organelle [24,25]. Notably, a cancer-specific functional arrangement between the ER and mitochondria as ATP supplier has been discussed and an improved inter-organelle shuttling of Ca²⁺ from the ER into the mitochondria where it enhances the production of ATP is proposed [26]. Subsequently, increased mitochondrial ATP may be shuttled into the ER [27], where it meets the excessive ATP demand of the ER in a cancer cell [24,25]. The transfer of the ubiquitous second messenger Ca²⁺ into the mitochondrial matrix is an important signaling process that substantially contributes to physiological and pathophysiological pathways. For instance, metabolic activity is strongly influenced by Ca²⁺-dependent enzymes of the tricarboxylic acid (TCA) cycle like pyruvate dehydrogenase, isocitrate dehydrogenase, and alpha-ketoglutarate dehydrogenase [28]. Furthermore, mitochondrial Ca²⁺ import causes changes in mitochondrial membrane potential [29]. Mitochondria also buffer local Ca²⁺ signals and, therefore, regulate Ca²⁺ concentration in cellular microdomains [30]. However, mitochondrial Ca²⁺ accumulation can have a harmful effect, as it may lead to the opening of the mitochondrial permeability transition pore (PTP), which causes the release of cytochrome c facilitating the formation of the apoptosome and the activation of caspases [31]. Thus, in view of the potential risk of mitochondrial Ca²⁺ overload leading to apoptosis and the signaling function of mitochondrial Ca²⁺ uptake to stimulate the organelle, mitochondrial Ca²⁺ uptake needs to be precisely regulated.

The predominant part of Ca²⁺ sequestered by mitochondria is released by inositol 1,4,5-trisphosphate (IP₃) from the internal Ca²⁺ store, the ER [32,33]. ER Ca²⁺ release generates Ca²⁺ hot spots that approache the mitochondrial surface [34]. This Ca²⁺ interplay between mitochondria and ER occurs mostly in specialized regions of the ER called mitochondria-associated ER membranes (MAMs) [35], where ER tubules are closely tethered to mitochondria by linking proteins like Mfn 2 [36,37], Grp75 [38], and PACS2 [39]. Within this highly specialized region, a sophisticated toolkit consisting of the IP₃ receptor, the sarco/endoplasmic reticulum Ca²⁺ ATPase (SERCA), the mitochondrial Ca²⁺ uniplex [40,41], and the mitochondrial Na⁺/Ca²⁺ exchanger (NCLX) [42] controls ER Ca²⁺ release and re-uptake as well as mitochondrial Ca²⁺ sequestration and extrusion, respectively. The mitochondrial Ca²⁺ uniplex consists of the pore forming proteins MCU [43,44], MCUb [45], EMRE [46], MCUR1 [47,48], MICU1 [49], and MICU2 [50]. Importantly, the activity of the mitochondrial Ca²⁺ uniplex is under the control of intermembrane Ca²⁺ that binds to the gatekeeping MICU1 [51] leading to its rearrangement [52] with a K_D of 4.4 µM free Ca²⁺ and, subsequently, the influx of Ca²⁺ into the mitochondrial matrix [53]. Accordingly, the activity of mitochondrial Ca²⁺ uptake from the MAM region correlates with the local Ca²⁺ concentration that is under the control of SERCA [54].

Considering the reported effects of resveratrol and piceatannol on the F1 subunit of ATP synthase, we hypothesize that these polyphenols trigger cancer cell death by turning the cancer cell-specific MAM arrangement into a suicide machinery. According to our hypothesis, the reduced ATP formation may decrease SERCA activity within MAMs leading eventually to excessive mitochondrial Ca²⁺ uptake triggering mitochondrial Ca²⁺ overload and, thus, the initiation of the apoptotic pathway in the cancer cell. The current study was designed to challenge this hypothesis by comparing the effects of resveratrol and piceatannol on cell survival and apoptosis, mitochondrial Ca²⁺ homeostasis, and (sub)cellular ATP dynamics of primary somatic cells with cancer cell lines.

Cell culture materials were obtained from PAA laboratories (Pasching, Austria). Oligomycin A, histamine, 2,5-di-t-butyl-1,4-benzohydroquinone (BHQ), and ethylene glycol tetraacetic acid (EGTA) were purchased from Sigma Aldrich (Vienna, Austria), thapsigargin, piceatannol, and resveratrol from Abcam (London, UK). Prior to experiments, cells were washed and maintained for 20 minutes in a HEPES-buffered solution containing 138 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 1 mM HEPES, 2.6 mM NaHCO₃, 0.44 mM KH₂PO₄, 0.34 mM Na₂HPO₄, 1 mM D-glucose, 0.1% vitamins, 0.2% essential amino acids, and 1% penicillin-streptomycin, the pH of which was adjusted to 7.4 with NaOH or HCl. During the experiments cells were perfused with a Ca²⁺-containing buffer, which consisted of 145 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 10 mM D-glucose and 10 mM HEPES, pH adjusted to 7.4, or with a Ca^2+-free buffer, in which CaCl₂ was replaced by 1 mM EGTA.

HeLa and Ea.hy926 cells were grown in Dulbecco's Modified Eagle Medium (Sigma Aldrich) containing 10% fetal bovine serum, 100 U/ml penicillin, 100 µg/ml streptomycin, and 2 mM glutamine (Gibco/LifeTechnologies, Vienna, Austria). For Ca²⁺ imaging, cells were plated on 30-mm glass coverslips and transiently transfected at 60 - 80% confluence with 1.5 µg plasmid DNA encoding the appropriate sensor or AKAP-CAAX-RFP as well as, if necessary, in combination with 100 µM siRNA using 2.5 µl of TransFast™ transfection reagent (Promega, Madison, WI, USA) in 1 ml of serum- and antibiotic-free medium. Cells were maintained in a humidified incubator (37°C, 5% CO₂, 95% air) for 16 - 20 hours. Afterwards, transfection mix was replaced by culture medium with supplements. All experiments were performed 48 hours after transfection. siRNAs were obtained from Microsynth (Balgach, Switzerland), and their sequences (5′-3′) were as follows: human MCU siRNA-1: 5' GCC AGA GAC AGA CAA UAC U dTdT 3'; human MCU siRNA-2: 5' GGA AAG GGA GCU UAU UGA A dTdT 3'; human UCP2 siRNA: 5' GCA CCG UCA AUG CCU ACA A dTdT 3'; human UCP3 siRNA: 5' GGA ACU UUG CCC AAC AUC A dTdT 3'; human Letm1 siRNA-1: 5' UCC ACA UUU GAG ACU CAG U dTdT 3'; human Letm1 siRNA-2: 5' AUG UUC CAU UUG GCU GCU G dTdT 3'.

Human umbilical vein endothelial cells (HUVECs) were cultured in endothelial cell growth medium (EGM-2) (Lonza, Basel, Switzerland). For imaging experiments HUVECs were seeded on Poly-L-Lysin coated 30-mm glass coverslips in 6-well plates and infected with BacMam 4mtD3cpv virus (Gibco/LifeTechnologies) following the CellLight protocol. Experiments were performed 48 hours after transfection. Prior approval was obtained for human cell and tissue sample collection from the Institutional Review Board of the Medical University of Graz (protocols 19-252 ex 07/08, 18-243 ex 06/07, 21.060 ex 09/10). Umbilical cords were collected after written informed consent by the mothers after full-term pregnancies in accordance with the Declaration of Helsinki.

Total RNA was isolated using the PEQLAB total RNA isolation kit (PEQLAB Biotechnologie GmbH, Erlangen, Germany), and reverse transcription was performed in a thermal cycler (PEQLAB Biotechnologie GmbH) using a cDNA synthesis kit (Applied Biosystems, Foster City, CA, USA). Expression of MCU, UCP2, UCP3, Letm1, and AKAP-RFP-CAAX in HeLa cells was examined by RT-PCR. A QuantiFast SYBR Green RT-PCR kit (Qiagen, Hilden, Germany) was used to perform real time PCR on a LightCycler 480 (Roche Diagnostics, Vienna, Austria), and data were analyzed by the REST Software (Qiagen). Relative expression of specific genes was normalized with GAPDH as a housekeeping gene. Primers for real time PCR were obtained from Invitrogen (Vienna, Austria), and their sequences (5′-3′) were:

Human MCU forward,TCCTGGCAGAATTTGGGAG;

Human MCU reverse, AGAGATAGGCTTGAGTGTGAAC;

Human UCP2 forward, TCCTGAAAGCCAACCTCATG;

Human UCP2 reverse, GGCAGAGTTCATGTATCTCGTC;

human UCP3 forward, AGAAAATACAGCGGGACTATGG;

human UCP3 reverse, CTTGAGGATGTCGTAGGTCAC;

human Letm1 forward, TGTTGCTGTGAAGCTCTTCC;

human Letm1 reverse, TGTTCTTCAAGGCCATCTCC;

mouse AKAP1 forward, CTCCTTGGCTGGTCCTCCTT;

mouse AKAP1 reverse, ACTCCTTGATGACGTCCTCG.

Dynamic changes in [Ca²⁺]_mito, [Ca²⁺]_ER, [ATP]_cyt, and [ATP]_mito were followed in cells expressing the 4mtD3cpv, D1ER, cytAT1.03, and mtAT1.03 (NGFI, ngfi.eu, Graz, Austria), respectively. Culture medium was removed and cells were kept in a HEPES-buffered solution described above. Single cell measurements were performed on a Zeiss AxioVert inverted microscope (Zeiss, Göttingen, Germany) equipped with a polychromator illumination system (VisiChrome; Visitron Systems, Puchheim, Germany) and a thermoelectric-cooled CCD camera (CoolSNAP HQ; Photometrics, Tucson, AZ, USA). Transfected cells were imaged with a 40× oil-immersion objective (Zeiss). Excitation of the fluorophores was at 440 ± 10 nm (440AF21; Omega Optical, Brattleboro, VT, USA), and emission was recorded at 480 and 535 nm using emission filters (480AF30 and 535AF26; Omega Optical) mounted on a Ludl filterwheel. Devices were controlled and data were acquired by VisiView 2.0.3 (Visitron Systems) software and analyzed with GraphPad Prism version 5.0 for Windows (GraphPad Software, La Jolla, CA, USA). Results of FRET measurements are shown as (R_i - Background) + [(R_i - Background) - (R₀ - Background)] (whereas R₀ is the basal ratio) to correct for photobleaching and/or photochromism.

Ea.hy926 and HUVECs were seeded on 24-mm glass coverslips and transfected with D1ER using TransFast™ transfection reagent or BacMaM D1ER virus (Gibco/LifeTechnologies), respectively, according to manufacturer's instructions. After two days, Ea.hy926 and HUVECs were stained for 10 min with 200 nM MitoTracker® Red CMXRos (Invitrogen) and imaged with 488 nm laser excitation (D1ER) and an 561 nm laser with a CFI SR Apochromat TIRF 100×oil (NA1.49) objective mounted on a Nikon-Structured Illumination Microscopy (N-SIM; Nikon, Tokyo, Japan) system equipped with an Andor iXon3 EMCCD camera. 3D-SIM was used in both channels with very short 30 ms exposure time per image. SIM images where reconstructed using Nis-Elements (Nikon). Images were background corrected with an ImageJ-Plugin (Mosaic Suite, background substractor). Further, the ImageJ/Fiji tool coloc2 was used to determine Mander´s coefficients whereby images were thresholded with Costes automatic threshold to determine Mander´s 2 coefficient (overlap of mitochondria with endoplasmic reticulum).

HeLa cells were seeded on 30-mm round coverslips and transfected after two days with mtDsRed/D1ER (1:1) or mtDsRed/D1ER/mAKAP-RFP-CAAX (1:1:3) using TransFast™ transfection reagent. After two days confocal image stacks were acquired with a Zeiss Observer Z.1 inverted microscope equipped with a Yokogawa CSU-X1 Nipkow spinning disk system, a piezoelectric z-axis motorized stage (CRWG3-200; Nippon Thompson Co., Ltd., Tokyo, Japan), and a CoolSNAP HQ2 CCD Camera (Photometrics). Cells expressing D1ER and mtDsRED were excited with 488 nm and 568 nm laser lines (Visitron Systems) with exposure times of 50 - 300 ms using an alpha Plan-Fluar 100x/1.45 Oil M27 (Zeiss). Stacks of 40 - 70 images with 0.15 µm step width were acquired. Stacks were automatically background corrected and blind-deconvolved using Huygens 2.4.1 (Scientific Volume Imaging (SVI), VB Hilversum, The Netherlands). Volume rendering of the stacks were performed with UCSF Chimera 1.10. 3D-Colocalisation analysis was performed with ImageJ (NIH, Bethesda, MD, USA) to determine Mander´s 2 coefficient.

24 h after transfection in 10 cm dishes, cells were seeded in 96-wells plate at a density of 5,000 cells per well. Incubation with piceatannol, resveratrol, oligomycin, and DMSO was started 48 h after transfection and lasted for 36 h. Cell viability was measured using CellTiter-Blue assay (Promega) and apoptotic caspase activity via Caspase-Glo® 3/7 assay (Promega) following the standard protocols.

Data shown represent the mean ± SEM. ‘n' values refer to the number of individual experiments performed. EC₅₀ values are given as mean plus 95% confidential interval in parenthesis. For life cell imaging, numbers indicate the numbers of cells/independent repeats. If applicable analysis of variance (ANOVA) was used for data evaluation and statistical significance of differences between means was estimated by Bonferroni post hoc test or two-tailed Student's t-test assuming unequal variances, where applicable using GraphPad Prism 5.0f (GraphPad Software, La Jolla, CA, USA). The level of significance was defined as P < 0.05.

The effects of resveratrol and its derivate piceatannol on cell survival and apoptosis were compared in somatic short-cultured human umbilical vein endothelial cells (HUVEC) with the endothelial/epithelial cancer cell hybrid EA.hy926. Resveratrol and piceatannol had only a small effect on cell viability and caspase 3/7 activity in somatic HUVEC cells (Fig. 1A). In contrast, a 36 h treatment of the cancerous EA.hy926 cells with resveratrol or piceatannol decreased cell viability by more than 60 % and around 70%, respectively (Fig. 1A). Consistently, the activity of apoptotic caspases 3/7 upon treatment with either resveratrol or piceatannol remained unchanged in HUVEC while was increased by more than 7- and 8-fold in EA.hy926 cells (Fig. 1B).

Beside the endothelial-cancer hybrid cells (EA.hy926), resveratrol and piceatannol significantly decreased viability of the homo sapiens cervix adenocarcinoma cells (HeLa) by 64.5 ± 1.1 (n = 3) and 53.7 ± 1.6% (n = 3), respectively. In line with these findings, caspase 3/7 activity of HeLa cells incubated for 36 h with either 100 µM resveratrol or 100 µM piceatannol was increased app. 2.5-(n = 3) and 2.5-fold (n = 3), respectively.

Since resveratrol and piceatannol were reported to block the F1 subunit activity of mitochondrial ATP-synthase [17,55,56], we next tested whether the polyphenols' effect on cancer cell viability is due to their inhibitory effect on mitochondrial ATP synthase. Therefore, the effect of the ATP synthase inhibitor oligomycin A on cancer cell viability and apoptosis was tested. Similar to resveratrol and piceatannol, oligomycin A (10 µM) reduced viability of EA.hy926 (Fig. 1A) and HeLa cells by 74.6 ± 7.6 (n = 3) and 74.3 ± 4.8% (n = 3), respectively. Likewise, in agreement to previous reports obtained in HepG2 cells [57] as well as in breast-, pancreatic-, and lung-cancer cells [58], oligomycin A enhanced caspase activity in EA.hy926 (Fig. 1B) and HeLa cells (n = 3) by more than 10- and 3.7-fold, respectively.

In line with the other two ATP-synthase inhibitors described above (i.e. resveratrol, piceatannol), oligomycin A had no effect on cell viability (Fig. 1A) and the activity of caspases 3/7 of short-termed cultured HUVECs (Fig. 1B).

Because mitochondrial Ca²⁺ overload is known to represent a hallmark in the initiation of apoptotic caspase activity, we investigated the effect of the polyphenols and that of oligomycin A on mitochondrial Ca²⁺ uptake. After incubation with resveratrol, piceatannol, or oligomycin A mitochondrial Ca²⁺ uptake in response to IP₃-generating agonists was strongly increased in the cancerous cell lines (Fig. 2A, B). In contrast, resveratrol, piceatannol, or oligomycin A had much less or no effect on mitochondrial Ca²⁺ uptake to intracellular Ca²⁺ release in short-term cultured HUVECs (Fig. 2C).

The concentration response correlation for the effect of resveratrol on histamine-triggered mitochondrial Ca²⁺ uptake in EA.hy926 cells revealed an EC₅₀ of 49.5 (29.9-84.8) µM with an Hill slope of 0.86 ± 0.10 and a maximal potentiation of 221.5 ± 5.3% at 300 µM (n = 3) (Fig. 3).

The inter-organelle Ca²⁺ crosstalk between the ER and mitochondria takes place in regions of MAMs [59] and establishes the Ca²⁺ control on mitochondrial ATP production [60,61]. On the other hand, the correct distance between ER and mitochondria is of crucial importance and, if the two organelles are too closely connected, the risk of uncontrolled mitochondrial Ca²⁺ overload yielding initiation of apoptosis occurs [62,63]. In order to seek differences in the organization of mitochondrial Ca²⁺ handling between cancer and non cancer cells as basis of their opposite sensitivity to resveratrol, we investigated the MAM structure of EA.hy926 in comparison to short-term cultured HUVEC cells. Super-high resolution structural illumination microscopy revealed strongly increased ER-mitochondria contact in EA.hy926 compared to HUVECs (Fig. 4A & B). To clarify whether or not the effect of resveratrol and piceatannol on mitochondrial Ca²⁺ uptake in EA.hy926 and HeLa cells is indeed due to the enhanced stability of MAMs, we overexpressed mAKAP-RFP-CAAX construct that artificially tethers mitochondria to the plasma membrane, thus, disrupts the connection between ER and mitochondria(Fig. 4C & D). Destabilization of contact sites between ER and mitochondria by overexpression of mAKAP-RFP-CAAX caused a loss of the effect of resveratrol, piceatannol, and oligomycin A on the mitochondrial Ca²⁺ uptake in response to intracellular Ca²⁺ release in EA.hy926 (Fig. 4E) and HeLa cells (Fig. 4F). These data indicate that the effect of the ATP synthase inhibitors on mitochondrial Ca²⁺ uptake in cancer cells relates to the inter-organelle Ca²⁺ organization between the mitochondria and the ER.

Because resveratrol, piceatannol, and oligomycin A showed similiar results in cell viability and mitochondrial Ca²⁺ uptake experiments, we next compared their effect on mitochondrial and cytosolic/global ATP levels. Measuring cytosolic and mitochondrial ATP by respectively-targeted fluorescence ATP probes [27,64] revealed for all ATP-synthase inhibitors a small reduction in the cytosolic ATP level of app. 20% of the decrease observed by the combined addition of 2-deoxy-D-glucose and oligomycin A (Fig. 5A). In contrast, resveratrol, piceatannol, and oligomycin A strongly affected mitochondrial ATP levels of more than 50% of the decrease obtained by the addition of 2-deoxy-D-glucose and oligomycin A (Fig. 5B). These data indicate that three compounds, in the concentration used in this study (resveratrol 100 µM, piceatannol 100 µM, oligomycin A 10 µM), exhibited similar inhibitory potential and predominantly affected mitochondrial ATP levels while their effect for lowering global ATP was very small.

Our data described above indicate that cancer cells establish an enforced ER-mitochondria coupling and are highly susceptible to resveratrol, piceatannol and oligomycin A that enhance mitochondrial Ca²⁺ sequestration specifically in cancer cells. Considering the importance of SERCA activity for the control of the Ca²⁺ concentration within the MAMs, we speculate that the inhibition of mitochondrial ATP-synthase may affect the activity of SERCA activity within the MAMs and, thus, hamper Ca²⁺ reuptake into the ER yielding increased mitochondrial Ca²⁺ sequestration. Therefore, the impact of resveratrol, piceatannol, and oligomycin A on SERCA activity was assessed by measuring the kinetics of ER Ca²⁺ refilling of previously depleted ER. All ATP-synthase inhibitors reduced ER Ca²⁺ uptake kinetic indicated as slope of ER refilling by about 70% (Fig. 6A&C).

To challenge our hypothesis that resveratrol, piceatannol, and oligomycin A affect SERCA activity by decreasing ATP within the ER-mitochondria junction, MAMs were disrupted by expressing mAKAP-RFP-CAAX construct that fixes mitochondria at the inner plasma membrane [65]. In cells with disrupted MAMs, the ATPase inhibitors failed to affect SERCA activity (Fig. 6B&C), thus, demonstrating that the close connection between ER and mitochondria is prerequisite for the effect of polyphenols to affect SERCA activity.

In HeLa cells, mitochondrial Ca²⁺ uptake is sensitive to a knock-down of MCU and UCP2/3 but not Letm1 (Fig. 7A). An inhibition of SERCA is known to impact mitochondrial Ca²⁺ uptake and to shift mitochondrial Ca²⁺ uptake route to become UCP2/3-independent but Letm1-dependent [66]. Accordingly, we next tested whether the redirection of the mitochondrial Ca²⁺ uptake route also occurs by incubation with resveratrol, piceatannol, and oligomycin A as consequence of their inhibition of SERCA due to the inhibition of ATP-synthase by these compounds. In line with the reported effect of thapsigargin, all three ATP-synthase inhibitors abolished the contribution of UCP2/3 to mitochondrial Ca²⁺ uptake of intracellularely released Ca²⁺ (Fig. 7B). Hence, under condition of an inhibition of the altered mitochondrial Ca²⁺ uptake route by a depletion of Letm1 with respective siRNA resveratrol, piceatannol, and oligomycin A failed to enhance mitochondrial Ca²⁺ sequestration (Fig. 7C) that entirely depended on the core mitochondrial Ca²⁺ uniporter protein MCU under all conditions (Fig. 7D). These data indicate that thapsigargin as well as the ATP-synthase inhibitors resveratrol, piceatannol, and oligomycin A shift the MCU-established mitochondrial Ca²⁺ uptake from UCP2-dependent to an UCP2/3-independent but Letm1-dependent route.

Next we investigated whether or not resveratrol-, piceatannol-, or oligomycin A-triggered cancer cell death is due to increased mitochondrial Ca²⁺ uptake as a consequence of reduced SERCA activity due to the inhibition of mitochondrial ATP-synthase. Therefore, the effects of these ATP-synthase inhibitors on cell viability and the activity of the apoptotic pathway were tested in cancer cells (HeLa) depleted from the above verified two mitochondrial Ca²⁺ transporters involved in mitochondrial Ca²⁺ uptake under conditions of hampered SERCA activity due to ATP depletion, MCU and Letm1. Diminution of mitochondrial Ca²⁺ uptake by transfection with specific siRNAs against MCU and Letm1 rescued HeLa cells from initiation of apoptotic pathways (Fig. 8A) and reduced cell viability (Fig. 8B) in response to treatment with either resveratrol, piceatannol, or oligomycin A.

It is well documented that resveratrol and its derivate piceatannol induce cell death exclusively in cancer cells, including human breast cancer cell line MCF-7 [67,68], human oral cancer cells OC2 [69], human promyelocytic leukemia cells HL-60 [70], human prostate carcinoma cells LNCaP, and HeLa cells [58]. In line with these reports, in the present work the hybrid-cell line EA.hy926 [71] and HeLa [72] were found to be sensitive to resveratrol-and piceatannol-induced cell death as well. In contrast to the endothelial hybrid-cell line EA.hy926, short term cultured endothelial cells were insensitive to resveratrol and piceatannol, thus, supporting the concept that these polyphenols specifically trigger cancer cell death but not that of respective somatic cells. Hence, resveratrol- and piceatannol-induced strong caspase 3/7 activity, indicating the initiation of the apoptotic pathways in cancer cells. The underlying reason(s) of the cancer cell specificity remains elusive. In fact, several mechanisms obviously involved in resveratrol-induced cell death have been solidly described but at the first glance it is not clear why an inhibition of ATP formation [15] [16,17], ROS formation [18,19], dissipation of mitochondrial membrane potential [13], or Ca²⁺ release from the endoplasmic reticulum (ER) [22] should be an option to trigger specifically cancer cell death and not that of somatic cells. However, resveratrol, instead of triggering cell death in somatic cells, counteracts pathological phenomena like cellular senescence [8] and inflammation [7]. Therefore, we hypothesize that the polyphenols' cytotoxic effect on cancer cells depends on cancer cell-specific settings and, thus, intended to reveal such mechanisms within this work.

Mitochondrial Ca²⁺ overload is a hallmark of the initiation process of the apoptotic cell death pathway [73]. Our findings that resveratrol and piceatannol do not exhibit a direct effect on basal mitochondrial Ca²⁺ levels while they strongly augment mitochondrial Ca²⁺ uptake upon intracellular Ca²⁺ release in cancer cells (EA.hy926 and HeLa) but not in the somatic HUVEC cells, indicate that the action of the polyphenols is due to their modulatory effect on a cancer-specific setting in the Ca²⁺ link between the ER and the mitochondria. While the concentrations of the polyphenols used in this study are rather high, the concentration response relationship for mitochondrial Ca²⁺ uptake matches that reported on cancer cell death [10,11,12,13] and plasma levels found upon resveratrol treatment in vivo [74,75].

Notably, a comparative analysis of the density of mitochondria-ER junctions (i.e. MAMs) in cancer and somatic cells using super-resolution microscopy revealed more inter-organelle junctions in cancer cell lines. These findings are in line with the recent report highlighting the importance of the ER-mitochondrial Ca²⁺ transfer in cancer cells [26] to boost mitochondrial ATP production that, in turn, meets the strongly elevated ATP demand of the ER in cancer [24,25].

Our findings that a disruption of this tight mitochondria-ER coupling by fixing mitochondria at the inner side of the plasma membrane [65] prevented resveratrol/piceatannol-induced enhancement of mitochondrial Ca²⁺ uptake upon intracellular Ca²⁺ release, supports our concept that the polyphenols trigger cancer cell death by acting within the cancer-specific enforced mitochondria-ER junctions. Hence, these data indicate that resveratrol and piceatannol do not affect mitochondrial Ca²⁺ uptake per se but may foster inter-organelle Ca²⁺ flux. Notably, the inter-organelle Ca²⁺ transfer between the ER to the mitochondria mainly depends on the amount of intracellularely released Ca²⁺, the mitochondrial Ca²⁺ uptake machinery and the amount of Ca²⁺ re-sequestration from the MAM back into the ER by SERCA [54,66]. While the first two mechanisms appeared to be unaffected by resveratrol or piceatannol, a modulation of SERCA activity by these polyphenols appears likely due to their great inhibitory effect on mitochondrial ATP formation. Hence, our data revealed that resveratrol and piceatannol only marginally affected cytosolic ATP levels, while their impact on mitochondrial ATP content was strong and comparable with the known ATP synthase inhibitor oligomycin A that is known to cause a slight drop in cytosolic [76] and a massive drop in mitochondrial ATP levels [77].

Remarkably, oligomycin A, that mimics the impact of resveratrol and piceatannol on mitochondrial ATP formation, exhibited identical effects on specific cancer cell death and mitochondrial Ca²⁺ uptake, thus, indicating that the decrease in mitochondrial ATP formation is the reason for cancer cell death induced by resveratrol and piceatannol (and oligomycin A) . This conclusion is further supported by our findings that resveratrol, piceatannol and oligomycin A clearly affect SERCA-mediated Ca²⁺ sequestration into the ER in HeLa cells while no effect on ER depletion was found. Importantly, a disruption of the tight mitochondria-ER coupling by fixing mitochondria at the inner side of the plasma membrane abolished the inhibitory effect of the polyphenols and oligomycin A on SERCA-mediated ER Ca²⁺ refilling. These data suggest that the inhibition of mitochondrial ATP formation yields reduced SERCA activity within the cancer cell-specific enforced mitochondria-ER coupling.

Inhibition of SERCA activity was shown to shift mitochondrial Ca²⁺ uptake routes from an UCP2/3-dependent and Letm1-independent towards an UCP2/3-independent and Letm1-dependent pathway [66]. In line with this report, our findings reported herein demonstrate that the ATPase inhibitors yielded loss of UCP2/3-dependency but the engagement of Letm1 in MCU-dependent mitochondrial Ca²⁺ uptake as a consequence of the reduction of SERCA activity due to the lack of local ATP supply. Notably, because SERCA activity counteracts the inter-organelle Ca²⁺ transfer from the ER to the mitochondria its inhibition by the lack of ATP due to the inhibition of ATP synthase by resveratrol, piceatannol, or oligomycin A yields enhanced mitochondrial Ca²⁺ accumulation. Hence, our findings that the siRNA-mediated knock-down of MCU and Letm1 prevented resveratrol-, piceatannol-, and oligomycin A-triggered initiation of the apototic pathway and cancer cell death, support our hypothesis of an enhanced mitochondrial Ca²⁺ sequestration within the enforced ER-mitochondria coupling in cancer cells as molecular mechanisms of resveratrol- and piceatannol-induced cancer cell death.

Our data presented herein demonstrate that the inhibition of ATP synthase by resveratrol, piceatannol, or oligomycin A specifically affect the tight inter-organelle coupling of mitochondria and ER in cancer cells, yielding enhanced mitochondrial Ca²⁺ accumulation, the initiation of apoptotic pathways, and, ultimately, cancer cell death (Fig. 9).

EA.hy926 (hybridoma cell line of primary human umbilical vein cells fused with the adeno-carcinomic human alveolar basal epithelial (human lung cancer) cell A549); EMRE (essential MCU regulator); ER (endoplasmic reticulum); GRP75 (glucose-regulated protein 75); HeLa (homo sapiens cervix adenocarcinoma cells); HUVEC (human umbilical vein endothelial cells); Letm1 (leucine zipper/EF-hand-containing trans-membrane domain 1); MCU (mitochondrial calcium uniporter); Mfn-2 (mitofusin 2); MICU1 (mitochondrial calcium uptake 1); MICU2 (mitochondrial calcium uptake 2); MCUR1 (MCU regulator protein 1); NCLX (mitochondrial Na⁺/Ca²⁺ exchanger); PACS2 (phosphofurin acidic cluster sorting protein 2); SERCA (sarco/endoplasmic reticulum Ca²⁺ ATPase); SIM (structural illumination microcopy); UCP2/3 (uncoupling protein 2 and 3).

We thank Dr. Nicole A. Hofmann, Institut of Experimental and Clinical Pharmacology, Medical University of Graz, Austria and Prof. Dr. Dirk Strunk, Experimental and Clinical Cell Therapy Institute, Paracelsus, Medical University of Salzburg, Austria for kindly providing the short-term cultured HUVECs, Dr. C.J.S. Edgell (University of North Carolina, Chapel Hill, NC) for the EA.hy926 cells, and Prof. Dr. György Hajnózky (Thomas Jefferson University, Philadelphia, Pennsylvania, USA) for kindly providing us the AKAP-RFP-CAAX construct. C.T.M. and C.K. are fellows of the Doctoral College “Metabolic and Cardiovascular Disease” (FWF W1226-B18) and were funded by the Medical University of Graz. BG and this work are funded by the FWF (W 1226-B18, DKplus Metabolic and Cardiovascular Disease) and P 28529-B27. W.P. was a fellow from the Austrian academic exchange service (ÖAD). Microscopic equipment is part of the NIKON-Center of Excellence, Graz that is supported by Nikon, Inc. (Austria, Netherlands, Japan), BioTechMed Graz and the Federal Ministry of Science, Research and Economy in course of the HSRM 2013 initiative.

The authors declare no conflict of interest.