Selective vulnerability of cancer cells by inhibition of Ca²⁺ transfer from endoplasmic reticulum to mitochondria

SUMMARY

In the absence of low-level endoplasmic reticulum-to-mitochondrial Ca²⁺ transfer, ATP levels fall and AMPK-dependent, mTOR-independent autophagy is induced as an essential survival mechanism in many cell types. Here we demonstrate that tumorigenic cancer cell lines, transformed primary human fibroblasts and tumors in vivo respond similarly, but autophagy is insufficient for survival, and cancer cells die while their normal counterparts are spared. Cancer cell death is due to compromised bioenergetics that can be rescued with metabolic substrates or nucleotides, and caused by necrosis associated with mitotic catastrophe during their proliferation. Our findings reveal an unexpected dependency on constitutive Ca²⁺ transfer to mitochondria for viability of tumorigenic cells and suggest that mitochondrial Ca²⁺ addiction is a feature of cancer cells.

Graphical abstract

INTRODUCTION

Inositol 1,4,5-trisphosphate receptors (InsP₃Rs) are a ubiquitous family of Ca²⁺ release channels present primarily in the endoplasmic reticulum (ER) (Foskett et al., 2007). Ca²⁺ release through the InsP₃R regulates numerous cell functions including transcription, proliferation, secretion and motility, among others (Cardenas et al., 2005; Foskett et al., 2007). InsP₃R-mediated Ca²⁺ signals also regulate cell metabolism, primarily by supplying released Ca²⁺ to mitochondria where it stimulates production of reducing equivalents by pyruvate dehydrogenase (PDH) and two Ca²⁺ dependent dehydrogenases in the tricarboxylic acid (TCA) cycle, alpha-ketoglutarate dehydrogenase (α-KGDH) and isocitrate dehydrogenase (IDH) (McCormack and Denton, 1979), as well as activities of respiratory chain components to promote oxidative phosphorylation (OXPHOS) and ATP production (Murphy et al., 1990; Territo et al., 2000). Low-level constitutive InsP₃R-mediated Ca²⁺ release is essential for maintaining basal levels of OXPHOS and ATP production in a wide variety of cell types (Cardenas et al., 2010). In the absence of constitutive ER-to-mitochondrial Ca²⁺-transfer, ATP levels fall and AMPK-dependent, mTOR-independent autophagy is induced (Cardenas et al., 2010; Mallilankaraman et al., 2012a; Mallilankaraman et al., 2012b), as an essential survival mechanism (Cardenas et al., 2010). In all cell types examined, inhibition of constitutive mitochondrial Ca²⁺ uptake induced a bioenergetic crisis that resulted in a reprogramming of metabolism reminiscent of that induced by nutrient starvation despite nutrient availability and enhanced nutrient uptake.

A hallmark feature of cancer cells is a re-programming of their metabolism even when nutrients are available (Boroughs and DeBerardinis, 2015; Jones and Thompson, 2009; Jose et al., 2011). All major tumor suppressors and oncogenes have connections with metabolic pathways (Deberardinis et al., 2008; Koppenol et al., 2011; Levine and Puzio-Kuter, 2010; Vander Heiden et al., 2009). Warburg suggested that cancer originates from irreversible injury in mitochondria followed by a compensatory increase of glycolysis (Warburg, 1956), but increasing evidence indicates that mitochondrial function is essential for cancer cells (Koppennol et al., 2011). A continuous supply of metabolic intermediates from the TCA cycle fuels lipid, nucleic acid and protein biosynthesis and provides redox power essential for cancer cell proliferation (Boroughs and DeBerardinis, 2015; Deberardinis et al., 2008). Many tumor cells require OXPHOS to maintain growth (Birsoy et al., 2014; Caro et al., 2012) and for the majority of their ATP production (Jose et al., 2011). Mutations in OXPHOS genes are tumorigenic (Bayley and Devilee, 2010) and mitochondrial inhibitors have antitumor activity (Cheng et al., 2012; Momose et al., 2010; Zhang et al., 2014).

Accordingly, here we asked what role constitutive mitochondrial Ca²⁺ uptake, important in normal cell bioenergetics, plays in cancer cell metabolism and viability. Using tumorigenic breast and prostate cancer cell lines and genetically transformed isogenic primary human fibroblasts, we found that interruption of constitutive ER-to-mitochondrial Ca²⁺ transfer elicited effects similar to those observed in normal cells, including diminished OXPHOS, AMPK activation and induction of autophagy. Whereas autophagy was sufficient for survival of normal cells, it was insufficient in cancer cells, which responded strikingly with massive death while their normal counterparts were spared. Furthermore, inhibition of InsP₃R activity strongly suppressed melanoma tumor growth in mice. Addition of metabolic substrates or nucleotides rescued the lethal effect of inhibiting mitochondrial Ca²⁺ uptake, suggesting that cell death was induced by compromised bioenergetics. Cell death was caused by necrosis associated with mitotic catastrophe at daughter cell separation during ongoing proliferation of the cancer cells. Our findings reveal a fundamental and unexpected dependency InsP₃R-mediated Ca²⁺ transfer to mitochondria for viability of cancer cells.

Results

Inhibition of InsP₃R Activity Creates a Bioenergetic Crisis in Both Tumorigenic and Non-tumorigenic Cell Lines

We examined the effects InsP₃R inhibition on metabolic responses of human breast and prostate cancer-derived tumorigenic cell lines. As controls, we examined non-tumorigenic lines derived from normal tissues. XeB (5 μM, 1h), a specific InsP₃R inhibitor (Jaimovich et al., 2005), reduced basal and maximal oxygen consumption rates (OCR) (Figure 1A), enhanced AMPK phosphorylation (P-AMPK) (Figure 1B) and induced autophagic flux in both non-tumorigenic and tumorigenic breast (Figure 1C and Figure S1A–B) and prostate (Figure S2A and B) cell lines. The effects of XeB on bioenergetic parameters were dose dependent, with 5 μM the minimum concentration required to strongly induce autophagy and significantly reduce basal and maximal OCR in both MCF10A and MCF7 cells (Figure S1C–F). Similar responses to InsP₃R inhibition, including decreased OCR, increased P-AMPK and induction of autophagy were observed in mouse melanoma B16F10 cells (Figure S2F–H) demonstrating that this is a non species-specific general phenomenon. Thus, XeB generates similar bioenergetic crises in tumorigenic and non-tumorigenic cell lines. XeB inhibited constitutive and spontaneous Ca²⁺ transients in both MCF10A and MCF7 cells (Figure 1D, Figure S1G and Supplemental Video 1 (MCF7) and Supplemental Video 2 (MCF10A)). Furthermore, simultaneous genetic knock-down (KD) of types 1 and 3 InsP₃R (~50–70% efficiency; Figure S1H), the main isoforms expressed in these cells (Figure S1I–K) created a similar bioenergetic crisis in normal and tumorigenic cells (Figure 1E and F).

Figure 1. Inhibition of InsP₃R Activity Reduces Oxygen Consumption and Activates AMPK-dependent Autophagy.

(A) Basal and maximal oxygen consumption rate (OCR) of non-tumorigenic and tumorigenic breast cell lines incubated with 5 μM XeB for 1h. N=3, mean ± S.E., *p<0.05, **p<0.01 (t-test).

(B) Representative Western blots of phosphorylated (P-AMPK) and total AMPK in non-tumorigenic and tumorigenic breast cell lines incubated with 5 μM XeB for 1h. Bar graph: P-AMPK/AMPK expressed as average fold increase over basal levels (control cells). N=5, mean ± S.E., **p<0.01, ***p<0.001 (t-test).

(C) Representative Westerns blots of autophagy marker LC3 or tubulin as loading control in non-tumorigenic and tumorigenic breast cell lines treated with 5 μM XeB (1h) and quantification of LC3-II/tubulin expressed as fold increase over basal levels (control cells). N=5, mean ± S.E., **p< 0.01. ***p<0.001 (t-test).

(D) Representative fluorescence recordings of Ca²⁺ release events in unstimulated MCF7 and MCF10A cells treated or not with 5 μM XeB. See also Supplemental Videos 1 and 2.

(E) Representative Western blots of AMPK in breast lines transiently transfected with a siRNA against both types 1 and type 3 InsP₃R (KD) or non-target scrambled (S) siRNA for 24h. Bar graph: P-AMPK/AMPK expressed as average fold increase over basal levels (NT cells). N=3, mean ± S.E, *p<0.05, **p<0.01 (t test).

(F) Representative Western blots of LC3 or tubulin in breast cells transiently transfected with a siRNA against types 1 and 3 InsP₃R (KD) or a non-target (S) siRNA for 24h and quantification of LC3-II/tubulin expressed as average fold increase over basal levels (NT cells). N=3, mean ± S.E, ***p<0.001 (t-test). See also Figure S1.

Inhibition of InsP₃R Activity Is Toxic Specifically for Tumorigenic Cells

Breast cells were treated with XeB for 24 h and cell viability assessed by flow cytometry. 2.5 μM XeB had little effect, except for tumorigenic HS578T cells in which death was enhanced by 8 ± 1% (Figure 2A). In contrast, 5 μM XeB induced substantial death by 43 ± 4% and 53 ± 2% in tumorigenic MCF7 and T47D cell lines, respectively, and over 22 ± 1% in HS578T cells. In contrast, non-tumorigenic MCF10A cells showed only limited sensitivity (5 ± 0.4% cell death). At 10 μM, XeB killed 63 ± 7% and 64 ± 5% of MCF7 and T47D cells, respectively, and 42 ± 6 % of HS578T cells, whereas MCF10A cells had only modest sensitivity (24 ± 1%) (Figure 2A). Similar XeB-mediated killing was observed in human cancer-derived prostate cell lines PC3 and DU145, whereas the non-tumorigenic PNT2 prostate line was little affected (Figure S2C). In agreement with results from flow cytometry, XeB caused striking LDH release in all tumorigenic breast lines compared with MCF10A cells (Figure 2B), as well as in mouse melanoma B16F10 cells (Figure S2I). As assessed by both flow cytometry and LDH release (Figure 2C and D), tumorigenic lines started dying as early as 6 h of exposure 5 μM XeB, while the non-tumorigenic cells remained insensitive. The observed differences in XeB sensitivity could not be accounted for intrinsic differences in proliferation rates between cell lines in untreated conditions (Figure S1L and Figure S2E). Tumorigenic breast (Figure 2E) and prostate (Figure S1D) cells exposed to XeB (5 μM for 24 h) displayed striking morphological changes, including rounding up and shrinkage, whereas similarly treated non-tumorigenic cells maintained normal morphology.

Figure 2. Inhibition of InsP₃R Causes Death of Tumorigenic Breast Cell Lines.

(A) Non-tumorigenic and tumorigenic human breast cell lines were treated with XeB for 24h and cell death determined by TOTO-3 incorporation by flow cytometry. N=3, mean ± S.E, *** p<0.001, NS: not significant (one-way analysis of variance followed by Dunnett’s multiple comparison post-test).

(B) Breast cell lines were treated with 5 μM XeB for 24h and cell death determined by LDH release. N=3, mean ± S.E, *p<0.05, ***p<0.001 (t-test).

(C) Time course of cell death determined by TOTO-3 incorporation by flow cytometry in human breast cell lines treated with 5 μM XeB. N=3, mean ± S.E, * p<0.05, *** p<0.001 (t-test).

(D) Time course of cell death determined by LDH release in breast cell lines treated with 5 μM XeB. N=3, mean ± S.E, *p<0.05, ***p<0.001 (t-test).

(E) DIC images of breast cell lines derived from non-tumorigenic (MCF10A) or tumorigenic (MCF7, T47D, HS578T) human tissue treated with 5 μM XeB for 24h. Bar: 10 μm. N = 3.

(F) Non-tumorigenic and tumorigenic breast cell lines were treated with 2 μM of the PLC inhibitor U73122 or its inactive analog U73343 for 12h and cell death determined by LDH release. N=3, mean ± S.E., ***p<0.001, NS: not significant (t -test).

(G) Cell death (LDH release) in human breast cell lines transiently transfected with a siRNA against type 1 and type 3 InsP₃R (KD) or a scrambled non-target (S) siRNA for 24h. N=3, mean ± S.E., ***p<0.001, NS: non significant (t-test). See also Figure S1.

These results suggest that tumor cell survival is dependent upon InsP₃R signaling. In agreement, inhibition of phospholipase C (PLC), the enzyme responsible for the generation of the InsP₃R ligand InsP₃, with U73122 (2 μM), but not the inactive analog U73343, caused striking LDH release in all tumorigenic breast cell lines, with substantially less effects in the non-tumorigenic line (Figure 2F). Similarly, simultaneous genetic KD of types 1 and 3 InsP₃R caused comparable LDH release as XeB and U73122 in MCF7 and T47D cells, while non-tumorigenic MCF10A cells showed only a small increase (Figure 2G). Thus, tumorigenic cell lines appear to have an essential dependence on InsP₃R-mediated Ca²⁺ signaling for survival, unlike normal cells.

Inhibition of InsP₃R Activity Reduces the Proliferative Potential of Cancer Cell Lines

Colony formation is a hallmark of cancer cells that reflects their ability to proliferate indefinitely (Munshi et al., 2005). XeB (5 μM, 24h) diminished colony formation by tumorigenic breast MCF7 (Figure 3A) and T47D cells and prostate PC3 cells (Figure S3A and S3B). In normal cells, the bioenergetic crisis caused by inhibition of InsP₃R activity could be overcome by providing pyruvate as a mitochondrial substrate, presumably by mass action to drive flux through the system (Cardenas et al., 2010). Methyl-pyruvate (5 μM, 24 h) rescued the XeB-induced MCF7 colony formation defect (Figure 3B). Pyruvate has antioxidant properties (Andrae et al., 1985). However, XeB did not enhance ROS production (Figure S3C) and N-acetylcysteine (NAC) (5 mM, 24 h) did not rescue colony formation (Figure 3B), nor prevent XeB-induced AMPK activation (Figure S3D), autophagy (Figure S3E) or MCF7 cell death (Figure S3F). These results suggest that cancer cell killing and reduction of colony formation by inhibition of InsP₃R activity is a consequence of a bioenergetic crisis that can be rescued by a mitochondrial substrate but not by an antioxidant.

Figure 3. Inhibition of InsP₃R Signaling Impairs Colony Formation in the MCF7 Tumorigenic Breast Cell Line.

(A) Representative plates of MCF7 cells treated with 5 μM XeB or vehicle (top) and quantitative analysis of colony numbers (bottom). N=3, mean ± S.E, ***p<0.001 (t-test).

(B) Representative plates of MCF7 cells treated with 5 μM XeB, 5 mM methyl-pyruvate, 5 μM XeB plus 5 mM methyl-pyruvate or 5 μM XeB plus 5 mM n-acetylcysteine (NAC) (top) and quantitative analysis of colony numbers (bottom). N=3, mean ± S.E., ***p<0.001, **p<0.001, NS: not significant (t-test). See also Figure S2.

XeB Induces Cell Death in Transformed But Not Isogenic Control Human Fibroblasts

Although the cell lines are well-validated and broadly used models, the fact that the control lines are also transformed complicates the results. We therefore turned to an isogenic primary human fibroblast model. Primary human fibroblasts were transformed by incorporation of oncogenic HRas^G12V while bypassing retinoblastoma (Rb)-mediated G1 cell cycle restraint by enforced expression of cyclin-dependent kinase-4 (Cdk4) (Ridky et al., 2010). This model (HrasG12V-CDK4) has classic cancer features, including invasive properties. HrasG12V-CDK4 fibroblasts showed elevated glucose consumption (Figure S3G) and lactate release (Figure S3H), consistent with acquisition of a Warburg-type metabolism. Nevertheless, control and transformed cells had comparable OCR, suggesting similar OXPHOS (Figure 4A). Furthermore, OCR was reduced by XeB to a similar extent in HrasG12V-CDK4 and isogenic control cells (Figure 4A), and a bioenergetic crisis ensued, as evidenced by increased P-AMPK and induction of autophagy (Figure 4B and 4C). These results suggest that mitochondria remain important for maintenance of cell bioenergetics in these primary transformed cells.

Figure 4. Pharmacological and Genetic Inhibition of InsP₃R Causes Death of Primary HRasG12V-Cdk4 Transformed Human Fibroblasts.

(A) Upper: representative traces of OCR in normal (black) or transformed (red) fibroblasts treated (●,▼) or not (■, ▲) with 5 μM XeB for 1h. a: injection of 1 μM oligomycin; b: 300 nM FCCP; c: 100 nM rotenone. Bottom, summary of basal and maximal OCR in fibroblasts treated or not with 5 μM XeB for 1h. N=3, mean ± S.E., ***p< 0.001.

(B) Representative Western blots of AMPK in normal (N) and transformed (T) fibroblasts treated with 5 μM XeB (1h). Bar graph: P-AMPK/AMPK expressed as average fold increase over basal levels (control cells). N=6, mean ± S.E., ***p<0.001 (t-test).

(C) Representative Western blots of LC3 and tubulin in normal (N) and transformed (T) fibroblasts treated with 5 μM XeB for 1h. Bar graph: LC3-II/tubulin expressed as average fold increase over basal levels (control cells). N=6, mean ± S.E., ***p<0.001 (t-test).

(D) Determination of cell death by LDH release in normal and transformed fibroblasts treated with 2.5 or 5 μM XeB (24h). N=3, mean ± S.E., ***p<0.001, NS: not significant (t-test).

(E) Representative Western blots AMPK in normal (N) and transformed (T) fibroblasts transiently transfected with a siRNA against type 1 and type 3 InsP₃R for 24h or siRNA against MCU for 72h. Bar graph: P-AMPK/AMPK expressed as the average fold increase over basal levels (NT cells). N=3, mean ± S.E., ***p<0.001 (t-test).

(F) Representative Western blots of LC3 and tubulin in normal (N) and transformed (T) fibroblasts transiently transfected with a siRNA against types 1 and 3 InsP₃R for 24h or siRNA against MCU for 72h, or scrambled sRNAs (S); LC3-II/tubulin expressed as average fold increase over basal levels (S cells). N=3, mean ± S.E., ***p<0.001 (t-test).

(G) Determination of cell death by LDH release in normal and transformed fibroblasts transiently transfected with a siRNA against types 1 and 3 InsP₃R for 24h or siRNA against MCU for 72h. N=3, mean ± S.E., *p<0.05, **p<0.01, ***p<0.001 (t-test).

(H) Determination of cell death by LDH release in normal and transformed fibroblasts treated simultaneously with 5 μM XeB and either 5 mM methyl-pyruvate (Pyr) or 5 mM dimethyl alpha-ketoglutarate (Keto) for 24h. N=3, mean ± S.E., *p<0.05, **p<0.01, ***p<0.001, NS: non significant (t-test). See also Figure S4.

As in the cancer cell lines, XeB induced profound morphological changes and reduced the number of HrasG12V-CDK4 cells per field, whereas isogenic controls behaved similarly to vehicle-treated cells (Figure S3I). LDH release was enhanced 6-fold in HrasG12V-CDK4 cells compared with isogenic controls (Figure 4D). The sensitivity of HrasG12V-CDK4 cells compared with isogenic cells could not be accounted for by differences in their normal rates of proliferation (Figure S3J). Simultaneous KD of types 1 and 3 InsP₃R, the main isoforms in these cells (Figure S4A–D,) enhanced P-AMPK and autophagy in both HrasG12V-CDK4 and isogenic control fibroblasts (Figure 4E and 4F), but it caused significant death specifically in the HrasG12V-CDK4 cells (Figure 4G). Thus, in primary human cells, as in breast and prostate cell lines, inhibition of InsP₃R activity is strongly toxic specifically for cancer cells.

Killing of Cancer Cells by Inhibition of InsP₃R Activity is Mediated by Altered Mitochondrial Function

In normal cells, inhibition of mitochondrial Ca²⁺ uptake phenocopies effects of InsP₃R inhibition on cell bioenergetics, suggesting that the primary role of InsP₃R activity in preserving normal cellular bioenergetics is to provide Ca²⁺ to mitochondria (Mallilankaraman et al., 2012a). The similar bioenergetic responses to inhibition of InsP₃R activity of tumor cell lines and HrasG12V-CDK4 cells suggested that cancer cells may have a similar dependence on constitutive mitochondrial Ca²⁺ uptake. To test this, we knocked down in HrasG12V-CDK4 and isogenic control cells the expression of MCU, the pore forming subunit of the mitochondrial uniporter Ca²⁺ channel, the major Ca²⁺ uptake pathway in mitochondria (Foskett and Philipson, 2015). MCU expression was similar in human normal versus transformed fibroblasts or between the MCF10A and MCF7 cells (Figure S4E and 4F). MCU knockdown (Figure S4E) increased P-AMPK and induced autophagy (Figure 4E and 4F), as in normal cells (Mallilankaraman et al., 2012a; Mallilankaraman et al., 2012b). Importantly, MCU KD enhanced killing preferentially of the HrasG12V-CDK4 transformed cells (Figure 4G). In contrast, overexpression of MCU was unable to protect MCF7 cells from death induced by InsP₃R inhibition with XeB (Figure S4G and S4H). These results suggest that InsP₃R-mediated Ca²⁺ release and its transfer to mitochondria through MCU is essential to maintain cancer cell viability.

The TCA cycle is regulated by product inhibition, substrate availability and cofactors including Ca²⁺. PDH, ICD and α-KGDH activities are all critically regulated by mitochondrial matrix Ca²⁺ (Glancy and Balaban, 2012). Inhibition of InsP₃R-mediated Ca²⁺ release reduces PDH activity, resulting in decreased OCR and ATP production (Cardenas et al., 2010). Cancer cells rely on mitochondria for production TCA intermediates needed for high growth and proliferation (Boroughs and DeBerardinis, 2015). We speculated that inhibition of mitochondrial function by absence of sufficient matrix Ca²⁺ cannot be compensated for specifically in cancer cells because of this dependence on the TCA cycle. Addition of dimethyl-α-ketoglutarate or methyl-pyruvate, two substrates that energize mitochondria (Cardenas et al., 2010; Wise and Thompson, 2010) to XeB treated (5 μM, 24 h) fibroblasts strongly reduced cell death (Figure 4H). Protection was unrelated to antioxidant properties of the substrates since NAC (5 mM) was without effect on XeB-induced P-AMPK (Figure S4I and 4K), autophagy (Figure S4J and 4L) or killing (Figure S4M).

Although enhanced P-AMPK indicates that inhibition of InsP₃R-mediated Ca²⁺ transfer to mitochondria lowers cellular ATP (Cardenas et al., 2010), glycolysis is believed to provide sufficient ATP for cancer cell functions (Deberardinis et al., 2008). However, as highly proliferative cells, cancer cells rely on the TCA cycle for production of metabolic intermediates for generation of fatty acids and nucleotides. At least three mitochondrial pathways are involved in pyrimidine and purine nucleotides synthesis, and the rate-limiting step in de novo synthesis of pyrimidines occurs in mitochondria and is linked to respiratory chain activity (Desler et al., 2010). To determine if mitochondrial hypo-function in response to InsP₃R inhibition compromises nucleotide metabolism that may contribute to cancer cell death, HrasG12V-CDK4 fibroblasts were treated with 5 μM XeB for 24 h in media supplemented with nucleosides. Notably, nucleoside supplementation was strongly protective of XeB-treated HrasG12V-CDK4 cells (Figure 5A) despite enhanced P-AMPK phosphorylation and autophagy activation (Figure 5B – 5E). Thus, with nucleoside supplementation, cancer cells behave as normal cells, surviving in the face of an unfavorable bioenergetic state.

Figure 5. Nucleoside Rescue of Cancer Cell Killing and Cell Cycle Defects by Inhibition of InsP₃R Activity.

(A) Normal and transformed fibroblasts were treated with 5 μM XeB in absence or presence of nucleosides (NUC) for 24h and cell death determined by LDH release. N=3, mean ± S.E., **p<0.01 (t-test).

(B) Representative Western blots of AMPK in normal fibroblasts treated with 5 μM XeB for 1h in presence or absence of NUC. Bar graph: P-AMPK/AMPK expressed as average fold increase over basal levels (control cells). N=3, mean ± S.E., ***p<0.001, NS: not significant (t-test).

(C) As in (B), using transformed fibroblasts. N=3, mean ± S.E., ***p<0.001, NS: not significant (t-test).

(D) Representative Western blots of LC3 and tubulin in normal fibroblasts treated with 5 μM XeB for 1h in presence or absence of NUC. Bar graph: LC3-II/tubulin expressed as average fold increase over basal levels (control cells). N=3, mean ± S.E., ***p<0.001, NS: not significant (t-test).

(E) As in (D), using transformed fibroblasts. N=3, mean ± S.E., ***p<0.001 (t- test).

(F) BrdU incorporation in normal fibroblasts treated with 5 μM XeB for 12h. Top: BrdU immunofluorescence. Bottom: BrdU labeling expressed as percent of control. N=3, mean ± S.E., ***p<0.001 (t-test).

(G) As in (F), using transformed fibroblasts. N=3, mean ± S.E., *p<0.05 (t-test).

(H) Cell cycle profile determined by flow cytometry in MCF10A cells treated or not with 5 μM XeB for 24h, showing percent of cells in each phase. N=3. **p<0.001, NS: not significant.

(I) Cell cycle profile in MCF7 cells treated or not with 5 μM XeB for 24h, showing percentage of cells in each phase. N=3. *p<0.05, NS: not significant.

(J) As in (I), using HeLa cells. N=3. NS: not significant.

(K) Cell death determined by propidium iodide incorporation by flow cytometry of MCF7 cells treated or not with 5 μM XeB for 24h after a 48h treatment with 400 μM hydroxyurea (HU). **p<0.001 (t-test).

(L) DIC images of HeLa cells loaded with 8 nM TMRE. XeB (5 μM) added at time 0h0′; images collected at indicated times. Arrow shows dividing cell. Bar: 10 μm. See also Supplemental Video 3.

Nucleosides contribute to purine and pyrimidine generation necessary for DNA synthesis. In normal fibroblasts, XeB (5 μM, 10 h) reduced bromodeoxyuridine (BrdU) incorporation by >50% (Figure 5F), suggesting that the unfavorable bioenergetic state in the absence of Ca²⁺ transfer to mitochondria slowed cell proliferation, as expected. In marked contrast, XeB had minimal effects on BrdU incorporation into HrasG12V-CDK4 cells (Figure 5G). These results suggest that normal cells restrain entry into S-phase when conditions are unsuitable, but transformed cells do not recognize this checkpoint. Because of the enforced expression of Cdk4 in HrasG12V-CDK4 cells, it remained unclear whether other cancer cells would progress similarly into mitosis during a XeB-induced compromised bioenergetic state. XeB (5 μM, 24 h) increased the population of MCF10A cells in G1 that correlated with a decrease proceeding through S-phase (Figure 5H), as observed in normal human primary fibroblasts. In contrast, MCF7 cells showed no increase of the G1 population and only a slight decrease in cells in S-phase (Figure 5I). This suggests that MCF7 cells, like HrasG12V-CDK4 cells but unlike normal cells, proceed through the cell cycle despite their compromised bioenergetic state. Similar behavior was observed in HeLa cells, a common cervical tumor cell line (Figure 5J).

HeLa cells were optically imaged to observe progression through the cell cycle. Remarkably, XeB-treated cells entered mitosis but then collapsed catastrophically and died during daughter cell separation (n = 30 mitoses) (Figure 5L and Supplemental Video 3), a phenotype never observed in vehicle-treated cells (n = 50 mitoses). This suggests that in cells with InsP₃R activity inhibited or mitochondrial Ca²⁺ uptake otherwise blocked, cancer cells progress through the cell cycle despite their bioenergetic crisis, in contrast to normal cells that stop in G1, and their progression into mitosis appears to be a step that results in cell death. To probe this further, MCF7 cells were incubated with 400 μM hydroxyurea (HU) to arrest them in S phase of the cell cycle (Figure S5A), and then treated with 5 μM XeB for 24 h. As shown in Figure 5K, prevention of cell cycle progression strongly decreased cell death induced by XeB.

Cell Death Induced by Inhibition of InsP₃R Ca²⁺ Signaling is Mainly Necrotic

Our results reveal that inhibition of InsP₃R-mediated Ca²⁺ transfer to mitochondria causes death specifically in tumorigenic cells, but how they die is unknown. Autophagy inhibition with 3-methyladenine (3MA) (Figure S5B and S5C) did not prevent cell death in XeB-treated MCF7 and MCF10A cells (Figure 6A) and normal and HrasG12V-CDK4 fibroblasts (Figure 6B), suggesting that autophagy is not driving cell death. Although 3MA tended to increase cell death in XeB-treated MCF10A cells and normal fibroblasts, suggesting that autophagy plays a cell survival role, as previously suggested (Cardenas et al., 2010), knockdown of ATG7 (Figure S5D and 5E) did not influence the extent of cell death (Figure S5F). Inhibition of apoptosis with the pan-caspase inhibitor Z-VAD-FMK did not protect tumorigenic cells from XeB-induced death (Figure 6A and 6B). Furthermore, western blot and immunofluorescence analyses of activated cleaved caspase 3 in human normal and HrasG12V-CDK4 fibroblasts (Figure S5G) and in tumor samples (Figure S5H) similarly suggested that apoptosis is not a major feature associated with cell killing. Flow cytometry analyses using FITC-AnnexinV and propidium iodide (PI) labeling further confirmed the presence of only a small apoptotic cell population after XeB treatment (Figure S5I). Inhibition of necroptosis with necrostatin also did not prevent XeB-induced cell death (Figure 6A and 6B). The transcription factor HMGB1 is normally localized in the nucleus but is released into the cytoplasm during early stages of necrosis (Scaffidi et al., 2002). Most (>50%) MCF7 cells exposed to 5 μM XeB for 24 h lost nuclear labeling, while untreated cells and treated MCF10A cells overwhelmingly maintained nuclear HMGB1 label (Figure 6C and 6D). Together with the imaging data, these results suggest that in response to inhibition of Ca²⁺ transfer to mitochondria, cancer cells die by necrosis, at approximately the time of daughter cell separation in mitosis.

Figure 6. Inhibition of InsP₃R Activity Kills Cancer Cells by Necrosis but Not by Autophagy, Apoptosis or Necroptosis.

(A) Non-tumorigenic (MCF10A) and tumorigenic (MCF7) breast cell lines were treated with 5 μM XeB for 24h in presence or absence of the apoptosis inhibitor Z-VAD-FMK (20 μM), autophagy inhibitor 3-MA (10 mM) or necroptosis inhibitor nec-1 (20 μM) and cell death determined by LDH release. N=3, mean ± S.E., NS: not significant (t-test).

(B) As in (A) for normal and transformed primary human fibroblasts. N=3, mean ± S.E., NS: not significant (t-test).

(C) HMGB1 immunofluorescence in MCF7 and MCF10A cells treated with 5 μM XeB or vehicle for 24h. Note that most nuclear label is absent in treated MCF7 cells. Bar: 10 μm.

(D) Quantification of loss of nuclear HMGB1 expressed as percent increase over control. N=3, mean ± S.E., ***p<0.001 (t-test).

Inhibition of InsP₃R Activity Impairs Tumor Growth in Nude Mice

To evaluate whether InsP₃R inhibition affected tumor growth in vivo, subcutaneous tumors were generated using B16F10 melanoma cells. As described earlier, treatment of B16F10 cells in vitro with 5 μM XeB induced AMPK phosphorylation and activated autophagy (Figure S2F and S2G). OCR was inhibited (Figure S2H) and LDH release was increased in a dose-dependent manner by a 24 h exposure to XeB (Figure S2I). Thus, this cell line is appropriate to examine in vivo. B16F10 cells were injected subcutaneously into ventral flanks of 5–6 wk old nu/nu mice. When tumors were palpable, usually by day 5 post-injection, one tumoral flank was injected with 100 μl of a solution of 100 μM XeB and the other with vehicle (PBS + ethanol). One day later, mice were euthanized and tumors were extracted and weighed (Figure S6A, treatment 1). XeB reduced tumor size by nearly 60% compared with untreated tumors (Figure 7A). In a different paradigm, one tumoral flank was treated every other day for a week with 100 μl of a solution of 100 μM XeB (Figure S6A, treatment 2). XeB reduced tumor size by ~70% compared with untreated tumors (Figure 7B). Markers of bioenergetic stress normally observed after InsP₃R inhibition, including enhanced AMPK phosphorylation and autophagy, were observed in the XeB-injected tumor sections (Figure S6B and S6C). The vast majority of XeB-treated tumor cells lost HMGB1 nuclear localization, whereas control tumor cells had HMGB1 largely in the nucleus (Figure 7C). Ki67, a marker of cell proliferation, labeled nuclei equivalently in both XeB- and vehicle-treated tumors (Figure 7D), consistent with the idea that tumor cells remain committed to enter the cell cycle even when in a state of compromised bioenergetics caused by inhibition of InsP₃R activity.

Figure 7. Inhibition of InsP₃R Retards Tumor Growth In Vivo.

Tumors were generated by subcutaneous injection of murine melanoma-derived B16F10 cells into nude mice.

(A) Tumor weight in mice with tumors injected with 100 μl of a 100 μM XeB solution 1d before euthanization. N=9, mean ± S.E., *p<0.05 (t-test).

(B) Tumor weight in mice with tumors injected with 100 μl of a 100 μM XeB solution or vehicle every other day for 6d before euthanization. N=9, mean ± S.E., *p<0.05 (t-test).

(C) HMGB1 immunofluorescence in tumors injected or not with 100 μl of a 100 μM XeB solution 1d before euthanization. Control tumor shows strong nuclear labeling (arrows) while in XeB-treated tumor the nuclei appear empty (arrows) with remaining labeling cytosolic or extracellular. N = 3. Bar: 100 μm.

(D) Ki67 immunofluorescence in tumors injected or not with 100 μl of a 100 μM XeB solution 1d before euthanization. Left: panoramic view of control and XeB-treated tumors. Bar: 100 μm. Right: detail of control and XeB-treated tumors. Bar: 10 μm. N = 3. See also Figure S6.

DISCUSSION

The main result of the present study is the demonstration of a unique requirement of tumorigenic cancer cells for constitutive ER-to-mitochondria Ca²⁺ transfer for their survival. In the absence of this transfer, either as a consequence of blocking InsP₃R-mediated Ca²⁺ release or blocking mitochondrial Ca²⁺ uptake, all tumor cells examined experienced a bioenergetic crisis similar to normal cells, but activation of autophagy, which enabled normal cells to survive, was insufficient, and the tumor cells died by necrosis as they proceeded through mitosis. Specific cancer cell killing was observed in human tumorigenic cell lines derived from breast, prostate, cervix (HeLa), as well as, in data not shown, bone (143B), colon (SW480 and Colo320), kidney (RCCC), skin (WM3918, A231 and WM239a, B16F10) and brain (T98G) cancers. In addition, the same effect was observed in isogenic primary human fibroblasts transformed by expression of oncogenic Ras and enforced expression of Cdk4, a relevant combination found in spontaneous human cancers (Ridky et al., 2010). Furthermore, inhibition of InsP₃R activity had strong anti-tumor efficacy in vivo.

Increased InsP₃R expression and/or activity has been associated with cancer cell proliferation, growth, survival and migration (Davis et al., 2013; Kang et al., 2010; Messai et al., 2014; Shibao et al., 2010; Szatkowski et al., 2010). Furthermore, MCU is highly expressed in ER-negative and basal-like breast cancers (Sorlie et al., 2006). Our results suggest that the fatal fate of cancer cells after InsP₃R inhibition is a result of mitochondrial dysfunction, but it is independent of apoptotic mechanisms but rather appears linked to metabolic defects. Constitutive InsP₃R-mediated Ca²⁺ release is essential to support activity of PDH, a key enzyme the controls flux of carbon into the TCA cycle, through Ca²⁺ activated PDH phosphatase activity. Consequently, diminished TCA production of NADH leads to lower rates of OXPHOS and ATP levels decline (Cardenas et al., 2010). It is now recognized that most cancer cells rely on mitochondrial metabolism and use a significant fraction of glucose-derived pyruvate for ATP generation and for anaplerotic reactions to replenish TCA cycle precursors for macromolecular biosynthesis (Ahn and Metallo, 2015; Jose et al., 2011; Koppenol et al., 2011; Moreno-Sanchez et al., 2007; Porporato et al., 2014). OXPHOS is important for ATP production in many cancers (see (Vander Heiden, 2013)). Pancreatic cancer (Viale et al., 2014) and leukaemic (Lagadinou et al., 2013; Nakada et al., 2010; Shi et al., 2012; Skrtic et al., 2011; Yang et al., 2011) stem cells, diffuse large B cell tumors (Caro et al., 2012) and primary glioblastoma sphere cultures (Janiszewska et al., 2012) rely on OXPHOS for survival. Primary human glioblastomas oxidize glucose via PDH and the TCA cycle during aggressive tumor growth (Marin-Valencia et al., 2012). Inhibition of complex 1 inhibits respiration and human cancer cell growth (Wheaton et al., 2014). The TCA cycle is also critically important in cancer cells for generation of biochemical intermediates to sustain high rates of proliferation (Deberardinis et al., 2008). Most cancer cells derive lipogenic acetyl-CoA from pyruvate through Ca²⁺-dependent PDH (Hatzivassiliou et al., 2005; Metallo et al., 2012), and activation of PDH can promote cancer cell senescence (Kaplon et al., 2013). Some cancers have defective OXPHOS but nevertheless rely on the TCA cycle (Mullen et al., 2014; Mullen et al., 2012). Even in cancer cells with mutations in the TCA cycle, turnover of TCA cycle intermediates is maintained (Mullen et al., 2012). Most cancer cells use PDH as well as glutamine to fuel the TCA cycle (DeBerardinis et al., 2007), where it enters and then cycles through α-KGDH as the first step (Mullen et al., 2014). Importantly, α-KGDH activity is strongly Ca²⁺ dependent (Armstrong et al., 2014), with apparent affinity similar to that of PDH phosphatase, suggesting that it may also be reduced by interruption of ER-to-mitochondria Ca²⁺ transfer. Reductive carboxylation of α-KG is observed in many cancer cells with impaired OXPHOS where it maintains levels of biosynthetic intermediates. Notably, reductive carboxylation of α-KG requires its oxidation by Ca²⁺-dependent α-KGDH (Mullen et al., 2014). Furthermore, reduction of α-KG to isocitrate is mediated by ICD, the other Ca²⁺ dependent enzyme in the TCA cycle. It is of interest that nucleotide supplementation prevented cancer cell death. α-ketoglutarate is a precursor for glutamine and aspartate necessary for pyrimidine synthesis that requires mitochondrial dihydrooroate, which in turns requires flux through the electron transport chain (Grassian et al., 2014; Mullen et al., 2014). Together, these data suggest that cancer cells, by virtue of their reliance on OXPHOS and the TCA cycle for their viability, have a fundamental reliance on ER-to-mitochondria Ca²⁺ transfer via activation of Ca²⁺ dependent dehydrogenases that fuel key steps in the TCA cycle and OXPHOS.

Normal cells also rely on OXPHOS and the TCA cycle but unlike the cancer cells examined, interruption of constitutive ER-to-mitochondria Ca²⁺ transfer had little effect on their viability, in agreement with previous observations (Cardenas et al., 2010). Normal cells activate autophagy as a pro-survival mechanism under these conditions. The tumor cells also activated autophagy, and its inhibition tended to accelerate their cell death. What then distinguishes tumorigenic cells from non-tumorigenic cells that leads to mitochondrial Ca²⁺ addiction specifically in cancer cells? Inhibition of mitochondrial Ca²⁺ uptake in non-tumorigenic cells slowed proliferation, whereas it didn’t in the tumor cell lines. InsP₃-mediated Ca²⁺ signaling is important in cell cycle progression (Ciapa et al., 1994; Han et al., 1992; Miller et al., 1993; Resende et al., 2010). Conversely, its inhibition activates AMPK (Cardenas et al., 2010) and present study) which can phosphorylate and activate p53 to induce p21^waf1/cip1-mediated cell cycle arrest (Jones et al., 2005). In contrast, proliferation of the tumor cells examined here was not slowed by inhibition of InsP₃-mediated Ca²⁺ signaling. MCF7 and HeLa cells have mutations in Cdk9 and Cdkn2 (http://lincs.hms.harvard.edu/) that may enable them to by-pass Ca²⁺ regulated checkpoints. The human fibroblasts we used were transformed by introducing a constitutively active Cdk4 that phosphorylates Rb independent of Ca²⁺ signaling. Whereas several of the cancer cell lines have mutated p53 that could explain lack of cell cycle arrest in response to AMPK activation by InsP₃R inhibition, we examined a broad collection of cell lines with normal and mutated p53 (http://p53.free.fr/index.html) and they all displayed uninhibited cell cycle progression. Indeed, MCF7 and HeLa cells and transformed human fibroblasts that were used for most of our studies express wild-type p53. Although AMPK can suppress the mTOR pathway which plays a role in coordinating cell growth and the cell cycle (Cuyas et al., 2014), AMPK activation induced by InsP₃R inhibition does not affect mTOR activity (Cardenas et al., 2010). Thus, it remains unclear why proliferation was not reduced in tumorigenic cells in response to inhibition of InsP₃R activity. Importantly, progression into mitosis resulted in catastrophic cell death during cytokinesis of all the tumor cells examined. Necrosis as the mechanism of cell death was confirmed by biochemical and histological markers in the cell lines, primary cells and solid tumors. In contrast, neither apoptosis, autophagy nor necroptosis appeared to play significant roles. The mechanisms that lead to mitotic catastrophic cell death (Vitale et al., 2011) around the time of cytokinesis remain to be determined. Interactions of the cyclin-dependent kinase substrate CP110 with Ca²⁺ binding proteins calmodulin or centrin is essential for correct cytokinesis. Whereas presence of a CP110 mutant unable to bind CaM induces cytokinesis failure and multinucleated cells (Tsang et al., 2006), it is possible that cytokinesis failure in the face of a bioenergetic crisis results instead in necrotic mitotic catastrophe. The failure of the tumorigenic cells to slow cell proliferation despite the fact that they were in an energetically-compromised state appears to be the major mechanism that distinguishes them from non-tumorigenic cells and accounts for their susceptibility to cell killing.

In conclusion, using a variety of different tumorigenic cell lines and isogenic primary transformed human fibroblasts, we have demonstrated that constitutive transfer of InsP₃R-mediated Ca²⁺ transfer to mitochondria is essential to maintain the viability of cancer cells. In the absence of this signal, normal cells restrict their entry into the cell cycle, whereas cancer cells bypass the Ca²⁺ checkpoints and undergo a catastrophic necrotic collapse during cytokinesis. Our results reveal a unique mitochondrial addiction to Ca²⁺ and a susceptibility of cancer cells that may have therapeutic implications.

EXPERIMENTAL PROCEDURES

Reagents

Antibodies

From Cell Signaling Technology: Phospho-α1-AMPK (Thr172), α1-AMPK, LC3B, HMBG1, ATG7, caspase 3. From BD Laboratories: β-tubulin, InsP₃R3. From Sigma: MCU. InsP₃R1 antibody: Dr. R.W. Neumar (University of Michigan, USA). Secondary antibodies conjugated with peroxidase from Amersham (Piscataway, NJ).

Chemicals

3-methyladenine, methyl-pyruvate, dimethyl 2-oxoglutarate, FCCP, oligomycin, rotenone and Hanks’ balanced salt solution were purchased from Sigma. Z-VAD-FMK and necrostatin-1 were from TOCRIS Biosciences. U73122 and U73343 were from Calbiochem. TOTO-3, DAPI, cleaved caspase 3 conjugated with FITC and secondary antibodies conjugated with Alexa 488 were from Molecular Probes. Xestospongin B was extracted and purified from the marine sponge Xestospongia exigua as described (Cardenas et al., 2005).

Cell Culture and Transfection

Details regarding cell culture and transfection are available in Supplementary Information.

Western Blotting and Treatments

Details of Western blotting and cell treatments are available in Supplementary Information.

Microscopy

Details of microscopy are available in Supplementary Information.

Oxygen Consumption

Oxygen consumption rate (OCR) was measured at 37°C using an XF24 extracellular analyzer (Seahorse Bioscience). Details provided in Supplementary Information.

Clonogenic Assay

Colony formation was assayed as described (Franken et al., 2006), with details provided in Supplementary Information.

Cell Cycle Analysis

Details provided in Supplementary Information.

Cell Death

Cell death was determined by LDH activity release by a colorimetric assay according to the manufacturer’s instructions (Roche) and by Propidium Iodide (PI, final concentration 2.5 μg/ml) exclusion by flow cytometry (FACS Canto, BD Biosciences, San Jose, CA, USA).

Mouse Tumor Generation and Histochemistry

All in vivo experiments were conducted with approval of the Institutional Animal Care and Use Committee. Murine melanoma derived B16F10 cells (1 × 10⁵ cells) were subcutaneously injected into flanks of immunodeficient mice and after 1 wk visible tumors were observed. Tumors of one flank were injected with 100 μl of a 100 μM XeB solution and those on the contralateral side were left untreated or injected 1% EtOH in PBS. Mice were euthanized after 1 d or 6 d treatment with XeB and tumors extracted, weighed and processed for immunohistochemistry.

Lactate Measurements

Cell lactate was measured colorometrically (Abcam) according to the manufacturer’s instructions.

Glucose Uptake

Fibroblasts were washed with Krebs-BSA buffer, pre-incubated or not with cytochalasin B (10 min) and incubated for 10 min at 37°C with 10 mM ³H-2-deoxyglucose. The reaction was stopped by placing it on ice, the samples were spun, decanted, lysed in 1% triton x-100 and cell ³H quantified.

Analysis and Statistics

All data summarized as mean ± SEM; significance of differences was assessed using unpaired t-tests. Differences were accepted as significant at the 95% level (p < 0.05).

Highlights.

• Low-level endoplasmic reticulum-to-mitochondria Ca²⁺ flux maintains ATP levels in cells

• Absence of basal mitochondrial Ca²⁺ uptake induces autophagy

• Autophagy protects normal cells but is insufficient for survival of cancer cells

• Cancer cells fundamentally require basal mitochondrial Ca²⁺ uptake for survival