Downregulation of type 3 inositol (1,4,5)-trisphosphate receptor decreases breast cancer cell migration through an oscillatory Ca2+ signal

Breast cancer remains a research priority due to its invasive phenotype. Although the role of ion channels in cancer is now well established, the role of inositol (1,4,5)-trisphosphate (IP3) receptors (IP3Rs) remains enigmatic. If the three IP3Rs subtypes expression have been identified in various cancers, little is known about their physiological role. Here, we investigated the involvement of IP3R type 3 (IP3R3) in the migration processes of three human breast cancer cell lines showing different migration velocities: the low-migrating MCF-7 and the highly migrating and invasive MDA-MB-231 and MDA-MB-435S cell lines. We show that a higher IP3R3 expression level, but not IP3R1 nor IP3R2, is correlated to a stronger cell line migration capacity and a sustained calcium signal. Interestingly, silencing of IP3R3 highlights an oscillating calcium signaling profile and leads to a significant decrease of cell migration capacities of the three breast cancer cell lines. Conversely, stable overexpression of IP3R3 in MCF-7 cells significantly increases their migration capacities. This effect is completely reversed by IP3R3 silencing. In conclusion, we demonstrate that IP3R3 expression level increases the migration capacity of human breast cancer cells by changing the calcium signature.


INTRODUCTION
Most frequently occurring cancer in women, breast cancer also presents the highest death rate. Cancer mortality is not correlated to tumor growth (due to proliferation mechanisms) but, in 90% of cases, to formation of metastasis [1] by migrating and invading cells. One of the mechanisms behind this invasion and metastasis process in breast cancer is the epithelialmesenchymal transition (EMT) that allows epithelial cancer cells to dedifferentiate and undergo the rear-tofront polarization and to acquire high migratory capacity, invasiveness, enhanced resistance to apoptosis, and stem cell properties [2,3]. Thus, EMT provides an explanation for why epithelial cancers with poor differentiation status are generally more aggressive and prone to metastasize than more differentiated cancers [4,5]. Cell migration is a complex multistep process that involves protrusions of the leading edge of the cell, formation of adhesion complexes and the release of adhesions at the cell rear. Calcium (Ca 2+ ) is a key effector of these migratory mechanisms by modulating the focal adhesion turnover or the cytoskeletal organization [6]. Ca 2+ increases can occur in the form of waves, spikes or oscillations with various impact on cell migration progression [7]. Several plasma membrane channels that increase Ca 2+ into the cytosol, such as the transient receptor potential (TRP) channels [8] and Orai/STIM channels [9,10] have been described in cancer cell migration, but implication of inositol-(1,4,5)-trisphosphate (IP 3 ) receptors (IP 3 Rs) [11] Research Paper and ryanodine receptors (RyRs) [12] in such process remain fragmented. IP 3 R protein subtypes (IP 3 R1, IP 3 R2 and IP 3 R3) are encoded by three different genes in mammals, however the resulting proteins share high similarity in their primary sequences and are expressed to varying degrees in different cell types [13]. Interestingly, Miyakawa et al. [14] demonstrated that IP 3 Rs subtypes differ by a specific Ca 2+ signature, which is associated to various sensitivity to endogenous modulators such as IP 3 , Ca 2+ and ATP. Thus, activation of IP 3 R1 generates very rapidly damped Ca 2+ oscillations, IP 3 R2 stimulation produces regular and robust Ca 2+ oscillations, whereas IP 3 R3 functions as an anti-Ca 2+ oscillatory units with a Ca 2+ transient signature [14,15] and is able to modulate the spatiotemporal pattern of intracellular Ca 2+ signals induced by ATP [16,17].
Overexpression of IP 3 Rs has been demonstrated in various cancer types where a pro-apoptotic [18,19] and pro-invasive [20,21] roles of IP 3 Rs have been established. Concerning IP 3 R3 subtype, it appears as a key actor of carcinogenesis as its expression level is correlated with colorectal carcinoma aggressiveness [22], whereas its inhibition reduces breast cancer cell proliferation [17], migration, invasion and survival of glioblastoma cells [20]. Furthermore, we recently demonstrated that IP 3 R3 co-localizes and interacts, both at molecular and functional levels, with voltage-and Ca 2+ -dependent K + channels (BK Ca ). This interaction appears to specifically occur in cancerous cells and increases cancer cell proliferation [23]. Based on our previous works [17,23], we investigated the role of IP 3 R3 dependent Ca 2+ signaling in the migratory process of three human breast cancer cell lines with different migration capacities. Our results clearly show that silencing of IP 3 R3 (siR3) reveals an oscillating Ca 2+ signature and significantly decreases migration of invasive breast cancer cells (MDA-MB-231 and MDA-MB-435S). Conversely, IP 3 R3 overexpression in MCF-7 cells modifies the ATP-induced Ca 2+ response from an oscillatory into a sustained signal and significantly increases their migration capability. This effect is completely reversed by siR3. Thus, our results strengthen the involvement of IP 3 R3 as a key player in the migration of breast cancer cells through modulation of their Ca 2+ signature.

IP 3 R3 expression level is specifically correlated to migration capacity of breast cancer cell lines
We have previously reported that IP 3 R3 is the unique isoform to positively regulate the 17-beta estradiolinduced proliferation of the estrogen-dependent MCF-7 cell line [17]. Here, we investigated the expression levels of the three IP 3 Rs subtypes in three breast cancer cell lines with various degree of malignancy: a very low migrating-(MCF-7), a metastatic-(MDA-MB-231), and a highly metastatic (MDA-MB-435S) cell lines. We first realized cell migration measurements using the Boyden transwell chamber assays in order to characterize and compare the migration potential of each cell line. The relative rank order of cell migration capacity of MCF-7, MDA-MB-231 and MDA-MB-435S is reported in Figure 1A. Relative cell migration is 1 ± 0.66, 15.43 ± 1.33 and 25.48 ± 1.59 N = 3, for MCF-7, MDA-MB-231 and MDA-MB-435S, cells respectively. In parallel, we measured and compared the expression level of IP 3 R3 at the RNA ( Figure 1B) and at the protein ( Figure 1C) levels in each cell line. Interestingly, it appears that a higher RNA and protein IP 3 R3 expression level is correlated to a higher migration capacity of breast cancer cell lines. The relative IP 3 R3 RNA and protein expression levels are respectively in MCF-7 (1 ± 0.04 (N = 3) and 1 ± 0.06 (N = 3)); MDA-MB-231 (1.41 ± 0.08 (N = 3, p = 0.003) and 1.78 ± 0.18 (N = 3, p = 0.04)) and MDA-MB-435S (1.52 ± 0.06 (N = 3, p = 0.004) and 2.41 ± 0.28 (N = 3, p = 0.02)). Immunostaining with anti-IP 3 R3 antibody confirmed this marked labeling in highly migrating cells MDA-MB-435S compared to MDA-MB-231 and MCF-7 cells ( Figure 1D). This correlation between the cell migration potential and the IP 3 R3 expression is specific to IP 3 R3 subtype, since it is not observed with the others IP 3 R1 and IP 3 R2 subtypes ( Figure 2). To appreciate the IP 3 R3 expression compared to the two others IP 3 R subtypes, we also investigated RNA and protein expression levels of both IP 3 R1 (Figure 2Aa and 2Ab, Table 1) and IP 3 R2 (Figure 2Ba and 2Bb, Table 1) in the same batch of the three cell lines. Similarly to IP 3 R3, IP 3 R1 is predominantly expressed in MDA-MB435s (Table 1), whereas IP 3 R2 appeared as expressed in MCF-7 as in MDA-MB-435S cell lines (Table 1). Moreover, the rationalization of IP 3 R3 expression to IP 3 R1 (Figure 2Ac) or IP 3 R2 (Figure 2Bc) protein expression levels confirmed its overexpression in migrating cell lines compared to the low migrating MCF-7 cell line. The slight IP 3 R1 and IP 3 R2 levels measured in MDA-MB-231 cells accentuate the predominance of IP 3 R3 in these cells (Figure 2Ac and 2Bc), even if MDA-MB-435S cells show the highest IP 3 R3 expression level ( Figure 1B). Taken together, our results establish, for the first time, a specific correlation between IP 3 R3 expression level and the cell migration capacities in breast cancer cell lines.

IP 3 R3 silencing drastically decreases migration of breast cancer cell lines
In order to confirm that IP 3 R3 is implicated in cell migration, we realized cell migration assays in cells transfected with two siRNA targeting IP 3 R3 (siR3) compared to cells transfected with a control siRNA (siC). For each cell line, the migration was measured at different time duration considering the different migration rate of each cell line. IP 3 R3 being involved in breast cancer cell proliferation as we already described [17,23], cell migration was assessed during a time frame which does not exceed 24 h where the cell viability, assessed by MTT, is not yet altered. This allows us to throw off the effect of IP 3 R3 silencing on cell proliferation and to ensure that the observed effects are due solely to the change in migration capacities of breast cancer cells. Cell migration and viability were thus tested 24 h after seeding, in the transwell chamber, for MCF-7 cells and 16 h after seeding for both MDA-MB-231 and MDA-MB-435S cells. In both cases, this time corresponds to 72 h post-transfection with respective siRNA. IP 3 R3 gene silencing reduced cell migration of the three cell lines, with a greater effect in the highly migrating MDA-MB-231 and MDA-MB-435S cells (Figure 3Aa, 3Ba, 3Ca, Table 2 and Supplementary Figure 1). In all conditions, the cell viability was not affected by  Table 3). The efficiency and specificity of IP 3 R1, IP 3 R2 and IP 3 R3 gene silencing following the transfection by specific siRNAs were confirmed by quantitative real-time PCR (Supplementary Figure 2 and Table 4A) and Western-blot (Supplementary Figure 3 and Table 4B) on the three breast cancer cell lines. Nevertheless, IP 3 R1 or IP 3 R2 silencing remains less invalidating on cell migration compared to IP 3 R3 silencing ( Figure 3). Like for siR3 conditions, IP 3 R1 or IP 3 R2 silencing has no effect on cell viability ( Figure  4Ab, 4Bb, 4Cb and Table 3).

ATP-induced maintained Ca 2+ mobilization is an IP 3 R3-dependent signaling pathway
Migration processes implying Ca 2+ flux [24][25][26], we investigated whether siR3-induced migration inhibition was associated with a modification of the Ca 2+ homeostasis. Breast cancer cell lines transfected with siC or siR3 were loaded with Fura-2/AM (2 µM) for 45 min and ATP (5 µM) was applied, in a Ca 2+ -free solution, to prevent ionotropic/purinergic receptors activation protein expression level was rationalized to IP 3 R1 (Ac) and to IP 3 R2 (Bc) expression levels. Actin protein was used as loading control and, quantitative analyses of Western-blots are the average of three independent experiments. Values are reported as mean ± SEM normalized to the MCF-7 cells (N = 3). *p < 0.05, **p < 0.01. www.impactjournals.com/oncotarget [23,27]. Statistical analyses revealed that IP 3 R3 silencing do not modify neither basal intracellular Ca 2+ values ( Figure 5A) nor the percentage of responding cells to ATP ( Figure 5B) in all cell lines (Table 5). We have previously reported that the decrease of IP 3 R3 expression level changes the Ca 2+ signal profile from a plateau-type to a sinusoidal oscillatory-shaped signal, which is in favor of a diminution of MCF-7 cell proliferation [17]. In this study, we investigated if such signaling pattern could also be related in highly migrating cells transfected with siR3. Statistical analyses revealed that at 72 h post-transfection, the number of oscillating cells in response to ATP is much higher in siR3-transfected cells compared to siCtransfected cells ( Figure 5C) in all cell lines (Table 5). Figure 5D represents typical Ca 2+ signals measured at 72 h post-transfection after perfusion with 5 μM ATP in a Ca 2+free medium. In order to determine if this IP 3 R3 signaling profile modification is associated with a change of ATPinduced Ca 2+ mobilization, we measured the "area under curve" (AUC) for each trace in each cellular condition. Thus, the mean AUC values for Ca 2+ signals elicited in siCtransfected versus siR3-transfected cells are statistically different ( Figure 5E) in all cell lines ( Table 5) Figure 6B) ATP basal values as no significant differences could be detected for each condition. Moreover ATP induced-Ca 2+ oscillations are specific to IP 3 R3 silencing since it is not observed in IP 3 R1-or IP 3 R2-silenced cells ( Figure 7D). Indeed, IP 3 R1 silencing significantly decreased the resting ratio, whereas IP 3 R2 silencing increased it compared to control conditions ( Figure 7A; 1.34 ± 0.005 vs. 1.26 ± 0.007 (p = 1.10 -17 ) and 1.40 ± 0.01 (p = 3.10 -6 ) for siC, siR1 and siR2 conditions respectively, N = 160 and 129 cells respectively). Silencing IP 3 R1 or IP 3 R2 does not affect the percentage of responding cells to ATP (5 µM) perfusion ( Figure 7B). Interestingly, the global amount of Ca 2+ released into the cell remained unchanged in siR2 The values are mean of triplicate assays (N = 3). The significant differences are shown with superscript letters and p values are: a = 0.009; b = 0.008 and c = 0.021.  conditions ( Figure 7C), but is drastically increased in siR1 conditions (21.16 ± 0.81 vs. 38.06 ± 1.21, n =160 (p = 4.10 -28 ) and 21.03 ± 0.86, n = 129, for siC, siR1 and siR2 conditions respectively).

IP 3 R3 overexpression enhances the migration capacity of MCF-7 cell line
Based on our results showing that siR3 decreased cancer cell migration, we investigated the effect of a stable overexpression of IP 3 R3 in the low migrating MCF 7 cell line. The transfection of MCF-7 cells with a pcDNA(3.1) plasmid encoding IP 3 R3 [28] led to a 2.59 fold increase in IP 3 R3 expression level (2.59 ± 0.6 in pcDNA(3.1)/ IP 3 R3-transfected cells vs. 1 ± 0.4 in empty pcDNA(3.1)transfected cells; N = 3, p = 0.03) ( Figure 8A). In order to confirm that the effects recorded were correlated to IP 3 R3 level increase and not to a modification of IP 3 R1 and/or IP 3 R2 expression levels, RT-qPCR analysis of all IP 3 R isoform mRNAs was carried out. Our results show no significant change in the expression of IP 3 R1 and IP 3 R2 at the mRNA level in MCF-7 cells overexpressing IP 3 R3 ( Figure 8B). The relative expression level is 0.77 ± 0.17 (p = 0.15) of control (empty pcDNA(3.1)-transfected cells) for IP 3 R1 and 0.78 ± 0.51 (p = 0.32) of control for IP 3 R2, whereas it is significantly elevated to 11.96 ± 1.81 (p = 0.03) of control for IP 3 R3. We then compared migration ability of MCF-7 overexpressing IP 3 R3 vs empty pcDNA(3.1)-transfected MCF-7 cells ( Figure 8C). The relative migration of IP 3 R3 overexpressing MCF-7 cells is enhanced by more than a 2 fold factor (2.26 ± 0.37 vs. 1 ± 0.33; N = 3; p = 0.008). Furthermore, the migration of MCF-7 stably overexpressing IP 3 R3 is reduced to control values by siR3 (0.82 ± 0.19; N = 3; p = 0.01; IP 3 R3+siR3). This effect on cell migration is independent of any effect on cell viability ( Figure 8D). We also record intracellular  Figure 9A). The perfusion of ATP (5 µM) induced a plateau-type signal that appears more sustained in IP 3 R3 overexpressing cells ( Figure 9B) without modify the amount of Ca 2+ mobilized ( Figure 9C; 5.02 ± 0.37 vs. 4.72 ± 0.39 in empty and IP 3 R3 transfected cells respectively). Indeed, the measure of ∆R/R ratio to Thapsigargin (TG)-or ATP-induced Ca 2+ response shows no significant differences (Supplementary Figure 4B and 4C). This result indicates that IP 3 R3 overexpression modify rather intracellular Ca 2+ availability than intracellular Ca 2+ homeostasis. This Ca 2+ mobilization maintained during ATP application stimulates migration capacities of MCF-7 cells ( Figure  10A). Indeed, migration of MCF-7 cells transfected with the pcDNA(3.1)-empty plasmid is increased following ATP stimulation (2 µM) (8.41 ± 1.2 in ATP treated cells vs. 1 ± 0.24 in Ctrl; N = 3, p = 0.0001). This increase seems to occur independently of IP 3 R3 since IP 3 R3 silencing failed to decrease the ATP-stimulated migration. However, in MCF-7 stably overexpressing IP 3 R3, and was not correlated to a proliferative effect ( Figure 10B). ATP strongly increases cell migration (10 ± 1.24 in ATP treated cells vs. 1 ± 0.24 in Ctrl; N = 3, p = 0.03) that is abolished following IP 3 R3 silencing.

DISCUSSION
It is well known that Ca 2+ ions, together with many other Ca 2+ -dependent proteins and signaling pathways, play a crucial role in regulating cell migration [25,26,29,30]. For example, capacitive Ca 2+ entry has been shown to be involved in the migration of breast cancer cells [31], hepatocarcinoma cells [32] and in vascular smooth muscle cells [33]. Conversely, few studies have shown the role of the secretory pathway of Ca 2+ in cancer cell migration. Thereby RyRs have been found to control astrocyte cell migration [34] whereas the SPCA1 Ca 2+ -ATPase has been shown to be involved in cell migration during Caenorhabditis elegans embryonic development [35]. In this context, and based on previous works showing a role of Ca 2+ and IP 3 R Ca 2+ release channels in cell motility, we have studied the potential implication of IP 3 R3 in cell migration process in three different breast cancer cell lines showing distinct migration capacities. Our study demonstrates, for the first time, that IP 3 R3 by modifying the calcium signal profile can regulate different breast cancer cell migration capacities. Indeed we establish that (i) MCF-7, MDA-MB-231 and MDA-MB-435S breast cancer cell lines express different levels of IP 3 R3 protein, (ii) cells expressing IP 3 R3 in a larger amount migrate more extensively than the others, (iii) silencing of IP 3 R3 changes a sustained ATP-induced Ca 2+ increase to an oscillatory one, (iv) overexpression of IP 3 R3 increases the migration capacities of the non-invasive MCF-7 cell line and switch the ATP-induced transient Ca 2+ response to a sustained phase.
Interestingly, we were able to discriminate a modulation of ATP-induced Ca 2+ response specifically correlated to IP 3 R3 expression. Indeed, down-regulation of IP 3 R3 in high migrating cell lines reveals an oscillatory ATP-induced Ca 2+ signal, whereas overexpression of IP 3 R3 in low migrating breast cancer cell line modify the Ca 2+ signal in a sustained one. These results are in agreement with previous studies showing that IP 3 R3 forms an anti-oscillating unit, its down-regulation revealing   a Ca 2+ oscillating profile specific to IP 3 R2 activation [15]. Interestingly, Ca 2+ oscillations associated to IP 3 R3 silencing were associated to a reduction of the intracellular Ca 2+ concentrations in MCF-7 and MDA-MB-231, and an increase in MDA-MB435s cells. Thus, even if IP 3 R3 silencing systematically induces an oscillating Ca 2+ signature in breast cancer cells, the Ca 2+ amount involved in this response differ between the various cell lines. This discrepancy could be explained by differences in the expression levels of the purinergic receptors P2 [36][37][38][39] or activation of P1 adenosine receptors due to the ATP hydrolysis by ectonucleotidases such as CD39 or CD73 [40]. Moreover a variation of intracellular Ca 2+ availability through the Mitochondrial Calcium Uniport (MCU) activity [41] could also convey such difference between MCF-7 and MDA-MB-231 vs MDA-435s cells.
One of the major issues of the present work lies in the identification, for the first time, that IP 3 R3, by regulating the Ca 2+ homeostasis in an anti-oscillating profile, impact the migratory capacities of breast cancer cells. Our results are thus supported by other studies showing that IP 3 R3 is overexpressed in various cancer tissues. Indeed, IP 3 R3 expression is found to be increased in gastric carcinoma [21] and glioblastoma [20] where it seems to be involved respectively in peritoneal dissemination and in brain invasion. Kang et al. [20] elegantly demonstrate that caffeine, at concentration that inhibits IP 3 R3 preferentially to the two other IP 3 R subtypes, is able to inhibit migration and invasion of glioblastoma cells in vitro. Moreover, IP 3 Rs inhibition markedly reduces the migration of pancreatic adenocarcinoma cells by modulating the Ca 2+ signaling complexes [42]. We establish a link between an oscillating IP 3 -dependent Ca 2+ signal and the migration capacity of breast cancer cells. Indeed, if Ca 2+ oscillating signature has been described to modulate dendritic [43] or astrocytoma [44,45] cell migration; such correlation has never been characterized on cancer cells. These intracellular Ca 2+ fluctuations, by modulating the availability of free Ca 2+ , may regulate migration processes. A variety of intracellular signaling molecules and their associated biochemical pathways have been identified in the regulation of cell migration such as actin and myosin complexes [46,47], and Rho subfamily of small G proteins (Rho, Rac, and Cdc42) [48,49]. This Ca 2+ signaling has also been shown to activate myosin light chain kinase (MLCK) and control the local disassembly of focal adhesions [24,[50][51][52][53][54]. IP 3 R3-dependent Ca 2+ oscillations may be the consequences of binding partners like IRBIT [55], Ca 2+ binding proteins [56], or a close interaction between IP 3 Rs and the Mitochondrial Calcium Uniport (MCU) [57,58]. Further studies are needed to determine the implication of these actors in IP 3 Rs dependent oscillations genesis.
Altogether, our results clearly indicate that IP 3 R3 is involved in the migration of human breast cancer cells through a specific calcium signature. Overexpression of IP 3 R3 increases cell migration capacities by inducing a sustained Ca 2+ signature, whereas down-regulation of IP 3 R3 expression reveals an oscillating Ca 2+ signature along with a slowing down of cell migration. Thus, this work led us to put forward the hypothesis that IP 3 R3, by remodeling the Ca 2+ signal, is a key player in the migration of human breast cancer cells.

Cell culture
Human breast epithelial cancer cell lines MCF-7, MDA-MB-231 and MDA-MB-435S were obtained from the ATCC (American Type Culture Collection). Cell lines were cultured in the Minimum Essential Medium (MEM, Gibco, LifeSciences, Cergy Pontoise, France) complemented with 0.45% of sodium bicarbonate, Protein expression level of IP 3 R3 was evaluated by Western-blot (black column) and compared to control conditions (pcDNA(3.1)-empty plasmid) (white column). (B) RT-qPCR was carried out to investigate the impact of stable IP 3 R3 overexpression on IP 3 R1 and IP 3 R2 expression levels. (C) The effect of stable IP 3 R3 overexpression on the migration capacities of MCF-7 cells was evaluated by Boyden chamber migration assay and the specificity of IP 3 R3-overexpression effect was controlled by an siRNA targeting IP 3 R3 in the pcDNA(3.1)-IP 3 R3 transfected cells. In all experiments, the migration was always measured 72 h after cell transfection with the respective siRNAs and at a time at which there was no effect on cell viability (D). Values are reported as mean ± SEM normalized to MCF-7 cells transfected with empty plasmid (N = 3). *p < 0.05, **p < 0.01. 0.06% Hepes, 0.1% of non-essential amino acids, 2 mM L-glutamine, and 5% fetal calf serum (FCS, Lonza, Aubergenville, France). Cell lines were maintained at 37°C and 5% CO 2 in a humidified atmosphere. Cells were used up to 20 passages after ATCC vial defrosting.

Cell migration
Migration tests were performed in 8 µm pore sized membrane Boyden chamber (BD FALCON TM Cell Culture Inserts, BD Biosciences, Le Pont de Claix, France) according to the manufacturer's protocol. The upper compartment was seeded with 4.10 4 cells 48 h after transfection (siC and siR3 or pcDNA(3.1) emptyplasmid and pcDNA(3.1) IP 3 R3-plasmid) in MEM medium supplemented with 5% FCS. The lower compartment was filled with MEM medium supplemented with 5% FCS. After 16 h of further incubation at 37°C for MDA-MB-231 and MDA-MB-435S cells and 24 h for the low migrating MCF-7 cells (as previously described [59,60]), migrated cells which have passed to the lower side were washed by PBS, fixed in ice-cold methanol and stained by hematoxylin. The remaining cells were removed from the upper side of the membrane by scrubbing. Migrated cells were counted under  an inverted microscope (Nikon eclipse TS100 microscope, Champigny-sur-Marne, France) in duplicate (20 contiguous areas at 400 × magnification for each insert). For each experiment, the number of migrating cells per area for each condition (siC and siR3 or pcDNA(3.1) empty-plasmid and pcDNA(3.1) IP 3 R3-plasmid) was normalized to control (siC and pcDNA(3.1) empty-plasmid). For each experiment, a cell viability test was carried out in the same condition than the migration assays.

Total RNA isolation and quantitative RT-qPCR
After transfection, 5.10 5 cells were seeded in 100 mm petri dishes in MEM medium 5% FCS. Total RNA was harvested from the cells with the standard trizolphenol-choloroform protocol. Complementary DNA was synthesized from 2 μg of RNA with random hexamers and MultiScribe ™ Reverse Transcriptase (Applied Biosystems, Carlsbad, CA). Relative abundance of mRNA was quantified based on Ct difference on a LightCycler real-time PCR machine (Roche, Basel, Switzerland) using a mix containing SYBR green, Taq polymerase and specific primers (Table 6). Results were expressed as gene expression normalized to β-actin expression.

Cell transfection
Cell transfection was performed using the nucleofection technology according to the Amaxa Biosystems manufacturer's instructions. For the IP 3 R silencing experiments cells were transfected with siRNAs targeting the three IP 3 R isoforms (siR1, 2, 3, ON-TARGET plus, Dharmacon, GE Healthcare Life Sciences, Velizy-Villacoublay, France) or with scrambled siRNA as a control (siC; siGENOME Non-Targeting siRNA, Dharmacon, GE Healthcare Life Sciences, Velizy-Villacoublay, France). MCF-7 cell line was also transfected with plasmid encoding IP 3 R3 (pcDNA(3.1)-IP 3 R3 plasmid, gift from Professor Jan B. Parys [28]) or with the corresponding empty plasmid (pcDNA(3.1)empty plasmid) for the overexpression experiments. Briefly, 1.10 6 cells (MDA-MB-231 and -435 s) or 2.10 6 (MCF-7 cells) were transfected with 2 μg siRNA or pcDNA(3.1) plasmid. After the electroporation, 500 μL of prewarmed MEM medium supplemented with 5% FCS was added and cells were placed at 37°C for 15 min in a CO 2 incubator. Immediately after transfection, cells were cultured 48 to 72 h for Western-blot, RT-qPCR, cell migration and calcium imaging experiments.

Cell viability
In parallel to the migration assay, cell viability was evaluated for the different conditions. Briefly, 2.10 4 cells were seeded in 6-well plates in EMEM medium 5% FCS. After 16 to 24 h (according to the migration time for each cell line), cells were washed with phosphatebuffered saline (PBS) and incubated with a medium containing 0.5 mg/ml of 3-(4,5-dimethylthiazole-2-yl)-2,5diphenyltetrazolium bromide (MTT, Sigma, Saint-Quentin Fallavier, France) and incubated for 1 h at 37°C. Medium was removed and 800 μL of Dimethyl-sulfoxide (DMSO) was added to solubilize crystals. The optical density of each sample was read on the microplate reader Infinite F200 Pro (TECAN, Lyon, France) at 570 nm.

Immunofluorescence microscopy
Cells were cultured on coverslips for 72 h. After washing 3 times with PBS, cells were fixed in 3% paraformaldehyde for 10 min at RT. Cells were then washed twice in PBS, blocked with and permeabilized with PBS containing Triton (0.1%) and BSA (5%) for 30 min and incubated overnight with primary antibodies at 4°C. The following day, cells were washed four times in PBS and labeled with an Alexa Fluor 488-labeled goat anti-mouse and a DyLight549-conjugated goat antirabbit secondary antibodies for 1 h in the dark at room temperature. After labeling, cells were washed with PBS and incubated with 4′, 6-diamidino-2-phenylindole1% for 1 min. Cells were then washed two more times and mounted onto slides with ProLong ® Gold antifade reagent (Life Technologies, Villebon-sur-Yvette, France). Images were acquired with a Zeiss LSM780 confocal microscope (Marly-le-Roi, France) and analyzed with ZEN 2012 software.

Calcium imaging
After transfection, 7.10 4 cells were seeded in MEM medium 5% FCS in 35 mm petri dishes on glass cover slips. After 3 days, cells were loaded for 45 min with Fura-2/AM (2 μM in medium solution) at 37°C in a CO 2 incubator and subsequently washed with MEM medium. The cover slip was then transferred into a perfusion chamber of a fluorescence Zeiss inverted microscope (Marly-le-Roi, France). Fluorescence was excited at 340 and 380 nm alternately, using a monochromator (Polychrome IV; TILL Photonics, Planegg, Germany), and captured by a Cool SNAP HQ camera (Princeton Instruments, Evry, France) after filtration through a long-pass filter (510 nm). Background fluorescence was determined at 340 and 380 nm from an area of the cover slip free of cells. These values were routinely subtracted. Metafluor software (version 7.1.7.0, Molecular Devices, St. Grégoire, France) was used for acquisition and analysis.
All recordings were carried out at room temperature. Cells were continuously perfused with the saline solution, and chemicals were added via the perfusion system. The flow rate of the whole-chamber perfusion system was set at 10 mL/min, and the chamber volume was 1 mL. Recording solution had the following composition (in mM): NaCl (145), KCl (5), CaCl 2 (2), MgCl 2 (1), glucose (5) and Hepes (10) at pH 7.4 (NaOH). In experiments where Ca 2+ -free solution was used, Ca 2+ was omitted and EGTA (0.4 mM) was added to the solution.

ATP measurements
The luciferin⁄luciferase detection of ATP was performed with the microplate reader Infinite F200 Pro (TECAN, Lyon, France) with ATP bioluminescent somatic cell assay kit (FLASC, Sigma, Saint-Quentin Fallavier, France). Breast cancer cell lines MCF-7, MDA-MB-231 and MDA-MB-435S transfected with siRNA control (siC) or siRNA targeting IP 3 R3 (siR3) were seeded at 2000 cells/well in white 96-well Nunc dishes with clear bottoms 72 hours before ATP measurements. For extracellular ATP measurements, 100µl of supernatant were incubated with the luciferin⁄luciferase (FLAAM, Sigma) at a final concentration of 0.04%. For intracellular ATP measurements cells were incubated with 100 µl of ATP releasing reagent (FLSAR, Sigma), before incubation with the luciferin⁄luciferase (FLAAM, Sigma). To determine the amount of ATP released, a calibration curve was constructed using known concentrations of ATP in solution (1, 10, 100, 1000, 10 000, 100 000 pM). Control experiments were performed to eliminate any drug effect on luciferase activity.

Reagents
All the products were from Sigma (Saint-Quentin Fallavier, France) unless otherwise stated. Final concentrations were obtained by appropriate dilution of stock solutions so that the solvent never exceeded 1/1,000.

Statistical analysis
All data are expressed as mean ± SEM of at least three independent experiments. N refers to the number of experiments repeated and n to the number of tested cells. The Student's t-test and one-way analysis of variance (ANOVA) with Bonferroni post-hoc analysis were used to group comparison. Statistical significance is indicated in the figures (* P < 0.05; ** P < 0.01; *** P < 0.001).