Bcl-xL acts as an inhibitor of IP3R channels, thereby antagonizing Ca2+-driven apoptosis

Anti-apoptotic Bcl-2-family members not only act at mitochondria but also at the endoplasmic reticulum, where they impact Ca2+ dynamics by controlling IP3 receptor (IP3R) function. Current models propose distinct roles for Bcl-2 vs. Bcl-xL, with Bcl-2 inhibiting IP3Rs and preventing pro-apoptotic Ca2+ release and Bcl-xL sensitizing IP3Rs to low [IP3] and promoting pro-survival Ca2+ oscillations. We here demonstrate that Bcl-xL too inhibits IP3R-mediated Ca2+ release by interacting with the same IP3R regions as Bcl-2. Via in silico superposition, we previously found that the residue K87 of Bcl-xL spatially resembled K17 of Bcl-2, a residue critical for Bcl-2’s IP3R-inhibitory properties. Mutagenesis of K87 in Bcl-xL impaired its binding to IP3R and abrogated Bcl-xL’s inhibitory effect on IP3Rs. Single-channel recordings demonstrate that purified Bcl-xL, but not Bcl-xLK87D, suppressed IP3R single-channel openings stimulated by sub-maximal and threshold [IP3]. Moreover, we demonstrate that Bcl-xL-mediated inhibition of IP3Rs contributes to its anti-apoptotic properties against Ca2+-driven apoptosis. Staurosporine (STS) elicits long-lasting Ca2+ elevations in wild-type but not in IP3R-knockout HeLa cells, sensitizing the former to STS treatment. Overexpression of Bcl-xL in wild-type HeLa cells suppressed STS-induced Ca2+ signals and cell death, while Bcl-xLK87D was much less effective in doing so. In the absence of IP3Rs, Bcl-xL and Bcl-xLK87D were equally effective in suppressing STS-induced cell death. Finally, we demonstrate that endogenous Bcl-xL also suppress IP3R activity in MDA-MB-231 breast cancer cells, whereby Bcl-xL knockdown augmented IP3R-mediated Ca2+ release and increased the sensitivity towards STS, without altering the ER Ca2+ content. Hence, this study challenges the current paradigm of divergent functions for Bcl-2 and Bcl-xL in Ca2+-signaling modulation and reveals that, similarly to Bcl-2, Bcl-xL inhibits IP3R-mediated Ca2+ release and IP3R-driven cell death. Our work further underpins that IP3R inhibition is an integral part of Bcl-xL’s anti-apoptotic function.


INTRODUCTION
Inositol 1,4,5-trisphosphate receptors (IP 3 Rs) are tetrameric Ca 2 + -permeable channels, predominantly located at the endoplasmic reticulum (ER) membrane [1][2][3]. Ca 2+ release through IP 3 Rs plays fundamental roles in a plethora of cellular processes, including proliferation, gene transcription, protein secretion, neurotransmitter release, fertilization, and apoptosis [4]. To maintain fidelity and specificity of these processes the activity of IP 3 Rs is tightly regulated at multiple levels. Among the most common regulatory mechanisms are the modulation of channel expression, posttranslational modifications, and interaction with regulatory factors including Ca 2+ itself, ATP and protein partners [1,5,6]. These regulators target different IP 3 R regions, which are arranged as globular domains such that the controlled trypsinization of IP 3 R generates five reproducible fragments [7], which have proven an excellent tool for dissecting the binding sites of different IP 3 R partners [8][9][10][11][12].
The B-cell lymphoma 2 (Bcl-2) family of proteins is well known for its role in controlling mitochondrial apoptosis and mitochondrial dynamics [13,14]. Anti-apoptotic Bcl-2 family members neutralize pro-apoptotic family members, including Bax/Bak and pro-apoptotic BH3-only proteins [15]. At the molecular level, anti-apoptotic Bcl-2 family members use their hydrophobic cleft (formed by the BH3-BH1-BH2 domains) to bind the BH3 domains of pro-apoptotic Bcl-2 family proteins [16][17][18]. Bcl-2 proteins also act at the ER, where they impact Ca 2+ homeostasis [19]. Antiapoptotic Bcl-2 and B-cell lymphoma-extra large (Bcl-xL) have emerged as important IP 3 R modulators [20,21]. The present consensus is that Bcl-2 inhibits the IP 3 R channel activity [20,22]. At the molecular level, inhibition of channel activity prominently occurs through the BH4 domain of Bcl-2 (BH4-Bcl-2) that targets several regions of the IP 3 R channel. Initially, inhibition of IP 3 R by Bcl-2 was explained by the interaction between the BH4 domain of Bcl-2 and a stretch of 20 amino acids (a.a. 1389-1408 of mouse IP 3 R1) located in the central, modulatory region and more specifically in the third tryptic IP 3 R fragment (Fragment 3) [8,23]. Recently, the BH4 domain was also found to target the ligand-binding domain (LBD), particularly the IP 3 -binding core (a.a. 226-604) [24]. In addition, we revealed a critical role for the C-terminal transmembrane domain of Bcl-2 to recruit the protein near the 6 th helix of the transmembrane domain of the IP 3 R [11,12].
In contrast to Bcl-2, Bcl-xL has been suggested to sensitize IP 3 Rs [25], specifically to promote channel opening at lower concentrations of IP 3 . Indeed, in DT40 cells, Bcl-xL was reported to increase IP 3 single-channel activity and to promote Ca 2+ oscillations by sensitizing IP 3 Rs. This effect in turn was proposed to maintain cell survival by optimizing mitochondrial bio-energetics. Moreover, it was shown that this pro-survival function of Bcl-xL relies predominantly on IP 3 Rs, since Bcl-xL overexpression in DT40 cells lacking all three endogenous IP 3 R isoforms (3KO) was much less effective in protecting the cells against pro-apoptotic stimuli [25]. At the molecular level, Bcl-xL was proposed to act by targeting two BH3-like domains in the C-terminal part of the channel to account for IP 3 R sensitization [26]. In that study, Bcl-xL was reported to only weakly bind to the central, modulatory region, underlying inhibition of IP 3 Rs only at very high Bcl-xL concentrations.
We previously demonstrated a critical and unique role for the K17 residue in the BH4 domain of Bcl-2 to mediate the inhibition of IP 3 R activity. A lysine residue at this position is not present in the BH4 domain of Bcl-xL (BH4-Bcl-xL) and this domain fails to inhibit IP 3 R activity [27]. Notably, in the linear sequence of Bcl-2, K17 corresponds to D11 of Bcl-xL, and substitution of K17 to an aspartate residue in Bcl-2 abrogated the ability of the BH4 domain to inhibit IP 3 Rs. Conversely, switching D11 to a lysine residue in Bcl-xL rendered BH4-Bcl-xL capable of inhibiting IP 3 Rs. These data appear to provide a rationale for the distinct actions of Bcl-2 and Bcl-xL. Nevertheless, using structural modeling, we previously found that another positively-charged residue, K87, located in the BH3 domain of Bcl-xL, spatially resembled the K17 in the BH4-Bcl-2 [28]. This observation prompted us to revisit the idea that Bcl-xL is truly an IP 3 R-sensitizing protein. In contrast to existing literature about Bcl-xL, we found that Bcl-xL inhibits IP 3 R-mediated Ca 2+ release, when overexpressed in living cells, and IP 3 R singlechannel openings, when applied as purified protein. At the molecular level, we demonstrate that Bcl-xL binds to full-length IP 3 R, and could target the same IP 3 R regions as the one targeted by Bcl-2's BH4 domain, namely the ligand-binding domain and the central, modulatory region. Moreover, mutating K87 to D in Bcl-xL impaired its ability to bind to full-length IP 3 Rs, as well as to the ligand-binding and the central, modulatory regions. In line with these observations, the mutant Bcl-xL K87D failed to inhibit IP 3 R function in living cells and in single-channel recording. Furthermore, we show that K87 of Bcl-xL is critical to protect cells against staurosporine-induced apoptosis which is dependent on IP 3 R/ Ca 2+ signaling. Finally, we demonstrate that, in MDA-MB-231, a Bcl-xL-dependent breast cancer cell model, endogenous Bcl-xL can exploit this mechanism to suppress IP 3 R activity and to counteract Ca 2+ -driven apoptosis. Indeed, lowering Bcl-xL-protein levels in MDA-MB-231 cells resulted in augmented ATP-induced IP 3 R-mediated Ca 2+ release and in increased sensitivity to staurosporine (STS). Overall, our data challenge the current paradigm that Bcl-xL promotes cell survival by sensitizing IP 3 Rs to IP 3 . Instead, we demonstrate that Bcl-xL inhibits IP 3 R function through a conserved lysine residue in its BH3 domain, thereby protecting cells against IP 3 R/Ca 2+ -driven apoptosis.

RESULTS
Bcl-xL inhibits IP 3 R-mediated Ca 2+ release in living cells Bcl-xL has been reported to sensitize IP 3 Rs in living cells [26]. Here, we evaluated the effect of Bcl-xL overexpression on IP 3 R function by monitoring agonist-induced Ca 2+ release (Fig. 1, Fig. S1). First, we performed Ca 2+ measurements in a population-based assay, using the ratiometric fluorescent Ca 2+ probe Fura-2 (Fig. S1). We used trypsin, an efficient agonist of protease activated receptors 2 in HEK-293 cells [29], thereby triggering IP 3 formation. We elicited IP 3 R-mediated Ca 2+ release in Fura-2-loaded HEK-3KO cells with reconstituted rIP 3 R1 (HEK-rIP 3 R1) and we studied the impact of overexpressing Bcl-xL. For this, we transfected the cells with either P2A-mCherry or 3xFLAG-Bcl-xL-P2A-mCherry. The 3xFLAG-Bcl-xL-P2A-mCherry construct generates separate mCherry and 3xFLAG-Bcl-xL proteins due to its P2A self-cleaving sequence. In contrast to previous reports [25,26], Bcl-xL overexpression significantly reduced the amplitude (Fig. S1a, c) and the area under the curve of the Ca 2+ signals (Fig. S1b, d) induced by both low (0.1 µM) and high (1 µM) trypsin concentrations. Moreover, similarly to our findings related to Bcl-2 and IP 3 R function [24], we noticed that the inhibitory effect of Bcl-xL overexpression was more prominent at low agonist concentrations than at high agonist concentrations. This unexpected IP 3 R inhibition by Bcl-xL in population-based Ca 2+ measurements prompted us to validate the effect of Bcl-xL overexpression on Ca 2+ signals in single cells exposed to other agonists ( Fig. 1) and to compare it with Bcl-2 overexpression, an established inhibitory modulator of IP 3 Rs [20,22]. Single cell Ca 2+ imaging was performed in Fura-2-loaded HEK-293 cells transfected with either P2A-mCherry, 3xFLAG-Bcl-xL-P2A-mCherry or 3xFLAG-Bcl-2-P2A-mCherry, whereby only mCherry-positive cells were analyzed. Here, we used ATP (10 µM) to trigger IP 3 R-mediated Ca 2+ release ( Fig. 1a, b, c). Similarly, we observed that Bcl-xL overexpression reduced the percentage of responding cells (Fig. 1d) and the area under the curve as representative of the extent of Ca 2+ release (Fig. 1e). Interestingly, Bcl-xL appeared to dampen IP 3 R-mediated Ca 2+ release to a similar extent as Bcl-2. We also measured the effect of Bcl-xL/Bcl-2 overexpression on Ca 2+ signals elicited by another agonist, namely carbachol ( Fig. 1f, g, h). We found that similarly to Bcl-2, Bcl-xL also inhibited carbachol-induced Ca 2+ signals, although Bcl-xL appeared less potent than Bcl-2 ( Fig. 1i, j). These results indicate that, similarly to Bcl-2, Bcl-xL inhibits IP 3 R-mediated Ca 2+ release, irrespective of the extracellular agonist that is applied.
Full-length Bcl-xL, but not its BH4 domain, targets the LBD of IP 3 R1 Next, we elucidated the interaction between Bcl-xL and IP 3 R. First, we compared the interaction of Bcl-xL and Bcl-2 with full-length IP 3 R. We overexpressed 3xFLAG-tagged Bcl-xL or Bcl-2 in HeLa cells expressing endogenous IP 3 Rs, immunoprecipitated Bcl-xL or Bcl-2 using anti-FLAG-coupled beads and immunoblotted for IP 3 Rs (Fig. 2a). This co-immunoprecipitation (coIP) analysis revealed that Bcl-xL could immunoprecipitate IP 3 Rs to a rather similar extend as Bcl-2, indicating that Bcl-xL and Bcl-2 display quite similar IP 3 R-binding properties. Bcl-2 can bind to the central, modulatory region of IP 3 R1, specifically to tryptic Fragment 3 [8,12,27]. We demonstrated that Bcl-2 binding to this site is associated with inhibition of IP 3 R1 activity. In addition to this, we recently discovered that Bcl-2 could also bind to the LBD of IP 3 R1, indicating that multiple regions are involved in IP 3 R1/Bcl-2 complex formation and inhibition of channel activity [11]. We also found that Bcl-xL could target this central, modulatory region of IP 3 R1 though with lower efficiency than Bcl-2 [12]. However, given the prominent inhibition of IP 3 Rs by Bcl-xL and the observation that this inhibition appears dependent on the agonist concentration, we asked whether Bcl-xL could also target the LBD. Thus, we used lysates from COS-7 cells that overexpressed 3xFLAG-Bcl-xL in GST-pull down experiments against purified GST-LBD and GST-Fragment 3 (representing a major part of the central, modulatory region) of IP 3 R1 (Fig. 2b). Our analysis revealed that similarly to Bcl-2, Bcl-xL can bind to both regions (Fig. 2c,  Fig. S2a). Since GST-pulldowns are only semi-quantitative, we applied microscale thermophoresis (MST), a biophysical approach allowing to measure molecular interactions. This technique is based on detecting a change in fluorescence of a labeled target as a function of the concentration of a non-fluorescent ligand. The change in fluorescence reflects the thermophoretic movement of the fluorescent target subjected to a microscopic temperature gradient. We thus used MST to assess direct binding between purified GST-IP 3 R fragments and purified 6xHis-Bcl-xL to determine the binding affinity. Using MST, we demonstrated that both purified GST-LBD and GST-Fragment 3 could bind to wild-type 6xHis-Bcl-xL (Fig. 2d). The specificity of this interaction is underpinned by two negative controls, parental GST and GST-Fragment 5b that lacks the 6 th TMD, previously established to be critical for Bcl-xL binding [12]. Indeed, no binding between 6xHis-Bcl-xL and GST or GST-Fragment 5b could be detected. Furthermore, we obtained the dissociation constant for both domains with 6xHis-Bcl-xL, revealing a K d of~701 nM for 6xHis-Bcl-xL interaction with the GST-LBD and a K d of~495 nM for 6xHis-Bcl-xL interaction with GST-Fragment 3. This indicates that wild-type 6xHis-Bcl-xL binds to both the LBD and Fragment 3.
We previously characterized the binding characteristics of the BH4 domain of Bcl-2 and Bcl-xL with Fragment 3 in detail via surface plasmon resonance (SPR) [27]. This study showed that the BH4 domain of Bcl-2 but not the one of Bcl-xL could interact with the Fragment 3. Recently, we also identified a novel binding site for Bcl-2's BH4 domain in the LBD, but had not yet characterized its ability to interact with Bcl-xL's BH4 domain [24]. Thus, we examined the importance of the BH4 domain of Bcl-xL for binding to the LBD using SPR. Biotin coupled to a peptide encompassing BH4-Bcl-xL was immobilized on streptavidin chips and different concentrations of purified LBD were applied as an analyte. Background binding was determined using a peptide with a scrambled sequence and subtracted. Biotin-BH4-Bcl-2 was used as a positive control for detecting LBD binding. The association curves for 1.1 µM GST-LBD show prominent binding to BH4-Bcl-2 while its binding to BH4-Bcl-xL is much lower (Fig. 2e). Similarly to what was observed for the binding of Fragment 3 to BH4-Bcl-2 versus BH4-Bcl-xL [27], GST-LBD displayed a strong concentration-dependent binding to immobilized BH4-Bcl-2 [11], while its binding to immobilized BH4-Bcl-xL appeared much weaker (Fig. 2f).
Taken together, these data reveal that while both Bcl-2 and Bcl-xL target the same regions in IP 3 R, they employ different binding determinants for these interactions. In contrast to Bcl-2, which exploits its BH4 domain for binding to LBD [24] and Fragment 3 [27], Bcl-xL seems to interact with the same IP 3 R regions but via motifs located outside of the BH4 domain.
Residue K87 of Bcl-xL contributes to the interaction with IP 3 R and particularly to the binding to LBD and Fragment 3 Our previously published results [27] and the data reported here (Fig. 2f) indicate that, in contrast to the BH4-Bcl-2, the BH4-Bcl-xL could not be responsible for targeting LBD and Fragment 3 of IP 3 R1. We therefore aimed to elucidate the molecular determinants in Bcl-xL responsible for its interaction with IP 3 Rs. Since Bcl-xL targets the same IP 3 R regions as Bcl-2, we envisioned that a similar interaction surface could underlie this phenomenon. We previously showed that K17 located in the middle of the BH4-Bcl-2 was critical for binding and inhibiting IP 3 Rs [27]. In the BH4-Bcl-xL, the corresponding residue is not a lysine but an aspartate, preventing its ability to bind to IP 3 Rs. However, a previously performed in silico Bcl-2/Bcl-xL structure superposition revealed that K87, located in the BH3 domain of Bcl-xL (BH3-Bcl-xL), likely is spatially constrained in a similar manner to K17 of BH4-Bcl-2 ( Fig. 3a) [28]. Moreover, sequence analysis of Bcl-xL orthologs among main vertebrate lineages revealed that K87 is highly Fig. 1 Bcl-xL overexpression inhibits IP 3 R-mediated Ca 2+ release in living cells. Ca 2+ signals were measured in Fura-2-loaded HEK cells expressing empty vector (pCMV24-P2A-mCherry; black), Bcl-2 (pCMV24-3xFLAG-Bcl-2-P2A-mCherry; orange) or Bcl-xL (pCMV24-3xFLAG-Bcl-xL-P2A-mCherry; green). EGTA (3 mM) was added to chelate extracellular Ca 2+ . IP 3 R-mediated Ca 2+ release was evoked by ATP (10 µM) (a-e) or carbachol (Cch, 10 µM) (f-j). Ionomycin (iono, 5 µM diluted in 250 mM CaCl 2 ) was added to assess the maximal Ca 2+ response. Representative single-cell Ca 2+ responses obtained from one well containing about 20-40 cells are shown (a-c, f-h). For each condition, six to nine different wells obtained from two to three independent transfections were monitored. Percentages of responding cells (d, i) and areas under the curve (e, j) were calculated from the Ca 2+ traces. Data are represented as mean of wells ± SD (N = 6-9), each data point represents one well. Statistically significant differences were determined using a one-way ANOVA (*P < 0.05).
conserved (Fig. 3b) and thus the importance of this residue was examined further. Interestingly, this lysine is located on the opposite side of the binding pocket involved in the interaction with Bak and Bax.
First, we used confocal microscopy to assess whether altering K87 into an aspartate affected Bcl-xL's subcellular localization. We transfected HeLa cells to express a mitochondria-or an ERtargeted RFP and compared the localization of 3xFLAG-Bcl-xL versus 3xFLAG-Bcl-xL K87D using anti-FLAG-based immunofluorescence ( Fig. S3a-d). We calculated average Pearson's coefficients above 0.75 for all conditions (Fig. S3e, f), indicating a high colocalization of Bcl-xL and Bcl-xL K87D with both the mitochondria and the ER. We also calculated average Manders' M1 coefficient to quantify the fraction of Bcl-xL or Bcl-xL K87D overlapping with the mitochondria, being 0.8 for both Bcl-xL and Bcl-xL K87D , or with the ER, being 0.6 for both Bcl-xL and Bcl-xL K87D (Fig. S3g, h). The similar coefficients calculated for the wild-type Bcl-xL and the Bcl-xL K87D indicate that the K87D mutation in Bcl-xL did not alter its subcellular localization.
Second, we tested the effect of the K87D mutation on the interaction of Bcl-xL with full-length IP 3 R. We therefore overexpressed 3xFLAG-tagged Bcl-xL or Bcl-xL K87D in HeLa cells expressing endogenous IP 3 Rs (Fig. 3c), immunoprecipitated Bcl-xL or Bcl-xL K87D using anti-FLAG-coupled beads and immunoblotted for IP 3 Rs. This co-immunoprecipitation analysis revealed that Bcl-xL K87D binding to the IP 3 R channel is severely impaired compared to wild-type Bcl-xL (Fig. 3c, d). We also immunoblotted for Bax to determine Bax binding to Bcl-xL or Bcl-xL K87D . We found that both Bcl-xL and Bcl-xL K87D could bind Bax, though Bax binding to Bcl-xL K87D appeared slightly reduced compared to its binding to Bcl-xL (Fig. 3c, e).
Third, we performed GST-pull down experiments with lysates from COS-7 cells overexpressing 3xFLAG-Bcl-xL or 3xFLAG-Bcl-xL K87D (Fig. 3f, Fig. S2b). We compared their binding to purified Fig. 2 Bcl-xL, but not its BH4 domain, binds to IP 3 R1 involving LBD and Fragment 3. a Representative co-immunoprecipitation experiments using anti-FLAG performed in lysates from HeLa cells transiently overexpressing 3xFLAG-Bcl-2 or 3xFLAG-Bcl-xL. This experiment was performed three times using each time independently transfected and freshly prepared cell lysates. The samples were analyzed via western blot using antibodies against IP 3 R1 and FLAG. Total HeLa lysates were used as input (20 µg). PD: pull down; IB: immunoblot. b Linear representation of a mouse IP 3 R1 (mIP 3 R1) monomer. The three functional domains, including the ligand-binding domain (LBD), and the five tryptic fragments, including Fragment 3, are represented. Respective amino acids are indicated by numbers. TMDs: transmembrane domains. c Representative GST-pull down experiment for assessing the binding of 3xFLAG-Bcl-xL from COS-7 cell lysate to GST-fused IP 3 R1 fragments. The samples were analyzed via western blot. Total COS-7 lysate was used as input (0.1 µg). This experiment was performed four times utilizing each time independently transfected and freshly prepared cell lysates. PD: pull down; IB: immunoblot. The corresponding western blot for the GST-IP 3 R fragments was shown in Fig. S2a. d Binding curves showing the interaction of purified 6xHis-Bcl-xL with titrated GST-fused IP 3 R domains generated by MST. GST was used as a negative control. Concentration of the 6xHis-Bcl-xL was kept constant at 50 nM, whereas the GST-LBD, GST-Fragment 3, GST-Fragment 5b and parental GST proteins were titrated down from 15 µM to 5 nM. The unit of the left axis (ΔF norm ) is a ratio of normalized fluorescence. Data points represent mean ± SD from triplicate measurements. e Representative sensorgrams of SPR experiments showing the binding properties of GST-fused IP 3 R-LBD, applied at 1.1 µM, to biotin-BH4-Bcl-xL and biotin-BH4-Bcl-2 peptides. The biotin-BH4 peptides, immobilized on a streptavidin-coated sensor chip. Sensorgrams were obtained after background correction for binding to the scrambled peptides. Data are expressed in resonance units (R.U.) as a function of time. f Quantitative analysis of the binding properties of biotin-BH4-Bcl-2 and biotin-BH4-Bcl-xL peptides to GST-LBD measured by SPR. Values obtained from independent experiments were plotted as mean ± SEM (N = 4).
GST-LBD and GST-Fragment 3 of IP 3 R1. In comparison to wild-type Bcl-xL, the ability of Bcl-xL K87D to bind the LBD and the Fragment 3 appears significantly reduced (Fig. 3f, g).
Finally, we applied an in cellulo mammalian protein-protein interaction trap (MAPPIT) assay, which is based on the functional complementation of cytokine receptor signaling [30]. The MAPPIT data confirmed that Bcl-xL is able to interact with Fragment 3 and that the interaction was impaired by the introduction of the K87D mutation ( Fig. 3i). No binding was detected with the negative control, indicating that the interaction is specific. In this assay, Bcl-xL binding to LBD could not be observed, potentially due to interference of the fusion protein to establish a functional recomplementation of the cytokine receptor.
Taken together, our data demonstrate that the K87 residue is crucial for the interaction of Bcl-xL with the IP 3 R, where it is involved in its binding to both the LBD and the Fragment 3.

K87 residue is critical for Bcl-xL-mediated IP 3 R inhibition in living cells
Next, we examined the role of the K87 residue in Bcl-xL-mediated IP 3 R inhibition. We used Fura-2-loaded COS-7 ( Fig. 4a-c) and HeLa ( Fig. 4d-g) cells with overexpressed mCherry and either Bcl-xL or Bcl-xL K87D . We used mCherry to identify transfected cells. We first studied the effect of Bcl-xL and Bcl-xL K87D overexpression in COS cells on IP 3 R-mediated Ca 2+ signals elicited by 500 nM ATP, a relatively high concentration provoking a Ca 2+ response in about 75% of the cells. Extracellular Ca 2+ was chelated with EGTA, so the reported Ca 2+ signals only originate from internal stores. Under these conditions, ATP-induced Ca 2+ signals appeared as a single transient (Fig. 4a). While about 75% of the cells expressing the empty vector displayed a response to ATP, only 40% of the cells expressing Bcl-xL responded (Fig. 4b). Cells expressing Bcl-xL K87D displayed similar responsiveness to ATP as empty vectorexpressing cells with about 75% responding cells. Quantification of the amplitude of the ATP-induced Ca 2+ transient in the responding cells yielded similar trends (Fig. 4c). Overexpression of Bcl-xL provoked a decrease in the peak [Ca 2+ ] provoked by ATP, while overexpression of Bcl-xL K87D failed to do this.
Then, we aimed to study the effect of Bcl-xL in HeLa cells, wellknown to display long-lasting Ca 2+ oscillations in response to extracellular agonists [31,32]. Here, we exposed HeLa cells to a very low [ATP] (70 nM), thereby mimicking basal, pro-survival Ca 2+ oscillations and enhancing the likelihood of observing different Ca 2+ -signaling patterns. We could discriminate three distinct Ca 2+ -signaling profiles: single peak responses, long-lasting responses and baseline Ca 2+ oscillations (Fig. 4d). Consistent with an inhibitory effect of Bcl-xL on IP 3 Rs, we found that long-lasting responses were clearly impaired upon overexpression of Bcl-xL ( Fig. 4d, e). Interestingly, this effect was not observed upon overexpression of Bcl-xL K87D . Furthermore, the area under the curve (Fig. 4f) and the peak amplitude ( Fig. 4g) were reduced upon overexpression of Bcl-xL, but not Bcl-xL K87D . This demonstrates that Bcl-xL's inhibitory action on IP 3 Rs is critically dependent on the K87 residue.
Finally, we also determined that overexpressed Bcl-xL or its mutant did not alter the ER Ca 2+ store content, by monitoring ER Ca 2+ release in HeLa cells following sarco/endoplasmic reticulum Ca 2+ ATPase (SERCA) inhibition by 1 µM thapsigargin (Fig. S5). These data are consistent with the differences observed upon ATP stimulation not being as a result of altered ER Ca 2+ levels, but instead are due to the specific effect of Bcl-xL on IP 3 R-mediated Ca 2+ release.
Purified Bcl-xL can directly suppress IP 3 R single-channel opening As all our functional data were obtained in intact cells, we also wished to provide more direct evidence for IP 3 R inhibition by Bcl-xL through electrophysiology. This is important because in intact cell systems Bcl-xL may have other targets besides IP 3 Rs that impact cytosolic Ca 2+ signals. In addition, these experiments can be performed in tightly controlled conditions, including different IP 3 and Bcl-xL concentrations. Therefore, we aimed to study the Fig. 3 The K87 residue is critical for Bcl-xL interaction with IP 3 R. a In silico representations of Bcl-2 and Bcl-xL three-dimensional structures. The lysine residues of interest (K17 in Bcl-2 and K87 in Bcl-xL) are indicated. Image taken from our previously published work [28]; this work is licensed under a Creative Commons Attribution 4. 0 International License. b Alignment of the conserved amino acid motifs for Bcl-xL's BH3 domain in vertebrates. The conserved lysine (K87 in human) is highlighted (red rectangle). The number of species used for each motif construction is shown in parentheses. "Z" means glutamic acid or glutamine. Top numbers represent amino acid numbers in human Bcl-xL sequence. c Representative coimmunoprecipitation (coIP) experiments using anti-FLAG performed in lysates from HeLa cells transiently overexpressing 3xFLAG-Bcl-xL or 3xFLAG-Bcl-xL K87D . The samples were analyzed via western blot using antibodies against IP 3 R1, FLAG and Bax. Total HeLa lysates were used as input (10 µg). PD: pull down; IB: immunoblot. The immunoreactive bands from independent coIP experiments, using each time independently transfected cells and freshly prepared lysates, were quantified and normalized to the binding of IP 3 R1 (d) and Bax (e) to 3xFLAG-Bcl-xL. Data represent mean ± SD (N = 5). Statistically significant differences were determined using a one-way ANOVA (*P < 0.05). f Representative GST-pull down experiment comparing the binding of 3xFLAG-Bcl-xL vs. 3xFLAG-Bcl-xL K87D from COS-7 cell lysate to purified GST-fused IP 3 R1 fragments and parental GST control. The samples were analyzed via western blot using anti-FLAG. Total COS-7 lysates were used as input (0.1 µg). The corresponding western blot for the GST-IP 3 R fragments is shown in Fig. S2b. g The immunoreactive bands from independent GST-pull down experiments, using each time independently transfected cells and freshly prepared lysates, were quantified and normalized to the binding of 3xFLAG-Bcl-xL and 3xFLAG-Bcl-xL K87D to parental GST control, which was set as 1 for each experiment. The data are plotted as mean ± SD (N = 5). Statistically significant differences were determined using paired t test (*P < 0.05). h Binding curves showing the interaction of purified 6xHis-Bcl-xL and 6xHis-Bcl-xL K87D with titrated GST-fused IP 3 R domains generated by MST. Concentration of the 6xHis-Bcl-xL and 6xHis-Bcl-xL K87D targets was kept constant at 50 nM, whereas the GST-LBD and GST-Fragment 3 were titrated down from 15 µM to 5 nM. The unit of the left axis (ΔF norm ) is a ratio of normalized fluorescence. The binding curves of wild-type 6xHis-Bcl-xL with GST-fused proteins are represented from Fig. 2d and are shown as reference. The binding curve of 6xHis-Bcl-xL K87D with parental GST is shown in Fig. S4. Data points represent mean ± SD from triplicate measurements. i Left: Representative example of a MAPPIT experiment. The binding is shown as fold induction, calculated by dividing the average luciferase activity of erythropoietinstimulated cells by the average obtained in non-stimulated cells. Binding of Bcl-xL, the Bcl-xL K87D mutant or irrelevant prey control (SV40 large T antigen) to the IP 3 R Fragment 3 and as negative control the bait vector without Fragment 3 are shown. Fold induction values at least four times higher than the irrelevant prey control are considered as bona fide protein-protein interactions. Values are represented as the mean of triplicates ± SEM within one representative experiment. The experiment was independently performed four times (N = 4). Statistically significant differences were determined using a one-way ANOVA (*P < 0.05). Right: Odyssey western blot analyses for the FLAG tag of the prey vector containing Bcl-xL or the Bcl-xL K87D mutant fusion proteins (green) or for β-actin (red).
impact of purified Bcl-xL proteins on IP 3 Rs. We generated 6xHistagged versions of full-length Bcl-xL, Bcl-xL ΔTMD and full-length Bcl-xL K87D that enabled their purification from E. coli using NiNTA columns (Fig. S6a).
Next, we tested the effect of the different recombinantly expressed and purified 6xHis-Bcl-xL variants on IP 3 R1 singlechannel activity using the on-nucleus patch-clamp technique (Fig. 5). Channel opening in isolated nuclei obtained from DT40-3KO cells ectopically expressing IP 3 R1 was triggered by 1 μM of IP 3 (Fig. 5a). We used purified Bcl-2 ΔTMD as a benchmark (Fig. 5b), which we previously validated to inhibit IP 3 R1 single-channel openings [24]. We subsequently first tested whether Bcl-xL ΔTMD could inhibit the opening of IP 3 R1 channels induced by 1 µM IP 3 , but this protein failed to modulate (inhibit/sensitize) IP 3 R1 channels (Fig. 5c). Next, we assessed the effect of 1 µM fulllength Bcl-xL (Fig. 5d), a concentration previously proposed to have a stimulatory effect on IP 3 R [26]. Consistent with the data obtained in intact cells and similarly to 1 µM Bcl-2 ΔTMD (Fig. 5b), application of 1 μM full-length 6xHis-Bcl-xL resulted in a significantly decreased open probability (P O ) of IP 3 R1 channels in the presence of 1 µM IP 3 (Fig. 5d). Clearly, these results contradict previous data that reported that Bcl-xL sensitizes IP 3 R [25,26]. In these reports, the effect of Bcl-xL on IP 3 Rs was shown to exhibit a bell-shaped dependence with 1 µM of Bcl-xL optimally sensitizing IP 3 Rs [26]. Hence, to ensure that we did not apply Bcl-xL at too high concentrations, we also examined the effects of 300 nM (Fig. 5e) and 100 nM (Fig. 5f) full-length Bcl-xL. These lower Bcl-xL concentrations also inhibited IP 3 R1 single-channel opening, though with lower potency compared to 1 µM full-length Bcl-xL. Next, we examined the effect of 1 µM full-length Bcl-xL K87D protein on IP 3 R1 single-channel openings activated by 1 µM IP 3 (Fig. 5g). Consistent with our in vitro binding assays and our Ca 2+ -imaging studies in intact cells, Bcl-xL K87D failed to inhibit IP 3 R1 singlechannel activity. Quantification of all conditions is shown in Fig. 4 Bcl-xL, but not Bcl-xL K87D , overexpression inhibits IP 3 R-mediated Ca 2+ release in single cells. Calcium measurements obtained from Fura-2-loaded HeLa (a-c) and COS-7 cells (d-g) transfected with a Bcl-xL (pCMV24-3xFLAG-Bcl-xL-P2A-mCherry; green) or Bcl-xL K87D -coding vector (pCMV24-3xFLAG-Bcl-xL K87D -P2A-mCherry; red), or an empty vector (pCMV24-P2A-mCherry; black). ER Ca 2+ response is elicited by addition of 70 nM (HeLa) or 500 nM (Cos-7) ATP, after addition of 3 mM EGTA to chelate extracellular Ca 2+ . Ionomycin (5 µM) diluted in 250 mM CaCl 2 was added at the end of the experiment (not shown) to trigger a high Ca 2+ release and confirm all the cells are equally loaded with Fura-2. For each condition, three to five wells were monitored and about 20-30 cells were analyzed by well. a Representative traces of Ca 2+ release in COS-7 cells. Traces represent mean ± SEM of one representative measurement (one well, about 20-30 cells). Percentage of responding cells (b) and maximal peak amplitude (c) were calculated for each condition. Data represent mean ± SD of four independent experiments (N = 4). Statistically significant differences were determined using a one-way ANOVA (*P < 0.05). Non-responding cells are defined as cells in which fluorescence signal measured after ATP stimulation do not exceed the maximal fluorescence value + SEM measured before stimulation. d Representative traces of Ca 2+ release in HeLa cells. Traces represent diverse Ca 2+ release patterns for one single cell. Distribution of typical Ca 2+ release patterns (e), areas under the curve (f) and amplitudes of the maximal Ca 2+ peak (g) were calculated from the Ca 2+ traces of the responding cells. Data represent mean ± SD of four independent experiments (N = 4). Statistically significant differences were determined using a one-way ANOVA (*P < 0.05).   5h. These data further demonstrate that Bcl-xL has a direct inhibitory effect on IP 3 R activity and that K87 is critical for this effect.
Another potential explanation could be that the conditions in which we have measured IP 3 R1 opening favor the detection of inhibitory effects and we may have missed potential sensitizing effects. We therefore measured the impact of purified Bcl-xL proteins on IP 3 R1 single-channel openings induced by threshold concentrations of IP 3 (Fig. 5i-m). In the presence of 100 nM IP 3 , the P O was reduced to~0.005 (Fig. 5i), compared to a P O of~0.25 at 1 µM IP 3 (Fig. 5a). Such conditions, which initiate threshold IP 3 R1 opening, should favor the detection of any potential sensitization of the channel. Nevertheless, similarly to Bcl-2 ΔTMD (Fig. 5j), fulllength Bcl-xL provoked a complete inhibition IP 3 R1 opening (Fig. 5k), while Bcl-xL K87D failed to inhibit IP 3 R1 opening (Fig. 5l). Quantification of all conditions is shown in Fig. 5m.
Finally, we also performed electrophysiology experiments to assess IP 3 R inhibition by 6xHis-Bcl-xL with high IP 3 concentrations (Fig. 5n-p). Indeed, we previously demonstrated that high IP 3 concentrations could compete with Bcl-2 for binding to the LBD of IP 3 Rs, thereby alleviating IP 3 R inhibition by Bcl-2 [24]. Here, we have established that 6xHis-Bcl-xL is also able to interact with the LBD of IP 3 Rs, prompting us to test the effect of purified 6xHis-Bcl-xL on IP 3 R1 single-channel activity triggered by high concentration of IP 3 . With 10 µM IP 3 , the P O reached more than 0.65 (Fig. 5n), compared to a P O of about 0.25 at 1 µM IP 3 (Fig. 5a). In those conditions, 6xHis-Bcl-xL did not alter channel activity, reflecting a loss of capacity to inhibit the channel at high [IP 3 ] (Fig. 5o, p). These results suggest that, similarly to our observations made for Bcl-2, IP 3 might compete with Bcl-xL for the LBD of IP 3 Rs, thereby rendering Bcl-xL less effective in inhibiting IP 3 Rs at high IP 3 concentrations.
Given that our findings were diametrically opposite to the previously reported findings, we sought to validate that the purified full-length 6xHis-Bcl-xL proteins used were properly folded and displayed bona fide anti-apoptotic functions. We first determined the CD spectrum of both Bcl-xL, which indicated that wild-type Bcl-xL, Bcl-xL K87D and Bcl-xL ΔTMD had a proper α-helical folding (Fig. S6b). Moreover, we performed a thermal ramping experiment. Unfolding of wild-type Bcl-xL was characterized by two apparent melting temperatures Tm1: 67°C and Tm2: 55.47°C, which were shifted to the left for Bcl-xL K87D (Tm1: 46.11°C and Tm2: 53.2°C), indicating some destabilizing effect of the mutation. These observations very much resemble the effect of K17D mutation in purified Bcl-2 [27]. We also measured Bcl-xL ΔTMD , which was characterized by one melting temperature Tm1: 76°C, indicating that Bcl-xL ΔTMD is much more stable than wild-type Bcl-xL (Fig. S6c). Next, we employed an in vitro Bax-liposome permeabilization assay, where purified Bax is incubated with liposomes encapsulating both a quencher (DPX) and a fluorophore (ANTS) (Fig. 6). Bax-pore formation can be triggered by cBid (Fig. 6a) or Bim (Fig. 6b) proteins, two "activator" BH3-only proteins. Full-length 6xHis-Bcl-xL potently inhibited cBid and Bimtriggered Bax-pore formation (IC 50 of about 20 nM). Of note, 6xHis-Bcl-xL K87D too inhibited Bax-pore formation, but was less efficacious (IC 50 of about 80 nM) than 6xHis-Bcl-xL (Fig. 6c, d). This might relate to the reduced Bax binding observed in the coIPs using cell lysates (Fig. 3c). Consistent with previous reports [33], Bcl-xL ΔTMD failed to inhibit Bax-pore formation.
Overall, our electrophysiological studies provide strong evidence that recombinant Bcl-xL with validated anti-apoptotic properties directly inhibits IP 3 R1 single-channel opening with a critical role for K87 in Bcl-xL. Furthermore, the data suggest that Bcl-xL's TMD is not only important for inhibiting Bax [33,34] but also for inhibiting IP 3 R opening. Yet, the significance and the role of the TMD of Bcl-xL in a cellular context for IP 3 R modulation remains to be elucidated.

Bcl-xL K87D is impaired in protecting cells against staurosporine-induced apoptosis
Next, we wished to validate the importance of the IP 3 R/Bcl-xL interaction for the protective effects of Bcl-xL against Ca 2+dependent pro-apoptotic stimuli (Fig. 7, Fig. S7). We therefore used STS, which has been previously validated to provoke Ca 2+ - driven apoptosis [35][36][37]. Here, we assessed whether STS provoked apoptosis through IP 3 R-mediated Ca 2+ elevations. First, we measured long-term Ca 2+ dynamics in HeLa cells exposed to 0.5 µM STS for 1 hour (Fig. 7a). Using live single-cell Ca 2+ imaging in Fura-2-AM-loaded cells, we observed that STS triggered longlasting Ca 2+ elevations in wild-type HeLa cells. Contrary to IP 3 R activation with physiological agonists (Fig. 1, Fig. 4), this Ca 2+ release is rather slow on onset and prolonged over a long period of time. We then used a HeLa cell model in which all three IP 3 R isoforms have been knocked out (HeLa-3KO). In these cells, the STS-induced Ca 2+ -release events were virtually absent (Fig. 7a, b). Having validated that STS treatment in HeLa cells provoked longlasting IP 3 R-mediated Ca 2+ elevations, we determined whether IP 3 Rs contributed to STS-induced cell death in HeLa cells (Fig. 7c). We therefore monitored apoptotic cell death in HeLa cells exposed to 0.5 µM STS for six hours by determining the ratio of cleaved poly(ADP-ribose) polymerase (PARP) in relation to total PARP [38]. Strikingly, in wild-type HeLa, about 90% of the total PARP was converted to the cleaved form, while only 30% of the total PARP appeared in the cleaved form in HeLa-3KO cells, Fig. 7 Wild-type Bcl-xL, but not Bcl-xL K87D , protects HeLa cells against IP 3 R/Ca 2+ -driven cell death using staurosporine. a, b Ca 2+ measurements in Fura-2-loaded wild-type HeLa (black) and HeLa-3KO cells (blue). Cells were exposed to 0.5 µM staurosporine (STS), after addition of 3 mM EGTA to chelate extracellular Ca 2+ (not shown). Representative traces of Ca 2+ release are shown (a), along with areas under the curve (b) calculated for one hour following STS addition. For each experiment (N = 2), two wells were monitored per condition and about 20-30 cells were analyzed by well. Each trace and each point represent one cell. Statistically significant differences were determined using a t test (*P < 0.05). c, d Wild-type HeLa and HeLa-3KO cells were treated with 0.5 µM STS or vehicle (DMSO) for 6 h. The samples were analyzed via western blot (IB: immunoblot). Representative western blots assessing uncleaved (top band) and cleaved PARP (lower band) as well as vinculin (c). The immunoreactive bands from independent experiments, using each time freshly prepared cell lysates, were quantified (d) and PARP cleavage was calculated as the ratio of cleaved PARP over total PARP. The data are plotted as mean ± SD (N = 6). Statistically significant differences between the "+STS" conditions were determined using a t test (paired, two-tailed, *P < 0.05). e, f Ca 2+ signals were measured in Fura-2-loaded wild type HeLa expressing empty vector (pCMV24-P2A-mCherry; black), Bcl-xL (pCMV24-3xFLAG-Bcl-xL-P2A-mCherry; green) or Bcl-xL K87D (pCMV24-3xFLAG-Bcl-xL K87D -P2A-mCherry; red). ER Ca 2+ release is triggered as in a. The traces represent the average response of all cells ± SEM in one well containing about 20 cells (e). The individual Ca 2+ traces are shown in Fig. S7a. For each condition, 1 to 2 independent wells obtained from 4 different transfections were monitored. The areas under the curve were calculated for all individual cell during the 45 min following STS addition (f). Data are represented as mean of wells ± SD (N = 5 to 7), each data point represents one well. Statistically significant differences were determined using a one-way ANOVA (*P < 0.05). Another graphical representation is shown in Fig. S7b. g-j Wild type HeLa (g, h) and HeLa-3KO cells (i, j) transiently overexpressing 3xFLAG-Bcl-xL or 3xFLAG-Bcl-xL K87D were treated with 0.5 µM STS or vehicle (DMSO) for 6 h. The samples were analyzed via western blot. Representative western blots assessing uncleaved and cleaved PARP as well as total Bcl-xL (endogenous + overexpressed) and β-actin (g, i). The vertical line in panel g indicates that two different parts of the same gel and exposure time were merged together. The original uncropped picture is shown in Fig. S7c. The immunoreactive bands from independent experiments, using each time independently transfected and treated cells and freshly prepared cell lysates, were quantified (h, j) and PARP cleavage was calculated like in d. The data are plotted as mean ± SD (N = 5). Statistically significant differences between the "+STS" conditions were determined using a one-way ANOVA (*P < 0.05). k, l Wild-type HeLa and HeLa-3KO cells transiently overexpressing 3xFLAG-Bcl-xL or 3xFLAG-Bcl-xL K87D were treated with 25 µM venetoclax or vehicle (DMSO) for 24 h. The samples were analyzed via western blot. Representative western blots assessing uncleaved (top band) and cleaved PARP (lower band), as well as overexpressed Bcl-xL (FLAG) and βactin (k). The immunoreactive bands from independent experiments, using each time independently transfected and treated cells and freshly prepared cell lysates, were quantified and PARP cleavage was calculated as in d (l). The data are plotted as mean ± SD (N = 5). Statistically significant differences were determined using a one-way ANOVA (*P < 0.05).
indicating that IP 3 Rs are crucial for STS-induced cell death in HeLa cells (Fig. 7c, d). In a more general way, this is the first time that, to our knowledge, that IP 3 Rs were directly implicated in STS-evoked pro-apoptotic Ca 2+ flux and directly linked to cell death.
Next, using live, single-cell Ca 2+ imaging, we studied the impact of overexpressing Bcl-xL-P2A-mCherry and Bcl-xL K87D -P2A-mCherry on STS-induced Ca 2+ elevations in Fura-2-AM-loaded wild-type HeLa cells (Fig. 7e, Fig. S7a). Ca 2+ signals were measured in mCherry-positive cells. Strikingly, Bcl-xL overexpression strongly suppressed prolonged Ca 2+ elevations induced by 0.5 µM STS compared to empty vector-expressing cells, while Bcl-xL K87D overexpression was much less effective (Fig. 7e, f, Fig. S7a, b). We then examined whether IP 3 R modulation by Bcl-xL contributed to the anti-apoptotic action of Bcl-xL (Fig. 7g). We first confirmed that the transfection of the cells with the 3xFLAG plasmids did not provoke cell death by itself (Fig. 7g, "-STS" conditions). Bcl-xL overexpression strongly suppressed PARP cleavage in wild-type HeLa cells exposed to 0.5 µM STS for 6 h compared to empty vector-expressing cells (Fig. 7g, "+ STS" conditions). In contrast, Bcl-xL K87D overexpression was much less effective than wild-type Bcl-xL in suppressing PARP cleavage in wild-type HeLa cells (Fig. 7g, h). This suggests that Bcl-xL protects against STS through inhibition of IP 3 Rs, since Bcl-xL K87D is much less efficient in doing so.
We then focused on IP 3 R-independent cell death mechanisms. We have shown that Bcl-xL K87D binding to Bax appeared to be somewhat impaired compared to wild-type (Fig. 3c). Furthermore, the ability of Bcl-xL K87D to neutralize Bax pore formation also appeared attenuated (Fig. 6). Since STS partially acts independently of IP 3 Rs and since PARP cleavage also occurs in HeLa-3KO cells, though to a lesser extent (Fig. 7d), we wanted to discriminate Bcl-xL anti-apoptotic effect between IP 3 R inhibition versus IP 3 R-independent processes, such as Bax inhibition. Hence, we examined the effect of Bcl-xL and Bcl-xL K87D overexpression on STS-induced cell death in HeLa-3KO cells (Fig. 7i, j). Consistent with the ability of Bcl-xL to bind and neutralize Bax, we found that Bcl-xL could suppress STS-induced PARP cleavage in HeLa-3KO cells.
Of particular interest and in contrast to the results obtained in wild-type HeLa cells, Bcl-xL K87D was equally effective as wild-type Bcl-xL in dampening STS-induced PARP cleavage in HeLa cells lacking IP 3 Rs. This implies that although Bax binding/inhibition is somewhat affected by the K87D mutation in Bcl-xL, there is sufficient residual Bax-binding and -inhibition capacity of Bcl-xL K87D to prevent cell death in cellulo.
We wished to further validate that the K87D mutation does not affect the protection afforded by Bcl-xL towards IP 3 R-independent cell death triggers. Hence, we chose the BH3 mimetic venetoclax/ ABT-199, a selective Bcl-2 inhibitor [16] previously established to neither interfere with the ability of Bcl-2 to inhibit IP 3 Rs nor to alter Ca 2+ signaling [39,40]. Venetoclax (25 µM; 24 h) triggered~80% PARP cleavage in HeLa cells (Fig. 7k, l). The level of PARP cleavage was similar between wild-type HeLa and HeLa-3KO, thereby validating that venetoclax indeed acted in an IP 3 R-independent manner. Bcl-xL and Bcl-xL K87D were equally effective in counteracting venetoclax-induced PARP cleavage by about 40-50% (Fig. 7k, l). Moreover, the anti-apoptotic effect of Bcl-xL and Bcl-xL K87D was also comparable between wild-type HeLa and HeLa-3KO. These data strongly indicate that K87D mutation impairs Bcl-xL's protective effect against IP 3 R-dependent cell death but does not significantly affects its canonical anti-apoptotic function, thereby antagonizing Bax/Bak.

Bcl-xL protects breast cancer cells from IP 3 R-mediated cell death
Finally, by knocking down Bcl-xL in a Bcl-xL-dependent cell model, we examined whether also endogenous Bcl-xL could inhibit IP 3 Rs. We used a breast cancer model, the mammary gland adenocarcinoma cell line MDA-MB-231, in which Bcl-xL is important for survival [41] and migration [42]. We transfected MDA-MB-231 cells with a siRNA targeting Bcl-xL, thereby lowering its protein levels by about 50% (Fig. 8a, b). Interestingly, Bcl-xL knock-down in MDA-MB-231 cells did not induce apoptosis by itself (Fig. 8a, c). This was important to exclude that any potential changes in Ca 2+ signaling in cells with decreased Bcl-xL levels were a consequence of ongoing cell death rather than due to a decrease in Bcl-xLprotein levels. Using thapsigargin, we next validated in MDA-MB-231 cells that the ER Ca 2+ -store content is not altered following Bcl-xL depletion (Fig. 8d, e, f). Therefore, changes in agonistinduced Ca 2+ signaling in Bcl-xL-depleted cells are not an indirect consequence of changes in ER Ca 2+ loading. We then measured IP 3 R-mediated Ca 2+ release elicited by ATP (0.5 µM) in individual MDA-MB-231 cells pre-treated with extracellular Ca 2+ chelator EGTA, thereby ensuring that Ca 2+ signals only arise from internal stores. Compared to the cells transfected with a non-target siRNA, the cells treated with a siRNA targeting Bcl-xL displayed a strikingly higher ATP-induced Ca 2+ response (Fig. 8g, h). We calculated a significant increase in the number of responding cells (Fig. 8i), the area under the curve (Fig. 8j) and in the maximal peak amplitude (Fig. 8k) in MDA-MB-231 cells in which Bcl-xL-protein levels were lowered. To be certain that this effect was not due to a potential downregulation of the Bcl-xL-related Bcl-2 protein, which is a prominent inhibitor of IP 3 Rs, we analyzed the Bcl-2protein levels via western blotting (Fig. S8). However, Bcl-2-protein levels were not decreased. Instead, Bcl-2-protein levels appeared increased, potentially as a compensatory mechanism that could help sustain the survival of the cells in which Bcl-xL was downregulated. Nevertheless, the overall Bcl-2-protein levels remained extremely low in these MDA-MB-231 cells, when benchmarked against the Bcl-2-protein levels present in OCI-LY-1 cells, a Bcl-2 dependent diffuse large B-cell lymphoma cell line. In any case, these data indicate that endogenous Bcl-xL suppresses IP 3 R activity in breast cancer cells, independently of Bcl-2 levels. To determine whether Bcl-xL could also counteract IP 3 R-mediated apoptotic Ca 2+ release in those cells, we exposed the MDA-MB-231 cells to STS (0.5 µM) (Fig. 8l). In MDA-MB-231 cells transfected with a non-target siRNA, STS only provoked limited PARP cleavage, indicating that these cells are rather resistant to STS. However, cells treated with the siRNA against Bcl-xL displayed a prominent increase in STS-induced PARP cleavage resulting in about 50% PARP cleavage. This indicates that lowering endogenous Bcl-xL-protein levels rendered MDA-MB-231 cells very sensitive to STS-induced cell death (Fig. 8l, m). Altogether, these results reveal that endogenous Bcl-xL suppresses IP 3 R-mediated Ca 2+ release and confers cell death protection against Ca 2+ -dependent cell death stimuli.

DISCUSSION
The main finding of this study is that the anti-apoptotic Bcl-xL protein functions as an inhibitor of IP 3 R channels both in intact living cells and at the single-channel level. These data challenge the presumed role of Bcl-xL as an IP 3 R-sensitizing protein [25,26,43]. This supposition is strongly supported by several independent lines of evidence. Molecular studies reveal that Bcl-xL targets the same regions in IP 3 Rs as Bcl-2 (e.g., LBD and Fragment 3), which are responsible for inhibition of IP 3 Rs. Further, we demonstrate that Bcl-xL, in a similar fashion to Bcl-2, possesses a lysine residue that is critical for IP 3 R binding and inhibition. The critical lysine identified in Bcl-xL (K87) spatially resembles and substitutes for the previously identified critical lysine in Bcl-2 (K17) [27]. Mutation of K87 abrogates the ability of Bcl-xL to bind and inhibit IP 3 Rs. The findings are further underpinned by singlechannel recordings of IP 3 R1 channels, whose open probability is reduced upon exposure to purified Bcl-xL, but not Bcl-xL K87D . Moreover, by inhibiting IP 3 Rs, K87 in Bcl-xL is important for Bcl-xL's ability to protect cells against STS, a stimulus that triggers apoptosis in an IP 3 R/Ca 2+ -dependent manner. Finally, we demonstrate that also in MDA-MB-231, a Bcl-xL-dependent breast cancer cell model, endogenous Bcl-xL inhibits IP 3 R function.
Over the past two decades, several, mainly anti-apoptotic, members of the Bcl-2-protein family, have emerged as critical modulators of Ca 2+ homeostasis and dynamics [21]. The two most studied proteins are Bcl-2 and Bcl-xL, which are consistently reported to be localized in the ER and to control the flux through ER-resident Ca 2+ -release channels [20]. While reports suggest that Bcl-2 may also lower ER Ca 2+ -store content and thus the likelihood for pro-apoptotic Ca 2+ transfer to mitochondria [19,44,45], other evidence has emerged that anti-apoptotic Bcl-2 is a direct inhibitor of Ca 2+ flux through IP 3 Rs without markedly affecting the ER Ca 2+ -store content [46]. This results from Bcl-2 binding to IP 3 Rs [8]. Subsequent work revealed the interaction domains in both IP 3 Rs and Bcl-2 that are responsible for the complex formation. In Bcl-2, we identified the BH4 domain [23] and the C-terminal TMD [11] as critical for IP 3 R inhibition in intact cells. For IP 3 Rs, we found that the LBD [24], a stretch of 20 a.a. in the central, modulatory domain [8] and the C-terminus of IP 3 Rs [11] participate in Bcl-2 binding. Importantly, the hydrophobic cleft of Bcl-2 is not necessary for IP 3 R modulation [11,39] and, as a result, BH3 mimetic drugs do not impact IP 3 R modulation by Bcl-2. In various cancer cell models, disrupting the complex between IP 3 Rs and Bcl-2 was even sufficient to provoke cell death through Ca 2+ overload [47][48][49]. Our current model is that Bcl-2 acting via its BH4 domain inhibits IP 3 Rs by targeting the LBD and the central, modulatory domain. The occurrence of inhibition is aided by the "effective concentration" of Bcl-2 in the close proximity of IP 3 Rs as a result of the interaction between the C-terminal regions of both proteins.
Our new data demonstrate that, similarly to Bcl-2, Bcl-xL inhibits IP 3 R-mediated Ca 2+ release by targeting precisely the same regions as the BH4 domain of Bcl-2, namely the ligand-binding domain and the central, modulatory domain. This challenges the currently accepted concept that Bcl-xL sensitizes IP 3 Rs [25,26,43]. We have previously proposed that the distinct modulation of IP 3 R by Bcl-2 and Bcl-xL could be due to differences in their BH4 domain [27]. Yet, in silico superposition of Bcl-2 and Bcl-xL indicated that the K17 residue critical for Bcl-2's BH4 domain spatially resembled K87 of Bcl-xL [28]. Our functional results prompted us to revisit the modulation of IP 3 Rs by Bcl-xL, revealing that Bcl-xL inhibited IP 3 Rs through an interaction mediated by K87, an evolutionary conserved residue located in the BH3 domain of Bcl-xL. Moreover, similarly to Bcl-2, inhibition of IP 3 Rs by Bcl-xL was dependent on the agonist/IP 3 concentration, whereby high agonist/IP 3 concentrations abrogated the inhibitory effect of Bcl-xL on IP 3 Rs. This is consistent with our molecular studies showing that Bcl-xL can target the LBD, the region where IP 3 binds. Thus, similarly to Bcl-2 [24], Bcl-xL binding to LBD might be antagonized by IP 3 .
This interaction also accounts for Bcl-xL's protective effect against IP 3 R-mediated, Ca 2+ -driven apoptosis by using STS. Exploiting wild-type versus IP 3 R-knockout HeLa cells, we demonstrated that STS triggered long-lasting Ca 2+ rises that depended on IP 3 Rs and that STS-induced cell death was for a large part driven by IP 3 Rs, though not exclusively. Bcl-xL overexpression suppressed STS-induced Ca 2+ rises and cell death. We also observed that mutating K87 into aspartic acid in Bcl-xL also mildly impacted the ability of Bcl-xL to bind Bax and the potency of Bcl-xL to prevent Bax-pore formation. Therefore, we wanted to exclude that the impaired protection against STSinduced cell death could be due to the slightly weakened Baxbinding properties of Bcl-xL K87D . We therefore used the HeLa-3KO cells to exclude any IP 3 R-independent mechanism, revealing that wild-type Bcl-xL and Bcl-xL K87D were equally potent in protecting HeLa cells that lacked IP 3 Rs against STS. In addition, we found that Bcl-xL and Bcl-xL K87D were equally effective in protecting cells against IP 3 R-independent cell death stimuli, such as venetoclax [11,39,50]. Hence, this demonstrates that reduced anti-apoptotic properties of Bcl-xL K87D are related to reduced inhibition of IP 3 Rs rather than non-IP 3 R-related targets such as Bax or Bak.
Our understanding of the role of Bcl-xL in Ca 2+ signaling has been shaped by previous studies from the Foskett lab [25,26,43]. Therefore, the previous model is that antiapoptotic Bcl-xL proteins enhance IP 3 R-mediated Ca 2+ release by sensitizing the channels to IP 3 . In support of these ideas, Bcl-xL promoted IP 3 R-driven Ca 2+ oscillations to drive mitochondrial bio-energetics and ATP production [25,43]. In contrast to Bcl-2, Bcl-xL was proposed to bind IP 3 Rs via its hydrophobic cleft responsible for scaffolding pro-apoptotic Bcl-2-family members [26]. Thus, BH3-mimetic Bcl-xL inhibitors interfere with the ability of Bcl-xL to modulate IP 3 Rs. Moreover, the effect of Bcl-xL on IP 3 Rs was concentration-dependent with an optimal IP 3 R-sensitizing effect observed at 1 µM Bcl-xL protein. In contrast, our present data demonstrate that Bcl-xL inhibits, rather than sensitizes, IP 3 R channels. Furthermore, our live single-cell measurements show that Bcl-xL suppressed Ca 2+ signals, even when induced by very low concentrations of agonist. By doing so, Bcl-xL seems to shift the profile of Ca 2+ signals from long-lasting responses towards transient peaks and spontaneous oscillations (Fig. 4e). Thus, while Bcl-xL indeed increases the number of cells displaying Ca 2+ oscillations, we argue that this is due to IP 3 R inhibition.
The reason for the discrepancy with the earlier studies is not clear, but we can exclude a number of obvious factors. First, we exclude that differences might be attributed to different cell models (here: HEK-293 and HeLa cells; [25]: DT40 cells). In the present study, we also used permeabilized DT40 cells in our electrophysiology experiments and the results we obtained were consistent with the experiments we performed in intact HEK-293 and HeLa cells. Furthermore, by sensitizing IP 3 Rs, Bcl-xL was also reported to lower ER Ca 2+ levels in DT40 cells [51]. However, overexpressing Bcl-xL in wild-type DT40 cells did not lower the thapsigargin-induced Ca 2+ release, indicating that in our hands Bcl-xL did not affect ER Ca 2+ -stores in these cell models (Fig. S9). We thus assume that the discrepancy with earlier studies are not Fig. 8 Knocking down Bcl-xL in the Bcl-xL-dependent MDA-MB-231 breast cancer cell line enhances IP 3 R-mediated Ca 2+ release and apoptosis. MDA-MB-231 cells were transfected with either a siRNA targeting Bcl-xL (siBcl-xL, purple) or a non-target siRNA (siCtrl, black). 48 hours later, the cells were used for experiments. a-c Transfected cells were lysed and proteins were analyzed via western blot (IB: immunoblot). Representative western blots assessing total PARP, Bcl-xL and β-actin (a). Quantifications from independent experiments, using each time independently transfected cells and freshly prepared cell lysates, are shown in b and c. Data are represented as mean ± SD (N = 12). Statistically significant differences were determined using a t-test (unpaired, two-tailed, *P < 0.05). d-k Ca 2+ signals obtained from Fura-2-loaded MDA-MB-231 cells. Representative traces of single cell Ca 2+ release are shown (d, e, g and h). ER Ca 2+ content was determined by quantifying the thapsigargin (2 µM)-releasable Ca 2+ in the presence of EGTA (3 mM) (d-f). IP 3 R-mediated Ca 2+ release was determined by quantifying ATP (0.5 µM)evoked Ca 2+ release in the presence of EGTA (3 mM) (g-k). Ionomycin (5 µM) diluted in 250 mM CaCl 2 was added at the end of the experiment (iono.) to validate Fura-2 loading. Representative single-cell Ca 2+ responses obtained from one well containing about 40-60 cells are shown. For each condition, 8 to 10 independent wells obtained from 3 different transfections were monitored. Area under the curves (f, j), percentages of responding cells (i), and maximal peak amplitudes (k) were calculated for each condition. Data represent mean of wells ± SD (N = 8 to 10), each point represents one well. Statistically significant differences were determined using a t-test (unpaired, two-tailed, *P < 0.05). Transfected cells were treated with 0.5 µM staurosporine (STS) for 6 h or with vehicle (DMSO). The cells were then lysed and proteins were analyzed via western blot. Representative western blots assessing uncleaved (top band) and cleaved PARP (lower band) as well as Bcl-xL and β-actin (l). Quantifications from independent experiments, using each time independently transfected and treated cells and freshly prepared cell lysates, are shown in m. Data are represented as mean ± SD (N = 6). Statistically significant differences were determined using a t-test (unpaired, two-tailed, *P < 0.05).
related to different cell models. Second, we also controlled some other factors and validated that they can also be ruled out. i. It could be argued that in the present study, very high Bcl-xL levels were used, or that experimental conditions were not favorable to observe sensitization of Ca 2+ release. However, in our singlechannel recordings, we applied 1 µM Bcl-xL, a concentration previously reported to maximally cause IP 3 R sensitization [26]. Nevertheless, we also used even lower Bcl-xL concentrations (100-300 nM), which also inhibited IP 3 R1 single-channel openings.
ii. In intact cell experiments and in IP 3 R1 single-channel recordings, we also used low concentrations of extracellular agonist and IP 3 , which should prime the system to observe IP 3 R sensitization. iii. We validated that both overexpressed Bcl-xL in living cells and purified Bcl-xL proteins used in this study are bona fide anti-apoptotic proteins and can exert anti-apoptotic functions. iv. We mainly focused on IP 3 R1 in this study as it was the IP 3 R isoform that was analyzed in depth in the original reports [25,43]. We have not formally ruled out IP 3 R isoform-dependence in the inhibitory effect of Bcl-xL. Nevertheless, previous work indicated that all three IP 3 R isoforms could bind and were similarly sensitized by Bcl-xL in a similar fashion [25,43]. Future work will determine whether Bcl-xL differentially impacts IP 3 R1, IP 3 R2 and IP 3 R3 channels.
At the cellular level, Bcl-xL has been shown, beyond its canonical anti-apoptotic activity, to favor cell survival by enhancing mitochondrial metabolism. For instance, Bcl-xL can interact with and promote activity of the F-type ATPase [52]. Yet, in breast cancer cells, Bcl-xL improves the metabolic capacities by more efficiently coupling the mitochondrial proton motive force with ATP production [53]. Although the involvement of Ca 2 + signaling in those processes is still unknown, sensitization of IP 3 R by Bcl-xL has been shown to optimize mitochondrial bioenergetics, which may relate to Bcl-xL's ability to promote Ca 2+ flux to mitochondria [25,54]. But if Bcl-xL does not sensitize IP 3 Rs, then how does Bcl-xL promote mitochondrial bioenergetics? In our work, we demonstrate that Bcl-xL inhibits Ca 2+ release from the ER to the cytosol in various cell systems, including in breast cancer cells, thereby protecting the cells from IP 3 R-mediated apoptosis. Furthermore, our group has previously established that Bcl-xL inhibits the voltagedependent anion channel 1 (VDAC1) [55]. Recently, Bcl-xL has been reported to dampen VDAC1-mediated mitochondrial Ca 2+ uptake in breast cancer cells [56]. This mechanism has been proposed to alter mitochondrial ATP generation and increase ROS production, thereby promoting breast cancer cell migration [42]. Since Ca 2+ transfers between the ER and the mitochondria are tightly connected, we speculate that Bcl-xL could inhibit both VDAC1 and IP 3 Rs in breast cancer cells to promote cancer malignant features.
Finally, since the interaction profile of Bcl-2 and Bcl-xL for IP 3 R binding is very similar, it is possible that peptides similar to those disrupting IP 3 R/Bcl-2 complexes, such as the Bcl-2/IP 3 R disruptor 2 (BIRD-2) [57], can affect IP 3 R/Bcl-xL complexes. Disrupting such IP 3 R/Bcl-xL complexes could therefore result in Ca 2+ -driven cell death, as observed in several Bcl-2-dependent cancer in which Bcl-2 was displaced from IP 3 Rs [22,58] or antagonize breast cancer cell migration, a process controlled by Bcl-xL at the level of the IP 3 R [42].
Overall, this work reassesses the model and mechanism of antiapoptotic action of Ca 2+ signaling events modulated by Bcl-xL. In contrast to the previous model, we argue that Bcl-xL, in a similar manner to Bcl-2, inhibits IP 3 Rs and thereby can protect cells against apoptosis. Bcl-xL not only phenocopies Bcl-2 at the functional level, but also at the molecular level. This in-depth understanding of the similarities and differences in the mechanism of interaction and action of distinct anti-apoptotic Bcl-2 family members may ultimately be exploited for the design of novel therapeutics modulating apoptosis.
For 6xHis plasmids purification, cDNAs sequences coding for Bcl-2 and Bcl-xL were cloned in pET45b plasmids. Full-length or truncated sequences were inserted in the 6xHis-encoding reading frame. The Bcl-xL ΔTMD is deleted of amino acids R209 to K233 and the Bcl-2 ΔTMD is deleted of amino acids L217 to K239. Following cloning of the pET45b-Bcl-xL, the K87D mutant was obtained by PCR site-directed mutagenesis as previously described [28]. All constructs were verified by sequencing (LGC Genomics, Berlin, Germany). Proteins were then purified from BL21 Escherichia coli as described before [24].
For GST fusion proteins purification, BL21 Escherichia coli were transformed and amplified as described before [62], and proteins were purified as previously described [63].
Concentration of the purified proteins was determined using BCA Protein Assay Reagent (Thermo Fisher Scientific, Merelbeke, Belgium). The purity was examined by SDS-PAGE and Coomassie blue staining of the gels with the Imperial Protein Stain reagent (ThermoFisher Scientific). Quality and integrity of the proteins were confirmed by immunoblotting with anti-GST (Cell Signaling Technology, Leiden, Netherlands; #2622) and anti Bcl-xL (Cell Signaling Technology; #2764) antibodies. Western blots were performed as previously described [59].
Bax, Bim and cBid proteins were purified from BL21 or DH5α Escherichia coli as extensively described before [64].
Single-cell Ca 2+ imaging 20,000 HeLa, 50,000 HEK-293 or 50 000 MDA-MB-231 cells were plated in four-chamber 35-mm dishes. 24 h after seeding, cells were transiently transfected with 0.25 µg of pCMV24-P2A-mCherry, pCMV24-3xFLAG-Bcl-2-P2A-mCherry, pCMV24-3xFLAG-Bcl-xL-P2A-mCherry or pCMV24-3xFLAG-Bcl-xL K87D -P2A-mCherry constructs. X-tremeGene HP DNA (Roche) was used as a transfection reagent according to the manufacturer's instructions. 48 h after transfection, the cells were loaded with Fura-2-AM (AnaSpec) or Fluo-4-AM (Invitrogen) and Ca 2+ imaging was performed using an Axio Observer Z1 fluorescent microscope (Zeiss, Jena, Germany) as previously described [27]. Data were plotted either as F 340 /F 380 ratio, [Ca 2+ ], or as normalized (F − F 0 )/F 0 or F/F 0 whereby F = F 340 /F 380 at different time points and F 0 = F 340 /F 380 at the start of the experiment. EGTA (Sigma-Aldrich) was used to chelate extracellular Ca 2+ . ATP (Sigma-Aldrich) and carbachol (Sigma-Aldrich) were used to elicit IP 3 R-mediated Ca 2+ release. Thapsigargin (Alomone Labs, Jerusalem, Israel), an irreversible inhibitor of sarco/endoplasmic reticulum Ca 2+ ATPases, was used to assess ER Ca 2+ -store content. Ionomycin (Alomone Labs), a Ca 2+ ionophore, was used to assess the total intracellular Ca 2+ and validate adequate loading of fluorescent Ca 2+ indicators. Ca 2+ traces were analyzed with the Excel (Microsoft, Redmond, WA, USA) and Prism (GraphPad, San Diego, CA, USA) softwares. The number of non-responding cells was determined, whereby non-responding cells were defined as cells in which the maximal fluorescence signal measured after agonist stimulation do not exceed the baseline value + SEM. The Δ maximal amplitude and area under the curve were analyzed in the responding cells. The baseline was calculated as the average fluorescence between EGTA and agonist addition. The Δ maximal amplitude was calculated by subtracting the baseline from the maximal response value to the agonist. The area under the curve was calculated by integrating the responses to agonist after subtracting the baseline.

Surface plasmon resonance
The following peptides (purity >80%) were obtained from LifeTein (South Plainfield, NJ, USA) and dissolved in dimethyl sulfoxide to prepare 10 mM stock solutions.

Mammalian protein-protein interaction trap
MAPPIT experiments were performed as previously described [28]. Briefly, the Fragment 3 of mIP 3 R1 was cloned in a pSEL+2L bait vector, downstream of a chimeric cytokine receptor (Fragment 3 bait), consisting of the extracellular domain of the erythropoietin receptor fused to the transmembrane and cytosolic part of the leptin receptor. Bcl-xL or Bcl-xL K87D was cloned in a in the pMG1-GW plasmid, downstream of a part of the glycoprotein 130 receptor (Bcl-xL or Bcl-xL K87D prey). The interaction between Fragment 3 and Bcl-xL or Bcl-xL K87D is detected by a luciferase reporter assay driven by a STAT-responsive promoter, since the functional complementation of the chimeric cytokine receptor results in liganddependent downstream STAT signaling. We also used the SV40 large antigen T as an irrelevant prey to monitor the signal representing the nonspecific binding to Fragment 3. As an extra negative control, binding of the chimeric cytokine receptor without the Fragment 3 fragment (no bait) to the two Bcl-xL preys was also assessed.

Microscale thermophoresis
Both purified 6xHis-Bcl-xL protein and 6xHis-Bcl-xL K87D were fluorescently labeled using the Monolith His-Tag Labeling Kit RED-tris-NTA 2nd Generation (Nano Temper Technologies, Munich, Germany) and binding affinities were evaluated using microscale thermophoresis. Concentration of 6xHis-Bcl-xL and 6xHis-Bcl-xL K87D was kept constant at 50 nM, whereas the GST-LBD, GST-Fragment 3, GST-Fragment 5b and GST-control proteins were titrated down from 15 µM to 5 nM. Measurements were performed in steady-state conditions using premium capillaries and subsequently recorded on a Monolith NT automated instrument (Nano Temper Technologies) with a pico-red laser channel at 5% excitation power and "medium" MST power. All experiments were repeated three times for each measurement. The Prism software (GraphPad) was used to plot the data points, fit a nonlinear regression curve and calculate the dissociation constant (K d ) for each condition.

Confocal microscopy
100,000 HeLa cells were plated on 18 mm diameter coverslips coated with 0.1 mg/ml poly-L-lysin (Sigma-Aldrich) in 12-well plates and cultured as described above. Twenty-four hours later, the cells were co-transfected with 1 µg pCMV24-3xFLAG-Bcl-xL or pCMV24-3xFLAG-Bcl-xL K87D along with 1 µg of plasmid coding for an ER-targeted RFP or a mitochondriatargeted RFP (mito-RFP). The RR-RFP and mito-RFP were a gift from Professor P. Agostinis (Laboratory of Cell Death Research & Therapy, KU Leuven). X-tremeGene HP DNA (Roche) was used as a transfection reagent according to the manufacturer's instructions. Twenty-four hours after transfection, the cells were fixated in 4% paraformaldehyde for 10 min and permeabilized in 0.3% Triton X-100 for 15 min at room temperature. The blocking was performed with 5% bovine serum albumin (Sigma-Aldrich) in PBS for 30 min at room temperature. Primary antibodies diluted in blocking buffer were applied overnight at 4°C and secondary antibodies diluted in blocking buffer were applied for 1 h at room temperature. Mouse anti-FLAG antibodies (Sigma-Aldrich; F3165) were used as primary antibodies while mouse IgG1s (Dako/Agilent Technologies, Heverlee, Belgium) were used for negative controls, at the same concentration (5 µg/ml). Alexa Fluor 488-coupled goat anti-mouse (Molecular Probes/Thermo Fisher Scientific, Merelbeke, Belgium; A-11017) were used as secondary antibodies (4 µg/ml). The coverslips were eventually mounted on glass slides with the Faramount Mounting Medium Ready (Dako/Agilent Technologies). The cells were imaged using an Axiovert 100 M LSM 510 confocal microscope (Zeiss). Image processing was performed with the ImageJ software (National Institutes of Health, USA) and colocalization was measured with the JACoP plugin [69].

Phylogenetic analysis
Conserved amino acid motifs were calculated using the MEME suite tool. The multiple sequence alignment of the motifs was done in MEGA7. Orthologous sequences for the alignment were selected based on the phylogenetic analysis, performed in RAxML tool with the PROTGAMMALG matrix (Supp. file 1).

Statistical analysis
The Prism software (GraphPad) was used for statistical analysis. Data are expressed as mean ± SEM or SD. Two-tailed Student's t tests were used to compare two conditions and repeated-measure ANOVA with Bonferroni post-tests were performed when comparing three or more conditions. For small sample sizes, non-parametric tests were performed. Statistically significant differences were considered at P < 0.05 (*), P < 0.01 (**) and P < 0.005 (***).