The trans-membrane domain of Bcl-2α, but not its hydrophobic cleft, is a critical determinant for efficient IP3 receptor inhibition

The anti-apoptotic Bcl-2 protein is emerging as an efficient inhibitor of IP3R function, contributing to its oncogenic properties. Yet, the underlying molecular mechanisms remain not fully understood. Using mutations or pharmacological inhibition to antagonize Bcl-2's hydrophobic cleft, we excluded this functional domain as responsible for Bcl-2-mediated IP3Rs inhibition. In contrast, the deletion of the C-terminus, containing the trans-membrane domain, which is only present in Bcl-2α, but not in Bcl-2β, led to impaired inhibition of IP3R-mediated Ca2+ release and staurosporine-induced apoptosis. Strikingly, the trans-membrane domain was sufficient for IP3R binding and inhibition. We therefore propose a novel model, in which the Bcl-2's C-terminus serves as a functional anchor, which beyond mere ER-membrane targeting, underlies efficient IP3R inhibition by (i) positioning the BH4 domain in the close proximity of its binding site on IP3R, thus facilitating their interaction; (ii) inhibiting IP3R-channel openings through a direct interaction with the C-terminal region of the channel downstream of the channel-pore. Finally, since the hydrophobic cleft of Bcl-2 was not involved in IP3R suppression, our findings indicate that ABT-199 does not interfere with IP3R regulation by Bcl-2 and its mechanism of action as a cell-death therapeutic in cancer cells likely does not involve Ca2+ signaling.


INTRODUCTION
A hallmark of cancer cells is their ability to prolong cell survival by avoiding apoptosis. The family of B-cell lymphoma-2 (Bcl-2) proteins is a critical regulator of this process [1][2][3][4][5]. It consists of anti-apoptotic members, including Bcl-2 [6] and Bcl-Xl [7] and pro-apoptotic members like Bax [8]. All the members of the family share at least one of the four conserved α-helical motifs, known as Bcl-2 homology (BH1-4) domains [1,9]. Many of these proteins exist in more than one isoforms [7,[10][11][12][13] and Bcl-2 is not an exception. Two isoforms of Bcl-2, resulting from alternative splicing, were described: Bcl-2α and Bcl-2β [14]. Most of the work until now has been done with Bcl-2α, which is the long isoform and which in addition to the four BH domains contains a C-terminal extension with a putative trans-membrane domain (TMD) ( Figure 1A). In contrast, Bcl-2β has a much shorter C-terminus and lacks a TMD [14,15]. While Bcl-2β is mostly detected in cytosolic fractions, the TMD and a short preceding sequence target Bcl-2α to a variety of intracellular membranes including mitochondrial, endoplasmic reticulum (ER) and nuclear membranes [16][17][18]. Bcl-2α is the more abundant isoform in both healthy and cancer cells and it remains dominant in cancer cells up-regulating Bcl-2 protein [14,19]. In virtually all studies published to this date, Bcl-2 refers to Bcl-2α. The anti-apoptotic function of Bcl-2 oncogene was first characterized at the level of the mitochondria, particularly at the outer mitochondrial membrane, where it inhibits Bax/Bak-mediated apoptosis.

Research Paper www.impactjournals.com/oncotarget
The mechanism involves a BH3-dependent interaction, where the hydrophobic cleft of Bcl-2 formed by the BH3-BH1-BH2 domains sequesters the BH3 domain of the proapoptotic members. This prevents Bax/Bak activation and oligomerization and inhibits the consequent mitochondrial permeabilization and cell death [2,3,20,21].
Ca 2+ signaling is another important modulator in cell-fate decisions, which can serve as a survivor factor promoting cell proliferation, but also as a celldeath inducer [22][23][24][25]. Bcl-2 was shown to execute its pro-survival function not only via direct inhibition of pro-apoptotic proteins but also via suppression of proapoptotic Ca 2+ signals. This occurs by direct interaction with inositol 1,4,5-trisphosphate (IP 3 ) receptors (IP 3 Rs) [26][27][28][29], the main intracellular Ca 2+ release channels, located at the ER [30][31][32][33][34]. IP 3 R inhibition by Bcl-2 appears to be an important mechanism that contributes to the oncogenic properties of Bcl-2. Many cancer cells, including leukemia, lymphoma and lung cancer cells, are addicted to IP 3 R/Bcl-2-complex formation for their survival, since tools that disrupt this complex trigger cell death [35][36][37]. Over the last years, important insights in the regulation of IP 3 Rs by Bcl-2 (i.e. Bcl-2α) at the molecular level have been obtained. The suppression of IP 3 R-mediated Ca 2+ release by Bcl-2 was attributed to the interaction of the BH4 domain of Bcl-2 with a 20 amino acid region (a.a. 1389-1408) located in the central modulatory domain, more particularly in the Domain 3 (Dom 3) (a.a. 923-1581) of IP 3 R [38,39]. Previous studies, which exploited synthetic peptides covering the BH4 domain of Bcl-2 (BH4-Bcl-2), revealed that this domain is necessary and sufficient to bind to IP 3 R and to suppress its activity [26,27,39,40]. Nevertheless, the relatively low affinity of inhibition by the BH4 domain (measured in vitro IC 50 =30μM) [27,39] cannot explain the potent inhibitory effect of Bcl-2 full-length protein in physiological conditions. This Achilles' heel of the model suggests that additional domains in Bcl-2 could be responsible for an efficient in cellulo inhibition of IP 3 R. Interestingly, the C-terminal domain, containing the last 6 th TMD of the IP 3 R (C-term Dom, a.a. 2512-2749), which is in close proximity of the channel pore is also targeted by Bcl-2 [41,42], but the mechanism and significance of this interaction are not completely solved. The same C-term Dom of IP 3 R also appeared to be responsible for interaction with other members of the family: Bcl-Xl and Mcl-1 [41].
Here, we aimed to identify the molecular determinants in Bcl-2α responsible for its interaction with the C-term Dom of IP 3 R and to assess their functional impact on Bcl-2α-mediated inhibition of the channel. We especially focused on two important functional domains in Bcl-2α, i.e. the hydrophobic cleft, involved in BH3-dependent interactions and the C-terminal region, containing the TMD, involved in hydrophobic interactions within the membrane environment ( Figure   1A). Using genetic and pharmacological approaches, we could however exclude the hydrophobic cleft as a major player in the formation of the Bcl-2α/IP 3 R complex. In contrast, we found that Bcl-2α binding to the C-term Dom of IP 3 R1 depends on the presence of the C-terminus of Bcl-2α. This region of Bcl-2α is required for efficient inhibition of IP 3 Rs in a cellular context and for inhibition of staurosporine (STS) -induced apoptosis. Furthermore, we demonstrated a direct interaction between a peptide corresponding to the TMD of Bcl-2α (TMD-Bcl-2) and the purified C-terminal fragment of IP 3 R1. The TMD-Bcl-2 was able to suppress IP 3 -induced Ca 2+ release (IICR) when applied at high concentrations. These results suggest that the C-terminal region, and particularly the TMD, of Bcl-2α not only serves as an anchor for tethering Bcl-2α in the membranes, but is also an important functional regulator of IP 3 R activity. Since the TMD is only present in Bcl-2α, but not in Bcl-2β, this study is the first one hinting towards important functional difference between the two isoforms with respect to Ca 2+ -signaling regulation.

Despite the presence of BH3-domain features in the IP 3 R sequence, the hydrophobic cleft of Bcl-2α is dispensable for interaction with the receptor
We performed a sequence alignment of the BH3 domains of different Bcl-2 proteins with the fragment of the central modulatory domain of IP 3 R1 (Dom 3), shown in previous studies to bind Bcl-2 [27,38,42]. This analysis revealed the presence of a BH3 motif (a.a. 1332-1342) upstream of the previously described region in Dom 3 of IP 3 R targeted by the BH4 domain of Bcl-2 (a.a. 1389-1408) ( Figure 1A) [43]. Figure 1B depicts the presence of the conserved LxxxGD/E motif [44] in the Dom 3 of IP 3 R and the α-helical secondary structure of this motif as predicted by I-TASSER web server. To determine whether a BH3-dependent mechanism plays a direct role in the interaction between Bcl-2α and IP 3 R we used two different approaches to antagonize the hydrophobic cleft of Bcl-2α, genetic manipulation and pharmacological inhibition. The genetic approach is based on mutations in the BH1 domain (replacement of G145R146 by AA yielding Bcl-2 GR/ AA ) ( Figure 1A), which lead to disruption of the binding between Bcl-2α and Bax [45][46][47]. The second approach is based on the use of pharmacological inhibitors like the BH3-mimetic compounds [48,49], designed to occupy the hydrophobic cleft, thereby disrupting interactions between BH3 domain-containing proteins and anti-apoptotic Bcl-2 proteins [48,49]. Here, we applied ABT-199, a selective Bcl-2 inhibitor which does not target Bcl-Xl [50].
First, we validated that both, the GR/AA mutation or the incubation with ABT-199 (3 μM), prevent Bcl-2α binding to Bax in co-immunoprecipitation experiments.
The concentration of ABT-199 that we used in the experiments is well above the documented subnanomolar affinity of this compound for Bcl-2 (Ki < 0.01 nM) [50], thus maximizing the potential effect of ABT-199 on Bcl-2/IP 3 R interaction. 3xFLAG-tagged proteins (3xFLAG-Bcl-2 wt in presence and absence of ABT-199 or 3xFLAG-Bcl-2 GR/AA ) were overexpressed in COS-1 cells and immunoprecipitated from the cell lysates using anti-FLAG-loaded agarose beads. Immunoblots were stained for FLAG and Bax (Figure 2A).
Next we performed two different sets of GST pulldown experiments, using the two purified IP 3 R domains targeted by Bcl-2, GST-Dom 3 (a.a. 923-1581) and GST-C-term Dom (a.a. 2512-2749). To compare the binding properties of the wild-type Bcl-2 protein versus the mutant for these IP 3 R fragments we overexpressed 3xFLAG-Bcl-2 wt or 3xFLAG-Bcl-2 GR/AA in COS-1 cells. The binding of 3xFLAG-Bcl-2 wt to GST-Dom 3 was used as reference and all binding values were normalized to this control.
Our results show that 3xFLAG-Bcl-2 GR/AA remained fully capable of binding to both GST-Dom 3 and GST-C-term Dom to a similar extent as 3xFLAG-Bcl-2 wt ( Figure 2B).
As a second approach, we examined the interaction between Bcl-2 wt and the two GST-fused domains of IP 3 R in presence or absence of the BH3-mimetic compound ABT-199 (3 μM). Incubation with ABT-199 did not significantly affect the binding of 3xFLAG-Bcl-2 wt to the GST-Dom 3, nor to the GST-C-term Dom ( Figure 2C).
Taken together, these results suggest that the hydrophobic cleft of Bcl-2 is dispensable for its interaction with IP 3 R.

The hydrophobic cleft of Bcl-2α does not contribute to the inhibitory effect on IP 3 Rs
Bcl-2 overexpression results in dampened IP 3 Rmediated Ca 2+ release in intact cells [27,29,39,51], but whether this effect is mediated through the hydrophobic with its BH domains and the trans-membrane domain (TMD). The two functional domains of interest, the hydrophobic cleft formed by BH3-BH1-BH2 domains and the C-terminus (C), are indicated with black lines. The C-terminal region, containing the TMD, is present in Bcl-2α, but not in Bcl-2β. The G145 and R146 residues, located in the BH1 domain were mutated to yield Bcl-2 GR/AA . Bcl-2 was truncated at W214 residue to yield Bcl-2 ΔC , which correlates with Bcl-2β. A schematic representation of IP 3 R1 is depicted in green. The Bcl-2binding fragments of IP 3 R used in this study, the domain 3 (Dom 3) and the C-terminal domain containing the last TMD (C-term Dom), are indicated with black lines and the six TMDs are shown as black bars. The exact BH4-binding site in the Dom 3 is represented in light grey (BH4-BS). The BH3-like motif in the Dom 3 is represented in yellow. B. The Dom 3 of IP 3 R1 contains a BH3 motif. Left: Sequence alignment between the BH3 domains of Bcl-2 family members and the Dom 3 of IP 3 R1 reveals the presence of the conserved residues (LxxxGD/E, pointed in red), required for a typical BH3 motif (46). Right: A secondary structure prediction of the putative BH3 motif of IP 3 R1 present in the Dom 3 sequence, predicted by the I-TASSER web server and drawn using PyMol. The predicted BH3 motif within the Dom 3 is depicted in yellow and the conserved residues in red.

Figure 2:
Bcl-2 GR/AA and Bcl-2 wt exposed to ABT-199 fail to bind pro-apoptotic Bax, but remain capable of binding the Dom 3 and the C-term Dom of IP 3 R. A. Representative FLAG-co-immunoprecipitation experiment for detection of the 3xFLAG-Bcl-2/Bax interaction is shown. Overexpressed 3xFLAG-Bcl-2 wt (in absence and presence of 3 μM ABT-199) or 3xFLAG-Bcl-2 GR/AA was immunoprecipitated from COS-1 cell lysates by anti-FLAG-loaded agarose beads. The immunoreactive blots were stained with antibody against FLAG and Bax. 0.1 μg and 0.5 μg of total COS-1 lysates were used as input for the 3xFLAG-proteins and Bax respectively. The experiments were performed 3 times utilizing each time independently transfected cells and freshly prepared lysates. B. Representative GST-pull down experiments with COS-1 cell lysates for comparing the binding of overexpressed 3xFLAG-Bcl-2 wt and 3xFLAG-Bcl-2 GR/AA to the GST-Dom 3 and GST-C-term Dom are shown. The samples were analyzed via Western blot and stained with anti-FLAG antibody. The binding to the GST was used as a negative control. 0.1 μg of total COS-1 lysates was used as input. C. The immunoreactive bands from 3 independent experiments, utilizing each time independently transfected cells and freshly prepared lysates, were quantified and normalized to the binding of 3xFLAG-Bcl-2 wt to GST-Dom 3, which was set as 1. The data are plotted as mean ± S.E.M. D. Representative GST-pull down experiments with COS-1 cell lysates for comparing the binding of overexpressed 3xFLAG-Bcl-2 wt in absence and presence of 3 μM ABT-199 to the GST-Dom 3 and GST-C-term Dom are shown. The samples were analyzed via Western blot and stained with anti-FLAG antibody. The binding to the GST was used as a negative control. 0.1 μg of total COS-1 lysates was used as input. E. The immunoreactive bands from 3 independent experiments, utilizing each time independently transfected cells and freshly prepared lysates, were quantified and normalized to the binding of 3xFLAG-Bcl-2 wt to GST-Dom 3, which was set as 1. The data are plotted as mean ± S.E.M. cleft of Bcl-2 is not known. To address this question, we monitored the change in cytosolic Ca 2+ levels in response to an IP 3 R agonist, ATP, using the ratiometric fluorescent Ca 2+ dye Fura-2-AM. Similarly to the GSTpull down experiments, we used the mutation (Bcl-2 GR/AA ) or ABT-199 to antagonize the hydrophobic cleft of Bcl-2α. Intact COS-1 cells overexpressing 1) 3xFLAG-empty vector, 3xFLAG-Bcl-2 wt or 3xFLAG-Bcl-2 GR/AA and 2) 3xFLAG-empty vector or 3xFLAG-Bcl-2 wt in presence or absence of ABT-199 (3 μM), and co-transfected with mCherry plasmid were exposed to ATP (0.5 μM). The proper expression of the 3xFLAG-proteins in the COS-1 cells was assessed via Western blotting using anti-FLAG antibody (Supplementary Figure S1A and S1B). Importantly, the expression levels of 3xFLAG-Bcl-2 wt and 3xFLAG-Bcl-2 GR/AA proteins were similar, although 3xFLAG-Bcl-2 GR/AA tended to be expressed at slightly higher levels. In addition, only cells with similar intensity of mCherry, thus similar levels of 3xFLAG-proteins were subjected to measurement. To chelate the free extracellular Ca 2+ , the experiments were performed in the presence of BAPTA (3 mM), an extracellular Ca 2+ buffer, ensuring that the ATP-induced [Ca 2+ ] rise is only due to Ca 2+ release from intracellular stores. The ER Ca 2+ -store content was also assessed by applying thapsigargin (Tg, 1 μM), an irreversible SERCA inhibitor, in the presence of BAPTA (Supplementary Figure S1A and Supplementary Figure S1B). Consistent with our previous studies, overexpression of 3xFLAG-Bcl-2 wt inhibited ATP-induced Ca 2+ release without affecting the ER Ca 2+ -stores content [27]. In line with our GST-pull down experiments, neither the overexpression of 3xFLAG-Bcl-2 GR/AA ( Figure 3A-3C), nor the presence of ABT-199 ( Figure 4A-4D) prevented this effect. The quantitative analysis indicated that 3xFLAG-Bcl-2 wt , 3xFLAG-Bcl-2 GR/AA ( Figure 3D) and 3xFLAG-Bcl-2 wt in presence of ABT-199 ( Figure 4E) were equally potent in inhibiting IP 3 R-mediated Ca 2+ release.
Finally, to underpin that IP 3 R inhibition by Bcl-2α is not affected by ABT-199, we performed direct IP 3 R singlechannel measurements by using patch-clamp recordings on giant unilamellar vesicles (GUVs) prepared from the ER membrane fractions of native WEHI7.2 cells, which do  Figure 4F, 4G presents a comparison of the measured IP 3 R-mediated channel current after application of IP 3 (5 μM) and Ca 2+ (1 μM). The results demonstrate a significant inhibition of IP 3 R activity in the presence of Bcl-2α, measured as the open probability (NP o ). The NP o value of 0.89 ± 0.07 for the WEHI7.2-control cells decreased to 0.26 ± 0.09 for WEHI7.2 Bcl-2 cells. Application of ABT-199 (1 μM) could not alleviate the inhibitory effect of Bcl-2α on IP 3 R single-channel opening (NP o 0.08 ± 0.05).
Collectively, these functional experiments based on independent approaches exclude a major contribution of the hydrophobic cleft of Bcl-2α for inhibiting IP 3 Rmediated Ca 2+ flux.
The C-terminal region of Bcl-2α is critical for its interaction with the C-term Dom, but not with the Dom 3 of IP 3 R1 After demonstrating that the hydrophobic cleft of Bcl-2α is not involved in the binding to and inhibition of IP 3 R, we investigated whether the C-terminal region containing the TMD of Bcl-2 could serve as an IP 3 Rinteraction domain. We studied the binding of 3xFLAG-Bcl-2 lacking its C-terminal region (3xFLAG-Bcl-2 ΔC ) to purified GST-Dom 3 and GST-C-term Dom using GSTpull-down assays. In these experiments, consistent with our previous results, 3xFLAG-Bcl-2 wt bound with equal efficiency both IP 3 R GST-domains [42]. In line with previous data, showing that the BH4 domain of Bcl-2 is sufficient to bind to the Dom 3 [27,39], 3xFLAG-Bcl-2 ΔC remained capable to bind to this domain. Yet, the interaction with GST-C-term Dom was severely impaired ( Figure 5A, 5B). These results suggest that while the C-terminal region of Bcl-2α is not crucial for interaction with the Dom 3, it is essential for binding to the C-term Dom of IP 3 R. 3xFLAG-Bcl-2 wt , 3xFLAG-Bcl-2 GR/AA and 3xFLAG-Bcl-2 ΔC bind to the full-size IP 3 R 3xFLAG-Bcl-2 mutants seem to have differential binding properties for the different IP 3 R domains. However, the performed FLAG-co-immunoprecipitation experiments with lysates from COS-1 cells overexpressing 3xFLAG-empty vector, 3xFLAG-Bcl-2 wt , 3xFLAG-Bcl-2 GR/AA or 3xFLAG-Bcl-2 ΔC revealed that the wild type and both mutated proteins are able to interact with the endogenous IP 3 R1 (Supplementary Figure S2). These data are consistent with previous studies showing that the BH4 domain of Bcl-2 is the major determinant for binding to IP 3 Rs [27].
We also compared the binding properties of 3xFLAG-Bcl-2 wt , 3xFLAG-Bcl-2 GR/AA and 3xFLAG-Bcl-2 ΔC for endogenous pro-apoptotic Bax. As expected, 3xFLAG-Bcl-2 GR/AA failed to interact with Bax. The truncated Bcl-2 displayed equal efficiency for binding Bax as the wild type Bcl-2, confirming that the hydrophobic cleft is the major binding determinant in Bcl-2 interactions with pro-apoptotic proteins (Supplementary Figure S2).

The TMD of Bcl-2α directly interacts with the C-term Dom of IP 3 R1
As a next step we performed pull-down experiments using neutravidin-coated beads that captured the biotinylated peptides corresponding either to the TMD of Bcl-2 (biotin-TMD-Bcl-2) or to a control version in which several hydrophobic residues were substituted by charged amino acids (biotin-TMD-Bcl-2-CTR) in the presence of either purified parental GST or purified GST-C-term Dom of IP 3 R1. After incubation and washing steps, the resulting pull-down samples were analysed via immunoblotting using anti-GST antibody ( Figure 5C). This analysis revealed a direct interaction between the GST-C-term Dom of IP 3 R1 and biotin-TMD-Bcl-2.

The lack of the C-terminus leads to loss of Bcl-2α's ability to suppress IP 3 R-mediated Ca 2+ release
Next, we studied the role of the C-terminus in Bcl-2α's inhibitory function on IP 3 R-mediated Ca 2+ signaling. Similar experiments were performed as described in Figure 4, comparing the effect of 3xFLAG-Bcl-2 wt versus 3xFLAG-Bcl-2 ΔC overexpression on ATP-induced IP 3 Rmediated Ca 2+ release. In contrast to 3xFLAG-Bcl-2 wt , which reduced Ca 2+ release in response to ATP (0.5 μM), 3xFLAG-Bcl-2 ΔC was not able to suppress IP 3 R-mediated Ca 2+ release ( Figure 6). The 3xFLAG-proteins displayed similar expression levels (Supplementary Figure S1C), indicating that the failure of 3xFLAG-Bcl-2 ΔC to inhibit IP 3 Rs is not due to a lower expression level compared to the 3xFLAG-Bcl-2 wt protein. The ER Ca 2+ -store content was not changed in either of the conditions, pointing that the difference in ATP-induced Ca 2+ rise is not due to a decreased ER Ca 2+ -store content (Supplementary Figure  S1C).

The TMD of Bcl-2α suppresses IICR in permeabilized cells and in single-channel recordings
We demonstrated that the TMD of Bcl-2α directly binds to the C-term Dom of IP 3 R and that Bcl-2 ΔC does not inhibit IP 3 R-mediated Ca 2+ release. Next, we assessed whether the TMD of Bcl-2α by itself could affect the Ca 2+ -flux properties of the IP 3 R. Therefore, we performed unidirectional 45 Ca 2+ flux assays in saponin-permeabilized mouse embryonic fibroblasts (MEFs), in which nonmitochondrial Ca 2+ stores were loaded with 45 Ca 2+ . After loading, the unidirectional Ca 2+ flux was measured in the presence of EGTA (1 mM) and in presence of Tg (4 μM). We quantified 45 Ca 2+ release triggered by IP 3 (3 μM) in presence or absence of different concentrations of the synthetic peptides corresponding to the TMD of Bcl-2 or its mutated version. Peptides were applied 2 min before till 2 min after IP 3 application. All conditions were matched to the vehicle control (DMSO). Data are plotted as the fractional loss (%/2 min) over time. These experiments indicated that high concentrations of TMD-Bcl-2 (30 μM and higher), but not of TMD-Bcl-2-CTR, suppress IICR without affecting ER Ca 2+ level ( Figure 7A, 7B). To further underpin these observations, IP 3 R1 single-channel recordings were performed using the nuclear-membrane patch-clamp technique on isolated nuclei obtained of triple-IP 3 R-knockout DT40 cells ectopically expressing IP 3 R1 [52]. This approach allows a direct measurement of the activity of the IP 3 R1 channel. IP 3 R1 single-  Figure 7C shows different representative traces of IP 3 R1 single-channel openings at a pipette holding potential of -100 mV in control conditions or in the presence of TMD-Bcl-2 or TMD-Bcl-2-CTR peptides, both at 60 μM final concentrations. TMD-Bcl-2 decreased the P o of the IP 3 R1 channel from about 0.25 in the control conditions to about 0.15, whereas the TMD-Bcl-2-CTR peptide did not have any significant impact.

Bcl-2α requires its TMD to suppress STSinduced apoptosis
Finally, we studied the potency of overexpressed 3xFLAG-Bcl-2 wt , 3xFLAG-Bcl-2 GR/AA and 3xFLAG-Bcl-2 ΔC to protect against STS, an apoptotic trigger that acts in part through Ca 2+ signalling [53]. As a marker of apoptosis, we monitored the cleavage of poly-(ADP-ribose)polymerase 1 (PARP1), which is a downstream target of activated Caspase-3. Compared to the control cells (transfected with an empty vector), the overexpression of 3xFLAG-Bcl-2 wt significantly reduced the levels of cleaved PARP1 upon STS treatment (1 μM, 6h). 3xFLAG-Bcl-2 GR/AA failed to prevent PARP1 cleavage, in line with its failure to bind Bax (Supplementary Figure S2). Despite the fact that 3xFLAG-Bcl-2 ΔC was equally efficient as the 3xFLAG-Bcl-2 wt to bind endogenous Bax (Supplementary Figure S2), it was much less efficient in preventing STSinduced apoptosis in COS-1 cells (Figure 8).

DISCUSSION
Here, we demonstrate that the efficient in cellulo suppression of IP 3 R activity by Bcl-2α protein requires the C-terminal region, containing the TMD, but not the hydrophobic cleft of Bcl-2α. Consistent with this finding, Bcl-2α lacking the TMD is less effective to protect cells against Ca 2+ -dependent pro-apoptotic stimuli like Histogram depicting the open probability (P o ) ± SD for the IP 3 R1 under the previously described conditions. The total number of recordings for each condition is indicated within every bar. staurosporine. Since the TMD is present only in Bcl-2α, but not in Bcl-2β, our study is the first one that indicates a possible difference between the functional effects of the Bcl-2 isoforms on IP 3 R activity ( Figure 9) and thus on Ca 2+ -dependent apoptosis. Furthermore, our data indicate that BH3-mimetic compounds like ABT-199, which selectively antagonizes Bcl-2, do not interfere with the functional regulation of IP 3 Rs by Bcl-2.
Using genetic and pharmacological approaches, we firmly ruled out a major role for the hydrophobic cleft of Bcl-2 in inhibiting IP 3 R function, despite the presence of previously suggested [54] or identified throughout this study putative BH3 motifs within the IP 3 R sequence ( Figure 1B). This is in striking contrast to the regulation of IP 3 Rs by Bcl-Xl, very recently described to occur via a BH3-dependent mechanism, involving an interaction between the hydrophobic cleft of Bcl-Xl and two BH3 motifs in the C-term Dom of IP 3 R [54,55]. Disruption of these interactions resulted in diminished cell viability. The authors speculated that similar BH3-dependent interactions might underlie the Bcl-2/IP 3 R complex [55]. Therefore, our work suggests that despite the similarities in their structure and function as inhibitors of the canonical Bax/Bak-dependent apoptosis, Bcl-2 and Bcl-Xl target and regulate IP 3 Rs by different mechanisms. The data reported here might concede another striking difference in addition to the documented selective function of Bcl-2 versus Bcl-Xl in regulating IP 3 Rs at the level of their BH4 domains [27]. Of note, selective BH3-mimetic molecules that could occupy the hydrophobic cleft of Bcl-2, but not that of Bcl-Xl, have been developed, indicating important differences in the molecular determinants contributing to the hydrophobic cleft of Bcl-2 and Bcl-Xl [50]. Hence, the BH3 motifs present in the IP 3 R might be suited for binding the hydrophobic cleft of Bcl-Xl, but not the one of Bcl-2. In addition, the hydrophobic cleft of Bcl-2 was recently excluded as a major contributor in the inhibition of another family of intracellular Ca 2+ -release channels, namely ryanodine receptors (RyRs) [45].
Previously we reported that the absence of the 6 th TMD of IP 3 R results in impaired Bcl-2 binding to the C-terminus of the channel [42]. Here, we demonstrate that the TMD of Bcl-2α is also required for this interaction, which likely occurs within the ER membrane. We propose that the TMD of Bcl-2α provides a concentration effect of Bcl-2 and its BH4 domain. This indicates that the membrane-dependent interaction between Bcl-2α and IP 3 R is critical for effective in cellulo inhibition of IP 3 Rmediated Ca 2+ signaling and subsequent protection against Ca 2+ -dependent apoptosis. Our results also hint towards an unappreciated function of the TMD of Bcl-2α beyond its anchoring role for protein insertion into the membranes. Indeed, the TMD by itself is sufficient to inhibit IP 3 Rs as shown in unidirectional 45 Ca 2+ fluxes and single IP 3 Rchannel recordings, correlating with previous findings that TMDs of other Bcl-2 family members play an important role in the protein functioning [56,57]. However, it should be noted that even 60 μM TMD-Bcl-2 only partially inhibited IP 3 R-mediated Ca 2+ release in permeabilized cells, indicating that also other Bcl-2 domains, particularly the BH4 domain, are required for efficient IP 3 R inhibition by Bcl-2α. Of note, the binding of Bcl-2α lacking its C-terminal region to the purified Dom 3 or to the fulllength IP 3 Rs was not significantly disturbed, supporting the idea that the BH4 domain of Bcl-2 is sufficient for binding to the Dom 3 of IP 3 R and that this binding indeed occurs with relatively high affinity as documented via previous surface plasmon resonance analysis. We propose a model according to which the efficient IP 3 R inhibition relies on a complex multi-domain binding between Bcl-2α and IP 3 R, involving interactions between the BH4 domain of the former and the Dom 3 of the channel, and between the C-terminal regions of both proteins. We propose that the C-terminus of Bcl-2α ( Figure 9C), but not the one of the Bcl-2β ( Figure 9B), mediates the inhibitory effect of the BH4 domain by increasing its local concentration in the proximity of the Dom 3 of IP 3 R. In addition, the dual targeting of IP 3 Rs by Bcl-2α via its C-terminus and its BH4 domain might affect the conformational flexibility of the IP 3 R, by locking it in a rigid conformation and limiting the opening of the Ca 2+ -channel pore in response to IP 3 (Figure 9). A particularly challenging aspect of our model is that, based on the most recent cryo-electron microscopy high-resolution structure of the IP 3 R1 [58], the 6 th TMD of the IP 3 R may not be readily available for interaction with other proteins. Yet, the published structure is in the absence of IP 3 and thus likely represents the closed state. Hence, changes in the IP 3 R structure might arise in different IP 3 / Ca 2+ conditions impacting the accessibility of the 6 th TMD of the IP 3 R to proteins like Bcl-2. Vice versa, it is also possible that the structure of IP 3 Rs loaded with Bcl-2 is different from the structure of IP 3 Rs in the absence of Bcl-2, thereby impacting the structural environment of the 6 th TMD of IP 3 R. Finally, we also would like to note that the molecular foundation for this model is mainly based on binding studies, using IP 3 R1-expression constructs and the electrophysiological analysis of IP 3 R1 channels. However, some of the cell models used for the functional analysis mainly express IP 3 R3 and IP 3 R1 isoforms [59]. As such, we anticipate that the important role of Bcl-2's TMD for domain interaction between Bcl-2 proteins and IP 3 R. On the right side these interactions are depicted within the ER membrane environment. A. Without IP 3 present, IP 3 R is in closed conformation and no Ca 2+ release occurs. B. Upon stimulation, IP 3 binds to the N-terminal ligandbinding domain of IP 3 R and leads to change in the conformation of the channel from closed to open state. This results in IP 3 R-mediated Ca 2+ release. Bcl-2β, which similarly to our Bcl-2 ΔC , contains the BH4 domain, but lacks the TMD, might result in ineffective binding and regulation of the channel in cellulo. C. Efficient IP 3 R inhibition by Bcl-2α in cellulo requires multi-domain interaction between the two proteins, which involves binding of the BH4 domain of Bcl-2α to the Dom 3 of IP 3 R and binding between their C-termini. Here, we hypothesize that due to this multi-domain interaction, the IP 3 R is "locked" in a rigid conformation, leading to decreased Ca 2+ release through the channel even in presence of IP 3 . We propose a model, in which the interaction between the TMD of Bcl-2α and the C-term Dom of IP 3 R can "concentrate" the BH4 domain in the proximity of the Dom 3 by serving as an anchoring mechanism (indicated with an anchor). This "concentration effect" could overcome the inherent low affinity of inhibition by the BH4 domain. In addition to its anchoring role, the TMD of Bcl-2α has an inhibitory effect by itself. efficient IP 3 R inhibition is not limited to IP 3 R1 channels, but further detailed molecular and functional work would be needed to firmly proof this. Of note, the BH4-domainbinding site present in IP 3 R1 is completely conserved in IP 3 R2 and IP 3 R3 [60], consistent with Bcl-2's ability to bind to the central domain of all three IP 3 R isoforms [27].
The importance of the multi-domain interaction between IP 3 Rs and Bcl-2 is underpinned by the fact that peptides antagonizing Bcl-2 at its BH4 domain (like Bcl-2/ IP 3 Receptor Disrupter-2; BIRD-2) are able to trigger proapoptotic Ca 2+ signaling in a variety of cancer-cell models, including lymphoma, leukemia and lung cancer cells [35,37,61]. Thus, development of inhibitors targeting Bcl-2's TMD and interfering with the IP 3 R/Bcl-2 complex at the level of the TMD/C-term Dom interaction might further potentiate BH4-domain-antagonizing tools by helping to destabilize the Bcl-2/IP 3 R complex. Yet, given the hydrophobic nature of TMD/C-term Dom interactions, such small-molecule developments may prove to be very challenging.
We conclude that efficient IP 3 R regulation by Bcl-2α requires the TMD, a unique feature that discriminates Bcl-2α from Bcl-2β. Bcl-2α, via its TMD, likely "concentrates" its BH4 domain in the proximity of the central, modulatory domain of the IP 3 R, thereby facilitating its ability to efficiently suppress IP 3 R-mediated Ca 2+ signaling and subsequent apoptosis.

Peptides
The following peptides, obtained from Life Tein (Hillsborough, NJ, USA) with purity ≥ 85% were used: the peptide corresponding to the TMD of Bcl-2, Bcl-2-TMD: KTLLSLALVGACITLGAYLGHK (also used with biotin tag); the control peptide containing several mutations of hydrophobic residues, Bcl-2-TMD-CTR: KTRRSLADRGACRTRGAYDGHK (also used with biotin tag) and the peptide used to compete with the 3xFLAG tag, Anti-DYKDDDDK-tag peptide: MDYKDHDGDYKDHDIDYKDDDDK.

Antibodies
The
24 hours after seeding COS-1 cells were transiently transfected with empty p3xFLAG-Myc-CMV-24 (3xFLAG-empty) or with the same vector containing either Bcl-2 wt or the mutants Bcl-2 GR/AA or Bcl-2 ΔC . For co-IP and pull-down experiments JETPrime transfection reagent (Polyplus Transfections, Illkirch, France) was used according to the manufacturer's instructions. For single-cell cytosolic [Ca 2+ ] measurements COS-1 cells were seeded in two-chamber slides. The same construct, in combination with a pcDNA 3.1(-) mCherry-encoding vector at a 3:1 ratio as selection marker, were introduced 24 hours after seeding using X-tremeGene HP DNA (Roche, Basel, Switzerland) as a transfection reagent according to the manufacturer's instructions.

Single-cell cytosolic Ca 2+ imaging
Fura-2-AM [Ca 2+ ] measurements in COS-1 cells were performed as previously described [27]. The effect of ABT-199 was studied by incubating the cells with 3 μM of the compound or DMSO for 1 hour (during the incubation procedure with Fura-2 AM). BAPTA (3 mM) was added for 1 minute prior to the stimulation with ATP or Tg to chelate all free extracellular Ca 2+ . Cytosolic Ca 2+ rises in response to 0.5 μM ATP or 2.5 μM Tg were measured in mCherrypositive (excitation 546 nm, emission 610 nm) and Fura-2loaded cells. Intracellular cytoplasmic Ca 2+ concentrations were calculated as previously described [27].

Unidirectional 45 Ca 2+ -flux assay
The unidirectional 45 Ca 2+ -flux experiments were performed in permeabilized MEFs as previously described [27]. IICR was triggered during the unidirectional 45 Ca 2+efflux phase by the addition of 3 μM IP 3 for 2 min. Peptides were added 2 min before IP 3 till 2 min after IP 3 . IICR was plotted as fractional loss, representing the amount of Ca 2+ leaving the store in a 2-min time period divided by the total store Ca 2+ content at that time point as a function of time [64].

Preparation of GUVs and electrophysiological analysis
Isolation of the ER-containing membrane fractions from control and Bcl-2-expressing WEHI7.2 cells and preparation of the GUVs were carried out as described previously [65]. GUVs were prepared from the 1:5 mixtures of the ER-containing fraction with 10:1 diphytanoylphosphatidylcholine/cholesterol lipid combination (5 mM). The Patch-clamp experiments were carried out using Axopatch 200B amplifier and pClamp 10.0 software (Molecular Devices, Union City, CA) for data acquisition and analysis. Patch pipettes were fabricated from borosilicate glass capillaries (World Precision Instr., Inc., Sarasota, FL) on a horizontal puller (Sutter Instruments Co., Novato, CA) and had a resistance in the range of 7-10 MΩ. Prepared vesicles were immersed in a bath solution containing 150 mM KCl, 10 mM Hepes, 5 mM glucose, pH 7.2. Patch pipettes were filled with the same solution.

Isolation of nuclei and electrophysiological analysis
Isolated DT40 nuclei were prepared by homogenization as previously described [52]. A 3 μl aliquot of nuclear suspension was placed in 3 ml of bath solution which contained 140 mM KCl, 10 mM Hepes, 500 μM BAPTA and 246 nM free Ca 2+ , pH 7.1. Nuclei were allowed to adhere to a plastic culture dish for 10 min prior to patching. Single IP 3 R channel potassium currents (ik) were measured in the on-nucleus patch clamp configuration using pCLAMP 9 and an Axopatch 200B amplifier (Molecular Devices, Sunnydale, CA, USA) as previously described [66]. Pipette solution contained 140 mM KCl, 10 mM Hepes, 1 μM IP 3 , 5 mM ATP, and 200 nM free Ca 2+ as well as 60 μM TMD-Bcl-2 or TMD-Bcl-2-CTR peptides. Traces were consecutive 3 s sweeps recorded at −100 mV, sampled at 20 kHz and filtered at 5 kHz. A minimum of 15 s of recordings were considered for data analyses. Pipette resistances were typically 20 MΩ and seal resistances were >5 GΩ. Single channel openings were detected by half-threshold crossing criteria using the event detection protocol in Clampfit 9. We assumed that the number of channels in any particular nuclear patch is represented by the maximum number of discrete stacked events observed during the experiment. Only patches with one apparent channel were considered for analyses.

Sequence alignment and secondary-structure predictions
The amino acid sequences of the BH3 domains of Bcl-2 proteins and the Dom 3 of IP 3 R were taken from the National Center for Biotechnological Information's nonredundant database. The I-TASSER v 2.1 webserver [67,68] was used to predict the secondary structure of the BH3-like motif identified in the Dom 3 of IP 3 R1. I-TASSER builds protein models using iterative assembling procedures and multiple threading alignments from template structures libraries. An estimate of accuracy of the predictions is given by the confidence score. The most accurate I-TASSER model was downloaded as PDB file and imported in PyMOL, a molecular graphic software (http://www.pymol.org).

Statistical analysis
Two-tailed unpaired Student's t-tests were performed when two conditions were compared. When comparing three or more conditions a repeated measure ANOVA with Bonferroni post test was performed. * indicates significantly different results with p < 0.05.