Ryanodine receptors are targeted by anti-apoptotic Bcl-XL involving its BH4 domain and Lys87 from its BH3 domain

Anti-apoptotic B-cell lymphoma 2 (Bcl-2) family members target several intracellular Ca2+-transport systems. Bcl-2, via its N-terminal Bcl-2 homology (BH) 4 domain, inhibits both inositol 1,4,5-trisphosphate receptors (IP3Rs) and ryanodine receptors (RyRs), while Bcl-XL, likely independently of its BH4 domain, sensitizes IP3Rs. It remains elusive whether Bcl-XL can also target and modulate RyRs. Here, Bcl-XL co-immunoprecipitated with RyR3 expressed in HEK293 cells. Mammalian protein-protein interaction trap (MAPPIT) and surface plasmon resonance (SPR) showed that Bcl-XL bound to the central domain of RyR3 via its BH4 domain, although to a lesser extent compared to the BH4 domain of Bcl-2. Consistent with the ability of the BH4 domain of Bcl-XL to bind to RyRs, loading the BH4-Bcl-XL peptide into RyR3-overexpressing HEK293 cells or in rat hippocampal neurons suppressed RyR-mediated Ca2+ release. In silico superposition of the 3D-structures of Bcl-2 and Bcl-XL indicated that Lys87 of the BH3 domain of Bcl-XL could be important for interacting with RyRs. In contrast to Bcl-XL, the Bcl-XLK87D mutant displayed lower binding affinity for RyR3 and a reduced inhibition of RyR-mediated Ca2+ release. These data suggest that Bcl-XL binds to RyR channels via its BH4 domain, but also its BH3 domain, more specific Lys87, contributes to the interaction.

RyR via its BH4 domain results in an inhibition of RyR-mediated Ca 21 release. The Bcl-2 K17D mutant does not show a dramatic loss of binding to the RyR and is as potent as wild-type Bcl-2 in inhibiting RyR-mediated Ca 21 release. These results may indicate that in contrast to the IP 3 R, which is differentially targeted by Bcl-2 and Bcl-X L , RyRs might have a common interaction site for both proteins and do not distinguish between these two proteins for their regulation.
In this paper, we show that similarly to Bcl-2, Bcl-X L binds to the RyR via a site located in its central, modulatory domain, thereby inhibiting RyR-mediated Ca 21 release. Although the BH4 domain of Bcl-X L was sufficient for inhibiting RyRs, we found that in fulllength Bcl-X L not only the BH4 domain but also the BH3 domain contributed to Bcl-X L /RyR-complex formation. In particular, we identified Lys87, located in the BH3 domain of Bcl-X L , as an important contributor of Bcl-X L binding to the RyR.

Results
Bcl-X L binds to RyR3. Bcl-2 K17D is a Bcl-2 mutant based on a critical difference between the BH4 domains of Bcl-2 and Bcl-X L and is impaired in binding to and regulating IP 3 Rs 19 . However, this mutant still binds to and regulates RyRs with similar efficiencies as wild-type Bcl-2 16 , suggesting that Bcl-X L may also bind to and regulate RyRs. Hence, we performed co-immunoprecipitation studies using lysates from HEK293 cells stably overexpressing RyR3 (HEK RyR3). In these cells, transiently overexpressed 3XFLAG-tagged Bcl-X L coimmunoprecipitated with RyR3 indicating the formation of RyR3/Bcl-X L complexes ( Fig. 1A and Supplementary Fig. 1A for uncropped Western-blot images).
In our previous work we reported that the interaction between Bcl-2 and the RyR occurred via the BH4 domain of Bcl-2 and a central regulatory domain of the RyR (a.a. 22632263-2688 for mink 2688 for mink RyR3) 16 . To examine whether a direct interaction between RyRs and the BH4 domain of Bcl-X L exists and whether this interaction occurs via the same or similar domains, surface plasmon resonance (SPR) experiments were performed (Fig. 1B). A concen-tration-dependent binding between biotin-BH4-Bcl-X L immobilized to streptavidin coated SPR chips, and the purified GST-RyR3 domain (mink RyR3, a.a. 2263-2688) could be detected. In contrast, but consistent with our previous observations, purified GST-tagged IP 3 R1 domain 3 (mouse IP 3 R1, a.a. 9232263-2688 for mink 1581), which is known to bind to the BH4 domain of Bcl-2, failed to bind to biotin-BH4-Bcl-X L 19 . While biotin-BH4-Bcl-X L was able to bind to the GST-RyR3 domain, it seemed to be less effective than biotin-BH4-Bcl-2 16 . To confirm the proper loading of the biotin-BH4-Bcl-X L peptide to the sensor chip, we monitored the binding of an antibody directed against the BH4 domain of Bcl-X L , which caused a prominent increase in resonance unit (RU) values ( Supplementary  Fig. 2). Collectively, these results indicate that the interaction of Bcl-X L with the RyR3 is direct and that Bcl-X L via its BH4 domain targets the same domain as Bcl-2 on the RyR. However, the BH4 domain of Bcl-X L seems to have a lower affinity for the GST-RyR3 domain compared to the BH4 domain of Bcl-2. This could indicate that biotinylation of the BH4 domain of Bcl-X L influences its binding capabilities more than is the case for the BH4 domain of Bcl-2. Alternatively, other domains besides Bcl-X L 's BH4 domain may be involved in the interaction of full-length Bcl-X L with the RyR. Therefore, we wanted to identify if other domains besides the BH4 domain of Bcl-X L are important for interacting with the RyR.
Superposition of the 3D-structures of Bcl-2 and Bcl-X L reveals a spatial resemblance of Lys17 in the BH4 domain of Bcl-2 with Lys87 in the BH3 domain of Bcl-X L . To identify the contribution and involvement of other Bcl-X L domains for targeting RyR channels, an in silico superposition of the Bcl-2 (PDB-entry 4AQ3 20 ) and Bcl-X L (PDB-entry 1R2D 21 ) structures was performed with the aid of PyMOL (The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC.). This superposition allowed the comparison of corresponding residues in the 3D-structures of Bcl-2 and Bcl-X L (Fig. 2). This analysis revealed that the positively charged e-amino terminus of the side chain of Lys87 in Bcl-X L , located in the RyR3 cells transiently overexpressing 3XFLAG-Bcl-X L . RyR3 was immunoprecipitated from these lysates utilizing a pan-RyR antibody. An anti-FLAG-HRP conjugated antibody was used for detecting co-immunoprecipitated 3XFLAG-Bcl-X L . Immunoblot showing the immunoprecipitated RyR3 (top) and co-immunoprecipitated 3XFLAG-tagged Bcl-X L (bottom). Immunoprecipitations using non-specific IgG antibodies were applied as negative controls. All experiments were performed at least three times utilizing each time independently transfected cells and freshly prepared HEK RyR3 lysates. All samples were run using the same experimental conditions on the same gel/blot. The uncropped image is shown in Supplementary Fig. 1A. (B) Sensorgrams of the surface plasmon resonance experiments expressed in RU as a function of time. The biotin-BH4-Bcl-X L peptide and the scrambled peptide were immobilized on different channels of a streptavidin-coated sensor chip. The channels on the chip were exposed to the indicated concentrations of purified GST-fusion proteins (GST-IP 3 R1 domain 3 and GST-RyR3 domain). Binding of the GST-tagged proteins to the scrambled peptides was subtracted from each sensorgram. GST-IP 3 R1 domain 3 bound stronger to the scrambled peptide than to the biotin-BH4-Bcl-X L resulting in apparent negative values after this correction. The black arrow indicates the start of the association phase (addition of the GST-tagged proteins) and the grey arrow indicates the start of the dissociation phase (running buffer alone BH3 domain, is in the same spatial constraints as the positively charged e-amino terminus of the side chain of Lys17 located in the BH4 domain of Bcl-2. Furthermore, Lys87 did not seem to be part of the hydrophobic cleft of Bcl-X L , as it was directed towards the space facing the BH4 domain. The Bcl-X L K87D mutant is impaired in RyR3 binding. The relevance of Lys87 in Bcl-X L for RyR binding was addressed via mammalian protein-protein interaction trap (MAPPIT) 22 , an in cellulo proteinprotein interaction assay. MAPPIT is based on the functional complementation of cytokine receptor signaling. To study the possible existence of RyR/Bcl-X L complexes, the RyR3 domain was cloned downstream of a chimeric cytokine receptor (RyR3 bait), consisting of the extracellular domain of the erythropoietin (Epo) receptor fused to the transmembrane and cytosolic part of the leptin receptor. In the latter, three tyrosines were mutated to phenylalanine to down regulate receptor signaling. Bcl-X L or the Bcl-X L K87D mutant were cloned downstream of a part of the glycoprotein 130 receptor (Bcl-X L or Bcl-X L K87D prey). If the Bcl-X L and Bcl-X L K87D prey constructs interact with the RyR3 bait construct, functional complementation of the chimeric cytokine receptor occurs, leading to ligand-dependent downstream STAT signaling. The latter is monitored via a luciferase reporter assay driven by a STATsensitive promoter. We also used the SV40 large antigen T (irrelevant prey) as a prey to monitor the signal representing the non-specific binding to RyR3. As a negative control, binding of the chimeric cytokine receptor without the RyR3 fragment (no bait) to the two Bcl-X L preys was also assessed. These MAPPIT results confirmed the data obtained via SPR and co-immunoprecipitation experiments, showing that Bcl-X L could interact with the RyR3 domain in a cellular context (Fig. 3A, top). Moreover, the Bcl-X L K87D mutant was severely impaired in interacting with the RyR3 domain without affecting its expression (Fig. 3A, bottom panel and Supplementary Fig. 1B for uncropped Western-blot images). No binding was detected when the RyR3 domain was not present in the bait vector (Fig. 3A, top panel), indicating that the interaction was specific.
The impact of mutating Lys87 into Asp was also examined in the context of the full-length RyR3 protein using co-immunoprecipitation experiments. Consistent with the MAPPIT data, 3XFLAGtagged Bcl-X L K87D displayed a reduced affinity for full-length RyR3 channels ( Fig. 3B and Supplementary Fig. 1C for uncropped Western-blot images).
Taken together, these data indicate that Bcl-X L , similarly to Bcl-2, binds via its BH4 domain to the same regulatory domain on RyR3. However, whereas for Bcl-2 the BH4 domain appears to be the main determinant for complex formation with RyR channels, it seems that for Bcl-X L both the BH4 domain and the BH3 domain, likely via Lys87, contribute to the interaction with RyR channels.
Bcl-X L , but not Bcl-X L K87D , inhibits RyR3-mediated Ca 21 release. Driven by the fact that Bcl-X L can bind to RyR3, we examined whether Bcl-X L could modulate RyR-mediated Ca 21 release (Fig. 4). Single-cell cytosolic [Ca 21 ] measurements in HEK RyR3 cells loaded with Fura-2-AM were performed (Fig. 4A). An empty vector (pCMV24) control, 3XFLAG-tagged Bcl-X L or the 3XFLAGtagged Bcl-X L K87D mutant were transiently transfected into the HEK RyR3 cells. An mCherry coding plasmid was co-transfected (at a 1:3 ratio) to identify transfected cells. After chelating extracellular Ca 21 with BAPTA (3 mM), caffeine (1.5 mM) was applied to induce RyRmediated Ca 21 release. Overexpression of 3XFLAG-tagged Bcl-X L inhibited caffeine-induced Ca 21 release compared to the empty vector control. The Bcl-X L K87D mutant failed to inhibit caffeineinduced Ca 21 release (Fig. 4B), correlating with its poor RyR3binding properties. To exclude that the observed reduction in caffeine-induced Ca 21 release upon Bcl-X L overexpression would have been due to an indirect effect via lowering of the Ca 21 -filling state of the endoplasmic reticulum (ER), we determined the amount of thapsigargin (1 mM)-releasable Ca 21 . This irreversible SERCA inhibitor causes a depletion of the ER Ca 21 stores and provides a good measure for the ER Ca 21 -store content. The ER Ca 21 -store content was not affected by overexpression of 3XFLAG-tagged Bcl-X L (Fig. 4C). This supports the view that Bcl-X L , similarly to Bcl-2, suppresses RyR-mediated Ca 21 release.
The BH4 domain of Bcl-X L by itself seems sufficient to inhibit RyR-mediated Ca 21 release. In order to assess whether the BH4 domain of Bcl-X L is sufficient for inhibiting RyR-mediated Ca 21 release, Fluo-3-AM loaded HEK RyR3 cells were loaded acutely with the BH4 domain of Bcl-X L , a control peptide or the vehicle via electroporation (Fig. 5A). The BH4 domain of Bcl-X L , but not a control peptide, suppressed caffeine (1 mM)-induced Ca 21 release. The BH4 domain of Bcl-X L inhibited caffeine-induced Ca 21 release in a concentration-dependent manner (Fig. 5B). This indicates that the BH4 domain of Bcl-X L was sufficient for inhibiting RyRmediated Ca 21 release.
We also assessed whether the BH4 domain of Bcl-X L could inhibit endogenous RyR channels by using 14-to 18-day-old rat hippocampal cultures known to express different RyR isoforms 23 . The experimental set-up was identical to the one previously used for characterization of the effect of the BH4 domain of Bcl-2 on native RyRs 16 . Cytosolic [Ca 21 ] was monitored in GCaMP3-expressing hippocampal neurons. The BH4 domain of Bcl-X L , a control peptide or the vehicle were introduced into the neurons via a patch pipette. After loading the neuron for five minutes with the peptides or Bcl-X L and its BH4 domain directly inhibit RyRs at the level of the ER. Bcl-X L and its isolated BH4 domain as a synthetic peptide inhibit the caffeine-induced [Ca 21 ] rise in the cytosol. Bcl-X L has also been implicated in the control of mitochondrial Ca 21 transport at the level of VDAC1. Bcl-X L was shown to inhibit Ca 21 uptake into the mitochondria 6,24 . However, it was also reported that Bcl-X L could stimulate mitochondrial Ca 21 uptake 25 . The latter effect could result in a decrease in caffeine-induced [Ca 21 ] rise in the cytosol. Therefore, we set out to document whether the decrease in caffeineinduced Ca 21 release in the cytosol by Bcl-X L is due to a decreased Ca 21 release from the ER or to an increased Ca 21 accumulation into the mitochondria. Direct ER-Ca 21 measurements were performed in HEK RyR3 cells utilizing a recently described green fluorescent CEPIA1 protein, that is targeted to the lumen of the ER (G-CEPIA1er) 26 . HEK RyR3 cells were transiently transfected with the empty vector (pCMV24) as control or with 3XFLAG-tagged Bcl-X L in combination with the G-CEPIA1er-encoding vector (at a 351 ratio). G-CEPIA1er-positive cells were selected and measurements were performed as in Fig. 4A. A typical average trace of one experiment and the quantification of all performed experiments are shown in Fig. 6A and B, respectively. These results indicate that overexpression of 3XFLAG-Bcl-X L suppressed the caffeineinduced Ca 21 release from the ER, supporting a model in which the inhibitory effect of Bcl-X L on RyR-mediated [Ca 21 ] rise in the cytosol occurs at least in part due to inhibition of the Ca 21 release from the ER. Finally, we set out to directly measure the effect of the BH4 domain of Bcl-X L on caffeine-induced mitochondrial Ca 21 entry. Rhod-FF-loaded HEK RyR3 cells were electroporated with either the vehicle (DMSO) or the BH4 domain of Bcl-X L (10 and 20 mM) and then stimulated with caffeine. Caffeine stimulation resulted in an increase in mitochondrial [Ca 21 ] (Fig. 6C). Compared to the vehicle control however, the BH4 domain of Bcl-X L potently inhibited the mitochondrial Ca 21 entry (Fig. 6 C, D). Furthermore, the effectiveness of BH4-Bcl-X L to inhibit caffeineinduced [Ca 21 ] rise in the mitochondria seemed higher than for inhibiting the caffeine-induced [Ca 21 ] rise in the cytosol, because 10 mM BH4-Bcl-X L inhibited caffeine-induced Ca 21 release in the cytosol by about 50% but inhibited caffeine-induced Ca 21 uptake in the mitochondria by about 90%. Taken together these data suggest that BH4-Bcl-X L likely inhibits, rather than stimulates, mitochondrial Ca 21 accumulation. This is consistent with our recent findings showing that BH4-Bcl-X L directly interacts with VDAC1 and suppressed VDAC1-mediated Ca 21 transfer into the mitochondria 27 . These experiments indicate that Bcl-X L can directly inhibit the caffeine-induced Ca 21 release at the level of the ER and potently inhibit mitochondrial Ca 21 uptake under these  Supplementary Fig. 1B. (B) Co-immunoprecipitations were performed in HEK RyR3 cells transiently overexpressing 3XFLAG-Bcl-X L or 3XFLAG-Bcl-X L K87D similarly as in Fig. 1A. Non-specific IgG antibodies were applied as negative controls. These experiments were performed at least three times utilizing each time independently transfected and freshly prepared HEK RyR3 cell lysates. All samples were run using the same experimental conditions and were derived from the same gel/blot, i.e. 3-8% tris-acetate gels for RyRs and 4-12% bis-tris gels for 3xFLAG-Bcl-X L . The double lines indicate that an additional empty lane separating the immunoprecipitated samples and the input samples was removed for the 3XFLAG-Bcl-X L blot. The uncropped image is shown in Supplementary Fig. 1C. www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 9641 | DOI: 10.1038/srep09641 experimental settings. We therefore conclude that the observed decrease in caffeine-induced Ca 21 release in the cytosol (Fig. 4 and  5) is mainly due to a direct inhibition of RyR3.

Discussion
The main conclusion of this paper is that Bcl-X L binds to and regulates RyR3 channels. Similarly to Bcl-2, Bcl-X L targets the central modulatory domain of the RyR protein, thereby suppressing RyRmediated Ca 21 release. Moreover, the BH4 domain of Bcl-X L was sufficient to inhibit both over-and endogenously expressed RyR channels in HEK293 cells or primary rat hippocampal neurons respectively. Consistent with this, the BH4 domain of Bcl-X L could bind to the purified RyR3 domain. However, the RyR3-binding efficiency of the BH4 domain of Bcl-X L seemed much lower than that of the BH4 domain of Bcl-2. Via an in silico superposition of the Bcl-2 and Bcl-X L crystal structures, a spatial overlap was observed between Lys17 in the BH4 domain of Bcl-2 and Lys87 in the BH3 domain of Bcl-X L : the positively charged e-amino groups of their side chains coincide in space. Consistent with the moderate RyR3-binding properties of the isolated BH4 domain of Bcl-X L , we found that Lys87 from Bcl-X L played a prominent role in binding to and regulating RyR3. The association of Bcl-X L with RyR channels and its functional implications appear to be very similar as the ones observed for Bcl-2, since i) RyR3/Bcl-X L binding is direct; ii) the binding of Bcl-X L to RyR3 occurs, at least in part, via the BH4 domain; iii) Bcl-X L overexpression inhibits RyR-mediated Ca 21 release; and iv) the BH4 domain of Bcl-X L is also sufficient to suppress RyR activity. These  findings correlate with the fact that the Bcl-2 K17D mutant and BH4-Bcl-2 K17D remain capable of binding to and regulating RyR channels, although this mutation changes the lysine critical for binding to the IP 3 R into the Asp11 residue in the BH4 domain of Bcl-X L 16 . This lack of selectivity between Bcl-2 and Bcl-X L may illustrate an important difference between IP 3 R-and RyR-mediated Ca 21 release. However, the binding of Bcl-X L versus Bcl-2 to RyRs in native tissues expressing RyRs ought to be further explored. In particular, it will be important to carefully analyze the Bcl-2-and Bcl-X L -expression levels in the relevant tissues and to determine whether a preferential binding of Bcl-2 or Bcl-X L to RyR channels exists in cells expressing both Bcl-2 and Bcl-X L . Despite these similarities, the molecular determinants underlying RyR/Bcl-X L -complex formation do not seem identical to those of Bcl-2, because the BH4 domain of Bcl-X L by itself displays rather moderate RyR3-binding properties. As a consequence, additional domains seem to be involved in RyR/Bcl-X L -complex formation. Here, we identified Lys87, located in the BH3 domain of Bcl-X L , as a critical determinant contributing to binding to and regulating RyR channels. Despite the importance of Lys87, the BH4 domain of Bcl-X L alone was able to suppress RyR activity.
The BH4 domain of Bcl-X L has been implicated in numerous studies to display strong anti-apoptotic and protective effects against a wide variety of insults and triggers, including in the heart 28-30 , endothelial cells 31,32 , blood cells [33][34][35] , pancreatic islets 36 and neurons 37 . Many of the cell types and tissues reported to benefit from the BH4 domain of Bcl-X L for their survival endogenously express RyR channels (cardiomyocytes, lymphocytes, pancreatic islets and neurons). Furthermore, in many apoptotic paradigms, reactive oxygen species (ROS) are implicated. ROS can impact the redox state and activity of the RyR channels (reviewed by Ref. 38). Mild increases in ROS moderately increase RyR activity by increasing its sensitivity for Ca 21 39 . However, severe ROS production associated with oxidative stress (e.g. in the context of ischemia/reperfusion injury) can lead to a continuously opening of the RyR channels, provoking an excessive Ca 21 leak from the ER or sarcoplasmic reticulum 40 . In the context of the heart, ROS has been clearly implicated to cause unzipping of the interdomain interactions critical for RyR2-channel stabilization 41,42 . During oxidative stress conditions, the BH4 domain of Bcl-X L may thus inhibit excessive RyR-mediated Ca 21 release from the intracellular Ca 21 stores in addition to exerting its protective effects at the mitochondria, thereby providing additional protection against cell death.
RyRs have important physiological functions in a variety of excitable cells and tissues, including skeletal muscle, cardiac muscle, neurons and pancreatic cells [43][44][45][46] . Furthermore, dysregulation of RyRs, either by somatic mutations or by altered expression levels, has been implicated in a variety of pathophysiological conditions, including malignant hyperthermia and central core disease 47,48 , cardiac diseases [49][50][51] and neurodegenerative diseases like Alzheimer's disease 52-54 and Huntington's disease 55 . At this point, the existence and physiological relevance of RyR/Bcl-2-and RyR/Bcl-X L -complex formation in these tissues and their potential disturbance in RyRassociated pathophysiologies will require further research.
In conclusion, our data further expand the number of Bcl-2-family members that are able to form protein complexes with RyR channels, thereby underpinning their critical role in regulating intracellular Ca 21 dynamics at the level of intracellular Ca 21 -release channels.

Methods
Chemicals, antibodies and peptides. Unless otherwise specified, all chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA). The following antibodies were used: mouse monoclonal anti-actin antibody, anti-FLAG M2 antibody and HRPconjugated anti-FLAG M2 antibody (Sigma-Aldrich), mouse monoclonal anti-RyR antibody 34C (Thermo Scientific, Rockford, IL, USA, or Developmental Studies Hybridoma Bank, University of Iowa, Iowa, USA) and mouse monoclonal anti-Bcl-X L antibody YTH-2H12 (Trevigen, Gaithersburg, WV, USA). The sequences of the peptides used in this study were: Biotin-BH4-Bcl-X L : Biotin-MSQSNRELVVDFLSYKLSQKGYSW (also used without the biotin tag) Biotin-scrambled BH4-Bcl-X L : Biotin-WYSKQRSLSGLVMYVLEDKNSQFS Control peptide: WYEKQRSLHGIMYYVIEDRNTKGYR These peptides were synthesized by Life Tein (Hillsborough, NJ, USA) with a purity of at least 85%.
Plasmids, constructs and protein purifications. 3XFLAG-Bcl-X L was obtained as previously described 19 . The 3XFLAG-Bcl-X L K87D mutant was obtained by PCR sitedirected mutagenesis utilizing the following primers: forward: 59ATCCCCATGGCAGCAGTAGATCAAGCGCTGAGGGAGGCA39, and reverse: 59TGCCTCCCTCAGCGCTTGATCTACTGCTGCCATGGGGAT39. The pCMV G-CEPIA1er containing plasmid was a gift from Dr. Masamitsu Iino (Addgene plasmid # 58215) 26 . The GST-IP 3 R1 domain 3 construct and the GST-RyR3 construct were obtained and purified as described 16 .
Cell culture, transfections and dissociated hippocampal cultures. All media and supplements added to the medium used in this paper were purchased from Life Technologies (Ghent, Belgium). HEK293 cells stably overexpressing RyR3 were cultured at 37uC in a 5% CO 2 incubator in a-Minimum Essential Medium supplemented with 10% fetal calf serum, 100 IU/mL penicillin, 100 mg/mL streptomycin, 2 mM glutamax and 800 mg/mL G418 56 . HEK293 cells were grown in Dulbecco's Modified Eagle Medium containing 4500 mg/L glucose, 10% fetal bovine serum and 50 mg/mL gentamicin 57 .
24 hours after seeding, the 3XFLAG-Bcl-X L or the 3XFLAG-Bcl-X L K87D mutant construct were introduced into the HEK RyR3 cells utilizing JETPrime transfection reagent (Polyplus Transfections, Illkirch, France) according to the manufacturer's protocol. 48 hours later the cells were harvested and lysed utilizing a CHAPS-based lysis buffer (pH 7.5, 50 mM Tris-HCl, 100 mM NaCl, 2 mM EDTA, 50 mM NaF, 1 mM Na 3 VO 4 , 1% CHAPS and protease inhibitor tablets (Roche, Basel, Switzerland)). For single-cell cytosolic [Ca 21 ] measurements the same constructs or the empty pCMV24 vector were introduced 48 hours after seeding in the HEK RyR3 cells utilizing X-tremeGENE HP DNA transfection reagent (Roche) according to the manufacturer's protocol. A pcDNA 3.1(-) mCherry expressing vector was co-transfected at a 153 ratio as a selection marker. For direct ER [Ca 21 ] measurements, the G-CEPIA1er construct was co-transfected (ratio 351) and used as selection marker instead of the mCherry expressing vector. Dissociated hippocampal cultures were obtained as described previously 58 . All animal experiments were performed according to approved guidelines. SPR analysis. SPR analysis was performed using a Biacore T200 (GE Healthcare, Diegem, Belgium). Immobilization to the streptavidin-coated sensor chip (BR-1005-31; GE Healthcare) and SPR measurements were performed as described previously 19 . NaOH (50 mM) with 0.0026% SDS was used as a regeneration buffer.
Co-immunoprecipitation experiments. Co-immunoprecipitation experiments were performed utilizing a co-immunoprecipitation kit (Thermo Scientific). RyR antibody or mouse IgG control antibody (Santa Cruz Biotechnology, Heidelberg, Germany) was immobilized according to the manufacturer's protocol. Gelatine was removed from the IgG control antibody utilizing a Pierce Antibody Clean-up Kit (Thermo scientific). Precleared HEK RyR3 lysates containing the 3XFLAG-Bcl-X L constructs (150 mg) were added to the resin to which the antibodies were immobilized and allowed to incubate overnight at 4uC. The next day, the resin was washed at least five times utilizing the CHAPS-based lysis buffer. The immune complexes were eluted by boiling (95uC) in 50 mL 23 LDS (Life Technologies) supplemented with 1/200 bmercaptoethanol for 5 min.
MAPPIT. The RyR3 domain was amplified by PCR using the following primers, forward: 59TAGTTGTCGACGAAGAGAGAAGTCATGGAGGA39, and reverse: 59TAGTTGCGGCCGCCTATTTGGTCCTCTCCACA39, and cloned in the pSEL12L bait vector 59 , using the restriction enzymes SalI and NotI. Bcl-X L was cloned in the pMG1-GW plasmid (prey vector) 22 using the Gateway recombination technology as described by the manufacturer (Life Technologies). Utilizing the same primers as described before, the Bcl-X L K87D mutation was also introduced in this construct via site directed mutagenesis. The MAPPIT analyses were done as previously described 22 with minor changes. Briefly, HEK293 cells were seeded in 96well plates. Six wells per condition were transfected with the different combinations of bait, prey and reporter plasmid (rPAP1-luci) using the calcium phosphate method. The next day, half of the wells were stimulated with 5 ng/mL Epo while the other half were left untreated. 24 hours later the cells were lysed and after the addition of substrate the luciferase activity was determined using a luminometer. The fold induction was obtained by dividing the average value of the stimulated cells by the average value of the non-stimulated cells.
Electroporation loading. Electroporation loading of HEK RyR3 cells was performed as previously described 16,60 .
Single-cell ER Ca 21 imaging. The G-CEPIA1er construct was introduced into HEK RyR3 cells as described above. A Zeiss Axio Observer Z1 Inverted Microscope equipped with a 203 air objective and a high-speed digital camera (Axiocam Hsm, Zeiss, Jena, Germany) were used for these measurements. Changes in fluorescence were monitored in the GFP channel (480/520 excitation/emission). To chelate extracellular Ca 21 , 3 mM BAPTA (Alfa Aesar, Ward Hill, MA, USA) was added. One minute later 1.5 mM caffeine was added to trigger RyR-mediated Ca 21 release. All traces were normalized (F/F 0 ) where F 0 is the starting fluorescence of each trace.
Single-cell mitochondrial Ca 21 imaging. HEK RyR3 cells were loaded for 30 min with 5 mM Rhod-FF-AM. Subsequently, cells were subjected to de-esterification over 15 min. During this time the BH4 domain peptides were introduced into the cells using the in situ electroporation technique 60 . Fluorescence-intensity changes in mitochondria were analyzed with custom-developed FluoFrames software. For each individual trace, the relative change of fluorescence (DF/F) was calculated. DF/F equals [F t -F 0 /F 0 ], with F 0 denoting the fluorescence before stimulation with caffeine and F t the fluorescence at different time points after caffeine stimulation. Subsequently, relative mitochondrial [Ca 21 ] changes were quantified as the area under the curve of the various Ca 21 traces.