The C Terminus of Bax Inhibitor-1 Forms a Ca2+-permeable Channel Pore*

Background: Evolutionary conserved Bax inhibitor-1 (BI-1) protects against ER stress-mediated apoptosis. Results: We identified a Ca2+-permeable channel pore in the C terminus of BI-1. Critical pore properties are an α-helical structure and two aspartate residues conserved among animals, but not among plants and yeast. Conclusion: C-terminal domain of BI-1 harbors a Ca2+-permeable channel pore. Significance: BI-1 has Ca2+ channel properties likely relevant for its function in ER stress and apoptosis. Bax inhibitor-1 (BI-1) is a multitransmembrane domain-spanning endoplasmic reticulum (ER)-located protein that is evolutionarily conserved and protects against apoptosis and ER stress. Furthermore, BI-1 is proposed to modulate ER Ca2+ homeostasis by acting as a Ca2+-leak channel. Based on experimental determination of the BI-1 topology, we propose that its C terminus forms a Ca2+ pore responsible for its Ca2+-leak properties. We utilized a set of C-terminal peptides to screen for Ca2+ leak activity in unidirectional 45Ca2+-flux experiments and identified an α-helical 20-amino acid peptide causing Ca2+ leak from the ER. The Ca2+ leak was independent of endogenous ER Ca2+-release channels or other Ca2+-leak mechanisms, namely translocons and presenilins. The Ca2+-permeating property of the peptide was confirmed in lipid-bilayer experiments. Using mutant peptides, we identified critical residues responsible for the Ca2+-leak properties of this BI-1 peptide, including a series of critical negatively charged aspartate residues. Using peptides corresponding to the equivalent BI-1 domain from various organisms, we found that the Ca2+-leak properties were conserved among animal, but not plant and yeast orthologs. By mutating one of the critical aspartate residues in the proposed Ca2+-channel pore in full-length BI-1, we found that Asp-213 was essential for BI-1-dependent ER Ca2+ leak. Thus, we elucidated residues critically important for BI-1-mediated Ca2+ leak and its potential channel pore. Remarkably, one of these residues was not conserved among plant and yeast BI-1 orthologs, indicating that the ER Ca2+-leak properties of BI-1 are an added function during evolution.

The endoplasmic reticulum (ER) 3 is the main intracellular Ca 2ϩ store and it plays a central role in intracellular Ca 2ϩ homeostasis and dynamics (1). The Ca 2ϩ concentration in the ER ([Ca 2ϩ ] ER ) is dynamically regulated by Ca 2ϩ -uptake pumps of the sarco/endoplasmic-reticulum Ca 2ϩ -ATPase family, intraluminal Ca 2ϩ -binding proteins, like calreticulin and calnexin, and Ca 2ϩ -release channels, like inositol 1,4,5-trisphosphate receptors (IP 3 Rs) and ryanodine receptors (RyRs) (2). In a resting cell, the steady-state [Ca 2ϩ ] ER is ultimately determined by the equilibrium between the active uptake and the basal or passive leak of Ca 2ϩ from the ER. The control of [Ca 2ϩ ] ER requires a tight handling of these mechanisms, because an increase in [Ca 2ϩ ] ER renders cells hypersensitive toward agonist-induced Ca 2ϩ signaling, whereas a decrease in [Ca 2ϩ ] ER may provoke ER stress and the activation of the unfolded protein response (3,4). The passive Ca 2ϩ leak has been one of the most enigmatic paradigms in Ca 2ϩ signaling, given the poor understanding of the molecular mechanisms and candidates responsible for the passive Ca 2ϩ leak (5). Recent work from the laboratories of Reed and Chae (6,7) demonstrated that Bax inhibitor-1 (BI-1) affects passive Ca 2ϩ leak from the ER. BI-1 has originally been identified as a suppressor of Bax-induced cell death in yeast (8) and was shown to affect the passive Ca 2ϩ leak from the ER (9, 10). BI-1 seems to be strongly evolutionarily conserved and BI-1 orthologs from plants can substitute for mammalian BI-1 in regard to its anti-apoptotic function (11). Besides this, other diverse functions of BI-1 have been described. BI-1 is a negative regulator of the ER-stress sensor IRE1␣ (12), it interacts with G-actin and increases actin polymerization (13), enhances cancer metastasis by altering glucose metabolism and by activating a sodium-hydrogen exchanger (14), and it reduces production of reactive oxygen species through direct interaction with NADPH-P450 reductase (15), a member of the microsomal monooxygenase system.
Recently, the role of BI-1 in Ca 2ϩ signaling has been further explored. The effect of BI-1 on cell death seems to involve changes in the amount of Ca 2ϩ that is releasable from intracellular stores (reviewed in Ref. 16). BI-1 physically interacts with Bcl-2 and Bcl-xL in the membrane of the ER (9), and may modulate the ER Ca 2ϩ homeostasis downstream of Bcl-xL (6). BI-1 overexpression led to a decrease in [Ca 2ϩ ] ER , whereas BI-1 Ϫ/Ϫ cells displayed increased [Ca 2ϩ ] ER (9). Bcl-xL overexpression led to a decrease in [Ca 2ϩ ] ER as well, but only in the presence of BI-1. Furthermore, BI-1 increased the permeability of ER membranes for Ca 2ϩ and mediated, in a pH-dependent manner, a Ca 2ϩ release from proteoliposomes reconstituted with purified BI-1 (7). The C-terminal region containing a lysine-rich motif (EKDKKKEKK) has been identified as a pH sensor of the BI-1 Ca 2ϩ channel (7), being essential for tetramerization and the Ca 2ϩ -release properties of BI-1.
Given the central role of the ER in controlling cell-survival and cell-death responses (17)(18)(19)(20)(21)(22)(23), it is reasonable to assume that the activity of BI-1 as a Ca 2ϩ channel can be an important determinant in these processes. However, the exact topology of BI-1 in the ER membrane and the position of the Ca 2ϩ pore are still unresolved. In this work, we present data on the topology of BI-1 in the ER membrane and results of screening a set of peptides representing the C-terminal hydrophobic part of BI-1 for their potential Ca 2ϩleak activity in unidirectional Ca 2ϩ fluxes. We identified an ␣-helical 20-amino acid peptide that is responsible for the Ca 2ϩ leak from the ER independently of other endogenous ER Ca 2ϩ -release channels and Ca 2ϩ -leak mechanisms, such as IP 3 Rs, RyRs, translocons, and presenilins. The Ca 2ϩ -flux properties of the peptide were confirmed in lipid bilayer experiments. Using mutant peptides, we identified critical negatively charged aspartate residues contributing to the potential Ca 2ϩ -channel pore and we show that Ca 2ϩ release properties of the BI-1 peptide require its hydrophobic and ␣-helical structure. Using peptides corresponding to BI-1 from other organisms, we found that Ca 2ϩ -transport properties are conserved among animal orthologs, but not in plant and yeast orthologs. Hence, this study identified a peptide derived from the C-terminal part of BI-1 that may be involved in the formation of a Ca 2ϩ -channel pore. Importantly, we found that replacing one of the critical aspartate residues in the proposed Ca 2ϩ -channel pore in the full-length BI-1 was sufficient to abolish the BI-1-mediated Ca 2ϩ leak.

EXPERIMENTAL PROCEDURES
Cell Culture-Mouse embryonic fibroblasts (MEF cells) were cultured at 37°C in a 9% CO 2 incubator in DMEM/Ham's F-12 medium supplemented with 10% fetal calf serum, 3.8 mM L-glutamine, 85 units ml Ϫ1 of penicillin, and 85 g ml Ϫ1 of streptomycin. All media were obtained from Invitrogen. DT40 cells were cultured as previously described (24).
For the 45 Ca 2ϩ fluxes, MEF cells were seeded in 12-well clusters (Costar, MA, 4 cm 2 ) at a density of ϳ10 4 to 2 ϫ 10 4 cells cm Ϫ2 . Experiments were carried out with confluent monolayers of cells (7.5 ϫ 10 4 cells cm Ϫ2 ) between the 5th and 7th day after plating. Lvec control cells were seeded in 12-well clusters (Greiner) at a density of 4 ϫ 10 4 cells/well, whereas IP 3 R1-overexpressing L15 cells were seeded at a density of 6 ϫ 10 4 cells per well. These fibroblast L cell lines have been previously created and characterized (25,26). Experiments were carried out on confluent cell monolayers between the 6th and 8th day after seeding.
Peptides and Chemical Reagents-All synthetic peptides were obtained from ThermoFisher Scientific (Ulm, Germany) at a purity of more than 90%. The identity of the peptides was confirmed via mass spectrometry and the purifications were performed via HPLC. 2-Aminoethoxydiphenyl borate, ryanodine, and anisomycin were obtained from Sigma.
Topology Assay-Neuro2A cells were transfected with hemagglutinin-tagged (HA) BI-1 (HA-BI-1), BI-1-HA, ERp44-HA, or empty vector in 10-cm cell culture dishes at a confluence of 80 -90% using the Attractene Transfection Reagent (Qiagen) according to the manufacturer's protocol. 24 h later, cells were harvested and the cell suspension was divided into three fractions. These fractions of suspended cells were permeabilized with either 20 M digitonin or 0.1% Triton X-100 in PBS for 30 min on a rocking platform at room temperature or were kept unpermeabilized in PBS. For incubation with Triton X-100, cells were first fixed with 4% paraformaldehyde for 10 min at 37°C. After permeabilization, cells were washed twice with either PBS containing 2 M digitonin or 0.01% Triton X-100, or in PBS alone, and stained with Alexa Fluor 488-conjugated anti-HA tag antibody (Cell Signaling Technology) diluted 1:20 in wash buffer for 30 min on a shaker protected from light. After two wash steps cells were transferred to FACS tubes and analyzed on a FACS Calibur (BD Biosciences). 45 Ca 2ϩ Fluxes-Unidirectional 45 Ca 2ϩ flux experiments were performed essentially as previously described (27)(28)(29). The 12-well clusters were fixed on a thermostated plate at 25°C on a mechanical shaker. The culture medium was aspirated, and the cells were permeabilized by incubating them for 10 min in a solution containing 120 mM KCl, 30 mM imidazole-HCl (pH 6.8), 2 mM MgCl 2 , 1 mM ATP, 1 mM EGTA, and 20 g ml Ϫ1 of saponin. The non-mitochondrial Ca 2ϩ stores were then loaded for 45 min in 120 mM KCl, 30 mM imidazole-HCl (pH 6.8), 5 mM MgCl 2 , 5 mM ATP, 0.44 mM EGTA, 10 mM NaN 3 , and 150 nM free 45 Ca 2ϩ (28 Ci ml Ϫ1 ). Then, 1 ml of efflux medium containing 120 mM KCl, 30 mM imidazole-HCl (pH 6.8), and 1 mM EGTA was added and replaced every 2 min. The indicated [IP 3 ] or 10 M Ca 2ϩ ionophore A23187 were added for 2 min, after 2, 10, or 20 min of efflux. At the end of the experiment, all 45 Ca 2ϩ remaining in the stores was released by incubation with 1 ml of a 2% (w/v) SDS solution for 30 min. Concentration-response curves were fitted using Origin 7.0 (Northampton, MA) software using the Hill equation.
Mag-Fluo4 Measurements in DT40 and TKO Cells-These experiments were performed as described (30,31). Briefly, DT40 and triple IP 3 R-knock out (TKO) cells were grown to a density of about 1 ϫ 10 6 cells/ml Ϫ1 and loaded with Mag-Fluo4 AM (20 M) for 60 min at 20°C. Cells were permeabilized by incubation with saponin (10 g⅐ml Ϫ1 ) for 4 min at 37°C. After centrifugation, cells were resuspended in Mg 2ϩ -free cytosollike medium (140 mM KCl, 20 mM NaCl, 1 mM EGTA, and 20 mM PIPES, pH 7.0), supplemented with 375 M CaCl 2 and 10 M carbonyl cyanide p-trifluoromethoxyphenylhydrazone, which inhibits mitochondrial Ca 2ϩ uptake. Cells were dispensed into a 96-well black-walled assay plate pre-coated with poly-L-lysine and centrifuged to spin the permeabilized cells to the bottom of the wells. Cells were incubated with 90 l of fresh Mg 2ϩ -free cytosol-like medium, supplemented with 375 M and prepared for microscopy. For recording cytosolic [Ca 2ϩ ], we used a Zeiss AxioObserver.Z1 microscope with a Fluor ϫ20/ 0.75 objective, Sutter Lambda DG-4 monochromator and filter sets 21 and 62 (Zeiss) for Fura-2 and mCherry, respectively. Images were recorded with an AxioCamHsM camera in the AxioVision version 4.6 software physiology module. mCherrypositive cells were preselected for the recordings, and Fura-2 fluorescence was recorded every 2 s. For the chelation of Ca 2ϩ , we used 3 mM 1,2-bis(2-aminophenoxyl)ethane-N,N,NЈ,NЈ-tetraacetic acid (BAPTA) (Invitrogen), which rapidly chelates the extracellular Ca 2ϩ . Release of Ca 2ϩ from intracellular stores was induced with 4 M thapsigargin, which specifically inhibits the activity of sarco/endoplasmic reticulum Ca 2ϩ -ATPase pumps at low micromolar concentrations. Ca 2ϩ traces were calibrated by incubating the cells for 10 min, first with Krebs-Ringer solution containing 50 mM EGTA and 2 M ionomycin (LC Labs), and then with 250 mM CaCl 2 and 2 M ionomycin. Traces were then converted into [Ca 2ϩ ] using the formula, where K d ϭ 225 nM (32).
Western Blot-Cells from 2-chamber slides were trypsinized and lysed in RIPA buffer (20 mM Tris, 150 mM NaCl, 1.5 mM MgCl 2 , 0.5 mM DTT, and 1% Triton X-100, pH 7.5) with 1ϫ protease inhibitor cocktail (Roche Diagnostics). Protein concentration was determined with the Bradford method and 10 g of total protein lysate was prepared in urea sample buffer (200 mM Tris, pH 6.8, 5% SDS, 0.1 mM EDTA, 8 M urea, 0.1% bromphenol blue, and 100 mM DTT). Samples were heated for 10 min at 37°C before loading on a precasted NuPAGE 4 -12% BisTris gel in MOPS buffer (Invitrogen). Proteins were transferred in running buffer (25 mM Tris, 190 mM glycine, 10% methanol) on a PVDF membrane. The membrane was blocked in TBS (10 mM Tris, 150 mM NaCl, pH 7.5) supplemented with 5% milk powder and 0.1% Tween 20. Expression of the Myc-BI-1 fusion protein was detected using an anti-Myc antibody (Sigma). The membrane was stripped with a mild stripping buffer (ThermoFisher Scientific) and reblotted with an antibody against glyceraldehyde-3-phosphate dehydrogenase (Sigma) as a loading control.
Electrophysiological Recordings from Planar Lipid Bilayers-Planar lipid bilayer were formed (3:1 mixture of phosphatidylethanolamine/phosphatidylserine lipids (Avanti Polar Lipids) dissolved in decane) on a 200-m diameter aperture separating cis and trans chambers containing: (cis) 95 mM Tris-OH, 200 mM HEPES (pH 7.3, ϳ300 mosmol; [Ca 2ϩ ] free ϭ 10 nM) and (trans) 50 mM Ba(OH) 2 , 55 mM HEPES (pH 7.3, ϳ285 mosmol) solutions, respectively. Peptides dissolved in dimethyl sulfoxide were added to the cis chamber at gradually increasing concentrations, and stirred continuously between the periods of membrane current recordings that were done using the Axopatch 1 amplifier and pClamp6 software (Axon Instruments). The cis chamber was grounded. Experiments were done at room temperature. Holding potential (V h ) was 0 mV for all experiments except when indicated. Off-line current-trace filtering (500 Hz), calculations of the mean current passing by lipid bilayer (I mean ), and histogram calculations and statistics analysis were done using pClamp10 and OriginPro 7.5 (OriginLab) software.

RESULTS
The Seventh Transmembrane Domain of BI-1 Resembles a Pore-The attempt to attribute a Ca 2ϩ -channel function to BI-1 is hampered by unresolved exact topology of the BI-1 in the ER membrane. We have previously shown that the C terminus of BI-1 resides in the cytosol (16), which results in two different models (Fig. 1A). In model A, which is supported by TMpred, the N terminus is located in the ER lumen. In model B, which is bioinformatically supported by TMHMM, the N terminus is cytosolic and the semi-hydrophobic seventh transmembrane domain, which bears some homology to the loop domain of the voltage-gated sodium channel SCN5A, would dip into the membrane, creating a loop domain (Fig. 1B). Unfortunately, it is not possible to use the enhanced green fluorescent protein protease-protection assay used previously (16) to distinguish between these two models, because the addition of an N-terminal enhanced green fluorescent protein alters the function of BI-1, 4 indicating that the bulky enhanced green fluorescent protein moiety at the N terminus might alter the topology or intracellular localization of BI-1.
We therefore modified the assay by using cells overexpressing proteins tagged with the very short HA tag and selective membrane permeabilization. Digitonin makes epitopes in the cytosol, but not in the ER lumen, accessible to fluorescent antibodies, which can be quantified by flow cytometry. Treatment with Triton X-100, in contrast, permeabilizes all membranes and leads to staining of both luminal and cytosolic epitopes. We overexpressed HA-BI-1 or BI-1-HA in Neuro2a cells and used the luminal endoplasmic-reticulum protein (ERp) 44-HA as negative control (Fig. 1C). Both HA-BI-1 and BI-1-HA resulted in a similar increase in immunoreactivity after digitonin treatment, whereas ERp44 staining never exceeded the level apparent in cells transfected with an empty vector. Permeabilization with Triton X-100, on the other hand, dramatically increased HA immunoreactivity in ERp44-HA-transfected cells and had no further significant effect on HA-BI-1 and BI-1-HA. These results favor model B, although it cannot be completely excluded that the fusion with HA itself may alter the topology of BI-1. Fig. 1 suggested a possibility of the presence of a pore domain right before the proposed C-terminal domains of BI-1 interacting with IRE1␣ (12), G-actin (13), and NADPH-P450 reductase (15), and forming the pH-sensor domain (7). Hence, we developed a set of two overlapping peptides named CTP1 (HGDQDYIWHCIDLFLD-FITV) and CTP2 (CIDLFLDFITVFRKLMMILAMNEKD) covering the residues following the 6th transmembrane domain, but excluding the pH-sensor domain of BI-1 (Fig. 1B).

A C-terminal Peptide Representing the Putative Pore Domain of BI-1 Causes Ca 2ϩ Release from the ER Ca 2ϩ Stores-The model supported by the experiments in
The Ca 2ϩ -flux properties of these peptides were examined in unidirectional 45 Ca 2ϩ -flux assays in saponin-permeabilized MEF cells. This assay provides an accurate assessment of Ca 2ϩflux properties of channels under unidirectional conditions (28,29,33,34). In short, after permeabilization of the MEF cells, 45 Ca 2ϩ was loaded to a steady-state in the non-mitochondrial Ca 2ϩ stores, which correspond largely to the ER. After irreversible inhibition of the Ca 2ϩ pumps by thapsigargin, the Ca 2ϩ leak from these stores can be measured under unidirectional conditions. The Ca 2ϩ content of the stores gradually decreased with time under these conditions ( Fig. 2A). The released Ca 2ϩ divided by the amount of Ca 2ϩ that was present in the store at that time (ϭ fractional loss), was determined and plotted (Fig.  2B). This analysis provides an accurate quantitative analysis of the Ca 2ϩ leak from the ER. In control conditions, the Ca 2ϩ content of the stores gradually decreased ( Fig. 2A, vehicle), whereas addition of 60 M CTP1, but not of CTP2, accelerated the decrease in store Ca 2ϩ content ( Fig. 2A) and resulted in an increased Ca 2ϩ leak from the ER (Fig. 2B). The concentrationdependent effects of CTP1 on the Ca 2ϩ leak were obtained by adding different concentrations of CTP1 (5, 10, 20, 40, and 80 M) after 10 min of efflux (Fig. 2C). These effects were normalized for the maximal releasable Ca 2ϩ , obtained by the addition of 10 M A23187, a Ca 2ϩ ionophore. The dose-response curve (Fig. 2D) was fitted by a logistic curve fitting, yielding an EC 50 value of about 30 M and a power (steepness of the curve) of 3. This indicates that CTP1-induced Ca 2ϩ release is highly cooperative, which is compatible with an oligomerization-dependent process and suggests that at least three CTP1 peptides are required to form a functional Ca 2ϩ -channel pore. 3 Rs, RyRs, Presenilins, and Translocon-To elucidate whether the Ca 2ϩ release mediated by CTP1 was caused by modulation of BI-1 or other documented Ca 2ϩ -release mechanisms, we examined the Ca 2ϩ -flux properties of CTP1 in the absence of BI-1, IP 3 R, RyR, presenilin, and translocon activity. First, we used primary BI-1 knock-out MEF cells and found that the Ca 2ϩ release mediated by CTP1 did not differ between primary wild-type and BI-1 Ϫ/Ϫ MEF cells (Fig. 3A). Second, using inhibitors of IP 3 Rs (100 M 2-aminoethoxydiphenyl borate), RyRs (100 M ryanodine), and the translocon (100 M anisomycin), we found that the Ca 2ϩ -release properties of CTP1 were not affected by these inhibitory compounds (Fig. 3B). Third, a recent study proposed presenilins as important Ca 2ϩ -leak channels from the ER (35), although direct modulation of IP 3 R activity by presenilins may also contribute to the effect (36 -38). However, we excluded the contribution of presenilins by examining the Ca 2ϩ -release properties of CTP1 in presenilin-1/2 double knock-out MEF cells (Fig. 3C). CTP1 mediated similar increases in the Ca 2ϩ flux from the ER of wild-type and prese- A, permeabilized MEF cells loaded to steady-state with 45 Ca 2ϩ were incubated in Ca 2ϩ -free efflux medium and their Ca 2ϩ content, expressed in counts per min (cpm), monitored as a function of time. In the time period indicated by the line, vehicle, CTP1, and CTP2 were added for 8 min. CTP1, but not CTP2, led to a decrease in the Ca 2ϩ content. B, results obtained from the experiment described in A were plotted as "fractional loss" (%/2 min) as a function of time. CTP1, but not vehicle or CTP2, provoked an increase in the fractional loss. C, different concentrations of CTP1 were added during the efflux phase for 6 min. Results were plotted as fractional loss as a function of time. D, CTP1-induced Ca 2ϩ release from 45 Ca 2ϩ -loaded permeabilized cells was quantified as the difference in fractional loss between 2 min after application of CTP1 and before CTP1, i.e. the fractional loss at 10 min minus the fractional loss at 8 min. Concentrationresponse curve for CTP1-induced Ca 2ϩ release was obtained from three independent experiments. All results were normalized to the Ca 2ϩ released by the Ca 2ϩ ionophore, A23187, and expressed as % of total A23187-induced Ca 2ϩ release.

CTP1-induced Ca 2ϩ Release Is Independent of the Presence of BI-1, IP
nilin-1/2 double knock-out MEF cells, indicating that the Ca 2ϩrelease properties of CTP1 were not due to modulation or activation of presenilin Ca 2ϩ -channel activity. Fourth, the contribution of IP 3 Rs in mediating CTP1-induced Ca 2ϩ release was further directly assessed by using L15 fibroblasts overexpressing IP 3 R1 and DT40 TKO cells. By analyzing the CTP1- induced Ca 2ϩ release from the ER of permeabilized fibroblasts either expressing an empty vector (Lvec) or overexpressing IP 3 R1 (L15), we found that CTP1 mediated similar increases in the Ca 2ϩ -leak rate from the ER of control fibroblasts and IP 3 R1overexpressing fibroblasts (Fig. 3D). Next, we monitored the (Ca 2ϩ ) ER by using ER-trapped Mag-Fluo4 in permeabilized TKO cells (Fig. 3E) and TKO cells ectopically overexpressing IP 3 R1 using an automated Ca 2ϩ assay (Fig. 3F). After loading the ER Ca 2ϩ stores to steady-state by adding ATP for active Ca 2ϩ uptake, an unidirectional Ca 2ϩ leak from the ER was initiated by adding thapsigargin, and was measured by monitoring the ER Ca 2ϩ concentration using the low-affinity Ca 2ϩ dye Mag-Fluo4. Then, CTP1 was added and the change in the Ca 2ϩ -leak rate was measured. Consistent with our previous experiments, we found that CTP1 enhanced the Ca 2ϩ -leak rate in TKO cells and in TKO cells ectopically expressing IP 3 R1. These results rule out that the effect of CTP1 is mediated through IP 3 R Ca 2ϩ -release channels. Taken together, these experiments indicate that the mechanism of CTP1-induced Ca 2ϩ release does not involve BI-1 or other documented Ca 2ϩ -leak or release mechanisms, like IP 3 Rs, RyRs, translocon, or presenilins.
Source of CTP1-induced Ca 2ϩ Release Is IP 3 -sensitive Ca 2ϩ Store-The ER, the main intracellular Ca 2ϩ store, is IP 3 sensitive. Hence, we investigated whether depleting the IP 3 -sensitive Ca 2ϩ store by adding an activating concentration of IP 3 would abolish the CTP1-induced Ca 2ϩ release. Importantly, the ability of CTP1 to mediate Ca 2ϩ release was almost completely abolished after depletion of the IP 3 -sensitive stores (Fig. 4A), indicating that the main source of the CTP1-induced Ca 2ϩ release largely corresponds to the IP 3 -sensitive store. This was further supported by experiments in which we compared the CTP1-induced Ca 2ϩ release relative to the Ca 2ϩ release induced by a maximal concentration (50 M) of IP 3 (Fig. 4B). This analysis showed that the maximal amount of releasable Ca 2ϩ by a saturating concentration of CTP1 (80 M) was Ͼ80% of the maximal amount of releasable Ca 2ϩ by 50 M IP 3 .
Second, because the predicted secondary structure of the CTP1 peptide displays an ␣-helical structure (Fig. 5B) with a high hydrophobicity index (second half of CTP1; Fig. 5C), we examined whether the ␣-helical properties and residues in the hydrophobic stretch were important to promote Ca 2ϩ release (Fig. 5D). Hence, we analyzed the double mutant CTP1-L210A,F211A in the unidirectional 45 Ca 2ϩ -flux assay. This peptide contains two altered residues in its hydrophobic stretch, but retains its ␣-helical properties (Fig. 5B). The mutated residues slightly reduce the hydrophobicity score of the second half of CTP1 (Fig. 5C). We found that CTP1-L210A,F211A was not able to induce Ca 2ϩ release. Similar results were found with CTP1-L210G,F211G, which does not possess an ␣-helical structure (Fig. 5B). To examine whether the ␣-helical structure of CTP1 is essential for its activity, we created CTP1-C207G,I208G. This peptide contains the essential Leu-210 and Phe-211 residues, but it has a largely altered secondary structure (Fig. 5B). Importantly, this peptide was also not able to promote Ca 2ϩ release. Finally, we developed a control peptide CTP1-C207A, which retains its ␣-helical structure (Fig. 5B) and its hydrophobicity score (Fig. 5C). As a consequence, this peptide was able to mobilize Ca 2ϩ (Fig. 5D). These data indicate that both residues in the hydrophobic stretch as well as the ␣-helical structure of CTP1 are essential for its activity. Because the CTP1-L210A,F211A peptide still retains a very high hydrophobicity score, we hypothesize that the Leu-210 and Phe-211 residues may contribute to the oligomerization of the peptide in the membrane, which may be required for CTP1mediated Ca 2ϩ mobilization from the ER.

CTP1 Displays Channel-like Activity with Permeability for Divalent Cations-Our experiments indicated that CTP1
induces Ca 2ϩ -channel activity in ER membranes. Hence, to unequivocally demonstrate that CTP1 is capable to form divalent cation-permeable channel pores, we examined, using Ba 2ϩ as a permeant ion, whether the CTP1 incorporation into planar lipid bilayers would initiate ionic currents transmitted by divalent cations. In a separate set of experiments, either CTP1 or CTP1-L210A,F211A mutant peptides were added at 10 M increment concentrations to the cis compartment of the bilayer chamber containing a HEPES/Tris solution (free [Ca 2ϩ ] buffered at 10 nM), whereas the opposite trans chamber contained 50 mM Ba 2ϩ . The exposure of the lipid bilayer to CTP1 incrementing from 10 to 30 M for several minutes for each concen-tration was without effect, but the further increment to 40 M resulted in the appearance of a progressively growing Ba 2ϩ current (Fig. 6A, top). The current developed from zero to more than 20 pA within 3-8 min, with longer exposure resulting in disintegration of the bilayer, which prevented careful measurements of the concentration-response relationship at higher concentrations due to a continuously increasing nonstationary current. Such a current was not observed when the CTP1-L210A,F211A mutant peptide was tested at concentrations up to 50 M (Fig. 6A, bottom). The threshold-like appearance of the CTP1-induced current might be explained by a model where CTP1 accumulation to a critical concentration within a lipid bilayer initiates a chain reaction of formation of ion-conducting channel assemblies. The biophysical properties of the CTP1-forming channels were not studied in detail due to the limited period of observation of clearly distinguishable single- channel openings before appearance of the complicated fluctuating pattern of the CTP1-multichannel activity (Fig. 6B). The current amplitudes analyzed during the first minute of quasistationary observation of CTP1 channel activity yielded a broadly distributed shallow peak on a histogram compared with a zero-centered peak for the CTP1-L210A,F211A mutant peptide, which revealed no channel activity for the latter (Fig.  6C). To obtain a meaningful quantitative parameter, a mean current recorded during the first minute of observation of the channel activity in the presence of 40 M CTP1 was calculated in multiple bilayers and compared with the mean currents recorded in the presence of 50 M CTP1-L210A,F211A (Fig.   FIGURE 6. CTP1 displays ion channel-like activity in planar lipid bilayers. A, traces of currents from two representative experiments attempting incorporation of the CTP1 (top) and CTP1-L210A,F211A (bottom) peptides into the artificial lipid bilayers. The peptide-concentration increments are indicated by arrows. Breaks within traces correspond to 1-min cis chamber solution stirring intervals after peptide additions. B, higher resolution (4-s long) fragments of the CTP1-generated currents at the time points corresponded to the numbered circles in A. C, current sublevel-probability histograms obtained from continuous 60-s long fragments of traces presented in A during the first minute of the channel activity appearance for CTP1 (40 M) and during prolonged incubation with CTP1-L210A,F211A (50 M). D, values of the mean current (I mean Ϯ S.E.) calculated from 60-s long recordings and averaged for four experiments for each peptide. For CTP1, the 60-s recordings were taken immediately after appearance of the initial series of spikes indicating formation of the CTP1-assembled channels (*, p Ͻ 0.05). E, representative 4-s long recordings obtained at various membrane potentials (shown on the right) from the same planar lipid bilayer within an interval of quasistable basal current carried by CTP1-induced (40 M) ion channels in asymmetric bilayer ionic conditions (cis, HEPES/Tris; trans, HEPES/Ba 2ϩ , 50 mM, see "Experimental Procedures").

6D)
, showing a significant ion-transporting capability of the CTP1-forming channels. To confirm the identity of the current-carrying ions in our asymmetric ionic conditions, varying the polarity of the membrane voltage during an interval of a relatively stable level of the CTP1-induced (40 M) currents revealed a voltage-dependent current rectification, indicating a high permeability for Ba 2ϩ and negligible permeability for large cations (Tris) or for large anions (HEPES) by CTP1-formed ion channels (Fig. 6E). The observation of a threshold concentration and time dependence of these bilayer experiments suggest that accumulation in the membrane and subsequent oligomerization of the peptide may be required for its pore properties. To underpin these observations, we performed additional biochemical experiments, showing that N-terminal biotinylated CTP1 was able to accumulate in ER microsomal membranes from HeLa cells within 5 min (supplemental Fig. 1A). The Ca 2ϩ -flux properties of biotin-CTP1 were confirmed in 45 Ca 2ϩ -flux assays (supplemental Fig. 1B). As a control, we used a biotin-CTP1 mutant peptide where four hydrophobic residues were replaced by arginines and found that this peptide did not accumulate in ER microsomal membranes and did not provoke ER Ca 2ϩ release. Importantly, whereas biotin-CTP1 D209A did not cause ER Ca 2ϩ flux, it potently accumulated in ER microsomal membranes, indicating that its lack of Ca 2ϩflux properties is not caused by deficient accumulation in ER membranes, but rather due to its lack to form Ca 2ϩ -permeable channel pores. Finally, we also found that biotin-CTP1 peptides may form oligomers in ER microsomal membranes of HeLa cells (supplemental Fig. 1C). These observations correlate well with the high cooperativity (steepness coefficient ϭ 3) found in the CTP1 concentration-response curve in the 45 Ca 2ϩ -flux assay (Fig. 2, C and D).
Ca 2ϩ -flux Properties of CTP1 Are Evolutionarily Conserved in Animal but Not in Plant or Yeast Orthologs of BI-1-Because BI-1 is a highly evolutionarily conserved protein with orthologs in yeast, plants, and animals, we examined the Ca 2ϩ -flux properties of CTP1 obtained from its yeast, plant, and animal BI-1 orthologs (Fig. 7A). Using unidirectional 45 Ca 2ϩ -flux assays, we found that all CTP1 peptides derived from animal BI-1 orthologs were able to induce Ca 2ϩ release, whereas CTP1 pep-tides derived from plant or putative yeast BI-1 orthologs were not (Fig. 7B). Interestingly, CTP1 from the zebrafish Danio rerio was less potent than CTP1 from other vertebrates. This may be due to the fact that the predicted propensity of zebrafish CTP1 to form an ␣-helix was reduced compared with mammalian CTP1 (supplemental Fig. S2). These observations again underpin the importance of the ␣-helical properties to form functional Ca 2ϩ -channel pores. In addition, we also analyzed the Arabidopsis thaliana CTP1 and found a high predicted probability for ␣-helical structure formation. This may indicate that its lack to induce Ca 2ϩ release is not due to structural defects but presumably to the lack of the critical Asp residue at position 12 of the peptide. The CTP1 region of the animal BI-1 orthologs seems not to be conserved in the yeast BI-1 ortholog and a recent phylogenetic analysis also suggested that this protein is in fact closer related to the BI-1 paralog TMBIM4 (16). In plant BI-1 orthologs, the CTP1 region is ϳ65% conserved, but CTP1 plant orthologs were, nevertheless, not able to provoke Ca 2ϩ release. Hence, the Ca 2ϩ channel-like activity of the CTP1 domain of BI-1 seems to be a property acquired later in evolution and is only present in animals.
Asp-209 and Asp-213 Are Key Residues in CTP1 Ca 2ϩ -Channel Pore-CTP1 contains four negatively charged residues (Asp-200, Asp-202, Asp-209, and Asp-213). Importantly, CTP1-D202R,D213R and CTP1-D200R,D209R did not display Ca 2ϩ -flux properties (Fig. 8A). To have a systematic assessment of the importance of each aspartate residue for the Ca 2ϩ -mobilizing properties of CTP1, we analyzed four single CTP1 mutant peptides (40 M), in which each respective aspartate residue was changed into an alanine residue (Fig. 8, B and C). Remarkably, D209A and D213A mutations in the CTP1 peptide completely abolished its Ca 2ϩ -flux properties. D202A only partially inhibited the Ca 2ϩ -flux properties of CTP1, whereas Asp-200 did not contribute to the Ca 2ϩ -flux properties of CTP1. This indicates that the negative side chains of Asp-209 and Asp-213 are directly involved in the Ca 2ϩ flux through the ER membranes, whereas Asp-202 might serve as an acceptor residue at the cytosol/ER membrane interface. In fact, these models correlate well with the predicted BI-1 structure, in which Asp-209 and Asp-213 are located in the loop domain inside the ER Importantly, comparing animal CTP1 orthologs with plant CTP1 orthologs, we found that Asp-209, which is essential for the Ca 2ϩ -flux properties of human CTP1, is conserved among the animal CTP1 orthologs, but not among plant CTP1 orthologs, where it corresponds to a threonine residue. Therefore, we assessed human CTP1-D209T and found that this peptide also lacked Ca 2ϩ -flux properties (Fig. 8, B and C). Hence, these data indicate that the appearance of Ca 2ϩ -flux properties of animal CTP1 orthologs is probably due to the change of a critical plant threonine to an aspartate residue.
In Contrast to Full-length Wild-type BI-1, BI-1-D213R Is Not Able to Reduce ER Ca 2ϩ Content-Next, we assessed the importance of the critical Asp-213-residue in the CTP1 region of full-length BI-1 for its Ca 2ϩ leak and cell death-inhibiting function. For Ca 2ϩ analysis, we co-transfected mCherry plasmid with empty vector, myc-BI-1, or myc-BI-1-D213R in HeLa cells loaded with Fura-2-AM. Only Ca 2ϩ signals in cells containing the mCherry plasmid as a marker for successful transfection were analyzed. The Fura-2 ratio (F 340 /F 380 ) was calibrated to obtain [Ca 2ϩ ]. We also verified similar expression levels of Myc-BI-1 and Myc-BI-1-D213R using Western blotting with anti-Myc antibodies (Fig. 9A). In agreement with previous observations (6,7), wild-type Myc-BI-1 significantly reduced the amount of thapsigargin-releasable Ca 2ϩ (p Ͻ 0.005; n ϭ 68 cells), confirming that BI-1 lowers [Ca 2ϩ ] ER by acting as a Ca 2ϩleak channel (Fig. 9, B and C). Remarkably, similarly to mock transfected cells (n ϭ 61 cells), Myc-BI-1-D213R did not affect the thapsigargin-releasable Ca 2ϩ (n ϭ 64). These data indicate that Asp-213 is an essential residue for the Ca 2ϩ leak property of BI-1.

DISCUSSION
The main findings of this contribution are the identification and characterization of the presumed Ca 2ϩ -channel pore of BI-1, its molecular determinants, and its functional properties using a peptide-based approach. We found that the C-terminal part of BI-1 harbors a Ca 2ϩ -channel pore, mediating Ca 2ϩ leak from the IP 3 -sensitive pool of the ER into the cytosol. We propose that this action of the C-terminal part of BI-1 is independent of other Ca 2ϩ -release channels, because the Ca 2ϩ -flux properties of the C-terminal peptide of BI-1 were observed (i) in artificial lipid-bilayer experiments, (ii) in permeabilized DT40 cells lacking all three IP 3 R isoforms, (iii) in the presence of  Two-tailed t test shows significant difference between mock-and myc-BI-1 (p Ͻ 0.005) as well as myc-BI-1 and myc-BI-1-D213R-transfected cells (p Ͻ 0.005), but not between mock-and BI-1-D213R-transfected cells (p ϭ 0.80).
known inhibitors of intracellular Ca 2ϩ release/leak channels, like IP 3 Rs, RyRs, and translocon, (iv) in permeabilized MEF cells lacking presenilins-1/2 Ca 2ϩ -leak channels, and (v) in permeabilized BI-1 Ϫ/Ϫ MEF cells. Furthermore, replacing the critical Asp-213 by an arginine in full-length BI-1 completely abolished the ability of BI-1 to reduce the ER Ca 2ϩ content. Hence, we conclude that the C-terminal part of BI-1 is involved in the formation of the Ca 2ϩ -channel pore and is responsible for the Ca 2ϩ -channel activity previously reported in a number of studies. Importantly, this property is conserved in animal BI-1 orthologs, but not in the plant and yeast BI-1 orthologs, suggesting that the Ca 2ϩ -leak channel function has been added later on during the evolution of BI-1.
We had previously shown that BI-1 overexpression reduces cytosolic and mitochondrial Ca 2ϩ signals in response to thapsigargin or to extracellular agonists, such as ATP, that activate IP 3 signaling, which we attributed to a decrease in the steadystate [Ca 2ϩ ] ER . Importantly, this study was the first to indicate the importance of the C-terminal tail of BI-1, because overexpression of BI-1 lacking its C-terminal tail did not decrease steady-state [Ca 2ϩ ] ER (10). The regulation of basal [Ca 2ϩ ] ER was confirmed in other reports (6) where [Ca 2ϩ ] ER and Ca 2ϩleak rate were directly monitored, revealing an increased Ca 2ϩleak rate from the ER caused by BI-1 overexpression. These effects of BI-1 were downstream of the anti-apoptotic Bcl-xL, which has been reported to lower [Ca 2ϩ ] ER in some conditions or cell types and is independent of the presence of Bax/Bak (6). Finally, recent studies demonstrated that purified BI-1 displays Ca 2ϩ /H ϩ antiporter-like activity in reconstituted proteoliposomes and increases the permeability of ER membranes in a pH-dependent manner (7,39). The ability of BI-1 to increase the Ca 2ϩ permeability of ER membranes and protect against apoptosis was critically dependent on the last 9 amino acids (EKDKKKEKK), which resemble the lysine-rich pH-sensing domain of other ion channels and was proposed to act as the pH sensor of BI-1 (7,11). In addition, other stimulators of the BI-1 Ca 2ϩ /H ϩ -antiporter activity have recently been identified and include lipids, such as cardiolipin and phosphatidylserine, and protein domains, such as the BH4 domain of Bcl-2 and Bcl-xL (40). Their stimulatory action on BI-1-mediated Ca 2ϩ release seems to involve enhanced oligomerization of the BI-1 protein into tetrameric complexes. Nevertheless, until now, these studies did not reveal the Ca 2ϩ channel pore of BI-1. The CTP1 sequence identified in our study is a likely candidate, as for most ion channels the channel pore is formed by ␣-helical transmembrane domains or possesses at least a certain degree of hydrophobicity, allowing (partial) membrane insertion. In addition, for cation channels, the channel pore is lined by negatively charged residues allowing the passage of positively charged ions through electrostatic interactions. Finally, we also suggest that CTP1 needs to oligomerize for its pore-forming activity, given the high cooperativity (the coefficient of 3) of the CTP1-induced Ca 2ϩ release, which we quantitatively assessed in unidirectional 45 Ca 2ϩ fluxes. This was also reflected in lipidbilayer experiments, in which CTP1-channel activity was timedependent and required a threshold concentration. Therefore, we propose that CTP1 has all the characteristics and properties of a bona fide Ca 2ϩ -channel pore. Taken together, our findings are consistent with a 6-transmembrane model for BI-1 in which part of the C-terminal region behaves as a pore-forming domain and the N terminus is located in the cytosol. The N terminus of BI-1 was recently shown to interact in yeast with an anti-apoptotic protein from enteropathogenic Escherichia coli, NleH (41), which would also suggest a cytosolic localization of the N terminus assuming that NleH is a cytosolic protein.
Finally, changes in BI-1 expression have been associated with a variety of pathophysiological conditions and may have beneficial or detrimental effects (42,43). On the one hand, endogenous BI-1 seems to be important to protect neurons against glucose-oxygen deprivation and to limit tissue damage in liver upon ischemic reperfusion injury. Moreover, BI-1 expressed in the liver seems to protect against hepatic ER stress induced by obesity-associated insulin resistance and glucose intolerance by inhibiting IRE1␣ and downstream unfolded protein responses. As such, BI-1-gene transfer may be a therapeutic tool to target type-2 diabetes (42,43). On the other hand, up-regulation of BI-1 levels has been associated with a number of cancers, including cancers of the prostate (44), breast, ovary, and uterus (45), peripheral adenocarcinomas in the lung (46) and human nasopharyngeal carcinoma (47). Moreover, overexpression of BI-1 in NIH3T3 cells promotes cell growth, cell transformation, and tumorigenesis in vivo (48). Very recently, BI-1 was shown to suppress autophagy through controlling IRE1␣ activity (49). However, in many of these cases, it is not known whether these physiological and pathophysiological responses are related to changes in BI-1 expression and whether its function involves its Ca 2ϩ -channel activity. Thus, the identification of the Ca 2ϩchannel pore of BI-1 and the creation of the BI-1 mutant deficient for mediating a Ca 2ϩ leak from the ER allows to identify the Ca 2ϩ -dependent function of BI-1 in a plethora of paradigms, including its role in protection against ER stress, negative regulation of ER stress through IRE1␣, and apoptosis.