Research Article

Structural basis for a pH-sensitive calcium leak across membranes

See allHide authors and affiliations

Science  06 Jun 2014:
Vol. 344, Issue 6188, pp. 1131-1135
DOI: 10.1126/science.1252043

Allowing calcium to leak across a membrane

Cells maintain a balance between calcium in the cytosol and calcium stored in organelles—too much stored calcium kills cells. Transmembrane Bax inhibitor motif (TMBIM) proteins form channels in organelle membranes that allow calcium to leak out. Chang et al. show that this calcium leak is pH-dependent. A bacterial homolog of TMBIM proteins converts between an open channel at low pH and a closed channel at high pH. Although the channel is open at low pH, calcium leakage is low because the inside of the channel remains at a neutral pH. Thus, at physiological pH, these channels will be in equilibrium between the open and closed states, so that excess calcium can leak through.

Science, this issue p. 1131

Abstract

Calcium homeostasis balances passive calcium leak and active calcium uptake. Human Bax inhibitor–1 (hBI-1) is an antiapoptotic protein that mediates a calcium leak and is representative of a highly conserved and widely distributed family, the transmembrane Bax inhibitor motif (TMBIM) proteins. Here, we present crystal structures of a bacterial homolog and characterize its calcium leak activity. The structure has a seven-transmembrane-helix fold that features two triple-helix sandwiches wrapped around a central C-terminal helix. Structures obtained in closed and open conformations are reversibly interconvertible by change of pH. A hydrogen-bonded, pKa (where Ka is the acid dissociation constant)–perturbed pair of conserved aspartate residues explains the pH dependence of this transition, and biochemical studies show that pH regulates calcium influx in proteoliposomes. Homology models for hBI-1 provide insights into TMBIM-mediated calcium leak and cytoprotective activity.

Calcium (Ca2+) is a ubiquitous intracellular messenger that regulates cellular and physiological activities. Cytosolic Ca2+ is kept at a low level to assure responsiveness to Ca2+ signals, but subcellular organelles such as the endoplasmic reticulum (ER) and Golgi apparatus maintain calcium stores. Upon activation, Ca2+ is mobilized to cross membrane barriers through calcium-release channels and calcium-uptake pumps (1). Under resting conditions, intracellular and subcellular calcium homeostasis is dynamically regulated to equilibrate between active calcium uptake and passive calcium leak. Calcium homeostasis is cytoprotective (2, 3). An overloaded ER calcium content promotes cell death (4); inversely, lowering of ER calcium content by anti-apoptotic proteins Bcl-2, Bcl-xL, or Bax inhibitor-1 (BI-1) elicits a survival signal (57). Bcl-2, Bcl-xL, and BI-1 have been suggested to regulate ER calcium leak, either directly by forming a leaky pore or by modulating calcium-release channels such as inositol 1,4,5-trisphosphate (IP3) receptors (IP3Rs) (810).

Human BI-1 (hBI-1) was discovered as a human gene product that can block lethality of the pro-apoptotic Bax protein in yeast (8). hBI-1 is localized to the ER membrane, where, among other functions, it mediates a calcium leak downstream of Bcl-2 and Bcl-xL (8, 11). By sequence similarity to hBI-1, a highly conserved TMBIM (transmembrane Bax inhibitor motif) family was identified (12) and assigned the Pfam (13) name of Bax1-I (identification code PF01027). TMBIM proteins are present in prokaryotes, fungi, plants, and metazoans, including invertebrates and mammals (12) (fig. S1). Humans have six identified TMBIM proteins (TMBIM1 to 6), each containing seven presumed transmembrane helices (14) and with variations mainly in their N-terminal extensions (fig. S2). Besides hBI-1 (TMBIM6) in the ER membrane, human Golgi antiapoptotic protein (hGAAP/TMBIM4) is in the Golgi membrane, where it mediates Golgi calcium leak, providing another identified connection to calcium and apoptosis (15). Other human TMBIM proteins are diversely localized and less well characterized (12).

Accumulating evidence has demonstrated the calcium-leak activity of the TMBIM proteins and their regulatory roles in apoptosis (11, 15); however, little is known about the structure or mechanism of action for these proteins beyond recent topological studies on hBI-1 and hGAAP (16, 17). Seeking structural clues into the mechanism of calcium flux activity, we undertook structural studies of TMBIM proteins. Here, we present crystal structures of a bacterial homolog in interconvertible conformational states dependent on pH, we demonstrate pH-sensitive calcium permeation by this protein consistent with the calcium-leak activity of hBI-1 and hGAAP, and we build homology models of hBI-1 to provide structural insights into the calcium leak and antiapoptotic functions of the TMBIM family.

Structural Analyses

To address the structural challenge of the TMBIM family, we identified prokaryotic homologs of hBI-1 that might provide structural insights into function. After screening 51 bacterial relatives for expression in Escherichia coli, we identified YetJ from Bacillus subtilis (BsYetJ), a previously uncharacterized protein, as a family member with satisfactory biochemical properties. The detergent-extracted protein was purified and crystallized in two crystal forms.

Form-1 crystals grew at pH 8 in space group P6522 with one protein molecule per asymmetric unit. We solved this structure by native single-wavelength anomalous diffraction (SAD) (18) using relatively low energy x-rays (~6 keV) to enhance anomalous signal-to-noise ratios. The eight ordered sulfur atoms contributed a Bijvoet-diffraction ratio of ~1.4%. Diffraction data up to 2.8 Å spacings were measured from 12 crystals, and 10 of these met criteria for statistical equivalence (fig. S3A). Previously established analytical procedures (18) allowed both substructure determination and native-SAD phasing. The resulting electron-density map (fig. S3B) permitted automatic tracing of a nearly complete model, which was further refined at 1.95 Å resolution against a separate high-energy data set (table S1 and fig. S3C).

Form-2 crystals grew at pH 6 in space group C2221 and have one molecule per asymmetric unit. Native crystals diffracted x-rays only to ~4.5 Å with severe anisotropy. Attempts at structure solution by molecular replacement from the form-1 structure did not succeed, suggesting a different conformation. Fortunately, a platinum derivative diffracted better; the structure was determined by Pt-SAD phasing at 3.6 Å resolution (fig. S3D), and a conformationally distinct model was built with reference to the form-1 structure, mainly by displacing one helix. We also found that BsYetJ can undergo an intracrystalline transition when form-1 crystals, as grown, are soaked in medium at pH 6. The resulting low pH conformation is almost identical to that in the orthorhombic form-2 crystals (fig. S3E). The converted structure in the hexagonal form-1 lattice diffracted better and could be refined to 2.5 Å resolution (table S1), and this structure was used for further structural characterization.

Structural Features

The structure of BsYetJ in each of its conformations comprises seven transmembrane helices (7 TMs), which by structure-based alignment compares with anticipated helix boundaries for the entire family of TMBIM proteins (fig. S2). The higher-pH form has a compact, closed conformation (Fig. 1A and fig. S4), whereas the lower-pH form has an opened conformation with helix TM2 displaced (Fig. 1B). On the basis of topology assays of the human homologs (16, 17), TMBIM proteins have their N-termini in the cytoplasm, which is consistent with the positive-inside rule (19) as applied to the electrostatic potential surfaces of the BsYetJ structures (fig. S5). The tightly packed helices of the closed form give this structure a barrel-like shape. The barrel is about 56 Å long by 34 Å in diameter and, judged by electrostatics, its axis is tilted in the lipid bilayer by 9° so that 31 Å is embedded in lipid bilayer. The open conformation produces a pore through the lipid bilayer (fig. S5, E to H) that is wide open (11 Å) at the periplasmic side and narrows to a 5 Å–wide bottleneck near the cytoplasmic side.

Fig. 1 Structures of BsYetJ.

(A) Ribbon drawing of form-1 structure determined at pH 8. (B) Ribbon drawing of form-2 structure determined at pH 6. The views of (A) and (B) are from the membrane, and each structure has its seven TMs color-coded. All connecting loops are colored in gray. (C) A cylinder diagram of the form-2 structure to show the structural features from a periplasmic view. The coloring is as for (B).

The polypeptide folding in BsYetJ is topologically different from that of any known structures, as revealed by a DALI search (20). As best seen in the open conformation, conceptually the overall structure has three components: TM1 to 3, TM4 to 6, and TM7 (Fig. 1C). TM1 to 3 and TM4 to 6 are similar, each forming a triple-helix sandwich substructure: TM3 is clamped by helix-loop-helix TM1 and 2; TM6 is clamped by helix-loop-helix TM4 and 5 (Figs. 1C and 2, A and B). In forming the triple-helix sandwich, short side-chain hydrophobic residues on the TMs are crucial (21) because they provide sticky patches to allow close TM1-TM3 and TM4-TM6 contacts for clamping (Fig. 2, A and B, insets). TM3 and TM6 are each bent at the sticky patch. Although there is no obvious sequence similarity between TM1 to 3 and TM4 to 6, the two components are superimposable when inverted by a pseudo-twofold symmetry viewed from within the membrane (Fig. 2, C and D). C-terminal helix TM7 is in the center of the structure, parallel to TM3 and antiparallel to TM6; these three helices together form a central layer that is sandwiched by the four peripheral helices TM1, 2, 4, and 5 (Fig. 1C).

Fig. 2 Structural features.

(A and B) Two triple-helix sandwich substructures consist of TM1 to 3 (A) and TM4 to 6 (B). The color scheme is as Fig. 1B. Insets in (A) and (B) are, respectively, the close α-helical contacts between TM1 and TM3 and between TM4 and TM6. Cα-H⋅⋅⋅⋅⋅⋅O contacts between 2.3 and 3.5 Å were drawn as magenta dashes. A, Ala; D, Asp; F, Phe; G, Gly; H, His; I, Ile; L, Leu; N, Asn; S, Ser; T, Thr; V, Val; and Y, Tyr. (C) Overall pseudo-inverse symmetry with the triple-helix sandwiches in magenta for TM1 to 3 and green for TM4 to 6. The pseudo-twofold axis is on the middle of the red TM7. (D) Superposition of the two symmetric components.

As a 7-TM protein, BsYetJ compares with heterotrimeric guanine nucleotide–binding protein–coupled receptors (GPCRs) (22) and recent CAAX metalloprotease structures (23, 24). The fold of BsYetJ is distinct from these (fig. S6). The CAAX protease helices encompass an intramembrane chamber, and cross-sectional registrations of these helices are unrelated. The GPCR fold has TM3 somewhat central, which has been proposed to have functional importance in ligand binding and signal transduction (25). For BsYetJ, TM7 has the central position, and it too has functional importance, as suggested by sequence conservation and biochemical analyses on its human relative, hBI-1 (17). By topology, BsYetJ somewhat resembles insect olfactory receptors, a special family of 7-TM proteins that also have their N-termini inside (26); however, sequence alignments did not reveal any homology between the two families.

Regulation of Pore Opening and Closing by pH

The facts that the pore-closed and pore-open conformations were obtained at pH 8 and pH 6 and that the pore can be opened by intracrystalline transition (Fig. 3, A and B) suggest a role for pH in regulation of conformational transition. Superposition of the pore-open and pore-closed structures indicates substantial structural changes for TM2 and the two loops connecting it to TM1 and TM3 (Fig. 3, A and B). Relative to the closed conformation, TM2 in the open conformation swings away by as far as 13.5 Å to form a transmembrane pore bordered by it, TM5 to 7, and the presumed lipid bilayer (Fig. 3A and fig. S5F). In the closed conformation, TM2 is rather short (14 residues); whereas in the open conformation TM2 is extended by roughly one helical turn at its N terminus. The transformation from the closed to open conformation (rotation χ = 38.6°; translation tχ = 3.4 Å) displaces TM2 away from contact with TM6 and also moves it along its axis toward the cytoplasm (Fig. 3B).

Fig. 3 Pore opening and closing regulated by pH.

(A and B) Superposition of the two conformationally different structures at pH 8 (cyan and blue) and at pH 6 (gray and magenta). (A) is cytoplasmic view, and (B) is membrane view. (C) Electron density of the pH-7 structure showing alternative conformations of TM2. The 2FoFc electron densities were drawn as gray isomeshes at 0.8σ. The side chains of TM2 in closed and open conformations were drawn respectively as blue and magenta sticks. (D) Overall pH-7 structure with two alternative conformations. Deviated side chains are shown as sticks: cyan for closed conformation and gray for open conformation.

During the pH 6 to pH 8 transition, the c axis of the P6522 lattice shrinks from 289 to 276 Å, but the resulting open-conformation structure is essentially the same as that obtained in the C2221 lattice as grown at pH 6 (fig. S3E). To test for reversibility of this pH-driven conformational transition, we soaked the closed-form crystals at pH 6; took diffraction images to confirm unit-cell shrinkage, the hallmark of pore-opening upon soaking; and then back-soaked the exact crystals to pH 8. This reestablished pH-8 structure, which has a diagnostic c axis of 293 Å, is reclosed to be almost identical to the initial pore-closed conformation. Thus, pH can regulate the opening and closing of the pore.

To further test the pH-driven conformational changes, we moved the opened crystals from pH 6 to pH 7 and determined this pH-7 structure by molecular replacement, using the back-soaked pH-8 structure because their lattice parameters were nearly identical (table S1). Electron density was seen for TM2 in both closed and open conformations, consistent with alternate states in equilibrium (Fig. 3C). The ratio of closed:open component was determined as 60:40 by occupancy refinement with the program PHENIX (27). The pH-7 structure as refined with two alternative conformations shows differences propagated into neighboring helices with associated main-chain and side-chain shifts (Fig. 3D). Thus, the conformation of BsYetJ is pH sensitive, and we presume that the ratio of TM2 occupancies reflects the conformational equilibrium. On the basis of the 60:40 closed:open ratio at pH 7, the apparent pKa (where Ka is the acid dissociation constant) of the protein in the crystal may be estimated to be slightly below pH 7.

Di-Aspartyl pH Sensor

When closed at pH 8, TM2 and cytoplasmic loop L2,3 (TM2-TM3 connection) engage in several hydrogen bonds with residues from TM3, 6, and 7; and all of these are disrupted in the open conformation at pH 6 (Fig. 4A). A key interaction at the closed-conformation interface appears to be a latch of Arg60 with the TMBIM-conserved di-aspartyl unit Asp171-Asp195 (Fig. 4B). The guanidinium moiety from Arg60 forms a doubly hydrogen-bonded salt bridge with Asp171 when the protein is at pH 8 (Fig. 4B), but this is broken at pH 6 (Fig. 4C) and reestablished upon back-soaking to pH 8 (Fig. 4D). In all cases, the carboxylate groups of the two aspartates are hydrogen-bonded to one another. Electron density that is continuous through the di-aspartate unit, even for the 1.95 Å–resolution structure, and refined O-O distances of 2.55 Å at pH 8 and 2.71 Å at pH 6 demonstrate the hydrogen bonding and anomalous aspartate protonation.

Fig. 4 Di-aspartyl pH sensor.

(A) H-bond interactions within the pore of the closed-conformation structure. E, Glu; M, Met; and R, Arg. (B to D) Successive structures from intracrystalline transitions with superposed 2FoFc electron densities contoured at two levels, 1.2σ (gray) and 3.0σ (magenta). (B) Starting form-1 structure at pH 8. (C) Structure after soaking a form-1 crystal into a medium at pH 6, disrupting the interactions between Arg60 and Asp171. (D) Structure after reversal, from first soaking at pH 6 and then back-soaking to pH 8, thereby reclosing the pore and restoring the interactions between Arg60 and Asp171.

The structural presumption based on the pH-8 structure would have Asp195 as the protonated group, leaving Asp171 free to form its salt bridge with Arg60. This presumption is validated by pKa calculations from the program PROPKA (28). For the closed conformation, the pKa values of Asp171 and Asp195 are 3.1 and 11.2, respectively, one depressed and the other elevated resulting from a presumed coupling effect (28). For the open conformation, the respective values are 6.2 and 12.0. The increased pKa of Asp171 in the open conformation may account for the equilibrated conformational states seen in our pH 7 structure. Thus, in absence of the guanidinium interaction, Asp171 is more readily protonated; alternatively, Asp171 when protonated is incompatible with Arg60 engagement. We suggest that the pH control of Asp171 thereby regulates the conformational transition and pore opening or closing. Asp195 is protonated throughout pH 6 to 8. Aspartate residues at positions 171 and 195 are strictly conserved in the TMBIM family (fig. S2), suggesting that a di-aspartyl pH sensor may be a family trait. The latch partner Arg60 is also conserved in TMBIM1 to 4 but not in TMBIM5 and 6, opening a question of alternative latch partners.

Both Arg60-bearing loop L2,3 and loop L1,2, which connects TM2 to TM1, are quite flexible, as seen in B-factor plots (fig. S7). This mobility is consistent with ready displacement of TM2 when the latch to Asp171 is broken at lower pH. When TM2 is displaced from its van der Waals contacts with TM5, both it and TM5 have increased flexibility. Thus, we picture a pH-sensitive conformational equilibrium between a rather flexible pore-open state seen at lower pH and a pore-closed state seen at higher pH.

Calcium Leak

Both hBI-1 and hGAAP are able to mobilize calcium leak into the cytoplasm from stores in ER and Golgi compartments, respectively (7, 11, 15). To test whether our bacterial homolog has calcium-leak activity, we overexpressed BsYetJ in E. coli, added calcium extracellularly, and measured the intracellular calcium concentration with the fluorescent calcium dye Fura-2/AM. Upon addition of external calcium, intracellular calcium concentration increased steadily in cells overexpressing BsYetJ but not for controls of an empty plasmid or an unrelated membrane protein transporter (Fig. 5A). Thus, BsYetJ is a bona fide functional bacterial TMBIM homolog with calcium-leak activity.

Fig. 5 Structural and functional characterization of calcium leak.

(A) Calcium influx into bacteria overexpressing BsYetJ. An empty plasmid and a manganese transporter were used as negative controls. Error bars indicate ±SEM (n = 9 experiments). (B) Calcium influx into proteoliposomes. Error bars, ±SEM (n = 3). (C) Electrostatic surface of the closed-conformation structure showing charged surface concavities and internal cavities but a blocked pore. (D) Electrostatic surface of the open-conformation structure at pH 7.4, where the cleft is electronegative. (E) Electrostatic surface for the open-conformation structure at pH 6, where the cleft is more neutral. The contour level of the electrostatic surface is at ±5 kT/e. Red, negative potential; blue, positive potential. (F to H) A proposed model for pH-sensitive calcium leak. (F) At higher pH (e.g., 8), Asp195 is protonated and Asp171 is deprotonated. Asp171 forms two H bonds with positively charged Arg60, and the Arg60-Asp171 latch closes the pore. (G) When in the open conformation at a more neutral pH (e.g., 7.4), Asp171 may equilibrate between protonated and deprotonated states. Calcium passage occurs only when Asp171 is transiently deprotonated. (H) At lower pH (e.g., 6), the equilibration will favor more complete protonation of Asp171, disfavoring calcium passage due to pore neutralization. Cartoons (F) to (H) correspond to structures (C) to (E) directly above.

Having shown that BsYetJ can produce a calcium leak in bacteria and that pH can control pore opening in BsYetJ, we explored the effect of pH on calcium influx. We constructed BsYetJ proteoliposomes at various pHs, preloaded them with the Fura-2 dye, and measured calcium influx when exposed to externally added calcium. We observed pH-dependent influxes of calcium, typically accumulating more rapidly at the outset of calcium application and then slowing somewhat to steady-state levels after ~10 min (fig. S8). The steady-state calcium influx was substantially lower when the pH was lower (pH 6.5) or higher (pH 7.9) than when under near-neutral conditions (pH 7.0 or 7.4) (Fig. 5B). We conclude that pH-sensitive calcium-leak activity is intrinsic to BsYetJ because we prepared the proteoliposomes from pure components.

Multiple hydrophilic residues lie within the core of closed-conformation BsYetJ (Glu49, Arg205, Asp202, Asn198, Asp195, Asp171, and Arg60), lining up from the periplasmic side to the cytoplasmic side (Fig. 5C). These residues border concave surfaces that invaginate from either side and form two charged internal cavities along with these residues. This incipient passageway for calcium ions is structurally blocked in the closed conformation by hydrophobic residue Phe164 (Fig. 5C). When TM2 is moved away from the closed conformation, the separated concave surfaces and internal cavities unify to form a transmembrane pore, largely electronegative at a physiological pH (7.4; Fig. 5D) and less electronegative at a lower pH (6.0; Fig. 5E).

The pore at a more neutral pH appears conducive to calcium passage; however, we did not detect calcium ions in the pore of any BsYetJ crystal structure even though calcium was included in all crystallization and soaking experiments. Evidently, the structure does not feature discrete calcium binding sites in the pore but does allow calcium passage with only transient interactions within the pore. On the basis of the alternative conformations in our pH-7 structure, we imagine BsYetJ in the lipid bilayer to be in a facile equilibrium between open and closed conformations.

Why is the flux rate higher at pH 7, where the pore is in equilibrium, than at pH 6, where it is open? Fluctuations in the open/closed equilibrium may permit a calcium leak by opening a pore that is electronegative when at physiologically neutral pHs (Fig. 5D), as in our assays for uptake of calcium into bacteria and proteoliposomes (Fig. 5, A and B), whereas either closure at higher pH or reduced electronegativity at lower pH will counteract calcium passage. A working model for this pH-sensitive calcium-leak activity is given in Fig. 5, F to H. We propose that BsYetJ exists equilibrated among three states: closed, open deprotonated, and open protonated. At higher pH, Asp171 is predominantly deprotonated and forms a doubly hydrogen-bonded salt bridge with Arg60 (Fig. 5F); with this Arg60-Asp171 latch in place, the structure is closed and the pore is sealed (Fig. 5, C and F). At a more neutral pH, the open conformation becomes accessible, even if transiently (Fig. 5G); the opened pore is electronegative with Asp171 remaining deprotonated, and a calcium leak is then facilitated along the calcium gradient into the cytosol (Fig. 5, D and G). At lower pH, the protonated state of Asp171 is favored because its pKa is raised in the open conformation, which precludes formation of the Arg60-Asp171 latch (Fig. 5H). In this state, electronegativity of the pore is reduced and calcium influx is impeded (Fig. 5E).

We conclude that BsYetJ is a pH-sensitive calcium-leak channel. Its molecular architecture is distinct from that known for other calcium channels and exchangers (2933), and its pH-dependent changes in conformation and electrostatics are compatible with observed calcium-flux activities. hBI-1 also exhibits pH-sensitive calcium-leak activities, proposed to be mediated by aspartic acid residues on TM7 (34) or by the C-terminal lysine-rich motif (35). Our results are consistent with the TMBIM-conserved di-aspartyl pH sensor, but not the C-terminal lysine-rich motif, as a shared mechanism for the pH-sensitive calcium leak.

Insights into the BI-1–Mediated Calcium Leak and Antiapoptosis

To explore the structural linkage of BsYetJ to human homologs, we constructed homology models of hBI-1, the most studied of TMBIM proteins, by using structure-based sequence alignment (fig. S2) and the program MODELLER (36).

Because BsYetJ is homologous with its human TMBIM relatives (21% for TMBIM4 and 18% for TMBIM6), the hBI-1 models are highly similar to their BsYetJ templates but with instructive differences. The di-aspartyl pH sensor Asp188-Asp213 (Asp171-Asp195 in BsYetJ) is intact in both states; however, the Arg60 latch of BsYetJ (and TMBIM1 to 4) is replaced by His78 in hBI-1. In the closed conformation of hBI-1, His78 forms an alternative latch by hydrogen-bonding to Asp209 (Ser191 in BsYetJ) (Fig. 6A). In the open conformation, Asp209 is freed from its interaction with His78, and it will likely exacerbate the pKa elevations of Asp188-Asp213 expected by analogy with BsYetJ. At a suitably low pH, we expect that Asp209 will adopt an alternative conformation and form a hydrogen bond with protonated Asp188 (Fig. 6B). Analogous to residues in BsYetJ, side chains of Leu71, Thr74, and Met181 (Phe164 in BsYetJ) are expected to provide pore-sealing functions in the closed state, separating concave indentations from the opposing membrane surfaces. Although many residues that line the open-state pore differ in hBI-1 as compared with BsYetJ, the shape and electrostatic features of the two pores remain very similar; particularly, the narrow opening on the cytoplasmic side of the hBI-1 pore is like that of BsYetJ.

Fig. 6 Homology models of hBI-1.

(A and B) Homology models of hBI-1 in its closed (A) and open (B) conformations. The conserved di-aspartyl Asp188-Asp213 unit is shown within the membrane in each case, and a third aspartate, Asp209, is shown interacting with His78 in (A) and with Asp188 in (B). In (A), the indicated pore-sealing residues, Leu71, Thr74, and Met181, separate concave surfaces invaginating from opposite membrane face.

On the basis of topology studies, hBI-1 was proposed to have a 6-TM topology with both N- and C-termini in the cytoplasm, where the predicted TM7 segment is either disordered or hemi-penetrant back into the membrane (16, 17). On the other hand, TM7 is in the middle of the 7-TM structure of BsYetJ, where it forms extensive interactions with other TM helices. These include the conserved di-aspartyl pH sensor between TM6 and TM7 and a contact between TM4 and TM7 (Thr104/121 and Phe200/218 in BsYetJ and hBI-1, respectively). The hydrophobicity analysis, secondary structure predictions, and conservation also suggest that hBI-1 has a 7-TM topology like that in BsYetJ (fig. S9).

Besides calcium-leak activity, which has been observed for TMBIM proteins wherever tested, TMBIM interactions with other proteins have also been identified. For example, hBI-1 and hGAAP coimmunoprecipitate with IP3R and modulate IP3-induced Ca2+ release (9, 15), and hBI-1 interacts with Bcl-2 as shown by in vivo cross-linking and coimmunoprecipitation (8, 11). On the basis of the BsYetJ structures and the homology models for hBI-1, a plausible mode of TMBIM-mediated protein-protein interactions would have a TM from the partner protein taking the place of TM2 after its displacement in the open conformation of the TMBIM protein (figs. S10, A and B).

It has been reported that Bax and Bak activation and mitochondria outer membrane permeabilization (MOMP) are enhanced by overloaded calcium stores (2). The protective role of hBI-1 in decreasing ER Ca2+ concentration is expected to reduce MOMP and thus suppress the activation of Bax and Bak toward the initiation of apoptosis. We propose that human TMBIM proteins function in maintaining a dynamic homeostasis of stored Ca2+ concentration and cytosolic Ca2+ concentration through the pH-sensitive calcium-leak mechanism. The overexpression of hBI-1 in various cancers, including prostate, breast, glioma, uterine, ovarian, and lung, presumably reflects recruitment of its antiapoptotic activity (3740). Knockdown of hBI-1 expression by RNA interference has shown effectiveness in inducing spontaneous apoptosis of cancer cells in prostate and breast (37, 38). The structures of BsYetJ and its derivative hBI-1 models provide substantial insights into the functioning of TMBIM proteins and offer therapeutic prospects for intervention of anti-apoptotic functions in treatment of cancers.

Supplementary Materials

www.sciencemag.org/content/344/6188/1131/suppl/DC1

Materials and Methods

Figs. S1 to S10

Table S1

References (4163)

References and Notes

  1. Acknowledgments: We thank J. Schwanof and R. Abramowitz at NSLS beamlines X4A and X4C for their assistance in data collection, J. Love for help in initial high-throughput screening, M. Punta for help in initial target selection, M. Su for help with the phylogenetic presentation, and F. Mancia and L. Shapiro for discussions. This work was supported in part by NIH grants GM095315 and GM107462. Beamlines X4A and X4C of the NSLS at Brookhaven National Laboratory, a U.S. Department of Energy facility, is supported by the New York Structural Biology Center. Atomic coordinates and structure factor files have been deposited in the Protein Data Bank (PDB) under the accession codes 4PGR for closed-form at pH 8, 4PGS for open-form at pH 6 by soaking, 4PGU for closed/open-form at pH 7 by back-soaking, 4PGV for closed-form at pH 8 by back-soaking, and 4PGW for open-form in C2221 lattice. Q.L., Y.C., and W.A.H. are inventors on a patent application filed by the New York Structural Biology Center on uses of the three-dimensional structures of BsYetJ and homology models of hBI-1.
View Abstract

Stay Connected to Science

Navigate This Article