Ion and pH Sensitivity of a TMBIM Ca²⁺ Channel

Summary

The anti-apoptotic transmembrane Bax inhibitor motif (TMBIM) containing protein family regulates Ca²⁺ homeostasis, cell death, and diseases including cancers. The recent crystal structures of the TMBIM homolog BsYetJ reveal a conserved Asp171-Asp195 dyad that is proposed in regulating a pH-dependent Ca²⁺ translocation. Here we show that BsYetJ mediates Ca²⁺ fluxes in permeabilized mammalian cells, and its interaction with Ca²⁺ is sensitive to protons and other cations. We report crystal structures of BsYetJ in additional states, revealing the flexibility of the dyad in a closed state and a pore opening mechanism. Functional studies show that the dyad is responsible for both Ca²⁺ affinity and pH dependence. Computational simulations suggest that protonation of Asp171 weakens its interaction with Arg60, leading to an open state. Our integrated analysis provides insights into the regulation of the BsYetJ Ca²⁺ channel that may inform understanding of human TMBIM proteins for their roles in cell death and diseases.

Graphical Abstract

eTOC:

Guo et al. present structural, functional, and computational analyses on a bacterial TMBIM Ca²⁺ channel BsYetJ. They show that BsYetJ mediates Ca²⁺ flux in mammalian cells, and its Ca²⁺ binding is sensitive to protons and other cations. The conserved aspartyl dyad regulates Ca²⁺ binding, pH sensing, and the channel pore opening and closing.

Introduction

Transmembrane Bax Inhibitor Motif (TMBIM) containing proteins are a family of membrane proteins that control Ca²⁺ flux across membranes to maintain Ca²⁺ homeostasis in subcellular Ca²⁺ stores (Ahn et al., 2009; Rojas-Rivera and Hetz, 2015; Gamboa-Tuz et al., 2018). TMBIM proteins are highly conserved and ubiquitously expressed in both pro- and eukaryotic cells, as well as in viruses and archaea (Figure S1) (Henke et al., 2011; Chang et al., 2014). TMBIM proteins have been shown to be associated with the suppression of cell death in response to a broad spectrum of stresses in plants and intrinsic and extrinsic programmed cell death in mammals (Reimers et al., 2008; Hu et al., 2009; Henke et al., 2011; Ishikawa et al., 2011; Lisak et al., 2015). An increasing number of studies suggest that the suppression of cell death by TMBIM proteins is directly or indirectly linked to their roles in maintaining Ca²⁺ homeostasis (Carrara et al., 2015; Rojas-Rivera and Hetz, 2015; Lisak et al., 2016; Liu, 2017; Luganini et al., 2018).

TMBIM proteins were originally identified from phenotypic screening of anti-apoptotic human genes in yeast overexpressing Bax, a proapoptotic protein killing yeast cells (Xu and Reed, 1998). Although there are no direct interactions between Bax and TMBIM proteins, anti-apoptotic Bcl-2 family members, such as Bcl-2 and Bcl-xL, have been alluded to interact with TMBIM proteins by co-immunoprecipitation (Xu and Reed, 1998; Urresti et al., 2016). The interactions and functional overlapping of TMBIM and Bcl-2 family proteins may provide multi-layered and redundant protections of cellular survival in response to stress signals including Bax overexpression. In humans, there are six members (TMBIM1-6) localized in diverse cellular and subcellular membranes. These proteins are dysregulated in multiple cancers (Grzmil et al., 2006; Lu et al., 2015; Rojas-Rivera and Hetz, 2015; Carrara et al., 2017). TMBIM4 and 6 are upregulated in brain, breast, and prostate cancers; they are downregulated in leukemia and colorectal cancer (Carrara et al., 2017). As a contrast, TMBIM3 is upregulated in colorectal and gastric cancer, while TMBIM1 is downregulated in leukemia, breast, and sarcoma cancers (Rojas-Rivera and Hetz, 2015). In addition to their abnormal expression in cancers, TMBIM proteins have been described with additional cellular functions. TMBIM1 is upregulated in response to mechanical stress (Yoshisue et al., 2002; Shukla et al., 2011). TMBIM2 over-expression protects against transient brain ischemia (Reich et al., 2011) and increases axon growth (Merianda et al., 2013). TMBIM4 and TMBIM6 regulate cell adhesion and migration through Ca²⁺-dependent mechanisms (Lee et al., 2010; Saraiva et al., 2013; Carrara et al., 2015). Located in the mitochondrial inner membrane, TMBIM5 regulates mitochondrial fragmentation and cytochrome c release (Oka et al., 2008). TMBIM6 protects against acute brain injury (Krajewska et al., 2011), promotes autophagy (Sano et al., 2012) and enables immune cell survival and function (Lisak et al., 2016). TMBIM3 and TMBIM6 may interplay with each other and synergistically mediate unfolded protein response and autophagy (Rojas-Rivera et al., 2012).

Most TMBIM-mediated cellular functions relate to their core roles in Ca²⁺ flux and Ca²⁺ homeostasis (Lisak et al., 2015; Carrara et al., 2017). TMBIM6 is able to mediate the passive leak of Ca²⁺ from the ER, hence its classification as a Ca²⁺ leak channel (Kiviluoto et al., 2013; Bultynck et al., 2014). This property of TMBIM6 lowers steady-state ER Ca²⁺ levels in TMBIM6-overexpressing cells (Chae et al., 2004; Xu et al., 2008). Similarly TMBIM4 lowers the content of ER and Golgi Ca²⁺ stores and enhances store-operated Ca²⁺ entry (de Mattia et al., 2009; Saraiva et al., 2013; Carrara et al., 2015).

Thus, a common feature of TMBIM proteins in cells is their ability to prevent the overloading of intracellular Ca²⁺ stores. This feature reduces the likelihood for Ca²⁺ overload and excessive Ca²⁺ fluxes into mitochondria to occur, thereby promoting cell survival and rendering cells more resistant to cell death stimuli (Pinton et al., 2008). In TMBIM4 and 6, several critical aspartate residues were identified to establish an assumed Ca²⁺-permeable channel pore near its C-terminus (Bultynck et al., 2012; Carrara et al., 2015). Originally, it was proposed that TMBIM6’s basal channel function was promoted by low pH, suggesting it would act as a H+/Ca²⁺ exchanger (Kim et al., 2008). However, follow-up work showed that low pH is actually abrogating Ca²⁺ flux through the Ca²⁺-permeable pore (Kiviluoto et al., 2013). However, due to challenges in obtaining mammalian TMBIM structural information, the molecular determinants leading to pore opening and closing and thus the underlying Ca²⁺ flux properties of TMBIM proteins remained unclear (Bultynck et al., 2012; Bultynck et al., 2014; Carrara et al., 2015). This has recently changed with the determination of the structures of a bacterial homolog of TMBIM proteins from Bacillus subtilis, BsYetJ (Chang et al., 2014). These structures reveal a unique fold of 7 transmembrane helices (TM) with its second transmembrane helix (TM2) capable of opening and closing laterally within the membrane bilayer. An aspartyl dyad that is conserved among all TMBIM members (Asp171-Asp195 in BsYetJ) (Figure S1) was proposed to be a pH sensor regulating the pore dynamics and Ca²⁺ translocation (Chang et al., 2014).

Here we present crystal structures of BsYetJ wild-type and variants with mutations of Asp171 or Asp195. Together with functional and computational studies we show the ion- and pH-dependent modulation of BsYetJ activity and provides insights into the core mechanism of the conserved Asp171- Asp195 dyad being the structural determinant that leads to the opening of this Ca²⁺ leak channel.

Results

BsYetJ can mediate Ca²⁺ efflux from eukaryotic intracellular Ca²⁺ stores

We have previously solved structures of a bacterial homolog BsYetJ and demonstrated its Ca²⁺ flux activity in E. coli overexpressing BsYetJ and proteoliposomes with reconstituted BsYetJ (Chang et al., 2014). Sequence alignment studies revealed that the proposed pH-sensing aspartyl-dyad (Asp171- Asp195 in BsYetJ) is a conserved feature found in TMBIM proteins across all kingdoms (Chang et al., 2014). To study the functional conservation of BsYetJ in mediating Ca²⁺ flux in eukaryotic cells, we assessed whether purified BsYetJ protein could promote ⁴⁵Ca²⁺ efflux from the ER of human cells, similarly to what was previously observed for synthetic peptides derived from the TMBIM6 C terminus (Bultynck et al., 2012). We performed unidirectional ⁴⁵Ca²⁺ flux assays in permeabilized HeLa cells. After saponin-mediated permeabilization of the HeLa cells and loading the ER Ca²⁺ stores to steady state with ⁴⁵Ca²⁺, we washed the cells with thapsigargin, an inhibitor of the sarco/endoplasmic reticulum Ca²⁺-ATPase (SERCA), and monitored the unidirectional ⁴⁵Ca²⁺ efflux from the ER Ca²⁺ stores. The resulting data are plotted as Ca²⁺ release and ⁴⁵Ca²⁺ content as function of time (Figures 1A, B). Notably, addition of 2.5 μM BsYetJ provoked a major increase in the ⁴⁵Ca²⁺ release (Figure 1A), a phenomenon that was not observed in the control condition, i.e. the addition of buffer without BsYetJ. This could also be observed as a drop in the ⁴⁵Ca²⁺ content upon addition of the BsYetJ protein (Figure 1B). There was a delay in the BsYetJ-associated response, likely related to the time needed for BsYetJ protein to incorporate into the ER membranes. Therefore, bacterial BsYetJ, when added to permeabilized human HeLa cells, may reconstitute to the membranes of intracellular Ca²⁺ stores, thereby functioning as Ca²⁺-permeable channel-like proteins that mediate Ca²⁺ flux across membranes.

Figure 1. Ca²⁺ activity of BsYetJ.

(A-B) Ca²⁺ efflux mediated by BsYetJ. Permeabilized HeLa cells loaded to steady-state with ⁴⁵Ca²⁺ were incubated in Ca²⁺-free efflux medium. ⁴⁵Ca²⁺ release (% / 2 min) was plotted as a function of time (A) and normalized ⁴⁵Ca²⁺ content (%) was plotted as a function of time (B). In the time period indicated by the dark square, buffer or BsYetJ (2.5 mM) was added for 2 min. Data from a typical experiment are represented as mean ± SD (n= 2; i.e. duplicate technical repeats). The graph is representative for 3 independent experiments. (C) pH dependence of ⁴⁵Ca²⁺ binding. Binding of 10 μM ⁴⁵Ca²⁺ (746 Ci/mmol) to 50 ng of BsYetJ was measured in assay buffer with a pH ranging from 6.0 to 8.0. Data were fit to a Gaussian model, revealing a peak at pH 7.46 ± 0.01. (D) Apparent affinity of Ca²⁺ binding by BsYetJ at pH 7.5. 10 μM ⁴⁵Ca²⁺ binding was measured in the presence of increasing concentrations of non-labeled Ca²⁺, yielding an EC₅₀ of 65.4 ± 0.9 μM. (E) Ion sensitivity of BsYetJ. 10 μM ⁴⁵Ca²⁺ binding by BsYetJ was measured in the absence (−) or presence of 5 mM or 100 mM (*) of the indicated cations. Data are represented as mean ± SEM (n= 3).

Inhibition of Ca²⁺ binding by ions.

To further characterize the substrate specificity profile of BsYetJ, we used scintillation proximity assay (SPA)-based binding of ⁴⁵Ca²⁺ (Quick and Javitch, 2007; Zehnpfennig et al., 2015) in the presence or absence of several physiologically relevant cations. Whereas binding measurements performed in assay buffer containing high concentrations of Na⁺ (100 mM) revealed virtually no Ca²⁺ binding, in absence of Na⁺, however, robust binding of ⁴⁵Ca²⁺ was observed (Figure 1C). Ca²⁺ binding was pH dependent in bell-shaped fashion, with higher activity at pH 7.5 and lower activity at pH 6 and 8. This result is consistent with our Ca²⁺ flux assays in proteoliposomes reported previously (Chang et al., 2014). Fitting the pH dependence of ⁴⁵Ca²⁺ binding to a Gaussian model revealed a peak at pH 7.5 (Figure 1C). It is noteworthy to mention that even though the activity at 7.5 was higher than that observed at pH 6.0, substantial activity was still detectable at the latter pH, suggesting that BsYetJ activity is regulated by pH. Measuring the isotopic dilution of ⁴⁵Ca²⁺ bound to BsYetJ with increasing concentrations of non-labeled Ca²⁺ at pH 7.5 revealed a reduction of the bound ⁴⁵Ca²⁺ by 50 % (EC₅₀) at a concentration of 65.4 μM of CaCl₂ (Figure 1D). Since Ca²⁺ has an ionic radius of 0.99 Å which is very close to Na⁺ with an ionic radius of 0.98 A, we speculated that the concentration of Na⁺ that reduces ⁴⁵Ca²⁺ binding by 50% (IC₅₀) would be comparable to the apparent affinity of Ca²⁺. However, Na⁺, K⁺ and Li⁺ inhibited ⁴⁵Ca²⁺ binding with IC₅₀ of 18.5±5.7 mM, 14.0 ± 3.9 mM, and 16.1 ± 6.5 mM, respectively. The inhibitory effect of the monovalent ions Na⁺ and K⁺ at physiological concentrations suggests a regulatory mechanism by these ions on the Ca²⁺ leak activity of BsYetJ.

To further expand the ion regulation profile of BsYetJ, we measured ⁴⁵Ca²⁺ binding in the presence of 5 mM each of Mg²⁺, Mn²⁺, Gd³⁺, Yb³⁺, and Lu³⁺ by using the SPA. We found that Mg²⁺ and Mn²⁺ had no apparent inhibition of Ca²⁺ binding (Figure 1E), while Yb³⁺ and Lu³⁺, and most notably Gd³⁺, inhibited Ca²⁺ binding. Among these lanthanides, Gd³⁺ has a much stronger inhibitory potency than Yb³⁺ and Lu³⁺. Measuring the concentration dependence of Gd³⁺ on ⁴⁵Ca²⁺ binding revealed an IC₅₀ of 24.2 ± 2.8 μM. It is possible that in addition to competing with Ca²⁺ for binding to the proposed Ca²⁺ binding site, Gd³⁺ might bind specifically to a remote site which may induce conformational changes unfavorable of Ca²⁺ binding.

Alternative pH-sensor structure determined in lipidic cubic phase

To understand the delicate structure of the pH-sensing unit in a native-like lipidic environment, we crystallized wild-type BsYetJ in lipidic cubic phase (LCP) and determined its structure at 2.5 Å resolution by molecular replacement. Figure 2A shows the superposition of the BsYetJ structure crystalized in LCP (^LCPBsYetJ) on its structure crystallized in the detergent pentaethylene glycol monodecyl ether (C10E5) (^C10E5BsYetJ). Both structures are in a closed conformation with an RMSD of 0.64 Å. The relatively large rearrangements are the two loops connecting to TM2, consistent with the required flexibility of TM2 capable of lateral movement in the conformational transition between open and closed states (Chang et al., 2014). In addition, some side chains adopt different rotamer conformations (Figure 2A), likely due to either interactions with lipids or detergents, or lattice packing. The structure in closed conformation is featured by the proximity of residues Arg60, Asp171, and Asp195 (Figure 2B). To see structural difference of the proposed pH-sensing dyad (Asp171-Asp195) in ^LCPBsYetJ and ^C10E5BsYetJ, we superimposed the two structures based on their Cα atoms on TM7 (Figure 2C). In the ^LCPBsYetJ structure, the distance between the Asp171 and Asp195 carboxylates is 2.9 Å, indicating the coupling of the two carboxylates and the protonation of Asp195 as a proton donor in hydrogen bonding. This is consistent with the corresponding H-bond in the ^C10E5BsYetJ structure. However, surprisingly, the side chains of Asp171 and Asp195 in ^LCPBsYetJ deviate from their respective positions in ^C10E5BsYetJ. Consequently, the two H-bond distances between Arg60 and Asp171 in ^LCPBsYetJ are longer than these in ^C10E5BsYetJ. Considering that the ^LCPBsYetJ structure is in a native-like environment, these relative movements show structural flexibility of these three residues even in the closed state.

Figure 2. Crystal structure of BsYetJ in LCP.

(A) ^LCPBsYetJ structure superimposed with the structure crystallized from detergent C10E5 (^C10E5BsYetJ). (B) Electron densities for the aspartyl dyad pH sensor and the conformation latch Arg60. (C) Superimposition of ^LCPBsYetJ with ^C10E5BsYetJ for the aspartyl dyad and Arg60. Dash lines show distances between indicated residues, magenta for ^LCPBsYetJ and dark for ^C10E5BsYetJ.

To induce potentially more drastic changes, we attempted soaking ^LCPBsYetJ crystals at pH 6 and 7. However, soaked ^LCPBsYetJ crystals did not diffract X-rays or even disappeared. This might suggest that the crystallographic lattice (P2₁) is not compatible with associated conformational changes. Nevertheless, the alternative conformations of Arg60, Asp171, and Asp195 seeing in ^LCPBsYetJ and ^C10E5BsYetJ structures show a possible pH-dependent structural transition that leads to pore opening. We can imagine that decreasing pH may increase the protonation probability of Asp171, rendering a structure to be primed for Ca²⁺ flux.

Structure and activities of two pH sensor mutants

To further characterize the channel structurally and biochemically, we made D171E and D195E variant constructs by adding a CH₂ group to the side chains of the dyad residues Asp171 and Asp195. Note these mutations retain the overall charge. In spite of crystallization of both mutants in LCP and detergent, we found that crystals that diffracted to adequate resolutions only from detergent preparations. The two mutant structures are in their open conformations even though they were crystallized at a neutral pH 7.5 (^D171EBsYetJ) and 7.0 (^D195EBsYetJ) (Figures 3A, B). Both mutants were crystallized in space group C2. The solved structures indicate four molecules in asymmetric unit. In each mutant structure, the four molecules have a Cα RMSD of 0.30-0.71 Å for ^D171EBsYetJ and 0.230.61 Å for ^D195EBsYetJ. Superimposition of the four molecules indicates most conformational changes involve the movement of TM5 (Figure 3B). In both structures, Arg60 is away from forming a salt bridge with Asp171 (^D195EBsYetJ) or Glu171 (^D171EBsYetJ). The distances between the carboxylates at positions 171 and 195 range between 2.9 and 4.6 Å, suggesting a rather unstable coupling between the two carboxylate groups, and thus incapable of locking TM2 in a closed conformation through Arg60 engagement. Upon TM2 opening, TM5 may also undergo conformational changes, more dramatically at the C-terminal side with as much as 9° twist between two TM5 helices in ^D171EBsYetJ (Figure 3B, left panel). Similar TM5 twists were also seen in ^D195EBsYetJ but to a less extent (Figure 3B, right panel).

Figure 3. Structure and activity of two pH-sensitive mutants.

(A) Two views of structures of ^D171EBsYetJ and ^D195EBsYetJ. The two mutant structures were superimposed with ^LCPBsYetJ. Both mutant structures are open with TM2 unlocked from residue 171. (B) Superimpositions of the four molecules in crystallographic asymmetric unit indicate the flexibility of TM5 in ^D171EBsYetJ (left) and ^D195EBsYetJ (right). (C) pH-dependent Ca²⁺ binding by wild-type BsYetJ (■), ^D171EBsYetJ (♦), or ^D195EBsYetJ (Δ). (D) The EC₅₀ of Ca²⁺ binding by wild-type (■), ^D171EBsYetJ (♦) or ^D195EBsYetJ (Δ) was determined by isotopic dilution of 10 μM ⁴⁵Ca²⁺ with non-radiolabeled CaCl₂, yielding apparent affinity constants of 64.2 ± 1.3 μM, 1.2 ± 0.1 mM, and 1.7 ± 0.3 mM, respectively. Data are represented as mean ± SEM (n=3).

To test whether the mutations to the pH-sensing dyad affect the interaction of the proteins with Ca²⁺, we measured ⁴⁵Ca²⁺ binding at different pHs for BsYetJ wild-type as well as the two mutants by SPA (Figure 3C). In contrast to the bell-shaped curve observed for the wild-type, the two mutants displayed reduced Ca²⁺ binding and significantly different pH dependence. Interestingly, the highest activity was measured at the lowest pH tested (pH 6.0), and Ca²⁺ binding was progressively reduced with increasing pH. Performing isotopic dilution of 10 μM ⁴⁵Ca²⁺ binding by BsYetJ wild-type, ^D171EBsYetJ, and ^D195EBsYetJ at pH 6.5 revealed that the EC₅₀ of Ca²⁺ binding by BsYetJ (64.2 ± 1.3 μM) is virtually indistinguishable from that determined at pH 7.5 (Figure 1D). On the other hand, the EC₅₀ for ^D171EBsYetJ and ^D195EBsYetJ determined at pH 6.5 were determined as 1.2 ± 0.1 mM and 1.7 ± 0.3 mM, respectively (Figure 3D). Therefore, both mutants have much weaker Ca²⁺ affinity relative to the wild-type, suggesting that Asp171 and Asp195 play a critical role in BsYetJ function, either by directly interacting with protons and/or Ca²⁺ or by structurally impacting the recognition of these two ions indirectly. Thus, we propose that the change of an aspartate to glutamate on the pH-sensing dyad may have disrupted the sensing mechanism, resulting in constantly open channels with reduced Ca²⁺ affinities.

Protonation state of Asp171 is critical for pore closing and opening

To understand the dynamics and underlying regulation mechanism in response to pH sensing and protonation, we performed extensive and comparative molecular dynamics (MD) simulations. Starting from the closed-state structure, we compared four simulation conditions, with the Asp171 or Glu171 in either deprotonated or protonated state. In the control simulations of the Asp171-deprotonated condition, the Asp171 and protonated Asp195 form a stable dyad; and Asp171 persistently forms a strong salt bridge with Arg60 (Figure 4A). In the Asp171-protonated condition, the protonation of Asp171 not only disrupted the Arg60-Asp171 salt bridge, the conformation of the Asp171-Asp195 dyad is different from the Asp171-deprotonated condition as well (Figure 4B). In the Glu171-deprotonated state, while we observed that the Arg60 formed the salt bridge with Glu171, the additional CH₂ on the side chain of Glu171 resulted in a less optimal dyad with Asp195 and a slight TM2 rearrangement away from TM6 (Figure 4C). In the Glu171-protonated condition, all the interactions among Arg60, Glu171, and Asp195 are lost (Figure 4D). In both Asp171- and Glu171-protonated conditions, the absence of the salt bridge between Arg60 and Asp171/Glu171 caused TM2 to move downward and also away from TM6 (Figures 4B, D). Such movements are obviously in a direction towards the open state. Thus, the results of our MD simulations are consistent with that protonation of 171 facilitates the conformational transition from the closed to the open state.

Figure 4. Molecular dynamics simulations of the Asp171-Asp195 dyad.

The results of MD simulations show that the mutation D171E and/or the protonation of the residue at this position disrupt the tight interactions among Arg60, Asp171, and Asp195, and result in a rearrangement of TM2 away from TM6. Representative frames of the simulations are shown for WT Asp171-deprotonated (A), WT Asp171-protonated (B), D171E Glu171-deprotonated (C), D171E Glu171-protonated (D). Note the Asp195 is protonated in all conditions. The interactions that are within the H-bond distance are indicated by dotted lines.

Discussion

In eukaryotes, cytosolic [Ca²⁺] is kept very low in the 100 nM range, which enables cells to exploit Ca²⁺ to encode signals with distinct spatiotemporal properties by transiently and locally increasing cytosolic [Ca²⁺] (Berridge et al., 2003). Ca²⁺ at a low concentration acts as a second messenger and drives a broad variety of cellular processes in response to extracellular stimuli, including exocytosis, contraction, metabolism, transcription, fertilization and proliferation (Berridge et al., 2003). The low cytosolic [Ca²⁺] is achieved by Ca²⁺-binding proteins such as parvalbumin and calbindin and by Ca²⁺-transport systems that pump Ca²⁺ out of organelles such as the ER via SERCA and Golgi via SPCA or that extrude Ca²⁺ to the extracellular environment via plasma membrane Ca²⁺ ATPases (PMCA) or Na⁺/Ca²⁺ exchangers (Carafoli and Krebs, 2016). Consequently, free Ca²⁺ concentrations in the ER compartments and the extracellular environment are in the ~500 μM and low mM range, respectively, which are about four orders of magnitude higher than in the cytosol. These stores provide the cells with plenty of Ca²⁺ that is available to use for signaling purposes.

However, overloaded intracellular Ca²⁺ stores can result in excessive Ca²⁺ transfer between the ER and the mitochondria, which can initiate Mitochondria Outer Membrane Permeabilization (MOMP) and subsequent cell death (Mattson and Chan, 2003; Pinton et al., 2008; Leanza et al., 2013). In fact, several oncogenes and tumor suppressors appear to exploit control over the ER Ca²⁺-filling state as a means to impact cellular fate (Bittremieux et al., 2016). Thus, not surprisingly, the Ca²⁺-flux properties of TMBIM-family members have been implicated in cell death control (Rojas-Rivera et al., 2012; Kiviluoto et al., 2013; Carrara et al., 2015). Despite their role in these central cellular functions, the molecular events that govern the activity and/or regulation of the TMBIM Ca²⁺ leak channels are not well understood (Sokolov et al., 2007). In the present study, our experimental data show that a bacterial TMBIM homolog BsYetJ can also lead to Ca²⁺ flux from ER Ca²⁺ stores from mammalian cells, suggesting that the Ca²⁺ flux activity of BsYetJ is conserved. Indeed, we previously used yeast polar lipids for BsYetJ-reconstituted Ca²⁺ flux assays, suggesting that BsYetJ is also capable of mediating Ca²⁺ flux in yeast cells (Chang et al., 2014). Interestingly, in humans, there are six known TMBIM proteins localized to different cellular and subcellular membranes (Lisak et al., 2015; Rojas-Rivera and Hetz, 2015; Liu, 2017). Our data present evidence that a bacterial TMBIM homolog can mediate Ca²⁺ fluxes from intracellular compartments of human cells. Moreover, it provides direct evidence that the Ca²⁺-flux property of TMBIM-related proteins is highly conserved from bacteria to human, indicating that the ability to mediate Ca²⁺ flux is an inherent property of the TMBIM family occurring already very early in evolution (Figure S1). However, it remains to be tested whether BsYetJ could suppress cell death as its mammalian relatives do.

Structurally, BsYetJ and likely the entire TMBIM family are different from canonical Ca²⁺ channels where there is a well-defined oligomeric pore and selectivity filter (Carrara et al., 2017). It has been hypothesized that the C-terminus of human TMBIM6 may form a re-entry loop and harbor a pore for Ca²⁺ flux (Xu and Reed, 1998; Bultynck et al., 2012). In the closed-state ^LCPBsYetJ and ^C10E5BsYetJ structures, the C-terminus is a transmembrane helix and is wrapped around by six other transmembrane helices. Two species-conserved residues, Asp195 and Asp171 (Figure S1), forms a hydrogen bond and we have hypothesized their roles in pH sensing and Ca²⁺ activity (Bultynck et al., 2014; Chang et al., 2014). In this study, we mutated either residue to a glutamate and showed that such mutation, even though retaining the overall negative charge, was defective to bind Ca²⁺ and lost the pH-dependent profile as the wild-type (Figures 3C, D). Our biochemical and structural data demonstrate that the two aspartate residues are important in defining a Ca²⁺ passage pore, which is permeable for Ca²⁺. It is noticed that Ca²⁺ affinity can be inhibited by monovalent ions Na⁺, K⁺, and Li⁺. So it is possible that these ions may all affect the binding of Ca²⁺ to the two mutants. In human TMBIM6, mutating Asp213 (equivalent to Asp195 in BsYetJ) to an arginine affected Ca²⁺ flux (Bultynck et al., 2012). Moreover, in contrast to wild-type TMBIM6, its D213R mutant was unable to protect cells against thapsigargin-induced cell death (Kiviluoto et al., 2013). Similarly, in human TMBIM4, mutating Asp219 (equivalent to Asp195 in BsYetJ) to an asparagine also affected the Ca²⁺ flux. Importantly, this D219N mutant lost its protective role in apoptosis, thus firmly established the connection of Ca²⁺ flux to apoptosis (Carrara et al., 2015). Learning from human relatives, mutants on Asp195 and Asp171 should help to decipher physiological roles of BsYetJ and its homologs in prokaryotes.

In our SPA-based ⁴⁵Ca²⁺ binding assays we found that monovalent ions Na⁺, K⁺ and Li⁺ inhibited Ca²⁺ binding to BsYetJ at physiological concentrations. It is possible that monovalent cations may regulate the basal channel activity of BsYetJ. Whether BsYetJ is also leaky to these monovalent ions is unknown and will be an intriguing question for future research. Interestingly, divalent ions, Mg²⁺ and Mn²⁺ do not show apparent inhibition, suggesting that the channel is more specific to Ca²⁺ perhaps owing to the unique size of Ca²⁺. This size selection is supported by the inhibitory effects by several lanthanides (Gd³⁺, Yb³⁺, Lu³⁺) that are Ca²⁺ surrogates of similar ionic radius. Based on the interkindom- mediated functional complementation of Ca²⁺ flux in mammalian cells by BsYetJ, it is feasible to assume that all TMBIM proteins possess Ca²⁺ flux activity. Furthermore, it seems likely that TMBIM family members share an ion recognition pattern that is similar to that of BsYetJ, but this hypothesis requires experimental validation through studies targeting eukaryotic family members. In our SPA measurements performed at different pHs, we found that Ca²⁺ binding follows a bell-shaped curve with a maximum at around pH 7.5 (Figure 3C). At this pH, BsYetJ wild-type is much more active than the two dyad mutants, suggesting that the sensor can effectively detect environmental changes in pH, undergo protonation steps, and conduct Ca²⁺ flux. These observations correlate very well with recent work showing that Ca²⁺ flux through the C-terminal peptide corresponding to TMBIM6’s Ca²⁺-channel- pore domain displays a bell-shaped dependence towards pH (Kiviluoto et al., 2013).

In bacteria, the physiological roles of TMBIM-mediated Ca²⁺ leak remain elusive. Nevertheless, intracellular Ca²⁺ is key to bacterial chemotaxis, mobility, virulence, and adaptation to environment (Dominguez et al., 2015). Although there are no intracellular Ca²⁺ stores in bacteria, several components of the Ca²⁺ signaling toolkits are expressed, including Ca²⁺-binding proteins and Ca²⁺- transport systems, such as P-type ATPases, electrochemical potential driven transporters and channels (Dominguez, 2004; Shemarova and Nesterov, 2014; Dominguez et al., 2015). While basal cytosolic [Ca²⁺] in prokaryotes is kept low in the 100-300 nM range, cytosolic [Ca²⁺] can transiently rise in responses to nitrogen starvation, environmental stress, and production of metabolites (Dominguez et al., 2015). Consequently, the expression of hundreds of genes is affected transiently by the increased cytosolic [Ca²⁺], indicating a key role for Ca²⁺ signaling in the control of physiological processes in bacteria (Dominguez et al., 2015).

STAR*Methods

Contact for Reagent and Resource Sharing

Further information and reasonable requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Qun Liu (qunliu@bnl.gov).

Experimental Model and Subject Details

Cell Culture

Adherent HeLa cells were cultured in DMEM medium (Invitrogen) at 37 °C and 5% CO₂. The culture was supplemented with 10% FBS (Invitrogen). For maintenance, HeLa cells were sub-cultured twice a week and at least three passages were performed prior to experiments.

Method Details

Protein Production

Expression and purification of N-terminal truncation (Δ5BsYetJ), ^D195EBsYetJ, and ^D171EBsYetJ were performed using the previously established protocols for the wild-type BsYetJ (Chang et al., 2014). Briefly, ^D195EBsYetJ and ^D171EBsYetJ mutations were generated by using a QuikChange Site-directed mutagenesis kit. pMCSG7 plasmid harboring the wild-type YetJ gene (Chang et al., 2014) was used as template to generate the two mutations. For the ^D171EBsYetJ mutation, primers 5'- ttttccctttacattttgtatgagttgaaccaaatcaagcaccgcca-3' (F) and 5'- tggcggtgcttgatttggttcaactcatacaaaatgtaaagggaaaa-3' (R) were used. For the ^D195EBsYetJ mutation, primers 5'-ggcgttatcgctgtacctagagtttatcaacttgttcatcaac-3' (F) and 5'- gttgatgaacaagttgataaactctaggtacagcgataacgcc-3' (R) were used. Both mutations were verified by sequencing. Protein expression was conducted in Escherichia coli BL21 pLysS cells growing in Terrific Broth medium. All purification steps were performed at 4 °C. Frozen cell pellets were suspended in a suspension buffer that contains 50 mM HEPES, pH 7.8, 300 mM NaCl, and 20 mM imidazole, 5% glycerol, and 1 mM MgCl₂. Cells were lysed by using an EmulsiFlex-C3 homogenizer (Avestin, Ottawa, Canada); and cell debris was removed by centrifugation at 15,000g for 30 min. Membranes were collected by centrifugation at 150,000g for 60 min and solubilized by addition of n-dodecyl β-D- maltopyranoside (DDM) to a final concentration of 1.5% (w/v) for 2 hrs. Cleaned-up supernatant was loaded to a HisTrap™ FF column (GE Healthcare, Inc.) and the protein was eluted by a linear gradient concentration of imidazole from 75 mM to 500 mM in the buffer of 50 mM HEPES, pH 7.8, 300 mM NaCl, 5% glycerol and 0.05% DDM. The affinity tag was cleaved by overnight incubation with sufficient amount of tobacco etch virus (TEV) protease and the enzyme was then removed by a second passage through the nickel column. The flow-through was concentrated and detergent was changed to n- dodecyl-N,N-dimethylamine-N-oxide (LDAO) by gel-filtration on a Superdex 200 10/300 GL column (GE Healthcare, Inc) with an elution buffer containing 25 mM Tris-HCl pH 8.0, 400 mM NaCl, 0.05% LDAO (w/v). Purified protein was concentrated to 10-13 mg/mL with an Amicon Ultra-15 centrifugal filter (Milipore, Inc).

Crystallization

For LCP crystallization, we used the N-terminal truncations (Δ5BsYetJ). To make lipidic cubic phase, we mixed 10 μL of protein (10 mg/mL) with 15 μL of monoolein (1:1.5 v/v) to get 40% hydration. Crystals were grown from the condition by mixing 50 nL of LCP with 800 nL of precipitant that contained 0.1 M Tris, pH 7.8, 5% PGA-LM (w/v), and 30% PEG 400 (v/v). The LCP crystals were harvested and cryocooled for cryogenic data collection.

For crystallization of ^D171EBsYetJ and ^D195EBsYetJ mutants, equal volumes of protein (10-12 mg/mL) in detergent LDAO were mixed with precipitants that contained 0.1 M MOPS, pH 7.5, 38% PEG 400, and 0.2 M NaCl (for ^D171EBsYetJ) or 0.1 M ADA, pH 7.0, 30% PEG 600 (v/v) (for ^D195EBsYetJ). For both conditions, only needles crystals were obtained. These needles were harvested and cryocooled for cryogenic data collection.

Structure Determination

Diffraction data were collected at National Synchrotron Light source (NSLS) beamline X25 with a Quantum 315R detector and Advanced Photon Source (APS) NE-CAT 24-ID-C beamline with a Pilatus 6M detector under a cryogenic temperature of 100 K. All data sets were indexed and integrated by XDS (Kabsch, 2010) and scaled and merged by CCP4 program AIMLESS (Winn et al., 2011). Data collection and reduction statistics for single and multi-crystal data sets are listed in Table 1.

Table 1.

Data collection and refinement statistics

Data collection	LCPBsYetJ	D171EBsYetJ	D195EBsYetJ
Wavelength (Å)	0.979	0.979	1.1
Space group	P21	C2	C2
Cell dimensions	33.86, 62.04, 47.34	218.09, 117.30, 52.41	217.33, 118.20, 51.89
a, b, c (Å) β (°)	93.95	92.96	93.23
Solvent content (%)	44.0	65.0	65.1
Bragg spacings (Å)	37.39-2.50 (2.63-2.50)	47.15-3.10 (3.31-3.10)	39.96-3.1 (3.31-3.10)
Total reflections	46387	76248	92561
Unique reflections	6634	23721	23817
Completeness (%)	95.3 (78.4)	99.1 (99.7)	99.8 (99.9)
I/σ(I)	6.0 (2.4)	5.4 (1.1)	6.0 (1.1)
Rmeas	0.218 (0.939)	0.266 (1.429)	0.312 (2.124)
Multiplicity	7.0 (6.5)	3.2 (3.2)	3.9 (3.9)
CC1/2	97.3 (77.0)	98.1 (36.6)	98.6 (37.2)
Refinement
Resolution (Å)	2.5	3.1	3.1
No. reflections	6273	23641	23799
Rwork/Rfree	0.243/0.260	0.243/0.271	0.247/0.276
No. atoms	1624	6430	6416
Protein	1592	6430	6416
Water	31	-	-
Average B (Å2)	39.6	61.53	68.0
Protein	39.5	61.53	68.0
Water	44.0	-	-
R.m.s deviations
Bond length (Å)	0.007	0.002	0.002
Bond angle (°)	0.986	0.454	0.409
PDB code	6NQ7	6NQ8	6NQ9

Structures were determined by molecular replacement method by PHASER (Read and McCoy, 2011) with starting model of either the closed or open state (Chang et al., 2014). Specifically, for the LCP structure, we used the closed state structure at pH 8 (PDB code 4PGR) for molecular replacement; and for the two mutants, we tried both the closed (PDB code 4PGR) and the open state structures (PDB code 4PGS). The LCP crystal was crystallized in space group P2₁ and molecular replacement solution is performed with only one molecule in asymmetric unit. The two mutants were both crystallized in space group C2 with four molecules in an asymmetric unit. Molecular replacement with the open-state structure gave the highest Z score as reported by PHASER.

The three structures were refined by CCP4 program REFMAC5 and PHENIX (Adams et al., 2010; Murshudov et al., 2011). Model and solvent adjustment and validation were performed iteratively with COOT (Brown et al., 2015) and PHENIX (Afonine et al., 2012). For the two mutant structures, non- crystallographic symmetry was used for restraints and TLS parameters were refined to model anisotropy. The stereochemistry of refined structures was validated with PROCHECK (Laskowski et al., 1993) and MOLPROBITY (Chen et al., 2010) for quality assurance. Data statistics for refinements were listed in Table 1.

Unidirectional ⁴⁵Ca²⁺-flux assay in HeLa cells

Unidirectional ⁴⁵Ca²⁺ efflux assays were performed following the previously established protocols (Luyten et al., 2014). We performed the experiments in 24-well tissue culture plates of which only 12 wells were analyzed. Confluent monolayers of HeLa cells were obtained by seeding 30,000 cells per well and growing them for 7 days at 37 °C and 5% CO₂ in DMEM medium. The culture medium was removed, and the cells were permeabilized in a solution containing 120 mM KCl, 30 mM imidazole-HCl, pH 6.8, 2 mM MgCl₂, 1 mM Na-ATP, 1 mM K-EGTA, and 20 μg saponin ml⁻¹. After permeabilization for 10 mins at 25 °C, Ca²⁺ stores were loaded for 45 min with ⁴⁵Ca²⁺ in a solution containing 120 mM KCl, 30 mM imidazole-HCl, pH 6.8, 5 mM MgCh, 5 mM Na-ATP, 0.44 mM K-EGTA, and 150 nM free Ca²⁺ mixed with ⁴⁵Ca²⁺ (with a specific activity of 0.3 MBq/mL). Then, 200 μl of efflux buffer containing 120 mM KCl, 30 mM imidazole-HCl, pH 6.8, and 1 mM K-EGTA was added and removed every 2 min. The indicated BsYetJ (2.5 mM) or buffer was added. At the end of each experiment, the ⁴⁵Ca²⁺ remnant in the ER was determined by incubating with 200 μl of 2% (w/v) SDS solution for 30 min.

Scintillation Proximity Assay (SPA)

Binding of ⁴⁵Ca²⁺ by BsYetJ wild-type and its mutants was measured with the SPA. For the assay 50 ng of protein was incubated with 10 μM ⁴⁵Ca²⁺ (746 Ci/mmol) in the presence of 2.5 mg/mL YSi His tag SPA beads in 500 mM Tris/MES, with a pH ranging from pH 6.0 to 8.0, 20% glycerol (w/v), 0.1 % DDM (w/v) and 1 mM TCEP. Tris/MES was equimolarly replaced with indicated salts. The specific cpm of ⁴⁵Ca²⁺ binding was determined by subtracting the non-proximity-based cpm, determined in the presence of 400 mM imidazole which prevents binding of the His-tagged recombinant protein to the YSi His tag SPA beads, from the total signal). Data of representative figures are shown as the mean ± S.E.M of triplicate determinations with at least three independent repeats. Non-linear regression fitting was performed in Sigma Plot 13 (Systat software), and the results are shown as mean ± S.E.M. of the fit.

Molecular Dynamics Simulations

The molecular models were all constructed starting from the closed-state structure of the wild-type YetJ (PDB ID 4PGR). MD simulations of these models were performed in the explicit water and 1-palmitoyl- 2-oleoylphosphatidylcholine (POPC) lipid bilayer environment using Desmond MD System (version 4.5; D. E. Shaw Research, New York, NY) with the CHARMM36 force field (MacKerell et al., 1998; MacKerell et al., 2004; Klauda et al., 2010; Best et al., 2012) and TIP3P water model. The system charges were neutralized, and 150 mM NaCl was added. The average size of the simulation systems was ~68000 atoms. The Protein-membrane relaxation was carried out with a protocol modified from that developed by Schrodinger, LLC. Briefly, the initial energy minimization was followed by equilibration with restraints on all protein and ligand heavy atoms in the beginning for 1 ns, then with restraints only on the protein backbone for 6 ns. For both the equilibrations and the following unrestrained production runs, we used Langevin constant pressure and temperature dynamical system (Feller et al., 1995) to maintain the pressure at 1 atm and the temperature at 310K, on an anisotropic flexible periodic cell with a constant-ratio constraint applied on the lipid bilayer in the X-Y plane.

For each condition, we collected multiple trajectories: WT Asp171-deprotonated (2 trajectories, totally 3.6 μs), WT Asp171-protonated (8 trajectories, totally 10.8 pμs), D171E Glu171-deprotonated (10 trajectories, totally 6 μs), D171E Glu171-protonated (10 trajectories, totally 6 μs). Figure 4 consists the representative frames of each condition.

Quantification and Statistical Analysis

Unidirectional flux experiments were done in duplicates with purified protein sample. SPA experiments were done in triplicates with purified protein samples.

Data and Software Availability

Data Resources

Atomic coordinates and structure factor files have been deposited in the RCSB Protein Data Bank (PDB) under the accession code 6NQ7 for wild-type ^LCPBsYetJ structure, 6NQ8 for ^D171EBsYetJ structure, and 6NQ9 for ^D195EBsYetJ structure.

Highlights:

• BsYetJ mediates Ca²⁺ flux in mammalian cells

• The Ca²⁺-binding activity is sensitive to protons and other cations

• The Asp171-Asp195 dyad controls pore opening and pH-dependent Ca²⁺ binding

• Protonation of Asp171 leads to an open-state structure