Structure and mechanism of the mitochondrial calcium transporter NCLX

Protein expression and purification

The rat NCLX gene with a Strep tag at the C terminus was cloned to the BacMam vector⁶¹. The NCLX virus was generated and used to infect HEK293 cells at a cell density of 3 × 10⁶ cells per millilitre at 37 °C. After 12 h, the cells were cultured in the presence of 10 mM sodium butyrate at 30 °C. After another 48 h, cells were harvested and stored at −80 °C for further use.

To purify the NCLX protein under the high-Ca²⁺ condition (2 mM Ca²⁺), the cells were dounce-homogenized in buffer A (50 mM HEPES pH 7.4, 150 mM NaCl, 2 mM CaCl₂) with protease inhibitors and ribonuclease. The membranes were solubilized at 4 °C for 2 h with the addition of 1.5% LMNG (Anatrace) and cholesteryl hemisuccinate (Anatrace) mixture. Following spinning at 18,000 r.p.m. for 45 min, the supernatant was mixed with Strep-Tactin resin (IBA) and incubated for 2 h at 4 °C. After extensive washing using buffer B (150 mM NaCl, 2 mM CaCl_2, 20 mM HEPES pH 7.4) containing 0.05% glyco-diosgenin (GDN, Anatrace), the protein was eluted using buffer B with 0.01% GDN and 10 mM desthiobiotin. To polish the protein, the eluted sample was concentrated and subjected to further purification using a Superose 6 increase column (Cytiva) in buffer B containing 0.006% GDN. The peak fractions of NCLX were concentrated to 15 mg ml^–1 for preparing cryo-EM grids. To purify the NCLX protein in a Ca²⁺-free state at pH 7.4, the 2 mM CaCl₂ was replaced with 5 mM EGTA. To purify the protein in a calcium-free state at low pH, the protein was initially extracted in the same Ca²⁺-free pH 7.4 conditions as above and gradually exchanged into buffer C (20 mM buffer with sodium acetate-acetic acid pH 5.5, 150 mM NaCl, 5 mM EGTA) containing 0.006% GDN for preparing cryo-EM grids.

Cryo-EM sample preparation and data collection

Protein sample (3 µl) was used for each cryo-EM grid. The glow discharged grids (300 mesh, R2/1, Au, holey carbon, Quantifoil) with the sample were blotted using a Vitrobot Mark IV for 3 s (4 °C and 100% humidity) or a Leica EM GP2 for 2 s (4 °C and 96% humidity), before being plunge frozen in liquid ethane.

For the NCLX sample without calcium (pH 7.4), datasets were collected on Titan Krios (300 keV; with GIF-Quantum Energy Filter) at cEMc/Stanford or S2C2/Stanford. Images were recorded using K3 Summit detector (Gatan) (super-resolution counting mode; 105,000× magnification; 0.86 Å physical pixel size) by serialEM. The Multiple Record method was used for data collection (image shift; 3 × 3). A total of 50 video frames were collected.

For the NCLX sample with calcium, and the NCLX low-pH sample without calcium, datasets were acquired using EPU software, similarly on Titan Krios (cEMc/Stanford or S2C2/Stanford) with Falcon 4 detector (Thermo Fisher Scientific; counted mode) at 96,000× nominal magnification (0.82 Å physical pixel size) or 130,000× nominal magnification (0.946 Å or 0.95 Å physical pixel size). Forty movie frames were collected in MRC format, with a dose of ~50 electrons per Ų.

Cryo-EM data processing

For NCLX without calcium data (pH 7.4), a total of 4,264 movies were processed by cryoSPARC live⁶² v4.2 (Patch motion correction, Patch CTF estimation and template picking). Then, 821,498 particles were extracted with 80-pixel box and 4 × 4 binning from 3,241 images (with <8 Å CTF estimates). Heterogeneous refinement was performed over three rounds, using accurate and biased maps. This resulted in 95,304 particles for subsequent steps. Furthermore, 57,247 particles from 1,304 images were selected for Topaz (v.0.2.4)⁶³ training. Topaz repicking subsequently yielded 659,059 particles (extracted with 80-pixel box and 4 × 4 binning), which underwent 3D classification using a seed-facilitated multi-reference strategy⁶⁴. The reference sets included (1) non-uniform refinement accurate map; (2) maps with a resolution gradient (an accurate map along with low-pass filtered maps at 10 Å and 20 Å); (3) biased maps via ab initio reconstruction; and (4) maps with noise re-weight (accurate map and maps with micelles scaled down by 0.3 and 0.7). After removing duplicated particles, 208,064 particles were re-extracted with a 320-pixel box. Two rounds of heterogeneous refinement were performed with noise reweighted and noise accurate maps, respectively. Afterwards, 86,447 selected particles gave rise to a 3.29 Å map (class 1) through successive rounds of non-uniform, Local CTF and Local refinements. The map was sharpened by DeepEMhancer⁶⁵. Similarly, 34,185 particles underwent non-uniform refinement, yielding a 3.93 Å map (class 2) and 26,871 particles yielded a 4.29 Å map (class 3). The local resolution estimate was performed by BlocRes.

For NCLX with calcium data, datasets A and B were collected and processed via the same processing strategy; thus, dataset A is described in the following steps. For dataset A, 13,255 videos were motion corrected using MotionCorr2 (ref. ⁶⁶). CryoSPARC (v.4.2) was used to process the dose-weighted images, including Patch CTF estimation. Topaz picking resulted in 3,180,107 particles (80-pixel box and 4 × 4 binning) from 12,725 images (with <8 Å CTF estimates). Heterogeneous refinement was performed over three rounds with accurate and biased maps. This resulted in 390,406 seed particles for subsequent processing. Furthermore, the previous 3,180,107 particles underwent 3D classification using the seed-facilitated multi-reference strategy with similar multiple-references as described above. After removing duplicates, the resulting 948,917 particles were combined with the seed particles, yielding 1,021,442 particles with duplicate removal. Two rounds of heterogeneous refinement were performed using noise reweighted maps. Afterwards, the re-extracted particles (with 320-pixel box) were subjected to two rounds of heterogeneous refinement with noise reweighted and accurate maps, respectively. The outcome was two classes with 334,585 particles and 191,077 particles, which were combined with 155,219 and 230,218 particles from dataset B using the same processing strategy as dataset A, respectively. The combined particles underwent heterogeneous, non-uniform, local CTF and non-uniform refinements. Refinement of a high-quality subset of 150,233 particles then yielded a 3.31 Å map (class 3a), and 145,777 particles yielded a 2.97 Å map (class 2a). The selected 132,072 particles were refined, yielding a 3.10 Å map (class 4a) after 265,702 particles underwent 3D classification without alignment. All maps were sharpened by DeepEMhancer for improved density. For dataset C, 10,373 videos were motion corrected using MotionCorr2. The dose-weighted images were imported and estimated by Patch CTF estimation using cryoSPARC v4.4. After Blob picking from 500 images, 158,465 selected particles were subjected to 2D classification and ab initio refinement, yielding 97,913 particles as seeds. Simultaneously, following template picking for whole images, 9,112,846 particles (extracted with 80-pixel box, 4 × 4 binning) underwent 3D classification using seed-facilitated multi-references. After duplicate removal, 1,423,457 re-extracted particles (with 320-pixel box) underwent three rounds of heterogeneous refinement (one with accurate and biased maps, followed by two with accurate map). The outcome was two classes with 686,084 and 205,206 particles, respectively. Refinement of the selected subset of 686,084 particles then yielded a 2.15 Å map with C3 symmetry (class 4a). The other 205,206 particles were processed via heterogeneous refinement, yielding 115,207 particles, which were further refined to produce a 2.60 Å map (class 3a). Classes 3a and 4a exhibit higher resolution compared to those in datasets A and B. Consequently, these higher-resolution maps derived from dataset C are retained for use in the paper. The local resolution estimate was performed using BlocRes in cryoSRARC.

For NCLX at low-pH (without calcium) sample, a total of 9,653 videos were imported into cryoSPARC live v4.4 and processed through steps similar to those described above for the NCLX dataset without calcium at pH 7.4, including particle extraction with 80-pixel box and 4 × 4 binning. This resulted in 5,574,228 particles from 9,593 images with CTF estimate better than 8 Å. Three-dimensional classification using the seed-facilitated multi-reference strategy yielded 816,107 particles, following duplicate removal. In parallel, following topaz picking, 2,221,595 particles underwent 2D classification (two rounds) and heterogeneous refinement (three rounds), resulting in 115,569 particles as seeds. These 2,221,595 particles underwent 3D classification using the seed-facilitated multi-reference strategy, which resulted in 805,025 particles after duplicate removal. These two sets of particles were combined, resulting in 1,117,206 particles after duplicate removal. Following one round of heterogeneous refinement using noised-reweighted map, 683,191 particles were re-extracted with 320-pixel box. Finally, these particles were refined with C₃ symmetry, yielding a 2.62 Å map (class 4b). The local-resolution estimate was performed using BlocRes in cryoSRARC.

Model building and refinement

The initial model used for NCLX structure was obtained through AlphaFold2⁶⁷ prediction. Chimera was then used to fit the initial model into the cryo-EM map, which was manually rebuilt using Coot⁶⁸. The rebuilt model was then subjected to refinement in Phenix⁶⁹ to optimize its geometry and stereochemistry and was assessed by MolProbity. The two structures, including class 1 from NCLX without calcium, and class 2a from NCLX with calcium, were manually adjusted and refined, the other structures used classes 1 and 2a as initial model and were manually adjusted and fitted into the corresponding map. Compared with other classes, density maps for classes 2 and 3 (in the Ca²⁺-free, pH 7.4 condition) are at lower resolution. In these two classes, side chains were removed in regions where the corresponding maps lack sufficient information, and their register was based on corresponding models in the Ca²⁺ bound forms. Figures were generated using PyMOL (Schrödinger)⁷⁰, UCSF Chimera⁷¹ and ChimeraX⁷².

System set-up for molecular dynamics simulations

We performed simulations of NCLX under five conditions: (1) simulations of cytosol-open NCLX initiated with a Ca²⁺ ion in the binding pocket, with residues D153 and D471 deprotonated; (2) simulations of matrix-open NCLX initiated with a Ca²⁺ ion in the binding pocket and residues D153 + D471 deprotonated; (3) simulations of matrix-open NCLX initiated with no Ca²⁺ ion in the binding pocket and residues D153 + D471 deprotonated; (4) simulations of cytosol-open NCLX initiated with a Ca²⁺ ion in the binding pocket and residues D153 + D471 protonated (neutral); and (5) simulations of matrix-open NCLX initiated with a Ca²⁺ ion in the binding pocket and residues D153 + D471 protonated. We initiated all simulations from the cryo-EM structures of NCLX (based on an earlier version of the cryo-EM maps): simulation sets (1) and (4) were initiated using chain A of the Ca²⁺-bound NCLX structure; simulation sets (2) and (5) were initiated using chain C of the Ca²⁺-bound NCLX structure; and simulation set (3) was initiated using the chain A of the NCLX structure with no Ca²⁺ bound. The Ca²⁺ ion in the Ca²⁺-bound NCLX structure was preserved for simulation sets (1), (2), (4) and (5). We performed five independent simulations for each simulation condition, each at least 1.4 µs in length. Initial atom velocities were assigned randomly and independently for each simulation.

For all simulation conditions, the protein structure was aligned with the Orientations of Proteins in Membranes⁷³ entry for 3V5U (NCX from Methanocaldococcus jannaschii²⁸) using PyMOL, and water molecules from 3V5U were incorporated. Prime (Schrödinger)⁷⁴ was used to add capping groups to protein chain’s termini. Protonation states of all titratable residues other than D153 and D471 were assigned at pH 7. Histidine residues were modelled as neutral, with a hydrogen atom bound to either the delta or epsilon nitrogen depending on which tautomeric state optimized the local hydrogen-bonding network. Using Dabble⁷⁵, the prepared protein structures were inserted into a pre-equilibrated palmitoyl-oleoyl-phosphatidylcholine bilayer, the system was solvated, and calcium and chloride ions were added to neutralize each system at a calcium concentration of 75 mM and a chloride concentration of 150 mM. The final systems comprised approximately 135,000 atoms, and system dimensions were approximately 120 × 140 × 110 Å (Supplementary Table 2).

Molecular dynamics simulation and analysis protocols

The simulation protocol was similar to that used in previous work⁷⁶, as those simulations proved sufficient to describe atomic-level interactions between a membrane transporter and its substrate. We used the CHARMM36m force-field for proteins, the CHARMM36 force-field for lipids and ions, and the TIP3P model for water molecules^77,78,79. All simulations were performed using the Compute Unified Device Architecture version of particle-mesh Ewald molecular dynamics in AMBER20 on graphics processing units.

Systems were first minimized using three rounds of minimization, each consisting of 500 cycles of steepest descent followed by 500 cycles of conjugate gradient optimization. Harmonic restraints (10.0 and 5.0 kcal mol⁻¹ Å⁻²) were applied to the protein and lipids for the first and second rounds of minimization, respectively. Then, 1 kcal mol⁻¹ Å⁻² harmonic restraints were applied to the protein for the third round of minimization. Systems were then heated from 0 K to 100 K in the NVT ensemble over 12.5 ps, and then from 100 K to 310 K in the NPT ensemble over 125 ps, using 10.0 kcal mol⁻¹ Å⁻² harmonic restraints applied to protein heavy atoms. Subsequently, systems were equilibrated at 310 K and 1 bar in the NPT ensemble, with harmonic restraints on the protein non-hydrogen atoms tapered off by 1.0 kcal mol⁻¹ Å⁻² starting at 5.0 kcal mol⁻¹ Å⁻² in a stepwise fashion every 2 ns for 10 ns, and then by 0.1 kcal mol⁻¹ Å⁻² every 2 ns for 20 ns. Production simulations were performed without restraints at 310 K and 1 bar in the NPT ensemble using the Langevin thermostat and the Monte Carlo barostat, and using a timestep of 4.0 fs with hydrogen mass repartitioning⁸⁰. Bond lengths were constrained using the SHAKE algorithm⁸¹. Non-bonded interactions were cut off at 9.0 Å, and long-range electrostatic interactions were calculated using the particle-mesh Ewald method with an Ewald coefficient of approximately 0.31 Å⁻¹, and fourth-order B-splines. The particle-mesh Ewald grid size was chosen such that the width of a grid cell was approximately 1 Å. Trajectory frames were saved every 200 ps during the production simulations. The AmberTools17 CPPTRAJ package was used to reimage trajectories⁸². Simulations were visualized and analysed using Visual Molecular Dynamics (VMD)⁸³ and PyMOL⁷⁰. The D153–Ca²⁺ distance is the minimum distance between a side-chain oxygen of D153 and a Ca²⁺ ion. The D471–Ca²⁺ distance is the minimum distance between a side-chain oxygen of D471 and a Ca²⁺ ion. The N467–Ca²⁺ distance is the minimum distance between the backbone oxygen of N467 and a Ca²⁺ ion. The N149–Ca²⁺ distance is the minimum distance between the backbone oxygen of N149 and a Ca²⁺ ion. In Extended Data Fig. 7a, to construct the probability distributions for these distance metrics, we used trajectory frames from all simulations under each condition and applied a Gaussian kernel density estimator.

Proteomic sample preparation for mass spectrometry analysis

The mass spectrometry (MS) sample preparation, data collection and data analyses follow similar protocols to those previously reported⁸⁴. To prepare NCLX for liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis, human NCLX with a C-terminal Strep-tag II was cloned into the BacMam vector and expressed and purified as described above for rat NCLX. Purified NCLX was then processed for proteomics analysis using an S-Trap Micro Column (C02-micro-80, ≤100 µg, Protifi). Briefly, purified proteins were reduced by incubating with 10 mM dithiothreitol at 95 °C for 10 min and subsequently alkylated using 40 mM 2-chloroacetamide at room temperature for 60 min. Phosphoric acid (final concentration = 1.2%) was added to acidify the denatured, non-digested proteins before loading the samples onto S-trap columns. The columns were washed three times with 100 mM triethylammonium bicarbonate (TEAB) in 90% methanol. To digest the bound proteins, each column was incubated with Trypsin (Promega, V5113) in 50 mM TEAB at a 1:20 trypsin-to-substrate ratio (w/w) at 37 °C overnight. Elutions were performed in succession using 50 mM TEAB (40 μl), 0.2% formic acid (FA) (40 μl), and then 50% acetonitrile (ACN) + 0.2% FA (40 μl) to elute the peptides from the column. The combined eluate was dried using a vacuum concentrator and resuspended in 10 μl of 0.2% FA, 3% ACN.

To analyse the presence of NCLX in WT and KO HeLa cells, the mitochondrial fraction was enriched by resuspending cells in mitochondrial resuspension buffer (5 mM HEPES-KOH pH 7.2, 250 mM sucrose). Cells were lysed using dounce homogenization, and large cellular debris were pelleted by centrifuging the lysate at 1,500 g for 10 min. The supernatants were then collected and further centrifuged at 13,000 g for 10 min, and the pellet was resuspended in mitochondrial resuspension buffer. This resuspension was then centrifuged at 13,000 g for 10 min to pellet mitochondria. To solubilize NCLX, the mitochondria were incubated with 1.5% m/v LMNG/cholesteryl hemisuccinate in 20 mM HEPES-NaOH, 150 mM NaCl, pH 7.4 supplemented with protease inhibitors and ribonuclease for 2 h at 4 °C. The detergent-solubilized mixture was then centrifuged at 18,000 r.p.m. (39,191 g) for 40 min to pellet any insoluble material. The resulting supernatant was collected and either directly run on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), or the supernatant was fractionated by size-exclusion chromatography on a Superose 6 column to isolate the fractions corresponding to the elution volume of NCLX, which were then run on SDS-PAGE. The gel was visualized with AcquaStain (Bulldog Bio, AS001000) and the gel fractions in the expected molecular mass range of NCLX were excised for MS analysis. The excised gel pieces were washed with H₂O and 100% ACN, reduced using 50 mM ammonium bicarbonate pH 8.5 (ABC) with 6.5 mM dithiothreitol for 60 min at room temperature, alkylated using 50 mM ABC with 54 mM iodoacetamide for 30 min in the dark at room temperature, and washed with alternating 100% ACN and ABC solutions. To digest the peptides, the dehydrated gel pieces were pre-incubated with digestion buffer (3 ng µl^–1 trypsin in ABC) for 60 min on ice. The excess digestion buffer was then replaced with an equivalent volume of ABC, and the samples were incubated overnight at 37 °C for protein digestion. The gel pieces were treated with 100% ACN to extract the peptides, which were dried in a vacuum concentrator and resuspended in 10 μl of 0.2% FA, 3% ACN for LC–MS/MS analysis with parallel reaction monitoring (PRM).

LC–MS/MS analysis

Data-dependent acquisition (DDA) analysis of purified, tryptic, human NCLX peptides was performed to build a spectral peptide library and to generate the PRM inclusion list yielding expected m/z and retention time windows for twelve unique NCLX peptides, which was subsequently used to examine the presence of NCLX in WT and KO HeLa samples. 2 µl of reconstituted peptides (~0.57 µg) were loaded onto a Dionex Ultimate 3000 RPLC nano system (Thermo Fisher Scientific) for liquid chromatography using mobile phases A (0.2% FA in water) and B (0.2% FA in ACN). The samples were first processed on an Acclaim PepMap 100 C18 trap column (Thermo Fisher Scientific, 164213) using 100% A at 5 µl min^–1 flow rate. The peptides were then separated using a nanoflow ultra-high-performance liquid chromatography column (Aurora Ultimate 25 cm, IonOpticks, AUR3-25075C18) with an elution gradient of 2–28% over 66.5 min at 300 nl min^–1. An Orbitrap Fusion Tribrid MS system (Thermo Fisher Scientific) was used to analyse eluted peptides. A Nanospray Flex Ion Source (Thermo Fisher Scientific) held at +2.2 kV was used to ionize precursors, with an inlet capillary temperature of 275 °C. The MS1 scans for both DDA and PRM methods were collected over 350–1,350 m/z (resolution = 60,000 at 200 m/z; automatic gain control (AGC) target = 400,000; RF lens = 30%; maximum injection time = 118 ms; normalized AGC target = 100%). The MS2 scans for both DDA and PRM methods were collected over 140–1,400 m/z (isolation window = 1.6 m/z; AGC target = 50,000; maximum injection time = 22 ms; Orbitrap resolution = 15,000 at 200 m/z; higher-energy collisional dissociation collision energy = 30%). The DDA method used a top 20 precursor pick for MS2 acquisition. The MS2 scans for the PRM method were collected using a targeted mass-triggered scan function using an inclusion list containing reference m/z and corresponding retention time windows (Supplementary Table 1).

LC–MS/MS data analysis

Datasets were imported into Skyline (v.23.1.1.520; MacCoss Lab software). The confidence of the peptide identity was assessed by the number of fragment ions, and their retention time compared with the purified NCLX peptides. Confidently identified peptides were determined by the presence of at least three fragment ions (<10 ppm error) within a 4 min retention time window from the purified NCLX peptides. Furthermore, transition ions were compared to library spectrum for the same peptide. Mass spectra for peptides, SLGVVFR and ALNPLDYMK, were obtained using FreeStyle 1.8 SP2 (Thermo Fisher Scientific).

Cell culture and molecular biology

Chinese hamster ovary, HEK293 and HeLa cells were cultured in DMEM with 10% fetal bovine serum. HCT116 cells were cultured in McCoy’s 5A medium with 10% fetal bovine serum. All of these vertebrate cell lines were incubated at 37 °C with 5% CO₂. Sf9 cells were cultured in the Sf-900 III SFM medium at 27 °C.

To express human NCLX in HEK cells, a gene encoding a C-terminally 1D4 (TETSQVAPA)-tagged full-length NCLX was cloned into the pcDNA3.1(+) vector. Site-directed mutagenesis was performed using the QuickChange II kit (Agilent) and verified using Sanger sequencing. Transient expression was achieved using Lipofectamine 3000 (Thermo Fisher Scientific) following the manufacturer’s instructions. HEK cells were harvested for experiments two days after transfection. The Bac-to-Bac baculovirus expression system (Thermo Fisher Scientific) was used to express proteins in Sf9 cells. Briefly, C-terminally 1D4-tagged human NCLX or an MCU–EMRE fusion protein⁸⁵ was cloned into the pFastBac1 vector, which was transformed into DH10Bac competent cells to produce bacmids. Sf9 cells were transfected with bacmids using the Cellfectin II reagent to generate the P1 baculovirus, which was then used to infect Sf9 cells to produce the P2 virus. Sf9 cells were harvested for experiments three days after being infected with the P2 virus.

To express human NCLX in Xenopus oocyte plasma membranes, the first 28 amino acids in the protein’s N-terminal mitochondrial targeting sequence were substituted with an MAGRQHGSGRLWALGG sequence from the mouse TMIE protein, which was found to show very high levels of oocyte plasma membrane expression. The DNA sequence was then optimized for Xenopus expression and cloned into a pOX vector for in vitro RNA synthesis using the mMESSAGE mMACHINE T3 transcription kit (Thermo Fisher Scientific) as described in our previous work⁸⁶.

CRISPR KO was performed as described in a past work⁸⁷. The sgRNA sequences used for HeLa and CHO cells are 5′-CCTGGATCTACCAACGGCAA-3′ and 5′-GGCTACCTGGACTACCTCGA-3′, respectively. To confirm gene KO, we cloned (1) the NCLX gene and (2) the NCLX complementary DNA produced by reverse transcription of the NCLX mRNA into pcDNA3.1(+) for Sanger sequencing to identify the exact gene modifications. qPCR was also performed in these NCLX KO lines, showing reduced NCLX mRNA levels, consistent with nonsense-mediated mRNA decay caused by CRISPR-induced premature stop codons (Extended Data Fig. 8g). Forward and reverse primers for HEK and HeLa cells: GGCTTCACTGGCTC TTTGCTT and CGAGAAGGCATCTCCAATGCTGTT, respectively; forward and reverse primers for CHO cells: GCATCATTTTCAATATCCTGGTGG and TGAGTTGGAAACACTGAAGC, respectively. In our hands, and as also observed in other laboratories^35,88, there are currently no useful antibodies against native NCLX. We were therefore unable to further verify NCLX KO using Western blot. As an alternative approach to validate NCLX KO at the protein level, NCLX peptides were detected by LC–MS/MS (Supplementary Fig. 2), appearing only in WT but not NCLX KO HeLa cells.

Mitochondrial extraction, protease digestion and Western blot

Protein expression in HEK or Sf9 cells was analysed via Western blot detection of proteins in mitochondrial lysates. HEK cell mitochondria extraction was performed using differential centrifugation as described before⁸⁵. To obtain Sf9 mitochondria, 2 × 10⁶ of Sf9 cells were spun down, resuspended in 1 ml of an Sf9 mitochondria isolation buffer (SMIB, 200 mM sucrose, 10 mM Tris, 1 mM EGTA, pH 7.5), and then lysed by passing through a 27-gauge needle for 25 times on ice in the presence of a protease inhibitor cocktail (Thermo Fisher Scientific, PIA32955). The lysate was spun down at 600 g for 10 min. The supernatant was then transferred to a new tube, and spun down at 7,000 g for 10 min. Finally, the pellet was resuspended in 500 μl of SMIB, and spun down again at 7,000 g for 10 min to obtain mitochondrial samples. NCLX expression in Xenopus oocytes was quantified by Western blot analysis of oocyte plasma membranes, which were isolated using our published protocols⁸⁶.

To conduct protease digestion experiments, mitochondria were extracted from HEK cells expressing NCLX constructs in a 10-cm dish, and were treated with 800 µl of a hypotonic shock buffer (5 mM sucrose, 5 mM HEPES, 1 mM EGTA, pH 7.2 KOH) on ice for 10 min to produce mitoplasts. After adding 200 µl of a high-salt buffer (750 mM KCl, 100 mM HEPES, 2.5 mM EGTA, pH 7.2 KOH), mitoplasts were pelleted at 17,000 g for 10 min. The isolated mitoplasts were resuspended in tris-buffered saline (TBS), with a small portion used for protein quantification using the bicinchoninic acid assay (Thermo Fisher Scientific, 23227). The samples were treated with 3 µg of proteinase K (Sigma, 70663) or TEV protease (produced in house) per 10 µg of mitoplast proteins. Protease digestion was performed at room temperature for 5 min, and was terminated by adding 0.5 mM of phenylmethylsulphonyl fluoride (Sigma, P7626) and 2 µg ml^–1 of leupeptin (Sigma, LEU-RO) and pepstatin (Sigma, PEPS-RO) at room temperature for 5 min. The samples were then denatured with an SDS loading buffer for subsequent SDS-PAGE.

To perform Western blot, 10 μg of mitochondrial or mitoplast proteins or membranes from 20 oocytes were separated using SDS-PAGE and transferred to low-fluorescence PVDF membranes (LI-COR), which were blocked in the LI-COR Intercept blocking buffer. The membranes were then incubated with primary antibodies diluted in TBST (that is, TBS + 0.075% Tween-20) at 4 °C overnight, followed by 1 h incubation with fluorescent secondary antibodies, diluted in TBST, at room temperature. Western blot signals were acquired using a LI-COR Odyssey CLx imager, and quantified using the ImageStudio software (v.5.0). Antibodies and dilutions: anti-1D4 (homemade, 100 ng ml^–1); anti-Tim23 (Santa Cruz, sc-514463, 1:1,000); anti-MCU (Cell Signalling, D2Z3B, 1:10,000); anti-actin (Santa Cruz, sc-68979, 1:2,000); anti-COX2 (Abcam, ab110258, 1:10,000); anti-Histone H3 (Millipore, 05-928, 1:10,000); IRDye 680RD goat anti-mouse secondary antibody (LI-COR, 925-68070, 1:15,000); and IRDye 800CW goat anti-rabbit secondary antibody (LI-COR, 926-32211, 1:10,000).

Mitochondrial and oocyte Ca²⁺ transport assays

Each mitochondrial Ca²⁺ flux experiment was performed using 2 × 10⁷ of HEK, HeLa, CHO or HCT116 cells, or 2.8 × 10⁷ of Sf9 cells. Cells were suspended in 10 ml of a wash buffer (120 mM KCl, 25 mM HEPES, 2 mM K₂HPO₄, 1 mM MgCl₂, 50 μM EGTA, pH 7.2 KOH), pelleted at 1,500 g for 3 min, and then resuspended in 2.2 ml of a recording buffer (120 mM KCl, 25 mM HEPES, 2 mM K₂HPO₄, 5 mM succinate, 1 mM MgCl₂, pH 7.2 KOH); 2 ml of the cell suspension was then transferred into a stirred quartz cuvette in a Hitachi F-7100 spectrophotometer (excitation = 508 nm; excitation-slit = 2.5 nm; emission = 531 nm; emission-slit = 5 nm; sampling rate = 2 Hz), with 250 nM of CG5N (Thermo Fisher Scientific, C3737) used to report extra-mitochondrial [Ca²⁺], and 30 μM of digitonin (Sigma, D141) used to permeabilize cells. Inhibitors used in these assays include Ru360 (synthesized in house) and CGP-37157 (Cayman, 1561110).

For oocyte Ca²⁺ uptake assays, stage V–VI oocytes were injected with 50 ng of NCLX mRNA, and incubated in an ND96 solution (96 mM NaCl, 2 mM KCl, 2 mM CaCl₂, 0.5 mM MgCl₂, 5 mM HEPES, pH 7.4 NaOH) for 3 to 4 days. To measure ⁴⁵Ca²⁺ uptake (Fig. 5d–g), oocytes were washed three times in an oocyte recording buffer (ORB, 100 mM NMG, 5 mM HEPES, pH 7.4 HCl), and finally placed in a density of ten oocytes per 400 µl ORB. To begin the assay, 400 µl of the oocyte-containing solution was mixed well with 100 µl of ORB containing 5 µCi of ⁴⁵Ca²⁺ (Perkin Elmer, NEZ013001MC). The assay was performed at room temperature. At desired time points, the reaction was terminated by transferring ten oocytes into 30 ml of ORB. After washing the oocytes two more times in fresh 30 ml ORB, each oocyte was lysed individually via pipetting and vigorous shaking in 10 ml of a scintillation cocktail for radioactivity measurements using a Beckman LS6500 scintillation counter. To obtain a data point, we first measured ⁴⁵Ca²⁺ in ten individual oocytes, and obtained the median reading. We then excluded oocytes with readings >5-fold higher than the median, which probably reflect sick oocytes with compromised membranes, as well as those with readings <20% of the median, which probably reflect oocytes with failed mRNA injection. The remaining readings, usually from 5–8 oocytes, were averaged and presented as an independent measurement. The results were discarded if there were fewer than five useful readings from ten oocytes.

To test H⁺-coupled ⁴⁵Ca²⁺ transport (Fig. 5h), oocytes one day after mRNA injection were incubated in a counter flux buffer containing 96 mM NaCl, 2 mM KCl, 0.5 mM CaCl₂, 2 mM MgCl₂, 2.5 mM HEPES, 5 µCi ml^{–1 45}Ca²⁺, pH 7.4 NaOH at a density of ten oocytes per 1 ml. After two days of equilibration, oocytes were transferred into microcentrifuge tubes, with the external solution reduced to 20 oocytes per 800 µl. To begin the reaction, 800 µl of the oocyte-containing solution was mixed well with 200 µl of counter flux buffer that contains high concentrations of pH buffers (200 mM MOPS for pH 6.8, 200 mM HEPES for pH 7.4, or 200 mM Tris for pH 8.2) to adjust the pH, BAPTA to reduce free [Ca²⁺] to 250 nM, and 5.2 µCi ml^{–1 45}Ca²⁺. Reaction termination and data analyses were performed as above, but with 20 oocytes used for each data point (experiments with fewer than ten useful oocyte readings were discarded).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.