Structural insights into mitochondrial calcium uniporter regulation by divalent cations

Summary

Calcium (Ca²⁺) flux into the matrix is tightly controlled by the mitochondrial Ca²⁺ uniporter (MCU) due to vital roles in cell death and bioenergetics. However, the precise atomic mechanisms of MCU regulation remain unclear. Here, we solved the crystal structure of the N-terminal matrix domain of human MCU, revealing a β-grasp-like fold with a cluster of negatively charged residues that interacts with divalent cations. Binding of Ca²⁺ or Mg²⁺ destabilizes and shifts the self-association equilibrium of the domain toward monomer. Mutational disruption of the acidic face weakens oligomerization of the isolated matrix domain and full-length human protein similar to cation binding and markedly decreases MCU activity. Moreover, mitochondrial Mg²⁺ loading or blockade of mitochondrial Ca²⁺ extrusion suppresses MCU Ca²⁺ uptake rates. Collectively, our data reveal that the β-grasp-like matrix region harbors an MCU regulating acidic patch that inhibits human MCU activity in response to Mg²⁺ and Ca²⁺ binding.

Graphical abstract

Mitochondria are organelles that synthesize adenosine triphosphate (ATP), establishing a large electrical gradient across the inner mitochondrial membrane (IMM) (i.e. ΔΨ_m~ −180 mV) (Marchi and Pinton, 2014) that is a major driving force for cation uptake (Carafoli, 2003). However, calcium (Ca²⁺) uptake must be precisely controlled by the mitochondrial Ca²⁺ uniporter (MCU) due to roles in vital signaling processes such as bioenergetics, cell death and shaping cytosolic Ca²⁺ transients. Under resting cytosolic Ca²⁺ levels, MCU is inactive; upon agonist stimulation of plasma membrane (PM) receptors that increase cytosolic Ca²⁺ levels, rapid and transient increases in matrix Ca²⁺ levels occur via MCU.

MCU (CCDC109A) was originally characterized as a Ca²⁺ selective ion channel which resides on the IMM (Kirichok et al., 2004) and an ion channel after self-association and arrangement of multiple TM domains into an ion pore (Baughman et al., 2011; De Stefani et al., 2011). Numerous binding partners that regulate MCU activity have been identified, suggesting that MCU functions as a heteromeric complex (Kamer and Mootha, 2015). Mitochondrial calcium uptake (MICU)1 and MICU2 are EF-hand containing proteins that serve as Ca²⁺-dependent gatekeepers of MCU activity (Csordas et al., 2013; Hoffman et al., 2013; Mallilankaraman et al., 2012b; Plovanich et al., 2013). Essential MCU regulator (EMRE) is a single IMM spanning protein of ~10 kDa; knockdown of EMRE in live cells attenuates mitochondrial Ca²⁺ uptake which cannot be rescued by MCU over-expression (Sancak et al., 2013). MCUb shares ~50% sequence similarity with MCU, but has no ability to constitute a Ca²⁺-permeable channel despite the presence of TM domains; rather, MCUb functions in a dominant-negative manner, inhibiting MCU activity (Raffaello et al., 2013). MCUR1 is a 40 kDa protein containing two putative TM domains; moreover, MCUR1 interacts with MCU, and over-expression greatly enhances while knockdown of MCUR1 decreases matrix Ca²⁺ levels in HeLa cells (Mallilankaraman et al., 2012a). Recently, we have revealed that MCUR1 is a vital scaffold factor that stabilizes the MCU complex (Tomar et al., 2016). SLC25A23, a magnesium (Mg²⁺)-ATP and PO₄³⁻ transporter protein found on the IMM also interacts with MCU and MICU1, enhancing mitochondrial Ca²⁺ uptake (Hoffman et al., 2014).

Despite the realization that MCU functions as an oligomer in complex with a suite of proteins, an understanding of the molecular regulatory mechanisms remains outstanding. Many Ca²⁺ channel proteins exhibit Ca²⁺-dependent feedback mechanisms including inositol 1,4,5-trisphosphate receptors (Bezprozvanny et al., 1991), ryanodine receptors (Meissner et al., 1986) and Ca²⁺ release activated Ca²⁺ channels (Hoth and Penner, 1993). Further, while a crystal structure of the N-terminal domain of MCU has been elucidated (Lee et al., 2015), it is unclear how divalent cations regulate MCU. Here, we report the atomic resolution structure of a conserved MCU matrix domain, which adopts a β-grasp-like fold containing an MCU regulating acidic patch (MRAP) that binds Mg²⁺ and Ca²⁺. Interaction of these divalent cations with MRAP destabilizes the protein and shifts the self-association equilibrium toward monomer. Disruption of MRAP by variation of a single Asp destabilizes the domain coupled with monomerization similar to Ca²⁺ or Mg²⁺ binding. Mutational disruption of MRAP in full-length human MCU perturbs oligomerization, markedly decreases agonist-induced mitochondrial Ca²⁺ uptake and attenuates basal mitochondrial Ca²⁺ levels. Importantly, we show that sustained elevation of matrix Ca²⁺ or bathing with extramitochondrial Mg²⁺ suppresses mitochondrial Ca²⁺ uptake. Together, our data establish that MRAP within the β-grasp-like matrix domain acts as an important regulatory component of human MCU that is Ca²⁺ and Mg²⁺ dependent.

Results

MCU contains a conserved β-grasp-like matrix domain

Human MCU is a ~40 kDa protein made up of 351 amino acids (NCBI, Q8NE86) (Fig. 1A). The mitochondrial signal peptide is localized in the first 49 N-terminal residues. MCU contains two predicted TM domains (TM1, residues 234-256; TM2, residues 266-283) close to the C-terminal region. Two putative coiled-coil (CC) domains can be identified prior to TM1 and after TM2, adjacent to the C-terminus. Both the N- and C-termini, which represents the bulk of the soluble regions and the overall majority of the protein, are localized within the matrix (Kamer and Mootha, 2015). Multiple sequence alignment of MCU proteins show a matrix-oriented region outside the CC and TM domains conserved among roundworm, fruit fly, fish, human and mouse sequences between residues 75 and 191 (Fig. S1). Thus, we engineered an MCU construct encompassing residues 72 to 189 (i.e. MCU_72-189). Far-UV circular dichroism (CD) spectroscopy showed that MCU_72-189 adopts an α-helix/β-strand mixture (Fig. S2A) and cooperatively unfolds with a high midpoint of temperature denaturation (T_m) > 60 °C with negligible ionic strength dependence (Fig. S2B), indicating a stably folded domain.

Fig. 1. Domain architecture and structural features of MCU.

(See also Fig. S1, Fig. S3, Fig. S7 and Table S1).

(A) Domain architecture of human MCU (NCBI, Q8NE86). Residue ranges and the prevailing topology are indicated at top. The relative location of the signal peptide (SP, blue), transmembrane domains (TM1, TM2, cyan), CC domains (CC1, CC2, yellow) and MCU_72-189 (green) are indicated.

(B) Backbone structure of MCU_72-189. The β-strands (β1-β6, arrows) and loop regions (L1-L6) are colored green while the α-helices (α1, α2) are pink. L2 is bounded by a dashed box.

(C) Zoomed view of L2. The R93 and E95 side chains are shown as red and blue sticks, respectively. The C97 and S92 residues are shown as yellow sticks.

(D) Electrostatic surface potential of MCU_72-189. The gradient shown is from −3 (acidic, red) to +3 (basic, blue) kT/e. Locations of acidic and basic residues are indicated.

(E) Location of the Mg²⁺ atom (yellow sphere) relative to the backbone structure and acidic residues (red sticks). The dashed box bounds the acidic patch region.

(F) Zoomed view of the acidic patch shown in (E) and centrally located Mg²⁺ atom (yellow sphere).

(G) Mg²⁺ atom coordination geometry. The 2Fo-Fc (2.0 σ) electron density map (grey mesh), protein chain (sticks), water molecules (orange spheres) and Mg²⁺ atom (yellow sphere) are shown. The Fo-Fc (9.0 σ) omit map (magenta mesh) shows unaccounted for density in the absence of the Mg²⁺ atom. Distances between the Asp (red sticks) Cγ atoms are indicated. The coordination geometry of Mg²⁺ is shown with distances to oxygen atom indicated.

(H) Electron density map of MCU_72-189 solved in the presence of LiSO₄. The 2Fo-Fc (2.0 σ) map (grey mesh) of the same region shown in (G) does not exhibit evidence for a cation, and the Fo-Fc (2.5 σ) map (magenta mesh) shows no unaccounted for density.

We crystallized an I141M/L146M MCU_72-189 double mutant to permit phasing by selenomethionine (Se-Met) incorporation. The human native MCU_72-189 protein was subsequently crystalized and solved by molecular replacement using the Se-Met structure. Native MCU_72-189 structures were crystalized under three different conditions (Table S1). The native structure (1.6 Å) revealed two central α-helices approximately perpendicular to one another, sandwiched between two 3-stranded β-sheets (Fig. 1B). The first β-sheet is formed with β1 (i.e. residues 76-80), β2 (i.e. residues 83-88) and β3 (i.e. residues 97-100) arranged in an antiparallel manner. One short loop (i.e. L1; residues 81-82) and one long loop (i.e. L2; residues 89-96) link the β-strands. The L2 conformation is stabilized by a salt bridge between the NH1 atom of R93 and the Oε1 atom of E95 (Fig. 1C). This salt bridge is in close proximity to a single Cys (i.e. C97) as well as a known phosphorylation site (i.e. S92) and may serve to stabilize the loop for post-translational modifications of these residues (Joiner et al., 2012; Lee et al., 2015). A centrally located α1 helix (i.e. residues 108-118) is connected to the N-terminal β-sheet through L3 (i.e. residues 101-107). The second β-sheet is located C-terminal to the central α1 and is comprised of β4 (i.e. residues 125-128), β5 (i.e. residues 149-153) and β6 (i.e. residues 156-160) strands arranged in an antiparallel orientation. The β4 and β5 strands are separated by a short α-helix (i.e. α2; residues 141-146) that is perpendicular to α1 and also centrally positioned relative to the two β-sheets. The α2 is linked to the β4 and β5 strands through one long 12-residue loop (i.e. L5; residues 129-140) and one short 2-residue loop (i.e. L6; residues 147-148). The C-terminal region of the MCU_72-189 construct (i.e. residues 168-189) does not show any appreciable electron density, suggesting the C-terminal tail is dynamic under these conditions.

Overall, the tertiary structure buries 33 residues that are at least 85% solvent inaccessible. The majority of these side chains are directed toward the inner core of the structure; however, V85 and V152, which are centrally located on each of the β-sheets, respectively, are positioned away from the inner core (Fig. S3A) and may play roles in mediating protein-protein interactions and/or stabilizing the β-sheets at a more local level. The electrostatic surface of the protein reveals a prominent negative patch near the C-terminal face with D131, D142, D147, D148 and D166 contributing the bulk of the acidic charge in this region (Fig. 1D). A second negative patch closer to the N-terminal face of the protein is primarily made up of E117, E118, D119, D124 and D155. Two basic patches on the surface of the protein are made up of R89, R93, R94 and R96 as well as R113, R124 and R134, respectively (Fig. 1D). The large degree of surface charge polarity exhibited by MCU_72-189 may facilitate a complementarity for higher order assembly of the domain.

A search for similar folds showed that the closest known structural homologue is ubiquitin-like protein 5 (ULP5) which adopts a β-grasp fold. Sequence identity between the crystallized region of ULP5 and MCU_72-189 is ~5%, yet the backbone root mean square deviation (RMSD) between the two proteins is 3.0 Å (Fig. S3B). While MCU_72-189 conforms to the core β-grasp structure consisting of a central α-helix that packs diagonally against a β-sheet with parallel and central N- and C-terminal β-strands (Burroughs et al., 2012), it also contains distinct divergences. First, MCU_72-189 contains a discontinuous sheet split into two halves due to lack of hydrogen (h)-bonding between the central β1 and β6 strands; second, a third β-strand is inserted in the N-terminal sheet (i.e. β3) prior to the central helix changing the orientation of α1 by ~90°; third, a short α-helix is inserted between β4 and β5, capping one end of the structure. This observed fold MCU_72-189 is not likely due to construct design since omission of the first two strands (i.e. leaving a 4-stranded core with the N- and C-terminal strands parallel) results in a destabilized conformation (Fig. S2C and S2D).

The β-grasp domain of MCU interacts with Mg²⁺ and Ca²⁺

MCU_72-189 crystallized in the presence of MgCl₂ revealing a Mg²⁺ atom near the negatively charged D131, D147 and D148 side chains that are primary contributors to the major acidic surface patch (Fig. 1E). Side chain OΔ1 and OΔ2 atoms of D147 are within 3.5 and 2.1 Å, respectively, of the Mg²⁺ ion which is centrally located in the acidic patch, while D131 and D148 side chains are positioned further away (i.e. minimum OΔ distance ~6.6 Å) (Fig. 1F). Mg²⁺ is coordinated by the D147 OΔ2 atom and 5 water molecules arranged in octahedral geometry (Fig. 1G). The arrangement of the 5 ligand water molecules and the D147 carboxylate around the Mg²⁺ is visible in the 2Fo-Fc electron density map contoured at 2.0σ; further, an Fo-Fc omit map contoured at 9.0σ clearly shows the unaccounted for density when the Mg²⁺ atom is removed (Fig. 1G). Within the crystal, two of the water molecules that coordinate the Mg²⁺ in the first shell are bridged by the D123 side chain of a symmetry mate in second shell coordination.

MCU_72-189 also crystallized with either LiSO₄ or Ca(CH₃COO)₂ (Table S1). In the presence of LiSO₄ (1.5 Å), the Fo-Fc omit map contoured at 2.5σ shows no unaccounted for electron density in the acidic patch, demonstrating specificity for the Mg²⁺ atom found in the presence of MgCl₂ (Fig. 1H). The structure crystallized with Ca(CH₃COO)₂ showed considerably lower resolution than the MgCl₂ or LiSO₄ structures (i.e. 2.7 Å versus 1.5-1.6 Å), possibly contributing to an inability to reliably locate a Ca²⁺ atom. A comparison of the Cα atom locations between the MgCl₂, LiSO₄ and Ca(CH₃COO)₂ structures reveals highly similar backbone conformations with only minor deviations at the N- and C-terminal ends; however, an all-atom comparison reveals larger variations for specific side-chains in the loop and structured regions (Fig. S3C and S3D). The precise significance of these differences requires further study.

Given the discovery of the Mg²⁺ ion bound to MCU_72-189 in the crystal state, we sought to characterize the cation binding in solution. We used changes in intrinsic fluorescence of MCU_72-189 as a readout of structural changes in response to MgCl₂ or CaCl₂ titration. In divalent cation-free buffer and using an excitation wavelength (λ_ex) of 280 nm, MCU_72-189 shows an intrinsic fluorescence emission maximum at a wavelength (λ_em) of ~305 nm, consistent with Tyr fluorescence and the absence of Trp in the domain; moreover, increasing concentrations of CaCl₂ or MgCl₂ systematically decreased the intrinsic fluorescence (Fig. S2E and S2F). Guanidine-induced denaturation was also found to decrease the intrinsic fluorescence of MCU_72-189 suggesting that the fluorescence change observed in our titrations may reflect a loss in structure associated with the cation (Fig. S2G). Plots of the changes in fluorescence intensity at the wavelength showing maximum fluorescence change (i.e. λ_ex=280 nm; λ_em=302 nm) versus CaCl₂ or MgCl₂ revealed saturable binding. Fitting the binding curves to a one site binding model that takes into account protein concentration indicated equilibrium dissociation constants for Ca²⁺ (K_d,Ca2+) and Mg²⁺ (K_d,Mg2+) of ~2.3 and ~1.7 mM, respectively (Fig. 2A and 2B).

Fig. 2. Cation binding, oligomerization and stability of MCU_72-189.

(See also Fig. S2, Fig. S4 and Tables S2-S4).

(A) and (B) Changes in intrinsic fluorescence (fl.) as a function of MgCl₂ (A) and CaCl₂ (B) concentration at 25 °C. Data were fitted to a one-site binding scheme (solid lines).

(C) and (D) MCU_72-189 secondary structure and stability sensitivity to MgCl₂ (C) and CaCl₂ (D). The (C) and (D) insets show far-UV-CD spectra in the absence (black line) and presence of 40 mM MgCl₂ (red line) or CaCl₂ (orange line) at 4 °C. Thermal stability in the absence (black circles) and presence of 40 mM MgCl₂ (red circles) or CaCl₂ (orange circles).

(E) SEC-MALS data in the absence (black) or presence of 40 mM MgCl₂ (red) or CaCl₂ (orange). The elution peaks are representative (lines) and the molecular weights are means ± SE (circles) of n=3 experiments at 4 °C. *P<0.001. Injections of BSA are shown (dashed lines) with a negligible change in molecular weight (circles) in the absence and presence of 40 mM CaCl₂.

(F) Equilibrium ultracentrifugation assessment of MCU_72-189. Data (open circles) are shown for a 1.4 mg mL⁻¹ sample acquired at 25,000 rpm and 25 °C. The red line shows the non-linearity in the data indicative of multiple species in solution.

(G) Thermal stability of native (black circles) and D147R MCU_72-189 (blue circles) in the presence of 40 mM MgCl₂ (red circles) or 40 mM CaCl₂ (orange circles).

(H) Thermal stability of native (black circles) and D131R MCU_72-189 (blue circles) in the presence of 40 mM MgCl₂ (red circles) or 40 mM CaCl₂ (orange circles). All data were acquired in 20 mM HEPES, 150 mM KCl, 2 mM DTT, pH 7.5 at 0.5 mg mL⁻¹ (A), (B), (C), (D), (G), (H) or ~2.0 mg mL⁻¹ (E). Mean ± SE, (n=3 independent experiments).

To test whether Mg²⁺ or Ca²⁺ binding to MCU_72-189 influences secondary structure or stability in solution, we first assessed the far-UV-CD spectrum of the protein as a function of increasing Mg²⁺ or Ca²⁺ levels. Increasing concentrations of MgCl₂ or CaCl₂ nominally changed the far-UV CD ellipticity suggesting that cation interaction with MCU_72-189 does not appreciably alter secondary structure (Fig. 2C and 2D, insets). Effects on domain stability were monitored by change in far-UV-CD signal at 222 nm as a function of temperature in the presence and absence of MgCl₂ or CaCl₂. In the absence of cations, MCU_72-189 had an apparent midpoint of T_m of ~66 °C. Addition of 40 mM MgCl₂ decreased the T_m of MCU_72-189 to ~62 °C (Fig. 2C). Remarkably, addition of 40 mM CaCl₂ destabilized the protein to a greater extent than MgCl₂ with a T_m value of ~59 °C (Fig. 2D).

Taken together, results from crystal structures, solution far-UV-CD and intrinsic fluorescence suggest that MCU_72-189 binds Mg²⁺ and Ca²⁺ with ~mM affinity on an acidic region of the protein surface resulting in destabilization of the domain.

Ca²⁺ and Mg²⁺ promote MCU_72-189 monomerization

With relatively small secondary structural changes associated with Mg²⁺ or Ca²⁺ binding, we sought to determine if other factors might contribute to the modest destabilization caused by cation binding. Size exclusion chromatography (SEC) was used to assess the quaternary structure of MCU_72-189. The protein eluted as a single major peak on a Superdex 200 10/300 column with an apparent molecular weight of ~24.0 kDa, equal to 1.7× the theoretical 13.9 kDa monomer mass suggesting a monomer-dimer equilibrium in fast exchange (Table S2). Consistent with this notion, we found that lower protein concentration shifted the apparent molecular weight toward monomer (Table S2). We next performed SEC with in-line multi-angle light scattering (MALS) to assess the effects of Mg²⁺ and Ca²⁺ ions on the self-association of MCU_72-189. The SEC-MALS-determined molecular weight of MCU_72-189 was 24.5 ± 0.2 kDa, consistent with the fast exchange equilibrium implied by the SEC experiments; moreover, the addition of 40 mM CaCl₂ or MgCl₂ shifted apparent molecular weight to 17.4 ± 0.1 or 15.3 ± 0.1 kDa, respectively (Fig. 2E). Thus, both Mg²⁺ and Ca²⁺ cations shift the self-association equilibrium of MCU_72-189 toward monomer.

We performed analytical ultracentrifugation (AUC) to confirm the tendency for MCU_72-189 dimerization and to quantify the strength of the self-association in the presence and absence of CaCl . A plot of ln(absorbance) versus radius² showed a non-linearity indicative of multiple species in solution (Fig. 2F). Global fitting of the equilibrium data to a single ideal species revealed molecular weights of ~22.3, 21.5 and 18.5 kDa for samples prepared at 1.2, 0.8 and 0.4 mg mL⁻¹, respectively (Fig. S4A, S4B and S4C; Table S3); moreover, these fitted molecular weights corresponded to 1.6, 1.5 and 1.3× the monomer mass, confirming the concentration dependence and fast exchange equilibrium as observed with SEC-MALS. The equilibrium dissociation constants for the monomer-dimer equilibrium (K_d,dimer) were also extracted using global fits at the three protein concentrations, and the K_d,dimer values were 58.6, 77.9 and 126 μM for the 1.2, 0.8 and 0.4 mg mL⁻¹ datasets, respectively. Consistent with SEC-MALS, CaCl₂ decreased the fitted molecular weights to ~20.0, 18.8 and 16.2 kDa, corresponding to 1.4, 1.4 and 1.2× the monomer mass for the 1.2, 0.8 and 0.4 mg/mL datasets, respectively (Fig. S4D, S4E and S4F; Table S3). Similarly, the globally fitted K_d,dimer values were 139, 197 and 369 μM, approximately ~2-3× higher (i.e. weaker affinity) than the K_d,dimer measured in the absence of the divalent cation.

Taken together, SEC-MALS and AUC data demonstrate that MCU_72-189 exists in a self-association equilibrium with sub-mM affinity undergoing fast exchange. Importantly, CaCl₂ weakens MCU_72-189 multimerization by ~2-3-fold suggesting a potential feedback mechanism of regulation.

Acidic patch mutations destabilize and monomerize MCU_72-189 similar to metal binding

To validate the notion that Ca²⁺ and Mg²⁺-induced destabilization is caused by binding to the acidic patch of MCU_72-189, we introduced D131R or D147R substitutions and assessed the effects of CaCl₂ and MgCl₂ supplementation on thermal stability. D147 is adjacent to the C-terminal end of the α2 helix and D131 is found within ~9.0 Å of D147 (i.e. Cγ distance) on the loop immediately following the β4 strand (Fig. 1E). D147R and D131R mutant proteins exhibited T_m values of ~61 °C and ~56 °C, respectively, compared to ~66 °C for the native protein, indicating that introduction of a positively charged residue in the acidic patch markedly destabilizes the domain, despite the loop localization of these residues (Fig. 2G and 2H). Addition of 40 mM MgCl₂ or CaCl₂ decreased the T_m of the D147R protein by ~2 °C, in contrast to the ~4 °C and ~7 °C destabilization observed for native protein by MgCl₂ and CaCl₂, respectively (Fig. 2G and Table S4). A suppressed destabilization compared to the native protein was also observed for the D131R sample in the presence of 40 mM MgCl₂ or CaCl₂ (i.e. ~0 °C and 3 °C, respectively) (Fig. 2H and Table S4).

To confirm that the mutation-induced destabilization of MCU_72-189 is associated with monomerization as observed with Mg²⁺ and Ca²⁺ binding, we performed SEC-MALS on the D147R protein. The SEC-MALS-determined molecular weight of the D147R protein with and without 40 mM CaCl₂ was 15.6 ± 0.1 and 15.7 ± 0.2 kDa, respectively (Fig. S2H). Thus, our MCU_72-189 mutations in the acidic patch caused a domain destabilization (i.e. ΔT_m between ~−5 and −10 °C) similar to the loss in stability caused by Ca²⁺ and Mg²⁺ binding to native protein (i.e. ΔT_m between ~−4 and −7 °C); further, these mutations suppressed the effects of cation binding on stability and promoted monomerization of the domain. These data suggest that the cation sensitivity may be an important regulatory feature of this domain.

MCU is functionally sensitive to acidic patch mutations

To test the idea that the acidic patch of MCU_72-189 plays a regulatory role in MCU activity we used genetically encoded Ca²⁺ sensors targeted to the mitochondrial matrix or the cytosol (i.e. mito-R-GECO1 or R-GECO1, respectively). We overexpressed full-length wild-type (WT) as well as D131A, D131R, D147A and D147R mutant proteins in HeLa cells and assessed changes in Ca²⁺ levels in response to the GPCR agonist, histamine. Reconstitution of human WT MCU and mutants did not alter the protein expression and mitochondrial localization (Fig. S5). HeLa cells overexpressing WT MCU showed a significantly increased maximal mitochondrial Ca²⁺ uptake compared to empty vector transfected cells (Fig. 3A and 3B). Each of the D131 and D147 mutants showed significantly reduced mitochondrial Ca²⁺ uptake compared to endogenous MCU and WT MCU overexpression, revealing a dominant negative effect. Concomitantly, HeLa cells expressing WT MCU also showed a significant decrease in maximal cytosolic Ca²⁺ levels after histamine stimulation compared to control cells owing to increased MCU activity, while ectopic constitution of D131A, D131R, D147A and D147R did not significantly alter the maximal cytosolic Ca²⁺ levels (Fig. 3C and 3D). To verify that the observed functional differences for D131 and D147 mutants were not due to altered ΔΨ_m we loaded the WT and mutant MCU-overexpressing HeLa cells with the fluorescent ΔΨ_m indicator, tetramethylrhodamine methyl ester (TMRM). Under unstimulated conditions, TMRM fluorescence was measured before and after addition of the ΔΨ_m uncoupler, carbonyl cyanide m-chlorophenyl hydrazine (CCCP). The TMRM fluorescence change was similar for WT and mutant MCU expressing cells suggesting intact ΔΨ_m (Fig. S6A and S6B).

Fig. 3. Live cell MCU functional activity.

(See also Fig. S5 and Fig. S6).

HeLa cells were transiently transfected with mito-R-GECO1 or R-GECO1. Cells were stimulated with 100 μM histamine after one minute baseline recording (arrowheads).

(A) Representative mean traces of mito-R-GECO1 fluorescence (f.a.u.) in HeLa cells transfected with empty vector (grey), MCU-WT-GFP (green), MCU D131A-GFP (magenta), MCU D131RGFP (blue), MCU D147A-GFP (yellow), and MCU D147R-GFP (red).

(B) Quantification of peak mito-R-GECO1 fluorescence (F_max/F₀) (mean ± S.E.). *P <0.05; **P <0.01; ***P <0.001.

(C) Representative mean traces of cytosolic R-GECO1 fluorescence (f.a.u) in HeLa cells transfected with empty vector (grey), MCU-WT-GFP (green), MCU D131A-GFP (magenta), MCU D131R-GFP (blue), MCU D147A-GFP (yellow), and MCU D147R-GFP (red).

(D) Quantification of cytosolic Ca²⁺ elevation as a function of R-GECO1 fluorescence (F_max/F₀) (mean ± SE). *P <0.05.

(E) HEK293T cells stably expressing pBSD (empty vector), MCU-WT, MCU D131A, MCU D147A or MCU D147R were permeabilized with digitonin in intracellular-like medium containing Fura2FF and thapsigargin (2 μM). Ru360 (1 μM), CGP37157 (1 μM), and CCCP (2 μM) were added prior to thapsigargin addition (arrowheads). Traces are means of n=3 independent experiments.

(F) Quantification of mitochondrial Ca²⁺ release ([Ca²⁺]_m) after CCCP addition for each stable cell line. Mean ± SE, (n=3 independent experiments). *P <0.05.

To assess mitochondrial Ca²⁺ uptake in a permeabilized system, we simultaneously measured ΔΨ_m and extramitochondrial Ca²⁺ clearance. HEK293T cells were suspended with JC-1 and Fura2FF for measuring ΔΨ_m and extramitochondrial Ca²⁺, respectively, in the presence of the SERCA pump blocker, thapsigargin to exclude ER Ca²⁺ uptake. As expected, the WT MCU expressing cells showed a rapid mitochondrial Ca²⁺ uptake after 20 μM external Ca²⁺ addition which is concomitant with loss of ΔΨ_m (Fig. S6C-S6F). Strikingly, cells harboring the MCU D131R mutation displayed a marked decrease in mitochondrial Ca²⁺ uptake with nominal ΔΨ_m change, fully consistent with the reduced MCU activity (Fig. S6C-S6F). These data reinforce the notion that acidic patch mutations of MCU decrease mitochondrial Ca²⁺ uptake.

To determine if the MCU acidic patch plays a role in basal mitochondrial Ca²⁺ homeostasis, we generated HEK293T cell lines stably expressing pBSD (i.e. Empty vector), WT, D131A, D147A or D147R MCU. Cells were bathed in Ca²⁺-free medium containing digitonin, Fura2FF and thapsigargin. To prevent mitochondrial MCU and Na⁺/Ca²⁺ exchanger (NCLX)-dependent Ca²⁺ flux, Ru360 and CGP37157 were added prior to permeabilization. Ca²⁺ released from the mitochondrial matrix was quantified by Fura2FF fluorescence after CCCP addition. HEK293T cells expressing WT MCU showed enhanced matrix Ca²⁺ release compared to empty vector expressing cells; however, each of the D131A, D147A and D147R expressing cells demonstrated similar or lower (i.e. D131A) mitochondrial Ca²⁺ release compared to control cells (Fig. 3E and 3F). The D131R substitution caused the greatest destabilization of MCU_72-189 in vitro (i.e. ΔT_m ~−10 °C) (Table S4), which is supported by maximal inhibition of agonist-induced mitochondrial Ca²⁺ uptake (Fig. 3B). Thus, depletion of basal mitochondrial Ca²⁺ in D131A-expressing cells may reflect greater destabilization of D131 mutants when compared to the D147 mutants and is correlated with an observed lethality for stably D131R-expressing cells.

Collectively, our live cell results suggest that D131 and D147 residues contribute to the regulatory region on the matrix domain (i.e. MRAP) that, if perturbed by mutation, inhibits MCU activity.

MCU is functionally sensitive to Ca²⁺ and Mg²⁺

Having observed an inhibition of MCU activity by mutations which destabilize MCU_72-189, we aimed to determine whether elevation of mitochondrial matrix Ca²⁺ can similarly effect MCU activity. Permeabilized HEK293T cells were treated with the NCLX blocker CGP37157 to attenuate Ca²⁺ efflux from the matrix and exposed to train of external Ca²⁺ pulses. The ΔΨ_m remained intact after four pulses of Ca²⁺ addition in control and CGP treated cells (Fig. 4A and 4B). Remarkably, the mitochondrial Ca²⁺ uptake rate was significantly reduced in cells treated with CGP when compared to control (Fig. 4C, 4D and 4E), which is consistent with Ca²⁺-dependent MCU inhibition mechanism.

Fig. 4. Matrix Ca²⁺-dependent MCU inactivation.

(See also Fig. S6).

Digitonin-permeabilized HEK293T cells bathed in intracellular-like solution containing thapsigargin were loaded with the Ca²⁺ indicator Fura2FF (1 μM) and the ΔΨ_m indicator JC-1, to which pulses of 10 μM extracellular Ca²⁺ were delivered before the addition of the mitochondrial uncoupler CCCP (2 μM) (arrowheads). Simultaneous measurements of ΔΨ_m and changes in bath Ca²⁺ ([Ca²⁺]_out) were monitored in the presence and absence of the NCLX blocker, CGP37157 (10 μM).

(A) Representative traces of JC-1 report ΔΨ_m from n=3 independent experiments.

(B) Quantification of the JC-1 ΔΨ_m before addition of CCCP. Mean ± SE, (n=3 independent experiments). ns, not significant.

(C) Representative traces of [Ca²⁺]_out from n=3 independent experiments.

(D) Magnified view of the kinetic changes in [Ca²⁺]_out representing mitochondrial Ca²⁺ uptake.

(E) Quantification of the [Ca²⁺]_m uptake rate taken from the reduction in bath Ca²⁺ after 3 pulses of 10 μM [Ca²⁺]_out addition. Mean ± SE. **P <0.01.

To test whether Ca²⁺ uptake via MCU can be regulated by Mg²⁺, we adapted our permeabilized cell system. HEK293T permeabilized cells were bathed with varying concentrations of MgCl₂ (i.e. 0, 1, 5 or 10 mM). A Mg²⁺ concentration-dependent decrease in the rate of mitochondrial Ca²⁺ uptake was observed after addition of 20 μM external Ca²⁺ (Fig. 5A). Importantly, no differences in ΔΨ_m were observed while simultaneously monitoring bath Ca²⁺ clearance (Fig. 5B and 5E). Significantly suppressed rates of mitochondrial Ca²⁺ uptake were observed in the presence of 5 mM and 10 mM extramitochondrial MgCl₂ compared to cells bathed in medium nominally free of MgCl₂ (Fig. 5C and 5D); moreover, even 1 mM Mg²⁺ significantly suppresses the slow MCU Ca²⁺ uptake rate assessed over a longer time frame (Fig. S6G amd S6H). As expected, after CCCP addition, reduced release of matrix Ca²⁺ was observed in the presence of 5 and 10 mM Mg²⁺ (Fig. 5F and 5G). These functional data show that Mg²⁺ reduces mitochondrial Ca²⁺ uptake in a concentration-dependent manner without altering ΔΨ_m. Although we cannot exclude the possibility that Mg²⁺ inhibits MCU from outside the matrix, consistent with the Mg²⁺ and Ca²⁺-dependent destabilization and monomerization of MCU_72-189 that we observed in vitro, increases in matrix Ca²⁺ as well as Mg²⁺ have an inhibitory effect on mitochondrial Ca²⁺ uptake.

Fig. 5. Mg²⁺-dependent MCU inactivation in live cells.

(See also Fig. S6).

(A) Permeabilized HEK293T cells were pulsed with 20 μM Ca²⁺ in the presence of MgCl₂ as indicated. Representative traces show changes in extramitochondrial Ca²⁺ concentrations ([Ca²⁺]_out); (n=3 independent experiments).

(B) Representative traces show ΔΨ_m (i.e. JC-1 ratio) in permeabilized HEK293T cells in response to Ca²⁺ and CCCP. The [Ca²⁺]_out and ΔΨ_m traces were acquired simultaneously; (n=3 independent experiments).

(C) Overlay of kinetic changes in [Ca²⁺]_out during 300 s in the presence of different MgCl₂ concentrations.

(D) The rate of mitochondrial Ca²⁺ uptake was calculated as 1/τ. Mean ± SE. *P < 0.05, ***P < 0.001; (n=3 independent experiments).

(E) Quantification of ΔΨ_m before the addition of CCCP. Mean ± SE. n.s., not significant; (n=3 independent experiments).

(F) and (G) Quantification of [Ca²⁺]_m after CCCP addition. Mean ± SE. *P < 0.05, ***P < 0.001; (n=3 independent experiments).

MCU_72-189 regulates higher order oligomerization of full-length human MCU

Our recent study using SEC and native PAGE suggested that human MCU forms complexes with molecular masses between ~44 and 670 kDa (Tomar et al., 2016). To elucidate whether ectopically expressed full-length human MCU forms oligomers, we used HEK293T cells expressing Flag-tagged MCU protein. The enrichment of human MCU-Flag protein by immunopurification resulted in both low and higher order oligomers that includes monomer, dimer, trimer, tetramer, and pentamers (Fig. 6A). We tested whether the conserved cysteines that reside in the N-terminal soluble region are essential for MCU oligomer formation by generating a single (i.e. C97A) and triple (i.e. C67A, C97A, C191A) MCU mutant constructs. Elimination of these cysteine residues did not affect oligomerization indicating that human MCU could form multimers independent of cysteine crosslinking (Fig. 6C).

Fig. 6. Assembly of human MCU multimers.

(A) HEK293T cells were transfected with MCU-Flag and lysates were subjected to immunopurification. The eluent was subjected to SDS-PAGE separation under reducing conditions followed by probing with anti-Flag antibody.

(B) HEK293T cells were transfected with MCU-Flag truncation constructs [i.e. MCU-Δ1 (residues 150-200), MCU-Δ2 (residues 291-320), MCU-Δ3 (residues 321-351) and MCU-Δ4 (residues 225-231)], immunopurified and detected for MCU complex formation as described above.

(C) MCU cysteine mutants (ΔCys1: C97A, and ΔCys3: C67A, C97A, C191A) were transfected in HEK293T cells, immunopurified and detected for MCU complex formation as described above.

(D) MCU N-terminal mutants (D131A, D131R, D147A and D147R) were transfected in HEK293T cells, immunopurified and detected for MCU complex formation as described above.

(E) HEK293T cells were transfected with MCU-Flag and lysates were subjected to immunopurification by Flag antibody in the presence of varying concentrations of MgCl₂. The eluent was subjected to SDS-PAGE separation under reducing condition followed by probing with anti-Flag antibody.

Images are representative of n=2-3 independent experiments.

Having observed full-length human MCU oligomerization in the absence of crosslinking, we next sought to identify regions of MCU that are determinant of the assembly. We generated a series of N-terminal and C-terminal MCU deletions (Fig. 6B). Remarkably, deletions on both sides of the MCU CC domain disrupted human MCU multimer formation (Fig. 6B). We next examined the effect of N-terminal domain MRAP mutations on full-length MCU complex formation. Our immunopurification revealed that the D131R mutant completely abrogated human MCU complex formation (Fig. 6D), consistent with the greatest suppression of MCU channel activity among all mutants in ectopically expressing HeLa cells (Fig. 3B). Finally, we tested whether Mg²⁺ cations could affect the oligomerization pattern of full-length human MCU, observing a marked MgCl₂ concentration-dependent decrease in MCU oligomer formation (Fig. 6E). Taken together, these data reveal that higher order oligomerization of human MCU requires both N- and C-terminal regions, and Mg²⁺ strikingly weakens MCU multimerization. A recent report suggested that the C. elegans MCU protein with an N-terminal domain deletion also forms a pentameric structure (Oxenoid et al., 2016). Our data show that the N-terminal domain contributes to human MCU oligomer formation and disruption of MRAP by mutation or divalent cations weakens this assembly.

Discussion

The β-grasp family of proteins exhibit a large structural diversity that can include as many as 10 or as few as 3 β-strands and up to 3 α-helices with the central α-helix crossing in either direction (Burroughs et al., 2012). Here, we have found that the β-grasp-like fold of MCU_72-189 not only plays a role as a protein-interaction domain through a tendency for self-association, but also has an intrinsic ability to bind metals in the form of Ca²⁺ and Mg²⁺. Interestingly, this region was also shown to associate with MCUR1 (Lee et al., 2015). Our MCU_72-189 structure shows a dynamic C-terminal tail (i.e. residues 168-189) that cannot be modeled in the electron density. A recent study found that the position of this C-terminus (i.e. residues 166-185) can be stabilized by a lipid-like molecule, ultimately placing a third α-helix close to α2 (Lee et al., 2015) (Fig. S7A). Considering the mobile nature of the C-terminal tail and the close proximity to MRAP, it is tempting to speculate that the 168-189 region could act as a dynamic cap, regulating the accessibility of cations and/or other larger molecules. The same study demonstrated that mutation of the proposed S92 phosphorylation site suppresses MCU activity, but showed an unfavourable distance for ionic interactions between R93 and E95 (Lee et al., 2015) (Fig. S7B). The formation of the R93:E95 salt bridge in our structure suggests this ionic interaction is also dynamic and, considering the close proximity to S92, may play a role in regulating the phosphorylation event that modulates MCU function.

We have found that the MCU_72-189 matrix region binds Ca²⁺ and Mg²⁺ with ~mM affinity in a distinct electronegative patch. The coordination of Mg²⁺ involves oxygen atoms derived from the D147 side chain and five water molecules. Interestingly, the cytosolic domain of the T. maritima Mg²⁺ transporter similarly coordinates Mg²⁺ with one Asp and five water molecules; however, when assembled into a pentamer one water molecule is replaced with an Asp oxygen from an adjacent subunit (Lunin et al., 2006). Thus, additional high resolution structural data on assembled MCU_72-189 multimers are needed to fully appreciate the precise roles of the MRAP residues in coordinating divalent cations in multiple shells and regulating self-association. Protein self-association is known to enhance stability, and thus, the destabilization of MCU_72-189 due to monomerization caused by cation binding is fully consistent with this innate feature of myriad proteins.

Next to K⁺, Mg²⁺ is the second most abundant cation in the cell with cytosolic and nuclear concentrations maintained in the ~mM range (Romani et al., 1993; Szanda et al., 2009). Mg²⁺ enters mitochondria via the Mrs2p channel in an electrogenic manner; hence, changes in ΔΨ_m can modulate Mg²⁺ uptake (Kolisek et al., 2003; Schindl et al., 2007). The free Mg²⁺ concentration in the mitochondrial matrix is ~0.5-1.5 mM and can change depending on the abundance of PO₄³⁻, ADP and ATP (Jung et al., 1990; Rutter et al., 1990). The weak binding affinity estimated for Mg²⁺ is well-suited to the high Mg²⁺ levels of the mitochondrial matrix as proteins that are structurally sensitive to cation binding have evolved affinities close to the concentration range of the cation in the compartment in which they function. Thus, while a fraction of MCU molecules could be loaded with Mg²⁺ under physiological conditions, the level of saturation and MCU inhibition may change dependent on the metabolic state of the cell. With the known role of MICU1 as a gatekeeper of MCU in the resting state (Csordas et al., 2013; Mallilankaraman et al., 2012b; Tsai et al., 2016), it is conceivable that both MICU proteins and divalent cations modulate MCU activity. Indeed, it has long been known that mitochondrial Ca²⁺ uptake can be inhibited by Mg²⁺ (Favaron and Bernardi, 1985; Prentki et al., 1983), and inhibition can occur after physiologically-relevant elevations in cytosolic Mg²⁺ (Szanda et al., 2009).

Conversely, global free Ca²⁺ levels in the matrix would likely never approach mM levels; nevertheless, based on a ΔΨ_m of −180 mV, the Nernst equation shows that Ca²⁺ levels in the mitochondria could reach ~71 mM at cytosolic levels of ~0.1 μM if permitted to freely equilibrate between compartments. Thus, the driving force exists in the highly negative ΔΨ_m to establish mM Ca²⁺ levels within the matrix. Additionally, diffusion theory predicts Ca²⁺ levels as high as ~0.1 mM at a distance of ~10 nm from the exiting region of Ca²⁺ channel pores (Tadross et al., 2013) and ~mM at the immediate base in nanodomains (Bauer, 2001; Chad and Eckert, 1984). Therefore, the ability of MCU_72-189 to bind Ca²⁺ will depend on the proximity to the open pore. Indeed, sustained cytosolic Ca²⁺ inactivates the mitochondrial Ca²⁺ uptake pathway (Moreau et al., 2006), and our observation that elevation of matrix Ca²⁺ inhibits the MCU-dependent Ca²⁺ uptake under normal ΔΨ_m maintenance (Fig. 4) is consistent with this earlier work; moreover, Ca²⁺-binding induced destabilization of MCU_72-189 may be an important contributor to this inhibition. We believe that Ca²⁺-binding induced inactivation of MCU may be a protective buffering mechanism for mitochondrial Ca²⁺ overload-linked cell death. A recent study found that the MCU channel can be active in the absence of the N-terminal domain (Oxenoid et al., 2016), despite our compelling studies here and previous work (Lee et al., 2015) showing a vital functional role for the N-terminal domain. Thus, deletion of the N-terminal domain may relieve all the regulatory constraints placed on the channel region similar to the effect of deletions observed with other proteins.

Taken together, our in vitro and live cell observations reveal a mode for MCU autoregulation mediated by Mg²⁺ and Ca²⁺ levels in the mitochondria. We demonstrate that Ca²⁺ and Mg²⁺ binding to MCU_72-189 destabilize the protein and shift the self-association equilibrium toward monomer. Binding of these divalent cations occurs on a negatively charged region that we have termed MRAP. Disrupting MRAP monomerizes and destabilizes MCU_72-189 similar to cation binding, decreases the sensitivity of the domain to Ca²⁺ and Mg²⁺, perturbs full length human MCU oligomerization and attenuates MCU uptake in live cells in a dominant negative manner. Blocking the Ca²⁺ extrusion path from the matrix decreases the rate of mitochondrial Ca²⁺ uptake without significantly altering ΔΨ_m; further, extramitochondrial Mg²⁺ similarly suppresses the rate of MCU-dependent mitochondrial Ca²⁺ uptake. Based on these observations, we propose a model of autoregulation where Mg²⁺ and/or Ca²⁺ binding to the MCU soluble matrix domain inhibits MCU activity via a destabilization that is concomitant with a suppression in the tendency for self-association (Fig. 7). MCU activity would be fine-tuned by the metabolic state of the cell and the abundance of ATP, ADP and PO₄³⁻ as well as Mrs2p uptake, where increases in free Mg²⁺ would heighten MCU inhibition. Additionally, local increases in Ca²⁺ levels within nanodomains close to MCU_72-189 established by an open MCU pore would also inhibit MCU activity in a negative feedback mechanism. It is tempting to speculate that the ~21 % fast inactivation of the peak inward currents observed during voltage step experiments of patched mitoplasts (Kirichok et al., 2004) is due to the Ca²⁺-dependent inhibition phenomenon we have described herein.

Fig. 7. Model of MCU autoregulation by Mg²⁺ and Ca²⁺.

(A) Activated MCU channel (open) due to increased IMS Ca²⁺ levels. Multimerized MCU_72-189 is shown on the matrix side.

(B) Feedback inhibited MCU channel (closed) due to the disruption of self-association by high Ca²⁺ levels in nanodomains within ~1-6 nm of the channel pore. Monomerized MCU_72-189 is shown relative to the closed C-terminal domain C. elegans MCU model (grey cylinders) (5id3).

(C) Inhibited MCU channel (closed) due to increased free Mg²⁺ levels entering the matrix through Mrs2p. In (A) – (C), Mg²⁺ and Ca²⁺ are depicted as yellow and orange spheres, respectively. The pentameric stoichiometry is based on our full-length human data (Fig. 6) and the recent C. elegans structural work (Oxenoid et al., 2016). Because our data shows that strong interactions under denaturing conditions are compromised and previous work under native conditions suggests the overall assembly between MCU subunits may be preserved after N-terminal deletion (Lee et al., 2015; Oxenoid et al., 2016), cation binding to MCU_72-189 is depicted in (B) and (C) to disrupt the open architecture and not the stoichiometry of MCU. Additional structural and dynamical information is required to reveal the precise stoichiometry of the MCU complex in the presence of regulators and the relative location of the MCU_72-189 region in an open and closed channel.

In summary, our structural and functional results reveal that MCU is autoregulated by prolonged high matrix Ca²⁺ and/or Mg²⁺ that binds to MRAP and destabilizes MCU. This regulatory model is consistent with a recent study demonstrating Mg²⁺-induced inhibition of MCU (Blomeyer et al., 2016) and work showing Ca²⁺-dependent inactivation of mitochondrial Ca²⁺ uptake (Moreau et al., 2006). Further studies are required to determine precisely how MRAP and MCU_72-189 is involved in heteromeric protein-protein interactions known to mediate MCU complex assembly and activity. Nevertheless, these findings provide new mechanistic insights into MCU regulation and avenues to the development of MCU inhibitors that control cell death.

Significance

MCU is a Ca²⁺ selective inwardly rectifying inner mitochondrial membrane resident channel that mediates Ca²⁺ flux into the mitochondrial matrix and is essential for bioenergetics. However, the mechanisms of MCU regulation by cations are unclear. We show for the first time the structural basis for divalent cation interactions with an extensive electronegative surface patch termed MRAP located on the soluble MCU_72-189 matrix domain. Further, we demonstrate that point mutations within MRAP can perturb the assembly of both the isolated MCU_72-189 region and the full-length human MCU protein similar to interactions with cations, thereby leading to channel inhibition. Consistent with our biochemical approaches, we demonstrate that blockade of Ca²⁺ extrusion from the matrix or extramitochondrial Mg²⁺ incubation significantly suppresses MCU activity. Our work reveals a mechanism for divalent cation regulation of MCU activity and provides a framework for the development of small molecule modulators of MCU function by targeting MRAP within MCU_72-189.

Experimental Procedures

Expression, purification and crystallization of MCU_72-189

Native and mutant proteins used for the in vitro experiments were expressed in BL21 (DE3) codon plus Escherichia coli and purified as detailed in the Supplemental Experimental Procedures. All crystals were grown at 18 °C using the hanging-drop vapor diffusion method. Purified proteins in 20 mM HEPES, 150 mM KCl, 2 mM DTT, pH 7.5 were concentrated to ~4.5 mg/mL. Crystallization was performed by mixing 1 μL of protein with 1 μL of precipitant and structures were determined as detailed in the Supplemental Experimental Procedures. The wwPDB coordinate codes are 5KUE (I141M/L146M), 5KUG (lithium), 5KUI (calcium) and 5KUJ (magnesium).

Far-UV circular dichroism spectroscopy

CD spectra and thermal melts were acquired on a Jasco J-810 Spectropolarimeter equipped with a Peltier temperature control system using a quartz cell with a path length of 0.1 cm. Details of the acquisition settings are given in the Supplemental Experimental Procedures.

Analytical ultracentrifugation

Sedimentation equilibrium studies were carried out using a Beckman Optima XL-A Analytical Ultracentrifuge. An An60Ti rotor and six-channel cells with Epon-charcoal centerpieces were used for the data acquisition. Absorbance measurements at 280 nm were collected in 0.002 cm radial steps and averaged over 10 readings. Details of the data analyses are given in the Supplemental Experimental Procedures.

Size exclusion chromatography and multi-angle light scattering

SEC was performed using a Superdex 200 10/300 GL column (GE Healthcare) connected to an AKTA pure FPLC system (GE Healthcare) at 4°C. Three injections of 125 μL each were loaded onto the column and separated at 0.5 mL min⁻¹. SEC-MALS was performed with a Superdex Increase S200 10/300 GL column in-line with a sixteen-angle Dawn Heleos II light-scattering instrument and Optilab TrEX differential refractometer (Wyatt Technologies). Analyses were performed as detailed in the Supplemental Experimental Procedures.

Ca²⁺/Mg²⁺ binding and chemical denaturation curves

Changes in intrinsic fluorescence as a function of CaCl₂, MgCl₂ or guanidine hydrochloride concentrations were taken as an estimate of binding affinity or conformational stability. Experiments were performed on a Cary Eclipse Spectrofluorimeter (Varian/Agilent). Details of these fluorescence measurements are given in the Supplemental Experimental Procedures.

Mitochondrial and cytosolic Ca²⁺ dynamics and confocal microscopy

HeLa cells were transiently transfected with genetically encoded mito-R-GECO1, mitochondrial targeted Ca²⁺ sensor or R-GECO1, cytosolic Ca²⁺ sensor along with Empty vector, MCU-WT-GFP, MCU D131A-GFP, MCU D131R-GFP, MCU D147A-GFP, and MCU D147R-GFP. After 48 hours, the change of mito-R-GECO1/R-GECO1 fluorescence was measured with 488 and 561-nm excitation on a Carl Zeiss META 510 confocal microscope equipped with a 40× oil objective. After 1 min of baseline recording, histamine (100 μM) was added, and the changes in mito-R-GECO1 and R-GECO1 fluorescence was recorded. Images were analyzed using ImageJ (NIH).

Measurement of resting mitochondrial Ca²⁺

Stable HEK293T cell lines were created using a pBSD plasmid which carries blasticidin resistance. HEK293T cells (~6 × 10⁷) stably expressing Empty vector, MCU-WT-Flag, MCU D131A-Flag, MCU D131R-Flag, MCU D147A-Flag, and MCU D147R-Flag were resuspended and permeabilized with digitonin (40 μg/ml) in 1.5 ml of intracellular medium composed of 120 mM KCl, 10 mM NaCl, 1 mM KH₂PO₄, 20 mM HEPES-tris (pH 7.2), 5 mM succinate, bath Ca²⁺ indicator Fura-2FF (0.5 μM), MCU blocker Ru360 (1 μM) and NCLX blocker, CGP (1 μM), and 2 μM thapsigargin to block the SERCA pump. To assess the resting [Ca²⁺]_m, after baseline recording of bath Ca²⁺ (i.e. [Ca²⁺]_out), CCCP (2 μM) was added to release the mitochondrial Ca²⁺. Further details are described in the Supplemental Experimental Procedures.

Ca²⁺ and Mg²⁺ -dependent inactivation of mitochondrial Ca²⁺ uptake

HEK293T cells were washed in Ca²⁺-free PBS, pH 7.4. Approximately 7×10⁶ cells were resuspended and permeabilized with 40 μg/ml digitonin in 1.5 ml of intracellular medium composed of 120 mM KCl, 10 mM NaCl, 1 mM KH₂PO₄, 20 mM HEPES-Tris, pH 7.2 and 2 μM thapsigargin to block the SERCA pump. All measurements were performed in the presence of 5 mM succinate. The simultaneous measurement of ΔΨ_m and extramitochondrial Ca²⁺ (i.e. [Ca²⁺]_out) clearance was achieved by loading the permeabilized cells with JC-1 (800 nM) and Fura-2FF (1 μM), respectively, in the presence and absence of NCLX blocker CGP37157 (10 μM). A series of Ca²⁺ boluses and mitochondrial uncoupler, CCCP (2 μM), were added at the indicated time points. The Mg²⁺-inhibition experiments were performed in a similar manner as the Ca²⁺ inhibition experiments in the absence of the CGP37157, using a single bolus of 20 μM CaCl₂ and varying concentrations of extramitochondrial MgCl₂ from 0 – 10 mM. Additional details are described in the Supplemental Experimental Procedures.

Immunopurification and human MCU oligomerization

HEK293T cells were transiently transfected with the indicated constructs. Flag-tagged proteins were immunopurified and enriched using Flag-antibody (Sigma) and equal amounts of protein were separated on 4-12% Bis-Tris SDS-PAGE under reducing conditions, transferred to a PVDF membrane, and probed with α-Flag-antibody. Further details are described in the Supplemental Experimental Procedures.

Statistical Analysis

Significance was evaluated via Student's unpaired t-test or one-way and two-way ANOVAs.

Highlights.

• The soluble MCU matrix domain contains an acidic patch on the β-grasp-like fold.

• Cation binding or mutation in the acidic patch perturbs MCU assembly and activity.

• Blockade of matrix Ca²⁺ extrusion or Mg²⁺ loading inhibits MCU uptake rates.

• The N-terminal acidic patch confers a mode for MCU autoregulation.