Introduction

Mitochondrial Ca2+ homeostasis is crucial for a variety of cellular processes, including energy metabolism, necrosis, apoptosis, and intracellular and intercellular signaling1,2,3. Impairment of mitochondrial Ca2+ homeostasis can lead to neurodegenerative disease, cardiovascular disease, hypertension, cancer, and diabetes4,5. Several ion channels and transporters, including voltage-dependent anion-selective channels (VDACs), Na+/Ca2+ exchangers (NCLX), leucine zipper EF-hand containing transmembrane protein 1 (LETM1), and the mitochondrial calcium uniporter (MCU) complex, precisely regulate mitochondrial Ca2+ homeostasis6,7. The MCU complex is a highly selective Ca2+ uptake channel complex in vertebrates, which includes the pore-forming MCU, its paralog (MCUb), essential MCU regulator (EMRE), MCU regulator 1 (MCUR1), and several mitochondrial Ca2+ uptake proteins (MICU1, MICU2, MICU3, and MICU1.1)8,9,10,11,12,13,14,15,16. The MICUs, which are localized within the mitochondrial intermembrane space, regulate Ca2+ uptake by the MCU through Ca2+ binding via their EF-hand motifs14,17.

In humans, MICUs include three paralogs and one splicing variant (HsMICU1, HsMICU2, HsMICU3, and HsMICU1.1), and the Ca2+ uptake threshold for the MCU complex in different tissues is dependent on the expression level of the MICUs15,16,17. In its Ca2+-free state, MICU1 acts as a gatekeeper by setting the Ca2+ uptake threshold of the MCU complex, while in the Ca2+-bound state it acts as an activator by potentiating Ca2+ uptake18,19. The molecular structure of MICU1 is a hexamer in the Ca2+-free state and a dimer in the Ca2+-bound state20. MICU2, which interacts with MICU1, increases the Ca2+ uptake threshold of the MCU complex and exhibits lower affinity for Ca2+ than the MICU1 homodimer17,21,22. MICU3 is primarily expressed in the central nervous system and skeletal muscle and also interacts with MICU1 but as an activator of the MCU complex15,23,24,25,26. MICU2 and MICU3 form a back-to-back (B-B) heterodimer in the Ca2+-free state and a face-to-face (F-F) heterodimer in the Ca2+-bound state27,28 (Supplementary Fig. 1). The MICU1-MICU2 F-F heterodimer can form a linear dimer of heterodimers through B-B dimer interaction between MICU2 proteins29. Notably, a dimer of the MICU1-MICU2 heterodimer was observed in an O-shaped MCU holocomplex (dimer of MCU-EMRE tetramer with a dimer of the MICU1-MICU2 heterodimer) in cryo-EM structures30,31,32. This suggests the MCU gating mechanism is regulated via Ca2+-dependent pore blocking by MICUs.

In metazoans, MICU1 and the EMRE are the major regulatory proteins controlling Ca2+ uptake into the mitochondrial matrix and interact directly with the MCU complex30,31. MICU1 interacts with the DXXE motif of the MCU in a Ca2+-dependent manner to control the opening and closing of the MCU, thereby regulating cellular metabolism and survival33,34,35. Expression of MICU1 in 138 sequenced eukaryotes, excluding fungi, has co-evolved with the MCU36,37. By contrast, the EMRE is highly conserved exclusively in metazoan MCU complexes, even though it plays an essential role in linking MICU1 to the MCU and activating Ca2+ uptake12,38,39. Dictyostelium discoideum serves as a minimal eukaryotic model for studying the MCU complex, as it possesses only a single MCU (DdMCU) and MICU (DdMICU)40. Unlike in metazoans, Ca2+ uptake by DdMCU is activated without the EMRE38,40, and cardiolipin plays an essential role in the process41. While the conformation of the N-terminal domain in DdMCU is entirely different from that in the human MCU, the Ca2+-dependent destabilization of oligomerization is similar42. Although structural and functional studies of the DdMCU N-terminal domain have been conducted42, a full understanding of the mechanism governing the DdMCU complex requires detailed structural and biochemical information on DdMICU, the sole regulatory protein in the DdMCU complex.

Here, we determined the crystal structure of Ca2+-bound DdMICU at 2.5 Å, which revealed that DdMICU consists of two lobes (N-lobe and C-lobe) and binds three Ca2+ ions via canonical EF-hand motifs (EFb, EF1, EF4). The Ca2+ binding affinity of these motifs was in the submicromolar range, as with HsMICU1. Structurally, DdMICU is comparable to human MICUs, with an additional Ca2+ binding site in the EFb motif. Like human MICUs, DdMICU forms a F-F dimer via conserved interface interactions but also adopts a head-to-head (H-H) dimer conformation. DdMICU exists in an equilibrium state between dimeric and tetrameric forms and can bind liposomes when the C-terminal helix is present. In our study, by combining the crystal structure of a Ca2+-bound DdMICU with biochemical analyses, we uncovered the molecular basis for the unique role of DdMICU in a MCU system lacking EMRE.

Results

Overall structure of DdMICU

We purified DdMICU (residues 105–461) and used it for crystallization and biochemical studies (Fig. 1a). Endeavoring to crystallize the DdMICU in Ca2+-bound and unbound states, we obtained diffraction-quality crystals in its Ca2+-bound state. The crystal structure of Ca2+-bound DdMICU was then determined by combining the molecular replacement method with single-wavelength anomalous diffraction (MR-SAD) phasing at 2.5 Å resolution with Rwork and Rfree values of 22.1% and 26.0%, respectively (Table 1). The asymmetric unit (ASU) consisted of a DdMICU dimer, and the overall structure of DdMICU comprised two lobes (N-lobe and C-lobe) connected by a bridge helix (Fig. 1b). The N-lobe contained two canonical EF-hand motifs (EFb and EF1) while the C-lobe contained one (EF4), and each of the three motifs was paired with an EF-hand-like motif (EFa, EF2, EF3).

Fig. 1: Overall structure of the DdMICU.
figure 1

a Schematic representation of DdMICU and human MICUs: MTS, mitochondrial targeting sequence; Kn, poly lysine; EF, EF-hand motif; B, bridge helix. The N-lobe, bridge helix and C-lobe are colored cyan, dark gray and pink, respectively. The Ca2+-binding canonical EF-hand motifs are highlighted in gray and marked with a red circle. The slashed area in HsMICU3 indicates an omission of 40 amino acids. b Stereo view of the cartoon representation of the DdMICU monomer structure. Missing regions are depicted with dotted ribbons. The Ca2+ bound in DdMICU is shown as a red sphere, and the N-lobe, bridge helix and C-lobe are colored greenish, grayish and pinkish, respectively.

Full size image
Table 1 Data collection and refinement statistics
Full size table

The EFb motif, which was first noticed in MICUs from plants, is not conserved in mammalian MICUs43. To determine whether the EFb motif is specific to plants or is also present in the protist Dictyostelium, we assessed the protein domain compositions of 650 MICU proteins in three eukaryotic groups: metazoans, protists and plants (Supplementary Data 1). With the exception of Caenorhabditis species, metazoan MICUs were predicted to be unable to bind Ca2+ to the EFb motif, while protists and plant MICUs were predicted to bind Ca2+ to the EFb motif. Subsequent alignment of selected MICU sequences representing each group with DdMICU showed the sequence diversity of the EF-hand motifs in MICUs (Supplementary Fig. 2). The canonical EF-hand sequence (X, Y, Z, −Y, and −Z position carboxyl or carbonyl residues) of the EF1 and EF4 motifs was conserved throughout the entire sample (Fig. 2a, Supplementary Fig. 2) and was also conserved in the EFb motif in non-metazoan MICUs. However, the EFb motif in metazoan MICUs lacked the aspartate residues at the X, Y and Z positions needed for Ca2+ binding.

Fig. 2: Ca2+-binding EF-hand motifs in the DdMICU.
figure 2

a Cartoon representation of the Ca2+ coordination geometry in the canonical EF-hand motifs of DdMICU. The gray and red spheres represent Ca2+ and water molecules, respectively. The Ca2+ coordinating residues are shown as ball-and-sticks, and interactions are depicted as yellow dotted lines. Below, multiple sequence alignments of EF-hand sequences are shown, with canonical EF-hand motif sequences indicated in salmon. b Dose response curves for Ca2+ binding to DdMICU EF-hand mutants, plotting normalized fluorescence against Ca2+ concentration. All experiments are done in triplicate; the circle and error bar indicate the mean and standard deviation. The calculated Kd value from each binding curve is shown at the upper right.

Full size image

Ca2+ ions bound to the three EF-hand motifs in DdMICU through pentagonal bi-pyramidal coordination geometry, which is the typical Ca2+ coordination in canonical EF-hand motifs (Fig. 2a). To compare the Ca2+ binding properties of each EF-hand in DdMICU with HsMICU1, we measured the Ca2+ binding affinity of EF-hand double mutants (DdMICUEFb, DdMICUEF1, and DdMICUEF4) in which only one of the three EF-hand motifs could bind Ca2+. The Kd values for Ca2+ binding to EFb, EF1 and EF4 were 0.36 ± 0.03 µM, 0.36 ± 0.03 µM and 0.24 ± 0.02 µM, respectively (Fig. 2b), which are similar to the apparent Ca2+ binding affinities in HsMICU1 (0.29 ± 0.01 µM in EF1, 0.37 ± 0.15 µM in EF4)22. Additionally, like HsMICU1 and HsMICU2, DdMICU exhibited significant stabilization upon Ca2+ binding to each EF-hand (Supplementary Fig. 3). The melting temperature (Tm) of DdMICU in the apo state (38.4 ± 0.1 °C) was comparable to that of HsMICU1 (40 °C). In the Ca2+-bound state, however, the Tm of DdMICU (76.6 ± 0.3 °C) was significantly higher than that of HsMICU1 or HsMICU2 (HsMICU1; 55 °C, HsMICU2; 63 °C)22. The Tm of the EFb single mutant (Tm of 57.8 ± 0.1 °C), which binds Ca2+ at EF1 and EF4, was similar to that of HsMICU1 and HsMICU2, suggesting that Ca2+ binding to the EFb motif further stabilizes the DdMICU. Human MICUs are stabilized by Ca2+ binding and set the Ca2+ uptake threshold of the MCU complex based on a Ca2+ binding affinity of about 0.3 µM. Given the present results, DdMICU would be expected to set a threshold for the DdMCU complex based on a similar Ca2+ binding affinity, and the EFb motif may help maintain the activation state of the DdMCU complex by stabilizing the Ca2+-bound state conformation of the DdMICU.

Comparisons of the DdMICU structure with HsMICUs

To assess structural differences, the Ca2+-bound DdMICU monomer and human MICU monomers were aligned and the root mean square deviation (RMSD) values were calculated (Table 2). The RMSDs between the DdMICU and human MICU monomers were approximately 3.0 Å. The RMSDs between the C-lobes were 1.7 Å or less, and between the N-lobes were 2.1 Å or more. We next compared the DdMICU structure with HsMICU1 in the HsMICU1-HsMICU2 dimer, which had the highest sequence homology with DdMICU among human MICUs and was a physiologically relevant formation (Fig. 3a). When we superimposed the DdMICU and HsMICU1 based on the C-lobe, the N-lobe of the DdMICU was rotated outward by approximately 20° (Fig. 3b).

Fig. 3: Structural comparison of DdMICU and HsMICU1.
figure 3

a Cartoon representation of the superimposed DdMICU and HsMICU1 (PDB code: 6XQO) monomer structures. b Structural comparison of the N-lobe between DdMICU and HsMICU1. The two structures are aligned based on the C-lobe, which is represented as a space-filling model. The tilt angle of the N-lobe is marked with a black arrow. c Superimposition of the N-lobe structures of DdMICU and HsMICU1. d Close-up view of the interaction residues involved in the EFb motif. e Superimposition of the C-lobe structures of DdMICU and HsMICU1. The N-lobe, bridge helix and C-lobe of DdMICU are colored cyan, green and pink; HsMICU1 is colored gray. The Ca2+ bound in DdMICU and HsMICU1 is shown as red and gray spheres, respectively.

Full size image
Table 2 RMSD values between DdMICU and human MICU structures
Full size table

To make a more detailed comparison, we separately superimposed the N- and C-lobes of DdMICU on corresponding lobes of HsMICU1. The N-lobe, which bound an additional Ca2+ in the EFb motif, differed somewhat from the N-lobe in humans (Fig. 3c). The EF-like motif of HsMICU1, which was aligned to the EFb motif of DdMICU, was loosened, and the α3-helix was located approximately 10 Å away. Notably, the hydrophobic patch at the α2, α4 and bridge helices interacted with phenylalanine residues in HsMICU1, which are conserved among HsMICUs (HsMICU1; Phe195, HsMICU2; Phe228, and HsMICU3; Phe209) (Fig. 3d). However, the canonical EF-hand loop in the EFb motif in DdMICU was fixed by Ca2+ binding to Asp191, and Phe187 interacted with Val153 on the exterior but not with the hydrophobic patch interior. The phenyl group of Phe187 was rotated about 120° in comparison to that residue in human MICUs. Simultaneously, the α7-helix of DdMICU was a turn shorter than in human MICUs due to the structural hindrance of the EFb motif. When we compared the C-lobe structure, it differed slightly from HsMICU1 in that the α11-helix of DdMICU was straight, whereas it had a short break caused by kink in HsMICU1 (Fig. 3e). Moreover, the α11-helices in HsMICUs in the Ca2+-bound state were all kinked or shortened. Taken together, these results show the overall structure of DdMICU to be similar to that of HsMICUs, though there are local differences.

Structural details in the F-F dimer interface of DdMICU

MICU F-F dimers are known to function as gatekeepers within the MCU complex, and their structure is common to all known MICUs20,27,28,29,30,31,32,44,45 (Supplementary Fig. 4). The F-F dimer in DdMICU, observed through crystallographic symmetry, was similar to that in HsMICUs other than HsMICU1 in the Ca2+-bound state, with a RMSD value of about 3.2 Å (Fig. 4a, Table 2). Despite a sequence homology of about 50% between DdMICU and HsMICU1, the RMSD between homodimers of DdMICU and HsMICU1 was high, about 13.6 Å (Supplementary Fig. 5, Table 2). Consistent with the RMSD values, the interface area of the F-F dimer for DdMICU (1370 Å2) was similar to those of F-F dimers of HsMICU1-HsMICU2, HsMICU2 or HsMICU3 (1400 Å2), but not in HsMICU1 (470 Å2) (Table 3 and Fig. 4b). Moreover, the calculated binding energy of the DdMICU F-F dimer (−10.6 kcal/mol) was comparable to that of the F-F dimers of HsMICU1-HsMICU2, HsMICU2 or HsMICU3 (−9.0 kcal/mol), but not HsMICU1 (−6.6 kcal/mol). When we assessed the sequence conservation on the surface of DdMICU, it clearly matched all the F-F dimer interface residues, which were highly conserved with conservation scores over 70% in the ConSurf server (Fig. 4c). We therefore suggest that the F-F dimer of DdMICU has a similar overall topology to HsMICU1-HsMICU2 and to the HsMICU2 and HsMICU3 dimers, reflecting a conserved dimer interface with a similar interface area and calculated binding energy.

Fig. 4: F-F dimer of DdMICU and comparison with HsMICU1.
figure 4

a Cartoon and space-filling representations of the F-F dimer of DdMICU. b Representations of the F-F dimer interface in DdMICU using space-filling models. The F-F dimer interface is colored slate. c Sequence conservation of DdMICU depicted with a space-filling model. Conserved sequences are colored from white (variable, grade 5) to burgundy (conserved, grade 9) according to ConSurf color coding. d–f Interactions at the DdMICU F-F dimer interface (d), HsMICU1-HsMICU2 (PDB code: 6XQO) interface I (e) and HsMICU1-HsMICU2 interface II (f). Mol A and Mol B of DdMICU are colored red and blue, respectively; HsMICU1 and HsMICU2 are colored orange and green.

Full size image
Table 3 Dimer interface analysis of MICU structures
Full size table

We next compared the interacting residues at the F-F dimer interface to confirm whether the essential interactions for dimerization in HsMICU1-HsMICU2 were conserved in DdMICU. With DdMICU, the F-F dimer interface interactions involved three pairs of electrostatic interactions (A: Ser216-Asp369, B: Lys222-Tyr391 and C: Tyr435-Asp375) as well as hydrophobic interactions (Fig. 4d). Notably, the interaction between Ser216 and Asp369 was conserved in HsMICU1-HsMICU2, where it corresponded to an Arg221(HsMICU1)-Asp330(HsMICU2) interaction, which is an important salt bridge at interface I in HsMICU1-HsMICU229 (Fig. 4e). Additionally, the hydrogen bond with Tyr391 was conserved with Arg352(HsMICU2) and was essential for heterodimerization at interface I in HsMICU1-HsMICU228,30. These two electrostatic interactions were not conserved at interface II in HsMICU1-HsMICU2 (Fig. 4f). Interestingly, Tyr435 and Asp375 formed a hydrogen bond through the side chain hydroxyl and carboxyl groups, and these interacting residues were conserved in Dictyostelium species. Thus, the electrostatic interactions in the DdMICU F-F dimer includes two hydrogen bonds that are similar to those observed in HsMICUs and one putative Dictyostelium-specific hydrogen bond.

In addition to the electrostatic interactions at the F-F dimer interface in DdMICU, we analyzed the hydrophobic interactions using the PDBePISA server. Based on the calculated solvation energies, which correspond to hydrophobic interactions, four I/L knob residues (Ile219, Ile223, Leu372, and Leu376) were the primary contributors to the hydrophobic interactions at the F-F dimer interface of DdMICU (Fig. 4d). The solvation energies of these four I/L knob residues were the highest among 21 hydrophobic interaction residues, with ∆iG values of 1.34, 1.79, 1.59 and 1.60, respectively. Leu372 and Leu376 in the α11-helix interacted with the hydrophobic cleft formed by the α4-helix and α5-helix, and Ile219 and Ile223 in the α5-helix interacted with the hydrophobic cleft formed by the α11-helix and α12-helix. Notably, Ile223 and Leu376 were conserved as methionine knobs in the HsMICU1-HsMICU2 structure and were similarly positioned within the hydrophobic pocket formed by opposing EF-hand helices29 (Fig. 4e). Ile223 and Leu372 also interacted with Tyr391 and Tyr435, which participated in electrostatic interactions. Collectively then, the F-F dimer structure of DdMICU is similar to that of human MICUs, and the electrostatic and hydrophobic interactions at the F-F dimer interface observed in human MICUs are well conserved in DdMICU.

Dimer formation and oligomeric states of DdMICU

Both the F-F and B-B dimers are crucial for gatekeeping in the human MCU complex, and an O-shaped holo-complex is formed in the high Ca2+ state through the B-B dimer interaction in HsMICU230,31,32. Although the B-B dimer was not observed in the DdMICU crystal, we identified a unique dimeric state, the “Head-to-Head (H-H)” dimer (Fig. 5a). The H-H dimer of DdMICU was observed in an asymmetric unit and was able to form a tetrameric structure through interaction with the F-F dimer (Fig. 5c). To analyze the oligomeric states of DdMICU in solution, we performed multi-angle light scattering coupled with size exclusion chromatography (SEC-MALS). The calculated molar mass (Mw) of the DdMICU in solution was 169.9 ± 2.2, 114.4 ± 7.9, 97.5 ± 9.5 and 92.4 ± 1.2 kDa at [NaCl] = 50, 100, 150 and 300 mM, respectively (Fig. 5d). These results suggest DdMICU is a tetramer at the low ionic strength of 50 mM, but as the ionic strength increases the equilibrium of the oligomeric state shifts to the dimer at ionic strengths of 150 mM or higher.

Fig. 5: Structural and biochemical analysis of the DdMICU H-H dimer and tetramer.
figure 5

a Cartoon and space-filling representations of the H-H dimer of DdMICU. b Close-up view of the H-H dimer interface interactions. The interaction residues are shown as sticks, with interactions depicted as dotted lines. Ca2+ is shown as a red sphere, a water molecule (w) as a white sphere. c Structure of the DdMICU tetramer. Mol-A, Mol-B, Mol-A’ and Mol-B’ of DdMICU are colored red, cyan, blue and gray, respectively. Ca2+ is shown as red spheres. The asymmetric unit of the DdMICU crystal is outlined with gray lines (H-H dimer), and the crystallographic two-fold symmetry is indicated by a gray-dotted line (F-F dimer). d SEC-MALS assay of DdMICU in solution with NaCl concentrations of 50 mM (black), 100 mM (red), 150 mM (green) and 300 mM (blue). The relative UV absorbance at 280 nm is shown as a line, with calculated molar masses represented by circles and indicated as the mean ± standard deviation.

Full size image

We also analyzed the interface residues of the H-H dimer of DdMICU in detail. The interface of the DdMICU H-H dimer included several hydrogen bonds (Fig. 5b). The mainchain carbonyl group of Gly228 and Asp229, located at the loop of the EF1 motif in Mol B, formed hydrogen bonds with Lys394 and Asp400 in Mol A. Asp229 participated in both the interface interaction of the H-H dimer and the Ca2+ coordination in the EF1 motif. Arg280 and Gln285 of Mol B participated in hydrogen bond networks with Arg295 and Asp355 of Mol A, which were mediated by one water molecule. To confirm whether the H-H dimer interface interactions contribute to tetramerization, we purified an H-H dimer mutant (DdMICU4Ala; Q285A/R295A/D355A/K394A) and conducted SEC-MALS assays in [NaCl] = 50 and 300 mM (Supplementary Fig. 6). The calculated Mw of the DdMICU4Ala in solution was 88.4 ± 1.5 and 83.7 ± 0.3 kDa at [NaCl] = 50 and 300 mM, respectively, indicating that the H-H dimer interface interaction is critical for tetramerization. To further assess the structural contribution of the H-H dimer interface, we analyzed its interface area, which was approximately 650 Å2, similar to that of B-B dimers (~700 Å2) but narrower than that of F-F dimers (~1200 Å2) in human MICUs (Table 3). The calculated binding energy of the DdMICU H-H dimer was −8.7 kcal/mol, which was similar to that of the F-F dimers (−8.4 kcal/mol) in human MICUs but lower than that of the B-B dimers of HsMICU2 and HsMICU3 (−7.5 kcal/mol). Thus, the H-H dimer interface in DdMICU includes several hydrogen bonds and may form an energetically stable dimer similar to the F-F dimers in human MICUs. This suggests the H-H dimer interaction could contribute to the oligomeric assembly of DdMICU, potentially influencing its regulatory function in the DdMCU complex.

Liposome binding properties of DdMICU through the C-terminal helix

The C-terminal helix (C-helix) of HsMICU1 is crucial for its gatekeeping function and directly interacts with the membrane and HsMCU30,46. Because the C-helix of DdMICU (444-TGIFSKFKKILKVLT-458) was poorly defined in the electron density maps due to the flexibility of the crystal lattice, we modeled the C-helix of DdMICU and analyzed its characteristics (Fig. 6a). We found that the C-helix of DdMICU is amphipathic, consisting of positively charged and hydrophobic residues on opposite sides (Fig. 6b). Four lysine residues gathered on one side to form a positively charged face, while seven hydrophobic residues gathered to form the opposite hydrophobic face. Comparison of the properties of the C-helices in DdMICU and HsMICU1 using the HELIQUEST47 server revealed that the mean hydrophobicity (H) of the C-helix of DdMICU is similar to that of HsMICU1 (0.58). In DdMICU, however, the net charge (z) was twice as high (DdMICU: 4, HsMICU1: 2), and the hydrophobic moment (µH) was stronger (DdMICU: 0.68, HsMICU1: 0.55) (Supplementary Fig. 7). These observations raise the possibility that the C-helix of DdMICU may also participate in membrane binding. We then performed liposome flotation assays using wild-type DdMICU and a C-helix deletion construct (DdMICU∆C) to investigate whether DdMICU would interact with liposomes serving as a peripheral membrane model (Fig. 6c, Supplementary Figs. 8 and 9). DdMICU bound to liposomes under both Ca2+ and EGTA conditions. This is consistent with earlier findings for HsMICU1, which also bound to liposomes, irrespective of the presence of Ca2+ or EGTA22,31. However, unlike HsMICU1, which bound to liposomes with or without its C-helix, DdMICU∆C did not bind to liposomes, despite the presence of cardiolipin. This suggests the C-helix of DdMICU plays a key role in membrane binding in a MCU system lacking an EMRE.

Fig. 6: Biochemical and structural characteristics of the C-helix in DdMICU.
figure 6

a Side view cartoon representation of the C-helix in DdMICU. Side chains are depicted as ball-and-stick models. Positively charged, hydrophobic and polar residues are labeled in blue, gray and black, respectively. b Front view cartoon representation of the C-helix in DdMICU. c Liposome flotation assay with DdMICU, DdMICU∆C, HsMICU1, and HsMICU1∆C. Proteins were resolved by 12% SAS-PAGE gel, with the size marker (M) indicated.

Full size image

Discussion

The MCU complex plays a key role in regulating mitochondrial Ca2+ homeostasis, which is essential for cellular function as well as mitochondrial metabolism, and is tightly regulated by essential regulatory proteins such as MICUs and the EMRE. Unlike in metazoans, the non-metazoan MCU complex is characterized by the MCU’s ability to take up Ca2+ without the EMRE38. This means that the Ca2+ uptake function of the MCU is directly regulated by a MICU. Despite the importance of non-metazoan MICUs being able to control MCU complexes without an EMRE, the molecular structure and underlying mechanism of non-metazoan MICUs were lacking. We therefore determined the crystal structure of DdMICU to study the non-metazoan MICU in an EMRE-free MCU complex system.

The overall structure of DdMICU is similar to that of human MICUs in that it is composed of an N-lobe, bridge helix, and C-lobe (Figs. 1b and 3a). DdMICU contains three canonical EF-hand motifs: the EF1 and EF4 motifs as well as a non-metazoan-specific EFb motif (Fig. 2a). These three motifs have conserved canonical EF-hand sequences with pentagonal bipyramidal geometry and a structure that is stabilized by Ca2+ binding (Supplementary Fig. 3). Moreover, the Kd values for Ca2+ binding to all three EF-hand motifs in DdMICU are approximately 0.3 µM, which is comparable to the Kd values of HsMICU1 (Fig. 2b). The Ca2+ uptake threshold of the human MCU regulated by MICUs is comparable to the Kd value of the EF-hand motifs in each MICU22. It is therefore likely that the Ca2+ uptake threshold of the DdMCU regulated by DdMICU is comparable to the Kd values of its EF-hand motifs. The DdMICU F-F dimer also showed structural similarity to the HsMICU1-HsMICU2 dimer (Fig. 4a and Supplementary Fig. 4), with conserved hydrophobic and electrostatic interactions at the dimer interface (Fig. 4b, c). Given the structural resemblance between DdMICU and human MICUs and conservation of key interface interactions, it is likely that the structural changes elicited by Ca2+ binding to the EF1 and EF4 motifs located at the dimer interface of DdMICU are similar to those observed in the HsMICU1-HsMICU2 dimer. Consequently, the Ca2+ uptake function of the DdMCU may be regulated by structural changes to DdMICU at the Ca2+ uptake threshold, which are likely comparable to those in HsMICU1.

Cryo-EM structures of the human MCU holo-complex revealed that the gating mechanism of the human MCU complex is dependent on interactions between MICU1-MICU2 and MCU-EMRE30,31,32,48. MICU1 acts as a gatekeeper for the MCU, blocking the IMS entrance at resting [Ca2+] levels and controlling its opening through Ca2+-dependent interactions with the MCU pore17,33,34,35. Notably, under high Ca2+ conditions, an O-shaped MCU holo-complex is formed through interactions between MICU1 and EMRE involving electrostatic interactions between the poly K domain of MICU1 and the poly D domain of EMRE, with no direct MCU-MICU interactions30,31,32. Furthermore, a working model of the human MCU complex suggests the apo HsMICU1-HsMICU2 heterodimer detaches from the HsMCU pore due to steric hindrance with the membrane as Ca2+ levels rise31. Similarly, if we aligned the model of the F-F dimer of the Ca2+-bound DdMICU structure with the apo state model with the same MCU pore interactions, DdMICU would also be expected to clash with the membrane surface and be dislodged from the DdMCU pore in the Ca2+-bound state (Supplementary Fig. 10). This suggests the mechanism by which the DdMICU regulates Ca2+ uptake shares similarities with that of HsMICUs, despite the lack of an EMRE. We therefore speculate that at low Ca2+ concentrations (less than 0.3 µM), DdMICU binds to the DdMCU through conserved MCU interaction motifs, blocking the DdMCU pore and acting as a gatekeeper. At higher Ca2+ concentrations (0.3 µM or above), Ca2+ binding to the EF-hand motifs of DdMICU induces a structural change that causes DdMICU to dissociate from the DdMCU (Fig. 7a). Despite detaching from the DdMCU pore, we suggest the amphipathic C-helix of DdMICU interacts with the membrane, preventing DdMICU from detaching from the mitochondrial inner-membrane and keeping it in close proximity to the DdMCU.

Fig. 7: Proposed model for Ca2+ uptake via the DdMCU-DdMICU complex.
figure 7

a Model of the DdMCU-DdMICU complex in low and high Ca2+ states. In the resting condition, DdMICU binds to the DdMCU pore and the membrane through conserved N-terminal residues and the C-helix. In the high Ca2+ state, DdMICU dissociates from the DdMCU, enabling Ca2+ uptake through the DdMCU. DdMICU remains anchored to the membrane via its C-helix. b Horizontal model of the DdMCU-DdMICU holo-complex. IMS, IMM, and Matrix indicate the intermembrane space, inner mitochondrial membrane, and mitochondrial matrix, respectively.

Full size image

We observed that DdMICU forms a H-H dimer, which further assembles into a tetramer through interactions with the F-F dimer interface (Fig. 5b, Supplementary Fig. 6). The tetramerization of the DdMICU was dependent on solution ionic strength, and DdMICU was present in an equilibrium between tetrameric and dimeric forms at physiological ionic strength (i.e., 100–150 mM) (Fig. 5d). The tetramer, including the H-H dimer, represents a distinct oligomeric structure that has not been observed with human MICUs. With DdMICU in the tetrameric state, its MCU interaction residues in the N-terminus and C-terminus helices are oriented in opposite directions. Consequently, the DdMCU holo-complex is likely to be oriented horizontally (Fig. 7b). Given that the HsMICU1 homo-hexamer is crucial for maintaining the cristae junction49, the DdMCU holo-complex may also be positioned at the cristae junction and contribute to its maintenance. Further studies will be necessary to ascertain whether DdMICU is involved in cristae junction formation and plays a key role in cristae separation via its multimeric states.

In summary, we report the structural characteristics of DdMICU in the Ca2+-bound state, highlighting its three canonical EF-hand motifs, which bind Ca2+, and its oligomeric states. We suggest DdMICU acts as a gatekeeper for the DdMCU, blocking the pore at Ca2+ concentrations below 0.3 µM and dissociating upon Ca2+ binding. The membrane-anchoring C-helix of the DdMICU may help to stabilize its proximity to the DdMCU. These structural insights suggest a Ca2+-dependent regulatory mechanism similar to that of metazoan MICUs, with additional regulatory features influenced by the C-helix of MICUs and its unique tetrameric structure. Given these findings, further studies will be needed to investigate the structural and functional features underlying the mechanism governing this MCU complex system, which lacks an EMRE.

Materials and methods

Cloning and purification of the DdMICU constructs

DdMICU (EAL63559.1, residues 105–461) was optimized with the Escherichia coli (E. coli) codon and synthesized (GenScript). DdMICU wild-type, EF-hand single mutants (DdMICU∆EFb; E202A, DdMICU∆EF1; E236A, DdMICU∆EF4; E425A), EF-hand double mutants (DdMICUEFb; E236A/E425A, DdMICUEF1; E202A/E425, DdMICUEF4; E202A/E236A), DdMICU∆C (residues 105-437), and DdMICU4Ala (Q285A/R295A/D355A/K394A) constructs were amplified and cloned into a modified pRSFDuet-1 (Novagen) expression vector that had a hexahistidine (6×His) tag at its N-terminus. The recombinant vectors were transformed into E. coli BL21 (DE3) and grown in Luria-Bertani (LB) medium containing 50 μg/mL kanamycin at 37 °C until the UV absorbance at 600 nm was 0.6. Overexpression of the recombinant protein was induced by addition of 0.2 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 20 °C for 18 h.

Cells were harvested by centrifugation (4000 × g) for 20 min at 4 °C, resuspended in buffer A [50 mM Tris-HCl pH 8.5, 500 mM NaCl, 5 mM imidazole, 5 mM CaCl2, 5 mM β-mercaptoethanol, and 5% (v/v) glycerol] and lysed by sonication. After centrifugation (15,000 × g) of the lysates for 1 h, the supernatants were loaded onto an Econo-Column (Bio-Rad) packed with Ni-IDA agarose resin (Elpis-Biotech) pre-equilibrated with buffer A. The bound protein was washed with buffer B [20 mM Tris-HCl pH 8.5, 500 mM NaCl, 40 mM imidazole, 5 mM CaCl2, 5 mM β-mercaptoethanol, and 5% (v/v) glycerol] and eluted with buffer C (20 mM Tris-HCl pH 8.5, 500 mM NaCl, 300 mM imidazole, 5 mM CaCl2, and 5 mM β-mercaptoethanol). The eluate was concentrated to 5 mL using an Amicon Ultra-15 30 K (Millipore, Merck) and further purified through size exclusion chromatography (SEC) on a HiLoad 16/60 Superdex 200 prep (Pharmacia) column pre-equilibrated with SEC buffer A (20 mM MES-NaOH pH 6.8, 100 mM NaCl, 5 mM CaCl2, and 5 mM DL-dithiothreitol). The single peak fractions were collected, concentrated up to 10 mg/mL, and stored at -80°C.

For seleno-L-methionine (Se-Met) incorporation, methionine-auxotrophic E. coli B834 (DE3) (Novagen) cells were transfected with a recombinant vector, pre-cultured in LB medium and harvested by centrifugation (4000 × g) for 20 min at 4 °C. The pellet was resuspended in M9 minimal medium [M9 medium supplemented with 50 μg/mL Se-Met, 0.4% (w/v) glucose, 0.004% (w/v) thiamine, 2 mM MgSO4, 0.2 mM CaCl2 and amino acids] to wash out the LB medium and harvested again by centrifugation. After the washing step, a resuspension of the cell pellet in M9 minimal medium was transferred to a fresh culture in 2 L of M9 minimal medium. The overall procedure to purify the Se-Met derivative of DdMICU was the same as for native DdMICU.

Crystallization of the DdMICU

To crystallize native and Se-Met DdMICU, initial screening was performed using the sitting-drop vapor diffusion method in a 96-well INTELLI-PLATE (Art Robbins Ins.) by mixing 0.2 μL of the protein and 0.2 μL of reservoir solution. After one to two weeks, native and Se-Met DdMICU formed microcrystals in reservoir solution containing PACT Premier (Molecular Dimensions) and Wizard Classic 3 & 4 (Rigaku Reagents). Additional crystallization trials were performed using the hanging-drop vapor-diffusion method in a 24-well VDX plate (Hampton Research) by mixing 2 μL of the protein and 1 μL of the reservoir solution. Final optimized native and Se-Met DdMICU crystals were obtained in a reservoir solution containing [0.1 M MIB buffer (Molecular Dimensions) pH 5.5, 20% (w/v) PEG 1500, 10 mM 3-((3-cholamidopropyl) dimethylammonio)-1-propanesulfonate (CHAPS), and 10 mM tris(2-carboxyethyl)phosphine (TCEP)] and [0.1 M MMT buffer (Molecular Dimensions) pH 7.0, 25% (w/v) PEG 1500, 0.1 M guanidine hydrochloride, 16 mM CHAPS and 10 mM TCEP], respectively. Crystals were cryoprotected by soaking for 10 min in mother liquor containing an additional 20% (v/v) ethylene glycol before flash freezing in liquid nitrogen.

Data collection, structure determination, and analysis

Native and SAD data sets were collected at a wavelength of 0.9794 Å on beamlines 5C and 7A at the Pohang Accelerator Laboratory (PAL) (Pohang, Korea). We collected the best resolution diffraction data for native DdMICU at 2.5 Å resolution and for Se-Met DdMICU at 3.5 Å resolution. Data indexing, integration and scaling were performed using the XDS package50 and HKL suite51. Both native and Se-Met DdMICU crystals were in the tetragonal space group I4122 (a = b = 103.9 Å and c = 297.5 Å for native DdMICU, a = b = 104.5 Å and c = 295.0 Å for Se-Met DdMICU). Because we were not able to solve the structure using the molecular replacement (MR) method with the human MICU structure, we used MR-SAD methods to solve the DdMICU structure. The 3.5 Å resolution data from Se-Met DdMICU were used to obtain the initial phase information and model from the anomalous dispersion of selenium with AutoSol and AutoBuild in the PHENIX package52. Using this initial model, we performed additional phase refinement and model building with Phaser-MR and Phaser-EP in the PHENIX package and with Coot53 using native DdMICU data. The final model was refined and modeled after iterative cycles with phenix.refine in the PHENIX package, REFMAC5 in the CCP4i suite54, and Coot. The coordinates and structural factors were deposited in the PDB RCSB with an accession code of 9K6M. PyMOL55 version 2.5.0 was used for all structural figures, and PDBePISA56 and PRODIGY57 were used for interface analysis. The C-helix of DdMICU and the apo state of the DdMCU-DdMICU complex were modeled using AlphaFold 358. The characteristics of the C-helix was analyzed with HELIQUEST47.

Sequence alignment and domain composition analysis

Multiple sequence alignments were performed using the Muscle algorithm in MEGA-X59 and graphically displayed using ESPript 3.060. Domain compositions of MICU proteins were identified with the PROSITE61 database using a list of 650 MICU sequences analyzed by Pittis et al.37. Sequence conservation of DdMICU was visualized using the ConSurf62 server.

Microscale thermophoresis analysis

The Ca2+ binding affinities of DdMICU EF-hand mutants were measured using MST assays. For those assays, purified DdMICU EF-hand double mutants were subjected to further gel filtration in SEC buffer B (20 mM HEPES-NaOH pH 7.0, 300 mM NaCl, 5 mM EGTA, and 5 mM DL-dithiothreitol) using a Superdex 200 10/300 GL column (GE Healthcare). To remove the remaining Ca2+ and EGTA, we dialyzed with SEC buffer B without EGTA for 48 h at 4 °C. The residual Ca2+ was measured after dialysis using the Ca2+-indicator fura-2 [non-acetoxymethyl ester (AM) form, Molecular Probes, Eugene, OR], which confirmed that the remaining Ca2+ concentration was less than 10 nM. The 6×His proteins (500 nM) were labeled using a His-Tag Labeling Kit-RED-tris-NTA 2nd Generation (NanoTemper Technologies) and incubated with serially diluted CaCl2 (12 pM-400 µM). To minimize surface adsorption and improve the homogeneity of the sample, we added 0.05% (v/v) Tween-20 and 0.1% (w/v) BSA to the buffer used to dilute all components. After combining the various components, the samples were incubated at room temperature for 30 min prior to being loaded into standard capillaries (NanoTemper Technologies). The MST assays were performed in triplicate (n = 3) using a Monolith NT.115 Pico instrument (NanoTemper Technologies, Germany), and the data were analyzed using MO.Affinity Analysis version 2.3 (NanoTemper Technologies)63,64.

Heat aggregation assay

The melting temperatures of DdMICU wild-type and its EF-hand mutants were determined spectrophotometrically as described previously65. Briefly, 100 µM purified protein in SEC buffer A was placed in PCR tubes, and the temperature was increased at the rate of 4 °C per 90 s using a Thermal Cycler Dice Gradient (Takara Bio, Inc.). Turbidity was measured based on the absorption at 470 nm using a Nanodrop-2000 spectrophotometer (Thermo Fisher Scientific, Inc.). The dose-response curve was fitted using OriginLab Pro 9.1, and the melting temperature (Tm) was labeled at the bottom right. All experiments were done in triplicate (n = 3). Data are presented as the mean ± 95% confidence interval (CI) based on the two-tailed t-distribution.

SEC-MALS of DdMICU

To characterize the oligomeric state of the DdMICU, size-exclusion chromatography coupled to multi-angle light scattering (SEC-MALS) was performed using a Superdex 200 10/300 GL column (GE Healthcare) connected to a MALS detector (MiniDAWN TREOS, Wyatt Technology) and a differential refractometer (Optilab rEX, Wyatt Technology). Aliquots of purified DdMICU in a volume of 200 µl at 2.0 mg/mLwere loaded into the column pre-equilibrated with SEC buffer A containing different NaCl concentrations normalized using bovine serum albumin. Weight-average molar masses were analyzed using the Zimm model for fitting and calculated using ASTRA 6 software (Wyatt Technology). The solvent refractive index and viscosity values were defined as 1.33 and 0.89 cP. The dn/dc value for all samples was 0.185 mL/g, a standard value for proteins.

Liposome preparation and flotation assays

A mixture of phosphatidylethanolamine (PE), phosphatidylcholine (PC) and cardiolipin (CL) (Avanti Polar Lipids, Inc.) at a ratio of 5:3:2 (w/w) was dissolved in chloroform and dried using an IKA RV 8 rotary evaporator (IKA, Inc.). The dried lipids were then hydrated in SEC buffer C (20 mM HEPES-NaOH pH 7.0, 300 mM NaCl, and 5 mM DL-dithiothreitol) to a final concentration of 10 mg/mL. All buffers contained either 5 mM CaCl2 or EGTA, depending on the experimental conditions. Following vortexing, liposomes were formed by freeze-thawing five times and extrusion ten times through a 200 nm filter using an Avanti Mini-extruder (Avanti Polar Lipids, Inc.) at 65 °C.

For the protein flotation assay, 500 µg of protein-free liposomes were mixed with 50 µg of protein in SEC buffer C to a final volume of 350 µl. The mixture was incubated for 10 min at room temperature, then mixed 1:1 with an 80% (w/v) HistoDenz™ (Sigma-Aldrich, Merck) solution in SEC buffer C. That mixture was added to the bottom layer of an open-top thick-walled polycarbonate tube (Beckman). A 600 µl layer of 30% HistoDenz™ solution in SEC buffer C was then added on top, followed by a 200 µl layer of SEC buffer C. The gradient was centrifuged at 259,000 × g for 3 h at 4 °C using an Optima MAX-TL ultracentrifuge with a TLS-55 rotor (Beckman). Liposomes were collected from the top layer, sampled and subjected to SDS-PAGE, followed by Coomassie staining. HsMICU1 (97–476) and HsMICU1ΔC (97–444) were used as positive controls. These proteins were purified using the same method described previously29.

Statistics and reproducibility

Error bars represent the mean ± standard deviation (SD) or mean ± 95% confidence interval (CI), based on at least three biologically independent experiments. The 95% CI was calculated based on the two-tailed t-distribution.

Reporting summary

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