The structure of the MICU1‐MICU2 complex unveils the regulation of the mitochondrial calcium uniporter

Abstract

The MICU1‐MICU2 heterodimer regulates the mitochondrial calcium uniporter (MCU) and mitochondrial calcium uptake. Herein, we present two crystal structures of the MICU1‐MICU2 heterodimer, in which Ca²⁺‐free and Ca²⁺‐bound EF‐hands are observed in both proteins, revealing both electrostatic and hydrophobic interfaces. Furthermore, we show that MICU1 interacts with EMRE, another regulator of MCU, through a Ca²⁺‐dependent alkaline groove. Ca²⁺ binding strengthens the MICU1‐EMRE interaction, which in turn facilitates Ca²⁺ uptake. Conversely, the MICU1‐MCU interaction is favored in the absence of Ca²⁺, thus inhibiting the channel activity. This Ca²⁺‐dependent switch illuminates how calcium signals are transmitted from regulatory subunits to the calcium channel and the transition between gatekeeping and activation channel functions. Furthermore, competition with an EMRE peptide alters the uniporter threshold in resting conditions and elevates Ca²⁺ accumulation in stimulated mitochondria, confirming the gatekeeper role of the MICU1‐MICU2 heterodimer. Taken together, these structural and functional data provide new insights into the regulation of mitochondrial calcium uptake.

Structural analysis of MICU1–MICU2 heterodimers reveals how Ca²⁺ binding modulates gatekeeping and activation of the mitochondrial calcium uniporter.

Introduction

Mitochondrial calcium uptake is critical in cellular processes such as stimulating ATP production, modulating cell metabolism by buffering cytosolic Ca²⁺, and preventing cellular Ca²⁺ overload, which would trigger apoptotic cell death, among other processes (Orrenius et al, 2003; Duchen et al, 2008; Balaban, 2009; Denton, 2009). Early studies revealed that Ca²⁺ enters the mitochondria through the mitochondrial Ca²⁺ uniporter, which is localized in the inner mitochondrial membrane (Kirichok et al, 2004; Kamer & Mootha, 2015). In vertebrates, the uniporter is a multicomponent complex that is highly selective for Ca²⁺ and consists of mitochondrial calcium uniporter (MCU) which is an ion transport pore (Baughman et al, 2011; De Stefani et al, 2011; Chaudhuri et al, 2013), MCU's negative regulatory subunit (MCUb) (Raffaello et al, 2013), metazoan‐specific essential MCU regulator (EMRE) with a single transmembrane domain (Sancak et al, 2013), two regulatory subunits mitochondrial calcium uptake 1 and 2 (MICU1 and MICU2) with canonical EF‐hand domains sensing Ca²⁺ concentrations of intermembrane space (Perocchi et al, 2010; Plovanich et al, 2013) and MICU3, a paralogue of MICU1 and MICU2, which is primarily expressed in the central nervous system (CNS) (Plovanich et al, 2013; Kamer & Mootha, 2015; Patron et al, 2018). In additional, MCU regulator 1 (MICUR1) is identified as required for MCU‐dependent mitochondrial Ca²⁺ uptake (Mallilankaraman et al, 2012a; Tomar et al, 2016).

MICU1 and MICU2 are localized in the intermembrane space and form a tight heterodimer to modulate uniporter Ca²⁺ uptake by functioning as a gatekeeper (Mallilankaraman et al, 2012b; Csordas et al, 2013; Ahuja & Muallem, 2014; Kamer & Mootha, 2014). MICU1 can interact with MICU2 and MICU3 in cells with the same cysteine site through disulfide bonds, while MICU2 and MICU3 cannot form complex as demonstrated by co‐IP (Patron et al, 2018). Additional, it has been shown that the MICU1 and MICU2 interactions also involve hydrophobic and hydrogen bonding interactions in the core domain except for disulfide bonds (Li et al, 2016; Kamer et al, 2017; Wu et al, 2019). The N‐domain topologies of MICU2 and MICU3 are different from that of MICU1, which dictate the distinct functions of these modulators. EMRE is a metazoan‐specific protein that is required for MCU‐mediated Ca²⁺ uptake (Sancak et al, 2013; Kamer & Mootha, 2015). EMRE associates with MCU through transmembrane hydrophobic interactions, while interacts with Ca²⁺ regulator MICU1–MICU2 through electrostatic pairing contributed by poly‐aspartate tail (EDDDDDD) in EMRE and polybasic sequence (KKKKR) in MICU1 (Tsai et al, 2016). It was reported that the MICU1 polybasic sequence also plays a role in the localization of MICU1 in the inner boundary membrane of mitochondria (Gottschalk et al, 2019). Several structures of MCU from fungi have been determined which demonstrate the tetrameric architectures and the Ca²⁺ selectivity mechanism (Baradaran et al, 2018; Fan et al, 2018; Nguyen et al, 2018; Yoo et al, 2018). Recently, the structure of the MCU‐EMRE complex showed that EMRE triggers the dimerization of MCU and revealed the mechanism of EMRE‐dependent MCU activation in metazoan (Wang et al, 2019). Although MCU and MCU‐EMRE structures provide important knowledge about the Ca²⁺ uptake, the MICU1–MICU2 heterodimer and its regulatory mechanism are critical missing pieces.

MICU1–MICU2 functional studies indicated that the heterodimer inhibits MCU in resting conditions and relieves inhibition with increased cytosolic Ca²⁺ concentrations (Kamer & Mootha, 2014, 2015; Patron et al, 2014; Matesanz‐Isabel et al, 2016). This Ca²⁺‐dependent inhibition‐activation uniporter switch is dictated by the Ca²⁺‐dependent interactions between MCU and MICU1 (Petrungaro et al, 2015; Phillips et al, 2019). MCU interacts with MICU1 through the aspartic acid ring in the DIME domain of MCU; however, the identity of interaction residues in MICU1 remains controversial (Paillard et al, 2018; Phillips et al, 2019). Crystal structures of MICU1 show that MICU1 is assembled as a trimer of dimers in the apo state and changes to multiple oligomers upon Ca²⁺ binding (Wang et al, 2014). The MICU2 structure displays resemblances to the MICU1 structure, except for some subtle but important differences in the N‐terminus domain (Kamer et al, 2019; Wu et al, 2019; Xing et al, 2019). In addition, Ca²⁺‐bound EF‐hands in MICU1 or MICU2 increase the inter‐helix angles of the EF‐hands and exposes the hydrophobic regions, which occurs mainly in EF‐hand 1 but not in EF‐hand 2 of both proteins (Wang et al, 2014; Wu et al, 2019). Furthermore, allosteric effects resulting from EF‐hands binding with Ca²⁺ (but not other divalent ions) are necessary for relieving the inhibition and triggering the MCU Ca²⁺ uptake (Kamer et al, 2017, 2018).

Although many MICU1–MICU2 complex studies have been documented, several fundamental questions remain to be answered. How do Ca²⁺ ions affect heterodimer conformational changes? How does EMRE interact with MICU1 in both Ca²⁺‐free and Ca²⁺‐bound conditions? How does the heterodimer activate the uniporter in high cytoplasmic Ca²⁺ concentrations and inhibit MCU under resting conditions? Herein, we present two pairs of different MICU1–MICU2 heterodimers crystal structures at 2.1 Å resolution, in which both Ca²⁺‐free and Ca²⁺‐bound EF‐hands are observed. Interface I defines electrostatic interactions crucial for the Ca²⁺‐free form, whereas interface II features hydrophobic interactions that are significant for the Ca²⁺‐bound form. Furthermore, we have identified a Ca²⁺‐enhanced alkaline groove in MICU1 that is responsible for interacting with EMRE. The discovery of such a Ca²⁺‐dependent interactions switch between MICU1‐MCU and MICU1‐EMRE is significant for understanding the conversion mechanism between different functions.

Results

Structure determination of Ca²⁺‐free and Ca²⁺‐bound MICU1–MICU2 heterodimers

We purified two MICU1–MICU2 complexes (His₆‐tagged MICU1 and tagged‐free MICU2, His₆‐tagged MICU2, and tagged‐free MICU1) under both EGTA and CaCl₂ conditions. The size‐exclusion chromatography (SEC) results indicate that the MICU1–MICU2 complex displays a monodisperse peak with a molecular weight of ~ 80 kDa under both 2 mM EGTA and CaCl₂ conditions (Fig EV1A, B, E, and F). Additionally, the MICU1–MICU2 heterodimer had a high purity with a stoichiometric ratio of ~ 1:1, as illustrated by SDS–PAGE (Fig EV1C, D, G, and H). Unfortunately, despite extensive efforts, the four aforementioned high‐purity samples did not yield crystal hits. To circumvent the difficulty of heterodimer crystallization, we introduced a flexible linker (GSGSGSGSGSGSGSGS) between the C‐terminus of MICU1 and the N‐terminus of MICU2. The MICU1–MICU2 chimaera displayed identical SEC retention volumes to the MICU1–MICU2 complex in both EGTA and CaCl₂ conditions (Fig 1A and B). The chimaera construct yielded crystal hits in the EGTA condition, which were optimized to produce diffraction‐quality crystals. Finally, the chimaera structure was determined by molecular replacement and refined at a resolution of 2.1 Å with R _work = 19.39% and R _free = 22.67%. The asymmetric unit contains two molecules of the MICU1–MICU2 chimaera, which pack in an antiparallel mode with a face‐to‐face configuration, forming a barrel‐like architecture with a large cavity in the middle (Fig 1C). Importantly, in one of the MICU1–MICU2 molecules, there are unambiguous electron densities in two EF‐hands (one in MICU1 EF‐hand 2 and the other in MICU2 EF‐hand 1; Fig EV2A and B). The two metal ions adopt a pentagonal bipyramid geometry with seven ligands and show a bidentate interaction with the Glu₁₂ (position 12 of the loop) carboxyl, which is typical for Ca²⁺‐bound EF‐hand proteins (Fig EV2C and D). This MICU1–MICU2 chimaera molecule represents a partial Ca²⁺‐bound MICU1–MICU2 heterodimer.

Figure EV1. The purified MICU1‐MICU2 complex in the presence of 2 mM EGTA and Ca²⁺ conditions.

• A, B
  Representative size‐exclusion chromatography profile of the MICU1–MICU2 complex (His₆‐tagged MICU1 and His₆‐free MICU2) in the presence of 2 mM EGTA (A) and CaCl₂ (B). In the co‐lysis purification, MICU1 was His₆‐tag labeled at the N‐terminus, while MICU2 did not have a fused tag.
• C, D
  Representative SDS–PAGE analyses of the purified MICU1–MICU2 complex with lanes corresponding to the peaks from the chromatography profiles shown in (A) and (B).
• E, F
  Representative size‐exclusion chromatography profiles of the MICU1–MICU2 complex (His₆‐tagged MICU2 and His₆‐free MICU1) in the presence of 2 mM EGTA (E) or CaCl₂ (F). In the co‐lysis purification, MICU2 was His₆‐tag labeled at the N‐terminus, while MICU1 did not have a fused tag.
• G, H
  Representative SDS–PAGE analyses of the purified MICU1–MICU2 complex with lanes corresponding to the peaks from the chromatography profiles shown in (G) and (H). The molecular weights of MICU1 and His₆‐MICU2 are too close to separate (ΔMw = 37 Da).

Figure 1. Biochemical and structural characterizations of the MICU1‐MICU2 chimaera.

• A, B
  Representative size‐exclusion chromatography profiles of the MICU1–MICU2 complex and chimaera in 2 mM EGTA (A) and 2 mM CaCl₂ (B) conditions.
• C
  Top (left) and side (middle and right) views of the overall MICU1–MICU2 chimaera structure with four molecules in a crossed arrangement. Ca²⁺‐free and partial Ca²⁺‐bound MICU1 structures are colored in pink and magenta, respectively. The Ca²⁺‐free and partial Ca²⁺‐bound MICU2 structures are colored in light blue and blue, respectively. The red, green, and orange spheres represent Ca²⁺, N‐, and C‐terminus.

Figure EV2. Structural analyses of Ca²⁺‐bound EF‐hands.

• A, B
  Electron density map of MICU1 EF‐hand 2 (A) and MICU2 EF‐hand 1 (B). Weighted 2F ₀ ‐F _c contoured at 2σ. Electron densities of the EF‐hands and Ca²⁺ ions are shown in cyan and orange meshes, respectively.
• C, D
  Ca²⁺ coordination in MICU1 EF‐hand 2 (C) and MICU2 EF‐hand 1 (D) illustrates the pentagonal bipyramid geometry. Red and green spheres represent Ca²⁺ and water, respectively.

MICU1–MICU2 heterodimers interfaces

The partial Ca²⁺‐bound and Ca²⁺‐free MICU1–MICU2 chimaeras feature an antiparallel packing mode in which EF‐hands 1 are localized in the heterodimer interface regions. The interactions of Ca²⁺‐free and Ca²⁺‐bound heterodimers are defined by two interfaces (Fig 2B). In the Ca²⁺‐free heterodimer, the interaction interface I involves extensive electrostatic interactions contributed by the α‐helixes of MICU1 EF‐hand 1 (referred to as M1_EH₁1 and M1_EH₁2) and the α‐helixes of MICU2 EF‐hand 2 pair (referred to as M2_EH_2p1, M2_EH_2p2, M2_EH_2p3, and M2_EH_2p4; detailed interaction analyses can be found in the Expanded View; Fig 2A left and middle). It is noted that Arg352 makes the greatest contribution to the interaction interface by forming hydrogen bonds with several key interfacial residues (Fig 2A left), substantiating the pull‐down experiment results from our previous report (Wu et al, 2019). Intriguingly, there is a lack of electrostatic interactions in interface II involving the MICU1 EF‐hand 2 pair α‐helixes (referred to as M1_EH_2p1, M1_EH_2p2, M1_EH_2p3, and M1_EH_2p4) and the MICU2 EF‐hand 1 α‐helixes (referred to as M2_EH₁1 and M2_EH₁2), while weak hydrophobic interactions exist (Fig 2A right; detailed interaction analyses can be found in the Expanded View).

Figure 2. Interfaces of MICU1‐MICU2 Ca²⁺‐free and partial Ca²⁺‐bound states.

1. Close‐up view of Ca²⁺‐free MICU1–MICU2 interfaces I (left and middle) and II (right). The atomic interactions at the interface I consist of mainly electrostatic (left) and weak hydrophobic (middle) interactions, while interface II mainly contains hydrophobic interactions (right).
2. MICU1 interacts with MICU2 at two major interfaces.
3. Close‐up views of the partial Ca²⁺‐bound MICU1–MICU2 interface II. The atomic interactions at interface II consist of weak electrostatic (left) and strong hydrophobic (middle) interactions. The red, green, and orange spheres represent Ca²⁺, N‐, and C‐terminus.
4. The surface‐rendered MICU2 hydrophobic pocket shows the interaction at interface II.

The partial Ca²⁺‐bound heterodimer interface I is similar to that in the Ca²⁺‐free heterodimer (Fig 2B right). In interface II, the hydrophobic interactions between hydrophobic residues Phe383, Met386 in MICU1, and the hydrophobic pocket in MICU2 EF‐hand 1 constitute the primary interaction mode (Fig 2C right), which correspond to our previous pull‐down experiments (Wu et al, 2019). In addition to the hydrophobic interaction, several hydrogen bonding or electrostatic interactions are also observed (Fig 2C left; detailed analyses can be found in the Expanded View). In many EF‐hand‐containing proteins, Ca²⁺ binding to the EF‐hand shortens the distance between Asp₁ (position 1 of the loop) and Glu₁₂ and expands the interhelical bottom angle, thus creating a hydrophobic pocket for recruiting hydrophobic residues or domains (Gifford et al, 2007). As shown by structural analyses, Phe383 inserts into the hydrophobic pocket that is created by Ca²⁺ binding to EF‐hand 1 of MICU2, just like embedding a wedge (Fig 2D). Based on the structural analyses, the interaction pattern of the partially Ca²⁺‐bound heterodimer appears to represent the “transition” conformation between the Ca²⁺‐free and Ca²⁺‐bound states.

Several key residues were previously screened by a scanning mutagenesis method using glutathione S‐transferase (GST) pull‐down experiments in both absence and presence of 2 mM CaCl₂ (Wu et al, 2019). The GST pull‐down experiments confirmed that the Glu242 in MICU1 and Arg352 in MICU2 are the crucial residues participating the interaction in the absence of Ca²⁺, whereas Phe383 in MICU1 and Glu196 in MICU2 contribute to the heterodimer formation in the Ca²⁺‐bound state (Wu et al, 2019). To further probe MICU1–MICU2 interactions, we created more mutants based on our structures (Fig EV3 and Appendix Fig S1). These mutants include MICU1_Q253A, MICU1_Q398A, MICU1_E427A, MICU2_K206A, and MICU2_Q336A. We first investigated whether the MICU1 mutants would affect the heterodimer interactions under the two Ca²⁺ conditions. Our results show that MICU2 can interact with MICU1_Q253A, MICU1_Q398A, and MICU1_E427A in both presence and absence of Ca²⁺ whether the MICU1 mutants or MICU2_WT construct were used as baits (Fig EV3A–C, Appendix Fig S1C–E). But MICU1_Q253A and MICU1_E427A mutants reduced the interactions in the Ca²⁺‐free condition compared with that in the presence of Ca²⁺ when GST‐MICU1 mutants were used as baits (Appendix Fig S1C and E). Next, we analyzed the effects of MICU2 mutants in Ca²⁺‐free and Ca²⁺‐bound heterodimers. GST‐MICU2_K206A pulled down MICU1_WT in both absence and presence of Ca²⁺, whereas MICU2_Q336A mutant displayed weakened interaction with MICU1 in these conditions (Fig EV3D and E). The reverse pull‐down experiments also yielded the same conclusions when MICU1_WT was used as the bait (Appendix Fig S1A and B). The results indicate that these residues might participate in the MICU1–MICU2 interaction in the absence of Ca²⁺, but are not the key sites.

Figure EV3. Specific MICU1‐MICU2 interaction sites in both EGTA and Ca²⁺ conditions were identified by GST pull‐down assays.

• A–C
  MICU2 interacts with MICU1_Q253A (A), Q398A (B), and E427A (C) in both conditions.
• D, E
  MICU2_K206A (D) and Q336A (E) mutants do not disturb the interaction of MICU2 with MICU1 in either condition.

Source data are available online for this figure.

Conformational changes in the Ca²⁺‐free and Ca²⁺‐bound heterodimers

We aligned the backbone atoms between the two states structures for better expounding the discrepancy of two structures. The overall structures of the two heterodimers (with and without Ca²⁺ bound) were similar with RMSD value of 2.045 Å. Evident from the overall structure alignment, the EF‐hand 1 of MICU2 and EF‐hand pair 2 of MICU1 display significant conformational changes (Fig 3A). It is evident that Ca²⁺ binding serves as a conformational switch to govern the EF‐hand close‐to‐open conformational transition by altering the interhelical angle (Gifford et al, 2007). Upon Ca²⁺ binding, the EF‐hand 1 α‐helixes in MICU2 rotate from an antiparallel arrangement in the apo state to an approximately perpendicular mode with an interhelical angle of 90° (Fig 3A middle, Appendix Fig S2B and Movie EV1). The EF‐hand pair 2 of MICU1 in the Ca²⁺‐bound state rotates toward MICU2 compared to the Ca²⁺‐free state (Fig 3A right and Movie EV1). On the other hand, the MICU1 EF‐hand 2, which coordinates Ca²⁺, maintains the same interhelical angle compared to that without the occupied Ca²⁺ (Appendix Fig S2A).

Figure 3. Structural alignments of MICU1‐MICU2 with different numbers of Ca²⁺ ions bound.

1. Structural alignment of Ca²⁺‐free and partial Ca²⁺‐bound MICU1–MICU2 heterodimers, as well as critical substructures with conformational changes indicated by numbered dashed circles. Middle and right, close‐up view of the MICU2 EF‐hand 1 (middle, circle 1) and MICU1 EF‐hand pair 2 (right, circle 2) conformational changes upon Ca²⁺‐binding. The important α‐helixes are highlighted as cylinders. The red sphere represents Ca²⁺.
2. Structural alignment of Ca²⁺‐free, partial Ca²⁺‐bound, and fully Ca²⁺‐bound MICU1–MICU2 model heterodimers and important substructures with conformational changes are indicated by numbered dashed circles. Three different MICU1 and MICU2 molecules are individually shown in similar colors, and the colors changing from light to dark correspond to an increase in Ca²⁺ numbers. Circles 1, 2, and 3: close‐up views of MICU2 EF‐hand 1, MICU1 EF‐hand pair 2, and MICU1 EF‐hand 1 conformational changes upon Ca²⁺‐binding, respectively. Important α‐helixes are highlighted as cylinders. The red sphere represents Ca²⁺.
3. Conformational changes of the alkaline groove between T4L fusion Ca²⁺‐free MICU1–MICU2 and partial Ca²⁺‐bound MICU1–MICU2 heterodimers. Inset, KRKK residues in the alkaline groove rotate upon Ca²⁺ binding.
4. Electrostatic surface potential of the alkaline groove in partial Ca²⁺‐bound structure. The gradient shown is from −3 (acidic, red) to +3 (basic, blue) kT/e.

Next, we calculated the buried surfaces in the two structures. The binding surface of the partial Ca²⁺‐bound heterodimer (~ 1,426 Å²) is larger than that of the Ca²⁺‐free heterodimer (~ 1,119 Å²), which is consistent with our structural analysis. The calculation of the buried surface area provides direct evidence to demonstrate that Ca²⁺ binding to EF‐hands enhances the compactness and, by extension, stability of the complex. Additionally, the Ca²⁺‐bound lobes extend ~ 58 Å in width, which is narrower than that in the Ca²⁺‐free lobes, which is ~ 64 Å (Appendix Fig S3); this result substantiates the notion that Ca²⁺ binding significantly tightens the heterodimer.

Our partial Ca²⁺‐bound structure features both Ca²⁺‐free and Ca²⁺‐bound characteristics that have helped shed light on the transition from the apo state to the Ca²⁺‐bound state. Next, we constructed a fully Ca²⁺‐bound model. The extensive structural information from numerous Ca²⁺‐bound EF‐hands in the literatures help us to readily build this model using the partial Ca²⁺‐bound heterodimer as a starting template. The comparison of the three heterodimers (apo, partial and fully Ca²⁺‐bound) shows that the conformational changes primarily occur in the EF‐hand domains, particularly in EF‐hands 1 in both MICU1 and MICU2, which expose the hydrophobic pockets upon Ca²⁺ binding (Fig 3B and [Link], [Link]). Furthermore, the EF‐hand pair 2 in MICU1 and the EF‐hand 1 in MICU2 rotate considerably in the partially or fully Ca²⁺‐bound heterodimers that may have implications in regulating the Ca²⁺ uptake in the uniporter (Fig 3A and B).

The alkaline groove of MICU1 mediates the interaction with EMRE

The polybasic sequence (KKKKR) of MICU1 mediates the interaction with EMRE (Tsai et al, 2016); however, this sequence in our structure is unresolved, likely due to its flexibility. To make the N‐terminus of the MICU1 construct more rigid, we introduced a fusion protein (T4 lysosome) in the MICU1 N‐terminus and determined the new T4L‐MICU1–MICU2 complex structure in the Ca²⁺‐free state at a resolution of 3.3 Å. The overall structure is similar to our previous complex devoid of the T4L fusion protein in which the two molecules of the heterodimers are assembled in a face‐to‐face configuration with a twofold symmetry (Appendix Fig S4A and B). Fortunately, the polybasic sequence (KKKKR) is now observed in MICU1, which is a short α‐helix (Appendix Fig S4C).

EMRE likely functions to anchor the MICU1–MICU2 heterodimer to prevent the dissociation from the channel with increasing Ca²⁺ concentrations in the cytosol (Petrungaro et al, 2015; Tsai et al, 2016). To better understand the spatial relationship among various components in the uniplex, we performed charge potential analyses of both the Ca²⁺‐free and partial Ca²⁺‐bound heterodimers. MICU1 α‐helixes (referred to as M1_CH1 and M1_CH2) are located at the protein edge and display an increased positive charge in the presence of Ca²⁺ (Figs 3D and EV4). This charge distribution is reflected by a large positively charged groove consisting of M1_CH1, M1_CH2, and the polybasic α‐helix (Appendix Fig S4C and D). The electron positivity of Ca²⁺ attracts the acidic residues in the EF‐hand loop toward the central loop area, whereas these acidic residues are repelled in the apo state. Because of the Ca²⁺ repulsion and lacking of attraction of negatively charged residues, the side chains of the KRKK motif (Lys340, Arg347, Lys350, and Lys351) in M1_CH1 helix and Arg325 in M1_CH2 are repelled toward the groove upon Ca²⁺ binding (Fig 3C). Considering that EMRE has the negatively charged poly‐aspartate tail, this alkaline groove becomes the most reasonable place for MICU1 and EMRE to interact. These conformational changes may provide a favorable condition for the interaction between MICU1 and EMRE. Upon Ca²⁺ binding, the KRKK motif in MICU1 further strengthens the interaction between EMRE and MICU1 by enhancing the positively charged areas to compensate for the loss of the interaction with the channel.

Figure EV4. Electrostatic representation of the alkaline groove in Ca²⁺‐free and partial Ca²⁺‐bound structure.

• A, B
  Side (A) and top (B) views of electrostatic surface potential of the alkaline groove in the Ca²⁺‐free MICU1 structure.
• C, D
  Side (C) and top (D) views of electrostatic surface potential of the alkaline groove in the partial Ca²⁺‐bound MICU1 structure.

Data information: The gradient shown is from 3 (acidic, red) to +3 (basic, blue) kT/e.

To confirm this speculation, we synthesized an EDDDDDD peptide to examine its interaction with MICU1 using microscale thermophoresis (MST) for understanding the process of Ca²⁺ signal from modulators to channel. We designed three mutants, including MICU1_poly_KE that replaced the polybasic motif with EEEEE sequence, MICU1__KRKKEEEE that replaced four curial residues (Lys340, Arg347, Lys350, and Lys351) with Glu, and MICU1_R325E. Our results show that the EMRE peptide could indeed interact with MICU1 in both conditions, although this interaction is weaker in the presence of EGTA (Fig 4C). Furthermore, the three MICU1 mutants all lost the interactions with the peptide in the presence of EGTA (Fig 4A). MICU1__KRKKEEEE perturbs the interaction with the EMRE peptide in the presence of Ca²⁺, while MICU1_poly_KE and MICU1_R325E retain the interactions with the EMRE peptide, similar to the wild‐type protein (Fig 4B). The MST results are in accordance with the structural analysis showing that the alkaline groove mediates the EMRE and MICU1 interaction and also enhances this electrostatic interaction in the presence of Ca²⁺. As revealed in the structure, KRKK motif and Arg352 rotate toward the groove interior to strengthen the EMRE‐MICU1 interaction and, as a result, these mutants in Ca²⁺‐binding state had little effect compared with the Ca²⁺‐free condition. Furthermore, the isothermal titration calorimetry (ITC) curve demonstrates that the EMRE peptide interacts with MICU1 at an approximate stoichiometric ratio of 1:1 (Fig 4D), in line with the ~ 100 kDa migration of MICU1–MICU‐EMRE complex in the uniporter assembly experiments in vivo (Opalinska & Janska, 2018).

Figure 4. The characterization of interactions between EMRE peptide and MICU1 in vitro and EMRE peptide competition in cells.

1. The binding of fluorescently (GFP) labeled MICU1 and mutants to EMRE peptide shown by MST in the presence of EGTA.
2. The binding of fluorescently (GFP) labeled MICU1 and mutants to EMRE peptide were analyzed with MST in the presence of Ca²⁺.
3. Comparison of binding of MICU1 and EMRE peptide in both conditions.
4. MICU1–MICU2 heterodimer interacts with the EMRE peptide at a 1:1 ratio.
5. Carbachol (CCh)‐induced representative traces of mitochondrial calcium uptake in HEK cells treated with mt‐EMRE and scramble peptides. Solid lines, mean; shaded regions, ± SEM, n = 3.
6. Resting mitochondrial Ca²⁺ levels in mt‐EMRE and scramble peptides treated cells, evaluated via ratiometric imaging of the mitochondrial targeted mt‐riG6m. Mean ± SD, Student's t‐test, **P < 0.01, n = 3.
7. Mean mitochondrial Ca²⁺ peaks response to 100 μM CCh. Mean ± SD, Student's t‐test, ***P ≤ 0.001, n = 3.
8. Representative [Ca²⁺]_c clearance traces of U2OS cells treated with PBS and mt‐EMRE peptide.
9. Quantification of mitochondria Ca²⁺ uptake rate in response to 100 μM histamine. Mean ± SD, Student's t‐test, n.s., not significant, n = 4.

Data information: Error bars for some points are too small to be visible.

Encouraged by the observation of the interaction in vitro, we next asked the question of whether this peptide would compete with the endogenous EMRE in cells. We synthesized a mitochondria‐targeting EMRE peptide (mt‐EMRE peptide) that contains a supplemental mitochondria‐targeting sequence at the N‐terminus of the peptide as previously reported (Cerrato et al, 2015). To detect the location of EMRE peptide, we used the mt‐EMRE peptide conjugated with fluorescence probe 5(6)‐carboxytetramethylrhodamine (TAMRA). Cells were transiently transfected mEmerald‐Mito‐7 that specifically targets the mitochondrial as a mitochondrial marker (Planchon et al, 2011) and treated with TAMRA‐mt‐EMRE peptide for fluorescence microscopy analysis. The merged image and quantitative co‐localization analysis illuminate the co‐localization between our peptide and mEmerald‐Mito‐7, demonstrating that mt‐EMRE peptide can be imported into mitochondria to exert its function (Appendix Fig S5). We further used the genetically encoded ratiometric Ca²⁺ indicator mt‐riG6 m to monitor the Ca²⁺ mitochondrial matrix signals (Li et al, 2020). Clearly, the fluorescence ratio of mt‐riG6 m in cells treated with the mt‐EMRE peptide was stronger than that in cells treated with the scramble peptide in resting conditions (Fig 4E and F). This result demonstrates that the mt‐EMRE peptide effectively competed with endogenous EMRE which probably weakened the interaction between the MICU1–MICU2 complex and the channel, resulting in the reduction of threshold in resting conditions. Moreover, in mt‐EMRE treated cells, the [Ca²⁺]_m peak evoked by carbachol was markedly greater, indicating elevated Ca²⁺ accumulation in mitochondria due to reduction of threshold (Fig 4E and G). With this in mind, we measured extra‐mitochondrial Ca²⁺ clearance using cytoplasmic Ca²⁺‐sensitive fluorescent dye. [Ca²⁺]_c measurements showed a marked enhancement of the transient rise evoked by stimulation with 100 μM histamine in the presence of mt‐EMRE peptide, similar to the control cells (Figs 4H and EV5A). The kinetics of Ca²⁺ clearance in both conditions showed negligible differences (Figs 4I and EV5B). In conclusion, these results demonstrate that mt‐EMRE peptide reduced uniporter threshold, resulting in consequently Ca²⁺ accumulation in mitochondria under resting and stimulation conditions. Interestingly, the peptide did not affect the kinetics of mitochondrial Ca²⁺ uptake in response to histamine.

Figure EV5. Mt‐EMRE peptide cytoplasmic Ca²⁺ clearance assays in HeLa cells.

1. Representative [Ca²⁺]_c clearance traces of HeLa cells treated with PBS and the mt‐EMRE peptide.
2. Quantification of mitochondria Ca²⁺ uptake rate in response to 100 μM histamine. Mean ± SD, Student's t‐test, n.s., not significant, n = 3.

Discussion

We have determined structures of two crystallographically independent MICU1–MICU2 heterodimers with barrel‐like configuration (Fig 1C), providing new insights into mitochondrial Ca²⁺ uptake modulation. In addition to the “barrel‐like” arrangement, the “linear” packing mode of the two complexes is observed (Appendix Fig S6). The adjacent MICU2 molecules in the middle are assembled in a back‐to‐back mode (Appendix Fig S6). In fact, the crystal packing shows that the MICU1 monomer is not within an interaction distance to associate with another MICU1 monomer, in neither a back‐to‐back nor a face‐to‐face configuration. Our structure appears to represent a “transition state” between apo and holo structures, which is informative since the hybrid structure simultaneously displays both Ca²⁺‐free and Ca²⁺‐bound characteristics (Fig 1C). Interfaces I and II seem to play different roles in prompting interaction modes; they are critical for the Ca²⁺‐free and Ca²⁺‐bound MICU1–MICU2 association through electrostatic and hydrophobic interactions, respectively (Fig 2). Recently, a human Ca²⁺‐free MICU1–MICU2 complex structure was reported, which displays the face‐to‐face configuration with salt bridge interaction sites between MICU1 and MICU2 (Park et al, 2020), in agreement with our chimaera structure. The superimposition between the Ca²⁺‐free complex and our Ca²⁺‐free chimaera structure shows negligible discrepancy, yielding RMSD value of only 1.28 Å (Appendix Fig S7). Thus, the previous pull‐down results and the recently reported Ca²⁺‐free MICU1–MICU2 complex structure substantiate the notion that our chimaera heterodimer does not appear to introduce any biases (Wu et al, 2019; Park et al, 2020).

We introduced T4L to afford some stability of polybasic KKKKR sequence in MICU1, and we observed at least one of its main conformations. Although T4L may affect the “native” position of the N‐terminus since its presence would reduce the dynamic range of conformational flexibility, it should not fundamentally affect the main orientation of KKKKR helix which protrudes out from the main structure and into the open space (Appendix Fig S4A–C). If KKKKR fragment were closely associated with the MICU1–MICU2 complex (i.e., not in the open space), its electron density would not be so poor and some (if not all) of its structure would have been observed. This standalone helix and other basic residues (KRKK motif and Arg325) constitute the alkaline groove that mediates the interaction of MICU1‐EMRE. The alkalinity of the groove is increased in the presence of Ca²⁺ which facilitates MICU1‐EMRE interaction, and the disruption of this interaction will lead to [Ca²⁺]_m disorder. We proceeded with assessing the uniporter threshold of Ca²⁺ uptake using mt‐EMRE peptide in competition assays. The mt‐EMRE peptide can effectively compete with the endogenous EMRE and destabilize the intrinsic EMRE‐MICU1 association, resulting in the reduction of threshold in resting conditions (Fig 4E and F), whereas the kinetic of mitochondria Ca²⁺ uptake was not affected (Figs 4I and EV5B).

Very recently, a cryo‐EM structure of the uniporter core complex (MCU‐EMRE‐MICU1–MICU2) was reported, revealing a rather unexpected stoichiometry of 4:4:1:1 (Fan et al, 2020). The super‐complex structures show that MICU1 directly blocks the Ca²⁺ pore in resting conditions through its arginine/lysine region interacting with aspartic acid ring in MCU channel entrance, while MICU1 dissociates from MCU and, together with MICU2, drifts to the side of channel in the active condition (Fan et al, 2020). Although the interaction region of EMRE‐MICU1 was unresolved in the super‐complex structure, the general direction of the interaction region points to our proposed alkaline groove pocket, providing support to our result. The interaction of MICU1‐MCU was significantly weakened in the presence of Ca²⁺ (Tsai et al, 2016; Phillips et al, 2019; Fan et al, 2020), while MICU1‐EMRE interaction was promoted in this condition. Thus, we put forward an interactions switch model that propels Ca²⁺ signal transmission from the regulator to the channel (Fig 5A). While weakly interacting with EMRE, MICU1 mainly associates with MCU in resting conditions to gate the Ca²⁺ channel and switches to dominantly binding to EMRE with the increasingly positive charge of alkaline groove at high Ca²⁺ levels, releasing Ca²⁺ into the mitochondria matrix (Fig 5A). In this switch scenario, the MICU1–MICU2 heterodimer primarily undergoes two allosteric conformational changes that (i) induces the hydrophobic pocket of MICU2 to attract Phe383 of MICU1 upon Ca²⁺ binding which may result in MICU1 dissociating from MCU and (ii) promotes the positive charge of the alkaline groove which transmits the Ca²⁺ signal to EMRE by tightening the MICU1‐EMRE interaction. The Ca²⁺‐dependent interactions switch model readily complements the transition mechanism between gatekeeping and cooperative functions, reflecting the conferred regulatory and scaffold functions of EMRE. Although the two super‐complexes structural characterizations can intuitively explain the gatekeeping and activation mechanisms to some extent (Fan et al, 2020), they have raised an important question—why there are varying and multiple copies of EMRE molecule in uniplex?

Figure 5. Model of the uniporter regulatory mechanism.

1. Model of MCU Ca²⁺ uptake modulatory mechanism by interaction switch. In resting conditions, MICU1–MICU2 heterodimer associates with MCU pore to prevent Ca²⁺ transporting. In the stimulated condition, MICU1–MICU2 heterodimer enhances the interaction with EMRE by increasing the positive charge of alkaline groove and disassociates form MCU, resulting in opening the gate and activating the Ca²⁺ channel. Inset shows modeled interaction between the EMRE C‐terminal poly‐aspartate tail and MICU1 alkaline groove.
2. Probability model for MICU1–MICU2 heterodimer to recognize and interact with the channel. As the number of EMRE molecules increases, the interaction sites between MICU1–MICU2 heterodimer and channel increase. Consequently, the probability for MICU1 to recognize and interact with the channel is increased, enhancing the regulatory function of MICU1–MICU2.

Recently, Payne et al (2020) reported that there exists variable number of EMRE per channel (1–4) in vivo and increasing number of EMRE enhances the gatekeeping function. According to the super‐complex structures (Fan et al, 2020), the tetrameric MCU channel possesses fourfold symmetry which is flat in the pore entrance and accommodates only one MICU1–MICU2 heterodimer. In the case of 4 molecules of EMRE, the heterodimer can locate in the any of the four positions by rotating 90° around the channel center, yielding equivalent interaction environment. When there are less copies of EMRE, the fourfold symmetry is perturbed and the heterodimer would be “guided” to interact with where EMRE(s) resides, as a result of the electrostatic interaction between the MICU1's alkaline groove and EMRE's poly‐aspartate tail which is weaker in resting conditions and stronger in activated state. Thus, the number of EMRE molecules corresponds to the number of probable locations of MICU1–MICU2 heterodimer. As the number of EMRE increases, the probability of recognition and interaction for MICU1–MICU2 heterodimer to associate with the channel also increases, which can improve the gatekeeping efficiency (Fig 5B) and is in agreement with the literature (Payne et al, 2020). Regardless the stoichiometric ratio of channel and EMRE, EMRE always tethers MICU1–MICU2 to MCU, maintaining the interaction switch to accomplish the opposite function transition.

In conclusion, we have determined the MICU1–MICU2 heterodimer complex structures in two Ca²⁺ states and revealed conformational changes from the Ca²⁺‐free condition to the Ca²⁺‐bound condition. The Ca²⁺‐dependent interactions switch has solidified the transition mechanism between opposite functions—gatekeeping and activation. Our results establish a framework for understanding the regulatory mechanism of MICU1–MICU2 and provided insights into EMRE's guiding and tethering function.

Materials and Methods

Protein cloning, expression, and purification

MICU1 (BC004190.2, residues 97–444) and MICU2 (BC031089.1, residues 85–406) were connected by a flexible linker (GSGSGSGSGSGSGSGS). The MICU1–MICU2 chimaera gene was cloned into the pET‐28b(+) plasmid with an N‐terminus His₆‐tag and transformed into the BL21(DE3) Escherichia coli strain. For the T4L‐MICU1–MICU2 chimaera, the same flexible linker was used between MICU1 and MICU2. The T4 lysozyme gene was cloned into the pET‐28b(+) plasmid between NcoI and NdeI restriction enzyme sites with an N‐terminus His₆‐tag. MICU1–MICU2 positioned after T4L was cloned between the NdeI and XhoI sites. The recombination proteins were expressed in 1 l Luria Broth (LB) medium at 37°C until the OD_600 nm reached ~ 0.8, and then, the cultures were induced with isopropyl β‐D‐1‐thiogalactopyranoside (IPTG) at the final concentration of 0.5 mM. Cells were cultivated at 16°C for 20 h before harvesting. The harvested pellet was resuspended in buffer I (20 mM HEPES pH 8.0, 300 mM NaCl, 0.3% Triton X‐100, and 1 mM phenyl methane sulfonyl fluoride with or without 2 mM CaCl₂) and sonicated for lysis. After centrifugation for 30 min at 18,000 g, the supernatant was loaded onto a Ni‐NTA gravity column pre‐equilibrated with buffer I. The protein was washed with buffer II (20 mM HEPES pH 8.0, 300 mM NaCl, and 30 mM imidazole with or without 2 mM CaCl₂) for 100 ml. Finally, the protein was eluted in buffer III (20 mM HEPES pH 8.0, 300 mM NaCl, and 500 mM imidazole with or without 2 mM CaCl₂). The eluted protein was concentrated until the total volume reached 1 ml and was then purified using SEC. Before running the ÄKTA purifier, the SEC column was equilibrated with buffer IV (20 mM HEPES pH 8.0, 300 mM NaCl, and 2 mM DTT with either 2 mM EGTA or 2 mM CaCl₂). The protein was detected at an absorption wavelength of 280 nm, and the sample was collected at the corresponding retention volume.

Crystallization and X‐ray data collection

The two proteins (MICU1–MICU2 and T4L‐MICU1–MICU2) were purified in buffer containing 20 mM HEPES pH 8.0, 300 mM NaCl, 2 mM DTT, and 2 mM EGTA and were concentrated to 4 mg/ml for crystallization. Crystallization trials were performed using the hanging‐drop vapor‐diffusion method at 20°C. The MICU1–MICU2 crystals were grown in 4 μl drops consisting of 2 μl protein and 2 μl reservoir solution containing 0.2 M ammonium chloride and 5% (w/v) polyethylene glycol (PEG) 3350. For the T4L‐MICU1–MICU2 protein, the crystallization condition was 2% tacsimate pH 5.0, 0.1 M sodium citrate pH 5.0, and 12% (w/v) PEG 3350. The drops were sealed in a 24‐well plate and equilibrated against 0.5 ml reservoir solution. Crystals of both proteins grew to diffraction size in 1 week and were cryo‐protected (by supplementing 30% glycerol in the case of MICU1–MICU2 and 20% glycerol in the case of T4L‐MICU1–MICU2 in their corresponding reservoir solutions) before flash freezing in liquid nitrogen. The diffraction data were collected at a wavelength of 0.979 Å under 100 K in the BL19U beamline at the Shanghai Synchrotron Radiation Facility (SSRF), and each dataset consisted of 720 frames with a 0.5° oscillation per frame(Wang et al, 2018).

Structure determination and refinement*

The X‐ray diffraction datasets of both structures were indexed, integrated, and scaled using the program HKL2000 (Minor et al, 2006). Data quality was subsequently assessed using phenix.xtriage (Adams et al, 2010). For the MICU1–MICU2 structure, the model was determined by molecular replacement (MR) using phenix.phaser (Adams et al, 2010) and MICU1 (PDB code: 4NSC) and MICU2 (PDB code: 6IIH) to simultaneously search. Although one of the heterodimer molecules in the asymmetric unit was successfully located, only MICU2 was found in the second heterodimer. Further MR efforts located the N‐lobe of MICU1. At this time, we resorted to the superimposition of the first MICU1–MICU2 molecule with the partial structure of the second molecule, which finally enabled the localization of the missing C‐lobe. For the apo T4L‐MICU1–MICU2 complex structure, the model was determined by MR simultaneously using the Ca²⁺‐free MICU1–MICU2 molecule and T4L probes. Relevant amino acid side chains were rebuilt based on the 2F_obs‐F_calc and F_obs‐F_calc maps using Coot (Emsley et al, 2010). The structures were refined by phenix.refine (Adams et al, 2010). Detailed data collection and structure refinement statistics are summarized in Table 1.

Table 1.

Data collection and refinement statistics for MICU1‐MICU2 heterodimers

Crystal name	MICU1–MICU2 (6LB7)	T4L‐MICU1–MICU2 (6LB8)
Data collection
Space group	P212121	P212121
Unit cell
a, b, c (Å)	a = 85.481	a = 79.339
	b = 107.721	b = 133.296
	c = 170.381	c = 178.023
α, β, γ (°)	α = 90	α = 90
	β = 90	β = 90
	γ = 90	γ = 90
Wavelength (Å)	0.978	0.978
R merge (%)a	0.1246 (1.192)	0.1947 (0.505)
I/σ/a	10.87 (1.35)	15.12 (4.43)
CC1/2	0.984 (0.775)	0.958 (0.937)
Completeness (%)a	0.98 (1.00)	0.99 (1.00)
Redundancya	13.0 (12.3)	13.0 (12.9)
Refinement
Resolution range (Å)a	47.3–2.101 (2.176–2.101)	49.06–3.283 (3.401–3.283)
Unique reflectionsa	92,168 (8,058)	29,560 (2,680)
R work/R free (%)a	(0.1939/0.2267) (0.2425/0.2451)	0.1835/0.2494 (0.2185/0.3460)
No. atoms
Protein	9,893	12,694
Ligand/ion	2	0
B‐factors
Protein	35.3	39.88
Ligand or (solvent)	28.50
R.m.s deviations (Å)
Bond lengths (Å)	0.021	0.014
Bond angles (°)	1.29	1.29

^aHigh‐resolution shell is shown in parentheses.

*Note that the MICU1–MICU2 complex structures published here were deposited by the authors on PDB on November 13, 2019 [Accession numbers 6LB7 and 6LB8]. The related structure published in IUCr J. (Park et al, 2020) was deposited on PBD on November 24, 2019 [Accession number 6LE5].

Modeling of the fully Ca²⁺‐bound structure

The fully Ca²⁺‐bound structural modeling of the MICU1–MICU2 complex was achieved using our partial Ca²⁺‐bound heterodimer structure (PDB code: 6LB7) and relevant Ca²⁺‐bound monomer structures. The partial Ca²⁺‐bound MICU1 was superimposed with the Ca²⁺‐bound MICU1 monomer (PDB code: 4NSD) by fixing the C‐lobe (Ca²⁺‐bound lobe) of the partial Ca²⁺‐bound MICU1. The partial Ca²⁺‐bound MICU2 was superimposed with the fully Ca²⁺‐bound MICU2 monomer (PDB code: 6IIH, chain B) by fixing the N‐lobe (Ca²⁺‐bound lobe) of the partial Ca²⁺‐bound MICU2. The fully Ca²⁺‐bound heterodimer was built by replacing the partial Ca²⁺‐bound MICU1 and MICU2 monomers with the fully Ca²⁺‐bound MICU1 and MICU2 monomers, respectively. The minimization of fully Ca²⁺‐bound model was achieved using Amber. The fully Ca²⁺‐bound system was solvated in the TIP3P water shell and neutralized with Na⁺. The energy minimization was performed in two stages. In the first stage, only water molecules were minimized with the protein and Ca²⁺ fixed using 50 kcal/(mol·Å²) restraints. In the second stage, the entire system was minimized with maximum minimization cycle set to 200,000 steps. The first 1,000 steps were performed using steepest descent algorithm, followed by the conjugate gradient algorithm to reach minimal energy. The energy minimization process converged after 31,173 steps.

GST pull‐down assays

The wild type and mutants of GST‐MICU1 and GST‐MICU2 were cloned into the pEGX‐6p‐1 vector between the BamHI and XhoI sites with an N‐terminus GST‐tag. Taking GST‐MICU1 wild type pulled down MICU2 mutants (referred to as MICU2^mut) as an example, GST‐MICU1 and MICU2^mut pellet were co‐lysed in buffer I (see the protein purification section) with 2 mM EGTA or CaCl₂. Cell lysis and centrifugation procedures were the same as described above in the protein purification section. The supernatant was incubated with GST magnetic beads for 2 h at 4°C, and the beads were washed with buffer V (20 mM HEPES pH 8.0, 300 mM NaCl, and 2 mM EGTA or CaCl₂) three times. The GST beads were resuspended in loading buffer, and SDS–PAGE was performed for analysis.

Isothermal titration calorimetry

The ITC assay was performed at 25°C using a VP‐ITC instrument (GE MicroCal). The EMRE peptide was dissolved in buffer containing 20 mM HEPES pH 7.5, 300 mM NaCl, and 2 mM DTT at a final concentration of 1 mM. The MICU1–MICU2 complex was purified using the same buffer as the EMRE peptide and concentrated to 50 μM. EMRE peptide aliquots (2 μl each) were injected into the sample cell (300 μl) containing the MICU1–MICU2 complex for 16 times, with time intervals of 300 s. The heat absorbed/released integration analysis was performed using the MicroCal Origin software package.

Microscale thermophoresis

Microscale thermophoresis experiments were performed as previously reported (Wienken et al, 2010). MICU1 (56–444 aa) was fused with GFP at the C‐terminus and purified in buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 2 mM DTT, and 2 mM EGTA or CaCl₂. Before the experiments, the MICU1‐GFP concentration was adjusted to achieve an optimal fluorescence range of 400–1,500 fluorescence units at the wavelength λ = 470 nm. The EMRE peptide was subjected to a serial twofold dilution starting from 1 mM for 12 times using the same buffer as MICU1‐GFP. Each peptide gradient (10 μl) was mixed with MICU1‐GFP (2× optimal concentration, 10 μl) at an equal volume. After waiting 5 min, the mixture was added to 12 capillaries. The thermophoresis experiments were measured using the LED excitation source at λ = 470 nm, and data analyses were performed using the software from the manufacturer.

Peptide competition assays in cells

To obtain the EMRE peptide that would target mitochondria, we introduced the following sequence to the N‐terminus of the EMRE peptide: d‐Arg–Dmt–Orn–Phe (d‐Arg, d‐arginine; Dmt, 2,6‐dimethyl‐l‐tyrosine; Orn, ornithine, referred to as mt‐EMRE) (Cerrato et al, 2015). We also conjugated the probe 5(6)‐carboxytetramethylrhodamine (TAMRA) in the mt‐EMRE peptide to detect the localization of EMRE peptide (referred to as TAMRA‐mt‐EMRE). HEK cells were grown on 24 mm coverslips. MEmerald‐Mito‐7 plasmid encoding for a mitochondria‐targeting sequence as mitochondrial marker (mEmerald‐Mito‐7 was a gift from Michael Davidson; addgene plasmid # 54160) (Planchon et al, 2011) and mito‐RFP‐linker‐GCaMP6 m (referred to as mt‐riG6 m) (Li et al, 2020) were transfected using the lipofectamine 3000 transfection reagent (Invitrogen) for peptide localization and competition assays, respectively. Six hours after the transfection, cells were treated with 20 μM TAMRA‐mt‐EMRE (for mitochondrial localization assay) or 100 μM mt‐EMRE and scramble (MARAMRRAFQK) peptides (for competition assay) and re‐incubated for 24 h in complete medium (DMEM supplemented with 10% FBS and 1% penicillin and streptomycin) at 37°C, 5% CO₂. Before imaging, cells were washed three times with a solution containing 20 mM HEPES pH 8.0, 10 mM NaCl, 140 mM KCl, and 1 mM MgCl₂. The dishes containing 1 ml solution were imaged using a ZEISS observer Z1 microscope with a 40× oil immersion objective lens (NA 1.3) as previously described (Li et al, 2020). MEmerald‐Mito‐7 and TAMRA fluorescence experiments used the following excitation/emission filter settings: 457–484 nm_Ex/580–678 nm_Em for mEmerald‐Mito‐7; 539–557 nm_Ex/580–678 nm_Em for TAMRA. To obtain mt‐riG6 m ratio signal, filters for GCaMP6 m (457–484 nm_Ex; 497–525 nm_Em) and RFP (539–557 nm_Ex; 580–678 nm_Em) were used. After establishing 40 s baseline in 1 mM Ca²⁺ bath solution, 100 μM carbachol was added and images were taken sequentially for 6 min. The corresponding fluorescence readings from regions of interest were exported from the Zen software and imported into Matlab 2014a (The MathWorks, Natick, MA, USA) to calculate the F _GCaMP/F _RFP ratio. Ratio data and statistics were then plotted with Prism7 software. At least three independent repeats were performed.

Measurement of cytoplasmic Ca²⁺ clearance

Measurement of cytoplasmic Ca²⁺ concentration was done by following the published protocol (Mallilankaraman et al, 2012b). U2OS or HeLa cells were grown on glass‐bottom imaging dishes. After culturing for 12 h, cells were treated with 100 μM mt‐EMRE peptides and re‐incubated for 24 h. Cells were washed with PBS three times and loaded with 5 μM fluo‐4 probe in PBS for 30 min. In order to facilitate entry of the fluo‐4 probe into the cells and avoid aggregation, 0.005% pluronic F127 was added. The dishes were mounted in an objective table at 37°C and imaged every 2 s at 488 nm excitation using a 60× oil objective (Nikon, A1). After establishing 50 s baseline, 100 μM histamine was added and image was taken sequentially for 3 min. Images were analyzed by NIS‐Elements analysis software.

Author contributions

WW, QS, JZ, and ZJ came up with the concept and designed the methodology. QS and WW collected data and determined the structures. WW and RZ performed the ITC, MST, and GST pull‐down interaction experiments. YW, WW, and ZQ preformed the cells experiments. All authors analyzed data. WW, JZ, and ZJ wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Axpanden View Figures PDF

Movie EV1

Movie EV2

Source Data for Expanded View and Appendix

Review Process File

The EMBO Journal (2020) 39: e104285

Data availability

The structures produced in this study are available in the following database: protein atomic coordinates data: PDB 6LB7 and 6LB8 (https://www.rcsb.org/structure/6lb7 and https://www.rcsb.org/structure/6lb8).