Variable Assembly of EMRE and MCU Creates Functional Channels with Distinct Gatekeeping Profiles

Summary

MCU is a Ca²⁺-selective channel that mediates mitochondrial Ca²⁺ influx. The human channel contains tetrameric pore-forming MCU, regulatory subunits MICU1/2, and EMRE that is required both for channel function and MICU1/2-mediated Ca²⁺ regulation. A structure of MCU with EMRE revealed a 4:4 stoichiometry, but the stoichiometry in vivo is unknown. Expression of tagged EMRE and MCU at a 1:10 ratio in cells lacking EMRE and MCU restored channel activity but not full channel gatekeeping. Increasing EMRE expression enhanced gatekeeping, raising the cytoplasmic Ca²⁺ concentration ([Ca²⁺]_c) threshold for channel activation. MCU-EMRE concatemers creating channels with 1EMRE:4MCU restored Ca²⁺ uptake in cells, whereas cells expressing concatemers that enforced a 4EMRE:4MCU stoichiometry demonstrated enhanced channel gatekeeping. Concatemers enforcing 2EMRE/4MCU recapitulated the activity, gatekeeping, and size of endogenous channels. Thus, MCU does not require four EMRE, with most endogenous channels containing two, but complexes with 1–4 EMRE have activity with full or partial gatekeeping.

Graphical Abstract

Highlights

• •Mitochondrial Ca²⁺ uptake is mediated by the MCU Ca²⁺ ion channel
• •EMRE is required for MCU activity and for MCU interactions with MICU proteins
• •The EMRE:MCU stoichiometry of an MCU channel in vivo is unknown
• •MCU does not require four EMRE, with most endogenous channels containing two

Molecular Biology; Molecular Structure

Introduction

Ca²⁺ uptake by energized mitochondria regulates bioenergetics, apoptosis, and global Ca²⁺ signaling. The primary pathway for mitochondrial Ca²⁺ uptake is the mitochondrial calcium uniporter (MCU), a Ca²⁺-selective ion channel in the inner mitochondrial membrane (IMM) (Kirichok et al., 2004). MCU-mediated Ca²⁺ uptake is driven by the large negative IMM potential (ΔΨ_m) generated by proton pumping into the intermembrane space (IMS) by the electron transport chain. Despite the large thermodynamic driving force, mitochondrial Ca²⁺ uptake is tightly regulated to maintain low matrix [Ca²⁺] ([Ca²⁺]_m) and prevent opening of the permeability transition pore, ΔΨ_m depolarization, and cell death, while meeting dynamic energy demands of the cell. The MCU channel exists as a protein complex that (in metazoans) comprises several proteins: MCU contains two transmembrane helices (TM1 and TM2) and comprises the channel pore-forming subunit (Baughman et al., 2011) (De Stefani et al., 2011); MICU1 and its paralogs MICU2 and MICU3 are IMS-localized elements that sense cytoplasmic [Ca²⁺] ([Ca²⁺]_c) to regulate channel activity (Perocchi et al., 2010) (Mallilankaraman et al., 2012) (Plovanich et al., 2013) (Csordás et al., 2013) (Kamer and Mootha, 2014) (Patron et al., 2014) (Payne et al., 2017); and EMRE is a single-pass transmembrane protein that both is essential for MCU channel activity (Sancak et al., 2013) (Kovács-Bogdán et al., 2014) (Yamamoto et al., 2016) and helps tether MICU proteins to the channel complex (Tsai et al., 2016). Although expression of each of these subunits is necessary for proper channel permeation, gating, and regulation, their stoichiometry in the functional MCU complex is largely unknown.

Although originally the pore was believed to assemble as a pentamer (Oxenoid et al., 2016), recent crystal- and cryo-EM structures of full-length fungal and metazoan MCU channels revealed that the pore assembles as a homo-tetramer, in which the transmembrane domains adopt a 4-fold symmetry that stabilizes the selectivity filter (Baradaran et al., 2018) (Yoo et al., 2018) (Wang et al., 2019) (Fan et al., 2018) (Nguyen et al., 2018). MICU1 forms homo-dimers as well hetero-dimers with its paralogs MICU2 (Patron et al., 2014) (Petrungaro et al., 2015) and MICU3 (Patron et al., 2019). MICU1 homo-dimers inhibit MCU channel activity at low (<500 nM) [Ca²⁺]_c, a process referred to as “gatekeeping.” They also contribute to cooperative activation of Ca²⁺ uptake when this threshold is exceeded by Ca²⁺ binding to its two EF-hand domains (Mallilankaraman et al., 2012) (Csordás et al., 2013) (Hoffman et al., 2013) (Patron et al., 2014) (Kamer and Mootha, 2014) (Wang et al., 2014) (Antony et al., 2016) (Liu et al., 2016) (Payne et al., 2017). MICU2 requires MICU1 to stabilize its expression and exists as an obligate hetero-dimer with MICU1 (Plovanich et al., 2013) (Patron et al., 2014) (Payne et al., 2017). MICU2 tunes the [Ca²⁺]_c threshold for relief of MICU1-mediated gatekeeping and the gain of cooperative Ca²⁺ activation of MCU open probability, mediated by Ca²⁺ binding to its EF hands that have different affinities from those in MICU1 (Kamer et al., 2017) (Payne et al., 2017). MICU1 and MICU2 are broadly expressed, whereas MICU3 is found exclusively in the central nervous system and appears to be structurally and functionally similar to MICU2 (Plovanich et al., 2013) (Patron et al., 2019) (Xing et al., 2019). MCU is likely regulated primarily by MICU1/2 dimers in most cells (Patron et al., 2014) (Payne et al., 2017). It has been suggested that MICU1 associates with the channel complex in part by a direct interaction with MCU by an electrostatic mechanism involving positively charged residues in MICU1 and negatively charged residues associated with the pore domain of MCU (Paillard et al., 2018) (Phillips et al., 2019). MICU2 lacks analogous basic residues, consistent with the fact that MICU2 requires dimerization with MICU1 for its association with the complex. The number of MICU1/2 dimers that associate with one MCU channel is unknown.

MICU1/2 association with the channel is also mediated by interactions with EMRE. A conserved polyacidic C terminus of EMRE faces into the IMS where it interacts with a polybasic region in the N terminus of MICU1 (but not MICU2) (Sancak et al., 2013) (Hoffman et al., 2013) (Yamamoto et al., 2016) (Tsai et al., 2016) (Vais et al., 2016). This interaction is required for the gatekeeping function of MICU1/2 since deletion of the polyacidic region or neutralization of its charge abolishes normal gatekeeping, causing mitochondrial Ca²⁺ overload (Tsai et al., 2016) (Vais et al., 2016). In addition, binding of EMRE to MCU is necessary for channel activity. In the absence of EMRE, MCU is expressed normally but it lacks function. The association of EMRE with the channel complex is mediated by transmembrane interactions with MCU (Tsai et al., 2016). A recent cryo-electron microscopy structure of human MCU channel in complex with EMRE (Wang et al., 2019) revealed a 4 EMRE:4 MCU stoichiometry, mediated largely by hydrophobic interactions between the transmembrane domain (TMD) of EMRE and TM1 in MCU. The structure revealed an asymmetric distribution of MCU N-terminal domains (NTDs) among subunits, in contrast to a dimer-of-NTD-dimers observed in fungal MCU channels (Baradaran et al., 2018) (Yoo et al., 2018) (Wang et al., 2019) (Lee et al., 2016). In addition, dimerization between NTDs of two assembled human MCU tetramers suggested that metazoan MCU channels may form dimers in situ.

An important question is whether this cryo-EM MCU-EMRE structure recapitulates the stoichiometry of MCU and EMRE in the functional channel complex in vivo. Furthermore, it is unknown whether the same or different numbers of EMRE subunits are required for MCU activity on the one hand and normal MICU1/2-mediated gatekeeping and cooperative activation on the other. EMRE (but not MCU) is rapidly turned over by degradation through the activity of IMM-associated matrix-AAA (mAAA) proteases (Tsai et al., 2017) (König et al., 2016). Such rapid turnover could result in variable relative steady-state expression levels of EMRE and MCU that could alter the stoichiometry in cells. If MCU requires four EMRE for its activity, then absence of a single EMRE subunit from the complex would lead to complete loss of channel activity. On the other hand, if fewer than four EMRE subunits per MCU tetramer are sufficient for activity but fail to properly tether MICU1/2 dimers to the complex, such channels may be functional but exhibit altered gatekeeping and cooperative-activation profiles.

We have addressed these questions by determining the consequences of altering the number of EMRE subunits in each complex on both channel activity and regulation by MICU1/2. First, we manipulated the relative MCU and EMRE expression ratios in cells and correlated them with functional features. Second, we created a series of MCU-EMRE concatemers with multiple concatenated MCU subunits fused to EMRE in order to “fix” their stoichiometries in each channel complex. These approaches enabled functional and biochemical comparison of channels with a range of EMRE:MCU stoichiometries to endogenous channel complexes. Our results suggest that an MCU channel largely exists in association with fewer than four and more than one EMRE subunit, likely two.

Results

Expression of EMRE and MCU in Double-KO Cells Restores Channel Activity but Not Gatekeeping

To create cells in which relative MCU and EMRE protein expression levels could be directly measured and correlated with functional outcomes, we first created a cell line in which all genomic copies of both MCU and EMRE were inactivated. This was accomplished by CRISPR/Cas9-mediated genetic knockout (KO) of MCU in HEK293T cells in which all EMRE genes had already been inactivated (provided by V. Mootha), hereafter referred to as MEKO cells. Inactivating mutations in the EMRE and MCU genes were verified by genomic sequencing. Western blotting revealed a complete loss of MCU protein expression (Figures 1A and 1B: red). MICU2 dimerizes with MICU1 under non-reducing conditions but does not form homo-dimers when expressed in MICU1 KO cells (Payne et al., 2017). Thus, MICU2 bands at ∼100 kDa observed in non-reducing western blots correspond to MICU1/2 hetero-dimers. MEKO cells exhibited reduced expression of MICU1/2 dimers relative to wild-type (WT) HEK293T cells (Figures 1A and 1C: red), which increased upon stable overexpression of MICU1-FLAG (MEKO + MICU1 cells) to levels equivalent to those in WT cells (Figures 1A and 1C: blue). Similarly, stable transfection of MEKO cells with MCU-V5 (MEKO + MCU cells) restored both MCU and MICU1/2 dimer expression to WT levels (Figures 1A–1C: yellow). We were able to express MCU-V5 at ∼10-fold higher levels than endogenous levels, even in MCU-KO cells. Similarly, we were able to express EMRE-V5 at high levels in EMRE-KO cells. However, we found that, when we co-expressed both MCU and EMRE in MEKO cells, neither protein expressed as strongly. Importantly, we screened numerous clones for each line to select the ones that had expression levels closest to the endogenous ones. Stable expression of EMRE-V5 in MEKO + MCU cells (MCU + EMRE cells) did not change MCU or MICU1/2 dimer expression relative to WT cells. EMRE in these stable lines was fully processed to its mature form (Figure S1A). The lack of available antibodies specific to the processed form of endogenous EMRE prevented comparison of EMRE levels between WT and MCU-EMRE cells. Consequently, expression levels of MCU and MICU1/2 in MCU + EMRE cells were similar to those in WT cells, whereas we do not know if the relative expression of MCU to EMRE in MCU + EMRE cells was similar to that in WT cells. Notably, EMRE-V5 expression was ∼10-fold lower than MCU-V5 expression in MCU + EMRE cells (Figures 1A and 1D). Given this EMRE:MCU ratio, binomial probability predicts that >98% of MCU tetramers would associate with zero, one, or two EMRE (Figure 1E).

Figure 1.

Expression of EMRE and MCU in Double-KO Cells Restores Channel Activity but Not Gatekeeping

(A) Western blots of MCU, V5-tag, MICU2, FLAG tag, and β-tubulin from whole-cell lysates of WT (HEK293T) cells, or MCU/EMRE KO (MEKO) cells +/− stable expression of MICU1-FLAG, MCU-V5, or MCU-V5 and EMRE-V5. Two bands for MCU-V5 protein due to low-level degradation of V5-tag, with upper band representing full-length MCU-V5. MICU2- and FLAG-blots performed under non-reducing conditions, such that MICU2 bands represent MICU1/2 hetero-dimers and FLAG bands represent MICU1-FLAG homo-dimers.

(B) Quantification of MCU western blot band intensity for cell lines in (A) from n ≥ 4 independent experiments. Mean pixel density corrected for β-tubulin intensity and normalized to endogenous MCU band intensity in WT cells for each experiment. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. p Values: pairwise comparisons with WT cells (∗p < 0.05; ∗∗p < 0.01; n.s., not different; Dunnett's multiple comparisons test).

(C) MICU2 western blot band intensity for cell lines in (A) from n ≥ 4 independent experiments under non-reducing conditions, quantified as described in (B).

(D) EMRE:MCU expression ratio in MCU + EMRE and MCU + EMRE + MICU1 cells (n = 9 independent experiments). Ratio calculated by dividing mean pixel density of EMRE-V5 band by that of MCU-V5 band for each experiment. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. EMRE:MCU expression ratio in MCU + EMRE cells is 0.09 ± 0.07 (mean ± SD).

(E) Probability distribution histogram of number of EMRE subunits per MCU channel in MCU + EMRE and MCU + EMRE + MICU1 cells based on mean EMRE:MCU expression ratio calculated in (D). Points: mean probability of obtaining 0–4 EMRE subunits per channel based on binomial distribution. Solid lines: spline fits of all data points. At this EMRE:MCU expression ratio, >98% of channels predicted to associate with 0, 1, or 2 EMRE subunits.

(F) Cytoplasmic (bath) [Ca²⁺] ([Ca²⁺]_c) of cell lines in (A) measured with FuraFF (K_D = 5.5 μM). Cells treated with digitonin (Dg), thapsigargin (Tg), and CGP37157 (CGP) as indicated. Bolus addition of Ca²⁺ to raise [Ca²⁺]_c to 8–10 μM at t = 700 s to stimulate MCU Ca²⁺ uptake. 2 μM CCCP added at t = 1,000 s to uncouple ΔΨ_m and release mitochondrial Ca²⁺ content. Representative traces from n ≥3 independent replicates for each cell line. Both MCU and EMRE required for Ca²⁺ uptake activity.

(G) Relative ΔΨ_m for experiments shown in (F). Transient depolarization observed in response to Ca²⁺ uptake, whereas addition of CCCP collapsed ΔΨ_m.

(H) Ca²⁺-uptake rates for WT and MCU + EMRE cells in response to bolus additions of CaCl₂ that increased [Ca²⁺]_c by ∼5 μM, n ≥ 3 independent experiments. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. p Values: pairwise comparison of MCU + EMRE with WT cells (n.s., not different; Welch's unpaired t test). Ca²⁺-uptake rates not different between WT and MCU + EMRE cells.

(I) Gatekeeping threshold for cell lines in (H) measured 300 s after bolus addition of ∼5 μM free Ca²⁺. Each point (hollow circle) represents independent experiment (n ≥ 3 for each condition). p Values: comparison of MCU + EMRE with WT cells (∗∗p < 0.01; n.s., not different; Welch's unpaired t test). Gatekeeping-threshold [Ca²⁺]_c after uptake in MCU + EMRE cells was reduced (∼500 nM) compared with WT levels (1.5–2 μM).

See also Figure S1.

To evaluate MCU function, we simultaneously measured MCU-mediated Ca²⁺ uptake and relative mitochondrial membrane potential (ΔΨ_m) in permeabilized cells as previously described using the cell-impermeant Ca²⁺-indicator dye FuraFF (K_D = 5.5 μM) (Mallilankaraman et al., 2012) (Payne et al., 2017) and tetramethylrhodamine ethyl ester (TMRE), respectively (Payne et al., 2017). Digitonin (Dg) was added to a suspension of cells to permeabilize the plasma membrane but leave the IMM intact (Figure 1F: t = 50 s). Thapsigargin (Tg) was added to inhibit the sarco/endoplasmic reticular calcium ATPase (SERCA), preventing Ca²⁺ uptake into the ER and releasing the ER Ca²⁺ stores into the bath (Figure 1F: t = 100–400 s). CGP37157 (CGP), an inhibitor of mitochondrial Na⁺/Ca²⁺-antiporter (NCLX)-mediated Ca²⁺ efflux, was added to isolate MCU as the only pathway for transport of Ca²⁺ between mitochondria and cytoplasm (Figure 1F: t = 400–700 s). Upon reaching a steady-state [Ca²⁺]_c, acute addition of a Ca²⁺ bolus to the bath (i.e., cytoplasm) immediately elevated [Ca²⁺]_c to 8–10 μM, which was followed by clearance of Ca²⁺ in WT cells that reduced [Ca²⁺]_c to a steady-state level between 1 and 2 μM (Figure 1F: t = 700–1,000 s; black trace). A transient increase in TMRE fluorescence was observed (Figure 1G) as Ca²⁺ influx depolarized ΔΨ_m, which recovered as [Ca²⁺]_c approached a steady state due to closure of MCU. This steady state represents the [Ca²⁺]_c threshold at which uptake ceases, a hallmark of MICU1/2-mediated MCU gatekeeping. Finally, the uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP) was added to dissipate ΔΨ_m (Figure 1G), leading to a rapid increase in [Ca²⁺]_c as total mitochondrial matrix Ca²⁺ content was released (Figure 1F: t = 1,000–1,100 s). TMRE fluorescence in CCCP was used to determine the relative ΔΨ_m between experiments.

Upon challenge with boluses of Ca²⁺ that increased [Ca²⁺]_c to ∼10 μM, MEKO cells exhibited no reduction in [Ca²⁺]_c (Figure 1F: red trace) or change in ΔΨ_m (Figure 1G: red trace), reflecting lack of mitochondrial Ca²⁺ uptake, as expected. MEKO + MICU1 cells (Figures 1F and 1G: blue traces) or MEKO + MCU cells (Figures 1F and 1G: yellow traces) also exhibited no MCU activity, because of the EMRE requirement for MCU channel activity (Sancak et al., 2013) (Kovács-Bogdán et al., 2014) (Yamamoto et al., 2016). In contrast, MCU + EMRE cells (Figures 1F–1H: green) exhibited mitochondrial Ca²⁺-uptake rates and transient ΔΨ_m depolarizations at [Ca²⁺]_c of 8–10 μM similar to WT cells (Figures 1F–1H: black). Surprisingly, the steady-state [Ca²⁺]_c achieved 200–300 s after uptake ceased was ∼1.5–2 μM in WT cells (Figures 1F and 1I: black), whereas it was reduced to ∼500 nM in MCU + EMRE cells (Figures 1F and 1I: green). Of note, reduction to levels lower than those achieved in WT cells is a feature of impaired gatekeeping (Mallilankaraman et al., 2012) (Csordás et al., 2013) (Kamer and Mootha, 2014) (Payne et al., 2017) (Patron et al., 2014) (Kamer et al., 2017). To explore this further, we measured MCU activity at low (100–600 nM) [Ca²⁺]_c using the high-affinity indicator Fura2 (K_D = 140 nM) following CGP addition. CGP was required because NCLX-mediated Ca²⁺ efflux can mask MCU-mediated influx even in the absence of full MICU1/2-mediated gatekeeping (Figures S1B and S1C: red traces). Acute CGP addition was without effect on [Ca²⁺]_c in WT cells (Figures 2A, 2B, and S1C: black) owing to MICU1/2 gatekeeping. A reduction in [Ca²⁺]_c following acute addition of CGP was observed in MICU1 KO cells, reflecting constitutive MCU activity (Figure S1C: red trace), as expected due to lack of gatekeeping (Mallilankaraman et al., 2012) (Csordás et al., 2013) (Kamer and Mootha, 2014) (Payne et al., 2017) (Patron et al., 2014). Notably, MCU + EMRE cells exhibited significant constitutive Ca²⁺ uptake (Figures 2A and 2B: green), confirming the lack of fully functional MICU1/2-mediated gatekeeping in these cells. MICU1/2 also mediates cooperative activation of MCU in response to higher [Ca²⁺]_c (Csordás et al., 2013) (Payne et al., 2017). Ca²⁺-uptake rates in response to acute additions of Ca²⁺ between 0.1 and 10 μM (Figures 2C and 2E), plotted on a log/log scale, yielded a slope of ∼2.8 in WT cells (Figure S1D: black), as previously observed (Payne et al., 2017). In contrast, the slope was ∼1.6 in MCU + EMRE cells (Figure S1D: green). However, this apparent loss of cooperative activation was due solely to constitutive channel activity at low (<1–2 μM) [Ca²⁺]_c and not a loss of cooperative activation at higher [Ca²⁺]_c as seen in MICU1 KO cells. Thus, gatekeeping and cooperative activation of MCU were uncoupled in MCU + EMRE cells, a phenomenon not previously seen in cells lacking MICU1 and/or MICU2 expression (Csordás et al., 2013) (Payne et al., 2017).

Figure 2.

Overexpression of MICU1 Only Partially Restores Gatekeeping in MCU + EMRE Cells

(A) [Ca²⁺]_c responses to acute inhibition of NCLX-mediated Ca²⁺ efflux for WT (black), WT + MICU1 (gray), MCU + EMRE (green), and MCU + EMRE + MICU1 (purple) cells measured with high-affinity indicator Fura2. Representative traces from n ≥ 3 independent replicates for each cell line. Overexpression of MICU1-FLAG in MCU + EMRE cells rescued gatekeeping at low (<300 nM) [Ca²⁺]_c.

(B) Ca²⁺-uptake rate below 600 nM [Ca²⁺]_c in response to inhibition of NCLX-mediated Ca²⁺ efflux for cell lines in (A). Each point represents initial Ca²⁺-uptake rate from an independent experiment (n ≥ 3 for each condition) determined by single-exponential fit to reduction of [Ca²⁺]_c. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. p Values: pairwise comparisons with WT cells ∗∗∗∗p < 0.0001; n.s., not different; Tukey's multiple comparisons test).

(C–F) Ca²⁺ uptake, measured using the paradigm depicted in Figure 1F, in suspensions of permeabilized WT (C), WT + MICU1 (D), MCU + EMRE (E), and MCU + EMRE + MICU1 (F) cells in response to acute increases of [Ca²⁺]_c to between 0.1 and 10 μM. Representative traces from independent experiments with different bolus additions shown.

(G) Ca²⁺-uptake rates for cell lines in (A) in response to bolus additions of CaCl₂ that increased [Ca²⁺]_c by ∼5 μM from n ≥ 3 independent experiments. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. p Values: pairwise comparison of each cell line with all other cell lines (n.s., not different; Tukey's multiple comparisons test). Ca²⁺-uptake rates similar for all cell lines shown.

(H) Steady-state [Ca²⁺]_c for cell lines in (A) measured 300 s after bolus addition of 0.1–10 μM Ca²⁺ plotted as function of initial peak [Ca²⁺]_c after addition. Each point (hollow circle) represents an independent experiment (n ≥ 12 for each condition). Solid lines: one-phase association fits of steady-state [Ca²⁺]_c versus the initial [Ca²⁺]_c after addition.

(I) Comparison of gatekeeping-threshold [Ca²⁺]_c from non-linear fits in (H). Bars: mean of the plateau parameter for each fit ± SEM. p Values: comparison of each cell line with all other cell lines (∗∗∗∗p < 0.0001; n.s., not different; Tukey's multiple comparisons test).

See also Figure S2.

Overexpression of MICU1 Only Partially Restores Gatekeeping in MCU + EMRE Cells

Although MICU1/2 dimer expression in MCU + EMRE cells was similar to that in WT cells (Figures 1A and 1C), we considered that, because of the low EMRE:MCU expression ratio in these cells (∼1:10), gatekeeping was disturbed because most channels lacked enough EMRE to tether sufficient MICU1/2 dimers to the channels. Because MICU1 can directly associate with MCU (Paillard et al., 2018) (Phillips et al., 2019), we reasoned that over-expression of MICU1 in MCU-EMRE cells might restore gatekeeping despite low EMRE expression, by directly interacting with MCU. To test this, MICU1-FLAG was stably expressed in MCU + EMRE cells (MCU + EMRE + MICU1 cells) and in WT cells (WT + MICU1 cells). Expression of MCU (Figures S2A and S2B) and the EMRE:MCU expression ratio (Figure S2C; Figures 1D and 1E: purple) were similar in MCU + EMRE + MICU1 and MCU + EMRE cells. Only MICU1-FLAG homo-dimers were observed in non-reducing western blots of MICU1-overexpressing cells (Figure S2D), because of its high expression relative to MICU1 in WT cells (Figure S2E) as we previously showed using the same antibodies (Payne et al., 2017). Nevertheless, MICU1/2 dimer expression in MCU + EMRE + MICU1 cells (Figures S2F and S2G: purple) was not different from that of WT, although a small reduction was observed relative to MCU + EMRE cells (Figures S2F and S2G: green). In contrast, it was increased 2- to 3-fold in the WT + MICU1 cells compared with WT (Figures S2F and S2G: gray). We measured MCU activity at low (100–600 nM) [Ca²⁺]_c as described above. As expected, acute CGP addition was without effect on [Ca²⁺]_c in WT + MICU1 cells (Figures 2A and 2B: gray) owing to MICU1/2 gatekeeping. MCU + EMRE cells exhibited significant constitutive Ca²⁺ uptake (Figures 2A and 2B: green), whereas gatekeeping at low [Ca²⁺]_c was restored in MCU + EMRE + MICU1 cells (Figures 2A and 2B: purple). This result suggests that gatekeeping can be reconstituted by EMRE-independent association of MICU1 with MCU. Ca²⁺-uptake rates in response to Ca²⁺ boluses that elevated [Ca²⁺]_c by ∼5 μM were not different between WT (Figures 2C and 2G: black), WT + MICU1 (Figures 2D and 2G: gray), MCU + EMRE (Figures 2E and 2G: green), and MCU + EMRE + MICU1 cells (Figures 2F and 2G: purple). There was an apparent increase in cooperative channel activation in MCU + EMRE + MICU1 cells (slope ∼2.1) (Figure S2H: purple) relative to MCU + EMRE cells (slope ∼1.6) (Figure S1D: green), but this could be attributed solely to the restoration of gatekeeping at low [Ca²⁺]_c. Cooperativity of MCU activation was similar between the WT (Figure S1D: black) and WT + MICU1 (Figure S2H: gray) cells (slope ∼2.8 for both). The normal threshold [Ca²⁺]_c for MCU activation was observed in WT and WT-MICU1 cells (∼1.5 μM; Figures 2C and 2D; H-I: black versus gray), despite enhanced expression of MICU1 and 2. Importantly, however, the threshold in MCU + EMRE + MICU1 cells was ∼800 nM (Figures 2F, 2H, and 2I: purple), still below that in WT cells. Thus, although MICU1 overexpression restored gatekeeping at low [Ca²⁺]_c in MCU + EMRE cells, presumably by directly interacting with MCU, it was insufficient to fully restore normal gatekeeping. We hypothesized that the low EMRE:MCU ratio in MCU + EMRE + MICU1 cells resulted in sub-stoichiometric association of EMRE with MCU that enabled channel gating but was insufficient to tether enough MICU1/2 dimers to reconstitute fully functional gatekeeping.

Increasing EMRE Expression Enhances MICU1/2-Mediated Gatekeeping

To test this, we increased the EMRE:MCU ratio by enhancing steady-state EMRE expression. Under normal conditions, mAAA proteases degrade EMRE (Tsai et al., 2017) (König et al., 2016). Substitution of a proline (P) residue with isoleucine (I) at amino acid position 76 in EMRE inhibits its degradation, increasing its expression without affecting channel function or gatekeeping (Tsai et al., 2017) (Figures S3A and S3B). We stably overexpressed C-terminal V5-tagged P76I-EMRE and MCU along with MICU1-FLAG (to ensure sufficient MICU1 to occupy all functional EMRE:MCU complexes) in the MEKO background (MCU + P76I-EMRE + MICU1 cells). EMRE expression was increased ∼6-fold in MCU + P76I-EMRE + MICU1 cells compared with MCU + EMRE + MICU1 cells (Figure S3C; Figure 3A), whereas expression of MCU (Figures S3D and 3B) and MICU1/2 (Figures S3E and S3F) was similar between the two lines. Assuming random assembly of MCU/EMRE complexes, >95% of MCU tetramers are predicted to associate with three or four EMRE subunits in MCU + P76I-EMRE + MICU1 cells (Figure 3C: orange), in contrast to the one or two in the cells expressing WT EMRE (MCU + EMRE + MICU1 cells, Figure 3C: purple).

Figure 3.

Increasing EMRE Expression Enhances MICU1/2-Mediated Gatekeeping

(A) EMRE:MCU expression ratio in whole-cell lysates of MCU + EMRE + MICU1 and MCU + P76I-EMRE + MICU1 cells from n ≥ 5 independent experiments. Ratio calculated by dividing mean pixel density of EMRE-V5 band by that of MCU-V5 band for each experiment. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. p Value: comparison of MCU + EMRE + MICU1 cells with MCU + P76I-EMRE + MICU1 cells (∗∗∗p < 0.001; unpaired t test with Welch's correction). EMRE:MCU expression ratio in MCU + P76I-EMRE + MICU1 cells is 0.92 ± 0.21 (mean ± SD).

(B) MCU western blot band intensity for cells in (A) from n ≥ 5 independent experiments. Mean pixel density corrected for β-tubulin intensity and normalized to endogenous MCU band intensity in WT cells for each experiment. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. p Value: comparison of MCU + EMRE + MICU1 cells with MCU + P76I-EMRE + MICU1 cells (n.s., not different; unpaired t test with Welch's correction).

(C) Probability distribution histogram of expected number of EMRE per MCU channel in MCU + EMRE + MICU1 and MCU + P76I-EMRE + MICU1 cells based on EMRE:MCU expression ratios calculated in (A). Points represent mean probability of obtaining 0–4 EMRE subunits per channel based on binomial distribution. Solid lines: spline fits of all data points. EMRE:MCU expression ratio in MCU + P76I-EMRE + MICU1 cells predicts that >95% of MCU tetramers associate with three or four EMRE subunits.

(D) [Ca²⁺]_c response to acute inhibition of NCLX-mediated Ca²⁺ efflux for WT + MICU1 (gray), MCU + EMRE + MICU1 (purple), and MCU + P76I-EMRE + MICU1 (orange) cells measured with Fura2. Representative traces from n ≥ 3 independent replicates for each cell line.

(E) Ca²⁺ uptake in MCU + P76I-EMRE + MICU1 cells in response to acute increases of [Ca²⁺]_c to between 0.1 and 10 μM. Individual traces from independent experiments with series of bolus additions shown. Gatekeeping threshold in MCU + P76I-EMRE + MICU1 cells restored to WT levels.

(F) Ca²⁺-uptake rates for MCU + EMRE + MICU1 (from Figure 2F) and MCU + P76I-EMRE + MICU1 cells in response to bolus additions of CaCl₂ that increased [Ca²⁺]_c by ∼5 μM from n ≥ 3 independent experiments. Replicate measurements: hollow circles; boxes: 25^th–75^th percentile with line at median; error bars: range of all measurements. p Values: pairwise comparison of MCU + EMRE + MICU1 cells with MCU + P76I-EMRE + MICU1 cells (n.s., not different; unpaired t test with Welch's correction).

(G) Ca²⁺-uptake rates between 0.1 and 10 μM [Ca²⁺]_c measured in (E) and Figure 2F plotted on log₁₀/log₁₀ scale. Each point (hollow circle) represents independent experiment. Solid lines: linear fits of (log₁₀) Ca²⁺-uptake rate versus (log₁₀) [Ca²⁺]_c. Slope of best-fit line, a measure of cooperativity of MCU activation, was similar between MCU + EMRE + MICU1 and MCU + P76I-EMRE + MICU1 cells.

(H) Comparison of gatekeeping-threshold [Ca²⁺]_c from non-linear fits (in Figure S3I). Bars represent mean of plateau parameter for each fit ±SEM. p Value: comparison of MCU + EMRE + MICU1 cells with MCU + P76I-EMRE + MICU1 cells (∗∗∗p < 0.001; n.s., not different; unpaired t test with Welch's correction). Gatekeeping threshold in MCU + P76I-EMRE + MICU1 was elevated relative to that in MCU + EMRE + MICU1 cells (from Figure 2F).

See also Figure S3.

In the low-[Ca²⁺]_c regime, no differences in Ca²⁺ uptake were observed between WT + MICU1, MCU + EMRE + MICU1, and MCU + P76I-EMRE + MICU1 cells (Figure 3D), indicating that MICU1/2-mediated gatekeeping at low [Ca²⁺]_c was intact. In the higher-[Ca²⁺]_c regime (8–10 μM) (Figure 3E), the rates of Ca²⁺ uptake (Figure 3F) and cooperative activation (Figure 3G) were similar in MCU + EMRE + MICU1 and MCU + P76I-EMRE + MICU1 cells. Importantly, the steady-state [Ca²⁺]_c achieved after uptake (gatekeeping threshold) was significantly higher in the MCU + P76I-EMRE + MICU1 cells (∼1.8 μM) than in MCU + EMRE + MICU1 cells (∼800 nM) (Figures S3G and 3H). These results indicate that increasing the EMRE:MCU ratio elevates the [Ca²⁺]_c threshold required for relief of MICU1/2-mediated gatekeeping. Together, the data imply that MCU can associate with fewer than four EMRE per channel, which is sufficient to reconstitute channel activity and cooperative channel activation but may be insufficient to fully reconstitute MICU1/2-mediated channel gatekeeping. These results suggest that MCU activity can be differentially modulated by the number of EMRE subunits associated with the channel complex, possibly by determining the number of MICU1/2 dimers associated with it. Accordingly, a 4 EMRE:1 channel stoichiometry may not be fixed in cells.

MCU-EMRE Concatemers Form Channels with Fixed EMRE:MCU Stoichiometries

Co-expression of MCU and EMRE at known ratios cannot determine stoichiometries of the two proteins in the functional channel complex. To address this, we created a system in which we could establish fixed EMRE:MCU stoichiometries. We expressed a synthetic protein in which a truncated EMRE (ΔN30 a.a.) peptide containing a C-terminal V5 tag was fused to the C terminus of MCU (Figure 4A). This construct was previously shown to reconstitute MCU channel activity in the absence of endogenous MCU and EMRE (Tsai et al., 2016). We stably expressed the MCU-EMRE fusion protein (hereafter referred to as ME) in MEKO cells. Western blotting for V5 demonstrated that the fusion protein was expressed at the expected size, but probing with an α-MCU antibody also revealed a smaller band at the approximate molecular weight of MCU (Figure S4A). We surmised that this was caused by proteolytic cleavage. To minimize this degradation, we introduced the P76I mutation into the ME concatemer (ME[P76I]). Degradation observed in the original construct was mitigated by expression of the mutant construct (Figure S4A).

Figure 4.

MCU-EMRE Concatemers Form Channels with Fixed EMRE:MCU Stoichiometries

(A) MCU-EMRE fusion protein comprised of full-length MCU fused to EMRE with the mitochondrial-targeting sequence truncated from its N terminus, followed by C-terminal V5 tag.

(B) Assembly of the ME concatemer (ME[P76I]) enforces a 1:1 EMRE:MCU stoichiometry in all channel complexes when expressed in MEKO cells.

(C) MME concatemer assembles as an MCU tetramer containing two EMRE subunits in MEKO cells.

(D) MMME concatemer cannot assemble as an MCU tetramer, but formation of hexamers or other higher-order oligomers may assemble as functional channels in MEKO cells.

(E) Expression of MMME in EMRE KO cells allows formation of channel complexes containing a single EMRE due to presence of endogenous MCU monomers.

(F) Western blot for MCU and Hsp60 in isolated mitochondria from representative clones of cell lines under reducing conditions.

(G) Quantification of molecular weight from normalized band intensity for α-MCU western blots in (F). Traces: mean of n = 3 replicate experiments. Single ∼40-kDa band observed in cells expressing ME[P76I]; MME exhibited bands at ∼80 and 160 kDa; MMME ran at ∼110 and 220 kDa. Band at ∼35 kDa in EMRE KO + MMME + MICU1 cells represents endogenous MCU.

(H) Immunofluorescence microscopy for V5 (MCU-EMRE concatemers), FLAG (MICU1), and DAPI in representative clones of cell lines in (F). All cell lines tested showed strong colocalization of both tags in mitochondria.

See also Figure S4.

Expression of ME[P76I] in MEKO cells is predicted to result in a 1:1 EMRE:MCU stoichiometry in all channel complexes (Figure 4B). We also created MCU-EMRE[P76I] fusion proteins with two, three, or four concatenated MCU subunits fused to P76I-EMRE (hereafter referred to as MME, MMME, and MMMME, respectively). MME is predicted to dimerize, forming complexes with two EMRE subunits/MCU channel (Figure 4C), and MMMME is predicted to form complexes with one EMRE subunit/MCU tetramer. In contrast, MMME should be unable to form tetramers in MEKO cells, although alternate assemblies (e.g., hexamers) with channel activity could possibly form (Figure 4D). MEKO cells were stably transfected with each concatemer together with MICU1-FLAG to ensure sufficient MICU1. Three clones were identified for each condition, and whole-cell lysates were analyzed for MCU and MICU1/2-hetero-dimer expression by western blot (Figures S4B–S4D). Surprisingly, the MMMME construct failed to express (Figure S4E). Thus, to create channels with a single EMRE subunit/MCU tetramer, the MMME concatemer and MICU1-FLAG were expressed in EMRE KO (HEK293T) cells (Figure 4E) and three stable clones were isolated as described. The presence of endogenous MCU in these cells should allow a single MCU monomer to associate with the MMME concatemer to form tetrameric channel complexes containing only one EMRE subunit. Indeed, the MMME concatemer only very weakly expressed in MEKO cells, whereas it expressed well in EMRE KO cells (Figures 4F and 4G), suggesting that endogenous MCU interacted with and stabilized MMME expression. To probe the integrity of each concatemer, mitochondria were isolated from selected clones and analyzed by western blot. The ME[P76I], MME, and MMME (in EMRE KO cells) concatemers were all predominantly expressed as full-length proteins (ME∼40 kDa; MME∼80 kDa, and MMME∼110 kDa, respectively, Figures 4F and 4G). Localization of each concatemer and MICU1-FLAG to mitochondria was confirmed by confocal immunofluorescence microscopy (Figure 4H). It was possible to compare MCU expression in ME[P76I] clones with that in WT cells as this concatemer exhibited a single band in western blots and contained a single MCU epitope. ME[P76I] was ∼2-fold overexpressed in one clone (#3), modestly (<2-fold) overexpressed in a second (#1), and slightly lower (∼2-fold) than WT cells in a third (#2) (Figure S4C). The presence of high-molecular-weight bands in cells expressing MME and MMME made it difficult to assess the relative MCU levels. This was further confounded by possible differences in MCU-antibody binding to fusion proteins containing multiple MCU subunits. We therefore employed V5-epitope ELISA assays on each clonal cell line. Based on the relationship in ME[P76I] clone #2 between western blot expression and the relative number of V5-binding sites per cell, the MCU expression level was estimated to vary by only 2-fold among the different concatemer clones (Figure S5A). MICU1/2 dimer expression in all clones was at least as high as in WT cells (Figure S4D).

We assessed mitochondrial Ca²⁺ uptake in each cell line at both high (5–8 μM) and low (<600 nM) [Ca²⁺]_c to determine (1) whether each concatemer was sufficient to form functional channels in the absence of endogenous MCU or EMRE expression and (2) if the channels exhibited normal MICU1/2-mediated gatekeeping. Each concatemer reconstituted MCU activity, with some variability among clones in the Ca²⁺-uptake rates (ranging from ∼100 to 400 nM/s) in response to 5-μM Ca²⁺ boluses (Figure S5B). Six of twelve clones displayed intact gatekeeping at low (<600 nM) [Ca²⁺]_c (Figure S5C). We selected individual clones for each concatemeric construct that had mean Ca²⁺-uptake rates at 5–8 μM [Ca²⁺]_c most similar to WT cells and intact gatekeeping in the low-[Ca²⁺]_c regime (Figures S5A–S5C, red arrows). For each, we measured Ca²⁺-uptake rates in response to 0.1–10 μM [Ca²⁺]_c and the subsequent steady-state [Ca²⁺]_c at which gatekeeping was re-established (Figures 5A–5D). Plotting uptake rate as a function of the peak [Ca²⁺]_c immediately after Ca²⁺ addition and fitting the data to a Hill equation revealed similar profiles for the different concatemers, with the rate for cells expressing ME[P76I] notably right-shifted to higher [Ca²⁺]_c (Figure 5E). Linear fits on a log/log scale showed that cooperativity in the transition of MCU from the closed to the open state was the same for each concatemer (slope = 2.5), and similar to WT + MICU1 cells (Figure 5F), suggesting that MICU1/2 regulation was properly coupled to the channel in each concatemer complex. Importantly, however, the threshold at which gatekeeping was re-established (Figure 5G) was elevated uniquely in cells expressing ME[P76I] (Figure 5H). This result is consistent with the right-shifted Ca²⁺ dependence of MCU activation observed in these cells, and it suggests that gatekeeping is particularly potent in MCU complexes containing four EMRE/channel. In contrast, gatekeeping thresholds in cells expressing concatemers predicted to form channels with fewer than four EMRE subunits were not different from those of WT (+MICU1) cells (Figure 5H). These results strongly suggest that channels containing four EMRE subunits tether more MICU1/2 dimers to the channel complex than are normally present in WT cells, consistent with the analysis above that also suggested that EMRE:MCU stoichiometry, and by analogy the MICU1/2 dimer:MCU stoichiometry, may not be 4:4 or invariant in cells.

Figure 5.

Function of MCU-EMRE Concatemers

(A–D) Mitochondrial Ca²⁺ uptake in MEKO + ME[P76I]+MICU1 (A), MEKO + MME + MICU1 (B), MEKO + MMME + MICU1 (C), and EMRE KO + MMME + MICU1 (D) cells in response to acute increases of [Ca²⁺]_c to between 0.1 and 10 μM. Individual traces from independent experiments with different bolus additions shown.

(E) Ca²⁺-uptake rates for cell lines in (A)–(D) in response to challenge with [Ca²⁺]_c between 0.1 and 10 μM. Each point (hollow circle) represents independent experiment. Solid lines: Hill equation fits of Ca²⁺-uptake rate versus initial [Ca²⁺]_c after bolus addition. ME[P76I]+MICU1 cells exhibited a right shift in the [Ca²⁺]_c dependence of MCU activation compared with all other cell lines.

(F) Ca²⁺-uptake rates between 0.1 and 10 μM [Ca²⁺]_c measured in (E) plotted on log₁₀/log₁₀ scale. Each point (hollow circle) represents independent experiment. Solid lines: linear fits of (log₁₀) Ca²⁺-uptake rate versus (log₁₀) [Ca²⁺]_c. Slope of the best-fit line as a measure of cooperative MCU activation similar among all cell lines.

(G) Steady-state [Ca²⁺]_c for cells in (E) measured 300 s after bolus addition of 0.1–10 μM Ca²⁺ plotted as a function of initial peak [Ca²⁺]_c. Each point (hollow circle) represents independent experiment (n ≥ 6 for each condition). Solid lines: one-phase association fits of steady-state gatekeeping-threshold [Ca²⁺]_c versus initial [Ca²⁺]_c after addition.

(H) Comparison of gatekeeping-threshold [Ca²⁺]_c 300 s after bolus additions of Ca²⁺ from non-linear fits in (G). Bars: mean of plateau parameter for each fit ±SEM. p Values: pairwise comparisons with WT + MICU1 cells (∗∗∗∗p < 0.0001; n.s., not different; Tukey's multiple comparisons test). Gatekeeping threshold was elevated in ME[P76I]+MICU1 cells compared with other cell lines.

See also Figure S5.

Biochemical Analyses of MCU-EMRE Complexes

To compare the relative sizes of MCU-EMRE concatemer and endogenous channel complexes, we performed BN-PAGE western blotting of MCU from isolated mitochondria. The MCU complex in WT cells ran as a single band at ∼440 kDa (Figures 6A, 6C, 6D, and 6F: black), in line with previous reports (Sancak et al., 2013) (König et al., 2016). Overexpression of MICU1-FLAG in WT cells had no effect on the size of the complex (Figures 6A, 6C, 6D, and 6F: gray). In addition, MCU ran at approximately the same size in MICU1-KO mitochondria as in WT mitochondria (Figures 6A, 6C, 6D, and 6F: red), with a minor band of lower molecular weight similar in size to that observed in EMRE-KO mitochondria (∼350 kDa) (Figures 6A, 6C, 6D, and 6F: orange). These results are consistent with a previous report that MICU1 may stabilize interactions between MCU and EMRE and that MCU-containing bands at ∼440 kDa do not contain MICU proteins (König et al., 2016). Re-probing the blots with an α-FLAG antibody revealed that MICU1 assembled in complexes of ∼250 and ∼400 kDa in the absence of MCU and EMRE (Figure 6B), further indicating that complexes at ∼440 kDa detected with the α-MCU antibody contained only MCU and EMRE subunits.

Figure 6.

The Size of Concatemeric Channels Is Proportional to the Number of EMRE

(A) BN-PAGE MCU western blot of mitochondria isolated from WT, MICU1 KO, EMRE KO, MEKO + MICU1 cells and representative clones of WT + MICU1, MEKO + ME[P76I]+MICU1, MEKO + MME + MICU1, MEKO + MMME + MICU1, and EMRE-KO + MMME + MICU1 cells. Endogenous MCU complexes observed as single band at ∼440 kDa.

(B) BN-PAGE western blot for FLAG (MICU1) in mitochondria isolated from representative clones as indicated. MICU1 in complexes at ∼250 and ∼400 kDa in cells lacking MCU and EMRE.

(C) Quantification of raw intensity and molecular weight from α-MCU BN-PAGE western blots in (A). Traces are mean of n = 3 replicate experiments. All MCU-EMRE concatemers showed bands between 300 and 600 kDa (similar to WT cells), as well as bands at >700 kDa that likely represent complexes containing >4 MCU subunits.

(D and E) Molecular weight profile of smallest band (200–600 kDa) from α-MCU BN-PAGE of cell lines in (C). Traces: mean of n = 3 replicate experiments; data normalized to band with maximal intensity for each condition. (D) MCU complexes in MICU1 KO cells (red) were similar to WT (+/− MICU1, black and gray), with a minor band similar in size to complexes formed in EMRE KO cells (orange). (E) MCU-EMRE concatemers exhibited molecular weights that varied in size proportionally to the predicted number of EMRE subunits per complex.

(F) Statistical analysis of molecular weight profiles in (D)–(E). Bars: mean calculated size of each complex derived from peak intensity of bands between 200 and 600 kDa ±SEM. Replicate measurements: hollow circles (n = 3). p Values: pairwise comparisons of each cell line with WT cells (∗p < 0.05; ∗∗p < 0.01; Dunnett's multiple comparisons test). Complexes with four EMRE subunits per channel (ME[P76I], blue) were significantly larger than those in WT cells (black), whereas complexes containing only one EMRE subunit (EMRE KO + MMME, green) were smaller.

Importantly, complexes from mitochondria expressing ME[P76I] with an enforced 1:1 EMRE:MCU stoichiometry (Figures 6A, 6C, 6E, and 6F: blue) were significantly larger (∼490 kDa) than those observed in WT or WT + MICU1 mitochondria. Accounting for the V5 tag, the size of the ME concatemer differs from that of one MCU and one fully processed EMRE subunit by ∼1 kDa. This small difference cannot account for the observed difference of ∼20 kDa between the WT or WT + MICU1 complexes compared with the ME[P76I] complex. This result suggests that endogenous MCU channels contain fewer than four EMRE subunits. A minor band at ∼700 kDa was observed in ME[P76I] mitochondria, possibly corresponding to a dimer of MCU tetramers. Bands at ∼440 and ∼700 kDa were also seen in mitochondria from cells expressing MME and MMME, although higher-molecular-weight oligomers were evident (Figures 6A and 6C: pink, gold, and green). This could indicate a tendency for the concatemers to form complexes containing more than one MCU channel. Bands for MMME were much stronger when the construct was expressed in EMRE KO cells than in MEKO cells, likely due to enhanced stability of MMME in the presence of MCU expression due to their biochemical interaction. The molecular weights of MME and MMME complexes were not different from that of the MCU complex in WT cells (Figures 6E and 6F: pink and gold). Importantly, MCU complexes in EMRE-KO cells expressing MMME, which are predicted to contain a single EMRE subunit, were significantly smaller than those in WT cells (Figures 6A, 6C, 6E, and 6F: green), suggesting that endogenous MCU channels associate with more than one EMRE.

Discussion

The first important conclusion from this study is that human MCU complexes with fewer than four EMRE subunits can assemble as fully functional channels. This is evidenced by the finding that mitochondria from cells that have ∼10-fold higher expression of MCU relative to EMRE took up Ca²⁺ at rates similar to WT cells at [Ca²⁺]_c >5 μM (Figure 1) and further by the fact that MCU-EMRE concatemers that form channels with less than four EMRE subunits were functional (Figure 5). BN-PAGE revealed higher-order MCU oligomers in mitochondria expressing MME or MMME concatemers (Figure 6), but bands corresponding to MCU tetramers with one or two EMRE subunits (250–550 kDa) predominated. The BN-PAGE western blots revealed bands for MMME in EMRE KO cells corresponding to both an MCU tetramer (presumably containing endogenous MCU subunits) and a larger species, possibly corresponding to a hexamer (as depicted in Figure 3D). Expression of the MMME construct in the cells lacking MCU was very weak compared with expression in the EMRE KO cells that expressed MCU, suggesting that the endogenous MCU interacted with the MMME protein to stabilize its expression. This indicates that the dominant functional species in MME expressing cells was likely an MCU tetramer containing one EMRE. Thus, under circumstances where endogenous EMRE expression is insufficient to occupy all four binding sites in each channel, complexes with fewer than four EMREs retain full channel activity when the gatekeeping threshold is exceeded. Thus, we conclude that a 4 EMRE:1 channel stoichiometry is not necessarily the channel stoichiometry as suggested (Wang et al., 2019).

Second, the endogenous MCU channel in HEK293T cells likely contains two EMRE subunits. No differences in the [Ca²⁺]_c dependence of MCU activity or in the gatekeeping threshold were observed between channels with two EMRE subunits (MEKO + MME + MICU1) and WT cells (Figure 5), suggesting that these channels recapitulate the stoichiometry of endogenous channels. Of note, however, the functional profiles of cells expressing the MMME concatemer, which are predicted to form channels containing a single EMRE subunit, were functionally indistinguishable from those of MEKO + MME + MICU1 (or WT) cells. This could be due to formation of higher-order oligomers in the MMME-expressing cells (Figure 6) that created channels containing more than one EMRE, as depicted in Figure 4D. The [Ca²⁺]_c threshold for MCU activity in these cells likely arises from channels with one or two EMRE subunits. BN-PAGE analyses further suggested that endogenous MCU complexes run as a single species containing more than one and fewer than four EMRE per channel (Figure 6). The size of complexes between 250 and 550 kDa depended on the number of EMRE subunits per channel but not on incorporation of MICU proteins as they were not present in the complexes. Importantly, channels formed by expression of MME concatemers were not different in size from those in WT cells, whereas complexes containing a single EMRE were significantly smaller. Thus, the size of the native complex in WT cells is consistent with a channel complex containing two EMRE subunits bound to each MCU tetramer.

The third major conclusion is that gatekeeping is enhanced in MCU channels containing more EMRE subunits. Because the EMRE[P76I] mutation increased the steady-state expression of EMRE without affecting MCU expression (Figure 3), differences in channel function and regulation between MCU + EMRE + MICU1 and MCU + P76I-EMRE + MICU1 cells were due solely to the elevated EMRE:MCU ratio in the P76I-EMRE cells. Both cell lines showed similar uptake rates and cooperativity of channel activation, but the right shift in plots of initial uptake rate versus [Ca²⁺]_c and the elevated gatekeeping threshold in MCU + P76I-EMRE + MICU1 cells indicate that MICU1/2-mediated gatekeeping in channels with three or four EMRE is enhanced compared with those associated with only one or two. These findings are reinforced by the functional profile of the ME[P76I] concatemer, in which all channels are predicted to contain four EMRE (Figure 5). Like MCU + P76I-EMRE + MICU1 cells, ME[P76I]+MICU1 cells showed a right shift in the Ca²⁺-uptake rate versus initial [Ca²⁺]_c relationship and an elevated gatekeeping threshold compared with WT + MICU1 cells, with no differences in uptake rate and cooperative activation.

These findings are relevant in light of the recent structure of human MCU that revealed four EMRE subunits bound to the MCU tetramer. We confirmed here that such a channel is functional with robust MICU1/2-mediated gatekeeping but that a 4:4 stoichiometry in not obligatory. Our results provide further evidence that EMRE performs two independent roles in MCU-channel regulation. First, it allows MCU to adopt an open channel conformation by binding only a single EMRE. In contrast, its role in tethering MICU1/2 to the complex to mediate channel gatekeeping is modulated by the number of EMRE subunits per channel. Such a phenomenon is reminiscent of the functional modulation of BK channels by auxiliary subunits γ1 and β (Gonzalez-Perez et al., 2018). Similar to the role of EMRE in MCU activation, γ1 has an “all-or-none” effect of the voltage dependence of BK-channel gating, in which a single subunit associating with the pore-forming subunits is sufficient to promote the full effect. On the other hand, each β-subunit that associates with the BK α-tetramer incrementally shifts the voltage-dependence of channel gating such that the full shift is observed only when four β-subunits are present in the channel complex.

Our studies suggest that the endogenous MCU-channel complex in HEK293T cells contains an MCU tetramer that, on average, associates with two or three EMRE subunits. We suggest that, under resting [Ca²⁺]_c, two apo-MICU1/2 dimers associate with the complex through interactions with the C terminus of two EMRE subunits and residues in MCU subunits, mediating channel inhibition (i.e., gatekeeping). When the local [Ca²⁺]_c rises, Ca²⁺ binding to MICU1/2 promotes dissociation of MICU1 from the pore and relieves MCU inhibition. Under these conditions, a role of EMRE in tethering MICU1/2 dimers to the complex may be essential to ensure that gatekeeping is re-established quickly enough to ensure that channel activity ceases at the proper [Ca²⁺]_c threshold after uptake has occurred. Future work to address the structures of MCU channels with fewer than four EMRE subunits and the MCU-EMRE complex in the presence of MICU1/2 dimers is necessary to confirm these hypotheses.

Limitations of the Study

This study was performed using an immortalized line rather than primary cells from different tissues. Creating double (MCU, EMRE) knockout lines in primary cells is desirable to determine the relevance of the findings here in other cell types. However, such an undertaking is beyond the scope of this manuscript. The purpose of this study was not to characterize the assembly of MCU channel complexes in different tissues but rather to show in a single system which complexes are sufficient for channel activity and how the relative stoichiometry of MCU and EMRE affects channel function.

MCUb has been shown to form complexes with MCU (Raffaello et al., 2013), and some evidence suggests that MICU1/2 (and possibly EMRE) do not associate with MCUb (Lambert et al., 2019). Thus, it is possible that the preferred 2:4 EMRE:MCU channel stoichiometry we observe is due to formation of complexes with two MCU, two MCUb and two EMRE subunits that retain channel activity.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: April 24, 2020