Systematic mapping of mitochondrial calcium uniporter channel (MCUC)-mediated calcium signaling networks

We devised a biochemical workflow to achieve an unbiased and comprehensive characterization of MCUC-specific PPIs in human cells under resting conditions (Fig. 1). To this goal, MCU, MCUB, or EMRE proteins were fused at the C-terminus to a StrepII-HA-His₆ tag and stably integrated into Flp-In T-REx HEK293 cells (parental line) under the control of a tetracycline-inducible promoter. This allowed the targeted integration of each bait protein at a single transcriptionally active genomic locus that is the same in every cell line and ensures a near-physiological expression level of the bait proteins. The latter is essential to minimize false positive associations and accurately define the endogenous protein composition and interaction network of MCUC. By encompassing all three membrane-spanning subunits of MCUC as baits, we aimed to increase the coverage of copurifying proteins across all mitochondrial sub-compartments. As shown in Figs. 2A and  EV1, we confirmed that tetracycline-dependent induction of baits did not affect the endogenous level of other MCUC subunits, except for EMRE. Upon expression of tagged MCU and EMRE, the level of endogenous EMRE was increased and decreased, respectively, which is consistent with previous findings indicating a tight regulation of its stability by MCU (Sancak et al, 2013; Tsai et al, 2017; König et al, 2016).

We employed a TAP strategy based on Strep and polyhistidine tags to enrich for interacting protein partners of MCUC. Immunoblot analyses validated that each bait could be efficiently purified from whole cell lysates (Fig. 2B). For a systematic and unbiased identification and quantification of proteins that co-precipitated with each bait, we performed LC-MS/MS analysis with at least four biological replicates. Importantly, we included TAPs from the parental line, hereby used as negative control (Ctrl-TAP) to discriminate specific interactions from spurious ones. As a result, we found that the relative abundance of MCU, MCUB, and EMRE was consistently and significantly higher in TAPs from bait-expressing cells than in control samples, when compared to the median intensity of all quantified proteins (Fig. 2C). Given our interest in characterizing the interaction partners of MCUC within mitochondria, we focused our analyses on mitochondrial preys. To define significant and specific MCUC interactors we filtered our dataset for mitochondrial proteins and ran a quantitative bait-prey co-enrichment analysis based on a two-tailed Welch’s t test, a within-group variance (s0) of 1, and a permutation-based false discovery rate (FDR) of either 0.05 (“high-confidence”) or 0.10 (“medium-confidence”) (Fig. 2D–F). In addition, Hein et al have previously demonstrated that the intensity profiles of interacting proteins are correlated (Hein et al, 2015). Therefore, as additional classifiers to identify MCUC interactors, we performed “local” and “global” correlation analyses, comparing the similarity of protein intensity profiles across either pairs of bait and control TAPs, or across all measured samples (Ctrl-TAPs, MCU-TAPs, MCUB-TAPs, and EMRE-TAPs), respectively.

With the term “interactor” or “prey” we refer to any protein that stably or transiently binds MCUC directly or indirectly. Indirect binding can occur when baits and preys are in close proximity without engaging in direct physical interactions. These associations might be facilitated by subunits of a protein complex or members of a pathway that directly interact with MCUC. Examples include connections between mitochondrial protease complexes and MCUC (König et al, 2016; Tsai et al, 2017, 2022) or between MCU and RCCI (Balderas et al, 2022). The joint analysis of all datasets yielded a total of 139 interactions among 95 mitochondrial proteins and successfully recovered all currently known subunits of the MCUC, namely MCU, EMRE, MCUB, MICU1, MICU2, and MICU3 (Fig. 2G; Dataset EV1). Among the MCUC interactors, 33 were found to be shared between at least two baits, and 11 were common to all three conditions. Interestingly, more than 50% of all preys were bait-specific, suggesting their involvement in the selective maturation and regulation of MCU, EMRE, or MCUB.

To globally analyze MCUC-specific PPIs, we assessed the biochemical, functional, and evolutionary properties of all identified protein partners (Fig. 3; Dataset EV1–3). Roughly 50% of our interactome consisted of proteins with at least one predicted transmembrane domain, for example subunits of the outer (TOM) and inner (TIM) membrane protein translocases and the oxidative phosphorylation (OXPHOS) machinery. The other half were soluble proteins including components of the MICUs family, the tricarboxylic acid cycle (TCA cycle), and the mitochondrial DNA (mt-DNA) maintenance and expression system (Fig. 3A). Accordingly, the uniporter engaged in interactions with proteins localized in all four submitochondrial compartments (Fig. 3B). MCU and EMRE, which represent the minimal functional unit of MCUC (Kovács-Bogdán et al, 2014; Pittis et al, 2020), sampled largely similar environments although exposing their tags on opposite sides of the IMM. Interestingly, they also mediated functional associations with proteins located at the outer mitochondrial membrane (OMM) and involved in mitochondrial morphology, apoptosis, and dynamics, such as FIS1, BNIP3L, and PARK7. Besides subunits of MCUC, the identified interactors encompassed a wide spectrum of mitochondrial functions, with several not yet known to be linked to Ca²⁺ signaling (Fig. 3C).

We then mapped the MCUC interactome as a protein network, whereby each node corresponds to a prey connected to a given bait through either evidence of co-enrichment or profile correlations (Fig. 3D). Reassuringly, each bait significantly interacted with itself and with all known MCUC members. To further corroborate the high quality and coverage of our dataset, we first searched the literature for experimental evidence of physical and functional associations between the uniporter and mitochondrial proteins in our network. Among shared interactions, we identified GHITM, also known as TMBIM5 or MICS1. This protein was recently characterized as a Ca²⁺/H⁺ exchanger of the IMM (Zhang et al, 2022), able to shape mt-Ca²⁺ cycling by inhibiting the activity of the m-AAA protease AFG3L2 (Austin et al, 2022; Patron et al, 2022). We also identified AFG3L2 and the YME1L proteolytic hub (YME1L1, PARL, and STOML2) (Wai et al, 2016), all previously implicated in MCUC processing, assembling and degradation (König et al, 2016; Tsai et al, 2017, 2022). To test the usefulness of the MCUC protein network as a resource for the identification of novel research directions and regulatory mechanisms, we also mined our dataset for associations between the uniporter and mitochondrial functions that are known to be controlled by Ca²⁺ but for which the molecular players remain poorly characterized. For instance, the MCUC interactome included several proteins (FIS1, KIF1B, GJA1, and MTUS1) involved in mitochondrial ultrastructural organization, shape and dynamics. This supports the notion that mt-Ca²⁺ signaling can affect the organelle’s morphology through the regulation of mitochondrial fusion and fission (Zhao et al, 2015). In addition, our results raise the hypothesis that members of the MCUC could participate in the regulation of contact sites between inner and outer mitochondrial membranes through interactions with members of the MICOS complex such as MICOS10 (Rampelt et al, 2022). Notably, MICU1 was recently found to bind the MICOS complex and regulate mitochondrial membrane dynamics independently of matrix Ca²⁺ uptake, while MCU was shown to relocalize to the cristae junctions upon an increase of [Ca²⁺] in the IMS (Tomar et al, 2023; Gottschalk et al, 2019). Since mt-Ca²⁺ signaling is a central regulator of oxidative metabolism and ATP production (Hajnóczky et al, 1995; Jouaville et al, 1999), we also looked for molecular players coupling MCUC with mitochondrial bioenergetics. Remarkably, we found that the MCUC network was especially enriched in PPIs with components of RCCI (e.g., NDUFA3, NDUFB10, NDUFS8), RCCIV (e.g., COX6C, COX7C, COX18) and the TCA cycle (PDHB and PDHX). Supporting our findings, the RCCI assembly subunit NDUFA3 was recently shown to bind MCU, an interaction between complexes that regulates uniporter level and activity to maintain bioenergetic homeostasis (Balderas et al, 2022). Furthermore, our biochemical approach detected interactions between MCU, the J subunit of the ATP synthase complex (ATP5MPL), and the ATP synthase assembly factor, TMEM70. The latter was recently described to interact with the ATP synthase subunit c (Kovalčíkova et al, 2019; Bahri et al, 2021), which was previously shown to bind MCU in trypanosomes and human cells (Huang and Docampo, 2020). To the same goal, during Ca²⁺ signaling the activation of SLC25A25 allows matching cellular energy demand and supply by regulating the matrix adenyl nucleotide pool (del Arco et al, 2016). However, no previous evidence of an interaction between the uniporter and SLC25A25 has been reported. Instead, we found SLC25A25 to be significantly co-enriched with the MCU and EMRE baits possibly allowing them to co-localize in the same Ca²⁺ microdomain and co-activate in response to rises in Ca²⁺ concentration ([Ca²⁺]).

Next, we surveyed properties of mt-Ca²⁺ signaling by analyzing the expression and evolutionary profiles of all MCUC interactors, as well as their involvement in human diseases (Fig. 3E–H; Datasets EV2 and EV3). First, we quantified the relative protein level of each MCUC binding partner across healthy human tissues using the GTEx (Genotype-Tissue Expression) database (Jiang et al, 2020). Interestingly, we observed a highly heterogeneous tissue distribution of MCUC interactors, which is consistent with the wide range of biological processes linked to mt-Ca²⁺ in our network (Fig. 3E). These proteins encompass a wide range of functions, from fundamental mitochondrial processes like OXPHOS and mt-DNA replication, to more specialized roles, such as regulation of organelle shape and dynamics, as well as activation of cell death pathways. As expected, based on their role as the minimal unit of the uniporter, MCU and MICU1 showed a ubiquitous expression compared to tissue-specific regulators such as MICU2, MICU3 and MCUB. EMRE was often undetected due to its small size and highly hydrophobic nature, which makes it difficult to detect by MS analyses of whole proteomes. Next, we assessed the evolutionary conservation of each interactor across eukaryotes by mapping orthology relationships across 120 eukaryotic species inferred from Protphylo (Cheng and Perocchi, 2015). Comparative genomics analyses have been instrumental in identifying MCUC components (Perocchi et al, 2010; Baughman et al, 2011; De Stefani et al, 2011), which are present in all major eukaryotic groups, but mostly absent in protists and fungi (Pittis et al, 2020). However, while coevolution can be used to predict functional associations, functionally related proteins do not necessarily co-evolve, calling for complementary approaches, like ours, to reach a comprehensive map of mt-Ca²⁺ signaling networks. Namely, most of the proteins in our network did not exhibit similar phylogenetic profiles to MCU (Fig. 3F). Overall, they showed a patchy evolutionary conservation, with mitochondrial functional modules of ancient origin being highly conserved, while others present only in higher vertebrates. The former group could represent processes that were placed under the control of mt-Ca²⁺ early in evolution; the latter would possibly spotlight species-specific functions linked to Ca²⁺ signaling after the acquisition of MCU. Finally, we found that half of the MCUC interactors were linked to over 300 different disease phenotypes, spanning from neurological and metabolic disorders to cancer (Fig. 3G), including disorders related to mt-DNA homeostasis, whose pathophysiology has not yet been connected to dysfunctions in mt-Ca²⁺ signaling (Fig. 3H). Altogether, our analysis provides a high-confidence and comprehensive dataset of endogenous MCU, MCUB, and EMRE protein binding partners.

To corroborate the quality of our resource, we experimentally validated a subset of specific interactions with different degrees of co-enrichment and correlation to MCU, EMRE, and MCUB baits. MICU3, which was not previously identified as a bona fide binding partner of MCUC in HEK293 cells (Sancak et al, 2013; Antonicka et al, 2020; Austin et al, 2022), resulted as a high-confidence interactor of all three baits in our dataset. Indeed, quantitative proteomic analyses of mitochondria have confirmed that MICU3 is expressed in HEK293 cells (Morgenstern et al, 2021), albeit in low amount when compared to other MCUC components (Fig. 4A). MICU3 is a mitochondrial protein located in the IMS and loosely attached to the IMM, which can form cysteine-mediated disulfide bonds with MICU1 and acts as an enhancer of MCUC thanks to its ability to sense [Ca²⁺] (Fig. EV2A–I) (Patron et al, 2019). To assess whether MICU3 also played a functional role in the regulation of mt-Ca²⁺ uptake in HEK293 cells, we stably expressed either control (pLKO) or shRNAs targeting different regions of MICU3 mRNA (Fig. 4B). Notably, mitochondria of digitonin-permeabilized sh-MICU3 cells showed reduced Ca²⁺ uptake capacity compared to pLKO upon consecutive addition of exogenous Ca²⁺ (Fig. 4C). The Ca²⁺ clearance phenotype was strongly correlated to MICU3 protein level and therefore not likely due to an off-target effect. Next, we investigated the link between MCU, Tafazzin (TAZ), and the lipid transfer protein PRELI domain-containing protein 1 (PRELID1). Both proteins participate in the biosynthesis of cardiolipin (CL) (Tatsuta and Langer, 2017), which is required for the assembly and stability of numerous IMM protein complexes, including MCUC. Whereas cardiolipin metabolism was shown to affect MCU-dependent Ca²⁺ uptake (Gottschalk et al, 2019; Ghosh et al, 2020), the functional interactions involved have not been reported yet. Notably, we observed that PRELID1 KD resulted in a dramatic reduction of MCU protein level at both monomeric (Fig. EV2J) and oligomeric states (Fig. 4D), confirming its role in the regulation of MCU stability. As a consequence of impaired MCUC assembly, histamine-stimulated si-PRELID1 HeLa cells showed a marked decrease in [Ca²⁺]_mt (Fig. 4E), without any obvious effect on [Ca²⁺]_cyt transients (Fig. EV2K) and mitochondrial membrane potential (Fig. EV2L).

As a third candidate to test we chose EFHD1, a poorly characterized EF-hands mitochondrial protein (Fig. 4F), whose role in mt-Ca²⁺ homeostasis remains controversial (Meng et al, 2023; Hou et al, 2016; Eberhardt et al, 2022). EFHD1 was recently shown to bind MCU in clear cell renal cell carcinoma (ccRCC) (Meng et al, 2023), and it therefore provides a great example of the predictive power of our resource. Indeed, we identified EFHD1 as a high-confidence and common interactor of all three baits suggesting a strong functional connection to MCUC. To test this hypothesis, we first used aequorin as a luminescent Ca²⁺ sensor to quantify [Ca²⁺]_mt and [Ca²⁺]_cyt transients in HeLa cells upon either stable or transient KD of EFHD1 by shRNA and siRNA treatment, respectively. Out of five distinct shRNAs targeting different regions of EFHD1 mRNA, two (sh1 and sh5) strongly decreased EFHD1 protein level (Fig. 4G) and caused a significant up-regulation of [Ca²⁺]_mt in response to histamine stimulation (Fig. 4H). Likewise, we observed higher [Ca²⁺]_mt upon transfection of HeLa cells with two distinct EFHD1-targeting siRNAs (si1, si2) compared to control (Scr) (Figs. EV3A and  4I). In both experimental set-ups, histamine-stimulated [Ca²⁺]_cyt responses remained unaffected by the silencing of EFHD1 (Fig. EV3B,C). We corroborated these results by quantitative and simultaneous single-cell analyses of [Ca²⁺]_mt and [Ca²⁺]_cyt using a mitochondrial matrix-targeted RCaMP and Fura-2 AM, respectively, as fluorescent Ca²⁺ probes. The measurements were performed in the absence of extracellular Ca²⁺ to study the ER-to-mitochondria Ca²⁺ transfer without the involvement of SOCE. Histamine stimulation in si-EFHD1 but not Scr cells triggered a significant increase in both peak and area under the curve (AUC) for the [Ca²⁺]_mt response, without affecting [Ca²⁺]_cyt transients (Fig. 4J). Similar results were observed in the presence of 1.8 mM Ca²⁺ in the extracellular medium, when both agonist-induced ER Ca²⁺ release and SOCE were activated (Fig. EV3D). To evaluate whether the histamine-stimulated mt-Ca²⁺ phenotype was due to EFHD1-dependent changes in the propagation of Ca²⁺ signals from the ER to the mitochondria, we also quantified [Ca²⁺]_mt and [Ca²⁺]_cyt transients upon opening of SOCE channels at the plasma membrane. To this goal, we pre-treated cells with thapsigargin, an inhibitor of the ER-localized Ca²⁺ ATPase (SERCA) pump, to deplete ER stores and activate SOCE. The addition of 1.5 mM of Ca²⁺ to the extracellular medium of si-EFHD1 HeLa cells resulted in [Ca²⁺]_mt transients of higher peak and amplitude compared to Scr, without a concomitant difference in [Ca²⁺]_cyt dynamics (Fig. 4K), consistent with the results obtained with histamine stimulation. Next, to understand whether the regulation of mt-Ca²⁺ uptake by EFHD1 was dependent on Ca²⁺ binding, we generated HeLa cells expressing either wild-type (WT) EFHD1 or a mutant in both EF-hand domains (EFHD1_EF1+2). Although the overexpression of WT EFHD1 did not affect histamine-stimulated [Ca²⁺]_mt response, EFHD1_EF1+2 cells phenocopied the effect of EFHD1 KD, causing an increase in [Ca²⁺]_mt (Fig. 4L). Importantly, we show that the overexpression of WT EFHD1, but not EFHD1_EF1+2, was sufficient to rescue the Ca²⁺ phenotype triggered by EFHD1 loss-of-function (Fig. EV3E). This suggests that the binding of Ca²⁺ to EFHD1 is required to exert an inhibitory function on MCUC.

Altogether, we corroborate a role for MICU3 in the modulation of MCUC activity in HEK293 cells, whereby all MICUs co-exist; we spotlight PRELID1 as a candidate protein linking membrane phospholipid metabolism and [Ca²⁺]_mt homeostasis; we identify EFHD1 as a negative modulator of mt-Ca²⁺ uptake.

To understand the mechanism of EFHD1-dependent inhibition of mt-Ca²⁺ uptake, we measured Ca²⁺ transients in mitochondria of digitonin-permeabilized mt-AEQ HeLa cells upon EFHD1 KD. The addition of exogenous Ca²⁺ triggered a dramatic increase of [Ca²⁺]_mt compared with control pLKO cells, and it was dependent on MCU activity given it was fully abrogated by pre-treatment with the ruthenium red derivative Ru360 (Fig. 5A). Because the impairment of IMM polarization and OXPHOS can also affect mt-Ca²⁺ uptake, we measured bioenergetics and mitochondrial membrane potential in EFHD1 knockdown HeLa cells. The oxygen consumption rate (OCR) of sh-EFHD1 cells at resting conditions, and after treatment with the RCCV inhibitor oligomycin and the uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP) was comparable to control cells, demonstrating that ATP-coupled respiration was unaffected and membrane potential was preserved (Fig. 5B). Similarly, when directly assessing glycolytic function, sh-EFHD1 and pLKO cells were indistinguishable (Fig. 5C) and the mitochondrial membrane potential measured with the potentiometric probe TMRM under both stable shRNA and transient siRNA-mediated silencing of EFHD1 was not affected (Fig. EV4A). These results are consistent with our findings that the basal [Ca²⁺]_mt was not affected by EFHD1 knockdown (Fig. EV4B). Next, we tested whether the inhibitory effect of EFHD1 on MCU-dependent Ca²⁺ uptake was due to potential regulation of MCUC abundance or assembly, as previously shown for other uniporter components (Garbincius and Elrod, 2022). However, we found that neither stable nor transient silencing of EFHD1 in HeLa cells affected the expression of known MCUC subunits (Fig. EV4C). Similarly, BN-PAGE of mitochondria from sh-EFHD1 HeLa cells did not reveal a significant change in the macromolecular profile of the complex compared to control cells (Fig. EV4D), indicating that both stability and assembly of MCUC are preserved in the absence of EFHD1.

To gain insights into the functional relationship between EFHD1 and MCUC, we characterized EFHD1 localization and topology. First, we performed a proteinase K (PK) assay on crude mitochondria isolated from HeLa cells (Fig. 5D). Proteins that are localized to the OMM and are exposed to the cytosol, such as TOM20, were immediately digested upon addition of PK, even in the absence of digitonin, while IMS-facing (TIM23) or matrix-localized (HSP60 and CypD) proteins were protected from PK until subsequent permeabilization of the OMM or IMM, respectively. EFHD1 showed a protein digestion profile comparable to that of TIM23 and consistent with an IMS localization. We obtained similar results when employing a PEGylation assay, in which OMM and IMM were selectively permeabilized by increasing concentrations of digitonin in the presence of maleimide functionalized polyethylene glycol (mPEG) (Fig. EV4E). This reagent selectively adds a ~5 kDa polyethylene glycol (PEG) polymer chain on free thiol groups of cysteines and is small enough to cross the OMM and directly react with IMS proteins. As it was reported for MICU1 (Tsai et al, 2016), EFHD1 was pegylated even in the absence of digitonin, in contrast to the mitochondrial matrix protein, SOD2. Consistent with the lack of a predicted transmembrane domain, we found that EFHD1 was only associated with, but not spanning the IMM, because it was recovered in the soluble fractions of mitochondria from HeLa and HEK293 cells upon carbonate extraction, at both low and high pH, as MICU1 (Fig. 5E). The presence of EF-hand domains, EF-hand-dependent mt-Ca²⁺ uptake regulation, and interaction with MCUC, together with evidence of an IMS localization, raised the hypothesis that EFHD1 functions is linked to other MICUs. Although EFHD1 was unable to oligomerize through cysteine-mediated disulfide bonds (Fig. EV4F) and to affect the formation of MICU1-MICU2 heterodimers, just like MICU2, it was cross-stabilized by MICU1 (Fig. 5F). Thus, we speculated that MICU1 could mediate the binding of EFHD1 to MCUC by affecting its intra-protein stability. Indeed, upon MICU1 silencing we found that the interaction between EFHD1 and MCU-tag was dramatically impaired, with a fourfold reduction in fold change and a loss of significance compared to control (Fig. 5G).

Since the sustained increase of [Ca²⁺]_mt has been shown to promote cell death, we speculated that high EFHD1 expression would protect cancer cells from pro-apoptotic triggers. As shown in Fig. 5H, the KD of EFHD1 in HeLa cells significantly decreased cell viability and sensitized cells to sub-lethal doses of apoptotic inducers such as C2-ceramide and paclitaxel (Fig. 5I). We next set out to evaluate the relevance of EFHD1 more broadly in cancer biology. Indeed, expression profiling of a subset of tumor types, such as clear cell renal cell carcinoma (ccRCC) and colorectal cancer, has recently proposed the expression level of EFHD1 as a prognostic factor and biomarker (Meng et al, 2023). However, a comprehensive assessment of its molecular signature in cancer cells and tissues is currently lacking. To this end, we analyzed publicly available cancer cell lines, primary tumors and patient datasets from the Cancer Cell Line Encyclopaedia (CCLE) of the Dependency Map Consortium (DepMap) (Ghandi et al, 2019), Gene Expression Omnibus of the National Centre for Biotechnology Information (NCBI-GEO), Cancer Genome Atlas (TCGA), Therapeutically Applicable Research to Generate Effective Treatments (TARGET), and Genotype-Tissue Expression (GTEx) repositories (Jiang et al, 2020). Compared to known components of MCUC such as MCU, EMRE, MICU1, MICU2, and MCUB, EFHD1 showed great variability in gene expression across 1450 cancer cell lines from 29 different lineages (Fig. EV5A) but was consistently upregulated in breast, uterine, ovarian, and cervical cancer cells, both at the RNA and protein (Fig. EV5B) level. Moreover, the expression of EFHD1 in primary breast tumor biopsies was significantly higher compared to the adjacent healthy tissues (Fig. EV5C), and breast cancer patients with the highest EFHD1 expression exhibited a decreased response to chemotherapy (Fig. EV5D) and a lower probability of survival (Fig. EV5E). Consistently, downregulation of EFHD1 in two breast cancer cell types, HCC1500 and EFM19, with the highest EFHD1 protein expression compared to the median of all CCLE lines (5.99-fold and 5.91-fold, respectively), resulted in a dramatic decrease in cell viability and proliferation (Fig. 5J,K) that was not accompanied by a change in invasive potential (Fig. EV5F).

Mapping of the MCUC network at resting conditions clearly highlighted that the molecular composition of MCUC and the functional associations between mitochondria and Ca²⁺ are far more complex than previously anticipated (Sancak et al, 2013). Moreover, to what extent genetic perturbations of MCUC impact function, structure, and regulation of mitochondrial processes has not been systematically assessed so far. To this goal, we sought to map the remodeling of MCUC interactome upon LOF of EMRE, MICU1, MICU2, and MCUB. We generated isogenic MCU-tagged HEK293 cell lines stably expressing either short hairpin RNAs (shRNAs) targeting MICU1, MICU2, EMRE and MCUB, or pLKO as a control. As shown in Fig. 6A, we achieved an almost complete KD of each target. As previously reported (Kamer and Mootha, 2014), we observed that MICU1 KD resulted in the concomitant reduction of MICU2 and EMRE protein levels in whole cell lysates, whereas MICU1 expression was not affected by silencing of EMRE, MICU2 or MCUB. Instead, neither MICU2 nor MCUB downregulation altered the overall stability of other MCUC members. We then performed TAP-MS analyses upon tetracycline-inducible expression of StrepII-HA-His₆ tagged MCU in stable pLKO and shRNA-expressing HEK293 cells in at least 4 biological replicates (Dataset EV4). Principal component analysis showed great reproducibility across biological replicates and a clear separation between the interactome of MCU in control (pLKO) versus EMRE and MCUB KD, while clustering together with MICU1 and MICU2 KD, suggesting there are minimal changes to the MCU interaction network upon MICUs KD (Fig. 6B). By quantitative bait-prey co-enrichment analysis and local correlation of protein intensity profiles, we mapped PPIs between MCU and 245 mitochondrial proteins (Dataset EV4). As expected, we observed that the interaction of MICU2 with MCU was dependent on MICU1 (Plovanich et al, 2013; Kamer and Mootha, 2014; Patron et al, 2014) and that silencing of EMRE resulted in the loss of MICUs (Sancak et al, 2013) as significant binding partners (Fig. 6C). Interestingly, although the MICU1-MCU interaction was not affected by MICU2 KD, MICU3 was not significantly co-enriched, suggesting that in absence of MICU2, the MICUs dimers are mostly formed by MICU1. On the contrary, the interaction of MCU with MCUB was always preserved, except upon silencing of MCUB, indicating that MCU and MCUB form a stable complex. Strikingly, the global analysis of all mapped MCUC interactomes spotlighted a dramatic remodeling of PPIs upon LOF of EMRE and MCUB (Fig. 6D,E). Downregulation of EMRE resulted in more than tenfold lower prey recovery. However, although the loss of EMRE leads to non-functional MCU channels (Sancak et al, 2013; Liu et al, 2020; König et al, 2016), we also identified novel and significantly enriched interactions between MCU and proteins of the TCA cycle that are known to be regulated by Ca²⁺, for example, the pyruvate dehydrogenase complex (Fig. 6F). It is tempting to speculate that such PPIs could mediate a compensatory metabolic response aimed at preserving mitochondrial energy production. This could explain why, despite both MCU and EMRE knockout (KO) mouse models displaying a complete loss of mt-Ca²⁺ uptake, only MCU KO mice showed impaired exercise performance (Liu et al, 2020; Pan et al, 2013). In the opposite direction, MCUB KD, which leads to increased [Ca²⁺]_mt, neither affected MCUC protein stability nor the interaction of EMRE and the MICUs with MCU. However, it substantially remodeled the MCU interactome, mostly by allowing the gain of novel interactions. These involved mitochondrial proteins that regulate stress response pathways, cell death, and mitochondrial translation. To corroborate these results, we tested whether a wider protein interaction network would also be reflected in the formation of MCU-containing complexes with higher MW upon sh-MCUB. Blue native-polyacrylamide gel electrophoresis (BN-PAGE) analysis of mitochondria isolated from pLKO MCU-tagged HEK293 cells identified the expected bands around 480 kDa and above 720 kDa and showed that silencing of MICU1 or EMRE caused a shift toward a lower MW (Plovanich et al, 2013; Sancak et al, 2013), while MICU2 LOF did not affect the overall assembly (Plovanich et al, 2013) (Fig. 6G). Instead, stable (shRNA) and transient (siRNA) MCUB KD resulted in the shift of MCU-containing complexes toward a higher MW, both in HEK293 (Fig. 6G) and HeLa cells (Fig. 6H). We observed the same phenotype in mitochondria isolated from the brain of MCUB KO mice, where we and others (Samaras et al, 2020; Hansen et al, 2024) could detect a clear expression of MCUB (Fig. EV6A–C). Moreover, two-dimensional blue native/SDS analysis of whole brain mitochondria from MCUB KO confirmed that MCU-containing complexes were shifted toward a higher MW (Fig. 6I). In summary, our results suggest that the extent to which MCU engages in PPIs within the organelle is determined by channel activity and [Ca²⁺]_mt and that MCUB can act as a protein barrier (Fig. 6J), by preventing PPIs between MCU and several mitochondrial complexes and pathways.

HeLa and HEK293 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS. HeLa cells stably expressing a mitochondrial matrix-targeted GFP-aequorin (mt-AEQ) (Arduino et al, 2017) were cultured in DMEM supplemented with 10% FBS and 100 µg/mL geneticin (Thermo Fisher Scientific, 10131027). EFM19 cells (German Collection of Microorganisms and Cell Cultures GmbH) and HCC1500 cells (American Type Culture Collection) were cultured in RPMI1640 media supplemented with 10% FBS. All cell lines were kept at 37 °C in an incubator with 5% CO₂.

The Flp-In T-REx system was used to generate isogenic, stable HEK293 cell lines exhibiting tetracycline-inducible expression of tagged bait proteins for tandem affinity purification (TAP). The open reading frame (ORF) of each bait protein (MCU, MCUB, and EMRE) was cloned into the pcDNA5/FRT/TO Flp-In expression vector, in frame with a C-terminal StrepII₂-HA-His₆-tag and under the control of a tetracycline-regulated CMV/TetO2 promoter. The resulting MCU, EMRE, or MCUB Flp-In expression vectors were co-transfected with the Flp recombinase vector, pOG44, into Flp-In T-REx HEK293 cells, which contained a single integrated FRT site and stably expressed the lacZ-ZeocinTM fusion gene (pFRT/lacZeo) and the tetracycline repressor (pcDNA6/TR). This allowed the targeted integration of each bait protein at a single transcriptionally active genomic locus that is the same in every cell line. Flp-In T-REx HEK293 cells were grown in DMEM with 10% FBS, containing blasticidin (15 µg/mL) and Zeocin (100 µg/mL). Stable transfectants were selected and maintained in DMEM with 10% FBS containing blasticidin (15 µg/mL) and 100 µg/mL hygromycin. The expression of tagged bait proteins was induced by the addition of tetracycline (1 µg/ml) to a growth medium lacking hygromycin and blasticidin, 24 h prior to harvest.

Human cell lines stably expressing shRNAs for gene-specific knockdown (KD) were generated as previously described (Perocchi et al, 2010). The following shRNA constructs were used for MCU (sh-MCU, TRCN0000133861); EMRE (sh-EMRE, TRCN0000145067); MICU1 (sh-MICU1, TRCN0000053370); MICU2 (sh-MICU2 TRCN0000055848); MICU3 (sh1-MICU3, TRCN0000056083; sh2-MICU3, TRCN0000056084; sh3-MICU3, TRCN0000056085; sh4-MICU3, TRCN0000056086; sh5-MICU3, TRCN0000056087); EFHD1 (sh1-EFHD1, TRCN0000056183; sh2-EFHD1, TRCN0000056184; sh3-EFHD1, TRCN0000056185; sh4-EFHD1, TRCN0000056186; sh5-EFHD1, TRCN0000056187); and MCUB (sh-MCUB, TRCN0000128550). Infected HeLa cells were selected with 2 µg/mL puromycin. EFM19 and HCC1500 cells were infected with pLKO and sh5-EFHD1 viruses and selected with 1 µg/mL puromycin. The same number of cells for each condition (50,000 for EFM19; 25,000 for HCC1500) were seeded in a 3 cm culture plate and imaged 10 days after seeding, with media changes every 48 h. For siRNA-mediated KD, 50 nM of negative control scramble or targeting siRNAs (Scr: MISSION siRNA #1 SIC001; si1-EFHD1: SASI_Hs01_00228164; si2-EFHD1: SASI_Hs01_00228165, si-MCUB: 5’-AUACUACCAGUCACACCAU-3’, si-PRELID1: 5’-CCCGAAUCCCUAUAGCAAA-3’, si-MICU1: 5’-UCUGAAGGGAAAGCUGACAAU-3’) were transfected using Lipofectamine RNAiMAX (Thermo Fisher Scientific, 13778075). Assays were normally carried out 48 h after transfection, unless noted otherwise. For the exogenous expression of MICU3 and EFHD1 ORFs in human cells, MICU3 (HsCD00296366) and EFHD1 (HsCD00719312) WT clones without a STOP codon in pDONR221 were purchased from DNASU, whereas MICU3 and EFHD1 EF-hand mutants were synthesized in the pUC57 vector (Thermo Fisher Scientific, SD0171) and cloned into a pDONR221 vector with the following primers: attB1-FW-MICU3, GGGGACAAGTTTGTACAAAAAAGCAGGCTTAGCCACCATGGTGGCTCGAGGGCT; attB2-RV-MICU3, GGGGACCACTTTGTACAAGAAAGCTGGGTTTCTGCTGTGAAGTTCTTTCTTCAGG; attB1-FW-EFHD1, GGGGACAAGTTTGTACAAAAAAGCAGGCTTAGCCACCATGGCCAGTGAGGAGCTG; attB2-RV-EFHD1, GGGGACCACTTTGTACAAGAAAGCTGGGTTTGTATTGAAGTTGGCCTTGAGTTT. Constructs were then cloned either into pcDNA-DEST40 Vector (Thermo Fisher Scientific, 12274015) or pLX304 for expression in mammalian cells in frame with a C-terminal V5-His₆ and V5 tag, respectively, by gateway cloning according to manufacturer instructions. Lentivirus production and infection were performed according to guidelines from the Broad RNAi Consortium and infected cell lines were selected 48 h post-transduction with the respective selection markers. Transient protein expression was performed using X-tremeGENE^TM HP DNA transfection reagent following the manufacturer’s instructions for a 1:3 DNA/reagent ratio. Assays were normally carried out 48 h after transfection, unless noted otherwise.

To monitor endogenous and overexpressed proteins, cells or isolated mitochondria were lysated in RIPA buffer (150 mM NaCl, 50 mM Tris, 1 mM EGTA, 1% NP-40, 0.1% SDS); after 30 min of incubation on ice, the lysates were centrifuged at 15,000×g for 10 min to remove debris. 20 µg of total proteins were loaded, according to BCA quantification (Thermo Fisher Scientific, 23225) for each lane. Proteins were reduced with Laemmli buffer (Bio-Rad, 1610747) supplemented with 2.75 mM β-mercaptoethanol (Thermo Fisher Scientific, 21985-023) and denatured for 5 min at 90 °C, unless otherwise specified. To visualize MICU1-MICU2 or MICU1-MICU3 heterodimers, protein samples were denatured at 70 °C for 10 min in LDS Sample Buffer (Thermo Fisher Scientific, NP0007) with or without 100 mM Dithiothreitol ( + DTT and -DTT, respectively), as previously performed (Patron et al, 2014). Proteins were separated by SDS-PAGE electrophoresis in 12% or 14% acrylamide gels, and transferred on nitrocellulose membranes (Amersham, 10600021) by semi-dry electrophoretic transfer (Bio-Rad). Accordingly with primary antibody datasheets, blots were blocked 1 h at room temperature (RT) with 5% non-fat dry milk (Carl Roth, T145.2) or 5% bovine serum albumin (BSA, Sigma-Aldrich, A7906) in TBS-Tween (0.5 M Tris, 1.5 M NaCl, 0.01% Tween) solution and incubated overnight at 4 °C with primary antibodies. Horseradish peroxidase-conjugated secondary antibodies (Bio-Rad, 1706515 or 1706516), diluted in TBS-Tween containing 0.5% BSA, were incubated for 1 h at RT followed by detection by chemiluminescence (Amersham, RPN2236). The expression level of specific proteins was detected by immunoblot analysis with the following antibodies: HA (BioLegend, MMS-101R), V5 (Thermo Fisher Scientific R96025), ACTIN (Sigma-Aldrich, A2228), MAIP1 (König et al, 2016), ATP5A (Abcam MS507), Lamin (Santa Cruz, sc-6217), VDAC (Abcam, ab14734), EFHD1 (Sigma-Aldrich HPA056959), HSP60 (R&D System MAB1800), SOD2 (Antibody Verify Inc. AAS29585C), TIM23 (BD bioscience, 611222), CyPD (Abcam, ab110324), GRP75 (Santa Cruz, sc-133137), MCU (Sigma-Aldrich, HPA016480), EMRE (Santa Cruz, sc-86337), MCUB (Santa Cruz, sc-163985), MICU1 (Sigma-Aldrich, HPA037479), MICU2 (Sigma-Aldrich, HPA045511), MICU3 (Sigma-Aldrich, HPA024048), TOM20 (Abcam, ab56783). Densitometry analysis of protein bands was performed with ImageJ by subtracting background signal and normalizing the area of each peak intensity to actin.

Isogenic, parental Flp-In T-REx HEK293 cells stably expressing tagged MCU, MCUB, or EMRE as baits, as well as Flp-In T-REx HEK293 cells expressing tagged MCU with stable knockdown of either MICU1, MICU2, EMRE, or MCUB were expanded into two expression plates (245 × 245 mm, NUNC, 9407400) to obtain roughly 250 million cells. For TAP, each cell pellet was resuspended in 2 mL lysis buffer (50 mM HEPES-KOH pH 7.4, 150 mM KCL, 5 mM EGTA, 5% Glycerol, 3% Digitonin) and incubated at 4 °C in rotation for 30 min. Lysates were centrifuged at 10,000×g for 5 min at 4 °C and the supernatants were incubated for 20 min at RT on a rolling shaker with 50 µL of a 500 mM HEPES solution at pH 8.0 containing avidin (10 µM final concentration). In the meantime, 150 µL of streptavidin resin was washed three times with 500 µL washing buffer (50 mM HEPES-KOH pH 7.4, 150 mM KCl, 5 mM EGTA, 5% Glycerol, 0.02% Digitonin), combined with the supernatant, and incubated 45 min at 4 °C in rotation. Resins were then washed three times, before eluting bound proteins with 250 µL of elution buffer (50 mM HEPES-KOH pH 8, 150 mM KCl, 5 mM EGTA, 5% glycerol, 10 mM Biotin), incubating on a rolling shaker at 900 rpm for 5 min at RT. The elution step was repeated 4 times and a total of 1 mL eluate was collected. Next, 50 µL of Ni-NTA resin was washed three times with 500 µL of washing buffer and incubated with the eluate for 45 min at 4 °C in rotation. The resin was then collected by quick spin and washed three times with 500 µL of washing buffer followed by two washing steps in washing buffer without detergent. Purified complexes were eluted through three steps of incubation with 35 µL of 1% RapiGest on a rolling shaker for 10 min at 900 rpm at RT. Eluates were pulled and subjected to acetone precipitation overnight. The next day, samples were centrifuged at 20,000×g for 30 min at 4 °C, the supernatant was removed, and the pellet was stored at −80 °C for LC-MS/MS analysis.

Eluates were resuspended in 50 µL denaturation buffer (6 M urea, 2 M thiourea, 10 mM HEPES pH 8.0, 10 mM DTT) for 30 min at RT before adding alkylation agent (55 mM iodoacetamide) and incubating at RT in the dark for 20 min. Proteins were digested at RT for 3 h by adding LysC at 1:100 ratio of enzyme:protein, before diluting the sample with 50 mM ammonium bicarbonate to reach a urea concentration of 2 M. Subsequently, samples were digested at RT overnight by adding Trypsin at 1:100 ratio of enzyme:protein. The next day, the digestion was stopped by acidifying the sample with 10 µL of 10% trifluoroacetic acid and the final peptides were cleaned up using SDB-RPS StageTips as described (Kulak et al, 2014). MS analysis was performed using Q Exactive HF mass spectrometers (Thermo Fisher Scientific, Bremen, Germany) coupled online to a nanoflow ultra-high-performance liquid chromatography instrument (Easy1000 nLC, Thermo Fisher Scientific). Peptides were separated on a 50 cm long (75 μm inner diameter) column packed in-house with ReproSil-Pur C18-AQ 1.9-μm resin (Dr. Maisch GmbH, Ammerbuch, Germany). Column temperature was kept at 50 °C. Peptides were loaded with buffer A (0.1% (v/v) formic acid) and eluted with a nonlinear gradient of 5–60% buffer B (0.1% (v/v) formic acid, 80% (v/v) acetonitrile) at a flow rate of 250 nL/min. Peptide separation was achieved by 120 min gradients. The survey scans (300–1650 m/z, target value = 3E6, maximum ion injection times = 20 ms) were acquired at a resolution of 60,000 followed by higher-energy collisional dissociation (HCD) based fragmentation (normalized collision energy = 27) of up to 15 dynamically chosen most abundant precursor ions. The MS/MS scans were acquired at a resolution of 15,000 (target value = 1E5, maximum ion injection times = 60 ms). Repeated sequencing of peptides was minimized by excluding the selected peptide candidates for 20 s.

Data analysis was carried out using the MaxQuant software package 1.5.5.2. The false discovery rate (FDR) cutoff was set to 1% for protein and peptide spectrum matches. Peptides were required to have a minimum length of 7 amino acids and a maximum mass of 4600 Da. MaxQuant was used to score fragmentation scans for identification based on a search with an initial allowed mass deviation of the precursor ion of a maximum of 4.5 ppm after time-dependent mass calibration. The allowed fragment mass deviation was 20 ppm. Fragmentation spectra were identified using the UniprotKB Homo sapiens database, combined with 245 common contaminants by the integrated Andromeda search engine. Enzyme specificity was set as C-terminus to arginine and lysine, also allowing cleavage before proline, and a maximum of two missed cleavages. Carbamidomethylation of cysteine was set as fixed modification and N-terminal protein acetylation as well as methionine oxidation as variable modifications. Both “label-free quantification (MaxLFQ)” with a minimum ratio count of 1 and “match between runs” with standard settings were enabled. Protein copy number estimates were calculated using the iBAQ algorithm, in which the sum of all tryptic peptides intensities for each protein is divided by the number of theoretically observable peptides. The mass spectrometric data have been deposited via PRIDE (Vizcaíno et al, 2013) to the ProteomeXchange Consortium under the accession number PXD040893.

Basic data handling, normalization, statistics, and annotation enrichment analysis was performed with the Perseus software package (1.6.15.0 release) (Tyanova et al, 2016), R (4.1.2 release) and GraphPad Prism (10.0 release). The label-free quantification (LFQ) module of the MaxQuant software (Cox and Mann, 2008; Cox et al, 2014) was used to define specific proteins that were quantitatively enriched with a given bait over all measured samples. Protein groups were filtered removing hits to the reverse decoy database and proteins only identified by modified peptides. Mitochondrial proteins were filtered using a curated list of 1276 proteins that integrates information from MitoCarta3.0 (Rath et al, 2021), Uniprot annotated mitochondria proteins, and IMPI (Smith and Robinson, 2019). It was required that each protein was quantified in at least two biological replicates from TAPs of each cell line to be considered for analysis. Protein LFQ intensities were log-transformed and missing values were imputed by values sampled from a normal distribution shifted 1.8 standard deviation and with a width 0.3 standard deviations from the distribution of all protein intensities within each sample as the background. Protein interactors were identified by volcano plot and protein correlation analyses. In each volcano plot, a quantitative bait-prey co-enrichment analysis was performed based on a two-tailed Welch’s t test using the multiple volcano (Hawaii) plot option of Perseus (version 1.6.2.0) and a permutation-based FDR cutoff of 0.05 (Class A) and 0.10 (Class B) and S0 = 1 (Hein et al, 2015). However, proteins that do not cross any FDR threshold can still represent true positive interactors. Therefore, to be comprehensive, we used the correlation coefficients of interacting proteins as an additional qualifier to the FDR-controlled confidence of the volcano plot. In protein–protein interaction studies, true positive interacting proteins often exhibit a good correlation of their intensity profiles across different samples, as demonstrated in our previous studies (Hein et al, 2015; Michaelis et al, 2023). Protein intensity profiles across each pair of bait and control TAPs were used to calculate a Pearson’s correlation coefficient between baits and preys (local correlation). A protein was defined as a specific interactor when having a correlation coefficient higher or equal than 0.6 based on the mean correlation of Class A interactors identified by volcano analysis (0.53) (Class C). Similarly, specific interactors were identified based on a Pearson’s correlation analysis across control, MCU-, EMRE- and MCUB-TAPs (global correlation) using a permutation-based FDR of 0.05 (Class D). For the remodeling of MCUC interaction network upon genetic perturbation, Class A, B, and C parameters were used to define MCU interactors. For sh-MCUB a permutation-based FDR cutoff of 0.025 and 0.05 for class A and B, respectively, and a correlation coefficient higher or equal to 0.8 for class C, were used.

MCU, MCUB, and EMRE protein–protein interactions were visualized using Cytoscape 3.10.0 (Shannon et al, 2003). For each MCUC interactor, biological function, the number of transmembrane domains, submitochondrial localization, and gene-disease associations were retrieved based on MitoCop (Morgenstern et al, 2021), literature-based manual curation, DeepTMHMM (Hallgren et al, 2022), MitoCarta3.0 (Rath et al, 2021), and DisGeNET (Piñero et al, 2021). By uncharacterized proteins, we refer to proteins that lack single gene-based experimental evidence but whose function is inferred simply through large-scale analysis or sequence similarity. Protein expression level of MCUC interactors across 201 samples from 32 different tissue types of normal human individuals was obtained from GTEx (Jiang et al, 2020). The tissue numbers were reduced by removing some samples (Artery–Coronary, Esophagus–Mucosa, Minor Salivary Gland, Nerve–Tibial, Pituitary, and Skin) and by averaging tissues belonging to the same organ (arteries, colon segments, esophagus segments, and heart compartments), obtaining 21 tissues in total. Orthologs of MCUC interactors were identified across 120 eukaryotic species and one prokaryotic species (E. coli) used as outgroup and are common to CLIME (Li et al, 2014) and ProtPhylo (Cheng and Perocchi, 2015). Orthologs were defined using OBH (one-way best hit) as in (Cheng and Perocchi, 2015). The NCBI taxonomy database and the R package taxize were used to build the species tree. Hierarchical clustering was performed using the Euclidean distance with complete linkage. Significant interactors were analyzed using gProfiler’s GOSt tool (Raudvere et al, 2019). Significance was established using the algorithm gSCS as a multiple-testing correction method with a significance threshold of 0.05. Only GO biological process driver terms significant in at least one condition were taken in consideration.

Measurement of extracellular Ca²⁺ clearance by mitochondria from digitonin-permeabilized HEK293 cells was performed using the membrane-impermeable Ca²⁺ indicator Calcium-Green-5N (Life technologies, C3737). Cells were harvested at a density of 500,000 cells/mL in growth medium supplemented with 20 mM HEPES (pH 7.2/NaOH). Cells were collected by centrifugation at 300×g for 3 min at RT, resuspended in extracellular-like media (145 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 10 mM Glucose, 10 mM HEPES, pH 7.4) containing 200 nM thapsigargin and incubated for 10 min under constant agitation at RT. Cells were then resuspended in intracellular-like buffer (140 mM KCl, 1 mM KH₂PO₄/K₂HPO₄, 1 mM MgCl₂, 20 mM HEPES, 100 µM EGTA, pH 7.2/KOH), supplemented with 1 mM Na⁺-pyruvate, 1 mM of equimolar solution of ATP/MgCl₂ and 2 mM Na⁺-succinate at a density of 2.5 × 10⁶ cells/mL and the plasma membrane was permeabilized by incubation with 60 µM digitonin for 5 min at RT under constant agitation. 100 µL of the cell suspension was seeded into a black 96-well plate (PerkinElmer) and Calcium-Green-5N fluorescence (excitation 506 nm, emission 531 nm) was monitored every 2 s at RT using a CLARIOstar microplate reader (BMG Labtech) after injection of 40 µM CaCl₂. The MCU inhibitor Ru360 (10 µM) was used as a positive control.

Measurements of Ca²⁺ transients in HeLa cells were performed using the luminescence Ca²⁺ indicator aequorin as previously described (Arduino et al, 2017). Briefly, HeLa mt-AEQ cells infected with lentivirus carrying the specific shRNA were seeded in white 96-well plates (PerkinElmer, 6005181) at 25,000 cells/well. After 24 h, aequorin was reconstituted with 2 µM native coelenterazine for 1–2 h at 37 °C. For measurements of Ca²⁺ kinetics upon siRNA-mediated EFHD1 knockdown, cells were transfected using a transfection mix including Lipofectamine RNAiMAX transfection reagent (Thermo Fisher Scientific, 13778075), 0.5 µg of mt-AEQ (Rizzuto et al, 1992) or cytosolic aequorin (Brini et al, 1995) cDNAs, and either a final concentration of 50 nM siRNA. Mt-AEQ-based measurements of Ca²⁺-dependent light kinetics were performed upon 100 µM histamine stimulation. Light emission was measured either using the luminescence counter MicroBeta2 LumiJET Microplate Counter (PerkinElmer) or the PerkinElmer Envision plate reader at 469 nm every 0.1 and 1 s, respectively. For measurements of Ca²⁺ kinetics upon EFHD1-WT or EFHD1_(EF1+EF2) overexpression, cells were transfected using TransIT-2020 Transfection Reagent (Mirus Bio) with 1 ug of EFHD1 cDNA and 0.5 µg of mt-AEQ or cytosolic aequorin. Ca²⁺ kinetics upon store-operated Ca²⁺ entry (SOCE) were measured upon pre-treatment with the irreversible SERCA inhibitor thapsigargin (100 nM) for 10 min in Ca²⁺-free modified Krebs-Ringer Buffer (135 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 0.4 mM KH₂PO₄, 1 mM MgSO₄, 20 mM HEPES, 600 μM EGTA, 10 mM glucose, pH 7.4 at 37 °C) followed by perfusion with the same medium without thapsigargin but supplemented with 1.5 mM CaCl₂. For the EFHD1 rescue expression experiment, HeLa cells were transiently transfected using a calcium phosphate transfection protocol. Briefly, for each well of a 6-well plate, a total of 8 µg of plasmid DNA (2 µg of mt-AEQ and 6 µg split between the other constructs) was diluted in 90 µL of water and 10 µL of 2.5 M CaCl₂ and mixed with an equal volume of 2xHBS (50 mM HEPES, 280 mM NaCl, 1.5 mM Na₂HPO₄, pH 7.06) to induce the formation of Ca²⁺ phosphate-DNA precipitates. The prepared transfection mix was applied to the cells for 8 h. Following this incubation, cells were washed three times with PBS to remove any residual precipitates. Twenty-four hours before the mitochondrial Ca²⁺ measurement, the transfected HeLa cells were trypsinized, and 25,000 cells were seeded into each well of a 96-well plate. The measurement of mt-Ca²⁺ uptake in digitonin-permeabilized mt-AEQ HeLa cells was performed as previously described (Wettmarshausen et al, 2018). Briefly, HeLa cells stably expressing mt-AEQ were harvested at a density of 500,000 cells/mL in growth medium supplemented with 20 mM HEPES (pH 7.4/NaOH) and the aequorin was reconstituted by incubation with 2 µM native coelenterazine n (Biotium, BOT-10115-1) for 2.5 h at RT. Cells were then centrifuged at 300×g for 3 min and the pellet was resuspended in an extracellular-like buffer (145 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 10 mM Glucose, 10 mM HEPES, pH 7.4) containing 200 nM thapsigargin and incubated for 20 min under constant agitation at RT. The cells were collected by centrifugation at 300×g for 3 min and the pellet was resuspended in an intracellular-like buffer supplemented with 1 mM Na⁺-pyruvate, 1 mM of equimolar solution of ATP/MgCl₂ and 2 mM Na⁺-succinate. The cells were permeabilized for 5 min with 60 µM digitonin, collected by centrifugation for 3 min at 300×g and resuspended in intracellular-like buffer at a density of 900 cells/µL. Then, 90 µL of cell suspension was dispensed into a white 96-well plate (PerkinElmer). Cells were incubated for 5 min at RT and light signal was recorded at 469 nm every 0.1 s using a luminescence counter (MicroBeta2 LumiJET Microplate Counter, PerkinElmer) after injection of CaCl₂ to achieve a free Ca²⁺ concentration of 5 µM in solution. Ru360 (10 µM) was used as a positive control. All the luminescence signals were converted in [Ca²⁺] values according to the Ca²⁺ response curve of aequorin, as previously performed (Brini et al, 1999).

For single-cell measurement of Ca²⁺ transient, HeLa cells were treated with siRNA for 48 h before the experiment. Cells were transiently transfected with mitochondrial matrix-targeted RCaMP to measure changes in [Ca²⁺]_mt then plated on Poly-D-Lysin coated 25 mm coverslips (Thermo Fisher Scientific, 25CIR-1.5). To measure changes in the [Ca²⁺]_cyt cells were loaded with 2 μM Fura2AM (Moleculer probes, F-1221) in 2% BSA containing extracellular medium (ECM, 121 mM NaCl, 5 mM NaHCO₃, 10 mM Na-HEPES, 4.7 mM KCl, 1.2 mM KH₂PO₄, 1.2 mM MgSO₄, 2 mM CaCl₂, and 10 mM glucose, pH 7.4) in the presence of 0.003% Pluronic F-127 (Thermo Fisher Scientific, P6867) and 150 μM sulfinpyrazone (Sigma-Aldrich, S9509) for 15 min at 35 °C. After dye-loading, cells were washed with fresh 0.25% BSA containing ECM with or without Ca²⁺ and transferred to the temperature-controlled stage (37 °C) of an Olympus IX81 motorized inverted epifluorescence microscope fitted with a Hamamatsu ORCA-Flash 4.0v3 sCMOS camera, high-speed excitation switching by Sutter Lambda 421 LED illuminator and UV-optimized Olympus UAPO/340 ×40/1.35NA oil immersion objective. For simultaneous measurements of [Ca²⁺]_cyt and [Ca²⁺]_mt Fura-2 fluorescence was recorded at 340 and 380 nm, and RCaMP at 577 nm excitations, using dual-band Chroma 59022bs dichroic and 59022 m emission filter. Image triplets were acquired every 0.33 s.

C57BL/6n WT mice or C57BL/6n MCUB KO mice (Feno et al, 2021) were housed in a pathogen-free, temperature- and humidity-controlled animal facility on a 12:12 h light-dark cycle. Diet consisted of standard laboratory chow and double-distilled water. All animal procedures were in accordance with the European Community Council Directive for the Care and Use of Laboratory Animals (86/609/ECC) and German Law for Protection of Animals and were approved by the local authorities. All experiments were performed with female mice that were at least 3 months old.

Mitochondria were isolated by nitrogen cavitation as previously described (Wettmarshausen and Perocchi, 2017). Briefly, HeLa and HEK293 cells were grown to confluency in 600 cm² cell culture plates. Culture medium was removed, and cells were rinsed with 30 mL PBS, scraped down and resuspended in 5 mL PBS. After 5 min of centrifugation at 600×g at 4 °C, the cell pellet was resuspended in ice-cold isolation buffer (IB; 220 mM mannitol, 70 mM sucrose, 5 mM HEPES-KOH pH 7.4, 1 mM EGTA-KOH, pH 7.4), supplemented with protease inhibitor (Sigma-Aldrich, 5056489001). Cell suspension was immediately subjected to nitrogen cavitation at 800 psi for 10 min at 4 °C. Nuclei and intact cells were pelleted by centrifugation at 600×g for 10 min at 4 °C. Supernatants were transferred into new tubes and centrifuged at 8000×g for 10 min at 4 °C. The resulting pellet containing crude mitochondria was resuspended in 50-200 µL IB for further analyses. For mouse brain mitochondria isolation, the tissue was homogenized with two strokes at 300 rpm using a loose-fitting Teflon homogenizer followed by nitrogen cavitation at 800 psi for 10 min in ice-cold isolation buffer supplemented with 0.5% fatty acid-free BSA (Sigma-Aldrich, A7030) and protease inhibitor (Sigma-Aldrich, 5056489001). Nuclei and intact cells were pelleted by centrifugation at 600×g for 10 min at 4 °C. Supernatants were transferred into new tubes and centrifuged at 12,000×g for 10 min at 4 °C. The centrifugation step was repeated, changing the buffer with IB without BSA. The final pellet was resuspended in IB without BSA and stored on ice for further use. During the isolation, whole cell lysate (WCL), as well as cytosolic (C), nuclear (N), and crude mitochondrial (M) fractions, were collected for immunoblot analysis.

Alkaline carbonate extractions at pH 10, pH 11, or pH 12 from crude mitochondria were performed as previously described (Wettmarshausen et al, 2018) to analyze membrane association and submitochondrial localization of proteins. Briefly, 100 µg of mitochondria were centrifuged at 8000×g for 10 min at 4 °C and then resuspended in 0.1 M Na₂CO₃ at pH of 10, 11, or 12 and incubated for 30 min on ice. Afterward, the samples were centrifuged at 45,000×g for 10 min at 4 °C. Pellets resulting from this process were resuspended in 100 µL of 2× Laemmli buffer, boiled at 98 °C for 5 min, and stored at −80 °C for subsequent use (referred to as the membrane sample). Supernatants were combined with 40 µL of 100% trichloroacetic acid and left to incubate overnight at −20 °C. The next day, the supernatants were centrifuged at 16,000×g for 25 min at 4 °C. Pellets were washed twice with cold acetone, air-dried for 20-30 min at RT, resuspended in 100 µL of 2× Laemmli buffer, and heated to 98 °C for 5 min (referred to as the soluble sample). SDS-PAGE analysis was performed on 25 µL of both the soluble and membrane samples. Antibodies against MICU1 (soluble, membrane-associated), ATP5a (soluble, membrane-associated), MCU (integral transmembrane protein) were used as positive controls. To determine the submitochondrial localization of MICU3 and EFHD1, 30 µg of crude mitochondria were exposed to increasing concentrations of digitonin or 1% Triton X-100, which sequentially permeabilize the outer and inner membranes. This was conducted in the presence of 5 mM membrane-impermeable maleimide functionalized polyethylene glycol (mPEG, Sigma-Aldrich, 63187). This compound attaches a polyethylene glycol polymer chain of approximately 5 kDa to free thiol groups of proteins. The reaction was carried out at RT for 30 min and was quenched with 100 µM cysteine on ice for 10 min. Immunoblot analysis was performed on the samples, with the intermembrane space protein MICU1, and the matrix soluble proteins HSP60 and SOD2, serving as controls. The proteinase K protection assay was performed by incubating 30 µg of mitochondria in 30 µL of isolation buffer in the presence of 100 µg/mL proteinase K with increasing concentrations of digitonin or 1% Triton X-100 to sequentially permeabilize outer and inner membranes. The reaction was carried out at RT for 15 min and was stopped by the addition of 5 mM phenylmethylsulfonyl fluoride (PMSF), followed by incubation on ice for 10 min. Immunoblot analysis was used to examine the samples, with TOM20 and TIM23 (integral outer and inner membrane proteins, respectively), along with cyclophilin D (CyPD) and HSP60 (soluble matrix proteins) as controls.

Blue-Native (BN) PAGE analysis was performed as described by Witting et al (Wittig et al, 2006) with some adaptations. Briefly, equal amounts of mitochondria (10 µg per lane, unless noted otherwise) were diluted at least 1:100 in ice-cold miliQ water with proteinase inhibitors (Sigma-Aldrich, 5056489001), the pellet was then collected at 20,000×g for 10 min at 4 °C, resuspended in ice-cold 1× BN Sample Buffer (Thermo Fisher Scientific, BN2003) with 1% (w/v) digitonin and incubated on ice for 15 min. Afterward, the sample was centrifuged at 20,000×g at 4 °C for 30 min and the supernatant was transferred into a new pre-chilled tube with 0.25% G-250 (Thermo Fisher Scientific, BN2004). BN-PAGE was performed at 4 °C on NativePAGE Novex 3-12% Bis-Tris Protein gels (Thermo Fisher Scientific, BN1001) according to manufacturer’s instructions, followed by overnight wet blot transfer at 30 V and 4 °C onto a 0.2 µM pore size PVDF membrane (Amersham, GE10600021). Immunoblotting was performed according to the standard protocol. Second dimension (2D) analysis was performed as described by Na Ayutthaya et al (Na Ayutthaya et al, 2020) with some modifications. Following the above described first dimension (1D) BN-PAGE, sample lanes for 2D were excised, incubated in 1× SDS sample buffer (Bio-Rad, 1610747) for 10 min and boiled shortly, followed by incubation in hot 1× SDS sample buffer for 15 min. As a control, one well was loaded with 5 µg of input mitochondria previously diluted in 5 µL of 2× LDS dye (Thermo Fisher Scientific, 84788) with 2.75 mM β-mercaptoethanol (Thermo Fisher Scientific, 21985-023) and RIPA buffer (total volume 10 µL) and boiled for 5 min at 95 °C. The 2D well of the pre-cast NuPAGE 4%-12% Bis-tris 2D well gel (Thermo Fisher Scientific, NP0326BOX) was washed and filled with 1× MOPS running buffer (Thermo Fisher Scientific, NP0001). Lanes were then fitted onto the 2D and overlaid with 1× SDS sample buffer, followed by electrophoresis according to the manufacturer’s instructions. The proteins were then transferred by wet blot transfer on 0.2 µM pore size PVDF membrane (Amersham, GE10600021) and immunoblotted using MCU (Sigma-Aldrich, HPA016480) according to the standard protocol.

RNA was isolated with the RNeasy Mini kit (Qiagen, 74104), according to manufacturer instructions. An equal amount of RNA from each sample was used to generate complementary DNA (cDNA) with the SuperScript™ III First-Strand Synthesis SuperMix kit (Thermo Fisher Scientific, 18080400). The resulting cDNA was diluted 1:8 in nuclease-free water and analyzed by real-time PCR using the following TaqMan assays: Hs00368816 (EFHD1) and Hs01003267 (HPRT1, used as control).

HeLa sh5-EFHD1 and pLKO mt-AEQ cells were seeded at a density of 20000 cells/well in a Seahorse XFe96 Cell Culture Microplate 24 h before the experiment in DMEM with 10% FBS. On the day of the experiment, medium was removed, and cells were washed twice with 200 µL of the respective assay medium for the mito-stress test (Agilent Base Medium supplemented with 1 mM pyruvate, 2 mM l-glutamine, 10 mM glucose, pH 7.4) or glycolysis (Agilent Base Medium supplemented with 2 mM l-glutamine, pH 7.4). Next, 180 µL of the assay medium for mito-stress test or glycolysis were added to each well and the plate was incubated for 60 min in a non-CO₂ incubator at 37 °C. For the mito-stress test, 10× port solution of the following compounds were injected to reach the specified final concentration: oligomycin (1.5 µM), carbonyl cyanide m-chlorophenyl hydrazone (CCCP; 1.5 µM), and antimycin A/rotenone (4 µM/2 µM). For the glycolysis assay, glucose (10 mM), oligomycin (1.5 µM), and 2-DG (100 mM). The CyQUANT® Cell Proliferation Assay Kit (Thermo Fisher Scientific, C7026) was used to normalize differences in cell density between wells. Briefly, growth media was removed, and the plate was frozen at −80 °C. Cells were thawed and lysed with 200 µL of dye/lysis buffer. After 5 min, fluorescence was measured at 480ex/520em using a CLARIOstar plate reader (BMG Labtech). The raw fluorescence of each well was divided by the plate average and the resulting factor was used to normalize the Seahorse data.

Mitochondrial membrane potential was measured as previously described (Vecellio Reane et al, 2016). Briefly, HeLa cells were incubated with 20 nM tetramethyl rhodamine methyl ester (TMRM, Thermo Fisher Scientific) for 30 min at 37 °C in Krebs-Ringer modified buffer (135 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 0.4 mM KH₂PO₄, 1 mM MgSO₄, 20 mM HEPES, 10 mM glucose, 1 mM CaCl₂, pH 7.4 at 37 °C). Images were taken on an inverted microscope (Zeiss Observer.Z1) equipped with an Apochromat ×40/1.1 NA water immersion objective, an AxioCam HRm, and a Zeiss HXP 120 C Fluorescence Light Source. TMRM excitation and emission were performed through the Zeiss filter set 43 HE (BP550/25 FT570 BP605/70). Images were taken every 5 s with a fixed 100 milliseconds (msec) of exposure time. In all conditions, 10 µM CCCP (carbonyl cyanide p-trichloro-methoxyphenylhydrazone), was added after ten acquisitions to completely collapse the electrical gradient across the IMM. After background correction, the fluorescence values after CCCP addition (i.e., TMRM fluorescence not due to membrane potential) was subtracted for each cell.

For the measurement of [Ca²⁺]_mt in resting conditions, sh5-EFHD1 HeLa cells were transfected with 4mtGCaMP6f. 48 h later, coverslips were transferred to an imaging chamber and incubated in 1 mL of modified KRB buffer supplemented with 5.5 M glucose and 1 mM CaCl₂. Imaging was performed on an inverted microscope (Zeiss Axiovert 200), equipped with a 40×/1.3 N.A. PlanFluor objective. Excitation was performed with a Deltaram V high-speed monochromator (Photon Technology International) equipped with a 75 W Xenon Arc lamp. Images were captured using an Evolve 512 Delta EMCCD (Photometrics), and the system was controlled by MetaMorph 7.5 (Molecular Devices), assembled by Crisel Instruments. As previously performed (Butera et al, 2021), we used the fluorescence measurement at the Ca²⁺-independent isosbestic point at 410 nm to normalize the differences in the expression of the probes and in the specimen focus plane. For this reason, the cells were alternatively illuminated every second at 485 and 410 nm, and fluorescence was collected through a 525/50 filter (Chroma) mounted on an OptoSpin25 wheel (Cairn research). The exposure time was 100 msec. Analysis was performed with the Fiji distribution of ImageJ. Data are presented as fluorescence ratio (485/410 nm) after frame-by-frame background correction.

In total, 12,500 cells/well from HeLa pLKO and HeLa sh5-EFHD1 were seeded into 96-well plates. Resazurin sodium salt (Santa Cruz Biotechnology, sc-206037) was used to assess cell viability. For untreated cells, culture media was changed 48 h after seeding to 0.004% resazurin in DMEM 10% FBS medium and incubated for 4 h at 37 °C after which the fluorescence was measured at 540ex/590em using a CLARIOstar Plate Reader (BMG Labtech). For treatments, 24 h after seeding, both cell lines were treated with 0.5 mM H₂O₂ (Sigma-Aldrich, 516813), 500 nM thapsigargin (Sigma-Aldrich, 586005), 40 µM C2-Ceramide (Santa Cruz Biotechnology, sc-201375) or 50 nM Paclitaxel (Abcam, ab120143-10mg), and 48 h after treatment cell viability was assessed as described above.

RNA-Seq and proteomics data were obtained from the DepMap portal (https://depmap.org/), RNA-Seq gene expression TPM values from DepMap Public 23Q2 release (Ghandi et al, 2019), and proteomics from the Nusinow et al, 2020 data release (Nusinow et al, 2020). Primary breast tumor and adjacent normal tissue gene expression pair data was analyzed with TNMplot.com (Bartha and Győrffy, 2021) using the RNA-Seq data as source. Kaplan–Meier plot to assess overall breast cancer patient survival was generated using GEPIA2 (Tang et al, 2019), selecting the BRCA dataset for analysis and setting the median expression value as cutoff for determining the high and low expressing EFHD1 cohorts. Gene expression data for chemotherapy responders and non-responders was analyzed from rocplot.org (Fekete and Győrffy, 2019) selecting the pathological complete response cohort, any chemotherapy, and JetSet only data as parameters.

Migration assays were performed using 24-well plates with uncoated polycarbonate membrane inserts (BD Biosciences, 353097). In total, 2 × 10⁵ cells were allowed to migrate for 24 h. Membranes were fixed with Methanol (Merck, 106009) stained by Hematoxylin solution (Sigma, 51275) and mounted on glass slides with Pertex mounting medium (#41-4010-00, MEDITE GmbH). Images were taken with an Olympus BX43 microscope.