Chemicals and reagents for solutions and buffers were purchased from Carl Roth (Karlsruhe, Germany). Histamine, monensin, and BHQ were obtained from Sigma-Aldrich (Vienna, Austria). Nigericin was from Tocris (Abingdon, United Kingdom) and ionomycin from Abcam Biochemicals (Cambridge, United Kingdom). For storage cells were maintained in a buffer (storage buffer) containing (in mM): 2 CaCl₂ 135 NaCl, 5 KCl, 1 MgCl₂, 1 HEPES, 2.6 NaHCO₃, 0.44 KH₂PO₄, 0.34 Na₂HPO₄, 10 D-glucose, 0.1% vitamins, 0.2% essential amino acids, and 1% penicillin/streptomycin pH adjusted with NaOH to 7.4 prior to imaging experiments. Therefore, cells were perfused in a physiological HEPES-buffered solution (2CaNa) composed of (in mM): 2 CaCl₂ 135 NaCl, 5 KCl, 1 MgCl₂, 1 HEPES, 10 D-glucose, pH adjusted with NaOH to 7.4 and stimulated with either BHQ or histamine in either a Ca²⁺ free solution or a nominal Ca²⁺ free solution containing EGTA instead of 2 CaCl₂.

In this study we exclusively used a HeLa cell model that was cultivated at passage >50 on low fluorescence glass cover slips with a diameter of either 15 or 30 mm in DMEM (Sigma-Aldrich, Vienna, Austria) containing 10% FCS (PAA, Pasching, Austria), penicillin (100 U/ml) and streptomycin (100 U/ml) in a humidified incubator (37°C, 5% CO₂/95% air). Cells were transfected in a serum-free medium using 3 μg/ml TransFast^TM Transfection reagent (Promega, Madison, WI, United States) together with 1–2 μg of plasmid(s) and/or 100 nM of the respective siRNA(s) per ml 2 days before experiments.

The plasmid encoding mutated ROMO1 at positions 15, 27, and 42 from potential ROS modulating cysteines to serines (ROMO1) was synthesized by General Biosystems Inc., (Morrisville, NC, United States) and the vectors encoding for GEMGeCO1, CARGeCO1, and mt-CARGeCO1 were purchased from Addgene (Cambridge, MA, United States) and used as vector templates for the N-terminal insertion via XhoI/BamHI of either MICU1^1–140, CVsue or ROMO1 using the primers as follows: MICU1 forward 5′-CCCTCTAGACTCGAG CATG TTTCGTCTGAACTCACT-3′; MICU1 position 420 reverse 5′- GGTTGGATCC TTCAAGGTGGCAAAATATCGG-3′; CVsue forward 5′-AAACTCGAG CATGGTGCCACCGGTGCAGGT- 3′; CVsue position 206 reverse 5′-CCCGGATCC TTTAATAT GCTGTCATCTTCTGCC-3′; ROMO1 forward 5′-AAATCTAG ACTCGAGCATGCCGGTGGCCGTGGGT-3′ and ROMO1 position 255 reverse 5′-CATAGGATCC CGGATGCCCATCCCA ATGG-3′. For SIM these targeting sequences were subcloned in frame with a C-terminal red fluorescent protein (mRuby). All siRNAs were purchased from Microsynth (Balgach, Switzerland) and their (sense strands, 5′-3′) sequences were: ACACUAGGCACCGAGAUUA (siVDAC1) for silencing hVDAC1 (Arif et al., 2019); GCCAGAGACAGACAAUACU (hMCU-si1) and GGAAAGGGAGCUUAUUGAA (hMCU-si2) for silencing hMCU; GAACUUUGCUGCUCUACUU for silencing hEMRE and UUCUCCGAACGUGUCACGU as a scrambled Control siRNA. It should be noted that the siRNAs for the knockdown of hMCU and hEMRE had been already verified and used before in the present HeLa cell type (Waldeck-Weiermair et al., 2015; Gottschalk et al., 2019) and thus a verification of the efficiency due to the siRNA mediated knockdown was only performed for VDAC1 siRNA in this work (Supplementary Figure S6).

Total cellular RNA was isolated from HeLa cells transfected with either siRNA against VDAC1 or the scrambled Control siRNA with the PEQLAB total RNA isolation kit (Peqlab, Erlangen, Germany), followed by reverse transcription to cDNA, performed in a thermal cycler (Peqlab) using the high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA, United States). The qPCR reaction was set up with the GoTag^® qPCR Master Mix (Promega, Mannheim, Germany) together with gene-specific primers (Invitrogen, Vienna, Austria). Experiments were performed on a LightCycler 480 (Roche Diagnostics, Vienna, Austria). Relative expression of specific genes was normalized to human GAPDH, as a housekeeping gene. Primer sequences were as follows: VDAC1 for: 5′-GGACTGAGTACGGCCTGACGTT-3′, VDAC1 rev: 5′-CAGTCCACGTGCAAGCTGATCT-3′, GAPDH (QuantiTect^® Primer Assay Hs_GAPDH, Qiagen, Hilden, Germany).

Ca²⁺ imaging was performed on a digital wide field microscope, the iMIC (Till photonics, Gräfelfing, Germany) equipped with a 40× objective (alpha Plan Fluar 40×, Zeiss, Göttingen, Germany) and an ultrafast switching monochromator, the Polychrome V (Till Photonics). Illumination of GEM-GeCO1 targeted sensors was performed at 430 nm excitation and emissions were collected with a dichrotome dual emission filter set (dichroic 535dcxr, CFP emitter 482/18 nm, and YFP emitter 535/3 nm). pH calibration was performed as described previously (Burgstaller et al., 2019) in a Ca²⁺-free or Ca²⁺ containing buffer supplemented with 10 μM ionomycin each to establish Ca²⁺ unbound and Ca²⁺ bound states of targeted GEMGeCO1. CARGeCO1 based Ca²⁺ indicators were excited at 575 nm and emitted at 600 nm. For simultaneous measurements, GEM- and CARGeCO1 targeted sensors were alternately excited for 400 ms each at 430 and 575 nm. Emissions derived from both sensors were taken in 3 s interval. Alternatively, we used an ultrafast switching mode where both Ca²⁺ indicators were excited for 150 ms in a 310 ms interval. During the measurements cells were continuously perfused by using a gravity-based perfusion system (NGFI, Graz, Austria) and images were recorded with a charged-coupled device (CCD) camera (AVT Stingray F-145B, Allied Vision Technologies, Stadtroda, Germany). Data acquisition and control of the digital fluorescence microscope was performed using the live acquisition software version 2.0.0.12 (Till Photonics).

Structured Illumination Microscopy experiments were performed as described previously (Gottschalk et al., 2018). In brief, transfected HeLa cells either expressing MICU1^1–140-mRuby, CVsue-mRuby or ROMO1-mRuby were loaded with 0.5 μM MitoTracker^TM Green (MTG, Thermo Fisher Scientific, Waltham, MA United States) in storage buffer for 40 min, once washed and imaged in 2CaNa with a super-resolution CFI SR Apochromat TIRF 100x-oil (NA 1.49) objective mounted on a Nikon-Structured Illumination Microscopy (N-SIM^®) System with standard wide field and SIM filter sets and equipped with two Andor iXon3^® EMCCD camera mounted to a Two Camera Imaging Adapter (Nikon Austria, Vienna, Austria). Co-localization of MTG fluorescence intensities with MICU1^1–140-mRuby, CVsue-mRuby and ROMO1-mRuby in the CM and IBM were measured via line scan analysis within a defined rectangle area. Normalized sum of intensities profiles derived from green (MTG) and red (mRuby) fluorescence were calculated to define the full width at half maximum (FWHM) as a measurement for the label latitude. MTG-staining was used as an internal reference for the IMM to determine the distance variation of sub-mitochondrial targeted mRuby at basis of the FWHM.

Statistical analysis for SIM experiments were performed using one way ANOVA and Bonferroni’s post hoc test. The acquired data of Ca²⁺ measurements were analyzed by the GraphPad Prism software version 5.01 (GraphPad Software, San Diego, CA, United States). Data are presented as mean ± standard error of mean (SEM) of independent experiments (n) throughout the whole manuscript. For comparisons between two groups, two-tailed Student t-test was used for evaluation of statistical significance and a P value between 0.01 and 0.05 (p Student’s t-test) was considered significant and indicated with “^∗”, P between 0.001 and 0.01 as very significant with “^∗∗,” and P < 0.001 as highly significant with “^∗∗∗.” For comparisons across multiple groups, one-way ANOVA with Bartlett’s test for equal variances and Bonferroni’s Multiple Comparison test were used for evaluating statistical significance expressed as described above.

We and others have predicted that the mitochondrial calcium uptake 1 protein (MICU1) is a specific IMS protein via its N-terminal targeting sequence (Petrungaro et al., 2015; Waldeck-Weiermair et al., 2015). Accordingly, we generated a potentially IMS-targeted mRuby variant using the first 140 amino acids that includes the polybasic region of MICU1, but none of the two EF-hand motifs or the multimerization sites. HeLa cells expressing MICU1^1–140-mRuby were subsequently loaded with the IMM specific dye MitoTracker^TM Green (MTG). Super-resolution structure illumination microscopy (SIM) revealed that MICU1^1–140-mRuby was exclusively directed to the inner boundary membrane (IBM) (Figures 1B,E and Supplementary Figure S1A). For targeting the sensor into the CL, we utilized a recently identified protein referred to as reactive oxygen species modulator 1 (ROMO1) that acts especially in the CL (Norton et al., 2014; Cogliati et al., 2016). Moreover, we used the subunit e of complex V (CVsue) that has been already shown to target within the CL (Rieger et al., 2014; Figures 1C,E and Supplementary Figure S1A) to structurally and functionally verify the targeting of ROMO1. In its size ROMO1 is comparable with CVsue and is also anchored with only one transmembrane domain into the cristae membrane. To disable ROMO1’s function as ROS modulator while preserving its feature as a targeting peptide we mutated all four ROS sensitive cysteine residues to rather ROS-inert serine residues at the amino acid positions 15, 27, 42, and 79. Fusion of ROMO1 to the N-terminus of an mRuby yielded predominantly CL localization that was comparable to that of CVsue targeting as revealed via SIM (Figures 1D,E and Supplementary Figure S1A). SIM data were randomly recorded under all conditions and line plot analyses were performed showing no influence of ROMO1-, CVsue- or MICU1^1–140-mRuby expression on mitochondrial thickness (Supplementary Figures S1B,C). Using these newly identified targeting sequences MICU1^1–140, CVsue, and ROMO1 we generated differentially targeted green and red GeCO1 biosensors aiming to visualize Ca²⁺ within the IMS or CL compartments (Figures 1A,F).

We generated two groups of differentially targeted Ca²⁺ biosensors including the ratiometric GEMGeCO1 (Zhao et al., 2011) or intensiometric CARGeCO1 (Wu et al., 2013; Supplementary Table S1). For imaging of the ratiometric GEMGeCO1 we used a standard CFP/YFP emission filter. Recording the single fluorescence intensities of the localized sensor constantly revealed a 5-fold higher signals in the YFP channel compared to the CFP emission at basal Ca²⁺ levels (Figures 2A,C,E). Analyses of all three sub-mitochondrial constructs resulted in comparable ratio signals under resting conditions (Figures 2B,D,F,G). However, the signals upon IP₃-mediated Ca²⁺ mobilization by histamine (100 μM) was more than 50% higher by using the IMS specific MICU1^1–140-GEMGeGO1 in comparison to both CL targeted probes (Figure 2H). Although the two Ca²⁺ indicators within the CL, CVsue-GEMGeCO1 and ROMO1-GEMGeCO1, performed almost identical, the expression level of the ROMO1 targeted probe was significantly higher enabling a better resolution of signals (Figure 2I). Moreover, testing pH sensitivity of MICU1^1–140 or ROMO1 localized GEMGeCO1 revealed relative pH insensitivity of the probes at neutral and higher pH as previously reported for the recombinant GEMGeCO1 protein (Zhao et al., 2011; Supplementary Figure S2). We also constructed the corresponding sub-mitochondrial targeted intensiometric red CARGeCO1 variant which displays a low fluorescent signal under resting conditions and greatly lights up in the presence of Ca²⁺. According to the low expression level together with the low fluorescence intensity of a CVsue-CARGeCO1 indicator we assumed that this probe is hardly detectable at basal Ca²⁺ levels and thus, visualization of CL localized CAR-GeCO1 was only performed by using the ROMO1 targeting sequence. In line with the results obtained by the ratiometric GEMGeCO1, statistical analyses of IMS or CL targeted CARGeCO1 derived signals resulted in comparable histamine responses indicating to distinct sub-mitochondrial Ca²⁺ concentrations upon Ca²⁺ mobilization with an IP₃ generating agonist (Supplementary Figure S3).

The distinct spectral properties of the green and red GeCO1 biosensors allow simultaneous Ca²⁺ recordings in two distinct (sub-)compartments. First, we tested the suitability of the biosensors for simultaneous measurements by co-expressing MICU1^1–140-GEMGeCO1 with MICU1^1–140-CARGeCO1 and alternately recorded the respective emission. Since both biosensors exhibit a similar k_D for Ca²⁺, i.e., 340 nM for GEMGeCO1 (Zhao et al., 2011) and 490 nM for CARGeCO1 (Wu et al., 2013), simultaneous measurements resulted in almost analog Ca²⁺ traces in both channels even at randomly occurring oscillatory Ca²⁺ spikes in one given HeLa cell (Supplementary Figure S4). These results demonstrate that both, the red and green biosensor retain their full functionality even in sub-mitochondrial regions and further demonstrate the suitability of these constructs to image Ca²⁺ in a multi-chromatic manner. Accordingly, we co-expressed each sub-mitochondrial targeted GEM-GeCO1 with either the cytosolic- (CARGeCO1) or the mitochondrial matrix-targeted (mt-CARGeCO1) red Ca²⁺ biosensor and imaged the responses upon histamine stimulation and store-operated Ca²⁺ entry within individual cells. Targeting of the various sensors is illustrated in Figure 3A and the respective sub-mitochondrial Ca²⁺ measurements simultaneously with cytosolic Ca²⁺ signals (CARGeCO1) are presented in Figures 3B,D,F. Furthermore, simultaneous measurements of the mitochondrial matrix Ca²⁺ with the newly targeted Ca²⁺ sensors for the IMS, and for both CL probes were shown in Figures 3C,E,G. Notably, the various sensors exhibited a distinct kinetics that further reflecting the individual Ca²⁺ signaling of the very sub-compartment of the mitochondria. Hence, co-imaging of IMS with cytosolic Ca²⁺ in a fast temporal sampling mode revealed almost simultaneous occurring Ca²⁺ increase (Supplementary Figures S5A,B), while mitochondrial matrix Ca²⁺ was clearly delayed upon histamine stimulation (Supplementary Figures S5C,D).

To verify the effect of an inhibition of mitochondrial Ca²⁺ uptake on the IMS and CL Ca²⁺ signals, the expression of MCU and its positive key regulator EMRE (Baradaran et al., 2018) were reduced by transfection with the respective siRNAs. Knockdown of MCU and EMRE resulted in an effective decrease of mitochondrial Ca²⁺ uptake upon intracellular Ca²⁺ mobilization (Figure 4A), while the respective Ca²⁺ signal in the IMS remained unaffected (Figure 4B). However, ablation of MCU and EMRE yielded enhanced Ca²⁺ signals in the CL measured with CVsue-GEMGeCO1 (Figure 4C) as well as with ROMO1-GEMGeCO1 (Figure 4D) upon stimulation with histamine. These results point to a specific Ca²⁺ accumulation that occurs especially within the CL under conditions where mitochondrial Ca²⁺ uptake is prevented by the knockdown of MCU/EMRE (Figure 4E).

Several studies have shown that VDAC1 serves as a gate for Ca²⁺ ions within the OMM (De Stefani et al., 2011). Hence, we performed siRNA mediated knockdown of VDAC1 that was validated on the mRNA level by qRT-PCR (Supplementary Figure S6). To monitor the Ca²⁺ transfer through the OMM we used the IMS targeted MICU1^1–140-GEM-GeCO1 in comparison with cytosolic CARGeCO1 or mt-CARGeCO1 and established a protocol where we exposed the biosensors to either slow/low or fast/high ER Ca²⁺ release and store operated Ca²⁺ entry by the re-addition of extracellular Ca²⁺ after excessive ER store depletion. Notably, in comparison to the IP₃-mediated ER Ca²⁺ release the SERCA inhibitor BHQ mediates a lower Ca²⁺ rise in the cytosol (Figure 5A) and the IMS (Figure 5B) that is marginally entering the mitochondrial matrix (Figure 5C). Knockdown of VDAC1 did not affect [Ca²⁺]_Cyto (Figure 5A). In contrast, depletion of VDAC1 resulted in a reduced Ca²⁺ elevation in the IMS (Figure 5B) and, much more pronounced, in the mitochondrial matrix (Figure 5C) under all conditions of cytosolic Ca²⁺ elevations.