Single and dual camera SIM imaging. The SIM-setup used is composed of a 405, 488, 515, 532, and a 561 nm excitation laser introduced at the back focal plane inside the SIM-box with a multimodal optical fiber. For super-resolution, a CFI SR Apochromat TIRF 100x-oil (NA 1.49) objective was mounted on a Nikon-Structured Illumination Microscopy (N-SIM^®, Nikon, Austria) System with standard wide field and SIM filter sets and equipped with two Andor iXon3^® EMCCD cameras mounted to a Two Camera Imaging Adapter (Nikon Austria, Vienna, Austria). At the bottom port a third CCD-camera (CoolSNAP HQ2, Photometrics, Tucson, United States) is mounted for wide field imaging. For calibration and reconstruction of SIM images the Nikon software (NIS-Elements, Nikon, Austria) was used. Reconstruction was permanently performed with the same robust setting to avoid artefact generation and ensures reproducibility with a small loss of resolution of 10% compared to most sensitive and rigorous reconstruction settings. Microscopy setup adjustments were done as described elsewhere (Gottschalk et al., 2019).

Cells expressing ERAT4.03 NA (NGFI, Graz, Austria) were stained with 100 nM TMRM in EH-loading buffer [135 mM NaCl, 5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 10 mM HEPES, 2.6 mM NaHCO3, 440 μM KH2PO4, 340 μM Na2HPO4, 10 mM D-glucose, 0.1% MEM vitamins (Gibco), 0.2% MEM amino acids (Gibco), 1% penicillin-streptomycin (100 U/ml), 1% streptomycin (100 μg/ml), and 1.25 μg/ml amphotericin B, pH adjusted to 7.4] (Gottschalk et al., 2019) for 20 min and imaged. Morphology parameters were measured automatically via a custom-made ImageJ macro using the following procedure. An additional background subtraction based on the rolling ball method was introduced to enhance contrast for later analysis. A global auto Otsu threshold using a stack histogram as well as a local auto Otsu threshold (radius of 640 nm) based on a single slice histogram were applied and merged. ERAT4.03 NA staining was Otsu thresholded, dilated, filled, and subsequently and used as a mask for TMRM stained mitochondria to exclude not transfected cells.

Binarized mitochondria were segmented using the ImageJ “Analyse particles” function. Mitochondrial area, aspect ratio, and form factor were analyzed using ferrets diameter and perimeter as previously described (Gottschalk et al., 2019).

To analyze the co-localization between ER and mitochondria, TMRM and ERAT4.03 NA were determined on a single cell level using ImageJ and plugin coloc2. The Pearson’s coefficient was calculated as previously described (Tawfik et al., 2022).

Cells expressing ERAT4.03 NA (ER-channel) (NGFI, Graz, Austria; www.ngfi.eu) were stained with 100 nM TMRM (Mito-channel) in EH-loading buffer for 20 min and imaged over time for 10 min. ER- and Mito-channel were background subtracted using the “rolling ball” background subtracter implemented in ImageJ. Pixel wise multiplication of ER- and Mito-channel was performed, followed by a pixel wise calculation to the potency of two. This was done to increase the contrast of ER and mitochondrial overlapping regions. Next, we used the Mosaic Particle Tracker (Sbalzarini and Koumoutsakos, 2005) out of the Mosaic suit plugin collection for ImageJ to identify and track single interactions between the ER and mitochondria. The Mosaic Particle Tracker Plugin was used to detect particles dependent on the contrast and size through an image stack and connects the particles afterwards to create the traces visible in the Supplementary Videos. Out of the x and y coordinates the plugin automatically calculates duration, velocity and a manifold of other results which can be extracted as txt-file.

HeLa or EAhy926 cells were grown under standard culture conditions until 50% confluence was reached, transfected in DMEM (without FCS and antibiotics) with 1.5 µg/well plasmids or 100 µM siRNA (MFN2 siRNA: 5′-CGG CAA GAC CGA CUG AAA U dTdT-3′; Control siRNA 5′-UUC UCC GAA CGU GUC ACG UTT-3′) using 2.5 µg/well TransFast™ transfection reagent (Promega, Madison, WI, United States). Next day, the medium was replaced with DMEM containing 10% FCS and 1% penicillin/streptomycin and kept for further 24 h prior experiments. Alternatively, plasmid transfection was done by using PolyJet™ In Vitro DNA Transfection Reagent (SL100688, SignaGen^® Laboratories, Frederick, MD, United States).

Dynamic changes in [Ca²⁺]_mito were analyzed in HeLa cells expressing the organelle-targeted FRET-based Ca²⁺ sensor 4mtD3cpv as previously described (Tawfik et al., 2022). [ATP]_ER was measured in cells expressing the ER targeted genetically encoded ATP sensor ERAT 4.01. Cells were stimulated with the IP3-generating agonist ATP (Sigma Aldrich, Vienna, Austria). Basal levels in [ATP]_mito were measured in HeLa cells expressing the organelle-targeted FRET-based ATP sensor mtAT1.03 (Depaoli et al., 2018). To determine mitochondrial membrane potential, cells were incubated in 20 nM of the fluorescent indicator tetramethylrhodamine (TMRM) (Invitrogen™ T668; Vienna, Austria) in EH-loading buffer for 20 min at room temperature, as described previously (Depaoli et al., 2018). A full disruption of the mitochondrial membrane potential by application of carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP) (Abcam, Cambridge, United Kingdom) was used to normalize the TMRM_mito/TMRM_nuc ratio.

Measurements were performed on an inverted wide-field microscope (Observer.A1, Carl Zeiss GmbH, Vienna, Austria) as described previously (Depaoli et al., 2018; Koshenov et al., 2021) that was equipped with an programmable gravity-based perfusion system (NGFI). In short, cells were imaged with a ×40 oil immersion objective (Plan Apochromat 1,3 NA Oil DIC (UV) VISIR, Carl Zeiss GmbH, Vienna, Austria) and a standard CFP/YFP filter cube. Emission was collected with a 505dcxr beam-splitter on two sides of the camera (CCD camera, Coolsnap Dyno, Photometrics, Tucson, AZ, United States). 4mtD3cpv, ERAT4.01, and mtAT1.03 were excited with a wavelength of 440 nm (440AF21, Omega Optical, Brattleboro, VT, United States), and emission was captured at 480 and 535 nm (480AF30 and 535AF26, Omega Optical, Brattleboro, VT, United States). TMRM was excited at 550 nm, and the emission was collected at 600 nm. A region of interest containing the mitochondrial TMRM fluorescence (TMRM_mito) and the cytosolic TMRM (generally the area of the nucleolus) fluorescence (TMRM_nuc) was selected, and a ratio TMRM_mito/TMRM_nuc was calculated over time. The data were recorded with the NIS-Elements AR software (Nikon, Vienna, Austria) and analyzed using GraphPad Prism (GraphPad Software, San Diego, CA). Measurements were background corrected using a background region of interest and corrected for bleaching using an exponential decay fit of the basal fluorescence extrapolated to the whole measurement. Results are shown as Δmax (TMRMmito/TMRMnuc) between basal level and maximal FCCP induced depolarization.

To measure [Ca²⁺]_cyto, cells were incubated with the fluorescent cytosolic Ca²⁺ indicator Fura-2 acetoxy-methyl-ester (Fura-2AM) (TEFLabs, Austin, TX) for 30 min in EH-loading buffer, as described previously (Tawfik et al., 2022). An NGFI AnglerFish C-Y7G imager equipped with a Nikon CFI Super Fluor 40 x oil NA 1.3 objective (NGFI) equipped with the NGFI PS9D perfusion system (NGFI, Graz, Austria) was used. Fura-2AM was illuminated with 340 and 380 nm, and emission was captured at 515 nm (495dcxru; Omega Optical, Brattleboro, VT, United States). The results of the measurements were recorded with live-acquisition software v2.0.0.12 (Till Photonics) and analyzed using GraphPad Prism (GraphPad Software 9.3.0, San Diego, CA). The measurement was background-subtracted using a background ROI and corrected for bleaching using an exponential decay fit. Results are shown as the ratio of F₃₈₀/F₃₄₀.

For live-cell imaging experiments, cells were seeded on 30 mm glass coverslips in 6-well plates and accordingly transfected with 1.5 μg of plasmid DNA and 100 μM of the respective siRNA. Before the measurement, cells were kept in EH-loading buffer at room temperature. During the measurement, the cells were continuously perfused with a Ca²⁺-containing buffer, which consisted of 145 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl2, 10 mM D-glucose, and 10 mM HEPES, pH adjusted to 7.4, using a perfusion system (PS9D, NGFI, Graz, Austria; www.ngfi.eu).

Total cellular RNA was isolated from HeLa cells using 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, United States). The qRT-PCR 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 reference gene. Primer sequences are as follows: MFN1 for: 5′-GCT​CTT​CTC​TCG​ATG​CAA​CT -3′, MFN2 rev: 5′- TGG​CGC​TCT​CCT​GGA​TGT​AG -3′, GAPDH (QuantiTect^® Primer Assay Hs_GAPDH, Qiagen, Hilden, Germany).

35,000 HeLa cells were plated on XF96 polystyrene cell culture microplates (Seahorse^®, Agilent, CA, United States) a day before the experiment. Thirty minutes before the measurement, the cell medium was changed to XF assay medium supplemented with 1 mM sodium pyruvate, 2 mM glutamine and 5.5 mM D-glucose and incubated in non-CO₂ 37°C incubator. The XF96 extracellular flux analyzer was used to measure oxygen consumption rate (OCR). OCR (pmol O₂/min) values were normalized to the protein content. Protein concentration of each well was determined with the PierceTM BCA Protein Assay Kit (Thermo Scientific, Rockford, IL, United States).

Western blots were performed according to standard protocols. Briefly, cell lysis was conducted with RIPA buffer (+1% Triton-X100) (Bio-Rad formulation) supplemented with protease inhibitor cocktail (1:50; Sigma Aldrich, Vienna, Austria). Samples were frozen in liquid nitrogen and thawed for three times and incubated for 30 min on ice. Protein amounts were measured with Pierce BCA Protein Assay Kit (ThermoScientific, United States). 40 µg protein samples were loaded on a 12.5% SDS-PAGE gel together with PageRuler™ Plus Prestained Protein Ladder (Fisher Scientific, Vienna Austria). Blots were blocked (5% milk) and antibodies diluted in 5% milk in TBS-Tween. The following antibodies were used: Mitofusin-2 (D2D10, 1:1,000, Cell Signaling Technology, MA, United States), and β-Actin (D6A8, 1:1,000, Cell Signaling Technology). HRP labeled goat-anti-rabbit (sc-2054, 1:5,000, Santa Cruz Biotechnologies) was used as secondary antibody. For visualization, the SuperSignal West Pico PLUS kit (Fisher Scientific) was used and detection was conducted on a ChemiDoc System (Bio-Rad Laboratories, Vienna, Austria).

Each exact n value and the number of independent experiments is indicated in each figure legend. Statistical analysis was performed using the GraphPad Prism software version 9.3.1 (GraphPad Software, San Diego, CA, United States) or Microsoft Excel (Microsoft Office 2016). Analysis of variance (ANOVA) with Tukey post hoc test and unpaired two sided t-test were used for evaluation of the statistical significance. p < 0.05 was defined to be significant. At least three experiments on three different days were performed for each experimental setup.

ER-mitochondrial interaction under ER stress is a widely studied topic that is predominantly investigated by high-resolution microscopy, co-localization studies (Bravo et al., 2011) and electron microscopy (Csordás et al., 2006; Bravo et al., 2011). We are aware that structured illumination microscopy has an inferior optical resolution compared to electron microscopy. Nevertheless, there are several advantages of live cell super resolution microscopy compared to electron microscopy. First and most importantly stands the fact that we are able to monitor the cells without fixation and accompanying artefacts. SIM also has a better resolution compared to conventional confocal microscopy without intense illumination and light induced cell toxicity. Besides, the approach has the big advantage of fast preparation and sample rate. Nevertheless, the dynamic assembly and disassembly of inter-organellar contact sites, called MAMs, under ER stress remains elusive, while representing an important qualitative characteristic of MAMs. Therefore, we investigated the dynamics of MAMs in response to the induction of ER stress by tunicamycin. We used ERAT4.01 NA expressing (Vishnu et al., 2014) (ER-targeted fluorescent protein) and TMRM (mitochondrial specific staining) to visualize MAMs. The overlap of the ER and mitochondria stainings was used to track single MAMs (Figure 1A). Induction of ER stress led to more stable MAMs, which could be tracked for longer time (Figure 1B; Supplementary Videos S1–S3). A detailed analysis showed that fractions of MAMs, which exist longer than 105 s, are enriched in cells treated with tunicamycin for 8 h (Figure 2A, Supplementary Figure S1). We set a threshold for the tracking duration of >105 s and observed that stable MAMs were significantly increased by approximately 50% during ER stress (Figure 2B). Additionally, we measured the MAM tracking velocity which was reduced after 8 h tunicamycin treatment (Supplementary Figure S2). The overall number of MAMs measured by the Pearson correlation coefficient of TMRM and ERAT4.01 remained unchanged. (Figure 2C). The mitochondrial morphology, revealed as aspect ratio and form factor, or the average mitochondrial size, were not affected by the early ER-stress (Figures 2D–F). We further evaluated EAhy926 cells regarding MAM duration, ER and mitochondrial colocalization and mitochondrial morphology upon treatment with tunicamycin. Likewise to HeLa cells tunicamycin treatment lead to longer MAM tracking times but unchanged ER-mitochondrial colocalization and morphology (Supplementary Figure S3).

[Ca²⁺]_mito homeostasis is greatly influenced by the dynamics and quantity of structural ER-mitochondrial interaction (Patergnani et al., 2011). Accordingly, we used HeLa cells stably expressing the mitochondrial-targeted genetically encoded Ca²⁺ sensor 4mtd3cpv to investigate if changes in the stability and lifetime of MAMs affect [Ca²⁺]_mito. We found increased basal [Ca²⁺]_mito levels in HeLa cells treated with tunicamycin for 2 and 8 h. In contrast, mitochondrial uptake after histamine-induced [Ca²⁺]_ER release was not influenced by tunicamycin induced ER-stress (Figures 3A–C). Increased basal [Ca²⁺]_mito changes might be the result of [Ca²⁺]_cyto elevations (Mukherjee et al., 2002). To exclude this possibility, we used the [Ca²⁺]_cyto indicator Fura-2 AM to determine basal and [Ca²⁺]_ER release triggered [Ca²⁺]_cyto levels. Neither basal [Ca²⁺]_cyto nor ∆[Ca²⁺]_cyto after [Ca²⁺]_ER release were affected by the treatment with tunicamycin (Figures 3D–F). Additionally, we observed an increased number of oscillations after ER Ca²⁺ release in tunicamycin treated cells (Figures 3G,H), pointing to an altered orchestration of ER and cytosolic Ca²⁺ signaling.

Oxidative phosphorylation (OXPHOS) is highly dependent on [Ca²⁺]_mito levels due to the importance of basal [Ca²⁺]_mito in the activity of the dehydrogenases of the tricarboxylic acid (TCA) cycle (Denton et al., 1975; Denton et al., 1978; Lawlis and Roche, 1981). Accordingly, we used tetramethylrhodamine methyl ester (TMRM) and Seahorse^® respirometer to analyze the effects of tunicamycin on mitochondrial membrane potential and oxygen consumption rate (OCR), respectively. We found significantly hyperpolarized mitochondria (Figures 4A,B) and increases in basal OCR, ATP-linked respiration, and maximal respiratory capacity after 8 h of tunicamycin incubation (Figures 4C–F). These data show a potential link between ER stress-induced MAM remodeling and increased mitochondrial OXPHOS via elevated basal [Ca²⁺]_mito.

Our data so far let us speculate that the increase in ER ATP during ER stress response (Yong et al., 2019) is achieved by a stabilization of MAMs that yield increased basal [Ca²⁺]_mito, and stimulated mitochondrial respiration. Accordingly, we measured the dependency of [ATP]_ER on oxidative mitochondrial respiration. For that purpose, we used the genetically encoded ER targeted ATP sensor ERAT4.01 (Vishnu et al., 2014) and challenged control and tunicamycin treated cells with oligomycin. We found increased basal [ATP]_ER and a stronger [ATP]_ER drop during oligomycin challenging in cells treated with tunicamycin (Figures 5A–C), pointing to an increased coupling of ER metabolism to mitochondrial oxidative phosphorylation. To further support these findings, we measured [ATP]_mito in mtAT1.03 stable expressing HeLa cells treated with and without tunicamycin. In our protocol, we first removed glucose, followed by glucose re-addition and challenging with oligomycin. The initial elevation of [ATP]_mito after glucose (Figure 5D) removal was shown to be linked to the activity of hexokinase (Depaoli et al., 2018) and cellular glycolytic activity. We could see an elevation of hexokinase activity (Figure 5F) while basal and minimum [ATP]_mito after glucose removal remained unchanged (Figures 5E,G), pointing to an increased aerobic glycolysis under tunicamycin treatment. The addition of oligomycin led to a reduction of [ATP]_mito in HeLa cells treated with tunicamycin (Figure 5H). Considering the changes of [ATP]_mito and [ATP]_ER upon tunicamycin treatment/ER stress, we conclude that increased mitochondrial oxidative phosphorylation fuels the increased availability of ATP in the ER during the early phase of ER stress.

MFN2 is known to mediate mitochondrial fission (Ngoh et al., 2012) and ER—mitochondria tethering (Naon et al., 2016; Casellas-Díaz et al., 2021). MFN2 is upregulated in response to tunicamycin induced ER stress (Ngoh et al., 2012). Knockout of MFN2 leads to increased ER stress response, including increased ER chaperone expression (Ngoh et al., 2012; Sebastián et al., 2012), ER expansion and enhancement of UPR branches (e.g., PERK, XBP-1, and ATF6) (Muñoz et al., 2013). Therefore, we investigated whether MFN2 might be involved in the formation of stable MAMs after tunicamycin treatment. We found that knockdown of MFN2 in HeLa cells prevented the formation of ER stress-induced stable MAMs (Figures 6A–D; Supplementary Figure S4; Supplementary Videos S4–S7).

Regardless, the overall proximity between ER and mitochondria calculated as Pearson correlation coefficient was not influenced by the knockdown of MFN2 (Figure 6E). The mitochondrial morphology was affected by silencing of MFN2. Aspect ratio and form factor were reduced, pointing to reduced mitochondrial interconnection, fragmentation and swelling (Figures 6F–H). Knockdown efficiency of MFN2 siRNA was measured via qRT-PCR and western blot and calculated to be approximately 80 and 71%, respectively (Figures 6I,J; Supplementary Figure S5).

The neutralization of ER-stress induced stable MAMs by reduction of MFN2 expression lead to the question whether the increased exchange of Ca²⁺ and ATP between mitochondria and the ER during ER-stress is affected likewise. Silencing of MFN2 prevented the tunicamycin-induced increase of basal [Ca²⁺]_mito (Figures 7A,B) and led to a general reduction of [Ca²⁺]_mito uptake capacity during [Ca²⁺]_ER-release (Figures 7A,C). To test the specificity of the used siRNA against MFN2 we performed a recovery experiment based on the [Ca²⁺]_mito uptake capacity during [Ca²⁺]_ER-release using a mCherry-MFN2 construct which we co-transfected with or without MFN2 siRNA. Basal [Ca²⁺]_mito was unchanged upon knockdown or expression of mCherry-MFN2 (Supplementary Figure S6). Both, knockdown and overexpression of MFN2, lead to a reduction of [Ca²⁺]_mito uptake capacity, which was recovered by simultaneous transfection with the MFN2 siRNA and mCherry-MFN2 construct (Supplementary Figure S6). These data prove the specificity of the siRNA used to silence MFN2.

Measurements of [ATP]_mito revealed increased basal ATP concentrations (Figures 7D,E) and a substantial increase in mitochondrial ATP in MFN2 depleted cells upon oligomycin (Figures 7D,F). These data indicate that MFN2 silencing caused a switch of the F₁F_OATPase into the reverse mode, consuming ATP to maintain the mitochondrial membrane potential. Since ER ATP is coupled to these events, we observed increased basal [ATP]_ER (Figures 8A,B) and a strong drop of [ATP]_ER during the perfusion with oligomycin in cells treated with tunicamycin (Figures 8A,C), as shown before (Figures 5A–C). Knockdown of MFN2 impedes that dynamic and negates the effect of tunicamycin on ER ATP in HeLa cells (Figures 8A–C), pointing to an inability of the ER to utilize ATP from mitochondria upon the induction of ER stress.