Mitochondria–ER contacts function as an iron supply hub Authors: Hijiri Oshio, Isshin Shiiba, Naoki Ito, Naozumi Okada, Fuya Yamaguchi, Yuto Ishikawa, Shun Nagashima, Yuuta Fujikawa, Keitaro Umezawa, Yuri Miura, Misaki Shimizu, Yoshiro Saito, Tomoyuki Yamaguchi, Ryoko Inatome, Shigeru Yanagi Published: 2026 ABSTRACT Mitochondrial iron dynamics are essential for cellular respiration and metabolic homeostasis, yet the molecular mechanisms governing iron supply to mitochondria remain poorly understood. Here we identify a pathway in which haem serves as an iron source for mitochondria, maintaining mitochondrial iron homeostasis and mitochondrial supercomplex integrity, regulated at mitochondria–endoplasmic reticulum contact sites (MERCs). We demonstrate that haem oxygenase 2 (HMOX2), an ER-resident enzyme, is also localized to MERCs and facilitates the supply of haem-derived iron to mitochondria. This process is orchestrated by the mitochondrial ubiquitin ligase MITOL (also known as MARCH5/MARCHF5), which ubiquitinates HMOX2 at K68 with K63-linked polyubiquitin chains, enhancing its haem-degrading activity. Notably, loss of HMOX2 or disruption of MITOL-mediated ubiquitination impairs mitochondrial iron homeostasis and mitochondrial respiration. These findings establish a paradigm in which MERCs function as an iron supply hub, integrating haem metabolism with mitochondrial iron utilization. Main Mitochondria–endoplasmic reticulum contact sites (MERCs) are specialized inter-organelle contact sites, defined as regions of 10–30 nm that facilitate communication between mitochondria and the endoplasmic reticulum (ER). MERCs play a critical regulatory role in maintaining metabolic homeostasis and have been implicated in both physiological and pathological conditions1,2. They function as a hub for the exchange of essential metabolites, including lipids and calcium ions, between mitochondria and ER3,4. Calcium transfer from the ER to mitochondria at MERCs is crucial for sustaining the activity of key mitochondrial matrix enzymes involved in the tricarboxylic acid (TCA) cycle and oxidative phosphorylation (OXPHOS)4,5. Additionally, the formation and function of MERCs are dynamically regulated in response to mitochondrial conditions such as oxidative stress and nutrient availability6–8. Given their profound impact on mitochondrial function, MERCs probably harbour additional regulatory mechanisms. The mitochondrial ubiquitin ligase MITOL (also known as MARCH5/MARCHF5) is a mitochondrial outer-membrane protein that regulates mitochondrial dynamics and quality control by ubiquitinating mitochondrial fission factors and misfolded proteins9,10. In addition, MITOL is detected in the mitochondria-associated ER membrane (MAM) fraction and plays a key role in maintaining MERCs integrity. Several MERCs proteins, including MFN2, IRE1α and RMDN3, have been identified as MITOL substrates, indicating that MITOL promotes MERCs formation and function through ubiquitination11–14. Recently, APEX2-MITOL-based proximity-labelling analyses identified numerous MERCs-associated proteins14, demonstrating that MITOL-mediated comprehensive analyses could be an effective strategy for discovering MERCs regulators. Iron homeostasis is strictly regulated by cellular uptake, excretion and storage via multiple regulatory systems15,16. Functional iron is primarily utilized as haem and iron–sulfur (Fe–S) clusters, both of which are biosynthesized in mitochondria17. Mitochondria serve as an iron-enriched organelle and a central site for Fe–S cluster synthesis and haem synthesis, and both iron-containing cofactors are indispensable as electron carriers in the TCA cycle and OXPHOS. Disruptions in mitochondrial iron balance—whether due to iron overload or iron deficiency—as well as defects in Fe–S cluster biogenesis are implicated in various human diseases18,19. Thus, a tightly controlled mechanism is required to ensure the efficient supply of iron to mitochondria to maintain cellular homeostasis. The ‘kiss-and-run’ mechanism, in which early endosomes transiently interact with mitochondria to directly translocate iron, has been proposed as a mitochondrial iron-acquisition pathway20,21. However, the mitochondrial proteins involved in sensing and responding to mitochondrial iron demand remain largely unidentified. Furthermore, alternative iron utilization pathways under iron-restricted conditions are being investigated, and the potential role of haem-derived iron has been suggested22. Haem oxygenases (HMOX), including the two functional isozymes HMOX1 and HMOX2, are the only enzymes in mammals that can directly catalyse intracellular haem degradation. Haem is metabolized into biliverdin, iron and carbon monoxide (CO) by HMOX, making this the only cellular pathway for generating haem-derived iron23. HMOX1 is well known as a stress-inducible antioxidant enzyme, but the physiological role of HMOX2 remains incompletely understood24,25. In this Article we demonstrate that HMOX2 serves as a regulator of haem-derived iron supply to mitochondria, with MITOL controlling this process through ubiquitination at the MERCs. Our findings establish a mechanistic link between MERCs, haem metabolism, and mitochondrial iron utilization, revealing a previously unrecognized pathway for iron supply to mitochondria. Results HMOX2 is a MERCs protein that interacts with MITOL To explore the functions of MERCs, we focused on mitochondrial ubiquitin ligase (MITOL/MARCH5/MARCHF5), which is a key regulator of MERCs and is also detected in the MAM fraction11. To identify a functional protein at MERCs, proximity-biotin labelling analysis was performed using TurboID (Fig. 1a). We generated HeLa cells stably expressing Myc-tagged TurboID fused to the N terminus of MITOL, and confirmed that the fused protein is correctly localized to mitochondria by immunofluorescence analysis (Extended Data Fig. 1a). The organelle membrane fractions were isolated by digitonin assay, and biotinylated proteins were purified by streptavidin pulldown assay for liquid chromatography mass spectrometry (LC–MS/MS) analysis (Fig. 1b,c). The LC–MS/MS analysis robustly identified established MAM proteins and known MITOL substrates—including MFN2 and RMDN311,14—thereby confirming the accuracy of our approach; notably, HMOX2 emerged as a candidate MERCs protein (Fig. 1d). Consistent with the LC–MS/MS analysis, western blot analysis showed that the proximal protein of MITOL was enriched in the pulldown samples of biotin-treated cells (Fig. 1e). HMOX, which includes the two isozymes HMOX1 and HMOX2, is localized to the ER and is the only enzyme that catalyses intracellular haem degradation, generating biliverdin, iron and CO. HMOX2 colocalizes with an ER marker protein, according to immunofluorescence analysis (Extended Data Fig. 1b), and subcellular fractionation analysis showed that HMOX2 is enriched not only in the ER fraction but also in the MAM fraction (Fig. 1f–h). To assess whether HMOX interacts with MITOL, an immunoprecipitation-immunoblotting (IP-IB) analysis was performed, which revealed that both HMOX1 and HMOX2 interact with MITOL under crosslinker treatment. (Fig. 1i). Furthermore, the HMOX2 mutant lacking the transmembrane domain (amino acids 294–316), which fails to localize to the ER (Fig. 1j), does not interact with MITOL (Fig. 1k). These results suggest that HMOX2 localizes to both the ER and MERCs, where it interacts with MITOL.Fig. 1HMOX2 is a MAM protein that interacts with MITOL.a, Schematic of proximity labelling using Myc-TurboID-MITOL. Biotin treatment leads to the biotinylation of proteins in proximity to MITOL. b, Schematic of the digitonin assay. Biotin-labelled cells were washed in PBS and treated with a low concentration (0.05%) of digitonin to extract the cytosol fraction, followed by a high concentration (1%) of digitonin to extract the organelle fraction. After digitonin extraction, the cells were lysed in RIPA lysis buffer containing 0.1% SDS to dissolve the digitonin-insoluble fraction. Biotinylated proteins were purified by pulling down with streptavidin beads from the organelle fraction. c, Representative streptavidin immunoblot showing biotinylated proteins in the organelle fraction and streptavidin pulldown (PD) samples from Myc-TurboID-MITOL-expressing cells with or without biotin treatment (Input). Data are representative of three independent experiments. e, Representative immunoblot showing enrichment of MITOL-proximal proteins (including HMOX2, MFN2 and RMDN3) in the streptavidin pulldown (PD) samples from biotin-treated organelle fractions (Input). Data are representative of three independent experiments. d, Identification of HMOX2 as a proximal protein of MITOL. Biotinylated proteins were purified from Myc-TurboID-MITOL-expressing cells using streptavidin beads and analysed by LC–MS/MS and label-free quantification (LFQ). The x axis represents the logarithm of a ratio of protein abundances in the biotin-treated and biotin-untreated samples (base 2), and the y axis represents log10(P value). Statistical significance was assessed using a Student’s t-test. The volcano plot shows differentially enriched proteins (log2(fold change) > 1.5, P < 0.01, n = 3). Purple dots indicate proteins markedly biotinylated upon biotin treatment. f, Schematic of the subcellular fractionation analysis to isolate a pure mitochondrial fraction and MAM enriched fraction. See Methods for details. g, HMOX2 is detected in the MAM and ER. LC–MS/MS-based heatmap showing the relative protein abundance across mitochondrial (Mito), MAM and ER fractions. h, HMOX2 is localized to the MAM and ER. Lysates from isolated fractions (f) were analysed by IB with the indicated antibodies. OMM, outer mitochondrial membrane; IMM, inner mitochondrial membrane. i, MITOL interacts with HMOX1 and HMOX2. HeLa cells transfected with the indicated vectors were treated with or without dithiobis(succinimidyl propionate) (DSP) crosslinker. Lysates were subjected to immunoprecipitation (IP) using anti-HA antibodies, followed by IB. Conditions labelled ‘(–)’ indicate transfection with an empty vector. j, HMOX2 ΔTMD mutant fails to localize to the ER. HeLa cells transfected with the indicated vectors were subjected to immunofluorescence analysis with anti-HA and anti-calnexin antibodies to visualize the localization of HMOX2 and ER, respectively. Three-slice images were reconstructed. Scale bar, 10 µm. ΔTMD denotes the transmembrane domain deletion mutant (aa 294–316). k, HMOX2 ΔTMD mutant fails to interact with MITOL. After DSP treatment, lysates from transfected HeLa cells were subjected to IP using anti-HA antibodies, followed by IB. Conditions labelled ‘(–)’ indicate transfection with an empty vector. α-tubulin serves as a loading control. Information on the number of independent experimental repeats for c,e,h–k is provided in the ‘Statistics and reproducibility’ section. Source numerical data and unprocessed blots are available as Source data. Illustrations in a, b and f created in BioRender; Oshio, H. https://BioRender.com/v093d2u (2026).Source data HMOX2 is a MAM protein that interacts with MITOL. a, Schematic of proximity labelling using Myc-TurboID-MITOL. Biotin treatment leads to the biotinylation of proteins in proximity to MITOL. b, Schematic of the digitonin assay. Biotin-labelled cells were washed in PBS and treated with a low concentration (0.05%) of digitonin to extract the cytosol fraction, followed by a high concentration (1%) of digitonin to extract the organelle fraction. After digitonin extraction, the cells were lysed in RIPA lysis buffer containing 0.1% SDS to dissolve the digitonin-insoluble fraction. Biotinylated proteins were purified by pulling down with streptavidin beads from the organelle fraction. c, Representative streptavidin immunoblot showing biotinylated proteins in the organelle fraction and streptavidin pulldown (PD) samples from Myc-TurboID-MITOL-expressing cells with or without biotin treatment (Input). Data are representative of three independent experiments. e, Representative immunoblot showing enrichment of MITOL-proximal proteins (including HMOX2, MFN2 and RMDN3) in the streptavidin pulldown (PD) samples from biotin-treated organelle fractions (Input). Data are representative of three independent experiments. d, Identification of HMOX2 as a proximal protein of MITOL. Biotinylated proteins were purified from Myc-TurboID-MITOL-expressing cells using streptavidin beads and analysed by LC–MS/MS and label-free quantification (LFQ). The x axis represents the logarithm of a ratio of protein abundances in the biotin-treated and biotin-untreated samples (base 2), and the y axis represents log10(P value). Statistical significance was assessed using a Student’s t-test. The volcano plot shows differentially enriched proteins (log2(fold change) > 1.5, P < 0.01, n = 3). Purple dots indicate proteins markedly biotinylated upon biotin treatment. f, Schematic of the subcellular fractionation analysis to isolate a pure mitochondrial fraction and MAM enriched fraction. See Methods for details. g, HMOX2 is detected in the MAM and ER. LC–MS/MS-based heatmap showing the relative protein abundance across mitochondrial (Mito), MAM and ER fractions. h, HMOX2 is localized to the MAM and ER. Lysates from isolated fractions (f) were analysed by IB with the indicated antibodies. OMM, outer mitochondrial membrane; IMM, inner mitochondrial membrane. i, MITOL interacts with HMOX1 and HMOX2. HeLa cells transfected with the indicated vectors were treated with or without dithiobis(succinimidyl propionate) (DSP) crosslinker. Lysates were subjected to immunoprecipitation (IP) using anti-HA antibodies, followed by IB. Conditions labelled ‘(–)’ indicate transfection with an empty vector. j, HMOX2 ΔTMD mutant fails to localize to the ER. HeLa cells transfected with the indicated vectors were subjected to immunofluorescence analysis with anti-HA and anti-calnexin antibodies to visualize the localization of HMOX2 and ER, respectively. Three-slice images were reconstructed. Scale bar, 10 µm. ΔTMD denotes the transmembrane domain deletion mutant (aa 294–316). k, HMOX2 ΔTMD mutant fails to interact with MITOL. After DSP treatment, lysates from transfected HeLa cells were subjected to IP using anti-HA antibodies, followed by IB. Conditions labelled ‘(–)’ indicate transfection with an empty vector. α-tubulin serves as a loading control. Information on the number of independent experimental repeats for c,e,h–k is provided in the ‘Statistics and reproducibility’ section. Source numerical data and unprocessed blots are available as Source data. Illustrations in a, b and f created in BioRender; Oshio, H. https://BioRender.com/v093d2u (2026). Source data MITOL mediates K63-linked polyubiquitination of HMOX2 at lysine 68 via MERCs Given that MITOL is an E3 ubiquitin ligase that interacts with both HMOX1 and HMOX2, we hypothesized that HMOX would be ubiquitinated by MITOL and performed an immunoprecipitation (IP)–immunoblot (IB) analysis to confirm this. We found that overexpression of wild-type (WT) MITOL markedly enhanced HMOX2 ubiquitination, whereas the control vectors and the MITOL C65/68S (CS) mutant, which lacks E3 ubiquitin ligase activity, showed no such effect (Fig. 2a). On the other hand, overexpression of WT MITOL did not enhance HMOX1 ubiquitination (Fig. 2a), suggesting that MITOL ubiquitinates HMOX2, not HMOX1. In addition, to examine ubiquitination of endogenous HMOX2, we generated 3×HA–HMOX2 knock-in (KI) HeLa cells, in which three tandem hemagglutinin (HA) tags were inserted at the N terminus of the endogenous HMOX2 locus (Extended Data Fig. 2a–d). Using these cells, ubiquitination of endogenous HMOX2 was detected by IP–IB analysis using the anti-ubiquitin antibody P4D1. Furthermore, ubiquitination was attenuated by knockdown of MITOL, indicating that endogenous HMOX2 ubiquitination is mediated by endogenous MITOL (Extended Data Fig. 2e). To clarify why MITOL ubiquitinates only HMOX2 and not HMOX1, we compared their amino-acid sequences26,27. Based on a unique difference in the C-terminal region, we generated a swapped mutant that localized to the ER (Extended Data Fig. 3a) and assessed its ubiquitination (Extended Data Fig. 3a,b). MITOL-mediated ubiquitination was not detected in either of the swapped HMOX mutants, similar to WT HMOX1 (Extended Data Fig. 3b), indicating that the C terminus of HMOX2 is necessary, but not sufficient, for ubiquitination by MITOL. It is known that MITOL adds some types of polyubiquitin chains to its substrates, including K6-, K48- and K63-linked polyubiquitination, to regulate the substrate fate11,12,14,28,29. K6- and K63-linked polyubiquitinations regulate substrate activity and/or subcellular localization, whereas K48-linked polyubiquitination promotes proteasomal degradation of the substrate. To clarify which types of polyubiquitin chain were added to HMOX2 by MITOL, a ubiquitin chain restriction assay using UbiCRest, which utilizes ubiquitin chain-specific deubiquitinating enzymes, was performed30. In the UbiCRest assay, polyubiquitin chains added to HMOX2 by MITOL were only digested in the presence of the universal deubiquitinating enzyme USP2 or the K63-linked polyubiquitin-specific deubiquitinating enzyme AMSH (Fig. 2b). These results indicate that MITOL adds K63-linked polyubiquitin chains to HMOX2. To further assess whether MITOL regulates HMOX2 protein stability or turnover, we performed cycloheximide (CHX) chase and MG132 treatment experiments (Extended Data Fig. 3c,d). These analyses showed that endogenous HMOX2 protein levels were unchanged under both MITOL knockout and MITOL overexpression conditions, suggesting that MITOL regulates HMOX2 function rather than its expression or degradation. Next, to identify the ubiquitination site of HMOX2 by MITOL, because ubiquitin is primarily conjugated to lysine residues on substrates, we generated a series of KR mutants in which each lysine (K) residue of HMOX2 was replaced with arginine (R). Compared to WT HMOX2, the ubiquitination of the HMOX2 K68R mutant was not enhanced by MITOL (Fig. 2c). The HMOX2 K68R mutant did not alter its subcellular localization (Fig. 2d). These data suggest that MITOL ubiquitinates HMOX2 at lysine 68. To examine whether MERCs formation is important for the ubiquitination of HMOX2 by MITOL, we knocked down PDZD8 and VAPB, which are MERCs anchoring proteins31,32, and checked ubiquitination of HMOX2 by MITOL. When MERCs formation was attenuated by knockdown of PDZD8 or VAPB, the MITOL-mediated ubiquitination of HMOX2 was much lower than that of the control (Fig. 2e). Additionally, the HMOX2 ΔTMD mutant, which lacks the transmembrane domain of HMOX2 and fails to localize to the ER (Fig. 1j), was not ubiquitinated by MITOL (Fig. 2f). These findings demonstrate that MITOL mediates K63-linked ubiquitination of HMOX2 at lysine 68, a process that requires intact MERCs integrity.Fig. 2MITOL mediates K63-linked polyubiquitination of HMOX2 at lysine 68 via MERCs.a, MITOL overexpression enhances HMOX2 ubiquitination. MITOL KO HeLa cells were transfected with the indicated vectors, and lysates were subjected to IP–IB analysis with the indicated antibodies. Conditions labelled ‘(–)’ indicate transfection with an empty vector. FLAG-Ub, FLAG-ubiquitin; WT, wild-type; CS, MITOL C65/67S mutant lacking ubiquitin ligase activity. b, MITOL mediates K63-linked polyubiquitination of HMOX2. MITOL KO HeLa cells transfected with the indicated vectors were fractionated to isolate the ER fraction. HMOX2 was immunoprecipitated from the ER lysate and subjected to a ubiquitin restriction assay using USP2, OTUD3, Cezanne, OTUB1 and AMSH (UbiCRest kit). IB analysis was performed with the indicated antibodies. Left: western blot analysis of the ER fraction (Input). Right: immunoprecipitated data (IP). Conditions labelled ‘(–)’ indicate transfection with an empty vector. DUBs, deubiquitinating enzymes. c, MITOL fails to ubiquitinate the HMOX2 K68R mutant. MITOL KO HeLa cells were transfected with each lysine mutant of HMOX2 and the indicated vectors. Lysates were subjected to an IP–IB analysis to compare the ubiquitination levels of the WT and KR mutant of HA-HMOX2. Conditions labelled ‘(–)’ indicate transfection with an empty vector. d, The K68R mutation does not affect HMOX2 subcellular localization. HeLa cells transfected with the indicated vectors were analysed by immunofluorescence. Enlarged images are of the areas outlined by boxes in the rows above. Scale bars, 10 μm. e, Knockdown of PDZD8 or VAPB inhibits MITOL-mediated HMOX2 ubiquitination. To attenuate the MERCs formation, HeLa cells were transfected with scrambled siRNA (sc), siPDZD8 or siVAPB for 48 h, followed by transfection with the indicated vectors 20 h before the IP–IB analysis. Conditions labelled ‘(–)’ indicate transfection with an empty vector. The α-tubulin serves as a loading control. PA-MITOL, PA-tagged MITOL. f, MITOL fails to ubiquitinate the HMOX2 ΔTMD mutant. MITOL KO HeLa cells were transfected with the indicated vectors. Lysates were subjected to an IP–IB analysis to compare the ubiquitination levels of the WT and ΔTMD mutant of HA-HMOX2. Conditions labelled ‘(–)’ indicate transfection with an empty vector. ΔTMD, transmembrane domain-deleted HMOX2 mutant (aa 294–316). Information on the number of independent experimental repeats for a–f is provided in the ‘Statistics and reproducibility’ section. Unprocessed blots are available as Source data.Source data MITOL mediates K63-linked polyubiquitination of HMOX2 at lysine 68 via MERCs. a, MITOL overexpression enhances HMOX2 ubiquitination. MITOL KO HeLa cells were transfected with the indicated vectors, and lysates were subjected to IP–IB analysis with the indicated antibodies. Conditions labelled ‘(–)’ indicate transfection with an empty vector. FLAG-Ub, FLAG-ubiquitin; WT, wild-type; CS, MITOL C65/67S mutant lacking ubiquitin ligase activity. b, MITOL mediates K63-linked polyubiquitination of HMOX2. MITOL KO HeLa cells transfected with the indicated vectors were fractionated to isolate the ER fraction. HMOX2 was immunoprecipitated from the ER lysate and subjected to a ubiquitin restriction assay using USP2, OTUD3, Cezanne, OTUB1 and AMSH (UbiCRest kit). IB analysis was performed with the indicated antibodies. Left: western blot analysis of the ER fraction (Input). Right: immunoprecipitated data (IP). Conditions labelled ‘(–)’ indicate transfection with an empty vector. DUBs, deubiquitinating enzymes. c, MITOL fails to ubiquitinate the HMOX2 K68R mutant. MITOL KO HeLa cells were transfected with each lysine mutant of HMOX2 and the indicated vectors. Lysates were subjected to an IP–IB analysis to compare the ubiquitination levels of the WT and KR mutant of HA-HMOX2. Conditions labelled ‘(–)’ indicate transfection with an empty vector. d, The K68R mutation does not affect HMOX2 subcellular localization. HeLa cells transfected with the indicated vectors were analysed by immunofluorescence. Enlarged images are of the areas outlined by boxes in the rows above. Scale bars, 10 μm. e, Knockdown of PDZD8 or VAPB inhibits MITOL-mediated HMOX2 ubiquitination. To attenuate the MERCs formation, HeLa cells were transfected with scrambled siRNA (sc), siPDZD8 or siVAPB for 48 h, followed by transfection with the indicated vectors 20 h before the IP–IB analysis. Conditions labelled ‘(–)’ indicate transfection with an empty vector. The α-tubulin serves as a loading control. PA-MITOL, PA-tagged MITOL. f, MITOL fails to ubiquitinate the HMOX2 ΔTMD mutant. MITOL KO HeLa cells were transfected with the indicated vectors. Lysates were subjected to an IP–IB analysis to compare the ubiquitination levels of the WT and ΔTMD mutant of HA-HMOX2. Conditions labelled ‘(–)’ indicate transfection with an empty vector. ΔTMD, transmembrane domain-deleted HMOX2 mutant (aa 294–316). Information on the number of independent experimental repeats for a–f is provided in the ‘Statistics and reproducibility’ section. Unprocessed blots are available as Source data. Source data MITOL-mediated HMOX2 ubiquitination facilitates the haem-degrading activity of HMOX2 K48-linked polyubiquitination is known as a signal for proteasomal degradation. Meanwhile, the K63-linked polyubiquitination regulates substrate activity. For example, we previously reported that MITOL adds K63-linked polyubiquitin chain to MFN2, IRE1α and RMDN3, and regulates its activity via MERCs11,12,14. Because MITOL adds K63-linked polyubiquitin chain to HMOX2, we hypothesize that MITOL modulates HMOX2 activity through ubiquitination. To quantify the intracellular HMOX activity, we focus on IFP2.0, a fluorescent protein known as a biliverdin reporter33, because biliverdin is the unique product of haem degradation catalysed by HMOX. We constructed IFP2.0-P2A-EGFP (enhanced green fluorescent protein) expression vector as an HMOX activity reporter, which enables polycistronic expression of IFP2.0 and EGFP (Fig. 3a). To eliminate the influence of endogenous biliverdin, we generated HMOX1 and HMOX2 double knockout HeLa cells (DKO cells) and expressed either empty vectors or HMOX activity reporters (Fig. 3b). The IFP2.0 signals in DKO cells were lower than in WT cells (Fig. 3c), indicating that the HMOX activity reporter works well for quantifying intracellular HMOX activity, and that this reporter assay is highly dependent on biliverdin levels generated by HMOX activity. Next, to measure HMOX2 activity, we constructed a dual-promoter vector that enables the simultaneous expression of HMOX2 and IFP2.0-P2A-EGFP (Fig. 3d). When the WT of the HMOX2 activity reporter was expressed in DKO cells, IFP2.0 signals increased. In contrast, the IFP2.0 signal was significantly lower in cells expressing the HMOX2 H45W/G159W (HW/GW) mutant, which lacks haem oxygenase activity (Fig. 3e,f)34. Similarly, the HMOX2 C265A mutant, which cannot bind haem to its haem regulatory motif (HRM), and the HMOX2 K68R mutant, which is not ubiquitinated by MITOL, also exhibited reduced IFP2.0 signals compared to HMOX2 WT (Fig. 3e,f). These HMOX2 mutants (HW/GW, G163H and C265A) were localized to the ER and were ubiquitinated by MITOL, supporting a role for MITOL-mediated ubiquitination at K68 in facilitating HMOX2-dependent haem degradation (Extended Data Fig. 4a,b). We also analysed HMOX2 activity using the reporter in which WT HMOX2 or the K68R mutant was expressed together with MITOL. Co-expression of MITOL markedly increased the IFP2.0 reporter signal in cells expressing WT HMOX2, but not in cells expressing the HMOX2 K68R mutant (Extended Data Fig. 5a). These results indicate that MITOL-mediated ubiquitination at K68 contributes to the enhancement of HMOX2 haem-degrading activity.Fig. 3MITOL-mediated HMOX2 ubiquitination facilitates the haem-degrading activity of HMOX2.a, Schematic of the vector to express endogenous HMOX activity reporter (top), and the endogenous HMOX activity reporter system (bottom). EGFP indicates transfected cells, and IFP2.0 serves as a reporter for intracellular biliverdin (BV) levels, reflecting the haem-degrading activity of HMOX. b, IB analysis of lysates from wild-type HeLa cells (WT) and HMOX1 and HMOX2 double knockout HeLa cells (DKO), transfected with the indicated vectors. Anti-GFP detects EGFP, reflecting the expression of the endogenous HMOX activity reporter. Anti-2A peptide antibody detects the P2A peptide sequence, confirming IFP2.0 expression. Vinculin serves as a loading control. EV, empty vector; *, nonspecific band; black arrowhead, endogenous HMOX1. c, The endogenous HMOX activity reporter reliably quantifies intracellular haem degradation. Cells transfected with an empty vector (EV) or the reporter were analysed by flow cytometry. IFP2.0-positive cell populations (%) were quantified from 4,000 EGFP-positive cells. Fold changes relative to WT HeLa cells (%) are shown (left), with representative flow cytometry data (right). Data are presented as mean ± s.d. from four independent experiments. Two-tailed Student’s t-test, ****P < 0.0001. d, Schematic of the vector to express HMOX2 activity reporter. e, IB analysis of DKO HeLa cells transfected with the indicated vectors. WT, wild-type; HW/GW, HMOX2 H45W/G159W mutant (lacking haem-degrading activity); C265A, HMOX2 C265A mutant (lacking haem transfer activity); K68R, HMOX2 K68R mutant (non-ubiquitinated mutant). Anti-GFP detects EGFP, reflecting the expression of the endogenous HMOX activity reporter. Anti-2A peptide antibody detects the P2A peptide sequence, confirming IFP2.0 expression. Vinculin serves as a loading control. EV, empty vector; Endo-reporter, endogenous HMOX activity reporter used in a–c. f, Comparison of haem-degrading activity between HMOX2 mutants with WT HMOX2. IFP2.0-positive cell populations (%) reflecting HMOX2 activity were quantified from 4,000 EGFP-positive cells. Fold changes relative to WT HeLa cells (%) (top) and representative flow cytometry data (bottom) are shown. Data are presented as mean ± s.d. from four independent experiments. One-way analysis of variance (ANOVA) with Tukey’s multiple comparisons, ****P < 0.0001. Source numerical data and unprocessed blots are available as Source data.Source data MITOL-mediated HMOX2 ubiquitination facilitates the haem-degrading activity of HMOX2. a, Schematic of the vector to express endogenous HMOX activity reporter (top), and the endogenous HMOX activity reporter system (bottom). EGFP indicates transfected cells, and IFP2.0 serves as a reporter for intracellular biliverdin (BV) levels, reflecting the haem-degrading activity of HMOX. b, IB analysis of lysates from wild-type HeLa cells (WT) and HMOX1 and HMOX2 double knockout HeLa cells (DKO), transfected with the indicated vectors. Anti-GFP detects EGFP, reflecting the expression of the endogenous HMOX activity reporter. Anti-2A peptide antibody detects the P2A peptide sequence, confirming IFP2.0 expression. Vinculin serves as a loading control. EV, empty vector; *, nonspecific band; black arrowhead, endogenous HMOX1. c, The endogenous HMOX activity reporter reliably quantifies intracellular haem degradation. Cells transfected with an empty vector (EV) or the reporter were analysed by flow cytometry. IFP2.0-positive cell populations (%) were quantified from 4,000 EGFP-positive cells. Fold changes relative to WT HeLa cells (%) are shown (left), with representative flow cytometry data (right). Data are presented as mean ± s.d. from four independent experiments. Two-tailed Student’s t-test, ****P < 0.0001. d, Schematic of the vector to express HMOX2 activity reporter. e, IB analysis of DKO HeLa cells transfected with the indicated vectors. WT, wild-type; HW/GW, HMOX2 H45W/G159W mutant (lacking haem-degrading activity); C265A, HMOX2 C265A mutant (lacking haem transfer activity); K68R, HMOX2 K68R mutant (non-ubiquitinated mutant). Anti-GFP detects EGFP, reflecting the expression of the endogenous HMOX activity reporter. Anti-2A peptide antibody detects the P2A peptide sequence, confirming IFP2.0 expression. Vinculin serves as a loading control. EV, empty vector; Endo-reporter, endogenous HMOX activity reporter used in a–c. f, Comparison of haem-degrading activity between HMOX2 mutants with WT HMOX2. IFP2.0-positive cell populations (%) reflecting HMOX2 activity were quantified from 4,000 EGFP-positive cells. Fold changes relative to WT HeLa cells (%) (top) and representative flow cytometry data (bottom) are shown. Data are presented as mean ± s.d. from four independent experiments. One-way analysis of variance (ANOVA) with Tukey’s multiple comparisons, ****P < 0.0001. Source numerical data and unprocessed blots are available as Source data. Source data MITOL-mediated HMOX2 ubiquitination maintains mitochondrial iron levels To elucidate the physiological role of MITOL-mediated HMOX2 activity facilitation, we tried to find the intracellular conditions that enhance MITOL-mediated HMOX2 ubiquitination. Using compounds that could modulate the intracellular haem-related metabolites, we found that ubiquitination levels increased upon iron depletion induced by deferoxamine (DFO), an iron chelator (Extended Data Fig. 6a–c). Knockdown of mitoferrin-1 (MFRN1), a mitochondrial inner-membrane iron importer35 that triggers mitochondrial iron deficiency, also enhanced MITOL-mediated HMOX2 ubiquitination (Extended Data Fig. 6d). Moreover, endogenous HMOX2 ubiquitination was increased upon DFO treatment, whereas this effect was not observed under MITOL knockdown condition (Extended Data Fig. 6e). Overall, these data suggest that MITOL-mediated HMOX2 ubiquitination is upregulated under mitochondrial iron-deficient conditions. It was suggested that the MITOL–HMOX2 axis may respond to and maintain mitochondrial iron homeostasis, so we next generated HMOX1 KO cells, HMOX2 KO cells and DKO cells to determine the effect of HMOX on mitochondrial iron levels (Fig. 4a,b). To quantify mitochondrial iron levels, we isolated mitochondria from each cell, and after microwave digestion, metal ion concentrations were quantified using inductively coupled plasma mass spectrometry (ICP–MS; Fig. 4a). From the ICP–MS analysis, mitochondrial iron concentration in HMOX2 KO and DKO cells was much lower than in HMOX1 KO and WT cells, even though there were no significant differences in Mn, Cu and Zn (Fig. 4c). Additionally, treating HMOX2 KO cells with excess haem (haemin) or ferric ammonium citrate (FAC) did not completely restore mitochondrial iron levels (Fig. 4d). These data suggest that HMOX2-derived iron is supplied to mitochondria. To determine whether the haem-degrading activity of HMOX2 is essential for maintaining mitochondrial iron levels, we generated stable expression cells in which HMOX2 WT or HMOX2 HW/GW mutants were re-expressed in HMOX2 KO cells, and confirmed their subcellular localization to the ER (Extended Data Fig. 7a,c). ICP–MS analysis revealed that mitochondrial iron levels, which were reduced in HMOX2 KO cells, were restored in WT HMOX2 re-expressing cells but not in HW/GW re-expressing cells (Fig. 4e), indicating that the haem-degrading activity of HMOX2 is crucial for maintaining mitochondrial iron levels. Because MITOL-mediated HMOX2 ubiquitination facilitates HMOX2 activity, we re-expressed the HMOX2 K68R mutant in HMOX2 KO cells and analysed mitochondrial iron levels to confirm whether ubiquitination affects mitochondrial iron levels (Extended Data Fig. 7b,c). Mitochondrial iron levels in K68R re-expressing cells remained at low levels, similar to those in HMOX2 KO cells (Fig. 4f). Because iron levels in the ER and cytoplasmic fractions were not affected (Extended Data Fig. 8a), MITOL-mediated ubiquitination of HMOX2 is specifically important for maintaining mitochondrial iron levels. Moreover, because MERCs formation serves as an important platform for MITOL-mediated ubiquitination of HMOX2, the effect of MERCs formation on mitochondrial iron homeostasis was assessed by measuring mitochondrial iron levels in PDZD8 KO and VAPB KO cells (Extended Data Fig. 7d,e). Mitochondrial iron levels in PDZD8 KO and VAPB KO cells were lower than those in WT cells (Fig. 4g). In addition, overexpression of an artificial MERCs linker8 partially rescued the reduction in mitochondrial iron levels observed in PDZD8 and VAPB KO cells (Extended Data Fig. 8b), suggesting that MERCs formation is essential for mitochondrial iron homeostasis. The mitochondrial iron measured by ICP–MS analysis cannot distinguish between haem, Fe–S clusters, labile iron and the protein-bound iron derived. To determine whether the reduction of mitochondrial iron by HMOX2 KO resulted from a decrease in haem, haem content was measured with a haemin assay kit, which showed no significant difference in either total haem content or mitochondrial haem content (Extended Data Fig. 8c,d). These findings establish that the MITOL–HMOX2 axis is crucial for maintaining mitochondrial iron levels, independent of haem levels.Fig. 4MITOL-mediated HMOX2 ubiquitination maintains mitochondrial iron levels.a, Schematic of sample preparation for ICP–MS analysis. Mitochondrial fractions were isolated, and microwave digestion was performed to release iron from iron–sulfur clusters and haem prior to total metal ion quantification by ICP–MS. b, Representative IB analysis of generated DKO HeLa cells, HMOX1 KO HeLa cells and HMOX2 KO HeLa cells. Each KO cell line was generated by the CRISPR–Cas9 system with two different gRNAs, as shown in Extended Data Fig. 7 and Supplementary Table 1. α-tubulin serves as a loading control. c, Mitochondrial ion content (Mn, Fe, Cu and Zn) in WT, HMOX1 KO, HMOX2 KO and DKO HeLa cells. d, Mitochondrial ion content in WT and HMOX2 KO cells following treatment with haemin or ferric ammonium citrate (FAC). HMOX2 KO cells were treated with 20 µM haemin or 10 µM FAC for 24 h before mitochondrial isolation. e, Mitochondrial ion content in HMOX2 KO cells reconstituted with WT HMOX2 or the HW/GW mutant. Re-expressing cell lines of HMOX2 WT, HW/GW mutant or K68R mutant was generated by stably expressing each HMOX2 in HMOX2 KO #1 cells (IB analysis in Extended Data Fig. 7b,c). f, Mitochondrial ion content in HMOX2 KO cells reconstituted with WT HMOX2 or the K68R mutant. g, Mitochondrial ion content in WT, PDZD8 KO and VAPB KO cells. PDZD8 KO cells and VAPB KO cells were generated by the CRISPR–Cas9 system with two different gRNAs shown in Supplementary Table 1 (IB analysis in Extended Data Fig. 7d,e). In c–g, mitochondrial ion content was quantified by ICP–MS and normalized to mitochondrial protein content (‘pg metal per μg protein’). Data are presented as mean ± s.d. from three independent experiments. Two-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001. Source numerical data and unprocessed blots are available as Source data. Illustrations in a created in BioRender; Oshio, H. https://BioRender.com/v093d2u (2026).Source data MITOL-mediated HMOX2 ubiquitination maintains mitochondrial iron levels. a, Schematic of sample preparation for ICP–MS analysis. Mitochondrial fractions were isolated, and microwave digestion was performed to release iron from iron–sulfur clusters and haem prior to total metal ion quantification by ICP–MS. b, Representative IB analysis of generated DKO HeLa cells, HMOX1 KO HeLa cells and HMOX2 KO HeLa cells. Each KO cell line was generated by the CRISPR–Cas9 system with two different gRNAs, as shown in Extended Data Fig. 7 and Supplementary Table 1. α-tubulin serves as a loading control. c, Mitochondrial ion content (Mn, Fe, Cu and Zn) in WT, HMOX1 KO, HMOX2 KO and DKO HeLa cells. d, Mitochondrial ion content in WT and HMOX2 KO cells following treatment with haemin or ferric ammonium citrate (FAC). HMOX2 KO cells were treated with 20 µM haemin or 10 µM FAC for 24 h before mitochondrial isolation. e, Mitochondrial ion content in HMOX2 KO cells reconstituted with WT HMOX2 or the HW/GW mutant. Re-expressing cell lines of HMOX2 WT, HW/GW mutant or K68R mutant was generated by stably expressing each HMOX2 in HMOX2 KO #1 cells (IB analysis in Extended Data Fig. 7b,c). f, Mitochondrial ion content in HMOX2 KO cells reconstituted with WT HMOX2 or the K68R mutant. g, Mitochondrial ion content in WT, PDZD8 KO and VAPB KO cells. PDZD8 KO cells and VAPB KO cells were generated by the CRISPR–Cas9 system with two different gRNAs shown in Supplementary Table 1 (IB analysis in Extended Data Fig. 7d,e). In c–g, mitochondrial ion content was quantified by ICP–MS and normalized to mitochondrial protein content (‘pg metal per μg protein’). Data are presented as mean ± s.d. from three independent experiments. Two-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001. Source numerical data and unprocessed blots are available as Source data. Illustrations in a created in BioRender; Oshio, H. https://BioRender.com/v093d2u (2026). Source data HMOX2-mediated iron homeostasis maintains mitochondrial respiration Mitochondrial iron is required for mitochondrial respiration. Because HMOX2 maintains mitochondrial iron levels, we examined the effect of HMOX2 on the mitochondrial oxygen consumption rate (OCR) using a flux analyser. Basal and maximal respiration remained unchanged in HMOX1 KO cells, whereas a significant reduction was observed in HMOX2 KO and DKO cells compared to WT cells (Fig. 5a,b). Under mitochondrial iron deficiency or conditions of mitochondrial Fe–S cluster impairment, previous studies have reported that the expression levels of several components of the mitochondrial respiratory chain are reduced18,36–38. Indeed, induction of iron deficiency by DFO treatment led to a reduction in succinate dehydrogenase complex iron–sulfur subunit B (SDHB; complex II) and Rieske (complex III), a component of the mitochondrial respiratory complex containing Fe–S clusters (Extended Data Figs. 6a and 9a,b). Consistent with the OCR results, SDHB and Rieske expression levels were reduced in HMOX2 KO and DKO cells (Fig. 5c,d). Additionally, the expression level of ALAS1, a key enzyme in haem synthesis, was decreased in HMOX1 KO and DKO cells but not in HMOX2 KO cells, indicating that intracellular haem biogenesis is mainly affected by HMOX1 (Fig. 5c). Although the respiratory reduction in HMOX2 KO cells was restored by re-expressing WT HMOX2, re-expression of the HW/GW or K68R mutants in HMOX2 KO cells did not restore respiration (Fig. 5e,f). Consistent with this, SDHB and Rieske levels were restored in WT HMOX2 re-expressing cells but not in cells re-expressing the K68R mutant, suggesting that HMOX2 loss or lack of ubiquitination disrupts mitochondrial iron homeostasis (Fig. 5g,h). Collectively, these findings demonstrate that HMOX2-mediated haem degradation and MITOL-dependent ubiquitination at K68 of HMOX2 are critical for SDHB and Rieske stability, thereby sustaining mitochondrial respiration.Fig. 5HMOX2-mediated iron homeostasis maintains mitochondrial respiration.a, HMOX2 maintains mitochondrial respiration. OCR was measured using the Seahorse Mito Stress Test in the indicated cell lines and normalized to protein content. b, Basal and maximal respiration rates, calculated from a. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. c, HMOX2 stabilizes Rieske and SDHB proteins. IB analysis of lysates from WT HeLa cells, DKO cells, HMOX1 KO cells and HMOX2 KO cells. Vinculin serves as a loading control. d, Rieske and SDHB protein levels were quantified and normalized by vinculin from c. Fold changes relative to WT HeLa cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ***P < 0.001, *P < 0.05. e, HMOX2-mediated haem degradation and MITOL-mediated HMOX2 ubiquitination are essential for mitochondrial respiration. OCR was measured using the Seahorse Mito Stress Test in the indicated cell lines. 50 µM DFO was added 24 h before measurement. f, Basal and maximal respiration rates, calculated from e. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ***P < 0.001, **P < 0.01, *P < 0.05. g, MITOL-mediated HMOX2 ubiquitination stabilizes Rieske and SDHB: IB analysis of lysates from the indicated cell lines, with vinculin as loading control. h, Rieske and SDHB protein levels were quantified, normalized by vinculin from g. Fold changes relative to WT HeLa cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; *P < 0.05. Source numerical data and unprocessed blots are available as Source data.Source data HMOX2-mediated iron homeostasis maintains mitochondrial respiration. a, HMOX2 maintains mitochondrial respiration. OCR was measured using the Seahorse Mito Stress Test in the indicated cell lines and normalized to protein content. b, Basal and maximal respiration rates, calculated from a. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05. c, HMOX2 stabilizes Rieske and SDHB proteins. IB analysis of lysates from WT HeLa cells, DKO cells, HMOX1 KO cells and HMOX2 KO cells. Vinculin serves as a loading control. d, Rieske and SDHB protein levels were quantified and normalized by vinculin from c. Fold changes relative to WT HeLa cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ***P < 0.001, *P < 0.05. e, HMOX2-mediated haem degradation and MITOL-mediated HMOX2 ubiquitination are essential for mitochondrial respiration. OCR was measured using the Seahorse Mito Stress Test in the indicated cell lines. 50 µM DFO was added 24 h before measurement. f, Basal and maximal respiration rates, calculated from e. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ***P < 0.001, **P < 0.01, *P < 0.05. g, MITOL-mediated HMOX2 ubiquitination stabilizes Rieske and SDHB: IB analysis of lysates from the indicated cell lines, with vinculin as loading control. h, Rieske and SDHB protein levels were quantified, normalized by vinculin from g. Fold changes relative to WT HeLa cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; *P < 0.05. Source numerical data and unprocessed blots are available as Source data. Source data Haem-derived iron is essential for mitochondrial respiration Although endosomal uptake is well established in intracellular iron dynamics20,21, the mechanism behind haem-derived iron via HMOX remains largely unknown. In particular, it is difficult to distinguish whether the mitochondrial iron is from non-haem iron or haem-derived iron. To clarify the importance of HMOX-mediated haem-derived iron for mitochondria, we utilized a haem-depleted (HD) medium. We depleted haem from fetal bovine serum (FBS) following established methods39,40, which depleted the extracellular haem supply and confirmed that only haem was depleted, and not iron from FBS (Extended Data Fig. 9c,d). To block the intracellular haem supply, we added succinylacetone (SA), an inhibitor of haem synthesis, to HD medium (Fig. 6a and Extended Data Fig. 9e,f), and also found that WT HeLa cells cultured in haem-depleted conditions, even though iron was sufficient, showed destabilization of SDHB and Rieske as well as decreased mitochondrial respiration (Fig. 6b–d). Furthermore, adding back haemin to the HD medium restored the reduced expression levels of SDHB and Rieske, as well as mitochondrial respiration, whereas adding back the iron derivative FAC did not, suggesting a haem-specific function for mitochondria (Fig. 6b–d). Similarly, supplementation with holo-transferrin (holo-Tf) also failed to rescue the reduced expression levels of SDHB and Rieske, despite the induction of a cytosolic iron overload response, as reflected by increased FTH1 expression (Extended Data Fig. 9g–i). We next used HMOX KO cells to evaluate the role of haem-derived iron to distinguish it from the direct function of haem. In WT cells and HMOX1 KO cells, the expression of SDHB and Rieske was restored by adding back haemin to the HD medium. However, the recovery of SDHB and Rieske stabilization was not observed by adding back haemin to HD medium in DKO cells and HMOX2 KO cells, suggesting that HMOX2-mediated haem-derived iron is important for the stability of SDHB and Rieske (Fig. 6e,f). These findings suggest that HMOX2-mediated haem-derived iron availability is essential for mitochondrial iron homeostasis and the stability of mitochondrial respiratory complexes, thereby sustaining mitochondrial respiration.Fig. 6Haem-derived iron is essential for mitochondrial respiration.a, Schematic of the experimental design to evaluate the role of haem-derived iron in mitochondria. Ctrl, control; HD, haem depletion; SA, succinylacetone; FAC, ferric ammonium citrate. b,c, Rieske and SDHB protein levels are stabilized in a haem-dependent manner. b, IB analysis of lysates from cells cultured under the four conditions shown in a. α-tubulin serves as loading control. c, Rieske and SDHB protein levels, quantified and normalized by α-tubulin from b. Fold changes relative to Ctrl cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001, **P < 0.01, *P < 0.05. d, Mitochondrial respiration is maintained in a haem-dependent manner. Basal respiration was measured using an Oxygraph-2k instrument in cells cultured under the four conditions shown in a. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; **P < 0.01, *P < 0.05. e, HMOX1-mediated haem degradation does not affect Rieske and SDHB protein stability upon haemin treatment. Left: IB analysis of lysates from cells cultured under the indicated conditions. SDHA, ATP5A and Tom20 serve as mitochondrial protein controls. α-tubulin and vinculin serve as the loading control. Right: Rieske and SDHB protein levels were quantified and normalized by vinculin from the left panel. Fold changes relative to Ctrl cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; **P < 0.01, *P < 0.05. f, HMOX2-mediated haem degradation stabilizes Rieske and SDHB protein levels upon haemin treatment. Top: IB analysis of lysates from cells cultured under the indicated conditions. α-tubulin and vinculin serve as loading control. Bottom: Rieske and SDHB protein levels were quantified and normalized by vinculin from the top panel. Fold changes relative to Ctrl cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ***P < 0.001, *P < 0.05. Source numerical data and unprocessed blots are available as Source data. Illustrations in a created in BioRender; Oshio, H. https://BioRender.com/v093d2u (2026).Source data Haem-derived iron is essential for mitochondrial respiration. a, Schematic of the experimental design to evaluate the role of haem-derived iron in mitochondria. Ctrl, control; HD, haem depletion; SA, succinylacetone; FAC, ferric ammonium citrate. b,c, Rieske and SDHB protein levels are stabilized in a haem-dependent manner. b, IB analysis of lysates from cells cultured under the four conditions shown in a. α-tubulin serves as loading control. c, Rieske and SDHB protein levels, quantified and normalized by α-tubulin from b. Fold changes relative to Ctrl cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001, **P < 0.01, *P < 0.05. d, Mitochondrial respiration is maintained in a haem-dependent manner. Basal respiration was measured using an Oxygraph-2k instrument in cells cultured under the four conditions shown in a. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; **P < 0.01, *P < 0.05. e, HMOX1-mediated haem degradation does not affect Rieske and SDHB protein stability upon haemin treatment. Left: IB analysis of lysates from cells cultured under the indicated conditions. SDHA, ATP5A and Tom20 serve as mitochondrial protein controls. α-tubulin and vinculin serve as the loading control. Right: Rieske and SDHB protein levels were quantified and normalized by vinculin from the left panel. Fold changes relative to Ctrl cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; **P < 0.01, *P < 0.05. f, HMOX2-mediated haem degradation stabilizes Rieske and SDHB protein levels upon haemin treatment. Top: IB analysis of lysates from cells cultured under the indicated conditions. α-tubulin and vinculin serve as loading control. Bottom: Rieske and SDHB protein levels were quantified and normalized by vinculin from the top panel. Fold changes relative to Ctrl cells are shown. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ***P < 0.001, *P < 0.05. Source numerical data and unprocessed blots are available as Source data. Illustrations in a created in BioRender; Oshio, H. https://BioRender.com/v093d2u (2026). Source data The MITOL–HMOX2 axis maintains mitochondrial iron dynamics and respiratory supercomplexes in C2C12 cells Recent studies have reported that HMOX2-mediated iron serves as a key source for cell proliferation22. Therefore, to analyse the cellular importance of HMOX2-mediated mitochondrial iron maintenance, we used C2C12 cells, which have a higher mitochondrial dependency than HeLa cells. To elucidate the role of MITOL-mediated HMOX2 regulation in C2C12 cells, we generated HMOX2 KO C2C12 cells and re-expressed HMOX2 WT or K68R mutant in HMOX2 KO C2C12 cells (Fig. 7a). Consistent with HeLa cell experiments, mitochondrial iron levels were significantly reduced in HMOX2 KO C2C12 cells compared to WT C2C12 cells, and re-expressing HMOX2 WT in HMOX2 KO cells restored iron levels, but not in cells re-expressing the HMOX2 K68R mutant (Fig. 7b).Fig. 7The MITOL–HMOX2 axis maintains mitochondrial iron dynamics and respiratory supercomplexes in C2C12 cells.a, Generation of mHMOX2 KO C2C12 cells and re-expressed lines with HA-tagged hHMOX2. mHMOX2 KO cells were generated by the CRISPR–Cas9 system with the gRNA shown in Supplementary Table 1. Lysates from the indicated C2C12 cell lines were analysed by IB using the indicated antibodies. α-tubulin serves as loading control. b, MITOL-mediated HMOX2 ubiquitination maintains mitochondrial iron levels in C2C12 cells. The abundance of metals in the indicated cells was measured by ICP–MS as described in Fig. 4a. Results were normalized to mitochondrial protein content (‘pg metal per μg protein’). Data are presented as mean ± s.d. from three independent experiments. Two-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001. c,d, MITOL-mediated HMOX2 ubiquitination stabilizes mitochondrial respiratory complex formation and activity. Mitochondrial fractions isolated from the indicated C2C12 cell lines were subjected to CN–PAGE followed by in-gel activity assays (c) or IB analysis (d). Coomassie Brilliant Blue (CBB) was used as loading control. SC, mitochondrial respiratory supercomplexes. e, MITOL-mediated HMOX2 ubiquitination sustains cell proliferation. The relative proliferation rate of the indicated C2C12 cell lines is shown as a line chart (left) and a bar graph comparing values at 24 h (right). Viable cells were quantified using the Cell Counting Kit-8 assay. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; *P < 0.05. Source numerical data and unprocessed blots are available as Source data.Source data The MITOL–HMOX2 axis maintains mitochondrial iron dynamics and respiratory supercomplexes in C2C12 cells. a, Generation of mHMOX2 KO C2C12 cells and re-expressed lines with HA-tagged hHMOX2. mHMOX2 KO cells were generated by the CRISPR–Cas9 system with the gRNA shown in Supplementary Table 1. Lysates from the indicated C2C12 cell lines were analysed by IB using the indicated antibodies. α-tubulin serves as loading control. b, MITOL-mediated HMOX2 ubiquitination maintains mitochondrial iron levels in C2C12 cells. The abundance of metals in the indicated cells was measured by ICP–MS as described in Fig. 4a. Results were normalized to mitochondrial protein content (‘pg metal per μg protein’). Data are presented as mean ± s.d. from three independent experiments. Two-way ANOVA with Tukey’s multiple comparisons; NS, not significant; ****P < 0.0001. c,d, MITOL-mediated HMOX2 ubiquitination stabilizes mitochondrial respiratory complex formation and activity. Mitochondrial fractions isolated from the indicated C2C12 cell lines were subjected to CN–PAGE followed by in-gel activity assays (c) or IB analysis (d). Coomassie Brilliant Blue (CBB) was used as loading control. SC, mitochondrial respiratory supercomplexes. e, MITOL-mediated HMOX2 ubiquitination sustains cell proliferation. The relative proliferation rate of the indicated C2C12 cell lines is shown as a line chart (left) and a bar graph comparing values at 24 h (right). Viable cells were quantified using the Cell Counting Kit-8 assay. Data are presented as mean ± s.d. from three independent experiments. One-way ANOVA with Tukey’s multiple comparisons; NS, not significant; *P < 0.05. Source numerical data and unprocessed blots are available as Source data. Source data To further validate the HMOX2-associated phenotype, we performed the same analyses in MITOL knockout cells. MITOL knockout in both HeLa and C2C12 cells resulted in reduced mitochondrial iron levels (Extended Data Fig. 10a,b). Re-expression of WT MITOL restored the iron reduction in MITOL KO C2C12 cells, whereas the E3-inactive MITOL CS mutant failed to do so (Extended Data Fig. 10b). Consistent with the changes in mitochondrial iron levels, SDHB and Rieske protein levels as well as OCR were reduced in MITOL knockout C2C12 cells and were restored by re-expression of WT MITOL but not by the CS mutant (Extended Data Fig. 10c–e). These results support a role for MITOL-dependent ubiquitination in promoting HMOX2 activity and downstream mitochondrial iron utilization and function. Clear-native polyacrylamide gel electrophoresis (CN–PAGE) and in-gel activity assays were performed to investigate the effect on the activity of mitochondrial respiratory complexes (Fig. 7c). Notably, the activity of high-molecular-weight mitochondrial complexes, known as mitochondrial respiratory supercomplexes (SC), including complexes I, III and IV and designated as ‘SC’ in the figure, was decreased in HMOX2 KO cells and HMOX2 K68R re-expressing cells compared to WT cells and HMOX2 WT re-expressing cells (Fig. 7c). IB data also show that mitochondrial complex formation, especially for high-molecular-weight supercomplexes, were disrupted in HMOX2 KO cells and HMOX2 K68R re-expressing cells compared to WT cells and HMOX2 WT re-expressing cells (Fig. 7d). Finally, cell proliferation was reduced in HMOX2 KO cells and was restored depending on mitochondrial function (Fig. 7e). Taken together, these findings demonstrate that MITOL-dependent HMOX2 regulation in C2C12 cells is essential for the supply of haem-derived iron to mitochondria, ensuring mitochondrial respiratory supercomplex formation, which is critical for cell proliferation. Discussion HMOX catalyses intracellular haem degradation, producing biliverdin, iron and CO. In particular, HMOX1 is well known as an antioxidant enzyme that is induced by haem accumulation and functions to reduce intracellular haem levels. Indeed, we observed a haem accumulation response, a decrease in ALAS1 expression level, which correlated with a reduction in HMOX1 (Fig. 5c). These data suggest that intracellular haem modulation is governed by HMOX1, but not by HMOX2. In addition, the correlation between HMOX1 and ferritin expression suggests that iron produced by HMOX1 is stored in ferritin41,42. However, the dynamics of haem-derived iron through HMOX2 remain poorly characterized. In this study we have found that HMOX2 specifically maintains mitochondrial total iron levels (Fig. 4c,e), but does not affect ER or cytosolic iron levels (Extended Data Fig. 8a). Our results show that mitochondrial total iron levels are significantly decreased in HMOX2 KO cells, whereas total and mitochondrial haem levels remain unchanged (Fig. 4c). Furthermore, our findings indicate that HMOX2 is critical for the stability of Rieske and SDHB, as well as for supercomplex formation. Rieske and SDHB are integral components of mitochondrial respiratory complexes as Fe–S cluster-containing proteins, and have also been reported to influence supercomplex formation43,44. Because supercomplex formation plays a role in enabling efficient electron transport in OXPHOS43–45, HMOX2 may contribute to supercomplex regulation, thereby enhancing mitochondrial energy production. MERCs play an essential role in regulating the metabolic processes necessary to maintain mitochondrial homeostasis2. It has been reported previously that MERCs are involved in the transport of calcium and lipids, but our study uncovers a role for MERCs in iron supply. These findings indicate that MERCs function as a regulatory hub for iron homeostasis, and further investigation of any additional functions is warranted. Iron is essential for mitochondrial function, but excessive iron leads to severe oxidative stress and mitochondrial dysfunction19. Because mitochondria are the primary source of reactive oxygen species (ROS), mitochondrial iron levels need to be tightly regulated. The MERCs-mediated iron supply system is regulated in close proximity to mitochondria, ensuring efficient iron delivery to mitochondria. Additionally, this process is controlled by MITOL, a mitochondrial outer-membrane protein, ensuring a responsive mechanism for mitochondrial iron demand. Recent studies have revealed that MERCs play a role in the regulation of cell division, and HMOX2 has been identified as an iron source for cell proliferation22,46. Taken together, these previous studies and our data (Fig. 7e) indicate a strong correlation among MERCs, HMOX2, mitochondrial iron utilization and cell proliferation. Our study indicates that MERCs are not merely a site of metabolite exchange, but function as a regulatory hub for mitochondrial iron utilization and cell proliferation. Future studies should investigate the molecular mechanisms of signalling at MERCs to contribute to the understanding of how the MITOL–HMOX2 pathway responds to mitochondrial iron demand. This study demonstrates that MITOL-mediated K63-linked polyubiquitination at K68 of HMOX2 plays a crucial role in maintaining mitochondrial iron homeostasis. HMOX1 is transcriptionally regulated, but HMOX2 is instead regulated at the post-translational level, suggesting that MITOL-mediated ubiquitination serves as a regulatory switch that modulates HMOX2 enzymatic activity in response to mitochondrial iron demand34,47,48. HMOX2 contains haem regulatory motifs (HRMs) in its C-terminal region, a feature absent in HMOX1. These HRMs contain Cys-Pro motifs, which undergo redox-dependent conformational changes that affect haem-binding and enzymatic activity26,49. HRMs are hypothesized to function as transient haem-binding sites that facilitate haem transfer to the catalytic domain to accelerate haem degradation. It has been established that the HMOX2 HW/GW mutant acts as an apo-HMOX2 mimic that allows us to study haem transfer dynamics, due to haem-binding at the HRM and the lack of haem in the catalytic core49. Additionally, the HMOX2 G163H mutant, which is an enzymatically inactive mutant, has high haem affinity in the catalytic core and may impair haem transfer ability49. In this study, we found that the HMOX2 HW/GW mutant is better ubiquitinated by MITOL than the HMOX2 WT or HMOX2 G163H mutants (Extended Data Fig. 4b). Although the HMOX2 C265A mutant is ubiquitinated by MITOL, we observed that its intracellular activity is similar to that of the HMOX2 K68R mutant (Fig. 3f), suggesting a relationship between MITOL-mediated ubiquitination and HRM-mediated haem transfer activity. One possibility is that ubiquitination enhances the structural flexibility of HMOX2, thus promoting haem transfer from HRMs to the catalytic site. However, the direct effect of MITOL-mediated ubiquitination on this process remains to be elucidated. Moreover, considering the effect of MITOL-mediated K63-linked ubiquitination on oligomer formation11,12,50, the ubiquitination of HMOX2 may recruit proteins, such as haem-binding proteins, to improve haem uptake efficiency. In addition, MITOL-mediated ubiquitination of HMOX2 is enhanced under iron deficiency, and the C-terminal region of HMOX2 is required for ubiquitination (Extended Data Figs. 3a,b and 6), suggesting that the C-terminal region of HMOX2 is a receptor of mitochondrial iron deficiency signals. However, the mechanism by which mitochondrial iron demand is communicated remains unexplored. Future studies should aim to identify the signalling pathways and molecular sensors involved in this process to further elucidate this regulatory mechanism. This study suggests that haem is a more critical source for mitochondrial iron utilization than non-haem iron-uptake pathways, thereby supporting the stability of Rieske and SDHB as well as mitochondrial respiration. There are two main iron-uptake pathways: haem-derived and non-haem iron pathways. The non-haem iron pathway is mediated by divalent metal transporter 1 (DMT1), duodenal cytochrome b (Dcytb), and others, but recent evidence suggests that the haem-derived pathway becomes essential when the non-haem iron-uptake pathway is impaired22,51. Furthermore, haem is believed to be more efficiently absorbed by cells than non-haem iron, and some reports indicate that haem uptake via haem transporters increases during iron deficiency52,53. These findings highlight the importance of haem-derived iron as a source for mitochondrial iron homeostasis and utilization. In contrast, our data also show that treatment with excess haemin or FAC in HMOX2 KO cells partially restored mitochondrial iron levels (Fig. 4d), suggesting that other pathways are also functioning. Further investigation is needed to clarify the interplay and balance between the MITOL–HMOX2 pathway and other mechanisms. This study identifies a mechanism through which HMOX2 supplies haem-derived iron to mitochondria via MITOL-mediated ubiquitination. The process of iron supply to mitochondria maintains protein stability, supercomplex formation, mitochondrial respiration, and cell proliferation. Future studies should focus on elucidating the molecular mechanisms of iron transport from MERCs to mitochondria and gaining a deeper understanding of the specific contribution of MITOL–HMOX2 to mitochondrial iron requirements. Furthermore, dysregulation of mitochondrial iron metabolism has been implicated in neurodegenerative diseases, mitochondrial diseases and cancer, among other conditions18,54–56. Investigating how the dysfunction of the MITOL–HMOX2 axis leads to mitochondrial impairment may provide insights into disease intervention and therapeutic development. Nevertheless, several mechanistic aspects remain to be clarified. Previous studies have reported that MITOL-mediated K63-linked ubiquitin chains facilitate substrate activity. However, in this study, we were unable to determine how MITOL-mediated ubiquitination influences HMOX2 enzymatic function or even whether it directly enhances haem degradation. It is plausible that ubiquitination facilitates haem transfer, but assessing intracellular haem transfer dynamics remains challenging. Further structural and biochemical analyses are necessary to precisely elucidate the impact of ubiquitination on HMOX2 activity. Due to the technical limitations of cellular haem tracking, the source of haem for HMOX2, which is degraded and utilized as an iron source for mitochondria, remains unknown. One possibility is that HMOX2 accesses a localized pool of haem at MERCs, where lipid and metabolite exchange are known to occur. In this context, the lack of detectable changes in bulk ER or cytosolic haem levels in our measurements does not exclude their contribution, as only a small fraction of the total cellular haem pool may be directed toward mitochondrial iron replenishment. Future studies will be required to directly trace the origin and dynamics of the haem pool utilized by HMOX2. Additionally, this study did not resolve the question of mitochondrial iron uptake at the outer membrane, including the specific molecules involved, which remain poorly understood. Although the molecular components at the outer mitochondrial membrane that may facilitate this process are not yet defined, iron import across the inner mitochondrial membrane is known to be mediated by mitoferrins (MFRN1/2), which form multiprotein complexes involved in mitochondrial iron metabolism. In this context, it is conceivable that HMOX2–MITOL activity could be functionally coupled to such inner-membrane transport machinery, although no direct interaction between MITOL and mitoferrins has been reported as yet. Furthermore, how this pathway is integrated with mitochondrial iron deficiency response mechanisms remains to be determined. Elucidating the molecular linkage between HMOX2–MITOL-dependent iron release and mitochondrial iron import will be an important direction for future studies. The physiological functions of HMOX2-derived metabolites, such as CO and biliverdin, also remain largely unexplored. Further investigations are needed to clarify the roles of CO and biliverdin in MERCs and the relationships with the MITOL–HMOX2 pathway. Methods Cell culture HeLa cells and C2C12 cells were cultured in Dulbecco’s modified Eagle medium (DMEM, 08458-16, Nacalai Tesque) supplemented with 10% FBS and penicillin/streptomycin. Cells were maintained in a humidified incubator at 37 °C with 5% CO2. Haem-depleted (HD) FBS was prepared as previously described, with modifications39,40. Briefly, after treatment with ascorbate for 8 h, three rounds of dialysis were performed at 8-h intervals. Haem depletion was confirmed by measuring absorbance at 405 nm, ensuring a reduction of more than 50%. HD medium was prepared by supplementing DMEM with 10% HD FBS, penicillin/streptomycin and 0.5 mM succinylacetone (SA). The cells were cultured in HD medium for 72 h to induce haem depletion. For haem repletion, 48 h after haem depletion, the medium was replaced with HD medium containing 20 µM haemin (HD + haemin) for an additional 24 h. For excess iron treatment, 48 h after haem depletion, the medium was replaced with HD medium containing 1 mM ferric ammonium citrate (HD + FAC) or 2 mg ml−1 holo-transferrin (HD + holo-Tf) for an additional 24 h. Cells were treated with cycloheximide (CHX, 5 µg ml−1, 24 h) and MG132 (30 µM, 3 h) before sampling. Transfection and reagents Cells were transfected using Lipofectamine 3000 (100022052, Invitrogen), FuGENE HD (E2311, Promega), Lipofectamine RNAiMAX (13778-150, Invitrogen) or PEI MAX (24765-100, Polysciences) according to the manufacturer’s instructions. The following siRNAs were used: siPDZD8, 5′-GTTTAAGCCGTTTTTTCCATT-3′; siVAPB, 5′-GCTCTTGGCTCTGGTGGTTTT-3′; siMFRN1(SLC25A37), 5′-GCCATTTCTTGGTCTGTCTAT-3′; siMITOL 5′-AACCAUUAUCUUAGCAUGGUA-3′. Scramble siRNA was purchased from Qiagen. siPDZD8 was obtained from Japan Bio Services, and siVAPB and siMFRN1 were as described previously57,58. Plasmids All expression vectors were constructed or mutated using the In-Fusion HD Cloning Kit (Takara Bio). HMOX1 (NM_002133.3) and HMOX2 (NM_001127204.2) were amplified from HeLa complementary DNA and cloned into pRK5 with an N-terminal HA tag. MITOL WT and the C65/68S mutant expression vectors were generated as previously described9. PA-tagged MITOL was generated as previously described14. Myc-tagged or V5-tagged MITOL was subcloned into pRK5. FLAG-ubiquitin was generated from HA-ubiquitin WT (Addgene 17608), a gift from T. Dawson via Addgene59. To generate an endogenous HMOX activity reporter or an HMOX2 activity reporter, HA or HA-tagged HMOX2, respectively, and IFP2.0-P2A-EGFP were constructed in an optimized pRK5 vector. IFP2.0-P2A-EGFP was derived from IFP2.0-C1 (Addgene 54784) and fused to EGFP via the P2A sequence (ATNFSLLKQAGDVEENPGP). The expression of IFP2.0-P2A-EGFP was driven by the EF-1α promoter, and HA or HA-HMOX2 was expressed under the phosphoglycerate kinase 1 (PGK) promoter. To generate Myc-TurboID-MITOL stably expressing cells, Myc-TurboID-MITOL was derived from pCDNA3 V5-TurboID-NES (Addgene 107169) and subcloned into the pCX4pur retroviral vector (GenBank ID AB086386.1)60. To generate hHMOX1-KO cells, hHMOX2 KO cells, DKO cells, hPDZD8 KO cells, hVAPB KO cells, mHMOX2 KO cells and mMITOL KO cells, CRISPR–Cas9 vectors were constructed. CRISPR gRNAs targeting human HMOX1, HMOX2, PDZD8 and VAPB and mouse HMOX2 and MITOL were designed using the online tool CRISPOR.org. The designed oligonucleotide pairs (Supplementary Table 1) were annealed and ligated into the PX459 vector (Addgene 48139)61. Generation of KO cell lines MITOL KO HeLa cells were generated as previously described62. HMOX1 KO, HMOX2 KO, DKO, PDZD8 KO and VAPB KO cells were generated from WT HeLa cells in this study. HMOX2 KO C2C12 cells and MITOL KO C2C12 cells were generated from WT C2C12 cells. To generate KO cell lines, HeLa cells or C2C12 cells were transfected with the PX459 vector ligated with each gRNA (Supplementary Table 1) using Lipofectamine 3000. After 24 h, cells were selected with 2 µg ml−1 puromycin for 48 h, followed by single-cell clone isolation in 96-well plates using the limiting dilution method. DKO cells were generated from single-cloned HMOX2 KO #1 cells by transfecting them with the PX459 vector targeting HMOX1 #2. Generation of 3×HA–HMOX2 KI cells Knock-in at the HMOX2 genome locus was performed using the PITCh method, as previously described63. Plasmid pX330A-HMOX2/PITCh, an all-in-one CRISPR-Cas9 vector for cutting the genomic HMOX2 exon 3 locus and the donor vector, was constructed using the pX330A-1×2 (Addgene 58766) plasmid and pX330S-2-PITCh (Addgene 63670). The oligonucleotides for the template of sgRNA targeting HMOX2 exon 3 are CRISPR-hHMOX2_sense (5′-caccgAGCAGCAATGTCAGCGGAAG-3′) and CRISPR-hHMOX2_antisense (5′-aaacCTTCCGCTGACATTGCTGCTc-3′). The donor vector, including PuroR, 3×HA and microhomologies, was constructed from pCRIS-PITChv2-FBL (Addgene 63672). HeLa cells were co-transfected with donor vector and pX330A-HMOX2/PITCh. After 72 h of transfection, cells were selected with 1 μg ml−1 of puromycin. Genotyping Genomic DNA of the 3×HA–HMOX2 knock-in cells used in this study were extracted with genotyping buffer (10 mM Tris-HCl pH 8.0, 10 mM ethylenediaminetetraacetic acid (EDTA), 100 mM NaCl, 0.5% sodium dodecyl sulfate (SDS), 0.4 mg ml−1 proteinase K), followed by NaCl-mediated protein precipitation and isopropanol DNA precipitation. DNA pellets were washed with 70% ethanol and resuspended in Tris–EDTA buffer, and then polymerase chain reaction (PCR) was amplified using a pair of primers (forward, 5′-ctgaggaaagcagaaccgacaggtatttg-3′; reverse, 5′-gctgggcccacctgattcagtgcag-3′) to confirm the insertion. PCR products were analysed by agarose gel electrophoresis in Tris–acetate–EDTA buffer. Generation of stable cell lines Stable cell lines were generated using a retroviral expression system, as previously described60. Briefly, Plat-A retroviral packaging cells were transfected with pCX4pur retroviral vectors using Lipofectamine 3000. After a few days, the culture supernatant containing retrovirus was collected and filtered through a 0.22-µm-pore filter (SLGSR33SS, Millipore). HeLa cells or C2C12 cells were infected with retrovirus in medium containing Polybrene (H9268, Sigma-Aldrich; 12996-81, Nacalai Tesque) for a few days, followed by selection with 1 µg ml−1 of puromycin. Myc-TurboID-MITOL stably expressing cells were generated from WT HeLa cells, and single-cell clones were isolated in 96-well plates using the limiting dilution method. HA-HMOX2 re-expressed HeLa cells were generated from single-cell cloned HMOX2 KO #1 HeLa cells, and single-cell clones were isolated in 96-well plates using the limiting dilution method. HA-HMOX2 re-expressed C2C12 cells were generated from single-cell cloned HMOX2 KO C2C12 cells, and polyclonal cells were used in this study. HA-MITOL re-expressed C2C12 cells were generated from single-cell cloned MITOL KO C2C12 cells, and polyclonal cells were used in this study. Antibodies The following primary antibodies were used: anti-HMOX2 (Cell Signaling Technology, 32790, 1:1,000), anti-HMOX1 (Proteintech, 10701-1-AP, 1:1,000), anti-calnexin (BioLegend, 699402, 1:1,000), anti-MITOL (LSBio (LifeSpan), LS-C164034, 1:1,000), anti-Myc (Cell Signaling Technology, 2276, 1:1,000), anti-Myc (Cell Signaling Technology, 2278, 1:1,000), anti-MFN2 (Santa Cruz Biotechnology, sc-100560, 1:1,000), anti-RMDN3 (Atlas Antibodies, HPA009975, 1:1,000), anti-PDZD8 (Proteintech, 25512-1-AP, 1:1,000), anti-α-tubulin (Sigma-Aldrich, T9026, 1:5,000), anti-IP3R-I/II/III (Santa Cruz Biotechnology, sc-377518, 1:1,000), anti-VAPB (Proteintech, 66191-1-Ig, 1:1,000), anti-Tom20 (Proteintech, 11802-1-AP, 1:1,000), anti-NDUFA9 (Abcam, ab14713, 1:1,000), anti-NDUFB8 (Abcam, ab110242, 1:1,000), anti-succinate dehydrogenase complex flavoprotein subunit A (SDHA; Abcam, ab14715, 1:1,000), anti-SDHA (Cell Signaling Technology, 11998, 1:500), anti-UQCRC2 (Abcam, ab14745, 1:1,000), anti-ATP5A (Abcam, ab14748, 1:1,000), anti-FLAG-horseradish peroxidase (HRP) (Sigma-Aldrich, A8592, 1:1,000), anti-ubiquitin (P4D1) (Santa Cruz Biotechnology, sc-8017, 1:1,000), anti-HA (Roche, 11867423001, 1:1,000), anti-HA (Covance, MMS-101R, 1:1,000), anti-HA (Cell Signaling Technology, 3724, 1:1,000), anti-HA-HRP (Roche, 12013819001, 1:1,000), anti-2A peptide (Sigma-Aldrich, MABS2005, 1:1,000), anti-vinculin (Sigma-Aldrich, V9131, 1:1,000), anti-GFP (MBL International, M048-3, 1:1,000), anti-Rieske (Santa Cruz Biotechnology, sc-271609, 1:500), anti-SDHB (Abcam, ab14714, 1:1,000), anti-ALAS1 (Abcam, ab154860, 1:1,000), anti-FTH1 (Cell Signaling Technology, 3998, 1:1,000), anti-MTCO1 (Abcam, ab14705, 1:1,000), anti-lipoic acid (Abcam, ab58724, 1:1,000), anti-ferrochelatase (FECH) (Santa Cruz Biotechnology, (A-3) sc-377377, 1:500), anti-aconitase 2 (ACO2) (Proteintech, 11134-1-AP, 1:1,000), anti-iron regulatory protein 2 (IRP2) (Cell Signaling Technology, 37135, 1:1,000), anti-mitoferrin-1 (MFRN1) (Proteintech, 26469-1-AP, 1:1,000), anti-heat shock protein 60 (HSP60) (Santa Cruz Biotechnology, sc-136291, 1:1,000), anti-cyclooxygenase 1 (COX1) (Thermo Fisher Scientific, PA5-26688, 1:1,000), anti-V5 (Cell Signaling Technology, 80076, 1:1,000), anti-V5 (Proteintech, 14440-1-AP, 1:1,000) and anti-transferrin (Proteintech, 17435-1-AP, 1:1,000). Immunoblotting Before SDS–PAGE analysis, samples were boiled in 2× Laemmli sample buffer (125 mM Tris-HCl, pH 6.8, 10% 2-mercaptoethanol, 4% SDS, 10% sucrose, 0.01% bromophenol blue (BPB)) for 5 min. Protein concentrations were quantified using the TaKaRa BCA protein assay kit (T9300A, Takara Bio) or protein quantification assay (74967.50/250, Takara Bio) if needed. Proteins were separated by SDS–PAGE or CN–PAGE and transferred onto Immobilon polyvinylidene fluoride (PVDF) membranes (IPVH00010, Millipore). Blots were probed with the indicated antibodies, and protein bands were visualized using Chemi-Lumi One L (07880-70, Nacalai Tesque) or Immobilon Western Chemiluminescent HRP substrate (WBKLS0500, Millipore). Band images were acquired using a LuminoGraph I imager (WSE-6100, ATTO), and relative band intensities were quantified using Fiji/ImageJ software. Immunofluorescence analysis Immunofluorescence was performed as previously described14. Cells were fixed with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 15 min at 37 °C, quenched with 50 mM NH4Cl in PBS for 10 min at room temperature (r.t.), permeabilized with 0.1% Triton X-100 in PBS for 10 min at r.t., and blocked with 1% bovine serum albumin (BSA) in PBS for 10 min at r.t. Cells were washed three times with PBS between each step. For immunostaining, cells were incubated with the indicated primary antibodies in PBS containing 0.5% BSA for 1 h at r.t, washed three times with PBS, and incubated with the appropriate secondary antibodies for 1 h at r.t. After washing three times with PBS, samples were mounted with fluorescence mounting medium (S3023, Dako) and analysed using an FV3000 confocal laser scanning microscope (Olympus). Three z-stack slices were acquired every 0.2 μm and reconstructed using Fiji/ImageJ software. Biotin labelling, digitonin assay and purification Biotin labelling was performed as previously described64. Myc-TurboID-MITOL-expressing cells were cultured in two 100-mm dishes. Cells were incubated in fresh medium containing 50 µM biotin or dimethylsulfoxide (DMSO) for 10 min at 37 °C for labelling, followed by quenching with ice-cold PBS washes. Digitonin assay and purification were performed as previously described14, with some modifications. Digitonin buffer was prepared by serial dilution of 1% digitonin in resuspension buffer (50 mM HEPES-KOH, pH 7.4, 150 mM NaCl). Cells were collected, washed by centrifugation at 1,000g for 3 min at 4 °C, resuspended in PBS, and pelleted by centrifugation at 500g for 3 min. Cells were then resuspended in 0.05% digitonin buffer and incubated for 10 min on ice. After centrifugation at 10,000g for 5 min at 4 °C, the supernatant containing cytosolic proteins was collected, and the pellet was washed with resuspension buffer. The pellet was resuspended in 1% digitonin buffer, incubated for 10 min on ice, and centrifuged at 10,000g for 5 min at 4 °C. The supernatant containing membrane proteins was collected, and the pellet was washed with resuspension buffer, centrifuged at 10,000g for 5 min at 4 °C, and resuspended in radio-immunoprecipitation assay (RIPA) lysis buffer. After incubation for 10 min on ice, the supernatant containing RIPA-solubilized components was collected by centrifugation at 10,000g for 5 min at 4 °C. Protein concentrations of each collected supernatant were determined using the TaKaRa BCA protein assay kit (T9300A, Takara Bio) and adjusted to 0.5 µg µl−1. Each supernatant was incubated with 10 µl of Dynabeads MyOne Streptavidin C1 (65001, Invitrogen) by rotation at 4 °C for 1 h. Beads were washed four times with RIPA lysis buffer and pelleted using a magnetic rack. The washed beads were separated for liquid chromatography tandem mass spectrometry (LC–MS/MS) analysis and IB assay. Biotin-labelled proteins were eluted by boiling in 2× Laemmli sample buffer for 5 min prior to IB analysis. Subcellular fractionation Subcellular fractionionation was carried out for ICP–MS analysis, CN–PAGE and the measurement of haem levels. Cells were homogenized in homogenization buffer (HB; 10 mM HEPES-KOH, pH 7.4, 250 mM sucrose, 1 mM EDTA) by passing through a 27-G needle with a 1-ml syringe five times on ice. The homogenate (whole fraction) was centrifuged at 800g for 10 min at 4 °C, and the supernatant was further centrifuged at 8,000g for 10 min at 4 °C. The resulting pellet (crude mitochondrial fraction, containing MAM) was collected, and the supernatant was centrifuged at 17,400g for 10 min at 4 °C to obtain the supernatant (ER and cytosol fraction). To isolate the ER fraction, the ER and cytosol fraction was centrifuged at 100,000g for 1 h at 4 °C. The resulting pellet (ER fraction) and supernatant (cytosol fraction) were collected. For ICP–MS analysis, HB was optimized without EDTA, and the crude mitochondrial fraction was resuspended in PBS. Protein concentrations were determined using the TaKaRa BCA protein assay kit (T9300A, Takara Bio). Isolating pure mitochondrial fraction and MAM enriched fraction The pure mitochondrial fraction and MAM enriched fraction were isolated as previously described65,66. WT HeLa cells were cultured in five 150-mm dishes. Cells were washed with PBS, collected, and pelleted by centrifugation at 300g for 2 min at 4 °C. The cell pellet was washed by resuspension in PBS, briefly pelleted by centrifugation at 13,000g for 30 s at 4 °C, and the supernatant was removed completely. The pellet was then resuspended in HB and homogenized as described above. The homogenate was centrifuged twice at 600g for 10 min at 4 °C to remove the pellet containing nuclear and debris fractions. The resulting supernatant was centrifuged at 8,000g for 10 min at 4 °C, yielding the crude mitochondrial fraction (pellet) and the supernatant, which was collected for ER and cytosol fraction isolation as described above. The crude mitochondrial fraction (pellet after centrifugation at 8,000g) was resuspended in isolation medium (IM; 5 mM HEPES-KOH, pH 7.4, 250 mM mannitol, 0.5 mM EDTA) and subjected to Percoll density gradient centrifugation. The crude mitochondrial suspension was layered on top of isolation medium 2 (IM2; 25 mM HEPES-KOH, pH 7.4, 225 mM mannitol, 1 mM EDTA) containing 30% Percoll and centrifuged at 95,000g for 30 min at 4 °C using a swing rotor. A low-density white layer (fraction 2; F2) was collected, diluted tenfold with IM, and centrifuged at 6,300g for 10 min at 4 °C, followed by further centrifugation at 100,000g for 1 h at 4 °C to purify the pelleted MAM enriched fraction. A high-density wide region (fraction 3; F3) was collected, diluted tenfold with IM, and centrifuged at 6,300g for 10 min at 4 °C to purify the pelleted pure mitochondrial fraction (without MAM contamination). Each fraction was lysed in NP-40 buffer for 20 min on ice and centrifuged at 15,000g for 10 min. The protein concentration of the supernatant was adjusted to ensure equal protein levels across fractions. Sample preparation of biotinylated proteins for LC–MS/MS analysis The protein-bound beads prepared by following the abovementioned pulldown purification were transferred to a new tube and then washed twice with 25 mM ammonium bicarbonate buffer. The beads were suspended in 500 µl of 25 mM ammonium bicarbonate buffer containing 5 mM dithiothreitol (DTT) and incubated at 65 °C for 20 min. After discarding the supernatant, 500 µl of 5 mM iodoacetamide (IAA) in 25 mM ammonium bicarbonate buffer was added to the beads, and the suspension was further incubated at r.t. for 30 min in the dark. After discarding the supernatant, the proteins on beads were digested by 0.5 µg of Trypsin Platinum (mass spec grade, Promega) in 500 µl of 25 mM ammonium bicarbonate buffer at 37 °C overnight with shaking. The supernatant was collected in a new tube and acidified with 50% trifluoroacetic acid to quench the reaction, and then desalted in a MonoSpin C18 column (GL Science) according to the manufacturer’s protocol. The desalted eluate was dried in vacuo and dissolved in 2% MeCN, 0.1% formic acid in distilled water for LC–MS/MS measurement. Biotin-treated and biotin-untreated samples were prepared in triplicates, respectively. Sample preparation of proteins in fractionated organelles for quantitative LC–MS/MS analysis The proteins in the resulted lysate were collected by acetone precipitation and dissolved in 25.5 µl of 50 mM ammonium bicarbonate buffer, and then 1.5 µl of 100 mM DTT in 50 mM ammonium bicarbonate buffer was added to the solution before incubation at 60 °C for 30 min. After cooling to r.t., 3 µl of 100 mM IAA in 50 mM ammonium bicarbonate buffer was added to the solution and incubated at r.t. for 30 min in the dark. Trypsin (0.5 µg) was added to the solution and incubated at 37 °C overnight. The reaction mixture was acidified with 10% trifluoroacetic acid (pH < 3) and then desalted by a GL-tip Monospin set-up (GL Science) according to the manufacturer’s protocol. The desalted eluate was dried in vacuo and dissolved in 2% MeCN, 0.1% formic acid in distilled water for LC–MS/MS measurement. Mitochondria, ER and MAM samples were prepared in triplicates, respectively. LC–MS/MS analysis LC–MS/MS analysis was performed on an Ultimate 3000 RSLCnano system (Thermo Fisher Scientific) coupled to a Q Exactive hybrid quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific) equipped with a nano electrospray ionization source. The nanoLC system was equipped with a trap column (C18 PepMap 100, 0.3 × 5 mm, 5 µm, Thermo Fisher Scientific) and an analytical column (NTCC-360/75-3-125, Nikkyo Technos). Peptide separation was performed with a 60-min gradient of water containing 0.1% formic acid (mobile phase A) and acetonitrile containing 0.1% formic acid (mobile phase B) at a flow rate of 300 nl min−1. The elution for biotinylated protein measurement was set as follows: 0–3 min, 2% B; 3–63 min, 2%–40% B; 63–65 min, 40%–95% B; 65–75 min, 95% B; 75–77 min, 95%–2% B; 77–90 min, 2% B. The elution for quantitative protein measurement in organelles was set as follows: 0–3 min, 2% B; 3–93 min, 2%–40% B; 93–95 min, 40%–95% B; 95–105 min, 95% B; 105–107 min, 95%–2% B; 107–120 min, 2% B. Measurements were performed in duplicate for proteomic analysis of biotinylated proteins, and in triplicate for quantitative proteomic analysis of the proteins in organelles, respectively. The mass spectrometer was operated in data-dependent acquisition (DDA) mode. MS parameters were set as follows: spray voltage, 2.0 kV; capillary temperature, 275 °C; S-lens RF level, 50; scan type, full MS; scan range, m/z 350–1,500; resolution, 70,000; polarity, positive; automatic gain control target, 3 × 106; maximum injection time, 100 ms. MS/MS parameters were set as follows: resolution, 17,500; automatic gain control target, 1 × 105; maximum injection time, 60 ms; normalized collision energy, 27; dynamic exclusion, 15 s; loop count, 10; isolation window, 1.6 m/z; charge exclusion, unassigned, 1, 8, >8. LC–MS/MS data processing For proteomic analysis of biotinylated proteins, the identification of peptides and proteins, and label-free quantification (LFQ) of the detected peptides were performed using Proteome Discoverer (version 2.4 SP1, Thermo Fisher Scientific). The analytical parameters used for the data processing of peptides were as follows: parent mass error tolerance, 10.0 ppm; fragment mass error tolerance, 0.02 Da; search engine, sequest HT; protein database, SwissProt (Homo sapiens); enzyme name, trypsin (full); maximum number of missed cleavages, 2; false discovery rate (FDR) confidence <0.01; dynamic modification, oxidation (methionine); static modification, carbamidomethyl (cysteine). Label-free quantification parameters for the detected peptides were set as follows: precursor quantification, precursor abundance based on area; normalization mode, non; and hypothesis test, t-test (background based). For quantitative analysis of proteins in fractionated organelles, the identification of peptides and proteins, and label-free quantification of the detected peptides, were performed by using Proteome Discoverer (version 3.0 SP1, Thermo Fisher Scientific) with a Chimerys node (Thermo Fisher Scientific). The analytical parameters used for the data processing of peptides were as follows: parent mass error tolerance, 10.0 ppm; fragment mass error tolerance, 0.02 Da; search engine, inferys 2.1; protein database, SwissProt (Homo sapiens); enzyme name, trypsin (full); maximum number of missed cleavages, 2; FDR confidence <0.01; dynamic modification, oxidation (methionine); static modification, carbamidomethyl (cysteine). Label-free quantification parameters for the detected peptides were set as follows: precursor quantification, precursor abundance based on area; normalization mode, non; protein ratio calculation, pairwise ratio based; N for TopN, 3; and hypothesis test, t-test (background based). Protein abundance was converted to percentage by dividing by the sum of protein abundance in each fraction to produce a heatmap (Fig. 1g). Immunoprecipitation To examine protein interactions, cells were washed with PBS and treated with 1 mM dithiobis(succinimidyl propionate) (DSP; 345-09121, Dojindo) in PBS for 30 min at r.t. before collection. Cells were lysed by pipetting in NP-40 lysis buffer (1% Nonidet P40 Substitute, 10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 10 mM NaF) containing a protease inhibitor cocktail (25955-11, Nacalai Tesque) and rotated for 20 min at 4 °C. Lysates were centrifuged at 15,000g for 10 min at 4 °C, and the supernatant was subjected to immunoprecipitation using the indicated antibodies and COSMOGEL Ig-Accept Protein G (02198-22, Nacalai Tesque). Beads were washed four times with NP-40 lysis buffer and boiled in 2× Laemmli sample buffer for 5 min. To evaluate protein ubiquitination levels, cells were lysed in NP-40 lysis buffer and sonicated for 30 s on ice. Lysates were centrifuged at 20,000g for 10 min at 4 °C, and the supernatant was subjected to immunoprecipitation using the indicated antibodies and Protein G Sepharose 4 Fast Flow (17-0618-02, Cytiva). Beads were washed four times with RIPA lysis buffer (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% SDS, 0.05% sodium deoxycholate, 1% Triton X-100, 5 mM EDTA) and boiled in 2× Laemmli sample buffer for 5 min. Ubiquitin restriction assay A ubiquitin restriction assay was performed using the UbiCREST kit (Boston Biochem) according to the manufacturer’s instructions. Five recombinant deubiquitinating enzymes (DUB)—OTUD3 (specific for K6-linked polyubiquitin), Cezanne (specific for K11-linked polyubiquitin), OTUB1 (specific for 48-linked polyubiquitin), AMSH (specific for K63-linked polyubiquitin) and USP2 (nonspecific)—were used for the reaction. To prevent inhibition of the DUB reaction, the isolated ER fraction was lysed in protease inhibitor-free NP-40 buffer. HA-HMOX2 was immunoprecipitated using an anti-HA antibody and SureBeads Protein G magnetic beads (161-4023, Bio-Rad Laboratories). HA-HMOX2-conjugated beads were divided into six aliquots and treated either without DUB or with one of the five DUBs for 30 min at 37 °C. After treatment, the beads were washed with RIPA buffer and boiled in 2× Laemmli sample buffer to elute the proteins. Flow cytometry To assess intracellular HMOX activity, cells were transfected with the endogenous HMOX activity reporter or HMOX2 activity reporter 48 h before analysis. Cells were washed with PBS, trypsinized, and collected into tubes for flow cytometry analysis. Flow cytometry was performed using a BD FACSCanto II system (Becton Dickinson). EGFP fluorescence was measured using a 502-nm excitation laser and a 530/30 emission filter, and IFP2.0 fluorescence was measured using a 735-nm excitation laser and a 780/60 emission filter. Data were analysed from flow cytometry standard (FCS) files using FlowJo software. ICP–MS analysis Iron levels were measured using an Agilent 8900 Triple Quadrupole ICP–MS instrument equipped with an autosampler (Agilent Technologies). Argon gas was used for plasma generation and nebulization. The instrument was operated in collision/reaction cell mode with He and O2 gases to minimize polyatomic interferences. Quantification was performed using the external calibration curve method with certified reference materials in combination with the internal standard method. Be, In, Y, Te and Bi were used as internal standards and injected online with the sample. The cell samples were mixed with 0.25 ml of a 70% nitric-acid solution and then digested using an ETHOS EASY microwave (Milestone General) at 160 °C for 30 min. After digestion and cooling, the solutions were diluted with ultrapure water to a final volume of 1.0 ml and adjusted to a nitric-acid concentration of 17.5% for Fe determination by ICP–MS. The monitored isotopes are as follows: 9Be+, 89Y+, 115In+, 125Te+, 209Bi+, 55Mn+, 56Fe+, 63Cu+ and 66Zn+. Measurement of haem levels Haem levels were measured using a haemin colorimetric assay kit (10191-568, BioVision) according to the manufacturer’s instructions. To measure haem levels in whole-cell lysates, cells were lysed in RIPA lysis buffer for 20 min on ice, followed by centrifugation at 15,000g for 10 min at 4 °C. Protein concentrations in the supernatant were measured, and 2.5 µg of total protein in RIPA lysis buffer was subjected to the assay kit. To measure mitochondrial haem levels, the crude mitochondrial fraction was isolated and lysed using the same procedure as for whole-cell lysates. Mitochondrial lysates containing 10 µg of mitochondrial protein in RIPA lysis buffer were subjected to the assay kit. As positive controls, whole-cell lysates from cells treated with 20 µM haemin for 24 h and mitochondrial lysates from cells treated with 1 mM 5-ALA and 1 mM FAC for 24 h were used. Lysates from cells cultured in HD medium for 72 h served as negative controls. Before measuring haem levels in FBS, the FBS was diluted tenfold in water. Measurement of iron concentration Iron concentration in FBS or whole-cell lysates was measured using an iron assay kit LS (Ferrozine Method) (FE31ME, Metallogenics) according to the manufacturer’s instructions. To measure iron concentration in whole-cell lysates, pelleted cells were lysed in optimized 2× RIPA lysis buffer (20 mM Tris-HCl, pH 7.4, 300 mM NaCl, 0.2% SDS, 0.1% sodium deoxycholate, 2% Triton X-100, 0.1 M HCl) for 20 min on ice, followed by centrifugation at 10,000g for 10 min at 4 °C. The supernatant was analysed using the iron assay kit LS or the TaKaRa BCA protein assay kit to determine protein concentration. Iron concentration was normalized to protein concentration after measurement. Measurement of OCRs The OCR in intact cells was monitored using an Oxygraph-2k instrument (Oroboros Instruments) or a Seahorse XFp analyser (Agilent Technologies) according to the manufacturer’s instructions. Oxygraph-2k Before analysis, cells were cultured in the indicated medium in 100-mm dishes. Cells were trypsinized, pelleted by centrifugation at 300g for 3 min at r.t., and resuspended in MiR05. The resuspended cells were counted and adjusted to 2.0 × 106 cells per 2.0 ml MiR05 in a chamber. After stable respiration was recorded, oligomycin (1 µM) was injected to measure adenosine triphosphate (ATP)-linked respiration. To measure the maximal respiration, a step-by-step titration with the uncoupler carbonyl cyanide m-chlorophenyl hydrazone was performed. Subsequently, antimycin A (1 µM) and rotenone (1 µM) were injected to measure the non-mitochondrial OCR. Basal respiration and maximal respiration were calculated as the difference from the non-mitochondrial OCR. ATP-linked respiration was calculated as the difference between steady-state OCR and stabilized OCR after oligomycin injection. Data were recorded and analysed using DatLab 7.4 software. Seahorse XFp analyser Cells were seeded in normal culture medium at a density of 20,000 cells well−1. After 24 h, the medium was replaced with XF base medium (containing 1.25 mM sodium pyruvate, 25 mM glucose and 4 mM glutamine) 1 h before the assay. OCR was measured after sequential exposure to 1.5 µM oligomycin, 0.5 µM carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone, 0.5 µM rotenone and 0.5 µM antimycin A using the Seahorse XF Cell Mito Stress test kit. OCR values were normalized to protein concentration, determined using the TaKaRa BCA protein assay kit. Basal respiration and maximal respiration were calculated after subtracting non-mitochondrial respiration, following the manufacturer’s instructions. CN–PAGE CN–PAGE was performed with modifications from a previously reported method67. The isolated crude mitochondrial fraction was resuspended in HB, and protein concentration was measured using the TaKaRa BCA protein assay kit. To prepare a 20-µl solubilized mitochondrial sample, 20 µg of mitochondrial protein (adjusted to 13 µl in HB) was mixed with 5 µl of 4× sampling buffer (200 mM Bis-Tris, 68 mM HCl, 200 mM NaCl) and 2 µl of 10% digitonin buffer (0.2 mg digitonin in 2 µl of water). This resulted in a final concentration of 1× sample buffer and 1% digitonin per microgram of mitochondrial protein in the sample. The suspension was gently mixed and incubated on ice for 30 min. Solubilized samples were centrifuged at 17,400g for 30 min at 4 °C, and 13.5 µl of the supernatant was mixed with 1.5 µl of 10× loading buffer (50% glycerol, 0.1% Ponceau S). The stock solution was stored at −80 °C. For CN–PAGE, 4 µl of the solubilized mitochondrial sample was loaded and separated on a precast NativePAGE 4–16% Bis-Tris protein gel (BN2112-BX10, Invitrogen). Electrophoresis was performed at 4 °C. After CN–PAGE, the gel was sectioned as needed and subjected to IB assay or in-gel activity assay as described. In-gel activity assay The in-gel activity assay was performed as previously reported67. Briefly, to visualize complex I activities, the gel was washed with 5 mM Tris-HCl (pH 7.4), submerged in 10 ml of staining solution A (0.25% nitro-blue tetrazolium chloride, 5 mM Tris-HCl (pH 7.4)), and gently shaken. Then, 100 μl of 10 mM β-nicotinamide adenine dinucleotide was added in staining solution. The gel was subsequently gently shaken until the purple signals became visible. To visualize complex IV activity, the gel was washed with 50 mM sodium phosphate buffer (pH 7.2) and submerged in 10 ml of staining solution B (0.075% diaminobenzidine, 50 mM sodium phosphate buffer, pH 7.2), and then 50 μl of 5 mM cytochrome c (in 100 mM Tris-HCl, pH 7.5) was added, and the gel was stained overnight by gently shaking in the dark. For loading control for mitochondrial protein, the gel was stained with Coomassie Blue. Cell viability assay Cell viability was assessed using a Cell Counting Kit-8 (347-07621, Dojindo) according to the manufacturer’s instructions. Statistics and reproducibility Statistical analysis was performed using Prism software (GraphPad Software) as indicated in each figure legend. All results are expressed as mean ± s.d. Statistical significance was defined as ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; NS, not significant. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications8. Data distribution was assumed to be normal, but this was not formally tested. No data were excluded from the analyses. The experiments were not randomized. The investigators were blinded to allocation during experiments and outcome assessment. All experiments were independently repeated at least three times with similar results unless otherwise stated. In experiments, representative images (including immunoblotting and immunofluorescence) are shown from experiments that were independently repeated at least three times with similar results, unless otherwise stated. Data in Figs. 1c,e,h–k and 2a–f come from experiments repeated three times independently with similar outcomes. Reporting Summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Online content Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41556-026-01974-0. Supplementary information Reporting Summary Peer Review File Supplementary Tables 1 and 2sgRNAs used in this study and light microscopy reporting table. Reporting Summary Peer Review File sgRNAs used in this study and light microscopy reporting table. Source data Source Data Figs. 3–7 and Extended data Figs. 8–10Statistical source data. Source Data Fig. 1Unprocessed western blots and/or gels. Source Data Fig. 2Unprocessed western blots and/or gels. Source Data Fig. 3Unprocessed western blots and/or gels. Source Data Fig. 4Unprocessed western blots and/or gels. Source Data Fig. 5Unprocessed western blots and/or gels. Source Data Fig. 6Unprocessed western blots and/or gels. Source Data Fig. 7Unprocessed western blots and/or gels. Source Data Extended Data Fig. 2Unprocessed western blots and/or gels. Source Data Extended Data Fig. 3Unprocessed western blots and/or gels. Source Data Extended Data Fig. 4Unprocessed western blots and/or gels. Source Data Extended Data Fig. 6Unprocessed western blots and/or gels. Source Data Extended Data Fig. 7Unprocessed western blots and/or gels. Source Data Extended Data Fig. 9Unprocessed western blots and/or gels. Source Data Extended Data Fig. 10Unprocessed western blots and/or gels. Statistical source data. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels. Unprocessed western blots and/or gels.