Abstract

   Mitochondrial oxidative phosphorylation (OXPHOS) fuels cellular ATP
   demands. OXPHOS defects lead to severe human disorders with unexplained
   tissue specific pathologies. Mitochondrial gene expression is essential
   for OXPHOS biogenesis since core subunits of the complexes are
   mitochondrial-encoded. COX14 is required for translation of COX1, the
   central mitochondrial-encoded subunit of complex IV. Here we describe a
   COX14 mutant mouse corresponding to a patient with complex IV
   deficiency. COX14^M19I mice display broad tissue-specific pathologies.
   A hallmark phenotype is severe liver inflammation linked to release of
   mitochondrial RNA into the cytosol sensed by RIG-1 pathway. We find
   that mitochondrial RNA release is triggered by increased reactive
   oxygen species production in the deficiency of complex IV.
   Additionally, we describe a COA3^Y72C mouse, affected in an assembly
   factor that cooperates with COX14 in early COX1 biogenesis, which
   displays a similar yet milder inflammatory phenotype. Our study
   provides insight into a link between defective mitochondrial gene
   expression and tissue-specific inflammation.

   Subject terms: Mechanisms of disease, Metabolic disorders, Biochemistry
     __________________________________________________________________

   Dysfunctions of the mitochondrial OXPHOS system lead to severe human
   disorders. Here, the authors analyzed a mouse mutant that mimics a
   mitochondrial disorder patient and find that reactive oxygen species
   trigger RNA release and inflammatory pathway in liver.

Introduction

   Mitochondria enable eukaryotic cells to efficiently utilize the
   chemical energy released from oxidation of nutrients. The mitochondrial
   oxidative phosphorylation system produces the bulk of cellular ATP. The
   four respiratory chain complexes are membrane bound enzymes that
   catalyze redox reactions, directing electrons from NADH and FADH[2] to
   molecular oxygen to produce water. This process establishes a proton
   gradient that drives synthesis of ATP by the F[1]F[o] ATP-synthase.
   Complexes I, III, IV, and V consist of subunits of dual genetic origin,
   encoded by both mitochondrial and nuclear genomes. Exceptional spatial
   and temporal coordination between different pathways of protein
   expression, transport, and assembly are required to enable proper
   biogenesis of these complexes^[82]1–[83]3.

   In the case of cytochrome c oxidase (complex IV), the terminal enzyme
   of the respiratory chain, the mitochondrial-encoded COX1 subunit
   represents the headstone of the biogenesis process. COX1 is
   cotranslationally inserted into the inner mitochondrial membrane with
   the aid of the OXA1L insertase^[84]4–[85]6. Already during synthesis,
   COX1 engages with proteins of an assembly intermediate termed MITRAC,
   which enables engagement of COX1 with nuclear-encoded subunits^[86]7.
   The early MITRAC subunits C12ORF62 (COX14) and MITRAC12 (COA3)
   associate with the ribosomes-COX1 nascent chain complex to assist in
   COX1 membrane insertion and stabilization of the nascent polypeptide.
   During the assembly process, COX1 passes through a series of assembly
   stages during which additional mitochondrial-encoded and
   nuclear-encoded proteins join to form the complex IV. These
   intermediates are stabilized by stage-specific chaperone-like assembly
   factors^[87]8. Interestingly, COX1 translation depends on the influx of
   nuclear-encoded complex subunits to adapt mitochondrial gene expression
   to cellular demands^[88]9. The cytochrome c oxidase is equipped with
   multiple nuclear-encoded tissue-specific isoforms for numerous
   subunits^[89]10. Since their function is still poorly understood, they
   are of broad interest in the field. It has been hypothesized that
   different COX isoforms regulate the activity and structure of the
   enzyme to meet different energy demands of distinct tissues^[90]11.
   Dysfunction of either structural or non-structural proteins associated
   with complex IV biogenesis have been linked to severe human disorders
   with a surprising variability of effects on different tissues^[91]12.
   This makes the comprehension of the pathophysiology challenging.

   In human mitochondria, C12ORF62 (COX14) and MITRAC12 (COA3) have been
   shown to stimulate the translation of COX1 in a ribosome nascent chain
   complex. In the absence of either protein, translation of COX1 is
   blocked leading to a loss of complex IV. Patients affected in both
   genes have been identified. The patient carrying a M^19I substitution
   in C12ORF62 exhibited severe metabolic acidosis and death in the
   neonatal period^[92]13. All major organ systems, prominently brain
   hypertrophy, hepatomegaly, hypertrophic cardiomyopathy, renal
   hypoplasia, and adrenal-gland hyperplasia were observed. Identical
   disease presentation was later observed in the two subsequent siblings
   in the same family. Biochemical analysis of patient fibroblasts showed
   reduced complex IV activity and COX1 translation. On the other hand,
   the patient carrying a Y^72C exchange in MITRAC12 showed a mild
   clinical course despite strong reductions in complex IV amounts^[93]14.
   The most severe symptoms were exercise intolerance and peripheral
   neuropathy. It is not clear how defects in the same assembly pathway
   leading to similar losses of complex IV display different pathological
   manifestations.

   Here we established two mice models, COX14^M19I and COA3^Y72C,
   recapitulating the amino acid exchange observed in the patients. The
   COX14^M19I mouse displayed multi-systemic pathology that worsened with
   age. Different magnitudes of gene expression defects across tissues
   could be attributed to differential stability of the mutant COX14.
   Despite the common defect in COX1 translation, the resulting complex IV
   deficiency and tissue pathology varied greatly. Interestingly, in
   liver, which displayed the highest reduction in complex IV, a strong
   inflammation was apparent. Furthermore, we observed an increased
   contact between mitochondria and lipid bodies in the mutant liver. Our
   data revealed that mitochondrial RNA release into the cytoplasm of the
   hepatocytes activates a mitochondrial RNA triggered Type I IFN response
   and local inflammation. Mitochondrial RNA release is triggered by
   increased mitochondrial ROS production in mutant mitochondria due to
   loss of complex IV. Accordingly, we reveal a pathology in which
   increased ROS production correlating with the loss of complex IV
   triggers mitochondrial RNA release and concomitant induction of
   inflammation pathways that contributes to hepatic failure.

Results

COX14^M19I leads to protein instability affecting complex IV assembly

   COX14 (C12ORF62 in human) mediates early steps of cytochrome c oxidase
   (complex IV) assembly by coordinating translation of mitochondrial COX1
   with early biogenesis steps. In order to analyze the molecular
   pathology of loss of COX14 we generated a mouse model by
   CRISPR/Cas-mediated patient mutation knock-in. We introduced a missense
   point mutation G88A into the Cox14 allele. This resulted in a single
   amino exchange at position 19 from methionine to isoleucine
   (Fig. [94]1a, Supplementary Fig. [95]1a). We were able to generate
   homozygous mice, COX14^M19I, which were used in the entire study. The
   altered nucleotide sequence in the Cox14 mRNAs did not compromise its
   stability across the different tissues tested (Fig. [96]1b). However,
   the mutant protein was detected to varying degrees in different organs
   of the mice (Fig. [97]1c, Supplementary Fig. [98]1b). In agreement,
   translation of COX1, assessed by in organello [^35S] methionine
   labelling of mitochondrial translation products, was significantly
   affected in mutant mitochondria (Fig. [99]1d). Subsequent pulse chase
   experiments showed a slight effect on the stability of the newly
   synthesized COX1 compared to the wild type situation (Fig. [100]1e).
   Thus, COX14^M19I apparently affected both translation and stability of
   COX1.

Fig. 1. COX14^M19I mice display reduced complex IV amounts.

   [101]Fig. 1
   [102]Open in a new tab

   a Cartoon depicting the generation of COX14^M19I mouse using
   CRISPR/CAS9 mediated double stranded cut in Cox14 allele and subsequent
   single stranded Oligonucleotide DNA (ssODN) mediated repair. b
   Expression of Cox14 mRNA in indicated tissues of 35-week-old COX14^M19I
   mice compared to wild-type (WT). Means ± SEM, n = 3, unpaired t test,
   ns = non-significant. c Western blot analysis and quantification of
   relative amounts of COX14 protein in indicated tissues from 22-week-old
   COX14^M19I mice (Supplementary Fig. [103]1b). Presented relative COX14
   amount as percent of WT after normalization to RIESKE. Means ± SEM,
   n = 4, unpaired t test, Brain p = 0.0175; Liver p < 0.0001; Muscle
   p = 0.0242; Kidney p = 0.0393. d [^35S] methionine labeling of
   mitochondrial translation products in isolated brain mitochondria from
   WT and COX14^M19I mice. Samples were analyzed by SDS-PAGE and digital
   autoradiography The figure is representative of n = 3. e Mitochondrial
   translation products pulse labeled with [^35S] methionine for 1 h in WT
   and COX14^M19I mice brain mitochondria and chased with non-radioactive
   methionine for indicated times. Samples were analyzed by SDS-PAGE and
   digital autoradiography. Quantification of COX1 presented. Means ± SEM,
   n = 3. f Volcano plot of mass spectrometric analysis of mitochondrial
   proteomes from WT and COX14^M19I mice liver mitochondria, n = 4. g
   Measurement of cytochrome c oxidase activity from indicated tissues of
   12-week-old WT and COX14^M19I mice plotted as percentage of the average
   of WT. Means ± SEM, n = 5, One-way ANOVA, Brain p = 0.0312; Liver
   p < 0.0001; Heart p < 0.0001; Muscle p = 0.0405. h Real-time
   respirometry analysis of 35-week-old WT and COX14^M19I mice liver
   mitochondria; oxygen consumption rates, OCR. Means ± SEM, n = 3. Source
   data are provided as a Source Data file.

   In steady state protein level analyses, COX14^M19I mitochondria
   displayed a decrease in both nuclear- and mitochondrial-encoded
   cytochrome c oxidase subunits (Fig. [104]1f). Moreover, complex IV
   activity and amount were significantly reduced in mutant mitochondria
   isolated from different tissues (Fig. [105]1g, Supplementary
   Fig. [106]1c). Among the tested tissues, liver appeared to be the most
   affected tissue. Furthermore, COX14^M19I liver mitochondria displayed a
   significant reduction in respiration (Fig. [107]1h). Hence, COX14^M19I
   mice allowed a detailed dissection of complex IV biogenesis defects in
   tissues of mice. In addition, we identified tissue-specific effects of
   the COX14^M19I mutation on COX1 translation and assembly.

Phenotyping COX14^M19I pathology

   To assess pathologies of COX14^M19I, the mice were subjected to a
   standard workflow following the international mouse phenotyping
   consortium (IMPC) pipeline. The animals exhibited a slight decrease in
   body weight after birth. Upon aging, the effect was only retained in
   COX14^M19I males which displayed a significant reduction when measured
   at 30 weeks of age (Fig. [108]2a). Moreover, 26% of the COX14^M19I
   animals experienced pre-weaning mortality as compared to that in the
   wild type. Once the animals were weaned, the life expectancies were
   quite similar in males whereas the females lived a shorter lifespan as
   compared to that of the wild type (Fig. [109]2b, Supplementary
   Fig. [110]2a). Interestingly, various blood parameters were altered in
   the COX14^M19I mice (Fig. [111]2c and Supplementary Data [112]1).
   Notably, we found elevated levels of alanine aminotransferase (ALT) and
   aspartate aminotransferase (AST), indicative of liver damage in mutant
   mice. Additionally, serum cholesterol and creatinine levels were high,
   suggesting dysfunctions of liver and kidney.

Fig. 2. Pathophysiology of COX14^M19I mice.

   [113]Fig. 2
   [114]Open in a new tab

   a Body weights of wild-type (WT) and COX14^M19I mice at 30 weeks of
   age. Means ± SD, n = 12, One-way ANOVA, ns non-significant, WT females
   vs. WT males p < 0.0001; COX14 ^M19I females vs. COX14 ^M19I males
   p < 0.0001; WT males vs. COX14 ^M19I males p = 0.0305. b Survival curve
   of WT and COX14^M19I female mice. c Serum biochemical parameters for
   16-week-old COX14^M19I mice compared to WT mice, n = 8, Unpaired t
   test. d Heatmap depicting steady state immune cell populations from WT
   and COX14^M19I mice spleen showing differences in ones with
   p-value < 0.05; One-way ANOVA; n = 6. e Representative H&E-stained
   sections from heart, skeletal muscle, liver and eye of COX14^M19I mice.
   Scale bar 200 μm. The figure is representative of n = 8. f
   Representative fundus images from WT and COX14^M19I eyes. Scale bar
   400 μm. The figure is representative of n = 8. g Retina thickness WT
   and COX14^M19I eyes at indicated age determined by Optical Coherence
   Tomography. Means ± SD, n = 24, One-way ANOVA, 5 weeks p = 0.0005; 15
   weeks p < 0.0001; 30 weeks p < 0.0001. h Echocardiography (left panel,
   left ventricle anterior wall thickness; right panel, ejection fraction)
   of WT and COX14^M19I mice. Means ± SD, n = 12, One-way ANOVA,
   ns=non-significant, left panel: WT females vs. COX14 ^M19I females
   p < 0.0001; WT males vs. COX14 ^M19I males p = 0.0018; right panel: WT
   females vs. COX14 ^M19I females p < 0.0001; COX14 ^M19I females vs.
   COX14 ^M19I males p = 0.0134. Source data are provided as a Source Data
   file.

   In recent years, mitochondrial dysfunction has been shown to affect
   development and function of immune cells^[115]15. We identified
   differences in several immune cell populations in the spleen of
   COX14^M19I animals (Fig. [116]2d). Resting immune cell populations such
   as the resting CD8 + T cells, resting CD4- NKT cells and resting NK
   cells were found to be increased. In contrast, effector NK cells and
   effector CD4- NKT cells were decreased. This suggested either a
   metabolic deficit to fuel the activation or exit of certain populations
   from spleen. However, no prominent structural changes were observed in
   spleen sections.

   Histological analysis of heart, skeletal muscle, liver, and eye
   indicated that these tissues were affected by the COX14 mutation
   (Fig. [117]2e). COX14^M19I heart sections identified groups of
   vacuolated and fragmented cardiomyocytes. In some sections,
   hypertrophic cardiomyocytes with enlarged nuclei were observed in the
   left ventricle. Liver sections displayed spots of focal necrosis
   demarcated by mononuclear cell infiltrate. The skeletal muscle sections
   showed few areas of fibers with more basophilic central nuclei
   indicating localized myopathy and regeneration. While COX14^M19I eye
   sections resembled the wild type at 16 weeks; at 30 weeks of age,
   thinning of the outer nuclear layer (ONL) was apparent in all sections,
   indicative of a lower number of photoreceptors cells. In 25% of the
   animals, the outer plexiform layer (OPL), the interface of bipolar
   cells and photoreceptors, was also found to be reduced in thickness. In
   addition, optical coherence tomography revealed morphological changes
   in the vasculature of the eye (Fig. [118]2f, Supplementary
   Fig. [119]2b). The retinal vessels were tortuous, branched, and bent in
   the superficial vascular plexus. Furthermore, a decrease in retina
   thickness was observed for the COX14^M19I animals with age, starting at
   5 weeks till the last measurements at 30 weeks (Fig. [120]2g). Analysis
   of the phototransduction by electroretinography showed a defect
   specifically in rod cells of the retina (Supplementary Fig. [121]2c).
   The COX14^M19I retinas also exhibited increased positive TUNEL signal
   in the outer nuclear layer (ONL) and the inner nuclear layer (INL) in
   animals already at 15 weeks of age (Supplementary Fig. [122]2d)
   indicating a gradual degeneration of the retina with age.

   Since the heart sections showed indications of hypertrophy, the animals
   were subjected to echocardiography. Similar to the results obtained for
   the eye, no statistically significant changes were observed at 16 weeks
   of age. However, upon aging these animals to 30 weeks, an increase in
   the ventricular wall thickness and fractional area shortening were
   observed (Fig. [123]2h, Supplementary Fig. [124]2e). A concomitant
   decrease in the ventricular diameter was detected. Yet, the ejection
   was slightly increased in case of female mice but not significantly
   altered in male mice. All of these indicated an altered heart function
   with age. In conclusion, similar to the patients with Leighs
   syndrome-like disorders, arising from a similar mutation, COX14^M19I
   newborn mice exhibited neonatal death (about 26%) and multisystemic
   pathologies that worsened with age.

Transcriptome analysis reveals dysregulation of signaling pathways

   Data on the cytochrome c oxidase, serum parameters, and histopathology
   showed that the liver was one of the most severely affected tissues in
   COX14^M19I mice. To gain further insights into the key pathways
   affected, we addressed changes in the gene expression pattern by RNA
   sequencing (RNA-seq) of wild type and COX14^M19I mice liver samples. An
   over-representation (pathway enrichment) analysis using terms from the
   KEGG pathway database indicated that pathways of respiratory electron
   transport chain, ATP synthesis, thermogenesis, plasma lipoprotein
   remodeling and chylomicron remodeling were downregulated in the mutant
   samples compared to the wild type (Fig. [125]3a). Conversely, pathways
   related to cholesterol biosynthesis, metabolism of steroids, antiviral
   IFN stimulated genes, class I MHC peptide loading and antigen
   presentation were enriched in COX14^M19I livers (Fig. [126]3b).
   Differential gene expression analysis revealed an increase in several
   SERBP-dependent cholesterol biosynthesis genes including the
   3-Hydroxy-3-methylglutaryl-CoA reductase (Hmgcr), Squalene epoxidase
   (Sqle), and Insulin induced gene 1 (Insig1) (Fig. [127]3c, d).
   Additionally, genes involved in the antiviral interferon alpha/beta
   pathway such as Interferon regulatory factor 7 (Irf7), Interferon alpha
   inducible protein 27 (Ifi27), Interferon induced protein with
   tetratricopeptide repeats 3 (Ifit3) and many more were found to be
   significantly upregulated in the context of COX14^M19I. Metabolic genes
   such glucose-6-phosphatase (G6pc); mitochondrial glutamate carrier,
   solute carrier family 25 member 22 (Slc25a22); and zinc transporter,
   solute carrier family 39 member 5 (Slc39a5) were found to be
   downregulated in COX14^M19I liver samples. To gain insights into a
   causal map of molecular interactions, gene regulatory networks were
   built around the two highly upregulated transcripts Irf7 (Fig. [128]3e)
   and Hmgcr (Supplementary Fig. [129]3). In case of Irf7, mostly
   second-degree contacts are perturbed which appear to be linked through
   the node of TRAF6, the mediator of RIG-I- and MDA5-mediated antiviral
   responses. However, the network of Hmgcr pointed to a dense network of
   first-degree contacts to be either upregulated or downregulated.
   Interestingly, many mitochondrial ribosomal proteins in this network
   map were downregulated. In summary, the investigation of alterations in
   gene expression revealed activation of inflammatory response pathways
   and indicated alterations in liver tissue metabolism.

Fig. 3. Altered gene expression in COX14^M19I mice.

   [130]Fig. 3
   [131]Open in a new tab

   a Over-representation analysis of genes significantly downregulated in
   COX14^M19I samples and b over-representation analysis of genes
   significantly upregulated in COX14^M19I samples. For both (c) and (d):
   Functional pathways were derived from the Reactome database, and
   statistically significant over-represented (enriched) pathways were
   determined with a cutoff of FDR < 0.05 (Benjamini–Hochberg procedure).
   Pathways are arranged in increasing order of FDR values from top to
   bottom, and the colour indicates the enrichment ratio. c Volcano plot
   presentation of differentially expressed genes between liver samples
   from WT and COX14^M19I mice. X-axis denotes fold change in expression
   (log2 scale), y-axis denotes adjusted p-value (negative log10 scale)
   for the analysed genes in the data set (each dot represents a single
   gene). Blue and red dots represent genes significantly downregulated
   and upregulated respectively in the COX14 mutant samples compared to WT
   (DESeq2; cut-offs used: adjusted p-value < 0.05 and absolute value of
   fold change <1.15). Non-significant genes are depicted in black.
   Selected genes involved in pathways relevant for the study are labelled
   in the plot. d Heatmap representing normalized expression (centered and
   scaled) for genes labelled in A in WT and COX14^M19I liver samples. e
   Irf7 interaction network, with first degree (direct) and second degree
   (neighbours of direct) connections of Irf7. (Green node: hub gene
   (Irf7); red nodes: genes upregulated genes in COX14^M19I samples; blue
   nodes: genes upregulated genes in COX14^M19I samples; grey nodes:
   interactors in the network which are not significantly up- or
   downregulated.) The different kinds of (directed) interactions
   considered were: protein-protein (PP), RNA-RNA (RR), transcription
   factor-protein (TP), transcription factor-RNA (TR) and
   RNA-transcription factor (RT) interactions. The size of the nodes
   denotes the degree (number of outgoing + incoming interactions).

Aberrant lipid droplet association with mitochondria in COX14^M19I livers

   The transcriptome analysis of liver indicated upregulated cholesterol
   biosynthesis. This observation led us to investigate cellular lipid
   profiles by mass spectrometry-based lipidomics. Yet, these analyses
   identified no significant alterations in any of the identified lipid
   classes in COX14^M19I liver (Fig. [132]4a and Supplementary
   Data [133]2). Furthermore, in order to substantiate these findings,
   primary COX14^M19I hepatocytes were cultured and subjected to BODIPY
   staining to detect differences in intracellular lipid organization. The
   number of lipid droplets in COX14^M19I hepatocytes was more in number
   than in the wild type (Fig. [134]4b, Supplementary Fig. [135]4a).
   Statistical analysis of lipid droplet size and number revealed a
   difference in lipid droplet number but not in the total lipid volume
   per cell (Fig. [136]4c). In addition, primary hepatocytes were
   subjected to transmission electron microscopy (TEM) analysis to assess
   lipid droplet distribution (Supplementary Fig. [137]4b). Interestingly,
   drastic alterations in mitochondrial morphology were observed.
   Mitochondrial cristae had lost their stacked organization in the
   COX14^M19I hepatocytes and the mitochondrial outer membrane appeared to
   be noncontiguous (Supplementary Fig. [138]4b). To address this in
   tissue, liver sections were subjected to TEM analysis (Supplementary
   Fig. [139]4c). Although the mitochondria exhibited wild type-like
   membrane organization, in COX14^M19I liver section we observed
   mitochondria that wrapped around a lipid droplet. Subsequent focused
   ion beam scanning electron microscopy (FIB-SEM) of liver samples
   demonstrated that, this phenomenon was frequently observed in every
   liver cell imaged from COX14^M19I. The mitochondria were in close
   contact with lipid droplets to varying degrees (Fig. [140]4d,
   Supplementary Fig. [141]4d). The FIB-SEM data set was subjected to
   visual inspection of a large portion of the cell volume of hepatocytes
   and reconstructed in 3D. We observed that the mitochondria wrapping the
   lipid droplets were not spatially isolated entities compartmentalized
   to a part of the cell but rather were present across the entire cell
   (Fig. [142]4e). The extent of mitochondria-lipid droplet contacts
   differed. In extreme cases the mitochondria engulfed the lipid droplet,
   while in some cases they just had a single point of contact
   (Supplementary Fig. [143]4e).

Fig. 4. Altered mitochondria lipid droplet contacts in COX14^M19I mice.

   [144]Fig. 4
   [145]Open in a new tab

   a Total amount of different classes of lipid species in wild-type (WT)
   and COX14^M19I mice liver samples analyzed by mass spectrometry.
   Means ± SEM, n = 8. b Cytochemistry of isolated primary hepatocytes
   from WT and COX14^M19I mice using Oil Red O. n = 3, scale bar 100 μm. c
   Statistical analysis of images from b, number of lipid droplets per
   cell (left) and average lipid droplet area per cell (right).
   Means ± SEM, n = 3 biological replicates, > 6 technical replicates pre
   mouse, Unpaired t test, left panel p = 0.0374; right panel p = 0.0691.
   d Representative TEM and FIB-SEM images of liver tissue samples from WT
   and COX14^M19I mice. n = 3, scale bar 1 μm. e Representative 3D
   reconstructions of interaction between mitochondrion and lipid droplets
   observed in FIB-SEM images of liver tissue samples from COX14^M19I
   mice. n = 3, scale bar 2 μm. Source data are provided as a Source Data
   file.

   To discern the functional implication of this contact, free
   mitochondria termed cytoplasmic mitochondria and lipid droplet
   associated peridroplet mitochondria were isolated from mice livers
   (Supplementary Fig. [146]5a, b). We subjected both populations to
   specific substrate-based real time respirometry analysis. However, we
   did not observe any preferential substrate utilization (Supplementary
   Fig. [147]5c). We concluded that despite of the apparent lipid droplet
   reorganization that was observed in the COX14^M19I mice the liver did
   not display a significant alteration in lipid metabolism.

mtRNA release to the cytosol from mitochondria

   The transcriptome analysis revealed that the antiviral interferon
   alpha/beta pathway was highly upregulated in COX14^M19I liver.
   Typically, these type I IFN responses or Interferon Stimulated Genes
   (ISGs) are expressed upon a wide variety of stimulants such as viral
   infections, cellular stress, ROS, or non-self-nucleic acid
   sensing^[148]16–[149]18. COX14^M19I liver showed a significant
   expression of ISGs both at RNA and protein levels (Fig. [150]5a, b). We
   observed that the mRNA expression of ISGs were also upregulated in
   other tissues such as the heart, skeletal muscle and brain but to a
   significantly lesser extent than in liver. In fact, the only other
   tissue to show a slight upregulation of ISGs at the protein level was
   kidney (Supplementary Fig. [151]6a). Thus, liver was the only organ
   with a high type I IFN response.

Fig. 5. Activation of inflammatory pathways in COX14^M19I and COA3^Y72C mice.

   [152]Fig. 5
   [153]Open in a new tab

   a qPCR analysis of the gene expression of nucleic acid sensing pathway
   target genes in total mRNA isolated from 12-week-old WT and COX14^M19I
   mice liver samples. Means ± SEM, n = 4. b Western blot analysis and
   quantification of tissue lysates from 12-week-old WT and COX14^M19I
   mice livers with indicated antibodies. Means ± SEM, n = 4, Unpaired t
   test, ISG15 p = 0.0005; OASL1a p < 0.0001; IFIT1 p = 0.0003. c
   Graphical presentation on the generation of the COA3^Y72C mouse line
   using CRISPR/CAS9-mediated double stranded cut in the Coa3 allele and
   subsequent single stranded Oligonucleotide DNA (ssODN) mediated repair.
   d Enzyme activity of cytochrome c oxidase from 60-week-old WT and
   COA3^Y72C brain, heart, liver and muscle plotted as percentage of
   average of WT. Means ± SEM, n = 3, One-way ANOVA, ns=non-significant,
   Brain p = 0.0028; Liver p < 0.0001; Muscle p = 0.0004 e qPCR analysis
   of gene expression of nucleic acid sensing pathway target genes in
   total mRNA isolated from 22-week-old WT and COA3^Y72C mice liver
   samples. Means ± SEM, n = 4. f Measurement of relative mRNA abundance
   in the cytosolic fractions of 12-week-old WT and COX14^M19I mice liver
   samples. Means ± SEM, n = 4. g Immunoblot analysis and quantification
   of tissue lysates from 24-week-old WT and COX14^M19I mice liver for
   indicated proteins. Means ± SEM, n = 4, Unpaired t test, RIG1
   p = 0.0002; MDA5 p = 0.0051; STING p = 0.0416; ZBP1 p = 0.0015. Source
   data are provided as a Source Data file.

   To address if the observed inflammatory phenotype was specific to the
   COX14^M19I mutation rather than a loss of complex IV activity, a second
   mouse model was generated using a similar strategy to target the
   functionally closely related gene COA3. Both COA3 (MITRAC12 in human)
   and COX14 act early in the COX1 assembly process and their loss of
   function results in decreased COX1 translation^[154]7. A missense point
   mutation A583G was introduced into the Coa3 allele. This resulted in
   the exchange of a single amino acid at position 72 from tyrosine to
   cysteine, similar to the amino acid exchange reported for a
   mitochondrial disease patient^[155]14 (Fig. [156]5c, Supplementary
   Fig. [157]6b). COA3^Y72C homozygous males and females did not display
   differences in life expectancies compared to the wild type
   (Supplementary Fig. [158]6d). COA3^Y72C sera displayed significant
   differences in HDL and total cholesterol levels (Supplementary
   Fig. [159]6c). Since ALT and AST levels in the serum were similar to
   the wild type and no morphological differences were observed by
   histochemistry, we concluded that the livers from COA3^Y72C mice did
   not display any apparent pathology. However, liver mitochondria showed
   a clear reduction in complex IV activity (Fig. [160]5d). Yet, the
   reduction in complex IV was not to the same extent as observed in
   COX14^M19I mice. Moreover, the expression of ISGs was upregulated in
   the COA3^Y72C mice but again to a lesser degree than in COX14^M19I
   (Fig. [161]5e). We concluded that the observed inflammatory pathology
   was specific to COX14^M19I mice and correlated with the defect on
   complex IV.

   Apart from viral infections, type I IFN inflammation is also induced by
   cytosolically sensed mitochondrial nucleic acids (Supplementary
   Fig. [162]7a)^[163]18,[164]19. Therefore, we analyzed if mitochondrial
   nucleic acids were present in the cytoplasm of COX14^M19I liver. Liver
   tissue homogenates were cleared of mitochondrial and nuclear fractions
   and subjected to RNA and DNA isolation. We detected both mitochondrial
   RNA (mtRNA) and mitochondrial DNA (mtDNA) by qPCR in significant
   amounts in cytoplasmic fractions of liver cells (Fig. [165]5f and
   Supplementary Fig. [166]6e). The presence of mtDNA in the cytoplasm in
   other organs such as heart, brain, and skeletal muscle was not
   increased above wild type levels. Accordingly, among the organs tested,
   liver was the only one to show a significant release of mitochondrial
   nucleic acids into the cytoplasm.

   To exclude that the observed detection of mitochondrial DNA was caused
   by nuclear-embedded mitochondrial DNA sequences (NUMTs) that were
   amplified by qPCR while quantifying mtDNA in the cytoplasm, an
   additional method was applied. Long-range multi-gene spanning probes
   covering three different regions of mouse mtDNA were designed to
   uniquely amplify the mtDNA (see Methods). These analyses further
   confirmed the presence of mtDNA in the cytosol of COX14^M19I mice liver
   cells.

   Based on the presence of mtDNA and mtRNA in the cytoplasm, we assessed
   which of the cytoplasmic sensors were involved in the detection of
   these nucleic acids. We found that RNA sensors such RIG-1 and MDA5 were
   significantly more abundant in liver cells than DNA sensors such as
   STING (Fig. [167]5g). Additionally, we also observe increase of the
   ZBP1 protein in the COX14^M19I mice samples. Thus, we concluded that in
   COX14^M19I mutant mice mitochondria released nucleic acids into the
   cytosol and thereby trigger a mtRNA mediated type I IFN inflammation in
   liver.

ROS imbalance leads to mitochondrial nucleic acid release

   Cellular pyrimidine deficiencies have been shown to trigger
   mtDNA-dependent immune responses in mouse retina and cells lacking the
   mitochondrial protease YME1L^[168]20. Hence a quantitative metabolomic
   screen with a focus on nucleotides, glycolysis, and TCA cycle
   intermediates was performed. Contrary to the case of YME1L deficiency,
   where a broad depletion of predominantly pyrimidine nucleotides was
   observed, COX14^M19I liver samples showed an enrichment of di- and tri-
   nucleotides (Fig. [169]6, Supplementary Fig. [170]7b). Increased levels
   of orotate and carbamoyl aspartate in COX14^M19I liver samples indicate
   enhanced pyrimidine synthesis. Decreased levels of metabolites of the
   oxidative and non-oxidative phase of the pentose phosphate pathway,
   which delivers ribose and reduction equivalents for purine
   biosynthesis, could indicate increased usage of these metabolites for
   nucleotide biosynthesis. Moreover, we found increased levels of citric
   acid cycle metabolites in COX14^M19I samples, which could reflect a
   boosted catabolic metabolism (Fig. [171]6, Supplementary Fig. [172]7b).
   Accordingly, nucleotide deficiency did not contribute to the release of
   mitochondrial nucleic acids in COX14^M19I mice liver.

Fig. 6. Metabolic analysis of COX14^M19I.

   [173]Fig. 6
   [174]Open in a new tab

   Levels of different classes of metabolites determined by mass
   spectrometry in wild-type (WT) and COX14^M19I mice liver samples.
   Means ± SD, n = 4 (WT); n = 6 (COX14^M19I), Unpaired t test, UTP
   p = 0.0039; ATP p = 0.0005; IMP p = 0.0003; Glucose-6-phosphate
   p = 0.0036; Citric acid p = 0.0001; Alpha-ketoglutaric acid p = 0.0107;
   Glutamate p = 0.0006; Carbamoyl-aspartatic acid p = 0.0005; Orotic acid
   p = 0.0014. Blue—similar amount in WT and COX14^M19I; red—decreased
   amount in COX14^M19I compared to WT; green—increased amount in
   COX14^M19I compared to WT. PPPE, early metabolites of the oxidative
   phase of pentose phosphate pathway not individually resolved by the
   analysis. Source data are provided as a Source Data file.

   To assess the extent of damage mitochondria exhibited in the COX14^M19I
   mice, primary hepatocytes were cultured and examined for different
   mitochondrial health parameters. The mitochondrial membrane potential
   in COX14^M19I samples was significantly lower than that of wild type
   controls (Fig. [175]7a). Interestingly, COX14^M19I hepatocytes elicited
   markedly increased mitochondrial ROS (Fig. [176]7b).

Fig. 7. mRNA release triggered by increased reactive oxygen species.

   [177]Fig. 7
   [178]Open in a new tab

   a Mitochondrial membrane potential measured in WT and COX14^M19I
   primary hepatocytes using TMRM staining. Means ± SEM, n = 5, Unpaired t
   test, p = 0.0003. b ROS production measured in WT and COX14^M19I
   primary hepatocytes using MitoSOX Red Mitochondrial Superoxide
   Indicator. Means ± SEM, n = 8, Unpaired t test, p < 0.0001. c qPCR
   analysis of gene expression of nucleic acid sensing pathway target
   genes in total mRNA isolated from WT and COX14^M19I primary hepatocytes
   treated with either vehicle control or NAC for 24 h. Means ± SEM,
   n = 3. d Measurement of relative mRNA abundance in the cytosolic
   fractions from WT and COX14^M19I hepatocyte fractionation samples.
   Means ± SEM, n = 4. e, f Western blot analysis of cell lysates from WT
   and COX14^M19I primary hepatocytes with indicated antibodies and
   quantification Means ± SEM, n = 3, Unpaired t test, ns =
   non-significant, e: IFIT1 p = 0.0047, OAS1a p = 0.0061, ISG15
   p = 0.0054; f IFIT1 p = 0.0113, OAS1a p < 0.0001. g ROS production
   measured in WT and COX14^M19I primary hepatocytes using MitoSOX Red
   Mitochondrial Superoxide Indicator. Means ± SEM, n = 6; p < 0.0001.
   Source data are provided as a Source Data file.

   Since COX14^M19I livers demonstrated an increase in SERBP-mediated
   cholesterol biosynthesis pathway as well as mitochondrial ROS
   production, we hypothesized that either of the two could trigger
   mitochondrial damage and subsequent nucleic acid release. To test this
   directly, we used respective inhibitors, Fatostatin and NAC, to assess
   if they inhibited type I IFN mediated inflammation. Additionally, known
   inhibitors of cytosolic nucleic acid sensing pathways were included in
   the analysis (Supplementary Fig. [179]8a). The treatment with
   Fatostatin did not alter the expression of ISGs, demonstrating that
   SERBP-mediated cholesterol biosynthesis did not contribute to the
   inflammatory response (Supplementary Fig. [180]8b). In contrast, NAC
   treatment caused the strongest reduction in ISGs expression amongst all
   tested reagents (Fig. [181]7c, Supplementary Fig. [182]8c–f).
   Furthermore, treatment with NAC prevented the release of mtRNAs in
   hepatocytes and subsequent upregulation of ISGs at the protein level
   (Fig. [183]7d, e). Upon NAC treatment, the reduction in IFIT1 protein
   levels matched the observed Ifit1 mRNA changes. For Isg15, we similarly
   observe a strong reduction. The protein levels of ISG15 further depend
   on the levels of USP18, which itself was reduced in the NAC treated
   sample. Unfortunately, due to unavailability of suitable specific
   primers, the mRNA levels of OAS1a could not be assessed. In addition,
   we could show that the NAC treatments reduced the mitochondrial
   superoxide levels (Fig. [184]7f). NAC treatment acts via cellular
   glutathione and thereby quenches ROS via the glutathione pool.
   Therefore, in a complementary approach we utilized Mito-TEMPO, which
   acts on superoxide as a superoxide dismutase mimetic. Similar results
   as seen for NAC were obtained with Mito-TEMPO (Fig. [185]7g and
   Supplementary Fig. [186]8g, h). Accordingly, increased mitochondrial
   ROS production was causal for the inflammatory phenotype observed in
   COX14^M19I livers.

   In conclusion, instability of the COX14^M19I protein and concomitant
   defects in complex IV assembly and function led to a perturbation of
   mitochondrial oxidative phosphorylation with increase in mitochondrial
   ROS production. The damage inflicted by ROS lead to mitochondrial
   damage and subsequent nucleic acid release. Sensing of the “foreign“
   mitochondrial RNAs in the cytoplasm instigated the inflammation and
   subsequent liver pathology.

Discussion

   Mice models are important assets to study the molecular pathology of
   OXPHOS-dysfunction in tissues. Here we assessed how defects in
   mitochondrial translation in a patient-resembling mouse model,
   COX14^M19I, affect cellular health and crosstalk of metabolic pathways
   in different tissues. Amongst the analyzed tissues, liver was the most
   affected. This pathology was already apparent at early age of the mice
   analyzed. It is interesting to note that despite the systemic defect in
   COX14 function, the impact on protein stability, COX1 synthesis and
   concomitantly amounts of complex IV differed significantly between
   tissues. The question of tissue specificity of mitochondrial disorders
   is a central and still unresolved topic of the field. It is likely that
   different metabolic demands, energy requirements thresholds for
   cellular damages contribute to tissue sensitivity. In case of
   COX14^M19I, the liver and heart defects correlated with reduced complex
   IV activity. Brain and muscle displayed similar complex IV reduction,
   yet a brain phenotype was not apparent in our analyses. Moreover, as we
   notice the most drastic reduction of the mutant COX14 protein in liver,
   one may argue that the reduced levels of COX14 in liver correlate with
   the complex IV defect. Yet, this correlation is not as apparent in the
   other tissues. Accordingly, the molecular pathology of mitochondria
   seems to affect different tissues to different extent. It is difficult
   to directly compare the mice phenotype to the patient pathology, as the
   death of the patient occurred already 24 hours after birth. Yet,
   cardiomyopathy and hepatomegaly were reported tissue dysfunctions.
   While cardiac defects were also apparent in the COX14^M19I mouse,
   information on inflammation and retinal state of the patient are not
   available.

   The liver plays key roles in many metabolic processes to which
   mitochondria contribute^[187]21. Accordingly, defects in liver
   mitochondria are linked to impaired carbohydrate and fatty acid
   metabolism, which affect energy production and increase fat
   accumulation leading eventually to fatty liver diseases^[188]22. A
   metabolomic screen of COX14^M19I mice liver samples showed a decrease
   of pentose phosphate pathway (PPP) metabolites and increased levels of
   TCA intermediates. Cellular redox homeostasis is linked with the PPP
   through generation of NADPH, which is also utilized for fatty acid
   synthesis. Furthermore, liver also derives NADPH from folate-mediated
   serine catabolism for lipogenesis^[189]23. Yet, both, serine and
   glycine, were found to be reduced in COX14^M19I livers. Additionally,
   transcriptional upregulation of SREBP-mediated cholesterol biosynthesis
   genes was apparent that did not translate to changes in cellular lipid
   species. In the light of these observation, the tight interactions
   between mitochondria and lipid droplets remains unexplained.

   Among transcripts that were upregulated in COX14^M19I liver, we
   identified such mRNAs that are linked to IRF7-induced inflammatory
   signaling pathways. Typically, these type I IFN responses or expression
   of interferon-stimulated genes (ISGs) can be very crudely categorized
   into either positive or negative regulators of inflammation. Among
   these, the positive regulators are subdivided into a wide array of
   genes depending on the mode of inhibiting viral
   replication^[190]24,[191]25. COX14^M19I livers displayed significant
   expression of Irf7, Ifit1, Ifit3, and Ifi27, which function towards
   binding viral RNA. Another set of positive regulators Gbp3 and Oas1a,
   which do not bind viral RNA rather restrict the GTP availability or
   promote RNAse activity respectively, were also found to be upregulated
   in mutant samples. However, the negative regulators Usp18 and Oasl1,
   which act on inhibiting the activity of ISG15 and IRF7, were also
   highly expressed. Typically, both the positive and negative responses
   are simultaneously active to limit the extent and duration of type I
   IFN responses^[192]26. Accordingly, COX14^M19I mice seem to exhibit a
   balance between both the responses thus limiting host toxicity and
   morbidity. However, persistent type I IFN signaling leads to other
   detrimental effects such as development of autoimmune diseases and
   decreased pathogen clearance^[193]27,[194]28. Interestingly, chronic
   exposure to IFNα also promotes T-cell differentiation, perturbs lipid
   metabolism and redox balance^[195]29,[196]30. It affects the metabolic
   fitness of CD8^+ T-cells as well^[197]31. Thus, the immune phenotype
   observed in COX14^M19I animals might be linked to the chronic type I
   IFN signaling from liver. Type I IFN signaling also affects the
   cholesterol flux in cells. It downregulates transcriptional activity of
   SREBPs^[198]32,[199]33. ISGs like RSAD2 have been also reported to
   inhibit cholesterol biosynthesis^[200]34. Surprisingly, COX14^M19I
   livers demonstrate the complete opposite phenotype. One possible
   explanation for this could be energetic insufficiency resulting from
   decreased complex IV activity.

   Our analyses show that the Type I IFN-mediated inflammation in
   COX14^M19I livers arose from RIG1-MDA5 based mtRNA sensing. mtRNA has
   been suggested to be released into the cytosol through herniation
   through BAK/BAX mediated pore formation^[201]35,[202]36. mtDNA has been
   proposed to be released through VDAC oligomers^[203]37, BAK/BAX
   macropores^[204]38, or mitochondrial-derived vesicles^[205]39. Yet, the
   triggers for these release processes appear to be highly variable
   ranging from physical injuries, cellular stress, metabolite imbalance
   to defects of mitochondrial health and dynamics^[206]19. In case of
   COX14^M19I liver mitochondria, increased mitochondrial ROS production
   triggered mtRNA release. Despite its importance as signaling molecule,
   ROS can affect mitochondrial structure by causing oxidative damage to
   the mitochondrial membrane with a concomitant decrease in integrity and
   function^[207]40. In isolated rat and mice mitochondria and MEF cells
   ROS was found to stimulate mtDNA release^[208]37,[209]41. Furthermore,
   ROS has been shown to trigger the formation of mitochondrial-derived
   vesicles^[210]42. Here we find that the instability of the COX14^M19I
   protein results in defective translation. The concomitant loss of
   complex IV leads to increased ROS production, mitochondrial damage, and
   mtRNA release especially in the liver (that displays the most severe
   complex IV phenotype), which induces type I IFN inflammation.
   Inflammatory patterns are also observed in other tissues, yet they
   appear to be not as pronounced as in case of the liver. Based on our
   observations on RNA release in liver tissue we concluded that a similar
   process is also in place in other tissues that display inflammatory
   responses. At the molecular level, our findings provide insights on how
   defective mitochondrial translation due to loss of COX14 function
   triggers ROS-induced type I IFN inflammation in liver leading to
   worsening pathology with time.

Methods

Animals

   The guidelines from the German Animal Welfare Act and approved by the
   Landesamt für Verbraucherschutz und Lebensmittelsicherheit,
   Niedersachsen, Germany (AZ: 33.9-42502-04-14/1720) were followed for
   the maintenance of animals and performance of animal experiments. The
   animals were housed either in high barrier (SPF-specified pathogen
   free) areas in IVC (individually ventilated caging) on standard rodent
   chow to WT C57BL/6 N mice, with restricted access for animal care staff
   only. The COX14^M19I and COA3^Y72C mice were created using genome
   engineering based on the CRISPR/Cas9 method. To obtain zygotes for
   microinjection, donor female mice of the C57BL/6 N strain were super
   ovulated. Microinjection was then performed using a mix containing the
   CAS9 enzyme, an sgRNA targeting the either the COX14 or COA3 locus, and
   a single-stranded oligonucleotide DNA repair template, using micro
   manipulators. The resulting embryos were surgically transferred into
   pseudo-pregnant female mice of the NMRI or C57BL/6 N strain. Founders
   with edited genomes were identified through Sanger sequencing of
   genomic DNA from ear biopsies. These mice were crossed with C57BL6/6 N
   for two generations to ensure germline transmission and eliminate any
   possible mosaicism. Heterozygous animals with the same modification
   were subsequently mated to produce homozygous offspring. Finally, PACE
   (PCR Allele Competitive Extension) fluorescent endpoint genotyping was
   used for subsequent genotypings.

Pronuclear injections for mice generation

   The COX14^M19I mice (C57BL/6N-Cox14^em83Cecad), in which methionine 19
   of the cDNA of Cox14 (Ensembl ENSMUST00000023761) is exchanged to
   isoleucine, was generated by pronuclear injection of CRISPR/Cas9
   reagents (PMID: 29723268)^[211]43. 400 nM Alt-R^TM guide RNA (5’-
   CACAGTGAGGAGCATCATCG −3’) targeting exon 2 of Cox14 and 500 nM standard
   desalted Ultramer™ single-stranded DNA repair oligonucleotides (5’-
   GCAGACCAACTCCAATCTGTCTT
   TGTAGGGGAAGACATGCCATCTGCCAAGCAGCTAGCCGATATCGGCTACAAGACCTTCTCTGCATCGATAA
   TGCTCCTCACTGTGTATGGGGGTTACCTCTGCA −3’) were injected with 200 nM
   Alt-R^TM SpCas9 protein and 30 ng/µl SpCas9 mRNA (TriLink, L-6125-20)
   in C57BL/6 N zygotes. All components were purchased from Integrated DNA
   Technologies (USA) and animals obtained from Janvier Labs (France) if
   not stated otherwise. Genome editing for this line was performed at the
   in vivo Research Facility of the CECAD Research Center (University of
   Cologne, Germany) in compliance with the European, national and
   institutional guidelines and approved by the State Office of North
   Rhine-Westphalia, Department of Nature, Environment and Consumer
   Protection (LANUV NRW, Germany; animal study protocol AZ
   84–02.04.2014.A372). Mice were anesthetized with ketamine (Ketaset,
   Zoetis) and xylazine (Rompun, Bayer) and euthanized by cervical
   dislocation. Carprofen (Rimadyl, Zoetis) was used as analgesic after
   surgery. A similar strategy was used for the COA3^Y72C mice, where the
   guide RNA (5’- CCTGAACCCACTCCGTCAGA −3’) targeted exon 2 of Coa3. The
   single-stranded DNA repair oligonucleotides
   (5’-AAGGGTGGCTAGACGAAAGGAATGCCCATGTGCTTTGTGGTTTCCTGAACTCAACTTCTGTCAGGTG
   GGTATACCTTCTACTCGGTGGCCCAAGAGCGTTTCCTTGATGAGCTGGAAGATGA −3’) were used
   for repair.

Mouse maintenance and handling

   The mice are fed ad libitum with Ssniff chow V1534-300. (ssniff® R/M-H
   autoclavable Complete feed for rats and mice - maintenance, fortified).
   The mice are maintained at a 12:12 dark-light cycle with lights on at
   4:30 am. The room temperature whetre the mice are kept is maintained at
   a constant temperature of 21 °C (+/−1 °C) and the humidity is about 50%
   (+/−5%). They are housed in a high level barrier (restricted entry;
   air-shower entry; gloves, mask, cap, change into sterilized scrubs and
   sterilized shoes; sterilization or disinfection of all supplies and
   equipment) in open cages under specific pathogen-free conditions. All
   mice of this husbandry unit are introduced into the unit via embryo
   transfer. The animal husbandry consists of five adjoining husbandry
   rooms. Hygienic monitoring by serological, microbiological and
   parasitological methods is performed according to the recommendations
   of the FELASA. Parasitology is carried out on every strain present in
   the room Staphylococci (including Staphylococcus aureus) are found
   sporadically. The mice are sacrificaed via cervical dislocation.

Isolation of primary hepatocytes and maintenance

   Hepatocytes were isolated by a collagenase perfusion technique with
   modifications. Briefly, the mice were sacrificed by cervical
   dislocation and quickly dissected ventrally to expose the peritoneal
   cavity. The inferior vena cava was cannulated with a 24-gauge 3/4-inch
   angiocatheter (BD) and the portal vein was cut. The liver was perfused
   via the IVC with 100 mL of KR-EGTA Medium (120 mM NaCl, 4.8 mM KCl,
   1.2 mM MgSO[4], 1.2 mM KH[2]PO[4], 24 mM NaHCO[3], 0.25 mM EGTA, pH
   7.4) at 37 °C, followed by perfusion with 70 mL of collagenase type IV
   (Worthington) in 1x PBS. After the liver was digested, it was dissected
   out and cut into small pieces and passed through a 100 mM strainer
   (BD). Hepatocytes were separated from nonparenchymal cells by low-speed
   centrifugation (100 g x 5 mins), and further purified by Percoll
   gradient centrifugation (50%, Sigma). The purified cells were washed
   once with Williams E Medium (supplemented with 10% FCS, 1% Gentamycin,
   10^−8M Insulin). Cells were then plated in either Collagen or
   poly-L-Lysine coated plates or coverslips. The media was changed after
   4 hours of plating to Maintenance Medium without FCS (William’s E
   medium containing 1% Glutamax, 1% Non-Essential Amino Acids, 1%
   penicillin/streptomycin, 0.2% normocin (Invivogen), 2% B27 (GIBCO), 1%
   N2 supplement (GIBCO), 10 mM nicotinamide (Sigma-Aldrich), 10 mM Y27632
   (Peprotech), 1 mM A83-01 (Tocris), 3 mM CHIR99021 (Peprotech), 25 ng/mL
   EGF (Peprotech), 50 ng/mL HGF (Peprotech) and 100 ng/mL TNFa
   (Peprotech)). The cells were treated with either 0.2 μM BX-795 (Sigma),
   0.8 μM RU.521 (Sigma), 20 μM Fatostatin (Sigma), 1 μM C-176 (Sigma), 50
   μM BC-1215 (Hycultec), 0.1 μM GSK690693 (Sigma), 0.1μM MRT67307 (Sigma)
   or 10mM N-acetylcysteine (NAC, Sigma) for 24 hours before harvesting
   them for end point readouts.

Mitochondrial isolation

Mitochondrial isolation from mouse tissues

   Isolation of mitochondria from mouse tissues (brain, heat liver and
   muscle) was performed: Freshly isolated or snap frozen tissue samples
   were used for the isolation of mitochondria. Heart and muscle samples
   were chopped into small pieces and preincubated in Digestion buffer (1X
   PBS + 10 mM EDTA + 0.05% Trypsin) for 30 min at room temperature.
   Mitochondrial isolation from brain, heart, liver and muscle was done
   via mechanical disruption at 16000 rpm using a Dounce homogenizer
   followed by differential centrifugation. Mitochondrial Isolation
   Buffer[HM] (67 mM Sucrose + 50 mM Tris/Cl pH7.4 + 50 mM KCl + 10 mM
   EDTA + 0.5% BSA; pH7.4) was used for the isolation of mitochondria from
   heart and muscle. Mitochondria Isolation Buffer[L] (10 mM Tris–MOPS,
   1 mM EGTA/Tris, 200 mM Sucrose; pH to 7.4) was used for the isolation
   of mitochondria from liver. Mitochondria Isolation Buffer[B] (75 mM
   Sucrose, 215 mM D-Mannitol, 1 mM EDTA pH 7.4, 20 mM HEPES; pH to 7.4
   (KOH)) was used for mitochondrial isolation from brain. The debris was
   pelleted at 700 × g and the supernatant containing the mitochondria was
   separated. Pelleting of mitochondria was done at 8000 × g. Finally, the
   mitochondria were washed Mitochondrial Isolation Wash Buffer (250 mM
   Sucrose + 3 mM EGTA/Tris + 10 mM Tris/Cl pH7.4; pH7.4)^[212]44.

RNA isolation

   Cells and tissues were subjected to RNA isolation using the TRIzolTM
   reagent, following the instructions provided by the manufacturer
   (Thermo Fisher Scientific). The RNA obtained was either used for
   reverse transcription into cDNA or subjected to RNA sequencing.

cDNA synthesis and qPCR

   First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) was used to
   transcribe isolated RNA using random hexamer primers, resulting in the
   generation of cDNA. The cDNA was then analyzed through qPCR using the
   SensiMix™ SYBR® Low-ROX kit (Meridian Bioscience) with each sample
   pipetted in triplicate. The reaction was carried out in a Quant studio
   6 flex Cycler. The primer sequences used for qPCR: Cox14 (FP:
   ATGCCATCTGCCAAGCAGCTAG; RP: CTGAGGTCTTCTGCTCTTCTGC), Isg15 (FP:
   AAGCAGATTGCCCAGAAGATTG; RP: CTGTACCACTAGCATCACTGTG), s18 (FP:
   CTCAACACGGGAAACCTCAC; RP: CGCTCCACCAACTAAGAACG), Ifi27 (FP:
   CAGTCTCTGGCACCATTCTA; RP: CAAAGCAGGTTCCTCTCCAATG), Eif4e3 (FP:
   ACCACTTTGGGAGGAGGAGAG; RP: CACTCACGCTGACTCCGATG), Gbp3 (FP:
   TGGTGATGTAGAAAAGGGTGAC; RP: CTCACTGGAGTCCTTCACTT TG), Oasl1 (FP:
   CAGCAGACCCCACCAACAAT; RP: GTGCTCTCTTCACCCTC CAG), Ifit1 (FP:
   ATGGGAGAGAATGCTGATGGTG; RP: AACTGGACCTGCTCTG AGATTC), Usp18 (FP:
   TCGGTGATACCAAGGAACAGT; RP: ATCCTTCCAGGTGACCCAAC), Actb (FP:
   AGATGACCCAGATCATGTTTG; RP: GAGCATAGCCCTCGTAGATG), Nd4 (FP:
   ACCCGATGAGGGAACCAAAC; RP: AGCGTCTAAGGTGTGTGTTGT), Nd6 (FP:
   GGGTTTGGTGGATCGTTTTTAGG; RP: TCCCCAAGTCTCTGGATATTCCT), Atp6 (FP:
   ACAGGCTTCCGACACAAACTA; RP: AGTAGCTCCTCCGATTAGGTGT), Atp8 (FP:
   CATCACAAACATTCCCACTGGC; RP: TGAGGCAAATAGATTTTCGTTCATT), Cox3 (FP:
   TGCAGGATTCTTCTGAGCGT; RP: AGTAGTGGGACTTCTAGAGGGTT), s12 (FP:
   TGCAGGATTCTTCTGAGCGT; RP: AACTGCAACCAACCACCTTC), Calr (FP:
   AGTTCTTGGACGGAGATGCC; RP: GGCGGACAGTGCGTAAAATC), s16 (FP:
   CGGCTAAACGAGGGTCCAAC; RP: TATTCTCCGAGGTCACCCCAA), Nd1 (FP:
   TCCCCTACCAATACCACACCC; RP: GCTACGGCTCGTAAAGCTCC), Cox2 (FP:
   AGACGAAATCAACAACCCCG; RP: GGCAGAACGACTCGGTTATCA), Nd2 (FP:
   ATAGGGGCATGAGGAGGACT; RP: TGAGTAGAGTGAGGGATGGGTT), Nd5 (FP:
   ACACATCTGTACCCACGCAT; RP: GGCGAGGCTTCCGATTACTA), Irf7 (FP:
   TGAGCGAAGAGAGCGAAGAG; RP: GCCCACAGTAGATCCAAGC).

SDS PAGE and western blotting

   Standard methods were utilized to perform SDS PAGE and gradient
   Tricine-SDS PAGE (10–18%). Tissue and cell lysis were performed using
   T-PER™ Tissue Protein Extraction Reagent (Thermo Scientific) along with
   1X Protease Inhibitor Cocktail and 1X PhosphoSTOP. The protein
   concentration was determined by utilizing a Bradford assay. For Western
   blotting, standard semi-dry transfer was employed. Antibodies used:
   Rabbit polyclonal Anti-IFIT1 antibody (ab111821) Abcam, dilution 1:100;
   Mouse monoclonal Anti-Oas1a Antibody (E-2): sc-365072 Santa Cruz,
   dilution: 1:1000; Mouse monoclonal Anti-ISG15 Antibody (F-9): sc-166755
   Santa Cruz, dilution: 1:500; Rabbit polyclonal anti-VDAC Home-made
   antibody, dilution: 1:1000 Rabbit polyclonal anti-ATP5 Home-made
   antibody, dilution: 1:2500; Rabbit polyclonal anti-RIESKE Home-made
   antibody, dilution: 1:2000 Anti-Rig-I (D14G6) Rabbit mAb#3743 Cell
   Signaling Technologies, dilution: 1:1000; Anti- MDA-5 (D74E4) Rabbit
   mAb#5321 Cell Signaling Technologies, dilution 1:1000; Anti- STING
   (D2P2F) Rabbit mAb #13647 Cell Signaling Technologies, dilution 1:100;
   Mouse monoclonal ZBP1 Antibody (H-9): sc-271483 Santa Cruz, dilution
   1:500; Rabbit polyclonal Anti-Lamin B1 antibody Abcam, dilution 1:500;
   Mouse monoclonal Anti-MTCO1 antibody [1D6E1A8] (ab14705) Abcam,
   dilution 1:1000;

Blue native PAGE

   Cardiac mitochondria were solubilized in a mitochondria solubilization
   buffer (40 mM Tris/Cl pH 7.4, 100 mM NaCl, 20% Glycerol, 0.2 mM EDTA pH
   7.4) containing 1% digitonin for 30 min at 4 °C, to a final
   concentration of 1 mg/ml. The solubilized samples were separated by
   4–13% polyacrylamide gradient gel and transferred using a standard
   semi-dry method.

In organello translation

   Mitochondria from freshly isolated liver tissue was used to perform in
   organello translation. 200 μg of isolated mitochondria were dissolved
   in freshly prepared 100 μl of Translation Buffer (25 mM sucrose, 100 mM
   Mannitol, 80 mM KCl, 5 mM MgCl[2], 1 mM Kpi pH7.4, 25 mM HEPES-KOH
   pH7.4, 6 mM ATP in H[2]O pH7.2 KOH, 0.75 mM GTP I H[2]O pH7,2 KOH,
   12 mM Creatine phosphate, 10 mM Na-Succinate, 0.625 mg/ml Creatine
   Kinase, 9 mM 2-Ketoglutarate, 0.15 mM Amino Acid mix, 100 μg/ml
   Emetine, 100 μg /ml Cyclohexamide, 2.5 mM Malate, 1 mg/ml BSA, 1X
   Protease Inhibitor Cocktail (PI)). The mitochondria were incubated at
   37 °C for 10 min. 1 μCi ^35S Methionine was added and the mitochondria
   were incubated for 1 hour at 37 °C while shaking. The reaction was
   stopped by adding 10 μM non-radioactive methionine. The mitochondria
   were pelleted down at 7000 xg, and washed once with Mitochondrial
   Isolation Wash Buffer. The samples were further used for steady state
   analysis.

   To visualize the signal of translation products labeled with ^35S
   Methionine, isolated mitochondria used for in organello translation,
   were loaded on a standard 10–18% Tris-tricine gradient gel and were
   further transferred on a PVDF membrane. The membrane was dried and was
   exposed for one day on a phosphor screen. The radioactive signal was
   visualized using Amersham Typhoon scanner (Amersham). The signal
   quantification was done using Image Quant LT (GE Healthcare).

Proteomics

Sample preparation for LC-MS/MS

   Hepatocytes from three biological replicates of WT and of COX14^M19I
   were lysed in 8% sodium dodecyl sulfate (SDS), heated for 10 min at
   99 °C and sonicated in Bioruptor (Diagenode) for 10 min using 30 s
   on/off cycles. Protein concentration in the lysate was determined using
   BCA assay kit (ThermoFisher Scientific) according to manufacturer’s
   protocol. Proteins were reduced by incubation with 10 mM Dithiothreitol
   (DTT) for 30 min at 60 °C followed by alkylation with 40 mM
   iodoacetamide (IAA) for 30 min at 37 °C, which was quenched by addition
   of 20 mM DTT and incubated for 5 min at 37 °C. Protein clean-up was
   performed according to the procedure by Hughes et al.^[213]45. After
   washing, beads with bound sample were resolubilized in 50 mM TEA-B
   buffer. DNA and RNA were digested with Universal Nuclease (Pierce, 250
   U/µl) for 30 min at 37 °C in presence of 1 mM MgCl[2]. Proteins were
   digested with trypsin (MS grade, Promega) using 1:20 (w/w)
   trypsin-to-protein ratio and incubated overnight at 37 °C. Peptides
   were dried in SpeedVac concentrator (Eppendorf). and then re-dissolved
   in 50 mM TEA-B. Labelling reaction with respective TMT6plex labelling
   reagent (ThermoFisher Scientific) was conducted according to
   manufacturer’s protocol. Labelled samples were pooled, dried in
   SpeedVac concentrator, and desalted using pre-packed C18-columns
   (Harvard apparatus). Desalted and dried peptides were dissolved in
   mobile phase A (10 mM NH[4]OH, pH 10) and fractionated using
   reversed-phase chromatography under basic pH. Fractionation by bRP-HPLC
   was performed on an Agilent 1100 series HPLC system. Elution was
   performed at a flow rate of 60 μl/min using mobile phase A and B (10 mM
   NH[4]OH in 80% ACN, pH 10) with a Waters XBridge C18 column (3.5-μm
   particles, 1.0 mM inner diameter, and 150 mM in length). The gradient
   was 5% B for 5 min, then changed to 10% B in 3 min, to 35% B in 34 min,
   to 50% B in 8 min, to 90% B in 1 min, to 95% in 5 min, back to 5% B in
   2 min, then held at 5% B for 6 min (64 min total runtime). Flow through
   was collected for 6 min and then fractions of 60 µl were collected for
   the rest of the gradient. Fractions were combined into a total of 24
   fractions plus the flow through, samples were dried in a SpeedVac
   concentrator, and stored at −20 °C for subsequent LC-MS/MS analysis.
   Labelling scheme for the comparison of cytoplasmic proteins in WT vs.
   COX14^M19I - (Sample, TMT label): 126, WT 1; 127, WT 2, 128, COX14^M19I
   1, 129, COX14^M19I 2, 130, WT 3, 131, COX14^M19I 3. Isolated
   mitochondria from three biological replicates for each WT and
   COX14^M19I were treated with a similar protocol with the following
   changes. Lysis was performed with lysis buffer (4% SDS, 100 mM
   ethylenediaminetetraacetic acid (EDTA) Hepes pH 8.0, 2x complete
   protease inhibitor cocktail) instead of 8% SDS. Heating was performed
   at 95 °C for 5 min. Protein reduction and alkylation were performed
   with 10 mM Tris(2-chlorehyl)phosphine (TCEP) (Sigma-Aldrich) and 40 mM
   chloroacetamide (CAA) (Sigma-Aldrich) for 30 min at 55 °C. Proteins
   were precipitated with acetone (1:4 v/v sample-to-acetone ratio)
   overnight at −20 °C. The next day, the protein pellet was washed three
   times with ethanol and air-dried before solubilization with 1% Rapigest
   (Waters) in 50 mM TEA-B and incubation at 95 °C for 3 min. Then, four
   times the volume 100 mM TEA-B were added and proteins were digested
   with trypsin (1:20 w/w trypsin-to-protein ratio) overnight at 37 °C.
   After digestion, sample volume was reduced in the SpeedVac concentrator
   to ~15–20 µl. Samples were then subjected to TMT labelling and bRP-HPLC
   fractionation as described above. Labelling scheme for comparison of
   mitochondrial proteins in WT vs. COX14^M19I – (Sample, TMT labelling):
   126, WT 1; 127, COX14^M19I 1; 128, WT 2; 129, COX14^M19I 2; 130, WT 3;
   131, COX14^M19I 3.

LC-MS/MS

   Dried peptides were solubilized in 2% (v/v) ACN 0.1% (v/v) TFA and
   injected onto a C18 PepMap100-trapping column (0.3 × 5 mM, 5 µm, Thermo
   Fisher Scientific) connected to an in-house packed C18 analytical
   column (75 µm x 300 mM; Reprosil-Pur 120 C18 AQ, 1.9 µm, Dr. Maisch
   GmbH). The columns were pre-equilibrated using a mixture of 95% buffer
   A (0.1% (v/v) formic acid in water), 5% buffer B (80% (v/v) ACN, 0.1%
   (v/v) FA in water). Liquid chromatography was operated by UltiMate 3000
   RSLC nanosystem (Thermo Fisher Scientific). Peptides were eluted over
   118 min using a gradient ranging from i) 5% to 8% buffer B over 4 min,
   ii) 8% to 32% buffer B over 72 min, iii) 32% to 50% over 30 min, and
   ending with a washing step at 90% of buffer B for 6 min and a
   re-equilibration step at 5% buffer B for 5 min. MS spectra were
   collected using an Orbitrap Fusion Tribrid mass spectrometer (Thermo
   Fisher Scientific). MS1 scans (350–1650 m/z) were acquired in positive
   ion mode with a resolution of 120,000 at 200 m/z, an automatic gain
   control (AGC) target of 250%, and with 50 ms maximum injection time.
   Precursor ions (allowed charge states 2–7, dynamic exclusion 40 s) were
   isolated using a 1.6 m/z isolation window and fragmented at normalized
   collision energy (HCD) of 30%. MS2 fragment spectra were acquired with
   a resolution of 30,000, AGC target of 500%, and 54 ms maximum injection
   time. The ten most intense fragment ions were selected for a subsequent
   synchronous precursor selection (SPS)-MS3 scan using an isolation
   window of 2 m/z and HCD collision energy of 55%. The SPS-MS3 spectra
   were acquired at a resolution of 30,000 and maximum injection time of
   54 ms.

LC-MS/MS data analysis

   Raw files were processed using MaxQuant version 2.1.4.0^[214]46,[215]47
   using default settings except selection of reporter ions in MS3
   (TMT6plex) for quantification. MS2 spectra were matched against
   canonical amino acid sequences of mus musculus proteins retrieved from
   Uniprot (August 2022). Differential expression analysis was conducted
   with Perseus software^[216]46. Corrected reporter intensity values were
   loaded into Perseus, contaminants and reverse hits were excluded, as
   well as protein hits with >70% missing values. Missing values were
   imputed from normal distribution with Perseus‘ default settings.
   Log2-transformed reporter ion intensities were then normalized using
   median polishing and subjected to statistical testing. Student’s T-test
   was executed with S0 = 0, Benjamini-Hochberg FDR of 0.05. Differences
   in protein group intensities between COX14^M19I mutant and wild type
   samples were expressed as log2 fold changes (COX14^M19I/WT). Protein
   groups with q-value < 0.05 were considered as differentially expressed.
   Plots were generated with R programming language and R Studio (version
   1.4.1106).

Complex activity assays

   Isolated mitochondria from mouse hearts were utilized to perform
   Complex IV activity assays using the Complex IV Rodent Enzyme Activity
   Microplate Assay Kit (ab109911). ELISA-based plate assays were
   performed in accordance with the manufacturer’s instructions (Abcam).
   To determine Complex IV activity, absorbance at 550 nm was measured. A
   Synergy H1 microplate reader (BioTek) was used to obtain the
   measurements.

Real time respirometry

Mito stress test

   For the determination of the Oxygen Consumption rate (OCR) in isolated
   mitochondria a XF96e Extracellular Flux Analyzer was used. To measure
   basal respiration in isolated mitochondria, a Mitochondrial Assay
   Buffer was used. The buffer contained 70 mM Sucrose, 210 mM Mannitol,
   5 mM HEPES, 1 mM EGTA, 10 mM KH[2]PO4, 5 mM MgCl[2], 0.5% BSA, and
   40 mM ADP (added freshly). Two different supplements were used: either
   10 mM Pyruvate, 4 mM Malate, and 4 mM ADP, or 10 mM Succinate and 4 mM
   ADP. Oxygen consumption was measured under different metabolic
   conditions by adding 3 µM Oligomycin, 4 µM CCCP, and 0.5 µM
   Antimycin/Rotenone.

Palmitate oxidation stress test

   For the purpose of measuring long chain fatty acid oxidation in
   isolated mitochondria from mouse liver, a XF96e Extracellular Flux
   Analyzer was used. The mitochondria were freshly isolated at 4 °C as
   described above and plated in a Seahorse plate resuspended in
   Respiration Buffer (214 mM Sucrose, 12.8 mM KCl, 0.86 mM EGTA, 4.3 mM
   MgCl[2], 25.7 mM K[2]HPO[4] 80μΜ Palmitoyl, 4.3μΜ Carnitine, 45 mM
   Malate, 3.7 mM ADP; ph 7.4) at room temperature. Addition of 0.2 mM ETO
   in the buffer was used as negative control. The basal Oxygen
   consumption rate was measured.

Serum biochemistry

   Blood samples were taken from isoflurane-anesthetized mice by
   retro-bulbar sinus puncture with non-heparinized glass capillaries.
   Samples were collected in two different types of coated tubes
   (Lithium/Heparin (KABE cat # 078028) and EDTA (KABE cat # 078035).
   After collection each sample was mixed by gentle inversion and then
   kept on RT until centrifugation. Samples were centrifuged within 1 hour
   of collection at 5000 × g, for 10 min at 8 °C. Plasma is analyzed on
   Beckman AU480 biochemical analyzer. All calibration and quality control
   samples are within expiration period and measured values fulfilled
   required parameters for each method used according to the IMPC
   (International Mouse Phenotyping Consortium) core pipeline
   ([217]https://www.mousephenotype.org/impress/index).

Immunophenotyping

   The methodology was based on standard immunophenotyping protocols of
   the Adult and Embryonic Phenotype Pipeline that has been agreed by the
   research institutions involved MPReSS -International Mouse Phenotyping
   Resource of Standardised Screens
   ([218]https://www.mousephenotype.org/impress/protocol/174/7). According
   to these guidelines, we use mouse spleenocytes for standard
   immunophenotyping.

Histology

   Tissues were fixed in 4% paraformaldehyde, processed on a Leica ASP6025
   automatic vacuum tissue processor, and embedded in a Leica EG1150 H + C
   embedding station. Sections (2 µm) were obtained using a Leica RM2255
   rotary microtome, then deparaffinized and hydrated in xylen and
   descending ethanol solutions before being stained with Picrosirius red
   (Direct red 80) (Sigma Aldrich) using 0.5 g picrosirius powder stain
   per 500 mL picric acid solution. Weigert’s haematoxylin kit (Sigma
   Aldrich) was used to counterstain the nuclei, following the
   manufacturer’s instructions. Stained slides were dehydrated, mounted
   with coverslips in an automatic staining system (Leica ST5010-CV5030),
   and digitalized in a slide scanner AxioScan Z.1 (Zeiss).

OCT and electroretinography

Optical coherence tomography (OCT)

   In order to visualize the retinal morphology, the spectral-domain
   optical coherent tomography imaging system adapted to mouse eye
   (Heidelberg Engineering, Germany) was used. Examined mice were
   anaesthetized similarly as is described for Electroretinography. The
   eyes were dilated by a drop of 0.5% Atropine (Ursapharm, Prague, Czech
   Republic). Transparent eye gel Vidisic (Vidisic, Bausch&Lomb, Prague,
   Czech Republic) was used to prevent eyes from drying. The eyes were
   fitted with +100 dioptric lenses (Roland Consult, Brandenburg, Germany)
   and the mice underwent an imaging session^[219]48.

   Retinal thickness was assessed by an automatic segmentation algorithm
   complemented with minor manual corrections. Animals were scanned
   repeatedly at the age of 5, 15 and 30 weeks.

Electroretinography

   Full-field single-flash electroretinography (ERG) was recorded under
   general anesthesia (i.m. Tiletamine + Zolazepam, 30 + 30 mg/kg,
   supplemented with 3.2 mg/kg Xylazine) using RETIanimal system (Roland
   Consult, Brandenburg, Germany). Animals were placed on a heated pad
   (38 °C), their pupils dilated by a drop of 0.5% Atropine (Ursapharm,
   Prague, Czech Republic) and eyes protected against drying by
   transparent eye gel (Vidisic, Bausch&Lomb, Prague, Czech Republic).
   Golden wire rings placed on the cornea served as active electrodes,
   reference subdermal electrode was in the central line near the snout
   and another subdermal electrode near the tail base represented the
   ground electrode. Scotopic ERG: mice were dark-adapted overnight;
   stimuli were single flashes of white light of luminances distributed in
   9 exponential steps between 0.001 and 10 cd.s.m − 2. Photopic ERG: mice
   were exposed to 30 cd.m − 2 white light for 2 min. and recording
   continued with this background illumination, stimuli were single
   flashes between 1 and 100 cd.s.m − 2 in 5 intensity steps. For each
   stimulation intensity, 20 responses were collected and averaged and the
   amplitude of a, b, and c- wave, respectively, was measured on the
   averaged signal^[220]48. Experiments were approved by the Animal Care
   and Use Committee of the Czech Academy of Sciences.

Echocardiography

   In order to visualize the small mouse heart and record the rapid heart
   rates, the cardiac ultrasound imaging is acquired using the Vevo 2100
   Imaging System (VisualSonics, Inc.) with a 30 MHz transducer (MS400)
   operating at a frequency that provides highly reliable and reproducible
   image quality. Echocardiography is performed on anaesthetized
   (isoflurane) mice and during imaging, the concentration of anesthesia
   is controlled to maintain a heart rate of 450–500 beats/min. The
   primary echocardiographic screen includes two-dimensional (2D) B-mode
   (“brightness” mode) imaging view of left ventricle (LV) along the
   parasternal long axis (PLAX) and M-mode (“motion” mode) image of the
   heart in short-axis view (SAX) for accurate linear measurements of left
   ventricular internal dimensions.

   A proper cardiac B-mode view allows visualization of the left atrium
   and ventricle, a small portion of the right ventricular wall and the
   output of aorta. The apex and the beginning of ascending aorta of the
   heart lay on the same horizontal line. M-mode echocardiography was
   performed at the ventricular level at the papillary muscle level and
   leads to a 1D high resolution temporal course of the diameter changes
   of the LV and the wall thickness of the anterior and posterior LV wall
   in systole as well in diastole. All M-mode echocardiography
   measurements follow the guidelines of the American Society of
   Echocardiography. The SAX M-mode images were used to measure left
   ventricular end-diastolic internal diameter (LVEDD), left ventricular
   end-systolic internal diameter (LVESD), and diastolic and systolic
   posterior wall thickness (LVPW) in three consecutive beats. Fractional
   shortening (FS) was calculated as FS% = [(LVEDD-LVESD)/LVEDD]x100.
   Ejection fraction (EF) was calculated as EF% = 100 x
   ((LVvolD-LVvolS)LVvolD) with LVvol = ((7.0/(2.4 + LVID) x LVID3). The
   corrected left ventricular mass (LV Mass Cor) was calculated as LV Mass
   Cor=0.8 (1.053 x ((LVIDD + LVPWD + IVSD)3 - LVIDD3)). The Stroke volume
   (SV) was the volume of blood pumped from one ventricle of the heart
   with each beat. The Stroke volume of the LV was obtained by subtracting
   end-systolic volume (ESV) from end-diastolic volume (EDV). The cardiac
   output (CO) was the volume of blood pumped by the heart per minute. In
   addition, heart rate and respiratory rate were calculated by measuring
   three systolic intervals, respectively three respiratory intervals.

RNA sequencing

   Prior to library preparation, RNA quality was assessed by measuring the
   RNA integrity number (RIN) using a Fragment Analyzer HS Total RNA Kit
   (Advanced Analytical Technologies, Inc.). Samples with a RIN over 8
   were selected. For library preparation, 200 ng of total RNA was used to
   create non-stranded mRNA-Seq libraries with the TruSeq RNA Library
   Preparation Kit (#RS-122-2001, Illumina). The size range of the final
   cDNA libraries was determined using the SS NGS Fragment 1-6000 bp Kit
   on the Fragment Analyzer, yielding an average size of 340 bp. Accurate
   quantification of the final libraries was performed using the DeNovix
   DS-Series System. The libraries were sequenced on the Illumina HiSeq
   4000 platform (SE; 1 × 50 bp; 30–35 million reads per sample).

Data processing and bioinformatic analyses

   To process the sequencing data, an in-house software pipeline was used.
   Illumina’s bcl2fastq (v2.20.0) converted the base calls in the
   per-cycle BCL files to the per-read FASTQ format from raw images and
   performed adaptor trimming and demultiplexing.

Mapping and normalization

   Initially, sequences were aligned to the reference mouse genome using
   the RNA-seq alignment tool (version 2.7.8a) allowing for two mismatches
   within 50 bases. Subsequently, read counting was performed using
   featureCounts. Read counts were analyzed in the R/Bioconductor
   environment (version 4.0.5) using the DESeq2 package version 1.31.5.
   Candidate genes were filtered using an absolute log2-fold change >1 and
   FDR-corrected P value < 0.05. Gene annotation was performed using mouse
   entries via biomaRt R package version 2.46.3.

   Furthermore, FastQC (v 0.11.9) was used for quality control of the raw
   sequencing data. STAR (v2.7.3a)^[221]49, was used to map the reads to
   the genome (mm10 for mouse data) with a maximum of two mismatches
   allowed per pair and default parameters. The featureCounts program from
   the Subread package (v2.0.0)^[222]50,[223]51 was used to count the
   number of aligned reads overlapping with exons for each gene.

   To identify differentially expressed genes, the unwanted sources of
   variation were identified and corrected using the svaseq function from
   the sva R package (v3.40.0)^[224]52. The DESeq2 R package
   (v1.32.0)^[225]53 was used to perform the differential expression
   analysis. Genes were considered significantly differentially expressed
   if they passed a cut-off of baseMean >= 50, fold change >1.15
   or < −1.15, and adjusted p-value < 0.05. The up and downregulated genes
   were analyzed for over-representation (pathway enrichment) using the
   WebGestaltR R package (version 0.4.4)^[226]54 with KEGG database
   terms^[227]55.

   Plots were created in R (version 4.1.0) using custom scripts.
   Over-representation (pathway enrichment) analysis of the up and
   downregulated genes was done using the WebGestaltR R package (version
   0.4.4)^[228]54. The Irf regulatory network was built using a custom
   script using the specific hub gene and deregulated gene lists, by
   parsing the following databases for interactions involving the input
   genes: RegNetwork^[229]56, STRING^[230]57, RISE^[231]58 and
   NPInter^[232]59. The different kinds of (directed) interactions
   considered were: protein-protein (PP), RNA-RNA (RR), transcription
   factor-protein (TP), transcription factor-RNA (TR) and
   RNA-transcription factor (RT) interactions. Cytoscape^[233]60 was used
   to build customized networks and visualizations.

Lipidomics

   Liver samples that were flash frozen underwent lipidomics
   analysis^[234]19: Lipid extractions were performed in the presence of
   internal lipid standards using an acidic liquid-liquid extraction (ABD)
   method using chloroform:methanol:HCl (20:40:0.3, vol:vol:vol)^[235]61,
   except for acid-labile plasmalogens, which were extracted under neutral
   conditions^[236]62. As controls, blank extractions and extractions in
   the presence of internal lipid standard without samples were performed.
   As quality control measure, standards were titrated into complex lipid
   mixtures of known composition. To ensure that similar amounts of lipids
   were subjected to extractions, a test extraction (ABD) was performed to
   determine the concentration of PC as a bulk membrane lipid and to adapt
   extractions volumes to similar total lipid amounts. On average, a total
   lipid amount of 3000 pmol was used for extractions. Lipid extractions
   were performed in the presence of internal lipid standards for each
   lipid class, with the standards resembling the structure of the
   endogenous lipid species. Following this approach, a relative
   quantification of lipid species was performed. Glass Pyrex tubes
   (16x100mM) and screw caps with PTFE-liner (Corning) were used for
   extractions. Glass pipettes and glass syringes were used for handling
   of samples in organic solvents. Lipid standards were added from stock
   solutions (1–10 µM, in chloroform) prior to extractions, using a master
   mix consisting of 50 pmol phosphatidylcholine (PC, 14:1/14:1,
   20:1/20:1; 22:1/22:1, Avanti Polar Lipids), 50 pmol sphingomyelin (SM,
   d18:1/13:0, d18:1/17:0, d18:1/25:0, semi-synthesized^[237]62, 100 pmol
   deuterated cholesterol (D[7]-cholesterol, Avanti Polar Lipids), 28 pmol
   phosphatidylinositol (PI, 17:0/ 20:4, Avanti Polar Lipids), 25 pmol
   phosphatidylethanolamine (PE) and 25 pmol phosphatidylserine (PS) (both
   14:1/14:1, 20:1/20:1, 22:1/22:1, semi-synthesized^[238]62), 25 pmol
   diacylglycerol (DG, 15:0/18:1-D[7], Larodan), 25 pmol cholesteryl ester
   (CE, 19:0, Avanti Polar Lipids, 9:0, 24:1, Sigma), and 24 pmol
   triacylglycerol (TG, D5-mixture, LM-6000, Avanti Polar Lipids), 5 pmol
   ceramide (Cer, d18:1 with N-acylated 14:0, 17:0, 25:0,
   semi-synthesized^[239]62 or Cer d18:1/18:0-D[3], Matreya) and 5 pmol
   glucosylceramide (HexCer) (GlcCer d18:1/17:0, Avanti Polar Lipids), 5
   pmol lactosylceramide (Hex2Cer, d18:1/17:0; Avanti Polar Lipids), 10
   pmol phosphatidic acid (PA, 17:0/20:4, Avanti Polar Lipids), 10 pmol
   phosphatidylglycerol (PG, 14:1/14:1, 20:1/20:1, 22:1/22:1,
   semi-synthesized^[240]62) and 5 pmol lysophosphatidylcholine (LPC,
   17:1, Avanti Polar Lipids). The phosphatidylethanolamine plasmalogen
   (PE P-) standard mix consisted of 22 pmol PE P-Mix 1 (16:0p/15:0-d3,
   16:0p/19:0), 31 pmol PE P- Mix 2 (18:0p/15:0-d3, 18:0p/19:0), 43 pmol
   PE P-Mix 3 (18:1p/15:0-d3, 18:1p/19:0). Semi-synthesis of PE P- was
   performed using deuterated fatty acids^[241]63. The final chloroform
   phase was evaporated under a gentle stream of nitrogen at 37 °C.
   Samples were either directly subjected to mass spectrometric analysis,
   or were stored at −20 °C prior to analysis, which was typically done
   within 1–2 days after extraction. Lipid extracts were resuspended in
   10 mM ammonium acetate in 60 µl methanol. Samples were analyzed on an
   QTRAP6500+ (Sciex) mass spectrometer, except for the measurement of
   cholesterol acetate, which was performed on a QTRAP5500 (Sciex), and TG
   and DG measurements, which were performed on a Q Exactive. Both MS
   systems were coupled to a robotic with chip-based (HD-D ESI Chip,
   Advion Biosciences) electrospray infusion and ionization via a Triversa
   Nanomate (Advion Biosciences). For QTRAP6500+ and 5500 measurements,
   2 µl aliquots of the resuspended lipid extracts were diluted 1:10 in
   10 mM ammonium acetate in methanol in 96-well plates (Eppendorf twin
   tec 96). For TG and DG measurements on a Q Exactive, 5 µl aliquots of
   the reconstituted lipid extracts were diluted with 15 µl 7.5 mM
   ammonium formate in isopropanol:methanol:chloroform (v:v:v, 4:2:1). For
   cholesterol determinations, the remaining lipid extract was again
   evaporated and subjected to acetylation^[242]64 All MS settings and
   scan procedures are listed in Supplementary Data [243]S1. Data
   evaluation was done using LipidView 1.3 beta (Sciex) for QTRAP
   6500 + /5500 data or LipidXplorer 1.2.8.1
   ([244]https://zenodo.org/record/3570469^[245]65,[246]66) and
   Peakstrainer^[247]67 for Q Exactive data, and in-house-developed
   software (ShinyLipids and ShinyLipidcountr_alpha4). The amount for
   endogenous molecular lipid species was calculated based on the
   intensities of the internal standards.

Oil O red staining

   Lipid bodies were stained using OilO according to the manufacturer
   protocol. Briefly, a stock solution of 10 mg/ml of OilO in isopropanol
   was thoroughly mixed and set aside shortly. 15 min later, a 3:2 V:V
   dilution was prepared in water and wet aside again for a few minutes.
   The diluted solution was filtered and used fresh for staining. Cells
   were fixed using 4% poly-formaldehyde solution in PBS for 30 min, then
   quenched for 15 min with a 100 mM NH[4]Cl[2] solution in PBS. Following
   two quick washes with water and one wash with 60% isopropanol, OilO was
   applied for 5 min. Several washes in water were followed and then the
   coverslips were fixed on glass slide using embedding medium containing
   Hoechst for nuclei staining.

   Images were captured using an inverted Nikon Ti epifluorescence
   microscope (Nikon Corporation, Chiyoda, Tokyo, Japan) fitted with a
   Plan Apochromat 60×, 1.4 NA oil immersion objective, an HBO-100W Lamp,
   and an IXON X3897 Andor (Belfast, Northern Ireland, UK) camera. The
   NIS-Elements AR software (version 4.20; Nikon) was utilized to operate
   the camera. Prior to image analysis, the images were deconvolved as
   follows. First, the point spread function (PSF) was generated using the
   PSF generator Fiji plugin via Born and Wolf 3D optical model. Then, the
   DevonvolutionLab2 plugin from Fiji was applied using Richardson-Lucy
   algorithm. The deconvolved images were analyzed using a built-in macro
   in Fiji. First, a threshold was applied to the OilO staining channel
   via Otsu method and the particles with 0.5–1 circularity were selected.
   Then, the acquired mask was used to measure the area and the average
   fluorescent intensity of individual particles. Nuclei stainings were
   used to count the number of cells in each image.

Electron microscopy

   For transmission electron microscopy mice at the age of 16 weeks were
   fixed by perfusion fixation^[248]68 or by immersion fixation using 2.5%
   glutaraldehyde, 4% formaldehyde and 0.5% NaCl in 0.1 M phosphate puffer
   pH7.3 as fixing solution. After dissection of the liver, small pieces
   of tissue were processed following a modified rOTO
   protocol^[249]69,[250]70. Briefly, samples were washed in 0.1 M
   phosphate buffer, incubated for 3 h at 4 °C in 2% OsO[4] (Science
   Services GmbH, Munich, Germany) and 0.25% K[4][Fe(CN)[6]] (Electron
   Microscopy Sciences), washed with H[2]O and incubated with 0.1%
   thiocarbohydrazide (Sigma-Aldrich) for 1 h at room temperature. For
   contrast enhancement the tissue was again with 2% OsO[4] for 90 min at
   room temperature. The samples were then washed with H[2]O, contrasted
   overnight at 4 °C with 2% uranyl acetate (SPI-Chem) and washed again
   with H[2]O, followed by dehydration in an increasing acetone series
   (30%, 50%, 75%, 90%, 3 × 100%). The tissue was infiltrated with
   increasing concentrations of Durcupan (Sigma-Aldrich, components A, B,
   C) in acetone. Fresh Durcupan with accelerator (component D) was added
   to the samples for 5 h before embedding the samples in resin blocks.
   The blocks were polymerized for 48 h at 60 °C. Ultrathin sections were
   obtained using a 35° diamond knife (Diatome AG, Biel, Switzerland) and
   an ultramicrotome (RMC PowerTome PT-PC, Science Services, Munich,
   Germany) and placed on 100 mesh hexagonal copper grids (Science
   Services GmbH, Munich, Germany). Electron micrographs were taken with
   an LEO912 transmission electron microscope (Carl Zeiss Microscopy
   Deutschland GmbH, Oberkochen, Germany) using a 2k on-axis CCD camera
   (TRS, Moorenweis, Germany).

   For volume EM by focused ion beam-scanning electron microscopy
   (FIB-SEM) the blocks were trimmed with a 90° rectangular shaped
   trimming diamond knife (Diatome AG, Biel, Switzerland) and then
   attached to the SEM stub (Science Services, Munich, Germany) using
   silver filled epoxy (Epoxy Conductive Adhesive, EPO-TEK, Electron
   Microscopy Service) and polymerized at 60° overnight. After coating the
   samples with a 10 nm gold layer with the sputter coater EM ACE600
   (Leica Microsystems, Vienna, Austria) at 35 mA current the sample stubs
   were placed into the Crossbeam 540 FIB-SEM (Carl Zeiss Microscopy
   Deutschland GmbH). In the chamber, a protective 500 nm platinum layer
   was deposited on top of the region of interest using a 3 nA current.
   Atlas 3D (Atlas 5.1, Fibics, Canada) software was used to collect the
   3D data. The block face was exposed with a 15 nA current, and a 7 nA
   current was used to polish the surface. The images were acquired at
   1.5 kV with the ESB detector (450 V ESB grid, pixel size x/y 5 nm) in a
   continuous mill-and-acquire mode using a current of 1.5 nA for milling
   (z-step 10 nm). Image post-processing was performed using the image
   processing package Fiji^[251]71: Alignment of images was performed
   using the SIFT algorithm. Then the images were cropped, the contrast
   inverted and smoothed using a Gaussian blur (sigma 1), and a local
   contrast enhancement was applied (CLAHE: blocksize 127, histogram bins
   256, maximum slope 1.5).

Metabolomics

   Samples were prepared and processed for LC-MS analysis of polar and
   lipid metabolites as detailed in Supplementary Methods. Anion-Exchange
   Chromatography Mass Spectrometry (AEX-MS) was used for the analysis of
   anionic metabolites. Whereas semi-targeted liquid
   chromatography-high-resolution mass spectrometry-based (LC-HRS-MS)
   analysis was performed for amine-containing metabolites.

Samples extraction of polar metabolites for metabolomics

   For the metabolite extraction 20–30 mg of snap frozen liver tissue were
   collected in 2 mL Eppendorf tubes containing a 5 mM metal ball. These
   tubes were kept in liquid nitrogen before grinding them to a fine
   powder using a TissueLyser (Qiagen) set to 25 Hz for 1 min.

   Immediately after homogenization 1 ml of −20 °C
   acetonitrile:methanol:water (4:4:2 (v:v:v)) mixture, containing 0.1 µL
   of U-^13C^15N amino acid mix (Cambridge isotopes MSK_A2-1.2), each
   0.1 µl of a 1 mg/mL stock solution of ^13C[10] ATP, ^15N[5] ADP and
   ^13C[10]^15N[5] AMP (Sigma) and 0.2 µL of a 100 µg/mL stock solution of
   deuterated citric acid (Sigma), was added. The extraction solvents were
   all (UPLC-grad, Biosolve). After addition of the extraction buffer, the
   samples were immediately vortexed for 10 seconds before incubating them
   for additional 30 min at 4 °C on an orbital shaker. To remove the
   insoluble material the metal balls were removed from each sample using
   a magnet and the tubes were centrifuged for 10 min at 21.000 × g at
   4 °C. The cleared supernatant was transferred to fresh 1.5 mL Eppendorf
   tube and immediately concentrated to complete dryness in a speed vacuum
   concentrator at room temperature. Samples were either stored at −80 °C
   or processed immediately for LC-MS analysis.

Anion-Exchange Chromatography Mass Spectrometry (AEX-MS) for the analysis of
anionic metabolites

   Extracted metabolites were re-suspended in 400 µl of UPLC/MS grade
   water (Biosolve), of which 200 µl were transferred to polypropylene
   autosampler vials (Chromatography Accessories Trott, Germany) before
   AEX-MS analysis.

   The samples were analysed using a Dionex ionchromatography system
   (Integrion Thermo Fisher Scientific)^[252]72. Five µL of polar
   metabolite extract were injected in push partial mode, using an
   overfill factor of 1, onto a Dionex IonPac AS11-HC column (2 mM ×
   250 mM, 4 μm particle size, Thermo Fisher Scientific) equipped with a
   Dionex IonPac AG11-HC guard column (2 mM × 50 mM, 4 μm, Thermo Fisher
   Scientific). The column temperature was held at 30 °C, while the auto
   sampler was set to 6 °C. A potassium hydroxide gradient was generated
   using a potassium hydroxide cartridge (Eluent Generator, Thermo
   Scientific), which was supplied with deionized water (Millipore). The
   metabolite separation was carried at a flow rate of 380 µL/min,
   applying the following gradient conditions: 0–3 min, 10 mM KOH;
   3–12 min, 10–50 mM KOH; 12–19 min, 50–100 mM KOH; 19–22 min, 100 mM
   KOH, 22–23 min, 100–10 mM KOH. The column was re-equilibrated at 10 mM
   for 3 min.

   For the analysis of metabolic pool sizes the eluting compounds were
   detected in negative ion mode using full scan measurements in the mass
   range m/z 77–770 on a Q-Exactive HF high resolution MS (Thermo Fisher
   Scientific). The heated electrospray ionization (HESI) source settings
   of the mass spectrometer were: Spray voltage 3.2 kV, capillary
   temperature was set to 300 °C, sheath gas flow 50 AU, aux gas flow
   20 AU at a temperature of 330 °C and a sweep gas glow of 2 AU. The
   S-lens was set to a value of 60.

   The semi-targeted LC-MS data analysis was performed using the
   TraceFinder software (Version 4.1, Thermo Fisher Scientific). The
   identity of each compound was validated by authentic reference
   compounds, which were measured at the beginning and the end of the
   sequence. For data analysis the area of the deprotonated [M-H^+]^−1 or
   doubly deprotonated [M-2H]^−2 monoisotopologue mass peaks of every
   required compound were extracted and integrated using a mass accuracy
   <5 ppm and a retention time (RT) tolerance of <0.05 min as compared to
   the independently measured reference compounds. These areas were then
   normalized to the internal standards, which were added to the
   extraction buffer, followed by a normalization to the fresh weight of
   the analyzed sample.

Semi-targeted liquid chromatography-high-resolution mass spectrometry-based
(LC-HRS-MS) analysis of amine-containing metabolites

   The LC-HRMS analysis of amine-containing compounds was
   performed^[253]73 using 50 µL of the available 400 µL of the above
   mentioned (AEX-MS) polar phase which were mixed with 25 µl of 100 mM
   sodium carbonate (Sigma), followed by the addition of 25 µl 2% [v/v]
   benzoylchloride (Sigma) in acetonitrile (UPC/MS-grade, Biosove,
   Valkenswaard, Netherlands). Derivatized samples were thoroughly mixed
   and kept at 20 °C until analysis.

   For the LC-HRMS analysis, 1 µl of the derivatized sample was injected
   onto a 100 × 2.1 mM HSS T3 UPLC column (Waters). The flow rate was set
   to 400 µl/min using a binary buffer system consisting of buffer A
   (10 mM ammonium formate (Sigma), 0.15% [v/v] formic acid (Sigma) in
   UPC-MS-grade water (Biosove, Valkenswaard, Netherlands). Buffer B
   consisted of acetonitrile (IPC-MS grade, Biosove, Valkenswaard,
   Netherlands). The column temperature was set to 40 °C, while the LC
   gradient was: 0% B at 0 min, 0–15% B 0–4.1 min; 15–17% B 4.1–4.5 min;
   17–55% B 4.5–11 min; 55–70% B 11–11.5 min, 70–100% B 11.5–13 min; B
   100% 13–14 min; 100–0% B 14–14.1 min; 0% B 14.1–19 min; 0% B. The mass
   spectrometer (Q-Exactive Plus, Thermo Fisher Scientific) was operating
   in positive ionization mode recording the mass range m/z 100–1000. The
   heated ESI source settings of the mass spectrometer were: Spray voltage
   3.5 kV, capillary temperature 300 °C, sheath gas flow 60 AU, aux gas
   flow 20 AU at 330 °C and the sweep gas was set to 2 AU. The RF-lens was
   set to a value of 60.

   Semi-targeted data analysis for the samples was performed using the
   TraceFinder software (Version 4.1, Thermo Fisher Scientific). The
   identity of each compound was validated by authentic reference
   compounds, which were run before and after every sequence. Peak areas
   of [M + nBz + H]^+ ions were extracted using a mass accuracy (<5 ppm)
   and a retention time tolerance of <0.05 min. Areas of the cellular pool
   sizes were calculated according to the AEX-MS method.

ROS and TMRM staining

   To measure mitochondrial membrane potential and superoxide anion
   production, Tetramethylrhodamin-methylester-perchlorat, TMRM and
   MitoSOX^TM Red respectively were utilized to stain isolated hepatocytes
   at a concentration of 10μM and 3μM each, following the manufacturer’s
   protocol (Invitrogen). BD-Canto flow cytometer (Becton Dickinson) was
   used to record 10,000 gated events per sample, which were analyzed
   using the FACS-Diva software to derive the final result.

Software

   Data collection was done and analysis was preformed using softwares
   associated with the instruments used for the specific experiments, as
   described in the correcponding methods sections. In short; Western blot
   quantifications: ImageStudioLite Software (Version 5.2.5); Numerical
   data presentation and statistical analysis: Graph Pad Prism (Version
   10.0.3); Proteomics: Perseus software and R Studio (version 1.4.1106);
   RNA sequencing: Illumina’s bcl2fastq (v2.20.0), FastQC (v 0.11.9), STAR
   (v2.7.3a), R (version 4.1.0); Lipidomics: LipidView 1.3 beta (Sciex),
   LipidXplorer 1.2.8.1 ([254]https://zenodo.org/record/357046923,24)

Study design

   Sample size for each experiment was determined based on prior
   experience and established protocols within the field for that type of
   experiment. No data was excluded unless statistically calculated for
   the specific experiment (see Source data). At least three biological
   replicates were done for each experiment. The exact number of
   biological replicates for each experiment is given in the Figure
   legends. Experiments were performed by two different researchers to
   assure randomisation. Male and female mice were used for all
   experiments and with comparable results. Both mutant and control groups
   were processed at the same time and using the same reagents. No
   blinding was used.

Quantification and statistics

   Prism 9 for macOS software (GraphPad Software, San Diego, CA) was used
   to calculate the Standard Error mean (SEM) or Standard Deviation (SD)
   and assess the significance of differences between groups, unless
   otherwise specified. The tests employed were either unpaired
   non-parametric or paired non-parametric t-tests and ANOVA analysis, as
   appropriate. All replicates in the experiments presented are biological
   replicates unless stated otherwise in the figure legends.

Reporting summary

   Further information on research design is available in the [255]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [256]Supplementary Information^ (10.1MB, pdf)
   [257]Peer Review File^ (4.1MB, pdf)
   [258]41467_2024_51109_MOESM3_ESM.pdf^ (85.8KB, pdf)

   Description of Additional Supplementary Files
   [259]Supplementary Data 1^ (28.8KB, xlsx)
   [260]Supplementary Data 2^ (472.4KB, xlsx)
   [261]Reporting Summary^ (2.9MB, pdf)

Source data

   [262]Source Data^ (2.9MB, xlsx)

Acknowledgements