Xinqidi,Your biological research partner
Xinqidi Biotech Co.,Ltd,Wuhan,China 2008-2021
R&D 13th year

Coding and non-coding roles of MOCCI (C15ORF48) coordinate to regulate host inflammation and immunit

Issuing time:2021-04-12 11:01


Mito-SEPs are small open reading frame-encoded peptides that localize to the mitochondria to regulate metabolism. Motivated by an intriguing negative association between mito-SEPs and inflammation, here we screen for mito-SEPs that modify inflammatory outcomes and report a mito-SEP named “Modulator of cytochrome C oxidase during Inflammation” (MOCCI) that is upregulated during inflammation and infection to promote host-protective resolution. MOCCI, a paralog of the NDUFA4 subunit of cytochrome C oxidase (Complex IV), replaces NDUFA4 in Complex IV during inflammation to lower mitochondrial membrane potential and reduce ROS production, leading to cyto-protection and dampened immune response. The MOCCI transcript also generates miR-147b, which targets the NDUFA4 mRNA with similar immune dampening effects as MOCCI, but simultaneously enhances RIG-I/MDA-5-mediated viral immunity. Our work uncovers a dual-component pleiotropic regulation of host inflammation and immunity by MOCCI (C15ORF48) for safeguarding the host during infection and inflammation.


Activation of the endothelium by inflammatory triggers initiates a cascade of signaling pathways with the aim of danger signal clearance and resolution of the inflammatory response, ultimately returning the endothelium to its basal state1. Failure to do so results in excessive vascular inflammation that leads to tissue damage in acute settings, and cardiovascular diseases such as atherosclerosis and fibrosis in chronic settings. Although broad immunosuppressive strategies such as tumor necrosis factor-alpha (α), interleukin-1beta (IL-1β), or interleukin-6 (IL-6) blockade with neutralizing antibodies have yielded promising results in clinical trials to reduce vascular inflammation2, their considerable side effects and limited efficacy indicate a need for alternative, more specific targets, potentially downstream of these cytokines, and preferably specific to the endothelium.

Mitochondria are multifaceted organelles most well-known for their roles in cellular energy production via oxidative phosphorylation and in apoptosis via cytochrome c release. Besides energetic homeostasis, the role of mitochondria in regulating the onset and outcomes of inflammation is becoming increasingly appreciated3. A classic example is the production of reactive oxygen species (ROS) from the respiratory chain. Once thought to be purely pathogenic, it is now appreciated that low or physiological levels of ROS play normative roles in both pro-inflammatory and pro-resolving signaling pathways, while excessive ROS can lead to chronic activation of pro-inflammatory mechanisms that cause tissue damage4. The mitochondria are also known to be a proximal site for the activation of interferon response (IR) via the localization of MAVS5 or the activation of inflammasome via the extrusion of cardiolipin6. Leakage of mitochondrial DNA can also trigger IR or inflammatory cell death through the c-GAS/STING pathway7. Furthermore, metabolites of the Krebs cycle such as succinate and itaconate are increasingly recognized as modulators of inflammation; for example, through the regulation of the NLRP3 inflammasome and the IR pathway8,9. Although mitochondrial dysfunction is widely implicated in the pathogenesis of cardiovascular diseases, the specific function of mitochondria in endothelial health and dysfunction has received relatively little attention. In particular, how endothelial mitochondria respond to and modulate an inflammatory response is poorly understood.

Previously, our lab reported a preponderance of small open reading frame (sORF)-encoded peptides, defined as proteins smaller than 100 residues that localize to the mitochondria (mito-SEPs)10. These peptides have functions in diverse processes in the mitochondria, including electron transport, lipid metabolism, and calcium homeostasis. We asked the question of whether mito-SEPs might play hitherto unappreciated roles in regulating inflammation. For instance, Fitzgerald and colleagues recently identified a mitochondrial peptide Mm47 that is required for the activation of the NLRP3 inflammasome11.

In this study, we perform a proteogenomic screen to find mito-SEPs in primary human aortic endothelial cells (HAECs) that can promote the resolution of inflammation. We report the discovery of modulator of cytochrome C oxidase during Inflammation (MOCCI), a mito-SEP encoded by C15ORF48. In concert with miR-147b in the 3′ untranslated region (UTR) of C15ORF48, MOCCI replaces its paralog NDUFA4 in Complex IV (CIV). This dampens CIV activity and protects host tissue against excessive immune pathology by reducing cell death and cytokine production. Furthermore, miR-147b exerts potent antiviral effects by enhancing the IR through the RIG-I/MDA5 pathway. Altogether, the coding and non-coding functions of the C15ORF48 synergize to protect the host during infection, illustrating how small peptides coordinate with their encoding transcripts to achieve maximal biological impact.


Proteogenomic screen of inflammation-associated Mito-SEPS (i-Mito-SEPs) identifies MOCCI, encoded by C15ORF48

Mito-SEPs are peptides encoded from sORFs in the nuclear genome that are imported into the mitochondria for various facets of mitochondrial function. While characterizing mito-SEPS involved in metabolic regulation, we uncovered an unexpected negative association between mito-SEPS and inflammatory pathways such as interferon signaling in transcriptomes of human hearts with dilated cardiomyopathy12, a setting of chronic vascular inflammation (Fig. 1a). This motivated us to devise a strategy to uncover hitherto uncharacterized mito-SEPs that are involved in regulating inflammation, specifically in the vascular system. To this end, we adopted an unbiased proteogenomic strategy based on ribosome-profiling (Ribo-seq) coupled with RNA sequencing (RNA-seq) of primary HAECs treated with IL-1β (Fig. 1b). We chose 45 min, 12 and 24 h post-treatment to capture the full course of the inflammatory response including resolution, as indicated by the downward trend of activated pro-inflammatory markers NF-κB activity, ICAM-1 and VCAM-1 (Fig. 1b). We used HAECs from two healthy male donors in triplicate to avoid donor-specific effects and performed paired Ribo-seq with RNA-seq. Our Ribo-seq protocol generated 28–30 nucleotide ribosome protected fragments (RPFs) that displayed an average in-frame periodicity of 81.98% across annotated protein coding ORFs (Supplementary Fig. 1a, b). We then used RiboTaper13 to identify all expressed ORFs from our dataset that belong to known consensus coding sequence (CCDS) regions of the genome. These were filtered to include only ORFs encoding peptides less than 100 amino acids and exclude those overlapping with larger proteins (Fig. 1c). Only those that meet the quality control metrics based on the parameters of known SEPs were retained (see methods). Of these, we filtered for those with significant differential expression at the RNA level during IL-1β stimulation (Supplementary Data 1). Lastly, candidate genes and peptides were passed through our mitochondrial protein motif and gene signature-based mito-SEP prediction pipeline as described previously10 to identify mitochondrial SEPs (Fig. 1c, d). A candidate was assigned a mitochondrial gene signature if it correlated with MitoCarta genes in two or more of four tissues with pathological or physiological inflammation (Fig. 1d, Table 1, and Supplementary Data 1). We identified 21 putative inflammatory mito-SEPS (i-Mito-SEP) (Fig. 1c). Finally, we triaged each surviving candidate according to its degree of differential expression and whether it was positive for mitochondrial motif, gene signature, or both (Fig. 1e). The top i-Mito-SEP candidate was an 83-aa peptide encoded by the C15ORF48 gene, also called NMES-1, which showed the largest upregulation following IL-1β treatment and which was predicted to be mitochondrial by both prediction methods (Fig. 1e). We also confirmed that Ribo-seq reads were increased upon IL-1β treatment, indicating that there is more translation during inflammation (Fig. 1f). In-frame P-site reads were also detected across the entire ORF (Fig. 1g). This allowed us to discriminate between alternate translation initiation sites and accurately determine the start site of the SEP.

Fig. 1: Proteogenomic screen in human endothelial cells (HAECs) identifies MOCCI as an inflammatory Mito-SEP (i-Mito-SEP).

a WGCNA-GSEA analysis of mito versus non-mito-SEPs in human failing heart reveals strong anti-correlation of mito-SEPs with inflammation. Each row represents one pathway and the pathways are separated into those that are involved in metabolism and those that are involved in inflammation. Color scale = NES score. Dataset used: European Genome-Phenome Archive EGAS0000100245412. bHAECs from two healthy male donors were treated with 1 ng/mL IL-1β for 45 min, 12 and 24 h. All timepoints including untreated controls were analyzed by RNA-seq and Ribo-seq in triplicate. NF-kb reporter activity, ICAM-1, and VCAM-1 expression illustrate the dynamics of the HAEC response: an increase in inflammation up to 12 h followed by resolution thereafter. Data for NF-kb reporter activity are presented as mean ±  SEM. n = 4 biological replicates for NF-kb reporter activity. cWorkflow to identify i-Mito-SEPs. Blue numbers indicate the number of SEPs called by RiboTaper and black numbers indicate the number of genes encoding those SEPs (see methods). d Summary of the mitochondrial gene signature workflow. (Bottom) UMAP of gene module association determination (G-MAD) scores of MitoCarta and random genes in human colon, skin, skeletal muscle (SKM), and PBMCs undergoing inflammation. Color scale indicates the percentage of mitochondrial genes in each cluster. Based on their G-MAD, the 240 candidate genes from c were assigned to a cluster in each dataset. (Top) Heatmap depicting the percentage of mitochondrial genes in a candidate’s cluster in each of the four datasets. Each column represents one candidate. Top scoring candidates were returned to the pipeline in c for further consideration. Datasets used were: GSE1122384, GSE1490583, GSE12086282, and GSE982081. e Volcano plot of the maximal fold change of the 240 candidate genes during the IL-1β treatment compared to untreated cells. The size of each dot represent the highest TPM reached at any timepoint, while its color indicates if candidate SEP was predicted to localize to the mitochondria by motif, G-MAD or both. The padjusted values were attained by applying the Wald chi-squared test, followed by correction for multiple testing using the Benjamini and Hochberg method through the DESeq2 pipeline. f MOCCItranscript and translation levels increase by 1000-fold after IL-1β treatment. TE translational efficiency. g Ribo-seq derived p-site read coverage across the MOCCI locus. In-frame (0) and out-of-frame (+1,+2) colored in orange, green, and blue respectively. Inset on the left shows that p-site reads are detected only at the start codon. Inset on the right summarizes the total number of in-frame (0) and out-of frame (+1,+2) reads in dataset.

Table 1 List of published datasets that were used.

Gene module association determination (G-MAD) uses a large cohort of expression data to predict function of a gene based on its expression pattern with other genes. In human tissues, C15ORF48 expression is correlated to genes involved in the regulation of cytokine activity, in particular IL-10; response to microbial infection, interferon signaling, and leukocyte chemotaxis (Supplementary Fig. 1c). Notably, in some tissues such as the artery, C15ORF48 is negatively associated with genes involved in cellular and viral mRNA translation (Supplementary Fig. 1c). These results suggest that C15ORF48 might play a role in regulating the innate immune response against pathogens such as bacteria and viruses. Indeed, single-cell RNA-seq (scRNA-seq) data of human peripheral blood mononuclear cells (PBMCs) show that C15ORF48 mRNA is highly expressed in monocytes at basal and interferon-stimulated state (Supplementary Fig. 2a). In non-immune cells however, C15ORF48 expression is low in healthy tissues, and upregulated in endothelial cells in conditions of chronic inflammation such as atherosclerosis (Supplementary Fig. 2b). Similarly, during influenza infection in mice, C15ORF48 is upregulated in granulocytes and monocyte precursors as well as lymphatic endothelial cells (Supplementary Fig. 2c). These data together point to a constitutive role of C15ORF48 in monocytes and an inducible function in endothelial and phagocytes cells during both acute pathogen-induced and chronic sterile inflammation.

MOCCI is a paralog of NDUFA4, 14th subunit of Complex IV

Consistent with our finding that C15ORF48 encodes a putative mito-SEP, G-MAD shows strong and positive association across all tissues between C15ORF48 and genes encoding respiratory chain CIV (red), in addition to inflammation (Supplementary Fig. 1c). The SEP encoded by C15ORF48—which we refer to hereafter as MOCCI–has previously been implicated in a large interactome study to interact with components of the electron transport chain (ETC), where it was overexpressed as a FLAG-tagged bait to capture interactors, although no definitive function was assigned to the protein14. MOCCI is conserved evolutionarily across vertebrates (Fig. 2a) but not present in fungi. MOCCI contains a sole B12D domain (NADH-ubiquinone reductase complex 1 MLRQ subunit) in an architecture that is found also in NDUFA4 and NDUFA4L2 (Fig. 2a, b and Supplementary Fig. 2d). Indeed, MOCCI shares a strong conservation at the amino acid level with NDUFA4/NDUFA4L2 proteins, which are also SEPs by definition (Supplementary Fig. 2d). NDUFA4 was originally assigned to be an accessory unit of Complex I (CI) (NADH:ubiquinone oxidoreductase) but was later found to be the 14th subunit of CIV15,16,17, which we confirmed by co-migration between NDUFA4 and COX4 using blue native polyacrylamide gel electrophoresis (BN-PAGE) of digitonin-solubilized mouse heart mitochondria (Supplementary Fig. 2e). To elucidate the function of MOCCI and verify the observed association with CIV, we first determined its subcellular localization in MOCCI-overexpressing HEK293T cells where MOCCI is detected by western blot in mitochondria-enriched fractions (Fig. 2c). Differential extraction assays of isolated mitochondria indicate that MOCCI is localized to the inner mitochondrial membrane (IMM) (Fig. 2d), consistent with the presence of a predicted transmembrane domain, and confirmed by Proteinase K (PK) protection assays (Fig. 2e). To verify that MOCCI is part of the ETC, specifically CIV, we overexpressed the ORF encoding mouse MOCCIFLAG by AAV9-mediated (AAV-MOCCI) delivery to the heart, an endothelial rich organ, and performed BN-PAGE analysis of isolated tissue mitochondria (Fig. 2f). Western blotting with anti-MOCCI antibody demonstrated that the MOCCIFLAG co-migrated with MTCO-1, the catalytic subunit of CIV (Fig. 2f). This was confirmed using an anti-FLAG antibody (Supplementary Fig. 2f). Furthermore, using two-dimensional BN-SDS PAGE, we confirmed the complete colocalization of MOCCI with MTCO-1, with the highest occupancy in CIV monomers and dimers, with some comigrating in supercomplexes (Fig. 2g).

Fig. 2: MOCCI is a subunit of Complex IV (cytochrome C oxidase).

a The MOCCI ORF is conserved across vertebrates and contains a predicted transmembrane (TM) domain and the B12D domain found only in NADH-ubiquinone reductase complex 1 MLRQ subunits. The MOCCI transcript is also the host gene for hsa-miR-147b (right), located in the 3′UTR of the transcript. The letters in blue are the sequence of the mature miRNA. b MOCCI is a paralog of NDUFA4 which also contains both the B12D and TM domains. The 3′UTR of the NDUFA4 transcript contains top target site of miR-147b-3p as predicted by TargetScan114, miRDB115, and miRBase116. It is not clear if the binding site in chicken Ndufa4 is functional, as it is not as well conserved. c MOCCI was found in mitochondria-enriched fractions of HEK293T transfected with MOCCI ORF-expressing plasmid (MOCCI-HEK293T). Values given indicates percentage of the total fraction loaded onto the gel. n = 2 biological replicates. Source data are provided as a Source Data file. d Differential solubilization of MOCCI-HEK293T mitochondria indicated that MOCCI was located in the inner membrane (IMM). TOMM70 and MTCO-1 serve as control for outer (OMM) and inner (IMM) mitochondrial membrane proteins, respectively. n = 1 biological replicate. Source data are provided as a Source Data file. e Proteinase K protection assay of MOCCI-HEK293T mitochondria further confirmed IMM localization of MOCCI. n = 1 biological replicate. Source data are provided as a Source Data file. f Blue native PAGE (BN-PAGE) of mouse heart mitochondria with AAV9-mediated MOCCI-FLAG expression (AAV-MOCCI). Probing with α-MOCCI and α-MTCO-1 antibodies, the core enzymatic subunit of CIV, reveal co-migration of MOCCI signal with CIV. AAV-GFP was used as a negative control. Coomassie blue stain of the gel is shown on the right. n = 4 biological replicates. gTwo-dimensional BN-SDS-PAGE to confirm co-migration of MOCCI and CIV. The lanes correspond to the bands annotated on the Coomassie blue-stained gel in (f). n = 2 biological replicates. Source data are provided as a Source Data file.

CIV Composition Switches from NDUFA4-to-MOCCI during Inflammation

In addition to the MOCCI ORF, the 3′UTR of C15ORF48 hosts hsa-miR-147b (Fig. 2a), an evolutionarily conserved miRNA that has previously been implicated to target NDUFA4mRNA via a conserved seed sequence in its 3′UTR (Fig. 2b). In HEK293T cells, transfection of miR-147b mimic downregulated NDUFA4 mRNA and NDUFA4 protein (Supplementary Fig. 3a, b). These data suggest that upregulation of the C15ORF48 mRNA has two outcomes: incorporation of MOCCI into CIV and miR-147b-mediated downregulation of the MOCCI paralog NDUFA4, essentially effecting a subunit switch in CIV. Since translation of an mRNA versus its processing into a pre-miRNA are mutually exclusive events, we first investigated if both processes were occurring in HAECs. In line with the HAEC Ribo-seq data, the MOCCI peptide is translated in about 25.7% of HAECs in response to IL-1 β (Fig. 3a and Supplementary Fig. 3c), where it shows clear localization to the mitochondria as judged by co-localization with TOMM20 (Fig. 3b). At the same time, IL-1β treatment in HAECs robustly induced the biogenesis of mature miR-147b (Fig. 3c). This suggests that upon IL-1β stimulation, HAECs generate both MOCCI and miR-147b, and together these might function synergistically to enforce the switch in CIV from NDUFA4-to- MOCCI. Importantly, HAECs demonstrated NDUFA4-to-MOCCI switch mediated by upregulation of endogenous C15ORF48 as visualized by intracellular flow cytometry in response to IL-1β (Fig. 3d). We validated this NDUFA4-to-MOCCI switch in A549 lung epithelial cells (Fig. 3e), which like HAECs upregulate both C15ORF48 transcript and miR-147b in response to IL-1β treatment (Supplementary Fig. 3d). Conversely, CRISPR/Cas9-mediated disruption of the MOCCI ORF within C15ORF48 (“MOCCI-KO”, Supplementary Fig. 3e) reduced MOCCI upregulation (Supplementary Fig. 3f) and NDUFA4 downregulation (Supplementary Fig. 3g) in HAECs treated with IL-1β. However, miR-147b production was also disrupted in the KO cells, despite normal levels of mutant C15ORF48 mRNA (Supplementary Fig. 3h). We were therefore unable to separate the effects of MOCCI and miR-147b using MOCCI-KO HAECs. To further clarify the relative contributions of the MOCCI peptide and miR-147b to the observed NDUFA4-to-MOCCI switch, we stably expressed the ORF of human MOCCI (hereafter referred to as “MOCCI”) and full length C15ORF48 mRNA (which produces both MOCCI and miR-147b and is hereafter referred to as “WT-mRNA”) via lentiviral transduction. Consistent with the findings from MOCCI-KO HAECs, we found that MOCCI alone was sufficient to lower NDUFA4 protein levels by 61.2%, while WT-mRNA reduced NDUFA4 protein levels by 80.6%, suggesting that MOCCI is equally, if not more effective than miR-147b at NDUFA4 attenuation (Fig. 3f and Supplementary Fig. 3i). Indeed, mutating the start codon of the MOCCI ORF within C15ORF48 mRNA to isoleucine (referred to hereafter as “ATGmut-mRNA”), while preserving miR-147b biogenesis (Fig. 3g), led to only 31.5% downregulation of NDUFA4 protein (Fig. 3f and Supplementary Fig. 3i). This was also validated by western blot, which showed NDUFA4 protein levels to be much lower in MOCCI and WT-mRNA HAECs while ATGmut HAECs showed only intermediate downregulation (Fig. 3h). While miR-147b from WT and ATGmut-mRNA expression reduced NDUFA4 mRNA by about 50%, MOCCI did not affect NDUFA4 mRNA levels (Supplementary Fig. 3j). These results together argue that MOCCI alone is sufficient to downregulate NDUFA4 protein level without affecting its mRNA levels. MiR-147b contributes by attenuating NDUFA4 mRNA and potentially synergizes with MOCCI to reduce NDUFA4 protein. Altogether, MOCCI and miR-147b exert different levels of control on NDUFA4 protein and mRNA levels (Supplementary Fig. 3k).

Fig. 3: MOCCI and miR-147b reduce NDUFA4 expression during inflammation.

a Intracellular flow cytometry staining for MOCCI in untreated (gray line) and IL-1β-treated (red line) human aortic endothelial cells (HAECs). Percentage of MOCCI-positive HAECs is 25.7 ± 5.2% (mean ± se), n = 4 biological replicates per condition. b Co-immunofluorescent staining for MOCCI and TOMM20, a mitochondrial protein in HAECs. Scale bar = 100 µm, n = 3 biological replicates. cC15ORF48 mRNA and miR-147b expression in untreated and IL-1β-treated HAECs. Data are presented as mean ±  SEM, n = 3 biological replicates per condition. d Intracellular flow cytometry staining for NDUFA4 and MOCCI in untreated or IL-1β-treated HAECs. e Intracellular flow cytometry staining for NDUFA4 and MOCCI in untreated or IL-1β-treated A549 cells. f Intracellular flow cytometry staining for NDUFA4 and MOCCI in HAECs overexpressing MOCCI, C15ORF48 WT-mRNA (both MOCCI and miR-147b), and C15ORF48 ATGmut-mRNA (only miR-147b), n = 3 biological replicates per condition. g Levels of C15ORF48 mRNA (primers binding to the 5′UTR of C15ORF48) or miR-147b by qPCR in HAECs with the indicated overexpression. Data were presented as mean ±   SEM, n = 3 biological replicates per condition. h Western blot of NDUFA4 in HAECs with the indicated overexpression. n = 3 biological replicates. Source data are provided as a Source Data file.

To confirm these in vitro observations, we analyzed the proteome of mitochondria isolated from AAV-MOCCI and AAV-GFP injected mice by quantitative mass spectrometry, and found that MOCCI was sufficient to downregulate NDUFA4 protein levels in vivo (Fig. 4a and Supplementary Data 2). We confirmed by PAGE analysis that AAV-MOCCI mitochondria had a 60% reduction in both total (SDS-PAGE) and CIV-associated (BN- and BN-SDS PAGE) NDUFA4 (Fig. 4b, c and Supplementary Fig. 4a). Levels of CIV (detected by COX4) and other complexes were unaltered (Fig. 4a, b and Supplementary Fig. 4b). These results confirm that MOCCI alone is sufficient to downregulate NDUFA4 protein levels, potentially by excluding NDUFA4 from CIV. miR-147b synergizes with MOCCI to mediate the switch by targeting NDUFA4 mRNA. Since the predicted structure of MOCCI and that of NDUFA4 are superimposable (Fig. 4d), we postulate that they occupy the same position in the CIV structure i.e., at the surface of the complex adjacent to MTCO-1 and in close proximity to the heme-copper binuclear catalytic core (Fig. 4e). To test if MOCCI and NDUFA4 are mutually exclusive in CIV complexes, we immunoprecipitated MOCCI-FLAG under native non-denaturing conditions in both total (Fig. 4f) and sucrose-gradient purified CIV monomeric complexes (Supplementary Fig. 4c) in AAV-MOCCI and AAV-GFP heart mitochondria. Pulldown of MOCCI, which enriched MTCO-1 by 231% failed to coimmunoprecipitate NDUFA4 (Fig. 4f, Supplementary Fig. 4c). These data imply that MOCCI and NDUFA4 are not present in the same CIV complexes and support a model in which NDUFA4 and MOCCI are incorporated into alternate assemblies of CIV, or CIV isozymes, raising the question of the functional significance of MOCCI versus NDUFA4-containing CIV complexes.

Fig. 4: NDUFA4 is exchanged for MOCCI in Complex IV during inflammation.

a Volcano plot of proteins detected by quantitative TMT-mass spec in AAV-GFP or AAV-MOCCI mouse heart mitochondria. p = two-tailed Student’s t test, n = 3 mice per AAV. b BN-PAGE (top) and SDS-PAGE (bottom) of AAV-MOCCI/GFP mouse heart mitochondria probed for the indicated respiratory chain proteins. Each lane represents mitochondria from one mouse. Quantification of NDUFA4 shown in Supplementary Fig. 4a. c Two-dimensional BN-SDS-PAGE to show downregulation of NDUFA4 in CIV complexes. n = 3 biological replicates. Source data are provided as a Source Data file. d Alignment of the predicted secondary structure of MOCCI (iTASSER model) and known structure of NDUFA4 in Maestro (Schrodinger). e The position of NDUFA4 (ribbon) in CIV cryo-EM structure (PDB 5Z62). All other subunits are depicted in tube representation. MTCO-1 (purple) and the redox centers (HEME-A and Cu2+) are highlighted to illustrate the relative position of NDUFA4. f Co-IP of NDUFA4 and MOCCI. Digitonin (1%)-solubilized isolated mouse heart mitochondria were immunoprecipitated with anti-FLAG and probed for the proteins as indicated. AAV-GFP served as negative control. Source data are provided as a Source Data file.

MOCCI reduces CIV activity, mitochondrial membrane potential, and ROS production

We initially sought to define the molecular function of MOCCI in THP-1 monocytes and their macrophage derivatives, which are most commonly employed to investigate mechanisms of inflammatory regulation. However, despite robust upregulation of the C15ORF48 transcript and mature miR-147b in M1 macrophages treated with IL-1β (Supplementary Fig. 5a, b), the MOCCI peptide is not translated in THP-1 monocytes or THP-1-derived M1 macrophages as detected by flow cytometry (Supplementary Fig. 5c) and western blotting (Supplementary Fig. 5d). Therefore, in vitro cultured human monocytes and macrophages produce only miR-147b from the C15ORF48 locus. We therefore carried out all subsequent experiments in HAECs and A549 cells. First, we examined the effect of the NDUFA4-to-MOCCI switch on CIV activity, which can be modulated by the incorporation of paralogous subunits (see Discussion). Since MOCCI alone is sufficient to recapitulate the switch, we measured oxygen consumption and electron flow through CIV using TMPD as an electron donor with heart mitochondria isolated from AAV-MOCCI mice. This experiment utilizes oxygen consumption by CIV as a readout of flux through CIV while CIII is inhibited by Antimycin A, uncoupled from ATP production by FCCP (Fig. 5a). AAV-MOCCI mitochondria had significantly lower CIV electron flux compared to control mitochondria (Fig. 5b), without a significant change in overall uncoupled rotenone-sensitive flux through CI (Supplementary Fig. 5e) or coupled respiration (Supplementary Fig. 5f). These results suggest that MOCCI incorporation lowers CIV activity without affecting respiratory output, which can be attributed to the known large excess capacity and low flux control coefficient of CIV in both isolated mitochondria18 and intact tissues19. Using permeabilized HAECs and A549, we found that MOCCI overexpression led to a small reduction in CIV activity (Fig. 5c). This again did not result in a reduction in overall coupled respiration in response to pyruvate (Supplementary Fig. 5g) in permeabilized cells and did not impact basal or maximal respiration in intact HAECs and A549 (Supplementary Fig. 5h), indicating that MOCCI expression does not compromise respiratory capacity in spite of the reduction in CIV activity. The attenuated CIV activity in MOCCI A549 cells was confirmed using spectrophotometric assays (RCA), which measures CIV enzymatic activity by kinetic measurements of azide-sensitive cytochrome c reduction20 (Fig. 5d). The reduction in CIV activity was accompanied by a reduction in membrane potential (∆Ψm), as measured using tetramethylrhodamine ethyl ester (TMRE)21 and ratiometric JC-1 in both HAECs and A549 (Fig. 5e). Lowering membrane potential is predicted to lower ROS production22,23. Indeed, we found that HAECs and A549 expressing MOCCI had lower total ROS as measured by 2′,7′-dichlorodihydrofluorescein (DCF) (Fig. 5f) and Amplex Red (Fig. 5g), and mitochondrial ROS as measured by MitoSOX™ Red (Supplementary Fig. 5i) compared to WT cells.

Fig. 5: MOCCI reduces CIV activity, membrane potential, and ROS production.

a (Top) Schematic uncoupled electron flow assay on Agilent’s Seahorse platform to measure CIV activity in isolated mitochondria using TMPD/ascorbate as electron donor (see Methods). (Bottom) Seahorse uncoupled electron flow analysis of AAV-MOCCI and AAV-GFP isolated mouse heart mitochondria. Oxygen consumption rate (OCR) was normalized with citric synthase (CS) activity. Dotted line shows the compounds injected at each time point. CIV activity = maximal activity after TMPD addition − minimal activity after azide injection. AA antimycin A, TMPD N,N,N′,N′-Tetramethyl-p-phenylenediamine dihydrochloride. Data were presented as mean and SEM of 3 biological replicates. b CIV activity of AAV-MOCCI and AAV-GFP isolated mouse heart mitochondria as measured by uncoupled electron flow assay on Seahorse shown in (a). Each column represents one mouse, and each dot represents one technical replicate. Data were presented as mean ±  SEM; P = two-tailed unpaired Student’s t test; n = 3 biological replicates with 6 technical replicates each. cCIV activity of digitonin (0.0025%) permeabilized MOCCI and control cells as measured by Seahorse. Data were presented as mean ±  SEM; P = two-tailed unpaired Student’s t test; n = 6 technical replicates each. d CIV activity of MOCCI and control A549 cells as measured by respiratory chain enzyme assay (RCA). Data were presented as mean ±  SEM; P = two-tailed unpaired Student’s ttest; n = 8 biological replicates each. e Membrane potential of MOCCI and control cells as measured by flow cytometry of TMRE or JC-1 dye incorporation. Data are presented as mean ±  SEM; P = one-way ANOVA; n = 4 biological replicates per condition from two repeats. f Total cellular ROS levels of MOCCI and control cells as measured by flow cytometry of DCF dye reaction. Data were presented as mean ±  SEM; P = one-way ANOVA; n = 3 biological replicates per condition. g Total cellular ROS levels of MOCCI and control cells as measured by Amplex Red assay. Data were presented as mean ±  SEM; P = one-way ANOVA; n = 4 biological replicates per condition.

miR-147b reduces CIV activity but not membrane potential and ROS production

Depletion of NDUFA4 has been previously reported to reduce CIV activity15 and impair respiration. Since miR-147b targets NDUFA4, we asked if miR-147b can recapitulate the effects of MOCCI. Transfection of a miR-147b mimic at similar levels as WT-mRNA cells (Supplementary Fig. 6a) reduced CIV activity to intermediate levels (Fig. 6a, b) without impairing respiration (Supplementary Fig. 6b). In contrast, siRNA-mediated knockdown of NDUFA4 significantly lowered CIV activity to an extent that affected basal respiration in intact cells (Fig. 6a, b and Supplementary Fig. 6b). Similarly, ATGmut-mRNA lowered CIV activity (Fig. 6c) without affecting respiration (Supplementary Fig. 6c). Hence both MOCCI and miR-147b tonically lower CIV activity without affecting the output of the respiratory chain, while complete NDUFA4 removal without replacement by MOCCI (i.e., siNDUFA4) reduces respiration (Supplementary Fig. 6b). Surprisingly, miR-147b did not reduce membrane potential or ROS production in HAECs and A549 (Supplementary Fig. 6d). In line with this, neither WT-mRNA nor ATGmut-mRNA HAECs or A549 had reduced membrane potential (Fig. 6d) or ROS production (Fig. 6e, f). Rather, expression of WT-mRNA—which produces both MOCCI and miR-147b—completely reversed the drop in membrane potential and ROS production caused by MOCCI (Fig. 6d, f). However, this antagonism is not due to miR-147b within WT and ATGmut-mRNA, since miR-147b transfection did not revert the membrane potential of MOCCI cells back to WT levels (Supplementary Fig. 6d) or increase ROS (Supplementary Fig. 6e), raising the possibility that C15ORF48 transcript might have additional non-coding roles to regulate mitochondrial membrane potential. Of note, knockdown of NDUFA4 did not decrease mitochondrial membrane potential either, suggesting that the incorporation of MOCCI specifically reduces membrane potential and ROS production rather than the removal of NDUFA4 (Supplementary Fig. 6d). These results illustrate a complex scenario with the following implications. First, the reduction in membrane potential and ROS production in MOCCI HAECs and A549 cells is not simply due to impaired mitochondrial integrity owing to protein overexpression, as WT-mRNA drives MOCCI expression to similar levels compared to MOCCI ORF (Fig. 3f). Secondly, the reductions in membrane potential and ROS generation in MOCCI cells may not be directly caused by reduction of CIV activity, since miR-147b reduce CIV activity without reducing membrane potential and ROS. Lastly, these results indicate that MOCCI incorporation is not functionally equivalent to removal of NDUFA4 via miR-147b or siNDUFA4, as summarized in Fig. 6g. This provides a conceptual explanation for the presence of miR-147b in the 3′UTR of C15ORF48, even though MOCCI alone is sufficient to mediate the NDUFA4-to-MOCCI switch at the protein level. While downregulation of NDUFA4 mRNA by miR-147b might serve to reinforce the switch, it is likely to have additional roles that MOCCI cannot fulfill. We next sought to define the functional effects of MOCCI and miR-147b expression on the inflammatory response.

Fig. 6: miR-147b reduces CIV activity but not membrane potential and ROS production.

a CIV activity of permeabilized A549 cells transfected with miR-control, miR-147b mimic, or siNDUFA4 as measured by Seahorse. Data were presented as mean ± SEM; P = one-way ANOVA; n = 8 technical replicates per condition. b Seahorse uncoupled electron flow analysis of CIV activity in permeabilized A549 cells transfected with miR-control, miR-147b mimic, or siNDUFA4, as shown in (a). Oxygen consumption rate (OCR) reads were normalized with protein levels quantified with BCA. Dotted line shows the compounds injected at each time point. CIV activity was measured as activity between the maximal activity after TMPD addition and the minimal activity after azide injection. AA antimycin A, TMPD N,N,N′,N′-Tetramethyl-p-phenylenediamine dihydrochloride, Oligo oligomycin. Data were presented as mean and SEM of eight biological replicates. c CIV activity of permeabilized WT or ATGmut-mRNA overexpressing- HAECs and A549 cells, as measured by Seahorse. Data were presented as mean ± SEM; P = two-tailed unpaired Student’s t test; n = 8 technical replicates per condition. d Membrane potential of HAECs and A549 cells with indicated transgenes as measured by flow cytometry of TMRE dye incorporation. Data were presented as mean ±  SEM; P = one-way ANOVA; n = 4 biological replicates per condition from two repeats. e Total cellular ROS levels of A549 cells with indicated transgenes as measured by Amplex Red assay. Data were presented as mean ± SEM; P = one-way ANOVA; n = 4 biological replicates per condition. Data for control and MOCCI were shown in Fig. 5g. f Total cellular ROS levels of HAECs and A549 cells with indicated transgene as measured by flow cytometry of DCF dye reaction. Data were presented as mean ± SEM; P = one-way ANOVA; n = 3 biological replicates per condition. g Effects of MOCCI, miR-147b mimic, siNDUFA4, and ATGmut-mRNA on CIV activity, membrane potential (MP), and ROS production.

MOCCI and miR-147b reduce pro-inflammatory cytokine production

To address this question, we first examined the effects of a NDUFA4-to-MOCCI switch mediated by MOCCI alone. C15ORF48 expression in human cells is upregulated in acute bacterial and viral infections (Supplementary Fig. 7a and Table 1), suggesting a role during an immune response. Even though ECs are not leukocytes, they constitute the first line of defense against bloodborne pathogens, and are prone to excessive inflammation triggered by infections. By clustering genes that have the same expression pattern into modules (see Methods), our IL-1β stimulated HAEC RNA-seq data show that C15ORF48 is co-expressed with genes in two modules (Modules 2 and 3 in Fig. 7a). These modules are enriched for negative regulation of cytokine production and activated T cell proliferation, and IL-10 production, which represent anti-inflammatory strategies aimed at host protection via limiting excessive immune activation during an infection. We, therefore examined the effects of MOCCI on cytokine production vis-à-vis WT and ATGmut-mRNA in HAECs following a viral challenge. We used the flaviviruses Dengue (DENV) and Zika (ZIKV) due to their ability to infect and replicate in ECs, and induce C15ORF48 expression (Fig. 7b). HAECs primarily secrete IL-6, IL-8, MCP-1 (CCL2), CXCL10, and Arginase when challenged with DENV/ZIKV. Levels of interferons and other common pro-inflammatory cytokines were undetectable (Supplementary Fig. 7b). MOCCI and WT/ ATGmut-mRNA significantly suppressed IL-6 and MCP-1 secretion (Fig. 7c, d), while MOCCI-KO HAECs showed the converse phenotype (Fig. 7e). Transfection of miR-147b mimic and siNDUFA4 also reduced MCP-1 secretion, confirming that the effects observed in WT and ATGmut-mRNA can be attributed to miR-147b (Supplementary Fig. 7c). These results together indicate that MOCCI and miR-147b exert a suppressive effect on pro-inflammatory cytokine production in HAECs. This effect was not limited to viral-induced cytokine production. Rather, we found that MOCCI, WT, and ATGmut-mRNA were also sufficient to lower the secretion of IL-6, MCP-1, and even IL-8 upon challenge with bacterial endotoxin LPS, which was more potent in inducing cytokine secretion in HAECs compared to DENV/ZIKV infections (Fig. 7f). Hence, we conclude that in cultured HAECs and A549 cells, both MOCCI and miR-147b are able to reduce MCP-1 and IL-6 production at basal and when challenged with inflammatory stimuli.

Fig. 7: MOCCI and miR-147b reduce MCP-1 and IL-6 secretion during viral infection.

a (Left) Differentially expressed genes in the HAECs treated with IL-1β across the timepoint series (Fig. 1b) were clustered into co-expression modules by dynamic tree cut. The log-fold change is relative to the untreated cells. C15ORF48 resides in Module 2. (Right) Panther GO analysis of genes in Module 2 and closely related Module 3 reveal significant enrichments in the ontologies depicted. Size of dot indicates the number of members in each GO annotation and the color indicates the pvalue of enrichment. b C15ORF48 mRNA levels in HAECs following DENV and ZIKV infection. Data were presented as mean ± SEM; n = 3 biological replicates per condition. c Levels of MCP-1 (CCL2) secreted by HAECs with indicated transgenes following DENV/ZIKV infection. Data were presented as mean ± SEM; P = two-way ANOVA; n = 6 biological replicates per condition. d Levels of IL-6 secreted by HAECs with indicated transgenes following DENV/ZIKV infection. Data were presented as mean ± SEM; P = two-way ANOVA; n = 6 biological replicates per condition. e Levels of MCP-1 (CCL2) secreted by WT and MOCCI-KO HAECs at basal and following DENV/ZIKV infection. Data were presented as mean ± SEM; P = two-way ANOVA; n = 3 biological replicates per condition. fLevels of MCP-1 (CCL2), IL-6, and IL-8 secreted by HAECs with indicated transgenes at basal and following LPS treatment after 24 h. Data were presented as mean ± SEM; P = two-way ANOVA; n = 3 biological replicates per condition.

Since both MOCCI and miR-147b lower CIV activity, our findings suggest that CIV attenuation is sufficient to suppress the secretion of MCP-1 and IL-6. This is consistent with recent findings that inhibition of CIV in human PBMCs with potassium cyanide results in widespread suppression of pro-inflammatory cytokine production24. We confirmed these findings in HAECs using potassium azide as a CIV inhibitor, which significantly reduced MCP-1 and IL-6 secretion without any cytotoxic effects during a 24 h incubation period (Supplementary Fig. 7d). Hence, CIV inhibition or attenuation suppresses pro-inflammatory cytokine production. In the case of MOCCI HAECs, reduction of ROS production may also contribute to attenuation of MCP-1 and IL-6 production. Treatment of HAECs with anti-oxidant N-acetylcysteine (NAC), which effectively lowers ROS in HAECs (Supplementary Fig. 7e), significantly reduced IL-6 and MCP-1 production at basal and following LPS exposure to levels comparable with that of MOCCI HAECs (Supplementary Fig. 7f). However, since miR-147b, WT, and ATGmut-mRNA suppress cytokine production without reducing membrane potential and ROS, this is unlikely to be the sole mechanism with which CIV attenuation reduces MCP-1 and IL-6 production.

MOCCI and miR-147b regulate the IR via RIG-I/MDA5

The co-expression of C15ORF48 with genes associated with type I interferon (Fig. 7a) prompted us to examine the effects of C15ORF48 expression on antiviral immunity. Intriguingly, MOCCI has a viral homolog in poxviruses25, suggesting that it has been evolutionarily selected and co-opted by viruses for immune evasion via modulation of the IR. The canonical pathway of type I IR is the IFN-stimulated gene factor 3 (ISGF3) pathway. ISGF3 is induced by the IFN-regulatory factor IRF9, leading to STAT1/STAT2-mediated activation of genes containing IFN-stimulated response elements (ISREs), also known as interferon-stimulated genes (ISGs)26. In line with the observed anti-inflammatory effects of MOCCI, RNA-seq analysis of HAECs 48 h postinfection showed that MOCCI HAECs suppressed the upregulation of DENV and ZIKV infection-induced genes, which are highly enriched for ISGs (Fig. 8a and Supplementary Fig. 8a). Enrichment analyses of differentially regulated ISGs suggest that STAT1/STAT2/IRF1/IRF8/ NF-κB target genes are tonically repressed by MOCCI, as well as genes that are regulated by the newly discovered anti-inflammatory transcription factor ETV7 and microRNA miR-146a (Supplementary Fig. 8a)27,28. These data suggest that MOCCI suppresses the IR following viral infection. On the contrary, WT-mRNA and ATGmut-mRNA showed the opposite trend of enhancing IR (Fig. 8a). As the IR is responsible for establishing an antiviral state, we tested if these observed trends resulted in functional consequences for viral replication, as measured by fold change of viral genome (vg) copy number 48 h postinfection. Consistent with the observed trends in IR and ISGs, MOCCI HAECs had a higher vg copy number while C15ORF48WT-mRNA HAECs and ATGmut-mRNA had lower vg copy number 48 h postinfection, with ATGmut-mRNA-mediated suppression reaching statistical significance (Fig. 8b). Importantly, we did not detect significant differences in vg copy number 2 h postinfection (Supplementary Fig. 8b), arguing that viral entry was not significantly affected. A similar trend was observed in A549, where MOCCI enhanced and ATGmut-mRNA suppressed viral copy number, with WT-mRNA showing an intermediate phenotype (Fig. 8c). Viral replication results correlated well with the levels of IFN-β mRNA in A549 cells, a measure of the magnitude of IR, which were repressed by MOCCI and enhanced by WT-mRNA (Supplementary Fig. 8c). Other ISGs were significantly enhanced by ATGmut-mRNA (Supplementary Fig. 8d). These data suggest that miR-147b exerts a viral-static effect that keeps viral replication in check by stimulating the IR. Indeed, transfection of miR-147b was sufficient to inhibit ZIKV replication (Fig. 8d and Supplementary Fig. 8e). Importantly, the antiviral effects of miR-147b are recapitulated by the transfection of siNDUFA4 (Fig. 8d) at levels that match miR-147b-mediated NDUFA4 mRNA downregulation (Supplementary Fig. 8f). These results indicate that miR-147b enhance the IR by downregulating NDUFA4 in the absence of replacement by MOCCI. MOCCI incorporation on the other hand reduced cell death following viral infection in both HAECs and A549 (Fig. 8e). Besides suppressing pro-inflammatory cytokine production, MOCCI therefore exerts a cytoprotective effect, albeit at the cost of a small elevation in viral replication.

Fig. 8: MOCCI and miR-147b coordinate to optimize host protection.

a Heatmap of a subset of DENV and ZIKV-induced genes that were downregulated in MOCCI-expressing cells compared to control, based on RNA-seq (see “Module Detection” in Methods for more details). P = one-way ANOVA; n = 3 biological replicates per condition. MOCCI = MOCCI ORF only; WT-mRNA = MOCCI + miR-147b; ATGmut-mRNA = miR-147b only. b HAECs with the indicated transgenes were infected with DENV and ZIKV (MOI = 1.0). Forty-eight hours later, viral replication was measured by qPCR-mediated viral genome quantification, compared to a baseline obtained at 2 h postinfection. Values are expressed as percentage of the control of each biological replicate. Data were presented as mean ± SEM; P = one-way ANOVA; n = 6 biological replicates per condition. cA549 cells with the indicated transgenes were infected with DENV and ZIKV (MOI = 0.1). Fold change was calculated as the same in “c”. Data were presented as mean ± SEM; P = one-way ANOVA; n = 3 biological replicates per condition. d HAECs transfected with miR-control, miR-147b mimic, or siNDUFA4 were infected with ZIKV (MOI = 1.0). Fold change was calculated as the same in “c”. Data were presented as mean ± SEM; P = one-way ANOVA; n = 3 biological replicates per condition. eQuantitation of cell death in HAECs and A549 cells with indicated transgenes following ZIKV infection as measured by lactate dehydrogenase (LDH) release. Data were presented as mean ± SEM; P = one-way ANOVA; n = 6 biological replicates per condition from two repeats for HAECs and 3 biological replicates per condition for A549 cells.

Lastly, we sought to address the mechanism of IR regulation by MOCCI and miR-147b. Because NDUFA4-to-MOCCI switch lowers ROS production, we examined if MOCCI suppresses the IR through ROS reduction. Contrary to our expectation, NAC treatment of HAECs and A549 reduced vg copy number (Supplementary Fig. 9a). MOCCI -miR-147b regulation of host antiviral response is unlikely to be directly related to ROS modulation. Hence, we sought to find additional mechanisms of miR-147b antiviral immunity. Following DENV and ZIKV infection, viral RNA genomes are detected by cytosolic RNA sensors RIG-I and MDA529, leading to the activation of the NF-κB and IRF3/7, transcriptional mediators of the IR (Fig. 9a)30. Following ZIKV infection, both ATGmut HAECs and A549 expressed higher levels of RIG-I and its downstream adapter MDA5 (Fig. 9b). WT-mRNA had higher levels than control but lower levels than ATGmut (Fig. 9b), correlating with the intermediate suppression of viral replication (Fig. 8b, c) and once again highlighting the antagonism between MOCCI and miR-147b with respect to the IR. Downstream of RIG-I and MDA5, ATGmut-mRNA cells had increased NF-κB activation as measured by phosphorylated p65 (Fig. 9b). Despite the enhancement in RIG-I levels, WT and ATGmut-mRNA cells maintain their ability to suppress pro-inflammatory MCP-1 production in response to a RIG-I agonist (Supplementary Fig. 9b). Altogether, miR-147b potentiates RIG-I/MDA5/NF-κB activity, enhancing the IR and providing host protection during a viral infection. MOCCI on the other hand dampens this response and prevents excessive cell death and hyperinflammation. Altogether, during an acute infectious challenge, C15ORF48 relies on both its coding and non-coding functions to finetune the balance between inflammation and host protection by combining the anti-inflammatory and cytoprotective effects of MOCCI and the antiviral effects of miR-147b (Fig. 9c). These two activities potentially lead to synergistic protection of the host during viral infection, on the one hand controlling viral replication and on the other controlling host production of pro-inflammatory cytokines and cell death that cause tissue damage.

Fig. 9: miR-147b works through RIG-I/MDA5 pathway to modulate interferon response.

a RIG-I/MDA5 pathway of viral RNA recognition and interferon response. ISG interferon-stimulated genes. b Western blot of RIG-I, MDA5, and phosphorylated p65 in uninfected and 24 h ZIKV-infected HAECs and A549 cells with the indicated overexpression. n= 2 biological replicates. Source data are provided as a Source Data file. c Model of how coding and non-coding functions of C15ORF48transcript coordinate to provide host protection during viral infection. The exact mechanism on how inhibiting CIV activity can modulate cytokine secretion and cell death remains an outstanding question.


By intersecting with cytosolic sensors of danger and damage such as RLRs and NLRs, mitochondrial integrity and functionality are key determinants of inflammatory activation31. Rising from our observations that mitochondrial SEPs display a programmatic anti-association with inflammatory gene signatures (Fig. 1a), we performed an unbiased proteogenomic screen to uncover mito-SEPs that regulate inflammation. This led to the identification of C15ORF48 as the top inflammatory mito-SEP or “i-Mito-SEP”. The C15ORF48gene is also known as NMES-1 as a cancer-associated gene with uncharacterized function32. Altogether, results uncover a mitochondrial program during acute inflammation induced by the mito-SEP which we name MOCCI encoded by the C15ORF48 gene. Activation of this gene following viral infection or pro-inflammatory stimulus IL-1β induces switching of the 14th subunit of CIV from NDUFA4-to-MOCCI, accompanied by the biogenesis of miR-147b in the cytosol (Fig. 9c). MOCCI tonically dampens CIV activity leading to lowered membrane potential and reduced ROS production. Lowered ROS levels, in turn, contribute to suppressing pro-inflammatory cytokine production and the IR, but this is mitigated by the strong antiviral activity of miR-147b. In this fashion, the synergistic and complementary roles of MOCCI and miR-147b in HAECs together provide maximal host protection following an acute viral infection by suppressing host inflammation and viral replication.

The respiratory chain consists of a series of five multi-subunit ETC complexes encoded by 13 genes located in mitochondrial (mt) DNA and over 80 genes in the nuclear DNA. The ETC resides in the IMM and is responsible for over 80% of cellular ATP production. Traditionally viewed as invariant machines that perform housekeeping functions, it is increasingly apparent that ETC complexes can have variable compositions across tissue types and cellular contexts. For instance, CI is comprised of 45 subunits, of which more than half are accessory subunits performing non-enzymatic roles, potentially in assembly and stability of CI and the respiratory chain supercomplexes33. The expression levels of some subunits vary across tissues, suggesting that CI assemblies across cell types are not invariant. In the same manner, CIV which contains 14 subunits and over 20 assembly factors, is unique among ETC complexes in that several of its core subunits have isoforms that are expressed in a developmental stage-specific or oxygen-responsive manner34. These CIV isozymes are thought to differ in their oxidation potential and proton pumping efficiency35 in response to respiratory control cues. For instance, since oxygen is a substrate of CIV, the activity of the CIV must be directly coupled to and modulated by the availability of oxygen. This is achieved through oxygen-regulated expression of CIV subunit variants, namely COX4-1 and COX4-2. While COX4-1 expression is ubiquitous, COX4-2 is highly specific to the lung epithelium and induced upon hypoxic stress, where it increases the activity of CIV and optimizes its efficiency including reducing ROS under low oxygen availability36. Similarly, COX7A2 is exchanged for COX7A1 under hypoxic stress37. Assembly factors of CIV, such as the hypoxia-induced HIGD1A, have also been reported to play a role in boosting CIV activity when oxygen is limiting38. The literature is hence requisite with examples of CIV isozymes being employed to meet cellular energetic requirements under distinct cellular contexts with respect to oxygen availability. Because of the prominent interactions between the ETC and OXPHOS with other areas of cellular biology such as the control of cell death and inflammation, an outstanding question is whether CIV isozymes are also involved in the control of such functions. MOCCI was previously identified by Pagliarini and colleagues in a high throughput interactome screen as a mitochondrial uncharacterized (x) protein (MXP) that interacts with components of the ETC14. Here, we demonstrate that MOCCI is a close paralog of NDUFA4. NDUFA4 was previously thought to be a subunit of CI, and later reclassified to be the 14th subunit of mature CIV complex15,16,17. Its controversial assignment as an integral CIV subunit, owing to its absence in several solved structures of bacterial, fungal, and mammalian CIV, can be explained by its sensitivity to cholate detergents17, which remove NDUFA4 from the rest of the complex. NDUFA4 is not required for CIV assembly15,16, but rather has been speculated to enhance CIV activity by preventing the dimerization of CIV monomers17. While NDUFA4 expression is constitutive and ubiquitous, MOCCI is induced transcriptionally and translationally upon inflammatory stimulation by IL-1β. We demonstrate that MOCCI is incorporated into mature CIV monomer and dimers, and to a smaller extent, supercomplexes. Furthermore, its inclusion into CIV occurs concurrently with the downregulation of NDUFA4, with which it shares an almost superimposable predicted secondary structure. On this basis, supported by the finding that MOCCI and NDUFA4 are not found in the same CIV complexes, we propose that NDUFA4 is replaced by MOCCI in what we term the NDUFA4-to-MOCCI switch. MOCCI is therefore a variant of the 14th subunit of cytochrome C oxidase. Whether NDUFA4 is actively displaced by MOCCI in existing CIV complexes, or CIV-NDUFA4 is replaced by newly synthesized CIV-MOCCI is currently unclear. While this paper was in preparation, Kurihara and colleagues reported their findings that MOCCI, which they named COXA4L3, is expressed in developing spermatids where it is incorporated into CIV39, providing independent confirmation that MOCCI is a CIV protein.

The switch from NDUFA4-to-MOCCI downregulates the activity of CIV. miR-147b-mediated NDUFA4 downregulation alone also reduces CIV activity, in line with previous findings15. On the basis that CIV inhibition can reduce pro-inflammatory cytokine production in human PBMCs24, we postulate that attenuation of CIV activity by both MOCCI and miR-147b results in lowered cytokine production. How CIV activity controls cytokine production is an open area of investigation. One possible mechanism, as seen with MOCCI expression, is the reduction of ROS production, but that does not appear to be the case since miR-147b suppresses cytokine production without reducing ROS. CIV inhibition could modulate cytokine production indirectly through retrograde signaling pathways that are activated during mitochondrial stress40. A growing consensus points to mitochondrial retrograde signaling as a type of integrated stress response41,42. An intriguing possibility is that changes in CIV activity are sensed and relayed to downstream effectors through similar mechanisms.

How the NDUFA4-to-MOCCI switch leads to dampening of CIV activity is also unclear, as neither MOCCI incorporation nor NDUFA4 depletion affects total CIV levels. One possibility is based on observations that CIV transitions between monomeric and dimeric forms43,44, which assemble distinct COX7 paralogues COX7A2 and COX7A, respectively45. CIV dimers have lower activity compared to monomers46 and are sensitive to inhibition by high ATP/ADP ratios. Dimerization of CIV monomers has been documented to lower CIV activity and is postulated to reduce ROS production44. Based on the structural observation that NDUFA4 resides in the interface between CIV monomers, Yang and colleagues postulated that NDUFA4 inhibits dimerization of CIV, thereby maintaining CIV in a maximally active form17. MOCCI incorporation and NDUFA4 depletion might favor CIV dimerization and in so doing lower CIV activity and ROS production. However, our BN-PAGE analyses do not provide evidence to support this hypothesis, potentially due to the non-physiological nature of detergent extraction and the need for more in-situ physiological methods such as structural comparisons of MOCCI versus NDUFA4-containing CIV complexes.

It is important to point out that the reductions in CIV caused by MOCCI versus miR-147b are not qualitatively the same. In the case of MOCCI, reduction of CIV activity is accompanied by lower membrane potential and reduced ROS generation. This is not the case for miR-147b. These results make it clear that NDUFA4 removal with or without replacement by MOCCI have different effects downstream of dampening CIV activity, for instance in differential regulation of the IR. How MOCCI lowers membrane potential and ROS production are open questions. One possibility is that MOCCI incorporation alters the intrinsic uncoupling of CIV, also known as “proton slip”, which is the regulated diminishing of the proton pumping efficiency of the ETC complexes I, III, IV, and V such that the H+/e ratio falls below 1.047,48,49. Among all ETC complexes, regulated proton slip is only considered to be physiologically relevant at CIV50 and is thought to confer respiratory control at high ∆Ψm to prevent excessive ROS generation51. It has been observed that tissue-specific isoforms of CIV containing COX6A-1/2 have different H+/e ratios, suggesting that other non-enzymatic subunits that exhibit specific expression patterns (such as MOCCI) might contribute to regulating the H+/e ratio of CIV in distinct cellular contexts52 to modulate membrane potential and ROS generation. Of note, NDUFA4L2, the other paralog of NDUFA4, also exhibits inducible expression by hypoxia, and is thought to inhibit the activity of Complex I, reducing flux through the ETC and thereby reducing ROS generation53. Hypoxia activates pro-inflammatory gene expression54. The induction of NDUFA4L2 upon hypoxia and the ensuing decrease in ROS production might, in a similar way as what we observe with MOCCI, serve to dampen hypoxia-induced inflammation. MOCCI may also function in a similar manner during IL-1β challenge to induce immune-dampening responses through uncoupling.

The 3′UTR of C15ORF48 contains the conserved microRNA miR-147b, indicating a strong evolutionary selection to keep both genes on the same transcript (Fig. 2a). Mature miR-147b-3p has previously been characterized as an anti-inflammatory microRNA in monocytes55, in agreement with our findings that MOCCI exerts an anti-inflammatory effect. In addition, the top predicted target of miR-147b is NDUFA4, which we validated experimentally here. A simple scenario is one which MOCCI and miR-147b function as a pair by mutually reinforcing the NDUFA4-to-MOCCI switch. Our data however indicate that MOCCI and miR-147b have both convergent and divergent functions. On one hand, both MOCCI and miR-147b lower CIV activity and reduce MCP-1 and IL-6 production. On the other hand, MOCCI reduced membrane potential and reduced ROS production, while miR-147b alone potentiated the IR pathway by increasing RIG-I levels. In both of these processes, MOCCI and miR-147b are mutually antagonistic i.e., miR-147b opposes MOCCI-mediated drop in ROS and membrane potential (Fig. 6d–f), while MOCCI opposes miR-147b enhancement of RIG-I (Fig. 9b). Consequently, the phenotypic effects of expressing MOCCI are not the same as that of expressing ATGmut-mRNA, and expression of the WT-mRNA had intermediate effects. For instance, while MOCCI suppressed the IR response, leading to increased viral replication following DENV/ZIKV infection, WT-mRNA and ATGmut-mRNA enhanced the IR leading to suppression of viral replication. Altogether, these data indicate that MOCCI and miR-147b, while potentially working in concert to effect a NDUFA4-to-MOCCI switch in CIV, are not functionally equivalent. Whilst the IFN-dampening effect of MOCCI may serve to limit immune pathology, it also potentially favors viral replication. This is in turn counteracted by miR-147b, which functions to control viral replication. In the presence of both MOCCI and miR-147b, WT-mRNA HAECs dually control viral replication and mitigate pro-inflammatory cytokine production.

Notably, our model assumes that both MOCCI and miR-147b are present in the same cell that expresses C15ORF48 mRNA. According to existing paradigms, processing of exonic miRNAs like miR-147b causes mRNA destabilization due to poly-A removal—a requisite step in miR biogenesis56. mRNA translation and processing of pre-pri-mRNA should in theory be mutually exclusive. Therefore, it is possible that endogenous MOCCI and miR-147b are not usually present in the same cell at the same time. Rather, they are active in different subsets of cells within a tissue where they play distinct functions. For instance, in human PBMCs, C15ORF48 mRNA is expressed constitutively at high levels in monocytes (Supplementary Fig. 2a). Results from our THP-1 model in vitro predict that these monocytes express high levels of miR-147b (Supplementary Fig. 5a–d), consistent with their main function of immune surveillance and viral defense. Endothelial cells, on the other hand, induce C15ORF48 and translate the MOCCI peptide efficiently due to the exquisite need for protection against ROS production, which leads to endothelial dysfunction and further vascular damage. An outstanding question is the mechanism that govern cell-type specific translation of MOCCI and biogenesis of miR-147b. Altogether, our study highlights how coding and non-coding functions of a gene can be elegantly coordinated to achieve a specific biological outcome.

In conclusion, starting from an initial observation that mito-SEPs correlate negatively with inflammation, we present here our finding that C15ORF48 and the mito-SEP it encodes—MOCCI—assembles a CIV isozyme that fine-tunes the ETC to dampen inflammation following acute infection. Our results suggest that the MOCCI-miR-147b axis or modulating CIV activity in general might be a feasible strategy for ameliorating viral-induced hyperinflammation, a possibility that calls for further investigation. Additionally, since our study only addressed the effects of C15ORF48 in acute inflammation, it will be interesting to investigate its ability to ameliorate chronic inflammation particularly in the setting of cardiovascular diseases. Lastly, this study should motivate further efforts to mine the mito-SEP peptidome for other regulators of the respiratory chain that function to coordinate a ubiquitous biological process of oxidative phosphorylation to achieve specific, context-dependent cellular outcomes.

Article classification: Biological abstract
Share to:
Add:Room A11-329, 1st Floor, No.1, SBI Venture Street, Optics Valley, East Lake
New Technology Development Zone, Wuhan, China.
Certificate NO.:U18Q28010569R0S