The epithelial C15ORF48/miR-147-NDUFA4 axis is an essential regulator of gut inflammation, energy metabolism, and the microbiome

Significance

Emerging evidence emphasizes the significance of mitochondrial metabolism reprograming in the gut epithelium for regulating intestinal inflammation. Nevertheless, the intricate mechanisms underpinning this connection remain elusive. We have uncovered the predominant expression of C15ORF48/miR-147 in the gut epithelium and conducted an in vivo investigation of its functions. Our findings suggest that epithelial C15ORF48/miR-147 safeguards against aberrant gut inflammation and dysbiosis. Mechanistically, through its two molecular products—C15ORF48 protein and miR-147-3p miRNA—C15ORF48/miR-147 silences NDUFA4, a mitochondrial complex IV subunit. Notably, NDUFA4 not only impacts mitochondrial respiration but also controls inflammation signaling. This establishes the C15ORF48/miR-147-NDUFA4 molecular axis as an essential controller of gut homeostasis, linking energy metabolism and inflammation. These insights hold promise for innovative therapies against gastrointestinal diseases.

Abstract

Chronic inflammation is epidemiologically linked to the pathogenesis of gastrointestinal diseases, including inflammatory bowel disease (IBD) and colorectal cancer (CRC). However, our understanding of the molecular mechanisms controlling gut inflammation remains insufficient, hindering the development of targeted therapies for IBD and CRC. In this study, we uncovered C15ORF48/miR-147 as a negative regulator of gut inflammation, operating through the modulation of epithelial cell metabolism. C15ORF48/miR-147 encodes two molecular products, C15ORF48 protein and miR-147-3p microRNA, which are predominantly expressed in the intestinal epithelium. C15ORF48/miR-147 ablation leads to gut dysbiosis and exacerbates chemically induced colitis in mice. C15ORF48 and miR-147-3p work together to suppress colonocyte metabolism and inflammation by silencing NDUFA4, a subunit of mitochondrial complex IV (CIV). Interestingly, the C15ORF48 protein, a structural paralog of NDUFA4, contains a unique C-terminal α-helical domain crucial for displacing NDUFA4 from CIV and its subsequent degradation. NDUFA4 silencing hinders NF-κB signaling activation and consequently attenuates inflammatory responses. Collectively, our findings have established the C15ORF48/miR-147-NDUFA4 molecular axis as an indispensable regulator of gut homeostasis, bridging mitochondrial metabolism and inflammation.

Aberrant inflammation plays a central role in the pathogenesis of various gastrointestinal (GI) diseases, including inflammatory bowel disease (IBD) and colorectal cancer (CRC). Although broad immunosuppressive strategies, such as TNFα or IL-12 inhibition, provide treatment options for IBD, their considerable side effects and limited efficacy highlight the need for improved, more specific therapies (1, 2). To enable the informed development of innovative targeted therapies for the treatment of IBD and CRC, a deeper understanding of the mechanisms regulating gut inflammation and mucosal immunity is required.

Unlike most other animal organs, gut tissue exists in a state of persistent, low-grade inflammation due to constant exposure to luminal microbiota (3). This “physiological” inflammatory environment in the gut necessitates the existence of a complex and highly nuanced control mechanism to prevent “pathological” inflammatory responses. The gut epithelium, consisting of a single layer of intestinal epithelial cells (IECs), is a cornerstone of this exquisite regulatory apparatus (4). Besides maintaining a direct barrier function, IECs also play an integral role in various aspects of the inflammatory response, including i) sensing of commensals and pathogens, ii) context-dependent regulation of tissue-resident immune cell responses, iii) elimination of invading pathogens, and iv) resolution of inflammation and tissue repair (5–9). Moreover, IECs shape the composition of gut microbiota by promoting the growth of some microbes and inhibiting others (10). Conversely, gut microbiota can modulate epithelial and immune cell responses through the production of various metabolites (11).

Increasing evidence suggests that reprogramming of mitochondrial metabolism plays an important role in the regulation of intestinal inflammation and microbiome homeostasis (12). During gut inflammation, the metabolic program of IECs undergoes significant changes to meet the high energy and nutrient demands of the immune response and tissue repair processes: IECs throttle down energy generation through oxidative phosphorylation (OXPHOS) and concomitantly increase the utilization of anaerobic glycolysis (9, 13). In addition, the altered mitochondrial activity of inflamed IECs results in the production of various metabolites such as reactive oxygen species (ROS) and succinate, which activate NF-κB signaling and promote inflammation (14). Interestingly, it has been postulated that mitochondrial dysfunction in IECs is a crucial driver in the pathogenesis of IBD, supporting the notion that changes in mitochondrial metabolism can directly shape the inflammatory response (12). Despite significant progress in the field, the precise molecular mechanisms through which mitochondria function as a central signaling hub in regulating intestinal inflammation remain inadequately understood.

In this study, we test the hypothesis that C15ORF48/miR-147 functions as a critical regulator of gut inflammation by controlling mitochondrial metabolism. C15ORF48/miR-147 is an inflammation- induced gene that produces two distinct molecular products: a small mitochondrial protein, C15ORF48, and a microRNA, miR-147-3p (15). These two molecules work cooperatively to regulate cell metabolism and inflammation by inhibiting NDUFA4, an accessory subunit of the mitochondrial complex IV (CIV) (16, 17). However, the role of C15ORF48/miR-147 in inflammation remains controversial with some studies suggesting that it serves as an inflammation promoter (17), whereas others indicate an inhibitory role (16). Furthermore, the physiological role of C15ORF48/miR-147 remains unclear due to the lack of appropriate animal models for testing its function.

Our findings using mouse genetic models suggest that C15ORF48/miR-147 is highly expressed in IECs under physiological conditions and functions as a molecular brake on intestinal inflammation by silencing NDUFA4 and subsequently dampening mitochondrial metabolism.

Results

C15ORF48/miR-147 Is Predominantly Expressed in IECs.

Our initial interest in C15ORF48/miR-147 stems from its role as the miR-147-3p host gene. This miRNA was originally identified as a lipopolysaccharide (LPS)-inducible miRNA that attenuates inflammatory responses in myeloid cells (15). However, prior efforts did not fully establish the physiological functions of this miRNA. In addition to miR-147-3p, its host gene is predicted to give rise to a small (83aa) protein, C15ORF48, which shares a high degree (53% similarity) of sequence homology with NDUFA4 protein (SI Appendix, Fig. S1 A and B). NDUFA4 was recently identified as the 14th subunit of the mitochondrial CIV and proposed to regulate CIV homodimerization (18, 19). NDUFA4 knockdown (KD) in mammalian cells attenuates CIV activity and inhibits proliferation in cells dependent on OXPHOS as their only energy source (19). Interestingly, miR-147-3p is predicted by the TargetScan algorithm to bind to the 3′UTR of NDUFA4 and attenuate its expression (20, 21) (SI Appendix, Fig. S1C).

To start defining the physiological functions of C15ORF48/miR-147, we performed expression profiling of several mouse tissues. Our qRT-PCR analysis of mature miR-147-3p expression (Fig. 1A) and western blot analysis of C15ORF48 expression (Fig. 1B) suggest that both molecular products of C15ORF48/miR-147 are predominantly expressed in the large intestine. Furthermore, immunofluorescence analysis of mouse colon tissue sections indicates that C15ORF48 is primarily expressed by EpCAM⁺ IECs (Fig. 1C). Interestingly, C15ORF48 is mainly expressed by terminally differentiated colonocytes in the apical part of the crypt structures, whereas the intestinal stem cells and proliferating progenitors at the base of the crypts are apparently devoid of C15ORF48 expression (SI Appendix, Fig. S2A), suggesting that C15ORF48/miR-147 may have a role in colonocyte differentiation and/or its expression is potentially controlled by a protein or chemical gradient along the apical–basal axis (22–24). In addition, our immunofluorescence analysis of human colon carcinoma HCT116 cells revealed that C15ORF48 protein is colocalized with a classical mitochondrial marker TOMM20 (SI Appendix, Fig. S2B), suggesting that C15ORF48 is a mitochondrial protein. This observation is in good agreement with prior findings (16). Based on the data from the Genotype-Tissue Expression (GTEx) portal (https://gtexportal.org/) and the Human Protein Atlas (https://www.proteinatlas.org), the expression pattern of human C15ORF48/miR-147 is largely congruent with the mouse expression profile (Fig. 1D and SI Appendix, Fig. S2C), indicating possible evolutionary conservation of C15ORF48/miR-147 functions in IECs.

Fig. 1.

The C15ORF48/miR-147 gene encodes two distinct molecular products—miR-147-3p miRNA and C15ORF48 protein—both of which are abundantly expressed in the large intestine. (A) qRT-PCR analysis of mature miR-147-3p expression in various tissues isolated from C57BL/6J mice (n = 2), including the heart, liver, spleen, esophagus, large intestine, small intestine, white adipose tissue (WAT), brown adipose tissue (BAT), muscle, and pancreas. The expression levels were normalized to snoRNA234, and miR-147-3p expression in the heart was arbitrarily set to 1. (B) Western blot (WB) analysis of C15ORF48 protein expression in different tissues isolated from the C57BL/6J mouse. GAPDH serves as loading control. (C) Immunofluorescence (IF) analysis of C15ORF48 (green) and EpCAM (epithelial cell marker, red) expression in mouse colon tissue. DAPI stain (blue) was used to define nuclei. (Scale bar, 10 μm.) (D) Violin plot of C15ORF48/miR-147 expression in normal human tissues. The data were compiled by querying the GTEx portal (https://gtexportal.org). Expression values are shown in transcripts per million reads (TPM). (E) C15ORF48/miR-147 expression in rectal mucosal biopsies from healthy controls (n = 20) and new-onset pediatric ulcerative colitis patients (n = 206). The previously published GSE109142 dataset (25) was used for the analysis. Expression values are shown in TPM reads. ****P < 0.0001 by Student’s t test. Data in (A), (D), and (E) are shown as mean ± SD.

Mitochondrial dysfunction in colonocytes is implicated as a disease driver in ulcerative colitis (UC) patients (25–27). Thus, to explore the potential role of C15ORF48/miR-147 in aberrant gut inflammation, we examined the expression levels of C15ORF48/miR-147 in rectal mucosal biopsies isolated from new-onset UC patients (GSE109142) (25) and found a significant C15ORF48/miR-147 down-regulation in comparison to normal controls (Fig. 1E), suggesting a potential role for C15ORF48/miR-147 in the pathogenesis of IBD.

Epithelial C15ORF48/miR-147 Protects Mice from Dextran Sodium Sulfate (DSS)-Induced Colitis.

To genetically dissect the functions of C15ORF48/miR-147, we generated germline C15ORF48/miR-147 knockout (KO) mice (hereafter C15ORF48/miR-147^−/− mice) that lack expression of both C15ORF48 and miR-147-3p in all tissues. This was accomplished by deleting a ~2 kb genomic fragment that encompasses exons 4 and 5 of C15ORF48/miR-147 using CRISPR/Cas9 gene editing technology (SI Appendix, Fig. S3A). The loss of C15ORF48/miR-147 expression in the KO mice was verified by western blot analysis using anti-mouse C15ORF48 antibody (SI Appendix, Fig. S3B) and qRT-PCR analysis of miR-147-3p (SI Appendix, Fig. S3C). C15ORF48/miR-147^−/− mice develop normally and do not display any overt morphological changes in the gut (SI Appendix, Fig. S3E). Of note, our analysis of the major immune cell populations in C15ORF48/miR-147^−/− mice using flow cytometry also did not reveal any gross defects (SI Appendix, Fig. S3 G and H).

To examine the role of C15ORF48/miR-147 in gut inflammation, we challenged C15ORF48/miR-147^−/− mice with DSS to induce acute colitis. The DSS-induced colitis mouse model was chosen in part because it resembles the pathophysiological features of human UC (28), which as discussed above displays C15ORF48/miR-147 dysregulation (Fig. 1E). Consistent with the reduced expression of C15ORF48/miR-147 observed in human UC patients, we found a decrease in C15ORF48 protein levels in the colons of DSS-challenged C57BL6/J mice (SI Appendix, Fig. S3F). Upon DSS administration, C15ORF48/miR-147^−/− mice develop an exacerbated colitis, as indicated by more profound body weight loss (Fig. 2A), an elevated disease activity index (DAI) (29) (Fig. 2B), more pronounced shortening of the colon (Fig. 2C), and considerably higher colon permeability (Fig. 2D) in comparison to wild-type (WT) control mice. Histopathological analysis of colon tissue from DSS-challenged C15ORF48/miR-147^−/−mice revealed a more severe disease in comparison to the control, as indicated by more extensive gland loss, epithelial layer erosion, edema, and immune cell infiltration (Fig. 2 E and F). Furthermore, based on immunofluorescence analysis of colon sections, C15ORF48/miR-147^−/− mice display a significant decrease in IEC proliferation and a remarkable accumulation of CD45⁺ immune cells (Fig. 2G). Additionally, production of proinflammatory cytokines, such as IL6, was significantly up-regulated in C15ORF48/miR-147^−/− colons (Fig. 2H). Together, our findings demonstrate that C15ORF48/miR-147 is required to keep gut inflammation in check.

Fig. 2.

Epithelial C15ORF48/miR-147 protects mice from fulminant DSS-induced colitis. (A) Body weight change in WT and C15ORF48/miR-147^−/− (KO) mice (n = 5) challenged with 3% DSS for 6 d. (B) DAI in WT and C15ORF48/miR-147^−/− (KO) mice (n = 5) challenged with 3% DSS for 6 d. The DAI was calculated on days 5 to 9 after the start of the DSS treatment as a sum of the body weight decrease, stool consistency, and rectal bleeding scores (0 to 4 scale). (C) Representative image of the colons and colon length in WT and C15ORF48/miR-147^−/− (KO) mice on day 10 after 3% DSS treatment (n = 5). (D) Analysis of colon permeability in WT and C15ORF48/miR-147^−/− (KO) mice on day 9 after 3% DSS challenge (n = 2). Serum concentrations of FITC-Dextran were determined 4 h after the oral administration of the compound. (E) Representative images of H&E-stained colon tissue sections from WT and C15ORF48/miR-147^−/− (KO) mice on day 10 after 4% DSS treatment. Note a more profound epithelial erosion, loss of glands, and infiltration of lymphoid cells in the KO sample. (Scale bar, 100 μm.) (F) Histopathology scores for WT and C15ORF48/miR-147^−/− (KO) mice on day 10 of the DSS-induced colitis model (n = 5). The score is a sum of the severity scores (0 to 5 scale) for gland loss, epithelial erosion, epithelial hyperplasia, edema, lymphoid accumulation, inflammation, and neutrophil infiltration. (G) IF analysis of EpCAM (cyan), Ki67 (proliferation marker; red), and CD45 (immune cell marker; green) expression in colon tissue sections isolated from WT and C15ORF48/miR-147^−/− (KO) mice on day 9 of the DSS-induced colitis model. Note a lack of epithelial cell proliferation and massive recruitment of immune cells in the KO samples. (H) ELISA analysis of IL6 production in colon homogenate samples from WT and C15ORF48/miR-147^−/− (KO) mice on day 9 of the DSS-induced colitis model. IL6 concentration is normalized to tissue weight. (I) Body weight change and (J) colon length in control Vil1^Cre and C15ORF48/miR-147^ΔIEC (KO^ΔIEC) mice (n = 5) after 3% DSS challenge for 6 d. (K) H&E-stained colon tissue sections from control Vil1^Cre and KO^ΔIEC mice at day 10 of the DSS-induced colitis model. (Scale bar, 100 μm.) Data in (A–D), (F), (H–J) panels are shown as mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by Student’s t test.

To determine whether aberrant inflammation in C15ORF48/miR- 147^−/− mice is driven by a defect in the immune or epithelial cells of the gut, we generated bone marrow (BM) chimeric mice (SI Appendix, Fig. S4A) and challenged them with DSS. We found that C15ORF48/miR-147^−/−hosts develop more severe colitis than WT hosts regardless of the source of donor BM as indicated by more pronounced body weight loss (SI Appendix, Fig. S4B), colon shortening (SI Appendix, Fig. S4C), and epithelial layer erosion (SI Appendix, Fig. S4D), suggesting that nonimmune (most likely epithelial) C15ORF48/miR-147 plays a protective role in gut inflammation. Interestingly, C15ORF48/miR-147 in immune cells might have an inflammation promoting role because C15ORF 48/miR-147^−/− host mice transplanted with C15ORF48/miR-147^−/− BM (KO→KO chimera) displayed less severe weight loss than C15ORF48/miR-147^−/− host mice transplanted with WT BM (WT→KO chimera) (SI Appendix, Fig. S4B).

To more precisely define the cell-intrinsic role of C15ORF48/miR- 147 in gut inflammation, we generated mice with an IEC-specific C15ORF48/miR-147 deletion. To that end, we first made a conditional C15ORF48/miR-147 allele (hereafter C15ORF48/miR-147^fl/fl) by inserting two loxP sites flanking exons 4 and 5 of the C15ORF48/ miR-147 gene using CRISPR/Cas9 technology (SI Appendix, Fig. S3 A and J). Like C15ORF48/miR-147^−/− mice, C15ORF48/miR-147^fl/fl mice are predicted to abrogate expression of both C15ORF48 and miR-147-3p after Cre-mediated locus excision. C15ORF48/miR-147^fl/fl mice were then crossed with Villin1-Cre (Vil1^Cre) deleter mice (30) to generate Vil1^CreC15ORF48/miR-147^fl/fl (hereafter C15ORF48/miR-147 ^ΔIEC) mice. Villin1 encodes an actin-binding protein whose expression is restricted to gut and renal epithelial cells (30). As expected, C15ORF48/miR-147 ^ΔIEC mice display an IEC-specific loss of C15ORF48 (SI Appendix, Fig. S3I). In agreement with our BM chimera studies, DSS challenge of C15ORF48/miR-147^ΔIEC mice results in a more severe colitis in comparison with control Vil1^Cre mice (Fig. 2 I–K). Thus, our mouse genetic studies establish C15ORF48/ miR-147 as an essential, IEC-intrinsic negative regulator of gut inflammation.

C15ORF48/miR-147 Deletion Up-Regulates IEC Metabolism and Inflammatory Responses.

To define the molecular mechanism by which C15ORF48/miR-147 controls gut inflammation, we performed global transcriptome analysis of purified C15ORF48/miR-147-sufficient and -deficient colonocytes from unchallenged and DSS-challenged mice using RNA-seq. We found that C15ORF48/miR-147 ablation alters expression of 128 genes in cells from unchallenged mice (SI Appendix, Fig. S5A) and 1,039 genes in the colonocytes from DSS-challenged mice (SI Appendix, Fig. S5C). Pathway analysis of differentially expressed genes (DEGs) in untreated C15ORF48/miR-147^−/− colonocytes identified a significant up-regulation of genes involved in inflammatory responses (type I and II IFN signaling) and energy metabolism (OXPHOS, glycolysis, and xenobiotic metabolism) (Fig. 3A). Mirroring the untreated cells, the DSS-challenged C15ORF48/miR-147^−/− colonocytes exhibited an up-regulation of innate immune (type I IFN, IL6 and TNF receptor pathways) and cellular metabolism (OXPHOS, mTOR signaling, and xenobiotic metabolism) genes (Fig. 3 D and E). We validated the aberrant activation of inflammatory responses in DSS-challenged C15ORF48/miR-147^−/− colonocytes using qRT-PCR analysis of the proinflammatory cytokine IL1β (Fig. 3F).

Fig. 3.

Loss of C15ORF48/miR-147 results in the dysregulation of inflammatory signaling and metabolic activities in colonocytes. (A) GSEA pathway enrichment analysis of up-regulated genes (fold change >1.5; P < 0.05; RPKM > 1) in unchallenged C15ORF48/miR-147^−/− (KO) colonocytes. WT colonocytes served as the baseline to determine the DEGs in KO colonocytes. NES, normalized enrichment score. (B) OCR profiles of purified primary colonocytes isolated from WT and C15ORF48/miR-147^−/− (KO) mice. The OCRs were measured sequentially under basal conditions, following inhibition of ATP synthase (with oligomycin), uncoupling the ETC with FCCP, and blocking complex I and III with rotenone and antimycin, respectively. The data were normalized using cell counts per well. (C) Baseline OCR to ECAR ratio of purified colonocytes isolated from WT and C15ORF48/miR-147^−/− (KO) mice (n = 3). P = 0.051 by Student’s t test. (D) GSEA pathway enrichment analysis of up-regulated genes (fold change >1.5; P < 0.05; RPKM > 1) in DSS-challenged (3% for 3 d) C15ORF48/miR-147−/− (KO) colonocytes. WT colonocytes served as the baseline to determine the DEGs in KO colonocytes. NES, normalized enrichment score. (E) Heatmap of gene expression values (z-score normalization) for several selected OXPHOS (Left) and inflammation-related (Right) genes in WT and C15ORF48/miR-147^−/− (KO) colonocytes (n = 3) after 3 d DSS challenge. (F) qRT-PCR analysis of IL1β expression in purified colonocytes isolated from WT and C15ORF48/miR-147^−/− (KO) mice after 3 d DSS challenge (n = 3). GAPDH was used for normalization. *P < 0.05 by Student’s t test. (G) GSEA pathway enrichment analysis of down-regulated genes (fold change < −1.5; P < 0.05; RPKM > 1) in IL1β (100 ng/mL for 4 h) challenged HCT116 cells ectopically expressing FLAG-tagged C15ORF48 protein. NES, normalized enrichment score. (H) OCR profiles of HCT116 cells stably expressing tetracycline-inducible pCW57.1-C15ORF48/miR-147 expression constructs. PCW is a vector control; WW plasmid gives rise to both C15ORF48 and miR-147-3p; WX plasmid was engineered to produce C15ORF48 alone; XW plasmid was engineered to produce miR-147-3p alone. (I) Heatmap of gene expression values (z-score normalization) for several selected inflammation-related genes in IL1β (100 ng/mL for 4 h) challenged HCT116 cells ectopically expressing FLAG-tagged C15ORF48 protein. Data in (B), (C), (F), and (H) panels are shown as mean ± SD.

To further investigate the role of C15ORF48/miR-147 in energy metabolism, we examined the rates of OXPHOS and anaerobic glycolysis in the purified C15ORF48/miR-147-sufficient and -deficient colonocytes using Seahorse extracellular flux technology. We observed that oxygen consumption rates (OCRs) associated with both basal and maximal mitochondrial respiration were significantly elevated in C15ORF48/miR-147^−/− colonocytes (Fig. 3B and SI Appendix, Fig. S5 D and E). Moreover, extracellular acidification rates (ECARs), which measure lactic acid levels generated during anaerobic glycolysis, were also significantly higher in C15ORF48/miR-147^−/− cells (SI Appendix, Fig. S5 B, F, and G). However, the baseline OCR/ECAR ratio is lower in C15ORF48/miR-147^−/− colonocytes (Fig. 3C), suggesting that C15ORF48/miR-147 deletion not only elevates the overall metabolic potential of IECs but also skews their energy metabolism toward glycolysis, a metabolic switch that is often observed during inflammation (31–33).

To dissect the role of C15ORF48 and miR-147-3p in the regulation of energy metabolism, we generated the C15ORF48/miR-147 expression constructs (Fig. 4B) that encode either C15ORF48 alone (WX construct) or miR-147-3p alone (XW construct) and compared their impact on mitochondrial respiration with C15ORF48/miR-147 cDNA (WW construct) that encodes both molecular products. First, we found that C15ORF48/miR-147 (WW construct) overexpression in gut epithelial and macrophage cell lines dampens their energy metabolism by attenuating both OXPHOS and anaerobic glycolysis (Fig. 3H and SI Appendix, Figs. S5J and S6). This finding is directly opposite to our observations with C15ORF48/miR-147^−/− colonocytes (Fig. 3B and SI Appendix, Fig. S5B). Additionally, both C15ORF48 and miR-147-3p can lower the ECAR, whereas the maximal OCR is significantly attenuated only by C15ORF48 (WX construct), but not miR-147-3p (XW construct) (Fig. 3H and SI Appendix, Fig. S5J). Collectively, these findings demonstrate that C15ORF48 and miR-147-3p function as suppressors of energy metabolism but have distinct roles in this inhibitory process.

Fig. 4.

C15ORF48 and miR-147-3p cooperatively suppress NDUFA4 in colonocytes. (A) Volcano plot showing expression changes for putative miR-147-3p target genes in purified colonocytes isolated from unchallenged (blue) and DSS-challenged (3% DSS for 3 d) (red) WT and C15ORF48/miR-147^−/− (KO) mice (n = 3). Note NDUFA4 as the most significantly induced miR-147-3p target under both conditions. (B) Schematic diagram of tetracycline-inducible pCW57.1-C15ORF48/miR-147 expression plasmids and their corresponding molecular products. C15ORF48/miR-147 cDNA was mutated to direct differential expression of C15ORF48 and miR-147-3p. WW gives rise to both C15ORF48 and miR-147-3p; WX produces C15ORF48 alone; XW produces miR-147-3p alone. (C) WB analysis of NDUFA4 and C15ORF48 expression in HCT116 cells stably expressing tetracycline-inducible pCW57.1-C15ORF48/miR-147 expression constructs. Relative NDUFA4 expression was calculated using densitometry and normalized to β-actin levels. NDUFA4 expression in empty vector cells (PCW) was arbitrarily set to 1. FC: fold change. Note a strong capacity of the WX construct (C15ORF48 alone) to suppress NDUFA4 expression. (D) The comparison of the predicted structure of C15ORF48 (green) and the crystal structure of NDUFA4 (red) (PDB ID: Q62425). Note a unique C-terminal α-helix in C15ORF48. FL: full-length C15ORF48; ΔC11: truncated version of C15ORF48 that lacks the C-terminal α-helix (11aa). (E) The relative positions of NDUFA4 (red) and C15ORF48 (green) in CIV cryo-EM structure (PDB ID: 5Z62). All CIV subunits are depicted in tube representation. Note the close proximity of the C-terminal α-helix of C15ORF48 to COX6B subunit, shown in the magnified Inset. (F) WB analysis of HCT116 cells expressing tetracycline-inducible FLAG-tagged C15ORF48 or its ΔC11 mutant using anti-FLAG and anti-NDUFA4 antibodies. β-actin was used for normalization. (G) Coimmunoprecipitation of endogenous COX6B with FLAG-tagged C15ORF48 proteins isolated from HCT116 cell mitochondria. (H) OCR profiles of HCT116 cells expressing tetracycline-inducible FLAG-tagged C15ORF48 or its ΔC11 mutant. Cells were induced with doxycycline for 48 h before the analysis. Note an elevated mitochondrial respiration in ΔC11 OE HCT116 cells. Data are shown as mean ± SD. DOX, doxycycline.

In addition, we observed that C15ORF48 protein overexpression down-regulates genes involved in the regulation of energy metabolism and inflammation. Gene set enrichment analysis (GSEA) (34) of DEGs in unstimulated C15ORF48 overexpressing (OE) cells revealed a down-regulation of cell metabolism (OXPHOS and glycolysis) pathways (SI Appendix, Fig. S5 H and I), while down-regulated transcripts in IL1β stimulated C15ORF48 OE cells were enriched in inflammatory response genes (type I and II IFN, TNF, and IL6 signaling) (Fig. 3 G and I). Of note, these findings are opposite to gene expression changes observed in C15ORF48/miR-147-deficient colonocytes (Fig. 3 A and D). Collectively, the opposite gene expression changes observed in C15ORF48/miR-147-deficient colonocytes and C15ORF48 OE cells suggest that C15ORF48/miR-147 may function as a negative regulator of inflammation and cell metabolism.

Cooperative Action of C15ORF48 and miR-147-3p Silences NDUFA4.

To further understand the molecular mechanisms of C15ORF48/miR-147, we examined the impact of its ablation on putative miR-147-3p targets. Our analysis of DEGs in C15ORF48/miR-147^−/− colonocytes suggests that NDUFA4 is the only miR-147-3p target gene (out of 1,070 putative targets) that is significantly up-regulated with or without the DSS challenge (Fig. 4A). Moreover, we also detected an elevated NDUFA4 expression by western blot analysis of total colon extracts and purified colonocytes from C15ORF48/miR-147^−/− mice (SI Appendix, Fig. S3D). Using a 3′UTR luciferase reporter assay, we confirmed the capacity of miR-147-3p to attenuate NDUFA4 expression by binding directly to an evolutionarily conserved “seed” sequence in the 3′UTR of NDUFA4 (SI Appendix, Fig. S7A).

In line with these findings, we observed that C15ORF48/miR-147 overexpression suppresses NDUFA4 protein expression (Fig. 4C and SI Appendix, Fig. S7C); however, this effect is more profound than the typical mild effect of miRNA-mediated regulation. Therefore, to further understand the NDUFA4 silencing capacity of C15ORF48/miR-147, we made use of the C15ORF48/miR-147 expression mutants (WW, WX, and XW constructs; Fig. 4B) that combinatorially express C15ORF48 and miR-147-3p molecules and were discussed above. Our analysis revealed that C15ORF48 protein has a stronger capacity to down-regulate NDUFA4, in comparison to miR-147-3p (Fig. 4C and SI Appendix, Fig. S7C). In contrast to miR-147-3p, C15ORF48 has no destabilizing effect on NDUFA4 mRNA (SI Appendix, Fig. S7D), suggesting that C15ORF48 triggers NDUFA4 protein degradation. We found that C15ORF48-mediated NDUFA4 degradation is proteasome independent because proteasomal inhibitors do not interfere with the process (SI Appendix, Fig. S7E). Taken together, our results imply that miR-147-3p and C15ORF48 act cooperatively to silence NDUFA4 at the posttranscriptional and posttranslational levels, respectively.

Because NDUFA4 and C15ORF48 proteins display significant sequence homology, we used the PyMOL algorithm to compare the predicted secondary structure of C15ORF48 with the deposited crystal structure of NDUFA4 (Fig. 4D) and docked both molecules into the published crystal structure of mitochondrial CIV (Fig. 4E). Our results indicate that the N-terminal portions of C15ORF48 and NDUFA4, which comprise the transmembrane domains of both proteins, are superimposable, suggesting the same docking position of C15ORF48 and NDUFA4 in CIV. On the other hand, C15ORF48 has a unique α-helical domain in its C terminus, which is positioned in close proximity to the COX6B subunit of CIV (Fig. 4E).

To investigate the role of the C-terminal α-helix in C15ORF48, we generated a truncated version of C15ORF48 that lacks this structural moiety (ΔC11 construct; Fig. 4D). Our immunofluorescence analysis suggests that deletion of the C-terminal α-helix has no impact on mitochondrial localization of C15ORF48 (SI Appendix, Fig. S7F) but interferes with the capacity of C15ORF48 to induce NDUFA4 protein degradation (Fig. 4F and SI Appendix, Fig. S7G). These results indicate that the C-terminal α-helical domain in C15ORF48 is required for NDUFA4 degradation. In agreement with our modeling studies (Fig. 4E), immunoprecipitation experiments revealed that the C-terminal α-helical domain in C15ORF48 is responsible for its interaction with the COX6B subunit of CIV (Fig. 4G). However, the significance of this interaction for C15ORF48 function is currently unclear. Additionally, in contrast to the full-length C15ORF48, overexpression of the ΔC11 mutant elevates mitochondrial respiration (Fig. 4H) and boosts the metabolic potential (SI Appendix, Fig. S7H) of colonocytes.

NDUFA4 Depletion Attenuates Cellular Metabolism and Inflammation.

Our findings indicate that C15ORF48/miR-147 functions by suppressing NDUFA4; however, the role of NDUFA4 in the regulation of gut inflammation is unclear. Thus, to facilitate our analysis of NDUFA4 functions, we silenced its expression in HCT116 cells using CRISPR/Cas9 gene editing technology. As expected, western blot analysis of several independent NDUFA4 KO clones revealed a complete loss of NDUFA4 expression (SI Appendix, Fig. S8A). In line with previous reports suggesting a role for NDUFA4 in the regulation of CIV activity (19), baseline and maximal mitochondrial respiration were significantly attenuated in NDUFA4 KO cells (Fig. 5A and SI Appendix, Fig. S8B). As a result, NDUFA4 KO cells become more glycolytic; however, their overall metabolic potential is reduced in comparison to control cells (SI Appendix, Fig. S8C). Similar trends were observed in RAW264.7 NDUFA4 KD macrophages (SI Appendix, Fig. S9 A–C), suggesting that the impact of NDUFA4 on mitochondrial respiration is tissue independent.

Fig. 5.

NDUFA4 is an essential positive regulator of energy metabolism and inflammation. (A) OCR profile of NDUFA4 KO (A2 clone) HCT116 cells. Control, parental HCT116 cells. (B) GSEA analysis of down-regulated genes (fold change < −1.5; P < 0.05; RPKM > 1) in IL1β (100 ng/mL for 4 h) challenged NDUFA4 KO (A2 and B2 clones) cells. NES, normalized enrichment score. (C) qRT-PCR analysis of TNFα, CXCL8, OAS1, and NOS2 expression in unchallenged and IL1β (100 ng/mL for 4 h) challenged parental control and NDUFA4 KO cells (A2 and B2 clones). GAPDH was used for normalization. Gene expression values in untreated parental control cells were arbitrarily set to 1. (D) WB analysis of IL1β (100 ng/mL for indicated time) stimulated parental control and NDUFA4 KO cells (A2 clone) using anti-phospho-IκBα and anti-total IκBα antibodies. β-actin was used as loading control. (E) Analysis of NF-κB activation in unstimulated and IL1β (100 ng/mL for 6 h) stimulated parental control and NDUFA4 KO cells (A2 clone) using NF-κB luciferase reporter assay. WT Firefly luciferase reporter contains three consensus NF-κB binding sites, which are mutated in the mutant reporter. Cotransfected Renilla luciferase reporter (pRL-SV40) was used for normalization. Data in (A), (C), and (E) are shown as mean ± SD. **P < 0.01; ***P < 0.005; ****P < 0.0001 using one-way ANOVA.

Global transcriptome analysis of NDUFA4 KO cells revealed perturbed expression of 2,010 genes in untreated cells (SI Appendix, Fig. S8D) and 2,217 genes upon IL1β challenge (SI Appendix, Fig. S8E). GSEA analysis of the DEGs identified a significant down- regulation of genes controlling inflammatory responses and hypoxia in both unstimulated (SI Appendix, Fig. S8F) and IL1β challenged NDUFA4 KO cells (Fig. 5B). Defective activation of inflammatory responses was also observed in LPS-treated RAW264.7 NDUFA4 KD cells (SI Appendix, Fig. S9 D–G), suggesting that control of inflammatory signaling by NDUFA4 is global and stimulus independent. We further examined the impact of NDUFA4 depletion on the activation of diverse signaling pathways in response to inflammatory stimuli and concluded that NDUFA4 promotes a broad range of innate immune effector responses, including proinflammatory cytokine and chemokine secretion (TNFα and CXCL8; Fig. 5C and SI Appendix, Fig. S9H), dsRNA sensing and IFN signaling (OAS1; Fig. 5C), ROS generation, and nitric oxide (NO) production (NOS2; Fig. 5C and SI Appendix, Fig. S9 I and J). Furthermore, we observed a notable reduction of IκBα phosphorylation and subsequent IκBα degradation in NDUFA4-depleted cells (Fig. 5D and SI Appendix, Fig. S9K). In line with these findings, activation of an NF-κB luciferase reporter was significantly diminished in NDUFA4 KO cells (Fig. 5E). Taken together, these observations indicate that NDUFA4 depletion leads to a substantial attenuation of NF-κB activation, which might contribute to the profound dysregulation of inflammatory signaling in these cells. Our findings also imply a potential role for NDUFA4 as a molecular link between mitochondrial metabolism and inflammation. Additionally, the opposite direction of energy metabolism and inflammatory response changes in C15ORF48 OE and NDUFA4 KO cells on one hand, and C15ORF48/miR-147 KO cells, on the other, supports our hypothesis that C15ORF48/miR-147 negatively regulates cell metabolism and inflammation by suppressing NDUFA4.

The Ability of C15ORF48 to Control Cell Metabolism and Inflammation Is Limited to Its NDUFA4 Silencing Capacity.

The capability of C15ORF48/miR-147 to silence NDUFA4 and the structural homology between C15ORF48 and NDUFA4 proteins raise a possibility that C15ORF48 can replace NDUFA4 in CIV and, thus, specifically reshape CIV activity. To test this notion, we ectopically expressed FLAG-tagged C15ORF48 and its ΔC11 mutant in NDUFA4 KO cells (SI Appendix, Fig. S10A). Interestingly, the maximal mitochondrial respiration capacity of NDUFA4 KO cells expressing the full-length C15ORF48 was not substantially altered (Fig. 6A and SI Appendix, Fig. S10B), indicating that C15ORF48 cannot fully rescue the impaired OXPHOS in NDUFA4 KO cells Fig. 5A. In contrast, the ΔC11 mutant significantly elevated the maximal respiration in NDUFA4 KO cells, suggesting that it can functionally substitute NDUFA4 in CIV (Fig. 6A and SI Appendix, Fig. S10B). As control, overexpression of FLAG-tagged NDUFA4 also rescued the OXPHOS defect in NDUFA4 KO cells (Fig. 6A and SI Appendix, Fig. S10B). Collectively, these findings support a hypothesis that the principal function of C15ORF48 is NDUFA4 silencing, and the C-terminal α-helical domain is critical for this activity. Deletion of the C-terminal moiety seems to transform C15ORF48 into an NDUFA4 functional homolog. This conclusion is further supported by our finding that ΔC11 mutant OE boosts the metabolic potential of NDUFA4-sufficient cells (Fig. 4H).

Fig. 6.

C15ORF48 overexpression is unable to fully rescue the metabolic and inflammatory defects in NDUFA4 KO colonocytes. (A) OCR profiles of NDUFA4 KO (clone A2) cells stably expressing FLAG-tagged NDUFA4, FLAG-tagged full-length C15ORF48 (FL), and its ΔC11 mutant. WT, parental HCT116 cells; A2, NDUFA4 KO (clone A2) cells. (B) qRT-PCR analysis of TNFα and NOS2 expression in unchallenged and IL1β (100 ng/mL for 4 h) challenged NDUFA4 KO cells OE FLAG-tagged full-length C15ORF48 (FL) and its ΔC11 mutant. GAPDH was used for normalization. Gene expression in untreated HCT116 cells (WT) was arbitrarily set to 1. Ctrl, NDUFA4 KO (clone A2) cells. Data are shown as mean ± SD. ns, no significance; ***P < 0.005; ****P < 0.0001 using one-way ANOVA.

We also examined how overexpression of the full-length C15ORF48 and its ΔC11 mutant impacts inflammation signaling in NDUFA4 KO cells. In contrast to a complete rescue of IL1β-driven activation of OAS1 and NOS2 by NDUFA4 (SI Appendix, Fig. S10C), we observed a partial rescue of OAS1 by both C15ORF48 and ΔC11 mutant, whereas the induction of TNFα and NOS2 was not significantly affected (Fig. 6B and SI Appendix, Fig. S10D). The specific rescue of the mitochondrial respiration defect, but not the innate immune defect by ΔC11 mutant OE in NDUFA4 KO cells indicates that the C15ORF48/miR-147-NDUFA4 axis regulates inflammation and cell metabolism through different mechanisms (SI Appendix, Fig. S10E).

Gut Dysbiosis Mediates Fulminant Colitis in C15ORF48/miR-147^−/− Mice.

Because the gut microbiota plays an important role in the pathogenesis of inflammatory GI diseases (35, 36), we examined how changes in gut microbiota impact aberrant inflammation in C15ORF48/miR-147^−/−mice. To this end, C15ORF48/miR-147^−/−and WT mice were treated with broad-spectrum antibiotics for 4 wk and then challenged with DSS. Analysis of body weight changes revealed no significant protection against excessive inflammation in WT mice (Fig. 7A). Furthermore, cohousing of C15ORF48/miR-147^−/− and WT mice for 4 wk yielded a similar experimental outcome (Fig. 7B), suggesting that gut dysbiosis is potentially mediating fulminant colitis in C15ORF48/miR-147^−/− mice. To determine the underlying dysbiotic changes, we performed 16S ribosomal RNA gene sequencing (16S rRNA-seq) of fecal DNA isolated from separately housed C15ORF48/miR-147^−/−and WT mice. Unsupervised principal coordinates analysis (PCoA) of Bray–Curtis distance indicated that C15ORF48/miR-147^−/−and WT microbiomes are very dissimilar (different beta diversity) (Fig. 7C). In contrast, no significant difference (P = 0.92) in the Shannon diversity index (alpha diversity; measures microbiome heterogeneity in each mouse strain) was observed (SI Appendix, Fig. S11A). Application of the Linear discriminant analysis Effect Size (LEfSe) method to define the most significant taxonomic changes in the two mouse strains, revealed just five operational taxonomic units (OTUs) (Lachnospiraceae_NK4A136, Dubosiella, Eubacterium xylanophilum, Turicibacter, and Blautia) that are significantly decreased in C15ORF48/miR-147^−/− mice (Fig. 7D and SI Appendix, Fig. S11B). The first three OTUs on this list are relatively abundant and comprise close to 20% of the WT gut microbiome (SI Appendix, Fig. S11 B and C). In C15ORF48/miR-147^−/− mice, the relative abundance of these OTUs plummets ~10-fold. The bacteria in the five OTUs are obligate anaerobes and belong to the Firmicutes phylum, which is known to play a protective, anti-inflammatory role in the gut (37–39). Of note, the Firmicutes to Bacteroidetes abundance ratio in C15ORF48/miR-147^−/− mice is trending lower (~50% drop; SI Appendix, Fig. S11D), mirroring findings observed in IBD patients (40). Because the most significant taxonomic changes in the C15ORF48/miR-147^−/− microbiome were depletions of bacterial OTUs rather than expansions, it is unlikely that fecal transfer of bacteria from C15ORF48/miR-147^−/− mice to cohoused WT mice can explain the exacerbation of colitis phenotype in WT mice after cohabitation. Alternatively, this observation suggests that a transfer of microbiota-shaping metabolites or bacteriophages, rather than the bacteria itself, triggers dysbiosis and subsequently drives severe colitis in cohoused WT mice.

Fig. 7.

Gut dysbiosis mediates severe colitis in C15ORF48/miR-147^−/− mice. (A) Body weight changes in antibiotic-treated WT and C15ORF48/miR-147^−/− (KO) mice (n = 5) upon DSS challenge. Mice were treated with broad-spectrum antibiotics (ampicillin: 1 g/L; neomycin: 1 g/L; metronidazole: 1 g/L; vancomycin: 0.25 g/L) in drinking water for 4 wk before colitis induction (3% DSS treatment for 5 d). (B) Body weight changes in single-house and cohoused WT and C15ORF48/miR-147^−/− (KO) mice (n = 5) upon DSS challenge. After being cohoused for 4 wk, mice were subsequently separated by their genotypes prior to inducing colitis using a 2% DSS treatment for 6 d. (C) Unsupervised PCoA of Bray–Curtis distance between WT and C15ORF48/miR-147^−/− (KO) fecal microbiomes (n = 5). (D) Linear discriminant analysis Effect Size (LEfSe) of significant taxonomic changes (LDA score > 2, P < 0.05) between WT and C15ORF48/miR-147^−/− (KO) microbiomes. LDA, linear discriminant analysis. Data in (A) and (B) are shown as mean ± SD. ****P < 0.0001 by one-way ANOVA.

Discussion

Despite recent recognition of C15ORF48/miR-147 as an important innate immunity regulator, its precise contributions to host homeostasis and inflammatory responses remain controversial and poorly understood. Here, we report that C15ORF48/miR-147 is abundantly expressed in IECs and functions as an essential cell-intrinsic regulator of gut inflammation and microbiome homeostasis. These functions of C15ORF48/miR-147 are mechanistically linked to its capacity to regulate mitochondrial metabolism at least in part by controlling NDUFA4, an auxiliary subunit of mitochondrial CIV. In line with these observations, we found that depletion of NDUFA4 in IECs results in attenuation of energy metabolism and inflammatory signaling. Collectively, our genetic studies establish the C15ORF48/miR-147-NDUFA4 molecular axis as an important regulator of cellular metabolism and intestinal homeostasis (SI Appendix, Fig. S12).

Our findings indicate that C15ORF48/miR-147 acts as a negative regulator of inflammation because its genetic ablation strongly exacerbates DSS-induced colitis in mice. Our results are in good agreement with a previous in vitro report demonstrating an anti-inflammatory role of C15ORF48/miR-147 in human aortic endothelial cells during viral infection (16). Conversely, Clayton et al. reported that C15ORF48/miR-147 can promote inflammation in primary human macrophages in response to LPS (17). These incongruous findings indicate that C15ORF48/miR-147 might play distinct roles in innate immune responses depending on the cellular context and/or inflammatory stimulus. Indeed, our BM transplantation experiments provide support for this notion: we found that C15ORF48/miR-147 in nonhematopoietic cells (most likely IECs) protects mice from aberrant gut inflammation, whereas immune C15ORF48/miR-147, in contrast, promotes it. Nevertheless, further research using conditional KO mice will be required to confirm the tissue-specific functional duality of C15ORF48/miR-147 in inflammation.

We posit that C15ORF48/miR-147 elicits its regulatory functions in gut inflammation primarily by silencing NDUFA4. The two distinct molecular products of C15ORF48/miR-147—C15ORF48 protein and miR-147-3p miRNA—act cooperatively to posttranscriptionally suppress NDUFA4. Several lines of evidence support this hypothesis. First, analysis of C15ORF48/miR-147 KO colonocytes revealed significant up-regulation of NDUFA4 mRNA and protein expression. Importantly, NDUFA4 was the only putative miR-147-3p target that was derepressed in C15ORF48/miR-147 KO colonocytes. Second, C15ORF48/miR-147 overexpression in HCT116 colonocytes drives profound NDUFA4 suppression, which is largely elicited by C15ORF48-mediated NDUFA4 protein degradation. In addition, we established NDUFA4 as a bona fide miR-147-3p target using the 3′UTR luciferase assay; however, the contribution of miR-147-3p to NDUFA4 down-regulation is relatively minor in comparison with C15ORF48, in line with the fine-tuning role typically ascribed to miRNAs. Third, we found that gene expression changes, including down-regulation of proinflammatory and mitochondrial metabolism signaling, in NDUFA4 KO IECs are very congruent to the gene expression changes in C15ORF48-expressing cells that undergo a marked NDUFA4 silencing. On the other hand, C15ORF48/miR-147 KO colonocytes display an up-regulation of proinflammatory and mitochondrial metabolism genes that is nicely correlated with the observed NDUFA4 derepression in these cells. The notion of NDUFA4 as a principal target of C15ORF48/miR-147 is also supported by the findings of two other reports that were published while our study was in preparation (16, 17). Despite the accumulating evidence of the importance of the C15ORF48/miR-147-NDUFA4 molecular axis, our data do not completely rule out the possibility that C15ORF48/miR-147 has additional targets through which it elicits its regulatory functions in the gut and other tissues. It is worth noting in this regard that miR-147-3p was previously reported to target HOXC6 and SLC22A3, and thus suppress cell proliferation and inflammatory responses, respectively (41, 42).

Our analysis of C15ORF48-mediated NDUFA4 silencing suggests that this process is transcription independent and does not require proteasomal activity. Furthermore, we found that a unique C-terminal α-helical domain that is present in C15ORF48 but not in its NDUFA4 paralog is necessary to trigger NDUFA4 degradation. Thus, we hypothesize that the incorporation of structurally distinct C15ORF48 paralog into mitochondrial CIV potentially displaces the NDUFA4 subunit and makes it susceptible for a protease attack. The nature of the protease that degrades NDUFA4 and the mechanism of its activation is currently unclear. However, the mitochondrion is known to harbor a number of organelle-specific mitoproteases, some of which (e.g., OMA1 and LONP1) were previously implicated in the control of NDUFA4 expression under stress conditions (43, 44).

The ability of C15ORF48 protein to potentially replace NDUFA4 in CIV raises a possibility that such paralog swap may result in the formation of a non-canonical CIV with distinct metabolic properties. Indeed, we and others have found that C15ORF48 overexpression results in significant dampening of mitochondrial OXPHOS (16). However, it is not clear from these experiments whether this energy metabolism change is driven by the incorporation of C15ORF48 into CIV or is simply the consequence of NDUFA4 loss. To address this question, we ectopically expressed C15ORF48 in NDUFA4-deficient IECs and found a partial restoration of the OXPHOS defect, specifically improving basal mitochondrial respiration capacity, while the maximal respiration capacity remained impaired in the reconstituted cells. Moreover, C15ORF48 overexpression also failed to rescue the compromised inflammatory responses in NDUFA4-depleted cells. Thus, our results suggest that the incorporation of C15ORF48 into mitochondrial CIV has limited effect on its activity and indicate that the principal function of C15ORF48 is the control of NDUFA4 expression. Besides, these findings provide additional support for the notion that NDUFA4 is the primary target through which C15ORF48/miR-147 mediates its control of mitochondrial metabolism and inflammatory responses.

Growing evidence suggests that NDUFA4 plays an important role in regulating CIV activity and by this virtue in controlling mitochondrial respiration (16, 19). However, its role and the mechanisms of action in regulating innate immune responses are still poorly understood. Loss of NDUFA4 in human-induced pluripotent stem cells was recently shown to induce mitochondrial stress, leading to mitochondrial DNA leakage and up-regulation of type I interferon responses (45). In our study, we found that NDUFA4 deletion in human IECs and mouse macrophages impairs activation of inflammatory responses, including production of mitochondrial ROS, nitric oxide, and proinflammatory cytokines. We also observed a significant defect in NF-κB activation in NDUFA4 KO cells that could explain the dysregulated innate immune responses. One hypothetical mechanism through which NDUFA4 can control activation of inflammatory signaling is the generation of immunomodulatory metabolites. For example, the dysregulation of CIV activity due to NDUFA4 loss may result in the accumulation of ROS, which is known to promote LPS-induced NF-κB activation and proinflammatory cytokine release (14). Indeed, we observed that NDUFA4 KD macrophages treated with LPS display a defect in ROS production, which becomes apparent several hours poststimulation and, therefore, is unlikely to mediate the NF-κB activation defect that has a much faster kinetic. In addition, our data suggest that overexpression of ΔC11 mutant (NDUFA4 mutant lacking C-terminal α-helix) in NDUFA4 KO IECs can effectively rescue the OXPHOS defect but has no effect on the impaired cytokine production in these cells, implying that the control of CIV activity and inflammation by NDUFA4 is mechanistically decoupled. Hypothetically, NDUFA4 might be directly involved in the activation of proinflammatory signaling by signal-induced translocation to the cytoplasm and triggering of NF-κB signaling via the inflammasome or IKK complex activation. Small mitochondrial proteins, for example, cytochrome C and DELE1, were shown to translocate to the cytoplasm to activate proapoptotic signaling and signal mitochondrial stress, respectively (46, 47).

C15ORF48/miR-147 was originally described as an inflammation- induced gene. However, we found it to be constitutively expressed in the intestinal epithelium, most likely because the gut exists in the state of low-grade, “physiological” inflammation due to constant exposure to the luminal microbiota. Moreover, our immunofluorescence analysis suggests that C15ORF48 is abundantly expressed in the apical epithelium but is absent in the deep crypt pockets of the mouse colon. This expression pattern may be attributed to the fact that apical colonocytes in contrast to crypt colonocytes are more exposed to the luminal bacteria. Alternatively, an increase in C15ORF48 expression in colon epithelium along the crypt–villus axis may indicate a role for C15ORF48/miR-147 in IEC differentiation: the crypt usually contains stem cells and undifferentiated colonocytes, whereas the apical epithelium is composed of fully matured colonocytes. This notion is supported by a previous report suggesting that miR-147-3p OE suppresses stem-like properties of colon cancer cells (48).

The C15ORF48/miR-147 gene gives rise to two molecular products—C15ORF48 and miR-147-3p—from the same RNA transcript. However, the biogenesis of mature miR-147-3p with its two endonuclease cleavage events likely renders the rest of the C15ORF48/miR-147 mRNA unsuitable for efficient C15ORF48 translation due to a poly-A tail loss and subsequent transcript degradation. Because we found that colonocytes express C15ORF48 along with miR-147-3p, we hypothesize the existence of two separate pools of C15ORF48/miR-147 transcripts: one that gives rise to miR-147-3p and another one that produces C15ORF48. Furthermore, the differential expression of these two molecules in the small intestine (exclusive C15ORF48 expression) and adipose tissues (only miR-147-3p expressed) (Fig. 1 A and B) suggests the presence of a regulatory mechanism that can control selective production of C15ORF48 and miR-147-3p molecules. Given our observation that C15ORF48 and miR-147-3p have distinct roles in the regulation of mitochondrial metabolism, selective expression of these two molecules might be linked to the tissue’s metabolic status.

Our findings imply a link between gut dysbiosis and the development of severe DSS-induced colitis in C15ORF48/miR-147 KO mice. We observed that WT mice cohoused with C15ORF48/miR- 147 KO mice exhibited a more severe colitis compared to single-housed WT mice, suggesting transmission of the fulminant colitis phenotype from C15ORF48/miR-147 KO to WT mice during cohabitation. Notably, we found that the trait transmission occurred prior to colitis development because cohoused WT and C15ORF48/miR-147 KO mice were kept separately after the DSS challenge. In line with these findings, microbiome analysis revealed significant dissimilarities in the fecal bacterial communities of WT and C15ORF48/miR-147 KO mice. Surprisingly, the five significantly different bacterial OTUs (Lachnospiraceae_NK4A136, Dubosiella, Eubacterium xylanophilum, Turicibacter, and Blautia) were more abundant in WT mice in comparison with C15ORF48/miR-147 KO mice. Thus, a fecal transfer of bacteria from C15ORF48/miR-147 KO to WT mice cannot possibly explain the development of severe colitis phenotype in WT mice after cohabitation. An alternative hypothesis that potentially can explain our mouse cohousing experiments is the fecal transfer of microbiome-shaping metabolites or specific bacteriophages from C15ORF48/miR-147 KO to WT mice. Fecal volatile organic compounds (VOCs) resulting from the metabolic activities of the intestinal mucosa and gut microbiota have been reported to contribute to GI disorders and response to treatment in IBD (49, 50). For instance, IBD patients exhibit elevated phenol concentrations in fecal samples compared to healthy controls. High phenol levels promote inflammation and induce proinflammatory cytokine production in murine macrophages. Therefore, fecal VOCs or other metabolites could potentially be responsible for the dysbiosis in C15ORF48/miR-147 KO mice and drive severe colitis in WT mice upon transfer from C15ORF48/miR-147 KO mice during the cohabitation.

In summary, our results have established the C15ORF48/miR-147 as a critical regulator of gut inflammation and microbiome homeostasis. C15ORF48/miR-147 mediates its function by controlling mitochondrial metabolism through NDUFA4 silencing. However, our study not only uncovered an elegant regulatory circuit that connects cellular metabolism and intestinal inflammation but may also open broad avenues for targeted therapies in autoimmune and other GI diseases.

Materials and Methods

Details about the generation of C15ORF48/miR-147^−/− and C15ORF48/miR-147^fl/fl mice are provided in SI Appendix, Materials and Methods. Mouse genotyping using genomic PCR, cell culture, generation of anti-mouse C15ORF48 antibody, induction and clinical evaluation of DSS-induced colitis, colon permeability assay, flow cytometry, histopathological analysis, immunofluorescence analysis, ELISA, colonocyte isolation, Seahorse extracellular flux assay, western blot analysis, bone marrow transplantation, gene expression plasmids, generation of NDUFA4 knockout and knockdown cells, lentiviral production and transduction, dual-luciferase reporter assay, RNA isolation and qRT-PCR, global transcriptome profiling by RNA-seq, coimmunoprecipitation experiments, total ROS measurement, mouse microbiota depletion with antibiotic treatment, mouse cohousing experiments, fecal microbiome analysis, and statistical analysis were performed according to the protocols described in SI Appendix, Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The source data for RNA-Seq have been deposited in the NCBI Gene Expression Omnibus (GEO) database (accession number: GSE246492 and GSE246493) (51).