The Mitochondrial-Derived Peptide MOTS-c Promotes Metabolic Homeostasis and Reduces Obesity and Insulin Resistance

Summary Mitochondria are known to be functional organelles, but their role as a signaling unit is increasingly being appreciated. The identification of a short open reading frame (sORF) in the mitochondrial DNA (mtDNA) that encodes a signaling peptide, humanin, suggests the possible existence of additional sORFs in the mtDNA. Here we report a sORF within the mitochondrial 12S rRNA encoding a 16 amino acid peptide named MOTS-c (mitochondrial open-reading-frame of the twelve S rRNA -c) that regulates insulin sensitivity and metabolic homeostasis. Its primary target organ appears to be the skeletal muscle and its cellular actions inhibit the folate cycle and its tethered de novo purine biosynthesis, leading to AMPK activation. MOTS-c treatment in mice prevented age-dependent and high-fat diet-induced insulin resistance, as well as diet-induced obesity. These results suggest that mitochondria may actively regulate metabolic homeostasis at the cellular and organismal level via peptides encoded within their genome.

Identification of a novel expressed short open reading frame (sORF) encoded within the mitochondrial genome Previously, cDNAs mapping to the 12S rRNA region have been cloned from human myeloblasts after stimulation with interferon, but the exact sequences have not been identified ( Tsuzuki et al., 1983 ). An in silico search for potential sORFs within the human 12S rRNA revealed one consisting of 51 base pairs with a strong Kozak sequence ( Harhay et al., 2005 ) that translates into a 16 amino acid peptide subject to several putative post-translation modifications ( Figures 1A and S1A-B ). We have named this peptide MOTS-c ( m itochondrial o pen-reading-frame of the t welve S rRNA type- c ) (GenBank accession# KP715230 ). MOTS-c peptide translation obligatorily occurs in the cytoplasm using the standard genetic code because mitochondrial translation, using the mitochondria-specific genetic code, yields tandem start and stop codons ( Figure 1A ). This suggests that its polyadenylated transcript ( Figure 1B ) is exported from the mitochondria. Mitochondrial RNAs, both tRNA and rRNA, are exported out of the mitochondria via mechanisms that are still poorly understood ( Amikura et al., 2001 ; Attardi and Attardi, 1968 ; Kashikawa et al., 2001 ; Kashikawa et al., 1999 ; Kobayashi et al., 1998 ; Kobayashi et al., 1993 ; Maniataki and Mourelatos, 2005 ; Nakamura et al., 1996 ). Multiple peptide sequence alignment of 14 species and its deduced phylogenetic tree suggest that MOTS-c is highly conserved, especially the first 11 residues ( Figure 1C-D ). To test if these 11 amino acids evolved over positive selection, we calculated ratios of non-synonymous (dN) to synonymous substitutions (dS), and their rate (dN/dS) for each residue in 14 species. We found that four residues were significantly undergoing positive selection (dN/dS > 1) at positions 4 (Q), 5 (E), 7 (G), and 9 (I), and two residues were undergoing purifying selection (dN/dS < 1) at positions 3 (W) and 10 (F) ( Figure 1E ). Positive selection was also detected by random effects likelihood (REL) analysis implemented in the HyPhy package with posterior probability greater than 0.95 and Bayes factor greater than 50 ( Figure S1C ) ( Pond and Frost, 2005 ). All further studies were performed based on the human MOTS-c sequence. We recognized the possibility that MOTS-c could also be of nuclear origin due to a phenomenon known as nuclear mitochondrial DNA transfer (NUMT) ( Ricchetti et al., 2004 ). Using the NCBI nucleotide basic local alignment search tool (BLASTN 2.2.29+) ( Altschul et al., 1997 ), we found that none of the putative NUMT-derived peptides shared complete homology with mitochondrial encoded MOTS-c ( Figure S1D ). Furthermore, a BLAST search using the human expression sequence tag (EST) database revealed 101 hits whose mRNA sequences were all completely homologous to the mitochondrial 12S rRNA locus ( Table S1 ). Interestingly, rats ( Rattus Norvegicus ) do not have any NUMT sequences for MOTS-c, making mitochondrial DNA its exclusive source. Because of the lack of methods for targeted knock-out or knock-down of mitochondrial genes, to further confirm its mitochondrial origin in humans, we selectively depleted mitochondrial DNA in HeLa cells (HeLa-ρ0) ( Hashiguchi and Zhang-Akiyama, 2009 ) and show the elimination of both 12S rRNA and MOTS-c transcripts ( Figure 1F ). HeLa-ρ0 cells had undetectable levels of mitochondrial-encoded cytochrome oxidase I and II (MT-COI/II) and MOTS-c, measured using specific MOTS-c antibodies ( Figure S1E ), whereas nuclear-encoded GAPDH expression was unaltered ( Figure 1G ). HeLa and HEK293 cells showed a certain degree of mitochondrial co-localization ( Figures 1H-I and S1F ); HeLa-ρ0 cells showed loss of MOTS-c expression and mitochondrial co-localization ( Figure 1H ). Furthermore, we also selectively depleted mitochondrial RNA using actinonin ( Richter et al., 2013 ), and show loss of MOTS-c and MT-COI expression in a time-dependent manner ( Figures 1J and S1H ). Therefore, using the ρ0 system and actinonin, we provide evidence of mitochondrial origin for MOTS-c at the DNA and RNA level, respectively. MOTS-c was detected in various tissues in mice and rats ( Figure 1K ), as well as in circulation in human and rodent plasma as determined with a MOTS-c specific ELISA ( Figures 1L and S1G-H ); again, NUMTs are absent in the rat making the mtDNA the only possible source of MOTS-c. Notably, fasting lowered endogenous expression of MOTS-c in certain metabolically active and mitochondria-rich tissues (skeletal muscle and testes) as well as in plasma, whereas homeostatic tissues (such as brain and heart) showed sustained levels ( Figure 1M-N ).

We identified a distinct metabolic signature from our global unbiased metabolomic profiling in MOTS-c-ST cells that strongly suggested its target of action as the folate-methionine cycle and the directly tethered de novo purine biosynthesis pathway ( Figure 3A ). We overlaid our microarray analysis with our metabolomics profiling and found that MOTS-c altered gene expression of enzymes involved in the folate-methionine cycle and de novo purine synthesis as early as 4 hours post-treatment ( Figure 3A-B ). We observed a concerted decrease in the levels of 5-methyl-tetrahydrofolate (5Me-THF), the most abundant form of activated folate, and methionine and elevated levels of homocysteine in MOTS-c-ST cells ( Figure 3A, C ). Notably, 5Me-THF levels declined prior to changes in methionine-homocysteine in HEK293 cells treated exogenously with MOTS-c, suggesting the regulation of the folate cycle to occur earlier ( Figure S2A ). Also, as expected, 5Me-THF depletion was coupled with the blockade of de novo purine biosynthesis ( Figure 3A, D ), resulting in an accumulation of endogenous AICAR (5-aminoimidazole-4-carboxamide ribonucleotide) to levels higher than 20-fold in MOTS-c-ST cells compared to control cells ( Figure 3A, E ), causing an expected significant decrease in purines ( Chan and Cronstein, 2010 ); this was also observed, but to a lesser extent, 72 hours after exogenous MOTS-c treatment ( Figure S2B ). Because de novo purine synthesis is feedback-regulated by purine products, our data suggest accelerated de novo purine synthesis in MOTS-c-ST cells indicated by increased levels of its precursor metabolites NAD + , glycolysis, and the pentose phosphate pathway (PPP), as discussed below, simultaneously with reduced levels of ADP-ribose and R5P ( Figures 3A, D and S2C ). A well described role of AICAR is to activate AMPK and stimulate fatty acid oxidation via phosphorylation-induced inactivation of acetyl-CoA carboxylase (ACC) that consequently alleviates allosteric inhibition of carnitine palmitoyltransferase 1 (CPT-1), and also enhance glucose uptake in muscle ( Steinberg and Kemp, 2009 ). MOTS-c treatment led to the phosphorylation of AMPKα (Thr172) and Akt (Ser473) in a time and dose-dependent manner ( Figure 3F ). Further, after 72 hours of MOTS-c treatment, increased phosphorylation of AMPKα2 (Thr172) and ACC (Ser79), and elevated CPT-1 protein levels were detected ( Figure 3G ). Notably AMPK activation occurred despite lower AMP levels and higher ADP and ATP levels ( Figure S2D ) similar to that achieved by salicylate ( Hashiguchi and Zhang-Akiyama, 2009 ), leptin ( Ohta et al., 2009 ), and metformin ( Fujii et al., 2006 ). Consistent with this notion, the folate cycle, specifically 5Me-THF, has recently been shown to be a target of the AMPK-activating drug metformin ( Cabreiro et al., 2013 ; Corominas-Faja et al., 2012 ). We also confirmed elevated levels of MOTS-c and phosphorylation of AMPKα2 (Thr172) in MOTS-c-ST cells ( Figure 3H ). We next investigated the effect of MOTS-c on relevant cellular metabolism including glycolysis, mitochondrial function, and fatty acid oxidation. MOTS-c appeared to stimulate glucose utilization evidenced by increased glucose clearance and lactate accumulation in culture media ( Figures 4A-B and S3A-C ) coupled with decreased intracellular glucose levels, along with other glycolytic intermediates ( Figure 4C ). Similar findings were observed in HEK293 cells within 24 hours of MOTS-c treatment ( Figure S3D ). Notably, our data suggest increased routing of glucose to the pentose phosphate pathway (PPP), the anabolic branch of glycolysis that provides ribose-5-phosphate as a precursor for de novo purine synthesis, as indicated by reduced intermediates leading to the accumulation of AICAR in MOTS-c-ST cells and also in HEK293 cells after exogenous MOTS-c treatment ( Figures 4C and S3D ). To directly test glycolytic flux in real-time, we measured extracellular acidification rate (ECAR). Upon glucose stimulation MOTS-c-ST cells showed significantly enhanced glycolytic response, reaching maximum glycolytic capacity only typically achieved after oligomycin treatment ( Figure 4D ). Because MOTS-c targets the folate cycle and depletes intracellular 5Me-THF levels, we were able to reverse the enhanced glycolytic response in MOTS-c-ST cells by supplementing the culture media with folic acid (100nM) ( Figure 4D ); similar results were observed in HEK293 treated with MOTS-c ( Figure S3E ). To confirm the specificity of the MOTS-c sequence, we substituted 2 highly conserved residues to alanine and created null mutants that did not show enhanced glycolytic response to glucose stimulation in HEK293 cells. These mutants were glutamic acid in position 5 (E5A) and glycine in position 7 (G7A) ( Figure S3F ). Furthermore, we generated a randomly scrambled MOTS-c peptide sequence that did not have any primary sequence homology to known peptides/proteins (M3S1) and showed lack of enhanced gly

As MOTS-c enhanced cellular glucose flux in vitro and acute treatment reduced glucose levels in mice fed a normal diet, we hypothesized its actions in vivo would be related to glucose handling and insulin sensitivity. We treated mice with intraperitoneal injections of MOTS-c for 7 days and then subjected them to a glucose tolerance test (GTT), and found significantly enhanced glucose clearance indicative of improved insulin sensitivity ( Figure 5A ). Next, we performed hyperinsulinemic-euglycemic clamp studies to quantify the effects of a 7-day MOTS-c treatment on whole body insulin sensitivity independent of changes in body weight that occur with extended treatment durations ( Figure 5B-D and Table S3 ). MOTS-c improved whole body insulin sensitivity as reflected by a ∼30% increase in the exogenous glucose infusion rate (GIR) required to maintain euglycemia during insulin stimulation ( Figure 5B ). Insulin promotes glucose disposal into peripheral tissues and suppresses hepatic glucose production (HGP) to maintain homeostasis during periods of increased glucose availability ( Kahn, 1994 ). Tritiated glucose was infused during the clamp to determine the tissue specificity of MOTS-c action on insulin sensitivity. We observed that MOTS-c treatment significantly enhanced the insulin-stimulated glucose disposal rate (IS-GDR) ( Figure 5C ), indicative of enhanced skeletal muscle insulin sensitivity, but the rate of hepatic glucose production (HGP) was comparable between the groups ( Figure 5D ). The insulin-stimulated skeletal muscles were collected at the end of the hyperinsulinemic-euglycemic clamp. MOTS-c treated mice showed enhanced ability of infused insulin to activate skeletal muscle Akt, concurrent with increased detection of MOTS-c ( Figure 5E ). Considering that 70-85% of insulin-stimulated glucose disposal is into skeletal muscle, MOTS-c actions to enhance insulin sensitivity and glucose homeostasis are likely mediated in this tissue. Because MOTS-c levels in skeletal muscle ( Figure 5F ) and in circulation ( Figure 5G ) decline concomitantly with the development of insulin resistance during aging in mice, we tested if MOTS-c could reverse age-dependent impairments in insulin action by measuring insulin-stimulated glucose (2-deoxyglucose) uptake into soleus muscles of middle-aged (12 months) and young (3 months) male C57BL/6 mice. Reduced insulin sensitivity becomes apparent around one year of age in C57BL/6 mice ( Pearson et al., 2008 ). Muscles from older (12 months old) mice were more insulin resistant than younger (3 months old), but 7 days of MOTS-c treatment restored sensitivity in the old mice to levels comparable to young animals ( Figure 5H ). To examine the glycolytic effects of MOTS-c on muscle in vitro, we generated L6 rat myocytes that stably over-express MOTS-c (L6-MOTS-c-ST). Following differentiation to mature myotubes, we measured glucose levels in the media (steady-state) and glucose-stimulated and maximum glycolytic capacity (flux). Similar to the MOTS-c-ST (HEK293) cells, L6-MOTS-c-ST cells showed accelerated media glucose clearance ( Figure 5I ), and enhanced glucose-stimulated and maximum glycolytic rate ( Figures 5J-K ).

Novel Peptides Derived from Mitochondrial sORFs as Signaling Molecules Our understanding of the molecular mechanisms involved in the communication signals from the mitochondria of particular cells to the rest of those cells and to distant organs is rapidly evolving. Signals that are sent from the mitochondria to the cell are largely categorized as mitochondrial retrograde signals and have been thought to consist of nuclear-encoded proteins that reside in mitochondria, secondary and transient metabolites, and fragments of damaged mitochondrial DNA (mtDNA). Notably, in the nematode C. elegans , a stress-induced tissue-specific signal originating from mitochondria has been shown to communicate with distant organs and extend lifespan; the exact identity of this signal, termed ‘mitokine’, is unknown and currently being investigated ( Durieux et al., 2011 ; Woo and Shadel, 2011 ). The demonstration of MOTS-c as a peptide encoded as a sORF in the mtDNA that is metabolically and developmentally regulated, which has endocrine like effects on muscle metabolism, insulin sensitivity, and weight regulation, establishes its role as a member of a new class of mitochondrial signals which can be called “MDPs” or mitochondrial-derived peptides ( Lee et al., 2013 ). The MDP humanin similarly acts in a systemic fashion to protect neuronal and vascular systems from disease processes and toxic insults ( Cohen, 2014 ; Lee et al., 2013 ). Technological advances have unveiled previously unknown properties of mitochondrial genetics that suggest the existence of sORFs in the mtDNA ( Mercer et al., 2011 ). MOTS-c may be a product of adaptation for effective bilateral communication between mitochondria in energetically sensitive tissues and cells of distant organs such as skeletal muscle. We provide here experimental evidence that MOTS-c, like humanin, is a mitochondrial signaling peptide encoded within the mtDNA that regulates global physiology ( Lee et al., 2013 ).

Mitochondria are thought to sense the cellular energetic status and are well recognized to modulate carbohydrate metabolism through incompletely characterized mechanisms ( Woo and Shadel, 2011 ). We propose MOTS-c as a novel key endocrine signal that originates from mitochondria and systemically regulates in vivo glucose metabolism and muscle insulin action. MOTS-c also shares some physiological similarities to the first-line diabetes drug metformin in terms of regulating glucose utilization, mitochondrial and fatty acid metabolism, and body weight ( Ferguson et al., 2007 ) which is thought to occur by targeting the folate cycle ( Cabreiro et al., 2013 ; Corominas-Faja et al., 2012 ) and signaling via cAMP ( Miller et al., 2013 ) and AMPK ( Shaw, 2013 ). One key difference between the actions of metformin and MOTS-c is that MOTS-c appears to directly target skeletal muscle as the key site of its activity, while metformin is thought to act primarily on the liver ( Diamanti-Kandarakis et al., 2010 ). Interestingly, the mitochondrial peptide humanin has been shown to regulate carbohydrate metabolism primarily by acting on hypothalamic and islet cells ( Kuliawat et al., 2013 ; Muzumdar et al., 2009 ); this suggests a highly regulated, tissue and context specific targeted MDP effect on whole body metabolism.

Cell Culture HEK293 and HeLa cells were routinely cultured in Dulbecco's modified Eagle's medium (DMEM), and L6 myoblasts were cultured in minimum essential media (MEM) supplemented with 10% fetal bovine serum (FBS) at 37°C with 5% CO 2 . ρ0 cells, which are devoid of mitochondrial DNA, were generated by culturing HeLa cells in low doses of ethidium bromide (EtBr; 100ng/mL) as described before ( Hashiguchi and Zhang-Akiyama, 2009 ). MOTS-c-ST and L6-MOTS-c-ST cells are HEK293 and L6 cells, respectively, stably overexpressing MOTS-c and were generated by transfecting MOTS-c expression clone (described below) followed by selection in and maintenance in G418 (500 μM; Sigma) in DMEM or MEM with 10% FBS. The control cells for MOTS-c-ST and L6-MOTS-c-ST cells were HEK293 or L6 cells, respectively, stably transfected with empty vector (EV) with the same selection and maintenance method. L6 myoblasts were differentiated to mature myotubes by culturing in MEM 2% media, once 80-90% confluence was reached, for 8-10 days (media replaced every 2-3 days).

RNA was isolated using the RNeasy kit (Qiagen, Valencia, CA) and then hybridized to BD-103-0603 Illumina Beadchips. Raw data were subjected to Z-normalization, as previously described ( Cheadle et al., 2003 ). Principal component analysis, performed on the normalized Z-scores of all of the detectable probes in the samples, was performed by using the DIANE 6.0 software: ( http://www.grc.nia.nih.gov/branches/rrb/dna/diane_software.pdf ). Significant genes were selected by the z-test < 0.05, false discovery rate < 0.30, as well as z-ratio > 1.5 in both directions and ANOVA p value < 0.05. Parametric analysis of gene set enrichment (PAGE) was analyzed as previously described ( Kim and Volsky, 2005 ). Gene regulatory network and canonic pathway analysis was performed using Ingenuity Pathways Analysis © (Ingenuity Systems; Redwood City, CA) (N=6). Microarray data are accessible in GEO database under accession code GSE65068 .