MicroRNAs Regulate Cellular ATP Levels by Targeting Mitochondrial Energy Metabolism Genes during C2C12 Myoblast Differentiation

In our previous study, we identified an miRNA regulatory network involved in energy metabolism in porcine muscle. To better understand the involvement of miRNAs in cellular ATP production and energy metabolism, here we used C2C12 myoblasts, in which ATP levels increase during differentiation, to identify miRNAs modulating these processes. ATP level, miRNA and mRNA microarray expression profiles during C2C12 differentiation into myotubes were assessed. The results suggest 14 miRNAs (miR-423-3p, miR-17, miR-130b, miR-301a/b, miR-345, miR-15a, miR-16a, miR-128, miR-615, miR-1968, miR-1a/b, and miR-194) as cellular ATP regulators targeting genes involved in mitochondrial energy metabolism (Cox4i2, Cox6a2, Ndufb7, Ndufs4, Ndufs5, and Ndufv1) during C2C12 differentiation. Among these, miR-423-3p showed a high inverse correlation with increasing ATP levels. Besides having implications in promoting cell growth and cell cycle progression, its function in cellular ATP regulation is yet unknown. Therefore, miR-423-3p was selected and validated for the function together with its potential target, Cox6a2. Overexpression of miR-423-3p in C2C12 myogenic differentiation lead to decreased cellular ATP level and decreased expression of Cox6a2 compared to the negative control. These results suggest miR-423-3p as a novel regulator of ATP/energy metabolism by targeting Cox6a2.


Introduction
MicroRNAs (miRNAs) are evolutionarily conserved, short (~22 nucleotides), single-stranded, non-coding RNA molecules that regulate gene expression often by degrading or repressing translation of target mRNAs. miRNAs contain a 5 0 "seed sequence" (nucleotide positions [2][3][4][5][6][7][8] used to predict binding sites on the 3 0 -UTR of target genes; however, the entire miRNA sequence can influence its binding affinity and effects. Functional miRNAs undergo biogenesis exclusion test and resuspended in a lysis buffer from the kit. Cell lysate was mixed with ATP assay solution, and bioluminescence was measured immediately with a DTX 880 Multimode Detector (Beckman Coulter). All measurements were performed in triplicate. Cellular ATP content was calculated from an ATP calibration curve generated from ATP standard solutions.

RNA isolation
Total RNA was extracted from cultured C2C12 cells using Tri-Reagent (Sigma-Aldrich, Germany) followed by an on-column DNase treatment. For small RNA isolation and enrichment, an miReasy Mini kit and an RNeasy MinElute Cleanup kit (Qiagen, Germany) were used Fig 1. Experimental scheme. (A) ATP measurement and expression profiling of microRNA and target mRNA. C2C12 myoblasts were maintained in growth medium (GM) for 2 days (80-90% confluence). Myogenic differentiation was then induced from D0 to D8 by switching to differentiated medium (DM). Cells were harvested for (1) intracellular ATP measurement (2) expression profiling of miRNA (Affymetrix GeneChip miRNA 3.0 Array) and target mRNAs (Qiagen RT2 Profiler PCR Array focusing on mitochondrial energy metabolism genes). (B) miR-423-3p overexpression by transfection of synthetic miR-423-3p mimic. C2C12 myoblasts were transfected with miR-423-3p mimic 24 hours prior to myotube induction (D -1), D3 and D7 post-induction and were then collected on D0, D4, and D8, respectively, for miRNA and target gene qPCR as well as ATP measurement. according to the manufacturer's protocols. RNA quality and quantity were assessed with an Agilent 2100 Bioanalyzer (Agilent) using an Agilent RNA 6000 Nano kit (total RNA) and an Agilent small RNA kit (small RNA).

MicroRNA-microarray analysis
MicroRNA expression profiling was performed using Affymetrix Gene Chip Micro 3.0 Array (Affymetrix, Inc, Santa Clara, CA, USA) containing 16,772 entries representing hairpin precursor (miRBase v17) (in total 19,724 probe sets for detection most of miRNA from 153 species), which provides >3-log dynamic range, with >95% reproducibility and 85% transcript detection at 1.0 amol, for a total RNA input of 130-500 ng. A total of 9 enriched small-RNA pools derived from D0, D4, and D8 post-induction of C2C12 myoblasts (three each) were used in the array hybridizations. Each RNA pool was generated from 5 individual RNA samples extracted from independent cultures. 200 ng of small RNA were used in sample preparation with a Flash-Tag Biotin RNA Labeling Kit for Affymetrix GeneChip miRNA arrays (Genisphere). The labeled RNA was then hybridized for 16 hours to an Affymetrix GeneChip miRNA array according to the manufacturer's recommendations (Affymetrix), washed and stained in the Affymetrix Fluidics Station 450, and scanned on the Affymetrix G3000 GeneArray Scanner. The image files were analyzed using the Affymetrix software (Expression Console), Robust Multi-array Average (RMA) background correction, log-2 transformations and quantile normalization methods implemented in JMP Genomics 5.1 were performed for data pre-processing, normalization, and statistical analysis.

RT 2 Profiler PCR Array
Expression levels of 89 genes functionally associated with mitochondrial energy metabolism were determined using Mouse Mitochondrial Energy Metabolism RT 2 Profiler PCR Array (Qiagen). The array is a pre-optimized qPCR panel of pathway-focused genes in a 96-well plate format including five standard housekeeping genes. Briefly, the real-time PCR was performed according to manufacturer's recommendations using SYBR Green PCR Master Mix and 20 ng (total RNA) cDNA/reaction well on a LightCycler 480 (Roche). The thermal cycler program was 10 min at 95°C, followed by 45 cycles of 15 sec at 95°C and 1 min at 60°C. Dissociation curve analysis was performed immediately after the last PCR amplification cycle.

Validation of microRNA-microarray
Expression differences for differentially-expressed miRNAs identified from microarray analysis (mmu-miR-423-3p, mmu-miR-128-3p, and mmu-miR-301a-3p) were validated by two-step real-time PCR. First, single-stranded cDNA was synthesized from total RNA using a Megaplex RT Primers kit, Rodent Pool Set v3.0 (Life Technologies) containing rodent-specific stem-loop primers of 641 and 373 unique microRNAs for mouse and rat, respectively (see S1 Table for additional primer information). Real-time PCR was performed using a standard LightCycler 480 SYBR Green I Master (Roche) on the LightCycler 480 system (Roche). U6 was used as an internal standard and the relative abundance of miRNAs was calculated using a comparative threshold cycle ΔΔCt method [17]. Validation of mmu-miR-423-3p mimic transfection was carried out using the same method.

mmu-miR-423-3p mimic transfection
Cells were seeded at a density of 1.5×105 cells/well on a 6-well plate in 2 mL of the growth medium. The C2C12 myoblasts were transfected with 150 nM mmu-miR-423-3p miScript miRNA mimic (5'-AGCUCGGUCUGAGGCCCCUCAGU -3', Qiagen) as double-stranded RNA oligonucleotides using the HiPerFect transfection reagent (Qiagen) at three different time points: 1 day prior to myotube induction and D3 and D7 post-induction ( Fig 1B). Transfected myoblasts were harvested 24 hours post-transfection on D0, D4, and D8 post-induction for RNA extraction and qPCR. A random-sequence double-stranded RNA oligonucleotide was used as a negative control in all transfections (Qiagen). Transfection conditions were pre-optimized using the siGLO Green transfection indicator (Thermo Scientific) according to the manufacture's recommendations.

Bioinformatic and statistical analysis
We predicted targets using the computational software RNAhybrid (http://bibiserv.techfak. uni-bielefeld.de/rnahybrid), which detects the most energetically favorable hybridization sites of a small RNA (miRNA) within a large RNA (mRNA) [18,19]. Here, we tested the miRNA probe sets with the following parameters: energy cutoff (mfe) = -20 kcal/mol, allowing the G:U wobble base-pair, 1 mismatch seed, and 5 internal bulging nucleotides in the seed region. Identified targets were located on the 3 0 -UTR of genes. Statistical analysis for microRNA microarray and RT 2 PCR array data was performed using a general linear model on JMP Genomics 6 (SAS Institute Inc., http://www.jmp.com). Adjustment for multiple tests across the Type 3 tests for the fixed effects was calculated using the post-hoc Tukey-Kramer test. For controlling false discovery rate, we chose the FDR according to Benjamini and Hochberg, 1995 [20]. Correlation coefficient analysis (r) between miRNA or mitochondria gene expression and ATP level was performed using SAS version 9.3. The expression data are available in the Gene Expression Omnibus public repository with the GEO accession number GSE52410: GSM1265694-GSM1265710

Luciferase reporter assay
A 256 bp fragment encompassing two miR-423-3p binding sites was amplified from genomic DNA using XhoI-integrated forward primer, 5 0 -gcactcgagaccggttatgagcacccttg-3 0 and NotI-reverse primer, 5 0 -taagcggccgctgcaggtggaaacatcacat-3 0 designed from NM_009943. The fragment was cloned into the multiple cloning region at downstream of the Renilla luciferase reporter gene using the psiCHECKTM-2 vector system (Promega). The construct was validated by Sanger sequencing. For luciferase assay, 100 ng of DNA construct and 50 nM of miR-423-3p mimic were co-transfected into 2 x 10 4 COS-7 cells seeded in a 96-well plate one-day prior to transfection using Attractene Transfection Reagent (Qiagen) according to the manufacturer's recommendation. Firefly and Renilla activities were determined 48 h post transfection using the Dual-Glo Luciferase Assay System (Promega) and a DTX 880 Multimode Detector (Beckman Coulter, Germany). AllStars Negative control (Qiagen) was used in pairwise co-transfection as a negative control and normalization for the effects of endogenous miRNAs.

siRNA transfection of Cox6a2
Synthetic siRNAs were pre-designed by Qiagen. A total of 4 pre-designed siRNAs (Qiagen) complementary to Cox6a2 were tested first and the most effective siRNA was used (Mm_Cox6a2_5). The average values of negative`non-silencing control siRNA´(AllStars Negative Control siRNA, Qiagen),`mock´, and`untreated´were used as control. The experiments were set up in triplicate. Transfection of siRNA was carried out using the HiPerFect transfection reagent (Qiagen) at 150 nM final concentration at three different time points: 1 day prior to myotube induction and D3 and D7 post-induction. Forty-eight hours after siRNA transfection, cells were rinsed 2 times with PBS. One set of the transfected cells was harvested for monitoring the effect of gene silencing. Another set was used for ATP level and ADP/ATP ratio determination. We obtained the level of knockdown of cDNA using quantitative PCR (qPCR) (Roche, Germany) and normalized data using Hprt1 and Ppia as internal controls. Therefore, the geometric means of raw threshold cycle (CT) values of these two genes were used for further calculations. All statistical analyses were performed using two-tailed Student's t-tests.

ADP/ATP Ratio Bioluminescent Assay
After knockdown of Cox6a2, in addition of ATP measurement, the ADP/ATP ratio was also determined by bioluminescent, which can be used as an indicator of cell viability, necrosis and apoptosis. ADP/ATP ration was measured during the induction of myogenic differentiation on D0, D4 and D8 using ADP/ATP Ratio Assay Kit according to the manufacturer's recommendations (Abcam). Briefly, C2C12 myoblasts were grown in 10 mL growth medium in a 75-cm2 flask at a density of 4×10 5 cells/flask. Cultured cells were harvested at 24-hour intervals from D0, D4 and D8 post-induction of siRNA and controls. The nucleotides were released from cell suspensions by adding nucleotide releasing reagent. The ATP levels were measured. After 2 minutes the ADP in the wells was converted to ATP by adding of ADP converted reagent. The ration of ADP/ATP was calculated.

Increased intracellular ATP during C2C12 myotube induction
To identify a critical time point for a functional study of ATP-miRNA regulatory pathways, intracellular ATP content was measured during the myotube induction of C2C12 myoblasts using a luciferase assay. At D0, cells were mononucleated and 80-90% confluent. After 4 days post-induction in low serum medium, they gradually differentiated and, by D8, were predominantly multinucleated myotubes (Fig 2A). Induction of differentiation was confirmed by measuring mRNA expression of myogenic markers Tnnt1, Myh1, and Myh3, all of which were up-regulated ( Fig 2B). Intracellular ATP content was determined in the differentiating cells at an interval of 24 hours during the course of induced differentiation (D0 to D8). ATP level increased gradually from D0-D5, increased sharply from D5-D6, and reached an approximate two-fold increase at D6-D8 compared to D0 ( Fig 2C). ATP levels are significantly different between day 0 and day 8 (p = 0.014) as well as between day 4 and day 8 (p = 0.029), whereas the increment of ATP levels between day 0 and day 4 was subtle and did not reach the significant threshold.

Identification of genes associated with cellular ATP content during C2C12 differentiation
To functionally link miRNAs to cellular ATP production, mRNA targets must be identified. We therefore performed an expression analysis of target genes on the same samples used in miRNA profiling. Here, the RT 2 Profiler PCR Array system containing a panel of 89 mitochondrial energy metabolism-related genes was used. Analysis of this array identified 35 and 49 genes that were differentially expressed at D4 and D8, respectively, compared to D0. Further, a correlation analysis of these differentially-expressed genes with ATP content, performed as with the miRNA analysis, narrowed down the gene list to 13 genes significantly positively correlated with ATP level (r = 0.63-0.92, p-value < 0.05) ( Table 1). Of these, the mRNA abundance of 7 out of 13 genes significantly increased as early as D4 and all 13 genes were ultimately up-regulated at D8. Among them 8 genes showed > 2-fold change at D4 and/or D8 post-myotube induction.

Associations between miRNAs and mitochondrial energy-metabolism genes
A pairwise correlation analysis was carried out between the expression level of miRNAs (5,040 filtered probes) and 89 focused genes derived from the RT 2 PCR array analysis. A Pearson correlation coefficient analysis revealed 4,260 significant correlation interactions (FDR < 0.05). Further filtering criteria were applied to consider only those interactions in which both miR-NAs and target mRNAs were significantly associated with differentiation and ATP level (FDR < 0.1) of C2C12 myotube induction. These resulted in 168 positive and 77 negative correlation interactions. Among these, 14 miRNAs (miR-423-3p, miR-17, miR-130b, miR-301a/b, miR-345, miR-15a, miR-16a, miR-128, miR-615, miR-1968, miR-1a/b, and miR-194) and 6 target mRNAs (Cox4i2, Cox6a2, Ndufb7, Ndufs4, Ndufs5, and Ndufv1) were identified. The miRNAs and mRNAs with high correlation interactions included miR-1 and Ndufv1 (r = 0.988, p = 5.486E-07) and miR-423-3p and Cox6a2 (r = -0.971; p = 1.401E-05). Although most miRNAs inhibit their target mRNAs, some, particularly those miRNAs encoded within mRNAs, are positively correlated with expression of their targets [21]. However, positive correlations could also indicate a connection via as-yet-undetermined indirect regulations. Therefore, we included both positive and negative correlation interactions in an IPA pathway analysis (Fig 6). Additionally, a computational bioinformatic web tool (RNAhybrid) [18,19] was used to scan significant correlation pairs of miRNA-mRNA to obtain additional evidence for their functional links (Table 2). This analysis identified a miR-423-3p as an miRNA of interest for potential interactions with several mRNAs in energy metabolism.
miR-423-3p as an ATP-regulating miRNA candidate Our microarray and correlation analysis indicated that expression of miR-423-3p decreased > 2-fold (FDR < 0.05) and was negatively correlated with the up-regulation of mitochondrial energy metabolism-related genes throughout the course of C2C12 myotube induction. In silico investigation of miR-423-3p showed that the miRNA is highly conserved across several species (Fig 7A) with middle abundance of the miRNA. Moreover, the target genes Cox6a2, Ndufb7, and Ndufs5 in our energy metabolism gene list (see also Table 1) each possess at least one predicted binding site for miR-423-3p in the 3 0 -UTR (RNAhybrid; Fig 7B). To our knowledge, miR-423-3p has not yet been associated with ATP metabolism. Therefore, miR-423-3p was selected for functional validation.

Overexpression of miRNA-423-3p mimic negatively regulated ATP level in C2C12 differentiation
To determine whether miR-423-3p regulates ATP metabolism, synthetic miRNA-423-3p mimics were transfected into C2C12 cells one day prior to differentiation induction and on D3 and D7 post-induction. A random-sequence double-stranded RNA oligonucleotide was used as a negative control and baseline at each time point. Gene expression analysis by real-time PCR for miR-423-3p, Cox6a2, Ndufb7, and Ndufs5 and ATP measurements (see experimental scheme, Fig 1B) were performed 24 hours after transfection. miR-423-3p expression was detected in cells after transfection, and endogenous miR-423-3p expression was significantly lower in the negative control (Fig 8A). ATP levels tended to decrease in miR-423-3p mimictreated cells compared to the negative control at D0 and D4 and were significantly lower at D8 (Fig 8B). Additionally, expression of Cox6a2, Ndufb7, and Ndufs5 was significantly down-regulated at D8 (Fig 8C). All the expression of Cox6a2, Ndufb7, and Ndufs5 in miR-423-3p mimictreated cells experiment was in the tendency lower than control except at D0 of Cox6a2. The fact is that the transcript abundance of Cox6a2 at D0 was very low with about 85 transcripts on average in the mimic experiment compared to control with about 59 transcripts. With these levels of transcripts, which was assumed as not expressed, the observed difference are biologically not relevant. The transcript abundant of Cox6a2 at D4 and D8 ranged from 7000 to 15000 transcripts. These results suggest that miR-423-3p modulates the ATP level partly by regulating the expression of Cox6a2, Ndufb7 and Ndufs5.

MiR-423-3p direct target validation
Luciferase reporter assay was used to experimentally validate Cox6a2, Ndufb7 and Ndufs5 as a direct target gene for miR-423-3p (Table 2). These 3 genes were selected due to its high fold change during myogenic differentiation in C2C12 cells and hence they are likely to contribute a significant impact on ATP regulation. Moreover, two miR-423-3p binding sites spaced by only a 92 bp stretch were predicted on the 3´-UTR of Cox6a2. The luciferase reporter assay results indicated the binding activity of the predicted target sites and its influence on translation of the reporter gene as shown in S1 Table. This data suggests Cox6a2 as a direct target of miR-423-3p. However, our luciferase reporter assay results could not confirm Ndufb7 and Ndufs5 as a direct target of miR-423-3p (data not shown) possibly influenced by background noises of COS-7 endogenous miRNAs and/or universal binding sites residing on the target sequence tested.

Knockdown Cox6a2 and ATP levels
To validate this gene, RNAi was used to knockdown Cox6a2 expression in vitro in the C2C12 murine muscle cell line. Subsequently, relative expression of Cox6a2 and ATP levels as well as ADP/ATP ratio were measured. siRNA targeting Cox6a2 significantly inhibited its expression to 43% (D0), 36% (D4) and 32% (D8) relative to control cells (Fig 9A). Further, Cox6a2 miR-1b UGGAAUGUAAAGAAGUAUGGGU n/a n/a n/a n/a 5 (-23.1 to -18.0) miR-1a UGGAAUGUAAAGAAGUAUGUA n/a n/a n/a n/a 1 (-19.1) miR-1-3p UGGAAUGUAAAGAAGUAUGGAG n/a n/a n/a n/a 2 (-20. The number of binding sites at the 3´-UTR of significant correlated mRNAs is presented with the free energy hybridization of miRNA and target in parenthesis. n/a denotes no binding site available (predicted) at the 3´-UTR. * denotes that binding sites were predicted while the negative correlation between miRNA and mRNA was not significant. inhibition resulted in significantly decreased ATP levels compare to control at D0 (p = 0.0007), D4 (p = 0.027) and-most significantly-on D8 (p = 0.0002) (Fig 9B). ADP/ATP ratios of Cox6a2 knockdown groups were higher than control groups of D0 and D4. At D8, ADP/ATP ratios was significantly higher than controls (p = 0.031) (Fig 9C).

Discussion
Using differentiating C2C12 myoblasts, we identified correlations between miRNAs, target mRNAs, and cellular ATP levels. Importantly, C2C12 myoblasts exhibited changing ATP levels during differentiation, gradually increasing at first (days 0-5), then rapidly (days 6-8) reaching 200% of the baseline level, demonstrating the increased production of ATP during induced differentiation. Previous work on ATP level and myogenic differentiation showed that extracellular ATP has an inhibitory effect on cell proliferation while simultaneously promoting myogenic differentiation of satellite cells [22]. The changing ATP level during myoblast differentiation provided a good model to search for potential regulators of ATP production. In the correlation analyses of miRNA expression changes and ATP level, we first considered down-regulated miRNAs that were negatively correlated with the increased ATP level. Of the miRNAs identified in this group, miR-15a shares its seed sequence with that of miR-15b, which has been reported to modulate cellular ATP levels and degenerate mitochondria by targeting the down-regulation of Arl2 in rat myocytes [23]. miR-301 and miR-17-3p function in mitochondrial metabolism by regulating ATPase and translocase [24]. Consistent with previous findings [25], the expression of miR-140, a chondrocyte-specific marker [26], decreased during myogenic differentiation. MiR-423-3p was found overexpressed in primary laryngeal carcinoma cell [27] and promotes cell growth and cell cycle in hepatocellular carcinoma cells [28]. Recently study reported that miR-423-3p was significantly correlated with the activity of caspase-3, an indicator of apoptosis in brain [29].
On the other hand, the miRNAs that were up-regulated and positively correlated with ATP level during C2C12 myogenic differentiation have not been well-described. Of these, miR-206 and miR-1 are muscle-specific miRNAs with important functions in muscle development, as regulators of cell proliferation and differentiation during myogenesis [30][31][32].
Because mitochondria produce ATP through oxidative phosphorylation to provide energy for cellular activities, like growth and differentiation, our microarray expression analysis focused on mitochondrial energy-metabolism genes. A number of genes were differentially expressed during C2C12 myogenic differentiation; however, only Cox6a2, Ndufb7, Ndufs5,  Ndufv1, Bcs1l, Ndufs4, and Cox4i2 were significantly up-regulated and simultaneously correlated with increased ATP throughout D0-D8.
Cox4i2 and Cox6a2 are nuclear genes localized in the inner mitochondrial membrane and specifically expressed in heart and muscle tissues. The COX complex functions in electron transfer in oxidative phosphorylation, which is responsible for 90% of ATP synthesis for muscle energy [33]. The COX4I2 protein encoded by Cox4i2 is up-regulated in a limited-oxygen environment to increase ATP levels and enhance the efficiency of cellular respiration [34]. Cox6a2, expressed in heart and skeletal muscle, encodes one of thirteen subunits of the respiratory chain complex IV protein, the COX6A subunit. It has been implicated in stimulating enzymatic activity of the functional complex IV and it directly affects ROS production in skeletal muscle [33]. Cox6a2 null mice have severely reduced skeletal muscle complex IV activity, decreased ATP levels, enhanced respiratory uncoupling, and increased thermogenesis, energy expenditure, mitochondrial biogenesis, and muscle-fiber type switching. These changes explain the unique phenotype of the null mice, which includes leanness and resistance to diet-induced obesity [33]. In this study knowndown of Cox6a2 (siCox6a2) in C2C12 muscle cell culture also resulted in decreased ATP levels. Additional, ADP/ATP ratio, an indicator for apoptosis, in siCox6a2 is significant higher than in control at D8.
Ndufb7, Ndufs5, Ndufv1, and Ndufs4 are members of the NADH dehydrogenase (ubiquinone) family of mitochondrial respiratory chain complex I, which represents the largest and first complex mediating electron transfer through the electron transfer chain. These enzymes transfer electrons from NADH to ubiquinone and through the respiratory chain, resulting in the generation of ATP, which is important for energy metabolism [35,36]. Ndufs4 and Ndufs5 are classified in the "iron-sulfur group" with a phosphorylation function. Wheras Ndufb7 is classified in the "hydrophobic group", Ndufv1 is classified in the "flavoprotein group" with NADH-binding and oxidizing properties [36,37].
A correlation analysis (considering both positive and negative relationships) between the expression of miRNAs and mitochondria-related mRNAs during myogenesis was used as the first line of evidence to build a miRNA-mRNA network. A negative correlation interaction between an miRNA and an mRNA could imply a direct regulation; in contrast, a positive correlation interaction could suggest an indirect regulation via (unknown) intermediate pathways or due to neighboring position in the genome. Additionally, positive correlation interactions can indicate direct positive regulation of mRNA transcript abundance by miRNAs [21]. The finding of significant negative correlations of miR-15a and miR-16a with Cox4i2, Cox6a2, Ndufb7, Ndufv1, and Ndufs4 is supported by previous work identifying miR-15 and miR-16 as ATP modulators affecting oxygen consumption [23] as well as direct regulators of Bcl2 to induce apoptosis through the regulation of mitochondrial function [23,38].

Conclusion
Our study provides a list of miRNA and mRNA which are cellular ATP regulators and which target genes involved in mitochondrial energy metabolism. Analyzing each miRNA-mRNA correlation interaction to identify mRNAs targets with 3 0 -UTRs containing potential binding sites for the corresponding correlated miRNAs revealed three mitochondria-related genes (Ndufb7, Ndufs5, and Cox6a2) sharing at least one binding site on the 3 0 UTR for the same miRNA: miR-423-3p. After validation of these predicted targets (Cox6a2, Ndufb7, and Ndufs5), Cox6a2 was confirmed as direct target of miR-423-3p. Additional, knowndown of Cox6a2 resulted in decrease ATP levels and increase ADP/ATP. We therefore concluded that Cox6a2 compose a potential miRNA-mRNA network for cellular ATP production and energy metabolism. miR-423-3p has been previously shown to involve in promoting cell growth and cell cycle progression [28]. This supports our hypothesis that down-regulation of miR-423-3p is required for C2C12 myogenic differentiation. Overexpression of miR-423-3p during C2C12 myogenic differentiation resulted in dramatically lower ATP levels and significantly lowers expression of Cox6a2, Ndufb7, and Ndufs5. Only Cox6a2 could be confirmed as a direct target of miR-423-3p. All together our results suggest for the first time that miR-423-3p modulates ATP production by targeting Cox6a2, as an indicator of cell apoptosis.
Supporting Information S1 Table. Luciferase reporter gene assay using the psiCHECKTM-2 vector system. (DOC)