Nuclear outsourcing of RNA interference components to human mitochondria

Abstract

MicroRNAs (miRNAs) are small non-coding RNAs that associate with Argonaute proteins to regulate gene expression at the post-transcriptional level in the cytoplasm. However, recent studies have reported that some miRNAs localize to and function in other cellular compartments. Mitochondria harbour their own genetic system that may be a potential site for miRNA mediated post-transcriptional regulation. We aimed at investigating whether nuclear-encoded miRNAs can localize to and function in human mitochondria. To enable identification of mitochondrial-enriched miRNAs, we profiled the mitochondrial and cytosolic RNA fractions from the same HeLa cells by miRNA microarray analysis. Mitochondria were purified using a combination of cell fractionation and immunoisolation, and assessed for the lack of protein and RNA contaminants. We found 57 miRNAs differentially expressed in HeLa mitochondria and cytosol. Of these 57, a signature of 13 nuclear-encoded miRNAs was reproducibly enriched in mitochondrial RNA and validated by RT-PCR for hsa-miR-494, hsa-miR-1275 and hsa-miR-1974. The significance of their mitochondrial localization was investigated by characterizing their genomic context, cross-species conservation and instrinsic features such as their size and thermodynamic parameters. Interestingly, the specificities of mitochondrial versus cytosolic miRNAs were underlined by significantly different structural and thermodynamic parameters. Computational targeting analysis of most mitochondrial miRNAs revealed not only nuclear but also mitochondrial-encoded targets. The functional relevance of miRNAs in mitochondria was supported by the finding of Argonaute 2 localization to mitochondria revealed by immunoblotting and confocal microscopy, and further validated by the co-immunoprecipitation of the mitochondrial transcript COX3. This study provides the first comprehensive view of the localization of RNA interference components to the mitochondria. Our data outline the molecular bases for a novel layer of crosstalk between nucleus and mitochondria through a specific subset of human miRNAs that we termed 'mitomiRs'.

Figure 1. Novel localization of Argonaute 2 (AGO2) to human mitochondria.

A. Purity assessment of mitochondrial fraction. Cyclin-dependent kinase 2 (CDK2) was assessed in the mitochondrial fraction by immunoblot analyses to check for nuclear and cytoplasmic contaminants relatively to mitochondrial protein ATP synthase subunit α (ATP5A1). The density of bands was measured using the ImageJ software and is represented as a relative intensity. Values are means±SD of three independent experiments. B. Western blot analysis of AGO2 in subcellular mitochondrial fraction from HeLa cells. AGO2 protein was detected using a rabbit polyclonal antibody anti-AGO2 in mitochondrial protein fraction (Mito) and total protein extracts (Lysate) from HeLa cells. ATP5A1 was used as a mitochondrial marker, actin as a cytosolic marker, and CDK2 as a nuclear/cytosolic marker. Representative image is shown of three independent experiments. C. Western blot analysis of AGO2 at mitochondrial membranous and soluble fraction from U2OS cells. Mitochondria were suspended either in isotonic mitochondrial buffer (MB) or in hypo-osmotic mitochondrial buffer (MB/10) alone or supplemented either with NaCl 1 M or Na₂CO₃ 0.1 M, and fragilized by freeze-thaw cycles and sonication when indicated. All samples were centrifuged (150,000× g) to separate the membrane pellet (p) from the soluble protein supernatants (s). 15 µg of protein extracts of each sample were subjected to Western Blot. The following proteins were immunodecorated on blots: voltage-dependent anion channel 1 (VDAC1) and NADH dehydrogenase (ubiquinone) 1 α subcomplex 9 (NDUFA9) as markers of mitochondrial membranes and cytochrome c (CYCS), as a intermembrane marker. Immunodetection of AGO2 is shown as the lower panel.

Figure 2. AGO2 protein co-localizes with mitochondria in human cell lines.

Cells were fixed and endogenous AGO2 detected with two distinct antibodies (rabbit polyclonal anti-AGO2, clone 7C6, A, D; mouse-monoclonal anti-AGO2, clone 4G8, B, C) and visualized by indirect immunofluorescence. Mitochondria were either stained with Mitotracker Red CMXRos (E–G) or immunostained with anti-ATP5A1 (H). Overlay images show colocalization of AGO2 and mitochondria (I–N). A sample of 21 cells was examined for dual channel co-localization for each cell type. Co-localization was also confirmed by Pearson's correlation coefficient (rp), whose value is indicated at the bottom of each overlay image. Cross-correlation functions (CCFs) were calculated for each co-localization. Representative plots obtained from the analysis of a single microscopic field are shown (O–R). Maximum CCF and pixel shift (dx) values are respectively indicated at the top and the bottom of each plot. Scale bar, 10 µm.

Figure 3. Co-immunoprecipitation of AGO2 and mitochondrial transcripts.

Either AGO2, or SLUG, which serves as a negative control were co-immunoprecipitated with associated mRNAs in HeLa protein extracts. Coimmunoprecipitated RNA was extracted with Trizol and subjected to RT-PCR amplification with the indicated primers: cytochrome c oxidase III (COX3), cytochrome b (cyt b) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Results are indicative of three independent experiments.

Figure 4. Isolation of mitochondrial and cytosolic miRNA.

A. Schematic workflow of the experimental design. B. Integrity, quality and purity analyses of RNA fractions. Electrophoretic images of mitochondrial and cytosolic RNAs were retrieved from analysis using the Agilent 2100 Bioanalyzer. Bands corresponding to ribosomal RNAs (rRNAs) 16S, 12S, 28S and 18S are indicated when present. Representative image is shown of three independent experiments. C. Purity assessment of mitochondrial RNA fraction. 16S rRNA and GAPDH were amplified by RT-PCR in each fraction as shown by electrophoretic image. GAPDH was assessed in the mitochondrial fraction to check for cytoplasmic contaminant relatively to mitochondrial 16S ribosomal RNA. Representative image is shown of three independent experiments. The density of bands was measured using the ImageJ software and is represented as a relative intensity. Values are means±SD of three independent experiments.

Figure 5. Evidence for a mitochondrial miRNA signature in HeLa cells.

A. Heatmap showing the 57 miRNAs differentially expressed in HeLa cells from mitochondrial and cytosolic RNA. Three independent microarray profiling experiments (microarray n1, n2 and n3) are shown and reproducibly revealed 13 miRNAs enriched in mitochondrial RNA. Log2 Hy5/Hy3 ratios are color scaled in gradient from green (low levels) to red (high levels) as indicated by the scale bar. B. Validation of microarray data by RT-PCR. Five genes (hsa-miR-494, hsa-miR-1974, hsa-miR-1275, 16S rRNA and GAPDH) were selected for microarray data validation of differential expression in either subcellular compartment. 16S rRNA was used as a positive mitochondrial control, while GAPDH as a negative mitochondrial control. Quantitative analysis of band intensities are shown for three independent RT-PCR experiments and indicated as arbitrary units (a.u.) in either mitochondria (grey) or cytosol (black). Error bars represent the standard error of the mean. Asteriscs indicate statistically significant differences as compared to the cytosol (hsa-miR-494, p-value = 3.32×10⁻⁵; hsa-miR-1974, p-value = 0.02; hsa-miR-1275, p-value = 6×10⁻⁴; 16S rRNA, p-value = 0.03).

Figure 6. Ontology enrichment analysis for target genes of hsa-miR-328, hsa-miR-494, hsa-miR-513 and hsa-miR-638.

The ExParser algorithm was used to compile datasets of genes whose expression patterns were comparable and statistically correlated to the expression patterns of the four mitomiRs. The datasets were uploaded into MetaCore™ and analyzed in respect to the Gene Ontology Process. Ten most significantly enriched processes for the genes targeted by hsa-miR-328, hsa-miR-494, hsa-miR-513 and hsa-miR-638 were scored and ranked in respect to the obtained p-values. Bars represent significance as −log(p-value) for hypergeometric distribution. Ontology enrichments were all filtered to allow no more than 5% false discovery rate.

Figure 7. Intrinsic features of mitomiRs versus a control sample of cytosolic-enriched miRNAs.

Values for each miRNA subgroup are shown in blue for the mitomiRs and red for the control miRNAs. Length of mature miRNAs (A) and of pre-miRNAs (B), and values of minimal folding free energy (MFE) (C) and adjusted MFE (AMFE) (D) are plotted for each miRNA from the 2 subgroups. Average values are shown as bars in each subgroup. Asterics indicate significant differences between mitomiRs versus control miRNAs (p-value<0.05). NS indicates not significant p-value.

Figure 8. Integrative model for subcellular localization of RNAi components.

Boxes represent the subcellular compartments. RNAi components are encircled. Arrows indicate the molecular fluxes documented herein in regards to the mitochondrial miRNAs (mitomiRs), and previously in P-bodies (PBs) , stress granules (SGs) , , multivesicular bodies (MVBs) and Golgi . Depending on the localization, different functions are ascribed to AGO2, such as post-transcriptional gene silencing or reversible translational regulation in P-bodies and stress granules , and transcriptional regulation in the nucleus . In mitochondria, the presence of mitochondrial genome adds another possible layer of regulation by AGO2.