Fig 1.
Construction of a mitochondrion-centric multi-tissue gene co-expression network.
A, 935 bona fide MPETs [10] were systematically queried using the CO-Regulation Database (CORD) to identify transcripts with similar expression signatures. The resulting correlate lists were merged to create a network where nodes denote transcripts and edges join transcripts with concordant expression signatures. For any node, the extent of centrality in the network correlates with increasing connectivity. For ease of viewing, only transcripts that correlate at R≥0.8 are displayed. B, Individual correlate lists derived from CORD were concatenated to generate a ranked master list where rank is based on the number of times a given transcript was present across all daughter lists. A running-sum (Enrichment Score) was then generated by descending the list and assigning a positive score (up-step) when a transcript was a seed and a negative score (down-step) when it was not. The barcode denotes positions in ranked list of the seed population. C, Cumulative distribution analysis of the ranked master list from B to quantify the extent of network intra-correlation, expressed as the percentage of the list containing half the seed population, reveals half of the seeds used to build the network were present in the top 13% of the ranked transcript list (seed50 = 13%). D, Comparison of seed50 values across all six major cell compartments following CORD-based network analysis. E, Data from B (mitochondrion) showing the maximum Enrichment Score (Max ES, ~1,400) and resulting leading edge (LE). The red lines denote the ESs calculated following 1000 random walks. F, Distribution of Pgc-1α-responsive transcripts within and beyond the LE sub-set. Matrix displays the number of transcripts within (+) and outside (-) the LE, regulated (+) or not (-) by Pgc-1α and the bar chart displays relative abundance. G, Distribution of proteins not documented as being mitochondrially-localised but which purportedly interact with known mitochondrial proteins, according to publicly accessible proteomics data, within and beyond the LE sub-set. Matrix displays the number of proteins within (+) and outside (-) the LE, purported to interact with known mitochondrial proteins based on analysis of BioGRID protein-protein interaction (PPI) data [40] and the bar chart displays relative abundance. MC+ and MC- denote MitoCarta and non-MitoCarta proteins, respectively. For example, there were 574 transcripts in the LE whose protein products have not been documented to produce mitochondrial proteins but which have been reported to interact with known mitochondrial proteins. The P values in F and G correspond to the result of a Chi-squared test and a Fisher Exact Test (FET), respectively.
Fig 2.
Weighted gene co-expression network analysis (WGCNA) of a human cardiac gene expression dataset and nomination of proteins potentially involved in mitochondrial and cardiac function.
A, Dendrogram representation of the 27 transcript modules derived from WGCNA of a human heart failure microarray dataset (comprising healthy (n = 16), ischemic heart failure (n = 86) and idiopathic heart failure (n = 108) samples). Horizontal lines (branches) represent modules and vertical lines (leaves) represent transcripts. Asterisk denotes an MPET-enriched (green) module. B, Green module from A where transcripts are represented as nodes (size denotes matrix membership (see Methods)) and edges denote significant transcript-transcript correlations. C, Schema illustrating candidate selection following CORD and WGCNA studies. Genes nominated by both analyses were re-processed with CORD to recover co-expressed genes and KEGG terms associated with those genes. In this illustrative example, all genes that correlate with the Recurrent gene apart from Gene b are linked to Hypertrophic Cardiomyopathy suggesting that the Recurrent gene may also be functionally linked. D, Scatter plot showing the sub-set of recurrent genes (24 of 78) that correlate with a disproportionately large number of genes linked to Hypertrophic Cardiomyopathy (HCM) and Dilated Cardiomyopathy (DCM). X and Y axes denote the EASE scores (a measure of ontological enrichment [44]), expressed as -log10 P values, corresponding to DCM and HCM KEGG terms, respectively. Grey dashed lines denote the P≥0.05 EASE threshold.
Fig 3.
LRRC2 localizes to the mitochondria and is regulated by Pgc-1.
A, HEK293 cells expressing a LRRC2-GFP fusion protein were incubated with MitoTracker Red and then imaged via confocal microscopy (left). Chart displays signal co-occurrence (green, LRRC2; red, MitoTracker Red) along the dashed line in unmerged images. Scale bar = 20μm. B, Approximate positions of potential ERRα binding sites in rat, mouse and human LRRC2 genes (top panel), Lrrc2 mRNA abundance in C2C12 and H9c2 cells following infection with either a GFP- or Pgc-1α-encoding adenovirus (bottom-left panel), and Lrrc2 mRNA abundance in the cardiac-specific inducible Pgc-1α/β double-knockout (DKO) mouse 3 weeks post-tamoxifen (TX) treatment (bottom-right panel). C, Transcript abundance of Lrrc2 and control transcripts (36b4, Tbp and Hprt) as well as Pgc-1α/β in C2C12 cells following adenovirus-mediated over-expression of Pgc-1α and Pgc-1α/β. D, Transcript abundance of Lrrc2 and control transcripts (36b4, Tbp and Hprt) as well as Pgc-1α/β in hearts of cardiac-specific Pgc-1α/β double knockout (DKO) mice and Cre only control mice. All QPCR data are represented as means ± s.e.m and are derived from three independent experiments. *, P≤0.05.
Fig 4.
LRRC2 modulates mitochondrial and cardiomyocyte function.
A, Lrrc2 mRNA abundance in PBS- and ET-1-treated neonatal rat ventricular myocytes (left panel), and left ventricle of mouse (middle panel) and rat (right panel) 4 weeks post-aortic banding compared to sham-operated controls. B, Correlation in gene expression between LRRC2 and hypertrophic biomarker MYH7 (β-MHC) in the human heart (left panel), covariate-corrected LRRC2 mRNA abundance in myocardial samples derived from dilated cardiomyopathy (DCM) patients (n = 97) and non-failing controls (n = 108) (middle panel) (P value calculated via Wilcoxon rank sum test), and effect of Lrrc2 loss-of-function in H9c2 cells on expression of hypertrophic biomarkers Bnp, Anf, and Et-1 (right panel). C, Effect of siRNA-mediated Lrrc2 loss-of-function in H9c2 cells on the abundance of total (left panel) and functional (middle panel) mitochondrial abundance, and mitochondrial protein-encoding transcript expression (right panel). All QPCR data are represented as means ± s.e.m and are derived from three independent experiments. *, P≤0.05; **, P≤0.001.