(A) Seed-matched genes are up-regulated and mitochondrial genes are down-regulated after inhibition of miR-122. Shown are the liver-expression profiles of mice treated twice weekly with anti-miR-122 for 4 weeks (top four profiles) or 1 week (bottom four profiles). Transcripts regulated in at least one experiment with P-value of regulation <0.05 are shown. Boxes identify consensus up-regulated and consensus down-regulated transcripts. Consensus up-regulated transcripts were enriched for 3′UTR matches to CACTCC, the central miR-122 seed hexamer. Consensus down-regulated genes were enriched for mitochondrially localized gene products. (B) Correlation to miR-122 in human HCC predicts ortholog regulation by anti-miR-122 in mouse livers. Shown is the relationship between transcript correlation with miR-122 levels in HCC and adjacent liver tissue and changes in expression in HCC tumor and adjacent non-tumor tissues, and in anti-miR-122-treated mice. Human/mouse ortholog pairs are binned by the correlation coefficient of the human ortholog with miR-122, at a rate of ∼1000 human orthologs per bin. The ortholog pairs within each bin are then restricted to those whose mouse ortholog is regulated with P<0.05 by anti-miR-122. For each bin, the mean correlation coefficient of the human ortholog is plotted against the mean-expression ratios in both human and mouse contexts. Left axis/blue dots: the log10 ratio of mean tumor expression to mean adjacent tissue expression is shown for the human orthologs in each bin. Right axis/red dots: the mean log10 expression ratio of regulation by anti-miR-122 is shown for the mouse orthologs in each bin. Purple box: putative primary miR-122 targets; blue box: putative secondary miR-122 targets. Source data is available for this figure at www.nature.com/msb.

(A) Target-focused human/mouse network of miR-122-associated genes. Genes in this network were either putative miR-122 primary targets up-regulated by anti-miR-122 in mouse livers and negatively correlated with miR-122 expression in human HCC, or putative miR-122 secondary targets down-regulated by anti-miR-122 in mouse livers and positively correlated with miR-122 in human HCC. Functional similarity (Yu et al, 2007) connections are shown between putative primary and secondary targets. Thicker blue lines emphasize the connections to PPARGC1A, the putative secondary target with the most connections to primary targets. Color of gene ovals indicates direction of correlation with miR-122 in HKU patient samples (magenta, negative correlation; cyan, positive correlation). (B) Over-expression of miR-122 in HCC enhances PGC-1α protein expression. PLC/PRF/5 cells were either mock transfected, transfected with vector alone, or transfected with vector expressing miR-122. Top panel: western blot of PGC-1α protein expression in the transfected cells. Bottom panel: expression levels of miR-122 in the transfected cells, as measured by qPCR. (C) Altered miR-122-expression level affects lactate accumulation in PLC/PRF/5 cells. Transfection of miR-122 mimetic or miR-122-expression vector caused a decrease in lactate content in PLC/PRF/5 HCC cells. Data shown are the representative set of two independent experiments and are mean±s.d. Source data is available for this figure at www.nature.com/msb.

miR-122 is the most abundant miRNA in adjacent non-tumor liver tissue, and its expression is reduced on average and more variable in HCC tumors. (A) Expression levels of 220 human miRNAs in tumor and adjacent non-tumor tissue; x axis: average expression levels in copies per 10 pg of mRNA; y axis: average of the log10 ratios of miRNA expression in each tumor to the mean-expression level for that miRNA in adjacent non-tumor tissue. (B) Expression levels of miR-122 in paired adjacent non-tumor (left) and tumor tissues from 96 HCC patients. Triangles represent individual subjects, whereas the box-plots show the median (horizontal bar), 25th–75th percentile range (box), 1.5 times the interquartile range (whiskers) and outliers (plusses) of the distribution of miR-122 expression. MiR-122 was significantly differentially expressed between tumor and adjacent non-tumor liver tissues (rank-sum P-value, 1.75e–13; miR-122 ranked 14th of 220 miRNAs).

miR-122 is the miRNA most strongly associated with seed-matched genes in tumor and adjacent non-tissues. Transcripts whose expression levels were correlated or anti-correlated with expression levels of miRNAs were tested for enrichment of 3′UTR hexamers (seed sequences) by the hypergeometric P-value. Shown are the E-values (Bonferroni-corrected P-values adjusted for testing of 4096 hexamers) for enrichment of the three hexamers matching bases 1–8 of the correlated miRNA (blue dots) or bases 1–8 of other known miRNAs (gray dots). Each column shows log10 E-values for a set of transcripts significantly positively or negatively correlated with a particular miRNA. The direction of correlation is indicated by ‘+' for positive and ‘−' for negative or anti-correlation. Only those transcript sets are shown for which the correlated miRNA has an E-value of <0.01, after further adjustment for the 241 transcript sets examined, for at least one of its 3′UTR seed hexamers in a correlated or anti-correlated transcript set. The red dashed line marks E<0.01 after adjustment for both 4096 hexamers tested per transcript set and for the 241 transcript sets examined.

(A) Up-regulation of seed-matched genes by anti-miR-122 is correlated with down-regulation of mitochondrial genes in individual mice. Shown is the mean regulation of miR-122 seed-matched genes up-regulated by anti-miR-122 compared with the mean regulation of mitochondrially localized genes down-regulated by anti-miR-122 in individual mice treated with anti-miR-122 or control oligonucleotides for 1, 3, or 4 weeks. Correlation coefficient of mean up-regulation versus mean down-regulation within anti-miR-122-treated mice is r=−0.66. (B). Up-regulation of seed-matched genes by anti-miR-122 is correlated with down-regulation of mitochondrial genes in individual HCC patients. To confirm the relation of the seed-matched genes and the mitochondrial genes derived from the 96 HCC patients and recapitulated in the mouse liver, we included additional transcriptome profiling data from 180 HCC, 40 cirrhosis, and 6 normal liver tissues for analysis. Shown is the mean regulation, in individual HCC patients, of miR-122 seed-matched genes anti-correlated with miR-122 and up-regulated by anti-miR-122 in mouse livers, compared with the mean regulation of mitochondrially localized genes correlated with miR-122 and down-regulated in anti-miR-122-treated mouse livers. Correlation coefficient of mean regulation of the two gene sets: tumor tissue, r=−0.73; adjacent non-tumor tissue, r=−0.52; cirrhotic tissue, r=−0.91; healthy tissue, insufficient data points to calculate correlation. (C) Treatments of miR-122 mimetic altered the expression levels of seed-matched primary targets and mitochondria-related secondary targets in HCC cells. The concurrent reduced expression of primary targets SMARCD1, MAP3K3, and CAT-1, and increased expression of secondary targets PPARGC1A, SDHA, and SDHB were observed after treatment with miR-122 mimetic in PLC/PRF/5/cells.

Expression of miR-122 secondary targets in HCC tumor and adjacent non-tumor tissue predicts patient outcome. 180 patients were classified by the mean-expression ratio in their tumors or adjacent non-tumor tissue of miR-122 secondary targets (Supplementary Table Ib). Lower expression of these genes correlates with lower expression of miR-122; however, directly measured miR-122 levels are less prognostic (P=0.092 for miR-122 in tumor tissue). Shown are Kaplan–Meier curves for survival of patients in the top versus bottom 50% of patients for miR-122 secondary target expression either in adjacent tissue (left panel) or in tumor tissue (right panel). Note that the median secondary target-expression level is 60% higher in adjacent non-tumor tissue versus tumor tissue.