SUPPLEMENTARY RESULTS Hypomethylated promoters are neither mutated nor deleted in HCC samples To rule out the possibility that the demethylation observed in gene promoters in our pyrosequencing assays (conversion of a C to T following bisulfite conversion) indicates a mutation of C to T rather than demethylation, we sequenced the unconverted DNA of AKR1B10, CENPH, MMP2, MMP9, MMP12, NUPR1, PAGE4, PLAU, and S100A5 promoter regions using the pyrosequencing SNP assay (Supplementary Fig. S2). We show that in all cases the fraction of cytosines in the unconverted sequence is similar in normal liver and in HCC. Therefore, the increase in the fraction of cytosines that were converted to thymidine in the tumor samples occurred only after bisulfite treatment and it was not due to mutations of Cs to Ts. Loss of signal for methylated DNA in hypomethylated promoters in HCC did not result from loss of DNA by deletions since our method internally controls for loss of DNA. Our MeDIP arrays are hybridized with both DNA immunoprecipitated with anti-5-methylcytosine antibody as well as total DNA. Thus, our assays measure both DNA methylation and DNA integrity. The DNA methylation signal reflects the ratio of signal for methylated DNA immunoprecipitation with the anti-5-methylcytosine antibody over total DNA in the sample at the indicated genome position. Loss of DNA by deletion would have increased the ratio of methylated DNA to total DNA to infinity and would have presented itself as hypermethylation rather than hypomethylation. Careful examination of the promoters that were demethylated in HCC provides evidence for the absence of deletions/amplifications in the genes that are hypomethylated in HCC patients. We quantile-normalized the background channel across all inputs and computed a per-probe differential statistic between cancer and normal using a t-test. The wilcoxon rank-sum test was used to identify enrichment of unusually high or low probe 1 differential statistics inside each demethylated promoter. The p-values from the rank-sum test were adjusted for multiple testing using the false discovery rate. Among 3,708 hypomethylated promoters, only 19 promoters corresponding to 12 genes (Supplementary Table S4) had a false discovery rate less than 0.05 indicating that the region may have been deleted/amplified. They have been excluded from the list of genes hypomethylated in HCC. Pathways controlled by epigenetically induced genes Functional analysis reveals that the epigenetically induced genes are enriched in pathways known to drive cellular transformation, cancer growth, angiogenesis and cancer metastasis (Supplementary Fig. S3A). Example pathways include the prometastatic TGF-beta signaling (1), the potent angiogenic and prometastatic VEGF signaling (2-3), the JAK-STAT signal transduction pathway that mediates cytokine triggered signals such as IL6 which plays an important role in cancer growth and metastasis (4), and participates in cross talk with the mitogenic and oncogenic RAS-MAPK and TGF-beta signal transduction pathways (5), the WNT signaling cascade that plays an important role in several cancers including colorectal, breast and prostate cancer and is upregulated in cancer metastasis (6-8), and finally the HEDGEHOG signaling cascade that was reported to be aberrantly reactivated in liver cancer and its blocking resulted in a decrease in cancer cell viability (9). Activation of both Jak-STAT and Ras/MAPK were previously shown to be enhanced in HCC samples, and their inhibition led to a strong apoptotic response (10). Enrichment of epigenetically induced genes in the gycolysis/gluconeogenesis pathway is consistent with the remarkable propensity of highly malignant, poorly differentiated tumors, including hepatomas, to utilize glucose at a much higher rate than normal cells (11). 2 In general, the epigenetically induced genes are enriched in biological processes that are known to be critical for tumor progression, survival and motility, differentiation, transcription regulation and signal transduction (Table 2). Many of the epigenetically induced genes in these critical processes are known to play a role in cancer while at least 20 of these genes, to our knowledge, play unknown roles (Table 2A and B). These 20 genes are enriched in biological processes such as histone binding, positively regulating MAPK pathway, methyltransferase activity and involvement in base-excision repair, suggesting that these new candidates play a role in liver cancer and are regulated by epigenetic mechanisms. The roles of several of these genes, CCDC138, KCTD2, PAQR4 and RNMT, are poorly established. Other genes such as CSPP1, FAM83D, EXOSC4 and NEIL3 have previously been linked to breast, colorectal cancer, renal carcinoma and glioblastoma through mutation and expression analyses (12-13). Overall our analyses support the hypothesis that hypermethylation and hypomethylation target distinct cell functions involved in resetting the cellular gene expression program from that of a normal untransformed state to that of a highly transformed and invasive cancer. Promoter hypermethylation inhibits genes that block cellular growth and metastasis whereas hypomethylation drives promitogenic, metabolic and prometastatic processes. 5-aza-2’-deoxycytidine (5-azaCdR) induces “HCC demethylated genes” in human untransformed hepatocyte cell culture (NorHep) In order to determine causal relationship between promoter hypomethylation and activation of genes epigenetically induced in HCC we determined whether the DNA methylation inhibitor 5- aza-2’-deoxycytidine (5-azaCdR) would induce these genes in a primary hepatocyte (NorHep) cell culture. To experimentally test causality between DNA methylation and expression we had 3 to resort to cell culture experiments. Since we used HepG2 hepatocellular carcinoma cell line as a model of transformed liver cells and NorHep cells as a model of normal liver cells we focused on genes that were hypomethylated (expressed) in HepG2 cells and hypermethylated (silent) in NorHep cells. We then tested whether these genes would be induced in NorHep cells following treatment with the DNA demethylating agent 5-azaCdR. Most of the studied genes were also hypomethylated in HCC patients. We examined by QPCR the expression of 84 genes following 5-azaCdR treatment of NorHep including 65 with high CpG-dense promoters (HCP). As shown in Fig. 2E, the most profound increase in gene expression among all promoters hypomethylated in HCC patients is observed for genes with HCP promoters where average expression differences are > 0. We therefore focused on this gene set for this analysis. Moreover, 38 out of these 84 genes belonged to the genes epigenetically induced in HCC patients that showed the most profound hypomethylation and/or induction and included genes known to be involved in cancer (e.g., CENPH, CKS2, IPO7, MAP3K4, MAPRE1, PPARG, PRPF6, RELB, SERF2) as well as new cancer gene candidates (e.g., CSPP1, RASAL2, SENP6, RNMT, TMX2, NEIL3, NENF) as shown in Table 2. Since the experiment was performed in a primary hepatocyte (NorHep) cell culture we also tested a group of 46 genes in addition to 12 genes from the 38 genes described above that are hypomethylated in HepG2 cells compared to NorHep cells and represent families important in cancer such as members of WNT signaling pathway (WNT16), semaphorins (SEMA4C), S100 calcium binding proteins (S100A5), ubiquitin-conjugating enzymes (UBE2E1/2), ephrin receptors (EPHA3/B1), transcription factors (E2F2), growth factors (EGF, FGF). 24 out of these 46 genes are also significantly hypomethylated in HCC patients. 4 71 out of the 84 examined genes were induced upon treatment with 5-azaCdR at 1.0 µM concentration for either 5 or 20 days (Supplementary Table S6). The highest increase was observed for FGF20, BMP4, CABYR, CCL20, WNT16, MMP2, PLAU, and S100A5 (4-50 fold, Supplementary Fig. S4B and Supplementary Table S6). As expected 5-azaCdR in NorHep cells led to global hypomethylation (Supplementary Fig. S4A) and a decrease in promoter methylation of MMP2, NUPR1, PLAU, and S100A5 genes (Supplementary Fig. S4C). Using microarray data, we also compared fold change in expression with the extent of hypomethylation for 230 genes epigenetically induced in HCC patients. For each gene, we compared differential gene expression to differential promoter methylation between cancer and normal tissue across all patients. If promoter hypomethylation does have impact on the increase in gene expression, then we expect the gene expression differences to be negatively correlated with the promoter methylation differences. For both differential expression and differential methylation, we used the log-fold change of the microarray probe that the most significantly differentiated between cancer and normal for the given gene. We then calculated Pearson's correlation coefficient between our measures of differential expression and differential methylation. As expected, we found that the resulting correlation coefficients from the 230 genes were on average less than zero (Wilcoxon rank-sum test, P ≤ 0.005) (see Supplementary Fig. S4D for a scatter plot). Details on genes demethylated in liver, ovarian and breast cancers Several of the genes that are demethylated in liver, ovarian and breast cancers have been implicated in cancer progression and metastasis. Examples include regulators of tight junction and cell-cell adhesion (CLAUDIN 4, CLDN4) (14), modulators of transmembrane signaling 5 systems (GUANINE NUCLEOTIDE-BINDING PROTEIN G(S)