Paired non-tumor and tumor tissue specimens of hepatocellular carcinoma (HCC) were collected from fifteen HCC patients diagnosed with stage I-IV pathologic tumor-node-metastasis (TNM) disease [17]. Normal liver specimens were obtained from cadaveric liver donors. All the samples were provided by the Tissue Bank of Divisions of Hepatobiliary and Pancreatic Surgery and Liver Transplantation at Department of Surgery, Queen Mary Hospital. Collection and storage of clinical specimens for the Tissue Bank has been approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority of Hong Kong. Normal hepatocyte and HCC cell lines included MIHA and L02 (normal hepatocytes); PLC (primary HCC); and MHCC97L (97L) and MHCC97H (97H), were derived from metastatic HCC [18]. Other non-HCC cancer cell lines included HeLa (cervix adenocarcinoma); MCF7 (breast adenocarcinoma); AGS (gastric adenocarcinoma); HCT116 (colorectal carcinoma); and K-562 (chronic myelogenous leukemia). The PLC, HeLa, MCF7, AGS, and HCT116 lines were purchased from the American Type Culture Collection (ATCC). L02 was obtained from the China Center for Type Culture Collection.

Total genomic DNA was isolated from peripheral blood mononuclear cells (PBMC), cell lines, and non-tumor and tumor liver specimens from HCC patients, using the QIAamp DNA mini kit (Qiagen).

Genomic DNA was processed for bisulfite conversion of unmethylated cytosines using the EpiTect Bisulfite kit (Qiagen). The bisulfite-modified DNA was used for PCR, with primers that recognize the converted DNA sequences (Figure 1A; Figure S1 in File S1; Table S1). PCR products were cloned into pGEM T-Easy vector (Promega Bioscience). Ten clones were randomly picked and sequenced. DNA sequences were analyzed with the BIO Analyzer software (http://biq-analyzer.bioinf.mpi-sb.mpg.de) for CpG reading. Percentage of methylation refers to the number of methylated CpGs over the total number of CpGs in the region analyzed.

The ChIP procedure was described in Current Protocols [19]. Briefly, ten to twenty million (1–2×10⁷) L02, HCT116 and 97L cells were collected and fixed with 37% formaldehyde. The sonicated cell lysates were subjected to immunoprecipitation with 5 μg of H3K4me3 or H3K27me3 antibodies (Abcam), and normal IgG (inputs as control), respectively, in the presence of protein G-Sepharose beads (GE Healthcare) at 4°C overnight. Beads were washed sequentially with FA lysis buffer, ChIP wash buffer, and TE buffer. Bound materials were eluted with ChIP elution buffer. Enrichment of histone modification was quantified by real time PCR, using SYBR Green (Applied Biosystems), with relevant ChIP primers (Figure 1A; Figure S1 in File S1; Table S1). Data are presented as fold enrichment, using a ChIP-qPCR analysis calculation formula (www.sabiosciences.com); inputs and mock IP were used for normalization.

Total RNA was isolated using the RNeasy Mini kit (Qiagen) and treated with DNase I, followed by reverse transcription with the Transcriptor First Strand cDNA Synthesis Kit (Roche). qRT-PCR analysis was performed using the pre-designed TaqMan probes for NANOG (Hs02387400_g1), POU5F1 (Hs00999634_gH), MYC (Hs00153408_m1), and 18S rRNA (4333760-0807027), which were selected from TaqMan Gene Expression Assays (Applied Biosystems). SYBR Green reagents (Applied Biosystems or TaKaRa Biotechnology) were used for detection of OCT4, MYC, KLF4, p53 and β-ACTIN gene expression (primer information in Table S2). Gene expression was calculated as CT and normalized to the level of 18S or β-ACTIN.

Human NANOG (NM_024865) and OCT4 (NM_002701) genes from human embryonic stem cell line were cloned into pWPI vector and were transfected with pCMV-dR8.91 and pMD2.G plasmids into the 293T packaging cell line. Viral supernatants were harvested 48 hours after transfection and precipitated using PEG-it Virus Precipitation Solution (System Biosciences). HCT116 p53−/− cells were infected with lentivirus expressing either OCT4 or NANOG in the presence of 8 μg/ml protamine sulfate. Cell sorting was carried out using FACSAria (BD Biosciences) for isolation of transduced HCT116 p53−/− cells.

Four to six weeks old BALB/cAnN-nu (Nude) mice were obtained and maintained at Laboratory Animal Unit of The University of Hong Kong. All mouse experiments were approved by The Committee on the Use of Live Animals of The University of Hong Kong. Lentivirus infected HCT116 p53−/− cells were injected subcutaneously into nude mice to generate primary tumors. The primary tumors were dissected into 1–2 mm³ cubes and then implanted into the cecum of other nude mice [20]. Colon to lung tumor metastasis was examined after 4 weeks of cecum implantation. The whole lungs from nude mice were sectioned and stained with H&E. The colon tumor lesions in the lungs were examined by microscopy.

CD133-PE (Miltenyi Biotec) labeled HCT116 cells were analyzed using FACSCalibur (BD Biosciences). CD133+^low, CD133+^high, and CD133− cells were subjected to cell sorting by FACSAria (BD Biosciences).

Data were presented as the mean ± SD and the Student's t test was performed using SPSS software. P<0.05 was regarded as statistically significant.

Previous studies have demonstrated the activation of downstream targets of NANOG, OCT4, SOX2 and MYC genes in aggressively growing human cancers [11]. We therefore investigated the CpG methylation pattern of NANOG (also known as NANOG1), OCT4, SOX2, KLF4 and c-MYC in human cancer cell lines by bisulfite sequencing approach (Figure 1A; Figure S1 in File S1). Compared to normal liver cells L02, which displayed a modest methylation level (39%), DNA hypomethylation of NANOG (Figure 1A–i) was observed in three liver cancer cell lines, namely PLC, 97L and 97H (18%, 8% and 14%, respectively), among which the metastatic liver cell line 97L showed nearly absence of DNA methylation (Figure 1B and 1C). While in leukemic K-562 cells, NANOG methylation (39%) showed a 2-fold decrease when compared to normal peripheral blood mononuclear cells (PBMCs) (81%; Figure 1B and 1C). Cancer cells from other tissue origins displayed differential NANOG promoter methylation; for example, hypomethylation was observed in HeLa (25%), whereas hypermethylation was shown in MCF7, HCT116 and AGS (71%, 92% and 91%, respectively) (Figure 1B and 1C).

We next examined the OCT4 promoter methylation pattern in cancer cells. The proximal promoter region (−503 to +7; Figure 1A–ii) of OCT4 was found hypermethylated in L02 (92%), MIHA (90%), HeLa (90%) and HCT116 (87%) cells, which is in contrast to the reduced levels observed in PLC and 97L cells (70% and 41%, respectively; Figure 1D). We extended the methylation analysis to a total of 50 CpGs that cover both the distal promoter and downstream region of the transcription start site (TSS, −2973 to +320; Figure 1A–ii). The reduced methylation pattern was conserved at the analyzed OCT4 gene region in PLC and 97L cells (Figure 1E). For the c-MYC gene, although the CpGs located upstream of TSS1 and TSS2 and within exon 2 remained unmethylated in both normal and cancer cells (Figure 1A–iii; Figure S2 in File S1), the methylation level of exon 3 was dramatically reduced and became hypomethylated in two liver cancer cell lines, PLC (34%) and 97L (6%), when compared to the moderate to high level of methylation (51% to 82%) in normal liver cells (Figure 1F). In contrast, cancer cells from other tissue origins, with the exception of HeLa, displayed hypermethylation status of c-MYC exon 3 (ranging from 72% to 99%; Figure 1F). On the other hand, CpG islands at the promoter regions of KLF4 and SOX2 genes remained unmethylated in all the examined cell lines (Figures S1, S3, S4 in File S1). Taken together, differential methylation pattern of pluripotency-associated genes was observed in human cancer cell lines from different tissue origins.

We next focused on the pluripotent gene expression in two cancer cell lines, 97L and HCT116, which showed opposite patterns of DNA methylation. qRT-PCR analysis demonstrated higher expression of NANOG, OCT4 and c-MYC genes in 97L cells but not in HCT116 cells, when compared to control L02 and PLC cells (Figure 2A–C; Figure S5A–C in File S1), suggesting the expression of these genes is negatively regulated by DNA methylation. For KLF4, increased gene expression was observed in both 97L and HCT116 cells (Figure 2D; Figure S5D in File S1) despite the fact that the CpG island methylation pattern was indistinguishable between normal and cancer cells (Figure S3 in File S1).

Epigenetic modifications on histone proteins also showed strong association with gene expression. We, therefore, determined the presence of two kinds of histone modifications, the active/permissive tri-methyl-H3K4 (H3K4me3) and the repressive tri-methyl-H3K27 (H3K27me3), in 97L and HCT116 cancer cells. ChIP analysis demonstrated a significant enrichment of active H3K4me3 mark at the promoter regions of NANOG and OCT4 and the exon 3 region of c-MYC in 97L cells, when compared to L02 and HCT116 cells (Figure 2E–G, left panels). This is in contrast to the low level of repressive H3K27me3 mark observed in these cells (Figure 2E–G, right panels). These results suggest that both histone modifications and DNA methylation function synergistically in regulating NANOG, OCT4 and c-MYC gene expression. On the other hand, H3K4me3 was found highly enriched at the KLF4 promoter of 97L cells, whereas HCT116 cells demonstrated a moderate level of H3K4me3 but with relatively lower level of repressive H3K27me3 (Figure 2H, right panel). This may account for the high level of KLF4 gene expression in both cell lines (Figure 2D), which is independent of DNA methylation.

Given that the differential NANOG methylation pattern in cancer cell lines can be a consequence of in vitro culturing rather than association with their tumorigenicity per se, we, therefore, investigated NANOG promoter methylation in primary HCC tumors paired with adjacent non-tumor tissues (n = 15), compared to normal liver samples (Figure 3A). Strikingly, 73% of non-tumor cases (11 out of 15) were more than 50% methylated, whereas 53% of tumor cases (8 out of 15) were less than 30% methylated at the NANOG promoter (Figure 3B and 3C). NANOG promoter methylation levels were significantly reduced in the paired HCC tumors compared with adjacent non-tumor tissues (p = 0.000), whereas no significant changes were detected between normal and non-tumor liver tissues (p = 0.301; Figure 3D). In accordance with the reduced promoter methylation, NANOG expression was found significantly higher in tumor tissue than in non-tumor tissue (p = 0.021; Figure 3E). It is worth noting that HCC samples at TNM stage III and IV, which is usually present with vascular invasion, lymph node and distant metastasis [17], was significantly associated with high NANOG (p = 0.011) and low p53 (p = 0.047) expression (Table 1; Table S3). The low level of p53 was also associated with vascular infiltration (p = 0.019; Table 1). These results suggest that tumor metastasis requires both the upregulation of NANOG via promoter hypomethylation and the suppression of p53 in HCC.

Following the identification of NANOG upregulation in metastatic HCC cells and tumor samples, we furthered our study to examine the in vivo function of NANOG in cancer progression. However, the strong expression of NANOG in HCC cells makes it difficult to be knocked down efficiently for in vivo functional studies (data not shown). We, therefore, selected the HCT116 cell line, which has low endogenous level of NANOG, for overexpression study. A stable NANOG expressing HCT116 p53−/− cells (Figure S6A in File S1) were injected into nude mice for tumor formation. Surprisingly, a minimum of 1×10⁴ cells of either HCT116 p53−/− or NANOG expressing HCT116 p53−/− (NANOG-HCT116 p53−/−) cells was sufficient to form subcutaneous tumors (Table S4), indicating that a higher level of NANOG expression could not further enhance the formation of primary tumors. However, using an orthotopic tumor implantation mouse model, we observed a significantly higher colon to lung metastasis by implanting xenograft tumors of NANOG-HCT116 p53−/− cells compared to the parental HCT116 p53−/− cells (Figure 4A–C). This provides strong in vivo evidence to support the function of NANOG in promoting cancer cell metastasis.

High expression of CD133 is one of the indications of CSC population in various types of cancers, including colon cancer [21], [22]. Since the three pluripotent genes NANOG, OCT4, and c-MYC were hypermethylated in colorectal carcinoma HCT116 cells, we asked whether the same pattern of methylation would be observed in the rare putative CSC population in HCT116 cells. We observed over 80% of HCT116 cells were CD133+, but only 1% of cells were regarded as CD133+^high (Figure 4D). Importantly, NANOG promoter methylation was maintained at high level in both CD133− and CD133+^low HCT116 cells (89% and 78%, respectively), while it was dramatically reduced to 42% in CD133+^high cells (Figure 4E). The reduced methylation was also found in accordance with a significant upregulation of NANOG expression in CD133+^high cells (Figure 4F). In addition, overexpression of exogenous NANOG in HCT116 cells resulted in an increase of CD133+^high population (Figure 4G), suggesting that demethylation of NANOG promoter contributes to the initiation of CSCs in tumor development.

OCT4, SOX2, and NANOG have been demonstrated to form a transcriptional regulatory loop in ESCs for the maintenance of pluripotency [23]–[25]. We asked whether such cross-regulation is conserved in cancer cells with aberrant epigenetic patterns. HCT116 cells with different p53 genetic backgrounds were used because p53 was reported as a barrier to the re-establishment of pluripotency [26]. Interestingly, overexpression of exogenous OCT4 significantly increased NANOG mRNA levels in HCT116 p53−/− cells, but not in HCT116 p53+/+ cells (Figure 5A; Figure S6B in File S1). The induction of NANOG expression was associated with a decreased promoter methylation in OCT4-HCT116 p53−/− cells (from 95% to 66%; Figure 5B). Moreover, overexpression of the exogenous NANOG significantly enhanced OCT4 mRNA levels in HCT116 cells regardless of their p53 status (Figure 5C). This suggests that the pluripotency regulatory circuit is conserved in cancer cells and it is partially restored in cancer cells via the epigenetic mechanism that reprograms NANOG promoter methylation.