Dynamic imbalance between cancer cell subpopulations induced by Transforming Growth Factor Beta (TGF-β) is associated with a DNA methylome switch
- Marion Martin1,
- Pierre-Benoit Ancey1,
- Marie-Pierre Cros1,
- Geoffroy Durand2,
- Florence Le Calvez-Kelm2,
- Hector Hernandez-Vargas†1Email author and
- Zdenko Herceg†1Email author
© Martin et al.; licensee BioMed Central Ltd. 2014
Received: 24 October 2013
Accepted: 28 May 2014
Published: 5 June 2014
Distinct subpopulations of neoplastic cells within tumors, including hepatocellular carcinoma (HCC), display pronounced ability to initiate new tumors and induce metastasis. Recent evidence suggests that signals from transforming growth factor beta (TGF-β) may increase the survival of these so called tumor initiating cells leading to poor HCC prognosis. However, how TGF-β establishes and modifies the key features of these cell subpopulations is not fully understood.
In the present report we describe the differential DNA methylome of CD133-negative and CD133-expressing liver cancer cells. Next, we show that TGF-β is able to increase the proportion of CD133+ cells in liver cancer cell lines in a way that is stable and persistent across cell division. This process is associated with stable genome-wide changes in DNA methylation that persist through cell division. Differential methylation in response to TGF-β is under-represented at promoter CpG islands and enriched at gene bodies, including a locus in the body of the de novo DNA methyl-transferase DNMT3B gene. Moreover, phenotypic changes induced by TGF-β, including the induction of CD133, are impaired by siRNA silencing of de novo DNA methyl-transferases.
Our study reveals a self-perpetuating crosstalk between TGF-β signaling and the DNA methylation machinery, which can be relevant in the establishment of cellular phenotypes. This is the first indication of the ability of TGF-β to induce genome-wide changes in DNA methylation, resulting in a stable change in the proportion of liver cancer cell subpopulations.
KeywordsHCC Tumor-initiating cells CD133 DNA methylation TGF-β pathway
Hepatocellular carcinoma (HCC) is the major form of primary liver cancer , and typically originates in a background of chronic inflammation caused by various factors, such as alcohol consumption, or viral infection (hepatitis B and hepatitis C) . Inflammation is an essential part of the wound-healing response to those risk factors. However, chronic inflammation favors the accumulation of mutations and epigenetic aberrations in hepatocytes, thereby promoting malignant transformation [3, 4]. This process is mediated by chemokines, cytokines, and growth factors secreted by the stromal components of the liver microenvironment . Among those secreted factors, the transforming growth factor beta (TGF-β) has been shown to have a key role that is cell-type dependent and variable during the hepatocarcinogenesis process . In established HCC, TGF-β overexpression is associated with poor prognosis [6–8]. However, characterization of the tumor cells targeted by TGF-β in HCC is still lacking.
As has been shown for other human malignancies, a subpopulation of cancer cells in HCC is known to display a higher tumorigenic potential [9–11]. These so called tumor-initiating cells (TICs), are defined by their self-renewal and differentiation capacity, and have been isolated based on their expression of several cell markers (EpCAM, CD133, CD90, CD44, CD24, CD13, and OV6) . Of these, the surface marker CD133/Prominin1[PROM1] has been one of the most consistently reported. CD133 is a transmembrane protein whose function is only partially known [12, 13], but that may represent a marker of a distinct cell subpopulation with defined characteristics. The functional characterization of these cells will increase our understanding of the mechanisms involved in promoting and sustaining liver cancer progression.
Several recent reports suggest a link between TGF-β signaling and liver TICs. Firstly, signaling pathways identified in liver cancer, including TGF-β, are active in isolated liver TICs . Secondly, TGF-β-induced epithelial-mesenchymal transition generates self-renewing cells, a process also implicated in a higher risk of tumor metastasis, as invasiveness and self-renewal are frequently shared features of stem cells, TICs and metastatic cells [15, 16]. Finally, a recent study showed that TGF-β is able to induce the expression of CD133 in liver cancer cell lines together with an increased tumor initiating ability in mice . Together, these studies point towards a specific role for TGF-β in inducing a TIC program in HCC.
DNA methylation is able to stably modify the cell phenotype through cellular division . Because of the relative stability of DNA methylation marks, DNA methylation is a strong candidate mechanism to translate the presence of TGF-β in the cellular microenvironment into persistent changes in phenotype. However, there is still limited evidence of a link between exposure to components of the tumor microenvironment and the induction of stable changes in DNA methylation in target cells.
In this study, we first defined the DNA methylome profile of CD133-expressing liver cancer cells. We then tested the potential association between DNA methylation and the induction of liver CD133+ cells by TGF-β. We show that TGF-β function in this context is intimately linked to a change in DNA methylation profiles, and that this may represent a key process in the establishment of chronic exposure imprints in liver cancer cells.
CD133- and CD133+ liver cancer cells differentially express DNA methylation genes
To serve as a basis for exploring a potentially different methylation program in CD133+ liver cancer cells, we studied the expression of genes coding for relevant players of the DNA methylation machinery. This included genes involved in maintenance of DNA methylation (DNMT1), de novo DNA methylation (DNMT3A and DNMT3B) and DNA demethylation (TET1 and TET2). Notably, DNMT3A was consistently and significantly overexpressed in both Huh7 and HepG2 cells progressively enriched for CD133 (Figure 1b). In addition, DNMT3B was overexpressed in HepG2 CD133-enriched cells, while TET2 displayed opposite differential expression in CD133-enriched Huh7 and HepG2 cells (Figure 1b). As noted above, the stable balance between the two cell fractions suggests no substantial difference in cell cycle rate between them. Therefore, significant differences in expression, although modest in magnitude, are compatible with true functional differences between the two subpopulations.
Together, these data suggests that CD133 positive and negative fractions grow in a constant proportion within liver cancer cell lines. They differentially express de novo DNA methylation genes (DNMT3A in both cell lines, and DNMT3B in HepG2) and a subset of genes involved in stemness (Additional file 1: Figure S1b). Functionally, expression of this marker has been associated with an increased tumor-initiating ability and ability to grow in non-attachment conditions, a well known surrogate measure of TIC-like activity. We found that MACS-sorted CD133+ Huh7cells were able to form spheres under non-attachment conditions, in contrast to their CD133- counterpart (Additional file 1: Figure S1c). This was not the case with HepG2 cells, where no sphere formation was observed, possibly due to the lower enrichment of CD133+ cells that was attained using MACS.
A differential DNA methylome distinguishes CD133- and CD133+ liver cancer cells
After having identified differentially methylated CpGs and the genes associated with these sites, we next aimed to identify the pathways that are specifically altered in CD133+ cells. To this end, we performed pathway analysis considering methylome profiles of both cell lines together or independently. Notably, in both cases there was enrichment in pathways previously associated with tumor initiating cell activity, such as Jak-STAT, Notch, Wnt and Akt (Additional file 3: Table S2). Other pathways included actin cytoskeleton, focal adhesion, and cell adhesion. In addition, there was a significant overrepresentation of inflammatory pathways, such as NFkB, p38, TNF, and TGF-β signaling pathways.
In summary, our data shows that CD133+ and CD133- liver cancer cells display a different DNA methylome. In spite of the cell line specific profiles, the data suggests a common CD133+ methylome signature, which includes the PROM1 gene itself. In addition, the methylome of CD133+ cells is characterized by a global reduction in DNA methylation relative to their CD133- counterpart, with an overrepresentation of non-CGI CpG sites. For those differentially methylated sites related to annotated genes (and mainly found in the gene bodies), there was an association with TIC- and inflammation-related pathways. These findings suggest that specific DNA methylation profiles are associated to the phenotype and functionality of these cell subpopulations.
TGF-β, but not IL-6, induces CD133 expression in a stable fashion
TGF-β is a member of a large family of pleiotropic cytokines that signal through a receptor complex comprising a diversity of type I and type II serine/threonine kinases. The recombinant TGF-β1 used in our assays is expected to bind the activin receptor-like kinase (ALK)5 (the TGF-beta type I receptor) . To rule out unspecific effects of this treatment, we used the small molecule inhibitor SB-431542, which targets ALK5 and ALK5-related type I receptors, with no effect on other family members that, for example, recognize bone morphogenetic proteins (BMPs) . By using this specific inhibitor of TGF-β pathway, we were able to abrogate the effect of TGF-β in inducing CD133 expression (as well as the morphological changes) in both cell lines (Figure 3b and Additional file 4: Figure S2e). Therefore, the ability to induce CD133+ cells is specific and fully dependent on TGF-β type I receptor signaling in both, Huh7 and HepG2 cells (Figure 3b).
Together, these findings suggest that TGF-β is able to specifically and stably induce CD133 expression (in contrast to the milder and transient effect of IL-6), an observation consistent with epigenetically-induced phenotype persistence.
De novo induction of CD133 by TGF-β is associated to an increased expression of DNMT3 genes
After having shown that TGF-β may be able to induce a de novo fraction of CD133+ cells, we asked whether this effect correlated with a differential expression of DNA methylation players, as we have shown that CD133+ cells overexpress DNMT3 genes in basal culture conditions (Figure 1b). DNMTs and TET2 displayed a significant increase in mRNA expression in at least one of the two cell lines, while TET1 was underexpressed after 4 days of release from TGF-β exposure (Figure 4b). As shown for the basal CD133-expressing cells (i.e. those isolated from untreated HCC cell lines), the most consistent finding was the statistically significant overexpression of DNMT3A in both cell lines after TGF-β treatment. Of note, in none of the conditions of study was IL-6 exposure able to induce statistically significant changes at the mRNA expression level of genes related to DNA methylation/demethylation (Figure 4b).Combined, these data shows the ability of TGF-β (in contrast to IL-6) to induce a stable de novo fraction of CD133-expressing cells in two independent liver cancer cell lines. This induction correlates with a functional characteristic of basal CD133+ cells, which is the increased ability to grow under non-attachment cell culture conditions (Figure 4c). Moreover, the persistence of CD133 induction by TGF-β (4 days after release from TGF-β treatment, and at least two cell divisions apart) suggests an epigenetic process is taking place, in contrast to the transient induction of CD133 by IL-6.
Cell subpopulation switch induced by TGF-β correlates with a differential DNA methylome
Differentially methylated regions (DMRs) after TGF-β exposure in two liver cancer cell lines and two time points (4 days treatment and 4 additional days post-release)
2.26E + 08
2.26E + 08
1.35E + 08
1.35E + 08
1.21E + 08
1.21E + 08
1E + 08
1.35E + 08
1.35E + 08
1.57E + 08
1.57E + 08
1.33E + 08
1.33E + 08
To gain a better insight on the consequences of TGF-β-induced methylome switch on the phenotype, we performed a whole genome expression analysis in both, Huh7 and HepG2 cells. We chose the 8-days time point (4 days of TGF-β treatment + 4 days post-release), considered in our model as the one defining long-term, stable changes induced by this cytokine. Expression analysis showed an expected profile of gene expression in both cell lines, including known TGF-β targets (Additional file 7: Table S4, Additional file 8: Figure S4c for qRT-PCR validations). In addition, “type I transforming growth factor beta receptor binding” was the first gene ontology category at the molecular function level (Additional file 7: Table S4). However, when intersecting the expression (two-fold change with an FDR < 0.05) and methylation (555 DMPs) significant gene lists, there was no significant overlap. As the effect of a specific methylation change on gene transcription is known to depend on the genomic location , we plotted all expression and methylation data, and analyzed separately CpG island and non-CpG island sites. As expected, no obvious correlation can be seen when plotting simultaneously all genes, independently of genomic location. However, hypermethylation within the gene bodies positively correlated with gene expression (Additional file 8: Figure S4a and S4b). In addition, we observed a small but significant overlap of 30 genes when intersecting the CD133 and the TGF-β methylation signatures (p = 0.0013).
Our data shows that the effect of TGF-β in liver cancer cell lines comes along with a remarkable switch of the DNA methylome at multiple loci. This reconfiguration is stable and common to two independent cell lines, and affects a significant proportion of enhancer regions and GC poor regions on gene bodies, which in some cases correlates positively with gene expression. The TGF-β “methyl-sensitive” signature described here includes DNA methylation players themselves and a number of TGF-β pathway loci, indicating a potential role for DNA methylation in establishing a TGF-β-induced phenotype switch in these cells. These results suggest that basal CD133+ cells and TGF-β-induced CD133+ cells only share a limited subset of their methylome. An important fraction of TGF-β methyl-sensitive CpG sites are not differentially methylated in CD133+ cells.
Kinetics of DNA methylation changes at TGF-β methyl-sensitive sites
In the second experiment we aimed to assess the stability of the methylation changes at longer time points. To this end, cells were treated with TGF-β during 4 consecutive days (Figure 6b). After this time cell culture medium was replaced by standard medium, and control and treated cells were followed for DNA extraction every 4 days until day 16 (corresponding to day 12 post-release from TGF-β exposure). First, we observed differences on the basal methylation of TRRAP when comparing to the previous experiments, especially in Huh7 cells (Figure 6a). This could be the result of differences on the efficiency of bisulfite conversion between experiments, but also differences between cell lines from various origins, as the Huh7 cells used for the second experiment were obtained from an independent laboratory. Regardless of basal difference, we observed that the increased methylation of TRRAP was replicated in both cell lines. Furthermore, significantly higher methylation levels persisted up to 12 days post-release from TGF-β, with no apparent change in its magnitude. Importantly, cell growth was not significantly affected after TGF-β release, and we estimate that at least 8 cell duplications took place during this time.
In summary, DNA methylation changes in a typical TGF-β methyl-sensitive locus in the TRRAP gene body can be observed as early as 1-2 days following TGF-β exposure, and persist through cell division for at least 16 days without apparent change in magnitude.
Silencing of de novo DNA methyl-transferases impair the effect of TGF-β at methyl-sensitive sites
Finally, we tested if, under the same conditions, DNA methylation changes in response to TGF-β were also impaired after DNMT3 silencing. TRRAP methylation is induced after 2 days of TGF-β treatment in HepG2 cells, an effect not influenced by the transfection of a non-targeting siRNA (Figure 7f). However, all further experimental conditions significantly reduce TRRAP methylation in the absence of TGF-β. In addition, the hypermethylation induced by TGF-β exposure was impaired, especially after transfection with siRNA against DNMT3B. Together, these results suggest that de novo DNA methyl-transferases are involved in the effect of TGF-β at different levels: morphology, distribution of cell subpopulations, and DNA methylation. The hypomethylation observed in absence of TGF-β also suggests a role for DNMT3A and DNMT3B in maintenance of DNA methylation at the TRRAP locus.
Discussion and conclusions
In the present report, we comprehensively describe the DNA methylome of CD133+ and CD133- liver cancer cells. We used two non-related HCC cell lines to isolate pure populations of CD133- and CD133+ cells for DNA methylome assays. As has been previously reported, CD133+ cells isolated from liver cancer cell lines (including those used in the present study), are functionally distinct cells with increased ability to induce tumors in animal models . These findings are in line with clinical studies reporting poor prognosis for those HCC cases displaying higher proportions of CD133-expressing cells. Although testing the tumor-initiating or metastasis-initiating ability of CD133+ cells was beyond the scope of our study, our results suggest that these cells display a differential DNA methylation signature. Whether DNA methylation is fundamental in establishing the cellular programs defining the main characteristics of these cells should be the focus of future studies.
Recently, the prognostic implications of TGF-β pathway activation in HCC have been linked to the ability of this signaling pathway to induce metastatic behavior in a fraction of HCC cells . An additional link between liver TICs and TGF-β in HCC has been the recent demonstration that TGF-β is able to increase the proportion of CD133+ cells in vitro . Here, we were able to reproduce and extend those previous observations. We showed that TGF-β is able to increase CD133 expression at the protein and mRNA level in two non-related HCC cell lines. The effect induced by TGF-β is stable, as opposed to the transient effect of the proinflammatory cytokine IL-6. We show that this effect depends on specific signaling through TGF-β type I receptor and is independent of increased cell proliferation of CD133+ cells. By using CD133-depleted cellular fractions, we suggest that TGF-β is able to induce de novo expression of CD133. Furthermore, this induction of CD133 cells correlates with an increased ability to grow in non-attachment conditions, a surrogate functional assay for stem/TIC properties.
Both, basal CD133-expressing cells and TGF-β-induced CD133+ cells, expressed increased levels of the de novo DNA methylation transcripts, DNMT3A and DNMT3B. This led us to further explore the ability of TGF-β to induce DNA methylation changes at the genome-wide level. We were able to show cell line-independent changes in DNA methylation induced by TGF-β in a stable fashion, suggesting an epigenetic mechanism involved in the establishment of a cellular program. The methylome of TGF-β -treated cells only partially overlapped with the methylome of CD133+ cells in basal conditions. This suggests that TGF-β may not only induce CD133 expression (or an increase in the CD133-expressing subpopulation), but also a defined DNA methylation profile. Further studies at longer time points and analyses of isolated CD133 negative and positive cells may shed light on the ability of TGF-β to imprint a DNA methylation signature independently of the induction of CD133 expression.
Notably, although TGF-β is known to induce DNA methylation changes at discrete loci [26–28], little evidence existed to date of a genome-wide level of TGF-β-methyl-sensitive loci. Specifically, several previous reports were focused on chromatin changes associated with epithelial-mesenchymal transition (EMT). EMT is a developmental process that involves actin cytoskeleton reorganization and loss of apical–basal polarity and cell-to-cell contact, and like other developmental processes it involves epigenetic reprogramming . However, both at physiological and pathological levels, EMT has been mainly linked to widespread changes of histone marks or histone modifiers, in addition to the well-known role of defined transcription factors [30–32]. Interestingly, gene-specific changes in DNA methylation have been correlated with the ability to maintain epigenetic silencing of critical EMT genes . In this sense, it has been suggested that DNA methylation is involved in the process of fixing the switch between epithelial and mesenchymal phenotypes. In our conditions, TGF-β induced a loss of E-Cadherin expression in both cell lines, but not an evident increase in N-Cadherin, which is a known marker of EMT (Additional file 4: Figure S2f). However, our results are consistent with a model of persistent changes in DNA methylation induced by TGF-β. Indeed, our experimental design was intended to reproduce an epigenetic process, by selecting only those changes in DNA methylation that survived cell division. Whether this effect of TGF-β is specific of transformed cells will require further studies.
Regardless of considerations on the targeted cells or the effect of the cell cycle or cell subtypes, we describe what seems to be a unique pattern of differential methylation. Our TGF-β signature (stable after TGF-β removal, and common to two liver cancer cell lines) was characterized by a reduced GC content, and enrichment at enhancer regions. Moreover, the under-representation of promoter CGIs and enrichment on gene bodies (together with the positive correlation with gene expression in a subset of these sites), may indicate specific mechanisms of gene regulation through DNA methylation. For example, changes in DNA methylation on these TGF-β methyl-sensitive loci may drive differential enhancer activity or alternative transcriptional events, as has been suggested in other contexts . Whether similar findings occur in response to other cytokines is not currently known.
In summary, our data support and reinforce several previous studies that have pointed to an association between TICs, and TGF-β. In addition, we provide a mechanistic insight into the process that may lead to the stable change in cancer cell subpopulations. Our study demonstrates that a key cytokine involved in HCC progression, TGF-β, is able to epigenetically induce a dynamic imbalance between cell subpopulations. The results reported here are in agreement with a model in which DNA methylation plays a pivotal role in establishing the cellular program of liver cancer cell subpopulations. The dynamics of a related process has recently been shown for CD44+ breast cancer stem cells . However, in our model, the effect of TGF-β is persistent (as compared to the effect of another cytokine, IL-6) and therefore epigenetically acquired. Motivated by these observations, the mechanistic evidence of an active interplay between TGF-β and the DNA methylation machinery should be an exciting focus of future studies.
Cell culture and treatments
Huh7 and HepG2 cells (American Type Culture Conditions) were cultured in DMEM medium (Gibco) at 37°C and 5% CO2, and were regularly tested for mycoplasma contamination (MycoAlert detection kit, Lonza).
For cytokine treatments, cells were plated and allowed to adhere before adding medium containing 10ng/ml final of IL-6 or TGF-β1 (recombinant human, Peprotech). For inhibition experiments, cells were treated with 2 μM SB-431542 (Sigma-Aldrich) alone or in combination with TGF-β1.
For spheres formation assay, hepatosphere medium was prepared as previously reported . Spheres were counted after 5 or 6 days.
siRNA non-targeting and pool siRNAs against DNMT3A and DNMT3B (Dharmacon, On-Target plus siRNA) were transfected at 20nM using RNAiMAX lipofectamine (Life Technologies) as recommended by the manufacturer. Cells were washed and medium was replaced 12 hours after transfection.
Fluorescence Activated Cell Sorting (FACS)
Cells were labeled with antibodies against CD44, CD133 (AC133), EpCAM, CD90 or TGFBRII (Additional file 9: Table S5). Secondary antibodies were conjugated alternatively with FITC, Cy3 or Alexa750.
To study cell cycle progression, bromodeuxyridine (BrdU) (Sigma) incorporation and DNA content were simultaneously assessed, as previously described . Fluorescent events were captured using FACS instrument (FACSCanto II, BD Biosciences), and analyzed using BD FACSdiva 6.0 (BD Biosciences software) and WinMDI software (version 2.9).
Magnetic Activated cell sorting (MACS)
Huh7 and HepG2 cells were depleted or enriched for CD133+ cells using magnetic-activated cell sorting (MACS, Miltenyi Biotec), with some adaptations to the manufacturer’s instructions. Cells were incubated 30 min at 4°C with FcR blocking reagent, followed by 15 min incubation with MicroBeads conjugated to monoclonal anti human CD133 antibodies. After washing, cell suspension was applied onto a pre-rinsed LS column placed in the magnetic field of a MACS separator. For CD133+ depletion, flow-through the LS column containing unlabelled cells was collected. For CD133+ enrichment, the column was removed from the separator and placed on a 15 ml collection tube. Labeled cells were collected by firmly pushing the plunger in the column. To increase purity of CD133+ cells, the eluted fraction was enriched a second time over a new LS column. For each experiment, aliquots were kept to test by FACS the efficiency of the enrichment.
Bisulfite modification and pyrosequencing
To quantify the percentage of methylated cytosine in individual CpG sites, we performed bisulfite pyrosequencing, as previously described . For samples processed for Infinium bead arrays, the conversion was performed on 600 ng of DNA using the EZ DNA methylation Kit (Zymo Research) and modified DNA was eluted in 16 ul of water. Quality of modification was checked by PCR using modified and unmodified primers for GAPDH gene. Pyrosequencing assays (primers for PCR, sequencing primers and regions) are described in Additional file 9: Table S6.
Bead array methylation assays
Methylation profiles of the different samples were analyzed using the HM450 Infinium methylation bead arrays (Illumina, San Diego, USA). Briefly the HM450 beadchip interrogates more than 480,000 methylation sites . The analysis on the bead array was conducted following the recommended protocols for amplification, labeling, hybridization and scanning. Each methylation analysis was performed in duplicate (for CD133+ versus CD133- samples) or in triplicate (for all other methylome analyses).
Whole genome expression array
Total RNA was isolated using the TRIzol Reagent (Invitrogen) according to the manufacturer’s instructions. RNA quantity and quality were assessed with a ND-8000 spectrophotometer and bioanalyzer. 500 ng of total RNA was used for each Human HT-12 Expression BeadChips (Illumina), as previously described . 10 candidate genes were selected for validation using quantitative RT-PCR. Four different housekeeping genes (HPRT1, GAPDH, SFRS4 and TBP1) were alternatively used for internal control. The different primers used are listed in Additional file 9: Table S7.
Immunoblotting and immunofluorescence
Protein extraction and immunoblotting was performed as previously described . Immunostaining was performed with anti-SMAD3, anti-phosphorylated SMAD3 and anti-tubulin/actin for loading control.
For immunofluorescence Huh7, HepG2, and 3T3 cells grown in cover slips were fixed in 4% formaldehyde for 20 min and then washed and stained with primary antibodies against E-Cadherin and N-Cadherin (Novus Biological). Alexa Fluor 488 and Alexa 555 (Life Biotechnologies) were used as secondary antibodies, and TO-PRO-3-iodide (Life Biotechnologies) for counterstaining. An Axiovert LSM 510 confocal microscope (Zeiss, Jena, Germany) was used for image collection. Images were analyzed using LSM image browser software (Zeiss).
Raw expression bead array data was exported from Genome Studio (version 2010.3, Illumina) into BRB-ArrayTools software (version 4.3.1, developed by Dr. Richard Simon and the BRB-ArrayTools Development Team), as previously described . Data was normalized and annotated using the R/Bioconductor package “lumi” . Class comparison between groups of bead arrays was done computing a t-test separately for each gene using the normalized log-transformed beta values. Only those probes with FDR < 0.05 and a fold-change of two were considered differentially expressed. WebGestalt (WEB-based GEne SeT AnaLysis Toolkit) web application was used for gene set enrichment analyses, including Gene Ontology, and pathways .
For methylome analyses we used a combination of R/Bioconductor packages (following recent guidelines for HM450 data mining) [40, 41]. “WateRmelon” package was used to load the raw data directly from idat files into a MethyLumiSet object [38, 42]. Data quality was inspected using boxplots for the distribution of methylated and unmethylated signals, and inter-sample relationship using multidimensional scaling plots and unsupervised clustering. Probes were filtered for low bead count, low quality (detection P value > 0.05), and recently described cross-reactive probes . Then, we performed color bias adjustment, followed by inter-sample quantile normalization, and probe bias correction with intra-array beta-mixture quantile normalization, as described . Methylation beta values were logarithmically transformed to M values, better suited for parametric statistical analyses . M values were used to determine batch effects using principal component analysis, and corrected with the surrogate variable analysis package (“sva”) . To obtain a common differential methylation between control and TGF-β-treated cells we used the “limma” package with TGF-β exposure as the main variable, using the cell line (Huh7 and HepG2) as a co-factor in the linear model . Differentially methylated positions (DMPs) were defined as those sites with a methylation difference (delta-beta) of 10% in any direction with an FDR-adjusted p value below 0.05. To study the genomic context of DMPs we used HM450 annotations, with hg19 as the human reference genome. Other genomic features were obtained from a recent publication using HM450 bead arrays . Finally, we used the “methyAnalysis” package to identify differentially methylated regions (DMRs) after transforming our dataset to a list of genomic regions followed by methylation data smoothing, as described (Pan Du and Richard Bourgon . methyAnalysis: DNA methylation data analysis and visualization. R package version 1.2.0). We defined DMR as a region with at least two differentially methylated probes, and a minimum gap of 100 bp.
To study differential methylation between CD133 negative and positive cells we performed a class comparison blocking by cell line status (Huh7 or HepG2). The analysis performed is an analysis of variance for a randomized block design. Two linear models are fit to the methylation data for each gene. The full model includes class variable and the block variable, and the reduced model includes only the block variable. Likelihood ratio test statistics are used to investigate the significance of the difference between the classes. In this comparison we set a p value threshold <0.001 and 5% difference in methylation. For all other comparisons, only those probes with FDR-corrected p values <0.05 were considered significant.
BRBArrayTools and R/Bioconductor packages were used for bead array analyses, as described above. For other comparisons, means and differences of the means with 95% confidence intervals were obtained using GraphPad Prism (GraphPad Software Inc.). Two-tailed student t test was used for unpaired analysis comparing average expression between classes. P values < 0.05 were considered statistically significant.
Bone morphogenetic protein
Fluorescence-activated cell sorting
Transforming growth factor beta.
We thank Fabienne Barbet from the ProfileXpert platform for the scan of the Illumina 450K arrays (service de Génomique & Microgénomique, Université Lyon 1, SFR santé LYON-EST, UCBL-INSERM US 7-CNRS UMS 3453, 8 av. Rockefeller 69373 Lyon Cedex 08)). This work was supported by la Ligue National (Française) Contre le Cancer, l’Association pour le Recherche Contre le Cancer (l’ARC), and the Agence Nationale de Recherches sur le SIDA et les Hépatites Virales (ANRS). PBA is supported by a fellowship from La Ligue National Contre le Cancer (La Ligue).
- Ferlay J, Shin H-R, Bray F, Forman D, Mathers C, Parkin DM: Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J cancer J Int du Cancer. 2010, 127: 2893-2917. 10.1002/ijc.25516.View ArticleGoogle Scholar
- Forner A, Llovet JM, Bruix J: Hepatocellular carcinoma. Lancet. 2012, 379: 1245-1255. 10.1016/S0140-6736(11)61347-0.PubMedView ArticleGoogle Scholar
- Martin M, Herceg Z: From hepatitis to hepatocellular carcinoma: a proposed model for cross-talk between inflammation and epigenetic mechanisms. Genome Med. 2012, 4: 1-13. 10.1186/gm300.View ArticleGoogle Scholar
- Hernandez-Gea V, Toffanin S, Friedman SL, Llovet JM: Role of the microenvironment in the pathogenesis and treatment of hepatocellular carcinoma. Gastroenterology. 2013, 144: 512-527. 10.1053/j.gastro.2013.01.002.PubMed CentralPubMedView ArticleGoogle Scholar
- Massagué J: TGFbeta in Cancer. Cell. 2008, 134: 215-230. 10.1016/j.cell.2008.07.001.PubMed CentralPubMedView ArticleGoogle Scholar
- Bedossa P, Peltier E, Terris B, Franco D, Poynard T: Transforming growth factor-beta 1 (TGF-beta 1) and TGF-beta 1 receptors in normal, cirrhotic, and neoplastic human livers. Hepatology. 1995, 21: 760-766.PubMedGoogle Scholar
- Abou-Shady M, Baer HU, Friess H, Berberat P, Zimmermann A, Graber H, Gold LI, Korc M, Büchler MW: Transforming growth factor betas and their signaling receptors in human hepatocellular carcinoma. Am J Surg. 1999, 177: 209-215. 10.1016/S0002-9610(99)00012-4.PubMedView ArticleGoogle Scholar
- Mima K, Okabe H, Ishimoto T, Hayashi H, Nakagawa S, Kuroki H, Watanabe M, Beppu T, Tamada M, Nagano O, Saya H, Baba H: CD44s regulates the TGF-β-mediated mesenchymal phenotype and is associated with poor prognosis in patients with hepatocellular carcinoma. Cancer Res. 2012, 72: 3414-3423. 10.1158/0008-5472.CAN-12-0299.PubMedView ArticleGoogle Scholar
- Ji J, Wang XW: Clinical implications of cancer stem cell biology in hepatocellular carcinoma. Semin Oncol. 2012, 39: 461-472. 10.1053/j.seminoncol.2012.05.011.PubMed CentralPubMedView ArticleGoogle Scholar
- Ma S: Biology and clinical implications of CD133(+) liver cancer stem cells. Exp Cell Res. 2013, 319: 126-132. 10.1016/j.yexcr.2012.09.007.PubMedView ArticleGoogle Scholar
- Visvader JE, Lindeman GJ: Cancer stem cells: current status and evolving complexities. Cell Stem Cell. 2012, 10: 717-728. 10.1016/j.stem.2012.05.007.PubMedView ArticleGoogle Scholar
- Grosse-Gehling P, Fargeas CA, Dittfeld C, Garbe Y, Alison MR, Corbeil D, Kunz-Schughart LA: CD133 as a biomarker for putative cancer stem cells in solid tumours: limitations, problems and challenges. J Pathol. 2013, 229: 355-378. 10.1002/path.4086.PubMedView ArticleGoogle Scholar
- Mak AB, Nixon AML, Kittanakom S, Stewart JM, Chen GI, Curak J, Gingras A-C, Mazitschek R, Neel BG, Stagljar I, Moffat J: Regulation of CD133 by HDAC6 promotes β-catenin signaling to suppress cancer cell differentiation. Cell Reports. 2012, 2: 951-963. 10.1016/j.celrep.2012.09.016.PubMed CentralPubMedView ArticleGoogle Scholar
- Marquardt JU, Thorgeirsson SS: Stem cells in hepatocarcinogenesis: evidence from genomic data. Semin Liver Dis. 2010, 30: 26-34. 10.1055/s-0030-1247130.PubMed CentralPubMedView ArticleGoogle Scholar
- Arzumanyan A, Reis HMGPV, Feitelson MA: Pathogenic mechanisms in HBV- and HCV-associated hepatocellular carcinoma. Nat Rev Cancer. 2013, 13: 123-135. 10.1038/nrc3449.PubMedView ArticleGoogle Scholar
- Mani SA, Guo W, Liao M-J, Eaton EN, Ayyanan A, Zhou AY, Brooks M, Reinhard F, Zhang CC, Shipitsin M, Campbell LL, Polyak K, Brisken C, Yang J, Weinberg RA: The epithelial-mesenchymal transition generates cells with properties of stem cells. Cell. 2008, 133: 704-715. 10.1016/j.cell.2008.03.027.PubMed CentralPubMedView ArticleGoogle Scholar
- You H, Ding W, Rountree C: Epigenetic Regulation of Cancer Stem Cell Marker CD133 by Transforming Growth Factor-beta. Hepatol. 2010, 51: 1635-1644. 10.1002/hep.23544.View ArticleGoogle Scholar
- Bergman Y, Cedar H: DNA methylation dynamics in health and disease. Nat Struct Mol Biol. 2013, 20: 274-281. 10.1038/nsmb.2518.PubMedView ArticleGoogle Scholar
- Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM, Delano D, Zhang L, Schroth GP, Gunderson KL, Fan J-B, Shen R: High density DNA methylation array with single CpG site resolution. Genomics. 2011, 98: 288-295. 10.1016/j.ygeno.2011.07.007.PubMedView ArticleGoogle Scholar
- Ohishi W, Cologne JB, Fujiwara S, Suzuki G, Hayashi T, Niwa Y, Akahoshi M, Ueda K, Tsuge M, Chayama K: Serum interleukin-6 associated with hepatocellular carcinoma risk: A nested case-control study. Int J cancer J Int du Cancer. 2014, 134: 154-163. 10.1002/ijc.28337.View ArticleGoogle Scholar
- Callahan JF, Burgess JL, Fornwald JA, Gaster LM, Harling JD, Harrington FP, Heer J, Kwon C, Lehr R, Mathur A, Olson BA, Weinstock J, Laping NJ: Identification of novel inhibitors of the transforming growth factor beta1 (TGF-beta1) type 1 receptor (ALK5). J Med Chem. 2002, 45: 999-1001. 10.1021/jm010493y.PubMedView ArticleGoogle Scholar
- Inman GJ, Nicolás FJ, Callahan JF, Harling JD, Gaster LM, Reith AD, Laping NJ, Hill CS: SB-431542 is a potent and specific inhibitor of transforming growth factor-beta superfamily type I activin receptor-like kinase (ALK) receptors ALK4, ALK5, and ALK7. Mol Pharmacol. 2002, 62: 65-74. 10.1124/mol.62.1.65.PubMedView ArticleGoogle Scholar
- Slieker RC, Bos SD, Goeman JJ, Bovée JV, Talens RP, van der Breggen R, Suchiman HED, Lameijer E-W, Putter H, van den Akker EB, Zhang Y, Jukema JW, Slagboom PE, Meulenbelt I, Heijmans BT: Identification and systematic annotation of tissue-specific differentially methylated regions using the Illumina 450k array. Epigenetics Chromatin. 2013, 6: 26-10.1186/1756-8935-6-26.PubMed CentralPubMedView ArticleGoogle Scholar
- Varley KE, Gertz J, Bowling KM, Parker SL, Reddy TE, Pauli-Behn F, Cross MK, Williams BA, Stamatoyannopoulos JA, Crawford GE, Absher DM, Wold BJ, Myers RM: Dynamic DNA methylation across diverse human cell lines and tissues. Genome Res. 2013, 23: 555-567. 10.1101/gr.147942.112.PubMed CentralPubMedView ArticleGoogle Scholar
- Sasaki A, Kamiyama T, Yokoo H, Nakanishi K, Kubota K, Haga H, Matsushita M, Ozaki M, Matsuno Y, Todo S: Cytoplasmic expression of CD133 is an important risk factor for overall survival in hepatocellular carcinoma. Oncol Reports. 2010, 24: 537-546.Google Scholar
- Dumont N, Wilson MB, Crawford YG, Reynolds PA, Sigaroudinia M, Tlsty TD: Sustained induction of epithelial to mesenchymal transition activates DNA methylation of genes silenced in basal-like breast cancers. Proc Natl Acad Sci United States Am. 2008, 105: 14867-14872. 10.1073/pnas.0807146105.View ArticleGoogle Scholar
- Papageorgis P, Lambert AW, Ozturk S, Gao F, Pan H, Manne U, Alekseyev YO, Thiagalingam A, Abdolmaleky HM, Lenburg M, Thiagalingam S: Smad signaling is required to maintain epigenetic silencing during breast cancer progression. Cancer Res. 2010, 70: 968-978. 10.1158/0008-5472.CAN-09-1872.PubMed CentralPubMedView ArticleGoogle Scholar
- Ding W, Dang H, You H, Steinway S, Takahashi Y, Wang HG, Liao J, Stiles B, Albert R, Rountree CB: miR-200b restoration and DNA methyltransferase inhibitor block lung metastasis of mesenchymal-phenotype hepatocellular carcinoma. Oncog. 2012, 1: e15-10.1038/oncsis.2012.15.View ArticleGoogle Scholar
- Wu C-Y, Tsai Y-P, Wu M-Z, Teng S-C, Wu K-J: Epigenetic reprogramming and post-transcriptional regulation during the epithelial-mesenchymal transition. Trends Genet. 2012, 28: 454-463. 10.1016/j.tig.2012.05.005.PubMedView ArticleGoogle Scholar
- Peinado H, Cano A, Esteller M, Ballestar E: Snail Mediates E-Cadherin Repression by the Recruitment of the Sin3A/Histone Deacetylase 1 (HDAC1)/HDAC2 Complex. Mol Cell Biol. 2004, 24: 306-319. 10.1128/MCB.24.1.306-319.2004.PubMed CentralPubMedView ArticleGoogle Scholar
- Hou Z, Rauscher FJ, Longmore GD, Langer EM, Yan K-P, Ayyanathan K, Peng H: The LIM Protein AJUBA Recruits Protein Arginine Methyltransferase 5 To Mediate SNAIL-Dependent Transcriptional Repression. Mol Cell Biol. 2008, 28: 3198-3207. 10.1128/MCB.01435-07.PubMed CentralPubMedView ArticleGoogle Scholar
- Abell AN, Jordan NV, Huang W, Prat A, Midland AA, Johnson NL, Granger DA, Mieczkowski PA, Perou CM, Gomez SM, Li L, Johnson GL: MAP3K4/CBP-regulated H2B acetylation controls epithelial-mesenchymal transition in trophoblast stem cells. Cell Stem Cell. 2011, 8: 525-537. 10.1016/j.stem.2011.03.008.PubMed CentralPubMedView ArticleGoogle Scholar
- Chaffer CL, Marjanovic ND, Lee T, Bell G, Kleer CG, Reinhardt F, D’Alessio AC, Young RA, Weinberg RA: Poised chromatin at the ZEB1 promoter enables breast cancer cell plasticity and enhances tumorigenicity. Cell. 2013, 154: 61-74. 10.1016/j.cell.2013.06.005.PubMed CentralPubMedView ArticleGoogle Scholar
- Haraguchi N, Ishii H, Mimori K, Tanaka F, Ohkuma M, Kim HM, Akita H, Takiuchi D, Hatano H, Nagano H, Barnard GF, Doki Y, Mori M: CD13 is a therapeutic target in human liver cancer stem cells. J Clin Investig. 2010, 120: 3326-3339. 10.1172/JCI42550.PubMed CentralPubMedView ArticleGoogle Scholar
- Hernández-Vargas H, Palacios J, Moreno-Bueno G: Molecular profiling of docetaxel cytotoxicity in breast cancer cells: uncoupling of aberrant mitosis and apoptosis. Oncogene. 2007, 26: 2902-2913. 10.1038/sj.onc.1210102.PubMedView ArticleGoogle Scholar
- Hernandez-Vargas H, Lambert M-P, Le Calvez-Kelm F, Gouysse G, McKay-Chopin S, Tavtigian SV, Scoazec J-Y, Herceg Z: Hepatocellular carcinoma displays distinct DNA methylation signatures with potential as clinical predictors. PLoS ONE. 2010, 5: e9749-10.1371/journal.pone.0009749.PubMed CentralPubMedView ArticleGoogle Scholar
- Ouzounova M, Vuong T, Ancey P-B, Ferrand M, Durand G, Le-Calvez Kelm F, Croce C, Matar C, Herceg Z, Hernandez-Vargas H: MicroRNA miR-30 family regulates non-attachment growth of breast cancer cells. BMC Genomics. 2013, 14: 139-10.1186/1471-2164-14-139.PubMed CentralPubMedView ArticleGoogle Scholar
- Du P, Kibbe WA, Lin SM: lumi: a pipeline for processing Illumina microarray. Bioinform. 2008, 24: 1547-1548. 10.1093/bioinformatics/btn224.View ArticleGoogle Scholar
- Wang J, Duncan D, Shi Z, Zhang B: WEB-based GEne SeT AnaLysis Toolkit (WebGestalt): update 2013. Nucleic Acids Res. 2013, 41 (Web Server issue): W77-W83.PubMed CentralPubMedView ArticleGoogle Scholar
- Marabita F, Almgren M, Lindholm ME, Ruhrmann S, Fagerström-Billai F, Jagodic M, Sundberg CJ, Ekström TJ, Teschendorff AE, Tegnér J, Gomez-Cabrero D: An evaluation of analysis pipelines for DNA methylation profiling using the Illumina HumanMethylation450 BeadChip platform. Epigenetics. 2013, 8: 333-346. 10.4161/epi.24008.PubMed CentralPubMedView ArticleGoogle Scholar
- Dedeurwaerder S, Defrance M, Bizet M, Calonne E, Bontempi G, Fuks F: A comprehensive overview of Infinium HumanMethylation450 data processing. Brief Bioinform. 2013, [Epub ahead of print]Google Scholar
- Pidsley R, Wong CCY, Volta M, Lunnon K, Mill J, Schalkwyk LC: A data-driven approach to preprocessing Illumina 450K methylation array data. BMC Genomics. 2013, 14: 293-10.1186/1471-2164-14-293.PubMed CentralPubMedView ArticleGoogle Scholar
- Chen Y, Lemire M, Choufani S, Butcher DT, Grafodatskaya D, Zanke BW, Gallinger S, Hudson TJ, Weksberg R: Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray. Epigenetics. 2013, 8: 203-209. 10.4161/epi.23470.PubMed CentralPubMedView ArticleGoogle Scholar
- Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero D, Beck S: A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics. 2013, 29: 189-196. 10.1093/bioinformatics/bts680.PubMed CentralPubMedView ArticleGoogle Scholar
- Du P, Zhang X, Huang C-C, Jafari N, Kibbe WA, Hou L, Lin SM: Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis. BMC Bioinformatics. 2010, 11: 587-10.1186/1471-2105-11-587.PubMed CentralPubMedView ArticleGoogle Scholar
- Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD: The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics. 2012, 28: 882-883. 10.1093/bioinformatics/bts034.PubMed CentralPubMedView ArticleGoogle Scholar
- Smyth GK: Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004, 3: Article3Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.