DNA Hypermethylation in Somatic Cells Correlates with Higher Reprogramming Efficiency§

Authors

  • María J. Barrero,

    1. Center for Regenerative Medicine in Barcelona, Barcelona, Catalonia, Spain
    Search for more papers by this author
  • María Berdasco,

    1. Cancer Epigenetics Group, Cancer Epigenetics and Biology Program (PEBC), Bellvitge Institute for Biomedical Research (IDIBELL), L'Hospitalet de Llobregat, Barcelona, Catalonia, Spain
    Search for more papers by this author
  • Ida Paramonov,

    1. Center for Regenerative Medicine in Barcelona, Barcelona, Catalonia, Spain
    Search for more papers by this author
  • Josipa Bilic,

    1. Center for Regenerative Medicine in Barcelona, Barcelona, Catalonia, Spain
    Search for more papers by this author
  • Marianna Vitaloni,

    1. Center for Regenerative Medicine in Barcelona, Barcelona, Catalonia, Spain
    Search for more papers by this author
  • Manel Esteller,

    Corresponding author
    1. Cancer Epigenetics Group, Cancer Epigenetics and Biology Program (PEBC), Bellvitge Institute for Biomedical Research (IDIBELL), L'Hospitalet de Llobregat, Barcelona, Catalonia, Spain
    2. Institució Catalana de Recerca I Estudis Avançats (ICREA), Barcelona, Catalonia, Spain
    • Manel Esteller, Cancer Epigenetics Group, Cancer Epigenetics and Biology Program (PEBC), Bellvitge Institute for Biomedical Research (IDIBELL), L'Hospitalet de Llobregat, 08908 Barcelona, Catalonia, Spain

      Juan Carlos Izpisua Belmonte, Salk Institute for Biological Studies, 10010 N. Torrey Pines Rd., La Jolla, CA 92037 USA

    Search for more papers by this author
    • Telephone: +34-93-2607253; Fax: +34-93-2607219

  • Juan Carlos Izpisua Belmonte

    Corresponding author
    1. Center for Regenerative Medicine in Barcelona, Barcelona, Catalonia, Spain
    2. Salk Institute for Biological Studies, La Jolla, California, USA
    • Manel Esteller, Cancer Epigenetics Group, Cancer Epigenetics and Biology Program (PEBC), Bellvitge Institute for Biomedical Research (IDIBELL), L'Hospitalet de Llobregat, 08908 Barcelona, Catalonia, Spain

      Juan Carlos Izpisua Belmonte, Salk Institute for Biological Studies, 10010 N. Torrey Pines Rd., La Jolla, CA 92037 USA

    Search for more papers by this author

  • Author contributions: M.J.B. and M.B.: conceived and designed the project, provided materials, collected and interpreted data, provided financial support, and wrote manuscript; I.P.: performed data analysis and interpretation, J.B. and M.V.: provided materials and collected data; M.E. and J.C.I.B.: conceived, designed and supervised the project, provided financial support, wrote manuscript, and final approval of manuscript. M.J.B and M.B. contributed equally to this article.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS May 31, 2012.

Abstract

The efficiency of somatic cell reprogramming to pluripotency using defined factors is dramatically affected by the cell type of origin. Here, we show that human keratinocytes, which can be reprogrammed at a higher efficiency than fibroblast [Nat Biotechnol 2008;26:1276-1284], share more genes hypermethylated at CpGs with human embryonic stem cells (ESCs) than other somatic cells frequently used for reprogramming. Moreover, pluripotent cells reprogrammed from keratinocytes (KiPS) are more similar to ESCs than those reprogrammed from fibroblasts (FiPS) in regard to DNA methylation levels, mostly due to the presence of genes that fail to acquire high levels of DNA methylation in FiPS cells. We propose that higher reprogramming efficiency correlates with the hypermethylation of tissue-specific genes rather than with a more permissive pluripotency gene network. STEM CELLS2012;30:1696–1702

INTRODUCTION

The process of somatic cell reprogramming to pluripotency entails dramatic changes in the epigenetic landscape of cells. DNA and histone modifications appear critical for this process, since histone deacetylases and DNA and histone methyltransferases inhibitors can modulate the efficiency of reprogramming [1, 2]. Moreover, the incomplete erasure of the epigenetic signature of somatic cells in early passage induced pluripotent (iPS) cells can influence their differentiation properties [3, 4]. Several studies have compared the genome-wide DNA methylation profiles of iPS cells and embryonic stem cells (ESCs) concluding that these patterns are very similar but some aberrations, both stochastic and shared between different iPS lines, can be found [5–7]. However, the comparison of the epigenome of iPS cells derived from somatic cell types with different reprogramming efficiency remains unexplored.

Certain cell types can be reprogrammed to pluripotency with higher efficiency, as judged by the appearance of a larger number of pluripotent colonies or by the possibility to reprogram using a lower number of factors. Reasons for this have been suggested to be mostly due to a more permissive environment of the regulatory regions of pluripotency genes or to the higher level of expression of certain reprogramming factors [8–10], although the epigenetic status of differentiation genes might also be playing a role. Here, we have approached this question by analyzing the DNA methylation patterns of human ESCs, induced pluripotent stem (iPS) cells reprogrammed from keratinocytes (KiPS) or fibroblast (FiPS), and their respective cells of origin. To avoid the interference of laboratory-specific signatures, we used pluripotent cells that were obtained or derived in the same laboratory, transduced with the same number of factors, and cultured under the same conditions.

MATERIALS AND METHODS

Cell Culture and Sample Processing

Pluripotent cell lines were grown in Matrigel-coated dishes and in mouse embryonic fibroblast (MEF)-conditioned HES media containing fibroblast growth factor. These included two human ESC lines (ES[4] and ES[2] [11]), two iPS cell lines reprogrammed from fibroblasts (FiPS4F7 and FiPS4F8 [12]), two iPS cell lines reprogrammed from foreskin keratinocytes (KiPS4F8 and KiPS4F1 [8]), and one iPS cell line reprogrammed from plucked hair keratinocytes ([H]KiPS4F1 [8]). Pluripotent cells were used at passages larger than 15, to ensure that the reprogramming process is complete. Primary cultures of fibroblast (lines HFF and F3) and keratinocytes (lines K1, K2, K3, and MMTA) were performed as previously described [13]. KiPS4F8 and MMTA are the only lines with identical genetic background. Total RNA and genomic DNA were extracted using the RNAeasy kit and the DNeasy Blood and Tissue kit from Qiagen (Duesseldorf, Germany), respectively.

DNA Methylation Profiling Using Universal Bead Arrays

Microarray-based DNA methylation profiling was performed with the HumanMethylation27 BeadChip Infinium Methylation Arrays (Illumina, Inc., San Diego, CA), which interrogates 27,578 CpG loci located in the regulatory regions of 14,495 genes and 110 microRNAs. Briefly, bisulfite conversion of 1 μg of genomic DNA was done using the CpGenomic DNA Modification Kit (Intergen Company, Purchase, NY). After bisulfite conversion, each sample was whole-genome amplified, enzymatically fragmented, purified and applied to the BeadChips. DNA methylation beta values are continuous variables between 0 (completely unmethylated) and 1 (completely methylated), representing the ratio of the intensity of the methylated bead type to the combined locus intensity.

Hierarchical Cluster Analysis and Definition of DNA Methylation Groups

After excluding 1,085 gender-specific CpGs and 289 low-quality CpGs, 26,259 CpGs were used in the subsequent statistical analyses. Hierarchical clustering was performed on all the studied samples using the Cluster Analysis tool of the BeadStudio software (version 3.2). Beta values of CpGs higher than 0.75 in at least 70% of the samples from each category were considered hypermethylated sequences.

Identification of Differentially Methylated Regions Between iPS Cells and ESCs

To increase the stringency of the analysis, we considered probes as differentially hypermethylated between categories when beta values of CpGs were higher than 0.75 in at least 70% of the samples from one category and lower than 0.75 in at least 70% of the samples from the other category. We considered probes as differentially hypomethylated between categories when beta values of CpGs were lower than 0.25 in at least 70% of the samples from one category and higher than 0.25 in at least 70% of the samples from the other category.

Definition of Promoter Classes

We classified the 27,578 probes included into the methylation arrays into three categories: HCP (high CpG content promoters), ICP (intermediate CpG content promoters), and LCP (low CpG content promoters) [14]. We determined the GC content and the ratio of observed versus expected CpG dinucleotides in a surrounding 500 bp window. The CpG ratio was calculated using the following formula: (number of CpGs × number of bp) (number of Cs × number of Gs). Based on their entire CpG contents across the genomic region, the following criteria was applied: HCP contains a 500-bp region with a GC content >0.55 and a CpG observed to expected ratio >0.6, LCP contains no 500-bp interval and with a CpG observed to expected ratio >0.4, and ICP contains CpG density between HCP and ICP.

Definition of Bivalent Domain-Containing Genes

Bivalent domain-containing genes have been extracted from three independent studies [15–17]. We considered bivalent genes those reported with bivalent domains in human ESCs in at least two of the three datasets. The enrichment statistical significance was determined using a hypergeometric test.

Comparison of DNA Methylation in Fetal, Extra-Embryonic, and Adult Tissues

To extend our comparison analysis of hypermethylated genes in ESCs and somatic cells, we collected previously published DNA methylation data performed on the Illumina HumanMethylation27 BeadChip: GSE25538, GSE24676, GSE27284. Probes with a detection p value >0.05 in at least one sample were excluded from the analysis, leaving 23,413 probes in total. For each probe, a median of Illumina Average beta values was calculated for each group of samples. A probe was considered as hypermethylated if it has a methylation level above 0.75.

Bisulfite Sequencing

The methylation status of specific genomic DNA sequences was established by bisulfite genomic sequencing. After polymerase chain reaction (PCR) and cloning, automatic sequencing of 10 colonies for each sequence was performed to measure the methylation status of every single CpG dinucleotide. Primer sequences for methylation analysis can be found in Supporting Information Table 1.

Genome-Wide Expression Data Analysis

For microarray hybridization, 100 ng of total RNA was labeled using Low Input Quick Amp Labeling kit (Agilent 5190-2305) following manufacturer's instructions and hybridized to the Agilent SurePrint G3 Human gene expression 8 × 60K microarray according to the manufacturer's protocol. The arrays were washed and scanned on an Agilent G2565CA microarray scanner at 100% photomultiplier tube (PMT) and 3 μm resolution. Intensity data were extracted using the Feature Extraction software (Agilent, Santa Clara, CA).

Alternatively, the GeneChip microarray processing was performed according to the manufacturer's protocols (Affymetrix, Santa Clara, CA). The amplification and labeling were processed as indicated in Nugen protocol with 25 ng starting RNA. For each sample, 3.75 μg single-stranded DNA (ssDNA) was labeled and hybridized to the Affymetrix HG-U133 Plus 2.0 chips. Expression signals were scanned on an Affymetrix GeneChip Scanner (7G upgrade). The data extraction was done by the Affymetrix GCOS software v.1.4.

Quantitative Polymerase Chain Reaction (qPCR) Analysis

Total mRNA was isolated using the RNAeasy kit from Qiagen, and 1 μg was used to synthesize cDNA using the Invitrogen Cloned AMV First-Strand cDNA synthesis kit. One microliter of the cDNA reaction was used to quantify gene expression of DNA methyltransferases (DNMTs) by qPCR using the primers included in Supporting Information Table 1.

RESULTS

Human iPS cells and ESCs Have Very Similar Expression and DNA Methylation Profiles

Unsupervised clustering of the samples exclusively using the methylation signals of the CpGs contained in the Infinium HumanMethylation 27K arrays from Illumina enabled the classification of all samples into two discrete groups: differentiated primary tissues (grouped in function of their cell type) and a second group containing both types of pluripotent cells (iPS cells and ESCs) (Fig. 1A). Scatter plots of representative samples from each group confirm these results and revealed a reduced similarity and overlap (r2 = 0.74–0.75) of CpG methylation between differentiated tissues and pluripotent cells (Supporting Information Fig. 1a), in contrast to the high similarity (r2 = 0.94–0.96) between the iPS cells and ESCs methylation patterns (Supporting Information Fig. 1b). Importantly, we found that the correlation between samples of different genetic background (KiPS4F1and MMTA r2 = 0.75, KiPS4F8 and K1 r2 = 0.74 or KiPS4F1 and K1 r2 = 0.74) was very similar to the correlation between samples of identical genetic background (KiPS4F8 and MMTA r2 = 0.75), suggesting that the genetic background does not represent an important bias in our analysis.

Figure 1.

Genome-wide CpG methylation profiling of ESCs, iPS, and somatic cells of origin. (A): Hierarchical cluster analysis and heatmap displaying the differential groups of CpGs according to their DNA methylation profile in all the analyzed samples. HumanMethylation27 BeadChip Infinium Methylation Arrays had been performed in ESCs (n = 2), iPS cells (n = 5), and differentiated cells from keratinocytes and fibroblast (n = 2 and n = 4, respectively). The methylation levels vary from fully methylated (red) to fully unmethylated (green). For the estimation of the methylation percentages in each condition, the total number of high-quality CpGs (n = 27,289) was considered. (B): Left, Venn diagram analysis of hypermethylated CpGs in ESCs, KiPS, and keratinocytes. Right, Percentages of hypermethylated CpGs unique to each cell type (ESCs, KiPS, and keratinocytes) and shared between cell types, identified from Venn diagrams. (C): Left, Venn diagram analysis of hypermethylated CpGs in ESCs, FiPS, and fibroblasts. Right, Percentages of hypermethylated CpGs unique to each cell type (ESCs, FiPS, and fibroblasts) and shared between cell types, identified from Venn diagrams. (D): Percentage of differentially hypermethylated probes found in FiPS, KiPS, and in both. Fail to methylate corresponds to regions that were hypermethylated in ESCs but not in somatic and iPS cells. Aberrant undermethylation corresponds to probes that were hypermethylated in both somatic cells and ESCs but not in iPS cells. Fail to demethylate corresponds to probes hypermethylated in somatic and iPS cells but not in ESCs. Aberrant hypermethylation corresponds to probes that were hypermethylated in iPS cells but not in ESCs and somatic cells. (E): Percentage of differentially hypomethylated probes in FiPS, KiPS, and in both. Fail to methylate corresponds to regions that were hypomethylated in somatic cells and iPS cells but not in ESCs. Aberrant hypomethylation corresponds to probes that were hypomethylated in iPS cells but not in ESCs and somatic cells. Fail to demethylate corresponds to probes hypomethylated in ESCs but not in somatic cells and iPS cells. Aberrant overmethylation corresponds to probes that were hypomethylated in ESCs and somatic cells but not in iPS cells. A CpG was considered as hypermethylated when its b value from BeadArray was higher than 0.75 and hypomethylated when lower than 0.25. (F): Methylation levels at the HIST1 cluster show memory of hypomethylation at the HIST1H3C regulatory regions in KiPS cells. HIST1H1A regulatory regions display undermethylation in iPS cells, which corresponds to memory in FiPS and is aberrant in KiPS cells. FiPS cells also show aberrant overmethylation along this cluster. Abbreviations: iPS cells, induced pluripotent stem cells; ESCs, embryonic stem cells; F, fibroblasts; FiPS, iPS cells reprogrammed from fibroblast; K, keratinocytes; KiPS, iPS cells reprogrammed from keratinocyte; DMR, differentially methylated CpG.

Comparison of gene expression profiles (Supporting Information Fig. 1c, 1d) revealed correlation patterns similar to those found in methylation. These data indicate that iPS cells and ESCs have very similar expression and DNA methylation patterns.

KiPS Cells Share More Hypermethylated Genes with ESCs than FiPS Cells

The comparison of hypermethylated probes between cell lines revealed that ESCs and KiPS cells are characterized by the highest number of hypermethylation events, however, FiPS cells showed approximately half of hypermethylated probes compared to the rest of pluripotent cells analyzed, suggesting incomplete establishment of DNA methylation patterns after the reprogramming of fibroblasts (Fig. 1B, 1C). A total of 1,780 regions failed to show the same levels of hypermethylation in iPS cells compared to ESCs, reflecting both memory (levels similar to the cell of origin but different from ESCs) and aberrant methylation (levels different from ESCs and the cell of origin) (Fig. 1D; Supporting Information Fig. 2a; Supporting Information Tables 2, 3). Interestingly, differentially hypermethylated regions most shared between KiPS and FiPS cells consist on failure to establish high levels of DNA methylation (Fig. 1D). This seems particularly conspicuous in FiPS cells, in which memory events consisting on genes that undergo methylation during reprogramming but fail to reach the hypermethylated levels found in ESCs represent 66% of all differential methylation in these lines. The comparison of hypomethylated probes in ESCs and iPS cells (Fig. 1E; Supporting Information Fig. 2b; Supporting Information Tables 4, 5) showed that KiPS cells tend to retain memory of a few hypomethylated regions, while FiPS cells exhibit a more prominent aberrant overmethylation of genes that are hypomethylated in ESCs.

iPS Cells Show Both Aberrant and Residual DNA Methylation Compared to ESCs

One of the most prominent differentially methylated regions between iPS cells and ESCs mapped to the gene HIST1H3C, which shows low levels of methylation in KiPS but not in FiPS cells or ESCs (Fig. 1F). Interestingly, somatic cells show hypomethylation at the regulatory regions of this gene, as is the case for most of the histone genes located in the HIST1 cluster, but in FiPS and ESCs this gene is hypermethylated. Located in the same cluster, the HISTH1A gene also fails in iPS cells to reach the levels of DNA methylation found in ESCs. Moreover, FiPS cells consistently present higher levels of methylation along the HIST1 cluster than the rest of the cell lines analyzed.

Most aberrant hypermethylation events found in iPS cells corresponds to probes located at the TCERG1L gene regulatory regions. This region displays very low methylation levels in ESCs, keratinocytes, and fibroblasts, and it is aberrantly hypermethylated in KiPS and FiPS cells (Supporting Information Fig. 3a). DNA hypermethylation strongly correlates with lower levels of expression in FiPS and KiPS cells compared to ESCs (Supporting Information Fig. 3b). This gene has been described to be aberrantly hypermethylated in other iPS cell lines [7, 18], and it is consistently expressed at lower levels in other published iPS cell lines [19] compared to ESCs (Supporting Information Fig. 3c). Finally, the analysis of DNA methylation at probes located on the X chromosome showed high levels of methylation in [H]KiPS4F1 cells, the only female cell line analyzed, when compared to the rest of the male pluripotent lines, suggesting that the X chromosome remains inactivated after reprogramming (Supporting Information Fig. 4).

iPS Cells Show Aberrant Levels of DNA Methylation at Certain Bivalent Domains Compared to ESCs

Next, we focused on the CpG sites that change methylation after reprogramming. We identified a total of 556 differentially methylated CpGs (R-DMR) between iPS cells and their cells of origin (Supporting Information Tables 6, 7; Supporting Information Fig. 5). Gain of DNA methylation events after reprogramming are more common than loss of DNA methylation (Fig. 2A, 2B). Despite this, we found that genes that become hypomethylated during reprogramming are significantly enriched in genes previously reported to contain bivalent domains in human ESCs [15–17] (p value = 7.8 E −13 in KiPS and p value = 9.3 E −13 in FiPS) versus those that become hypermethylated (p value = 0.99 in KiPS and p value = 0.93 in FiPS). As expected, the expression levels of genes marked with R-DMR inversely correlates with the levels of DNA methylation in each cell line (Fig. 2C, 2D). To see potential differences in the levels of methylation of bivalent domain-containing genes, we compared the level of methylation of bivalent genes that are hypomethylated in ESCs (1,404 out of 1,603 bivalent genes covered by the array) between the different cell types (Fig. 2E). Most bivalent genes remain hypomethylated in somatic cells and only a small percentage is hypermethylated. Interestingly, iPS cells exhibit higher levels of DNA methylation at some of these genes, more conspicuously in FiPS cells that present several bivalent genes overmethylated and in general higher levels of methylation at these genes. None of these genes show memory from somatic cells but show aberrant overmethylation. Importantly, the presence of DNA methylation at developmental genes after reprogramming has been correlated with impaired differentiation ability [3].

Figure 2.

Changes in DNA methylation and gene expression during reprogramming. (A): Number of CpGs that become hypermethylated or hypomethylated after the reprogramming of keratinocytes distributed according to their location at LCP, ICP, or HCP regions. (B): Number of CpGs that become hypermethylated or hypomethylated after the reprogramming of fibroblasts distributed according to their location at LCP, ICP, or HCP regions. A significant enrichment (*p value <2.2e−16; test ANOVA) into the LCP class of CpGs that change their methylation status during reprogramming (from fibroblast or keratinocytes). (C): Box plot representation of the expression levels of genes that become hypomethylated (top) or hypermethylated (bottom) after the reprogramming of keratinocytes. (D): Box plot representation of the expression levels of genes that become hypomethylated (top) or hypermethylated (bottom) after the reprogramming of fibroblasts. (E): Levels of DNA methylation at bivalent domain-containing genes that are hypomethylated in ESCs. Genes that show levels higher than b value 0.4 are indicated. Abbreviations: FIPS, iPS cells reprogrammed from fibroblast; HCP, high CpG content promoters; ICP, intermediate CpG content promoters; KiPS, iPS cells reprogrammed from keratinocytes; LCP, low CpG content promoters.

Keratinocytes Share More Hypermetylated Tissue-Specific Genes with ESCs Than Fibroblasts

Gain of DNA methylation is more prominent in KiPS than in FiPS cells, making KiPS more similar to ESCs than FiPS cells (Figs. 1B, 1C, 2A, 2B). Regions that become hypermethylated after reprogramming show low CpG content (Fig. 2A, 2B). Genes found hypermethylated in ESCs have been described to correspond mainly to tissue-specific genes [20, 21]. Keratinocytes share with ESCs a higher number of hypermethylated genes when compared to fibroblasts (Fig. 1B, 1C) and overall show higher levels of methylation at these genes (Fig. 3A). This might contribute to the higher efficiency of keratinocyte reprogramming (Supporting Information Fig. 6a). Alternatively, lower levels of DNA methylation at the regulatory regions of the pluripotency genes in somatic cells may cause a more permissive chromatin environment, thereby making these genes more prone to be activated during reprogramming. To address this possibility, we tested the methylation status of 249 genes belonging to the pluripotency network based on differential gene expression between pluripotent and somatic cells [19] and identified a cluster of pluripotency genes whose differential expression in pluripotent and somatic cells is clearly regulated by DNA methylation (Fig. 3B). Comparison of the methylation levels between keratinocytes and fibroblasts reveals that keratinocytes have higher methylation levels in these pluripotency genes (Fig. 3C), indicating that the higher reprogramming efficiency does not correlate with a more permissive chromatin environment at these genes.

Figure 3.

DNA hypermethylation correlates with higher reprogramming efficiency. (A): Box blot of the levels of DNA methylation found in genes that are hypermethylated in ESCs. (B): Hierarchical cluster analysis and heatmap displaying the DNA methylation profile of pluripotency genes that are regulated by DNA methylation. The methylation levels vary from fully methylated (red) to fully unmethylated (green). (C): Box plot of methylation levels of the genes shown in (B) in ESCs, K, and F. (D): Number of probes that are hypermethylated and those shared with ESCs in different cell types. Abbreviations: ESCs, embryonic stem cells; F, fibroblasts; HUVEC, human umbilical vein endothelial cells; K, keratinocyte.

Cell Types with Higher Reprogramming Efficiency Share More Hypermethylated Genes with ESCs

To further test our hypothesis, we considered the methylation status of other cell types. Cells from extra-embryonic origin, such as endometrium [22], amnion [23], and Human Umbilical Vein Endothelial Cells (HUVEC) [24], have been reported to reprogram more efficiently (Supporting Information Fig. 6a). Also, embryonic fibroblasts reprogram more efficiently than their adult counterparts [25, 26]. Loss of DNA methylation has also been reported while comparing ESCs with fetal and adult tissues [27, 28]. The analysis of genes hypermethylated in fetal, extra-embryonic, and adult cells revealed that those which reprogram more efficiently share more hypermethylated genes with ESCs (Fig. 3D). The analysis of the methylation levels at genes hypermethylated in ESCs (Supporting Information Fig. 6b) consistently shows higher levels of methylation in fetal and extra-embryonic tissues but no consistent pattern at pluripotency genes (Supporting Information Fig. 6c).

DISCUSSION

Our data are consistent with previous reports describing that iPS cells and ESCs have very similar expression and epigenetic profiles but show a certain number of consistent differences [6]. More specifically, we found two prominent hot spots for aberrant epigenetic reprogramming, the HIST1 cluster and the TCERG1L gene. We speculate that the redundancy of histone variants might be imposing a more relaxed selective pressure for proper epigenetic reprogramming of the histone cluster. On the other hand, TCERG1L has been described to be mutated and hypermethylated in colon cancer [18], suggesting that its silencing might confer a higher proliferative advantage during reprogramming. Moreover, several iPS cell lines have been also reported to have DNA hypermethylation and reduced expression of this gene [6].

In early passage iPS cells, the incomplete erasure of DNA methylation at differentiation genes has been correlated with impaired differentiation [3] and the sustained expression of somatic genes with tendency to differentiate back into the tissue of origin [4]. Studies that analyzed late passage iPS cells did not find significant differences in the DNA methylation levels of developmental genes compared to ESCs [6]. However, we found several bivalent genes that are hypomethylated in somatic cells that showed higher levels of DNA methylation in iPS cells compared to ESCs, especially in FiPS cells. In fact, most bivalent domains were found to be hypomethylated in somatic cells. These genes might be expressed, continue bivalent, or become silenced through mechanisms independent of DNA methylation in somatic cells. Genes that become hypomethylated and bivalent after reprogramming show higher levels of expression in iPS cells than in their corresponding cells of origin, suggesting that bivalency is more permissive to transcription compared to DNA methylation. This is in agreement with the finding that in ESCs bivalent genes have notable levels of poised RNA Pol II at transcriptional start sites that might result in low but detectable rates of transcriptional elongation [29, 30].

Compared to somatic cells, pluripotent lines showed a larger number of hypermethylated probes, which map to regions with low CpG content that have been described to correspond mainly to tissue-specific genes [20, 21]. These genes are not expressed in pluripotent cells but become induced during differentiation due to the action of early developmental transcription factors. Therefore, these regions seem more permissive to transcriptional activation than methylated regions with high CpG content. Interestingly, we found that both KiPS and FiPS cells failed to reach full levels of DNA hypermethylation in several of these regions, suggesting that regaining DNA methylation is a limiting step for reprogramming. Since several of these differentially methylated regions were shared between KiPS and FiPS cells, it appears that some genomic regions are recurrently resistant to gain DNA hypermethylation. Our results are in agreement with the work of Ohi et al. [7] describing that several human iPS cell lines failed to acquire DNA hypermethylation at certain somatic genes.

We found that cell types that have more DNA hypermethylation of tissue-specific genes show higher reprogramming efficiency. Moreover, keratinocytes, which share more hypermethylated genes with ESCs, also acquire closer levels of DNA hypermethylation to ESCs after reprogramming. However, we found no significant differences in the expression levels of DNA methyltransferases between keratinocytes and fibroblasts or between FiPS and KiPS cells (Supporting Information Fig. 7). Eventually, the existence of more hypermethylated regions in keratinocytes might boost the acquirement of DNA hypermethylation in these cells during reprogramming. Importantly, many factors can influence the efficiency of reprogramming and dissecting the specific impact of DNA methylation over other factors is not straight forward. Mechanistically, the relevance of the acquirement of DNA hypermethylation for reprogramming is also difficult to assess, since DNA methylation inhibitors affect both de novo and preexistent DNA methylation. Pluripotent cells express higher levels of DNMTs (Supporting Information Fig. 7) that might be involved in the acquirement or maintenance of DNA hypermethylated patterns, but the role of particular DNMTs in this process remains unexplored. Pawlak et al. [31] showed that Dnmt3a and Dnmt3b are dispensable for the reprogramming of MEFs, but the authors did not perform genome-wide DNA methylation analysis to confirm that DNA hypermethylation typical of ESCs cannot take place. Several papers have reported increased reprogramming efficiency when cells are treated with DNA methylation inhibitors [1, 2]. However, others have reported that these same compounds promote differentiation of ESCs [32]. This discrepancy can be explained by the fact that the treatment with the inhibitors enhances reprogramming only when done at very particular time frames during the reprogramming process, and more likely helps to overcome the chromatin barrier that impedes the function of the transfactors [3, 9, 33], which have been described to target hypomethylated CpG-rich regions at early stages of reprogramming [34]. In contrast, we found a positive correlation between DNA hypermethylation in areas with low CpG content and efficiency of reprogramming. These ES-like DNA hypermethylation patterns are not fully established until later stages of reprogramming and might be crucial for the acquirement of a definitive ESC identity [34]. Moreover, the recent findings that somatic cells can be reprogrammed to pluripotency by just overexpressing certain ESCs-specific miRNAs that target differentiation genes [35] suggest that the silencing of differentiation genes plays a far more relevant role in reprogramming than previously suspected.

CONCLUSION

We propose here that the analysis of global levels of DNA methylation could help to score for somatic cell types with higher levels of DNA methylation at tissue-specific genes and therefore more amenable to the process of reprogramming.

Acknowledgements

We thank M. Carrió and L. Casano for the establishment of primary cell lines and the histology and embryo micromanipulation platforms at the CMRB for the teratoma assays. This work was supported by grants RYC-2007-01510 and SAF2009-08588 to M.J.B. M.J.B and J.B are partially supported by the Ramón y Cajal and Juan de la Cierva programs, respectively. M.E is supported by EU FP6 ESTOOLS LSHG-CT-2006-018739, SAF2007-00027-65134, Consolider CSD2006-49, Lilly Foundation, Dr. Josef Steiner Cancer Research Foundation, Cellex Foundation, and European Research Council Advanced Grant EPINORC 268626. M.B. is supported by Fundació La Marató de TV3 111430/31 and COST TD09/05. J.C.I.B. is supported by grants from the G. Harold and Leila Y. Mathers Charitable Foundation, The Leona M. and Harry B. Helmsley Charitable Trust, Sanofi, MINECO, CIBER, and Fundacion Cellex.

DISCLOSURE OF POTENTIAL CONFLICT OF INTEREST

The authors indicate no potential conflicts of interest.

Ancillary