Establishment of human iPSCs
Human iPSCs derived from fetal lung fibroblasts (MRC5), amnion (AM), endometrium (UtE), placental artery endothelium (PAE) and menstrual blood cells (Edom) were independently established in our laboratory by retroviral infection of 4 genes (OCT-3/4
, and KLF4
) ( and Table S1
). These cells clearly showed human ES-like characters in terms of morphology; cell-surface antigens; gene expression of stem cell markers; teratoma formation in which these cells differentiated to various tissues including neural tissues (ectoderm), cartilage (mesoderm), and epithelial tissues (endoderm); growth (more than 20 passages); and DNA methylation patterns at OCT-3/4
promoter regions (Figures S1
). Short tandem repeat (STR) analysis showed clonality between the respective iPSC lines and their parent cells (Table S2
). Silencing of transgenes and normal karyotypes of iPSCs were also confirmed (Figure S4
and Table S3
Pluripotent stem cells are significantly more hyper-methylated than their parent cells.
Analysis of DNA methylation profiles
To investigate the dynamics of DNA methylation in pluripotent stem cells, we examined 5 ESC lines (HUESCs) 
, 22 iPSC lines, their parent cells and 201B7, using Illumina's Infinium HumanMethylation27 BeadChip. In total, 24,273 CpG sites in 13,728 genes were analyzed, along with 33 human cell lines (Table S1
). The iPSC line “201B7” was generated from human skin fibroblasts 
. Quantitative scores of DNA methylation levels were obtained as β-values determined from the Illumina analysis, ranging from “0”, for completely unmethylated, to “1”, for completely methylated. We also performed genome-wide gene expression analysis using the Agilent Whole Human Genome Microarray chips. As assessed by unsupervised hierarchical clustering analysis and scatter plot of DNA methylation and gene expression data, human iPSCs could be clearly discriminated from their parent cells and were similar to ESCs ( and Figure S5
). The distribution of DNA methylation levels shows that the degree of global methylation in pluripotent stem cells was higher compared to the parent cells (), suggesting that a global gain of DNA methylation occurs during reprogramming.
Identification of stem cell-specific differentially methylated regions (DMRs)
For further analysis, we defined DMR as representing a CpG site whose score differed 0.3 points or more from the β-value between the two groups. By comparison among ESCs (average from 5 lines), iPSCs (average from 22 lines), and parent cells (average from 5 lines), about 90% of the CpG sites (17,572 sites) examined did not show differential methylation among ESCs, iPSCs and parent cells (), suggesting that only a small number of the CpG sites is affected during reprogramming. The number of the CpG sites has been reported to be larger by genome-wide analysis 
Defining stem cell-specific DMRs as novel epigenetic iPS markers.
We then identified 220 sites that are pluripotent stem cell-specific DMRs (). The 174 sites (79.5%) of the stem cell-specific DMRs had significantly higher methylation levels in iPSCs/ESCs when compared to the parent cells (). Approximately 80% of the DMRs between the iPSCs and their parent cells changed to a “hyper-methylated” state from a “hypo-methylated” state in iPSCs. In contrast, 45 sites of the stem cell-specific DMRs are hypo-methylated in iPSCs/ESCs, compared with the parent cells. Gene ontology analysis indicates that the hypo-methylated stem cell-specific DMRs especially included genes related to mRNA transcription regulation (). Interestingly, the majority of the hypo-methylated stem cell-specific DMRs were located on CpG islands, whereas the majority of the hyper-methylated stem cell-specific DMRs were located on non-CpG islands (). No iPS-specific DMRs were detected. We extracted 3,123 sites that are differentially methylated in one or more parent-specific iPSCs, compared to their parent cells, because DMRs are dependent on parent cell types (Figure S6
). These DMRs are here designated as stem cell-required DMRs. Distribution analysis of the stem cell-required DMRs revealed a dispersed pattern rather than specific localization on the genome (Figure S7A
From the combined gene expression and DNA methylation data, we chose 27 genes in the stem cell-specific DMRs showing more than a 5-fold change in expression of human iPSCs/ESCs, as compared with those in the parent cells (Table S4
). Nine genes with hypo-methylated stem cell-specific DMRs were found in the group “genes significantly expressed in iPSCs/ESCs,” and 17 genes with hypo-methylated stem cell-specific DMRs belonged to the category “low expression or silenced in iPSCs/ESCs”. In addition, the methylation state and gene expression in EPHA1
were confirmed by quantitative combined bisulfite restriction analysis (COBRA) 
(), RT-PCR () and bisulfite sequencing ().
We also extracted genes with stem cell-required DMRs exhibiting high expression or suppression in human iPSCs/ESCs (Tables S5
). Interestingly, gene ontology analysis of the genes with stem cell-required DMRs showed that genes in the transcription factor category were detected only in the hypo-methylated stem cell-required DMRs (Table S7
). The top 20 transcription factor genes with hypo-methylated stem cell-required DMRs exhibiting high expression in human iPSCs are summarized in and include OCT-4/3
(also known as POU5F1
, and FOXD1
List of the top 20 out of 82 transcription factor genes with hypo-methylated stem cell-required DMRs exhibiting “high” expression in human iPS cells.
Aberrant and inherited methylation in iPSCs
Few changes in DNA methylation were detected between iPS and ES cells and these were not consistent among the different iPS lines (, Figures S6
). In further analyses, we compared the DNA methylation states of each iPSC line or each parent cell line with that of ESCs (averaged value) (). For the whole genome, the number of DMRs between ESCs and iPSCs (ES-iPS-DMRs) varied in the 22 iPSC lines (). A comprehensive analysis of methylation in ESCs and iPSCs identified 1,459 ES-iPS-DMRs covering 1,260 genes that were differentially methylated in one or more iPSC lines. ES-iPS-DMRs are composed of aberrant (iPS-specific) methylation sites, in comparison with ESCs and inherited methylation sites from the parent cells. The number of inherited sites as well as aberrant sites varied among iPSCs. Analysis of the ES-iPS-DMRs on each chromosome showed a characteristic distribution of the ES-iPS-DMRs on the X chromosome in XX-iPSCs ( and Figure S8
). Female XX-iPSCs demonstrate a tendency to carry a large number of ES-iPS-DMRs on the X chromosome, but male XY-iPSCs had few ES-iPS-DMRs on the X chromosome (, lower panel). While no ES-iPS-DMRs overlapped for all the iPSCs (), 20 ES-iPS-DMRs overlapped in more than 15 out of 22 lines (, inset). These 20 ES-iPS-DMRs include the genes for MPG
(N-methylpurine-DNA glycosylase isoform b), FZD10
(frizzled 10), IREX2
(iroquois homeobox protein 2) and ZNF248
(zinc finger protein 248), which are highly associated with aberrant methylation during reprogramming. Distribution analysis of the ES-iPS-DMRs across the genome did not show any specific localization (Figure S9
). We further compared overlapping ES-iPS-DMRs in reference to a genome-wide methylation analysis 
, and found that 72 gene promoters overlapped between our data and that of Lister et al..
Aberrant methylation in human iPSCs.
More than 70% of the ES-iPS-DMRs were hyper-methylated in each iPSC (), indicating that the iPSC genome is more methylated than the ESC genome. In addition, the majority of the ES-iPS-DMRs were located on CpG islands (), suggesting that aberrant methylation is biased towards CpG islands.
Effect of long-term culture on DNA methylation status in iPSCs
We investigated the effect of continuous passaging on the DNA methylation profile of human iPSCs. To address the effect, we subjected 7 iPSC lines to additional rounds of passaging under identical culture conditions, and obtained genomic DNA and RNA at passage 4 (P4) to P40 for DNA methylation and gene expression. The number of the ES-iPS-DMRs ranged from 80 in MRC-iPS-25 to 286 in UtE-iPS-11 at early passage (P10 to P20), whereas the number of the ES-iPS-DMRs dramatically decreased in all lines at late passage (P30 to P40) (, upper-left panel). The number of inherited and aberrant sites decreased to 30 and 70, respectively, at P30 to P40 (, upper-center and right panels). These decreases in the numbers of ES-iPS-DMRs indicate that iPSCs have become closer to ESCs in their DNA methylation profiles. In particular, XX-iPSC lines (AM-iPS-8, UtE-iPS-4 and -11, and Edom-iPS-2) showed decreases in the number of ES-iPS-DMRs with passaging. The XY-iPSC lines, such as MRC-iPS-25 and PAE-iPS-1, had only a small number of ES-iPS-DMRs. The number of ES-iPS-DMRs continued to decrease to approximately 100 ES-iPS-DMRs containing 30 inherited sites. Intriguingly, few ES-iPS-DMRs on the X chromosome were detected in XY-iPSCs throughout the passaging. In contrast, the number of ES-iPS-DMRs in XX-iPSCs ranged from 10 to 70 at the early passage (P4 to P20), and decreased to zero after P30 (, lower panels). We also investigated the effect of continuous passaging on the DNA methylation profile of the parent cells (UtE1104 and Edom22) (). The number of the DMRs between ESCs and parent cells (ES-parent-DMRs) increased with passaging. In addition, we also confirmed that the transgenes were silenced at each passage ( and Figure S4
), indicating that the decreasing number of the ES-iPS-DMRs in iPSCs occurred in the transgene-independent phase.
Effect of long-term cultivation on ES-iPS-DMRs.
Comparative analysis of ES-iPS-DMRs dynamics
We then compared each ES-iPS-DMRs with passaging. The UtE-iPS-11 had 286 ES-iPS-DMRs at P13, 194 sites at P18, 110 sites at P31, and 55 sites at P39. The ES-iPS-DMRs detected at P13 decreased with passaging (blue bars in upper-left panel in ). Interestingly, 66 de novo ES-iPS-DMRs appeared at P18, while at P13 these sites showed no differences between UtE-iPS-11 and ESCs (orange bars in upper-left panel in ). These 66 ES-iPS-DMRs also decreased with passaging (P31 and P39). The 29 additional ES-iPS-DMRs at P31 also appeared and decreased with passaging (P39) (green bars in upper-left panel in ) and 16 ES-iPS-DMRs at P39 (red bar in upper-left panel in ) appeared. Rapid appearance and gradual disappearance of ES-iPS-DMRs was a recurring theme, but the number of newly-appearing ES-iPS-DMRs decreased with passaging (, upper-left panel). The same change in ES-iPS-DMRs occurred on the X chromosome, but the number of the ES-iPS-DMRs approached zero at early passages (, upper-center panel). Intriguingly, this change also occurred at inherited sites, which was contrary to our expectations. The inherited sites also repeatedly appeared and disappeared, and the number of newly-appearing inherited sites decreased with passaging (, upper-right panel). The term “inherited” is here used to mean the same methylation state found in iPSCs and their parent cells, but the “inherited” regions behaved like “aberrant” regions that had multiple appearances and disappearances. These multiple appearances/disappearances of ES-iPS-DMRs were observed in all iPSC lines regardless of parental cell type. The ES-parent-DMRs were also analyzed. The de novo ES-parent-DMRs appeared as well as the ES-iPS-DMRs, but did not decrease with passaging ().
Number of the ES-iPS-DMRs and ES-parent-DMRs with passaging.
Most ES-iPS-DMRs were hyper-methylated in iPSCs
ES-iPS-DMRs can be categorized into two groups: a, hyper-methylated and b, hypo-methylated sites in iPSCs, as compared with ESCs. ES-iPS-DMRs that disappeared at the last passage (P39) (blue bars in ) in both UtE-iPS-11 and Edom-iPS-2 were extracted, and each methylation score of the extracted ES-iPS-DMRs is shown (, upper and middle panels). To compare methylation scores, a “difference value” was estimated by subtracting the scores of ESCs from those of each cell (, lower panels). Positive and negative difference values indicate that these sites are hyper- and hypo-methylated, respectively, when compared with ESCs. Difference values of the ES-iPS-DMRs showing aberrant methylation states in iPSCs at the early passage approached zero with passaging. It should be noted that the almost all difference values became largely positive in iPSCs at early passage (P13 or P22), even though they were negative in the parent cells, and then approached zero upon further passaging. This transiently-induced hyper-methylation was observed at each passage in all iPSC lines examined. The observed transient hypermethylation patterns during iPS reprogramming did not correspond to methyflated CpGs in the parental cells. However, this observation does not rule out that transient aberrant methylation could also be observed in some cases on sites that were methylated in the parental cells.
Hyper-methylation in the ES-iPS-DMRs and ES-parent-DMRs.