Silencing of the DLK1-DIO3 Gene Cluster Occurs Frequently in Early hESC Cultures
To investigate possible genetic and epigenetic changes during early hESC culture, we first performed karyotype analysis and genome-wide gene expression analysis of undifferentiated hESCs from both the initial passages (ihESCs, P4–P9) and early passages (ehESCs, P20–P30) of 12 karyotypically normal hESC lines obtained from our recently established hESC bank . All 12 cell lines were found to possess a normal karyotype during early culture (data not shown). Unsupervised clustering based on gene expression revealed that the hESC lines clustered mainly by their genetic background, rather than by passage number (Supporting Information Fig. S2). Only 0.10% ± 0.09% of the transcriptome (12–136 of the 38,500 genes/isoforms) showed significant changes (>twofold, t test p < .05) between ihESCs and ehESCs of the same line (Supporting Information Fig. S3), indicating that early culture has little impact on gene expression. Surprisingly, a common pattern of gene expression change between ihESCs and ehESCs was observed across different lines; that is, MEG3, SNORD114-3, and three other expressed sequence tags in the DLK1-DIO3 imprinted cluster were significantly downregulated in most ehESCs. All 12 ihESCs cell lines expressed high levels of these genes; however, among ehESCs cell lines, the expression of these genes was completely silenced in seven lines, and relatively high in only five lines (Fig. 1A). In contrast, the expressions of pluripotency-related genes, such as POU5F1 (also known as OCT4) and NANOG, were not significantly different between ihESCs and ehESCs (Fig. 1A).
Figure 1. Expression status of the DLK1-DIO3 cluster in initial- and early-passage hESCs. (A): Heat map showing the relative expression levels of selected transcripts in ihESCs and ehESCs. The colors indicate low (3green) to high (11red) absolute log2 expression of genes levels. The presence of multiple rows for MEG3 is due to the presence of multiple probes for the gene. Asterisk (*) indicates cell lines derived under 5% oxygen. (B): Schematic representation of the DLK1-DIO3 gene cluster, with maternally and paternally expressed transcripts shown in pink and blue, respectively. Black bars mark the positions of DLK1-DIO3 IG-DMR and MEG3 promoter DMRs (RI-RIII) analyzed by Bisulfite sequencing; CpG sites are represented by vertical bars. (C): Representative scatter plots of snoRNA and miRNA sequencing data in paired ihESCs (x-axis) and ehESCs (y-axis) from the chHES137 and 175 lines; miRNAs and snoRNAs expressed in DLK1-DIO3 cluster are highlighted in red. The expression values were based on normalized read counts (reads per million) and were log 10 transformed. MEG3on and MEG3off indicate that MEG3 is active or silenced (Supporting Information Fig. S4 for the scatter plots of chHES26, 45, 51). (D): Methylation status of the DLK1-DIO3 cluster. The bisulfite sequencing analysis indicates the degree of DNA methylation at IG-DMR and the MEG3 promoter DMRs in three paired hESCs: chHES45, chHES137, and chHES175 (Supporting Information Fig. S5 for the bisulfite sequencing reads in the three paired hESCs). Abbreviations: ehESCs, early-passage human embryonic stem cells; ihESCs, initial-passage human embryonic stem cells.
Download figure to PowerPoint
The human DLK1-DIO3 cluster in14q32 also encodes a cluster of microRNAs and snoRNAs, including SNORD114-3 [48, 49] (Fig. 1B), some of which are not covered by the gene expression arrays we used. We performed Solexa deep sequencing for small RNAs in the ihESCs and ehESCs from five of the lines described above to determine whether these additional genes have the same pattern of expression variation as MEG3 and SNORD114-3. In cell lines chHES26, 45, 51, and 137, in which MEG3 and SNORD114-3 were expressed in ihESCs but later silenced in ehESCs, a group of miRNAs and snoRNAs from the locus were significantly downregulated in ehESCs (Fig. 1C and Supporting Information Fig. S4, Tables S3, S4). In contrast, line chHES175, in which the expression of MEG3 and SNORD114-3 was unchanged, showed no significant change in expressions of miRNAs and snoRNAs between ihESCs and ehESCs (Fig. 1C, Supporting Information Tables S3, S4). Thus, the expression pattern of miRNAs and snoRNAs in the cluster correlated with that of MEG3 and SNOD114-3 from ihESCs to ehESCs, suggesting that genes in the cluster are subjected to a common expression regulatory mechanism.
Imprinting of the DLK1-DIO3 cluster is known to be regulated by a differentially methylated intergenic region (IG-DMR) and by DMRs in the MEG3 promoter region  (Fig. 1B). To further determine whether silencing of the DLK1-DIO3 cluster in ehESCs correlates with increased DNA methylation of these DMRs, we examined the DNA methylation status in both ihESCs and ehESCs using bisulfite sequencing. In the chHES45 and chHES137 lines that underwent MEG3 and SNORD114-3 silencing in ehESCs, the DMRs in the ihESCs showed partial (∼50%–70%) methylation, while in ehESCs they showed nearly complete methylation (>95%; Fig. 1D and Supporting Information Fig. S5). However, in line chHES175, in which the expression of MEG3 and SNORD114-3 was persistent, methylation of all DMRs did not change from ihESCs to ehESCs (both at ∼50%; Fig. 1D and Supporting Information Fig. S5). In order to evaluate the imprinting pattern of the DLK1-DIO3 cluster, we performed allelic expression analysis in two single nucleotide polymorphism (SNP) sites of MEG3. We found that the two SNP sites were expressed in a monoallelic pattern in ihESCs of chHES45 and chHES137 (Supporting Information Fig. S6), but totally silenced in their ehESCs. The above results suggest that the DLK1-DIO3 cluster was normally imprinted in ihESCs but became completely silenced during early culture.
To further test whether aberrant silencing of the DLK1-DIO3 cluster is common in other cell lines, we used real-time PCR to examine the expression of MEG3 and SNORD114-3 in an additional 20 hESC lines from our cell bank. The results showed that MEG3 and SNORD114-3 were expressed at relatively high levels in ihESCs of all 20 cell lines, but they became completely silenced or exhibited more than twofold downregulation in ehESCs of 16 lines (Fig. 2A, 2B). We then retrieved gene expression datasets of the most commonly used hESC lines and several induced pluripotent stem cell (iPSC) lines from different laboratories around the world via the GEO database. We obtained 64 datasets generated using HG-U133 plus2 microarrays, and then compared the expression pattern of genes to our representative ihESCs and ehESCs of chHES26 (Fig. 2C, Supporting Information Table S5). In these datasets, which are mostly for human pluripotent cells after 20 passages, the pluripotency marker genes POU5F1 and NANOG exhibited a similar high level of expression across all samples. However, the genes in the DLK1-DIO3 cluster, except for DLK1, were only expressed in two initial-passaged iPSC samples (P6) and four hESC lines at a level similar to that of P6 ihESCs of chHES26, while the other 58 cell lines show no expression similar to the P24 ehESCs of chHES26. Similarly, in another gene expression dataset from 32 cell lines between P17 and P65, the expression of MEG3 was only detected in five cell lines between P17 and P23 (Supporting Information Fig. S7 and Table S5). These results suggested that most hESC and hiPSC lines used in laboratories worldwide have undergone DLK1-DIO3 silencing during early in vitro culture.
Figure 2. Gene expression pattern of DLK1-DIO3 cluster in other hESCs and human induced pluripotent stem cells (hiPSCs). (A, B): Expression of MEG3 (A) and SNORD114-3 (B) in 20 paired ihESC (black) and ehESC (gray) lines by RT-PCR. Asterisk (*) indicates derived under 5% oxygen. Error bars indicate the calculated SE of the mean for the replicate values. (C): Heat map showing the relative expression levels of selected transcripts in hESCs and hiPSCs. Colors indicate low (3green) to high (11red) absolute log 2 expression levels. DLO indicates derived under 5% oxygen. hESC samples are marked in red and hiPSC samples are marked in black. See Supporting Information Table S5 for the list of samples, GEO numbers, PubMed IDs, and abbreviations. Abbreviations: ehESCs, early-passage human embryonic stem cells; ihESCs, initial-passage human embryonic stem cells.
Download figure to PowerPoint
We further examined the expression status of MEG3 in the ICM of human blastocysts to identify the expression status of the DLK1-DIO3 cluster in the original embryonic cells in vivo. Our result showed that the ihESCs had comparable MEG3 expression levels to the ICM, indicating the activation of the DLK1-DIO3 cluster in ihESCs is a normal status inherited from embryonic cells and that the silencing of MEG3 expression in ehESCs represents an abnormal variation acquired through early culture (Supporting Information Fig. S8).
Silencing of DLK1-DIO3 Is Irreversible
Normal expression of the Dlk1-Dio3 gene cluster has been related to the establishment of full pluripotency in mouse iPSCs [51, 52]. We compared the differentiation capability of ihESCs and ehESCs from three lines with an initial-on-late-off MEG3 expression pattern: chHES55, 60, and 137 to determine whether aberrant silencing of the DLK1-DIO3 cluster in ehESCs affects the pluripotency of hESCs. All three lines were able to differentiate into the three germ layers both in vitro and in vivo (Supporting Information Fig. S9). This is consistent with most published studies, which report that late-culture hESCs maintain their multilineage differentiation ability .
To assess whether the silencing of the DLK1-DIO3 cluster in ehESCs could be reversed during differentiation, we analyzed the expression level of MEG3 and SNORD114-3 in differentiated progenies from ihESCs and ehESCs. Compared to hESCs, the differentiated cells in day 15 embryonic bodies (EB day15) exhibited marked changes in cellular morphology and downregulation of NANOG, but showed similar expression patterns of MEG3 and SNORD114-3, with a statistically significant linear correlation (r = 0.921 for MEG3, p < .001; r = 0.789 for SNORD114-3, p < .001) (Fig. 3A–3D and Supporting Information Fig. S10). Upon directed differentiation toward the neural lineage, we observed the same inheritance tendency in differentiated ß-tubulin-positive neural cells (Supporting Information Fig. S11A, S11B). Taken together, our results indicate that differentiation does not reverse the silencing of the DLK1-DIO3 cluster.
Figure 3. The silencing of the expression pattern of DLK1-DIO3 during differentiation is irreversible. (A): Morphology of human embryonic stem cells (hESCs) and EBs, scale bars = 100 µm. (B): RT-PCR analysis of expression analysis of pluripotency marker NANOG expression in undifferentiated hESCs (ES) and EB day-15 derived from initial-passage hESCs and early-passage hESCs. (C, D): RT-PCR analysis of MEG3 (C), SNORD114-3 (D) expression in undifferentiated hESCs (ES), and EB day15 derived from ihESCs and ehESCs. Abbreviations: EB, embryoid body; ES, embryonic stem.
Download figure to PowerPoint
Early hESCs and Their Liver-Like Cell Derivatives Exhibit Attenuated Apoptosis Induced by DNA Damage
Downregulation of MEG3 in the DLK1-DIO3 cluster was previously shown to participate in hepatocarcinoma tumorigenesis, possibly by suppressing p53 to exert an antiapoptotic effect . To evaluate whether this effect occurs in ehESCs or their hepatocyte derivatives, we compared the apoptosis rate in ihESCs and ehESCs from the chHES56, 90, and 12 lines, treated with or without mitomycin C (MC). After MC-induced DNA damage, both ihESCs and ehESCs exhibited one- to two fold higher rates of apoptosis, as indicated by fluorescence-activated cell sorting analysis. However, in the ehESCs that were negative for MEG3 expression (MEG3off) from the chHES56 and chHES90 lines, the rates of MC-induced apoptosis were 20% lower than the rates observed in their ihESC counterparts with MEG3 expression (MEG3on), with a statistical significance (Fig. 4A and Supporting Information Table S6). By contrast, in the MEG3on ehESCs of the chHES12 line that was positive for MEG3 expression, the rate of MC-induced apoptosis was comparable to their MEG3on ihESC counterparts (Fig. 4A).
Figure 4. Analysis of apoptosis and p53 activity in human embryonic stem cells (hESCs) and HPLCS after DNA damage. (A, B): Apoptosis rates in hESCs (A), ihESCs, and ehESCs and in HPLCs (B) derived from initial-passage hESCs (ihESCs) and early-passage hESCs (ehESCs) with (MC) or without (control) MC treatment for 6 hours. Data are represented as mean ± SD (n = 3) using the percentage of apoptotic cells (FITC+/PI- and FITC+/PI+). * indicates p < .05; ** indicates p < .01 with respective control by t test. (C, D): MEG3 and TP53 mRNA expression in hESCs (C), ihESCs, and ehESCs or HPLCs (D) derived from ihESCs and ehESCs with (MC) or without (control) MC treatment for 6 hours. MEG3 and TP53 expressions were normalized to the initial control. Data are represented as mean ± SD (n = 3). *, p < .05; **, p < .01 with respect to the control (TP53) expression. The results of the independent sample t test and correlation analysis are shown in Supporting Information Table S6 and Figure S12. (E, F): Western blot analysis of p53 and phospho-p53 expression in hESCs (E), ihESCs, and ehESCs or HPLCs (F) derived from ihESCs and ehESCs with (+) or without (−) MC treatment for 6 hours; β-actin was used as an internal control. Abbreviations: HPLC, hepatocyte-like cell; MC, mitomycin C.
Download figure to PowerPoint
We further examined the effect of MEG3 silencing by differentiating both ihESCs and ehESCs from the chHES56, 90, and 12 lines into HPLCs. After a three-step induction, the cells exhibited the typical hepatocyte morphology and were positive for both the hepatocyte progenitor marker alpha fetal protein and hepatocyte nuclear factor 4 alpha (Supporting Information Fig. S1A, S1B). Similar to their original hESCs, after MC-induced DNA damage, HPLCs derived from MEG3off ehESCs exhibit significantly lower rates of apoptosis than HPLCs derived from MEG3on ihESCs (Fig. 4B and Supporting Information Table S6). Taken together, this data suggest that MEG3 silencing attenuates DNA damage-induced apoptosis in hESCs and their differentiated cells.
We next examined whether the attenuated apoptosis in MEG3off ehESCs and their differentiated derivatives were associated with suppression of the p53 pathway, as previously reported [47, 55]. Our data showed that TP53 mRNA levels were decreased by up to 30% in both control and MC-treated ehESCs from the chHES90 and chHES56 lines, which had up to 70% downregulation of MEG3 compared to the corresponding ihESCs (Fig. 4C). However, upregulation of TP53 mRNA was observed in MEG3on ehESCs from the chHES12 line, compared to the level in its ihESC. Increased p53 and phosphor-p53 protein levels were observed in all hESCs after MC treatment; however, ehESCs from MEG3off cell lines exhibited reduced phosphor-p53 levels (Fig. 4C, 4E). Likewise, HPLCs from MEG3off ehESCs inherited compromised p53 reactivity to MC-induced DNA damage in terms of mRNA and protein levels compared to its ihESC-derived counterpart (Fig. 4D, 4F). Taken together, our data suggest a positive correlation between the expression level of MEG3 and TP53 (Supporting Information Fig. S12 and Table S7), which indicates that the antiapoptotic effects observed in MEG3off ehESCs and the differentiated HPLCs may be caused, at least in part, by suppression of the canonical p53 pathway.
Oxygen at 5% Is Required to Maintain Persistent Expression of the DLK1-DIO3 Cluster
In searching for the causes for aberrant silencing of the DLK1-DIO3 cluster, we noticed that, among the 32 lines analyzed, six cell lines with persistent MEG3 expression were derived under 5% oxygen before shifting to atmospheric oxygen conditions for prolonged culture (Figs. 1A, 2A). From the GEO database, the lines WIBR1, 2, and 3, which were derived under 5% oxygen, also expressed high levels of MEG3 and SNORD114-3 (Fig. 2C) . This led us to speculate that the oxygen concentration might be a critical factor in maintaining normal expression of the DLK1-DIO3 cluster during culture, especially during the initial derivation and expansion stages. To test this hypothesis, we derived five new hESC lines from human blastocysts under 5% oxygen conditions, and divided them at P5 for subsequent culture either in 5% or 20% oxygen until P20 (Supporting Information Fig. S13). As expected, MEG3 and SNORD114-3 were stably expressed in culture under continuous 5% oxygen in all five lines, including those that were cultured to P20 under the same conditions. However, those lines that were transferred to 20% oxygen at P5 and cultured to P20 showed 1.5–3.2-fold downregulation of MEG3 and SNORD114-3 (Fig. 5A). Furthermore, we analyzed the methylation level in ihESCs and ehESCs (including samples under 5% oxygen and 20% oxygen). We found that the level of DNA methylation was approximately 50%–60% for both IG-DMR and MEG3 promoter DMRs in ihESCs and ehESCs under 5% oxygen, which is similar to the DNA methylation level of normal imprinting; however, the ehESCs under 20% oxygen showed significantly higher methylation level, approximately 70%, consistent with the downregulation of MEG3 and SNORD114-3 (Fig. 5B and Supporting Information Fig. S14). To exclude the possibility that loss of imprinting may also occur during early culture, we analyzed the allelic expression of MEG3 in ihESCs and ehESCs cultured under both 5% and 20% oxygen. The results indicated that MEG3 was expressed monoallelically in ihESCs and ehESCs (including samples under 5% oxygen and 20% oxygen) (Supporting Information Fig. S15). This data indicated that low oxygen levels during initial culture are important to maintain DLK1-DIO3 cluster expression during late culture, even when the culture is shifted to atmospheric oxygen conditions. To further test whether late exposure of ihESCs derived under 20% oxygen to 5% oxygen could recover the expression of the DLK1-DIO3 cluster, we transferred P5 ihESCs from four lines derived under 20% oxygen to 5% oxygen, and cultured the cells until P20. Out data showed that late exposure to 5% oxygen could not recover the expression of MEG3 and SNORD114-3 (Fig. 5C). Taken together, our results indicate that atmospheric oxygen is the main cause for the silencing of the DLK1-DIO3 cluster during early culture and further that an initial low oxygen concentration (5%) is essential for maintaining the normal expression of the DLK1-DIO3 cluster (Fig. 5D).
Figure 5. Effects of oxygen concentration on expression of the DLK1-DIO3 cluster in hESCs. (A): RT-PCR expression analysis of MEG3 (left) and SNORD114-3 (right) in P5 ihESCs under 5% oxygen (ihESCs-5%), and P20 ehESCs cultured under 5% (ehESCs-5%) or 20% oxygen (ehESCs-20%). (B): The DNA methylation level of the IG-DMR, the MEG3 promoter regions (RI, RIII), and the total methylation level combined all three regions (Total) in chHES268 and chHES283 with different passages and oxygen concentrations. The dotted line indicates 50% methylation. ** indicates p < .01 between ihESCS-5% or ehESCs-5% and ehESCs-20%. Supporting Information Figure S14 for the methylation details of individual bisulfite sequencing reads from these two cell lines. (C): Relative expression levels of MEG3 (top) and SNORD114-3 (bottom) in ihESCs cultured under 20% oxygen (ihESCs-20%) and ehESCs after transfer to 5% oxygen (ehESCs-5%). (D): A schematic summary of oxygen conditions on the expression of MEG3 through different stages of hESCs culture. Abbreviations: ehESCs, early-passage human embryonic stem cells; ihESCs, initial-passage human embryonic stem cells.
Download figure to PowerPoint