Linking Incomplete Reprogramming to the Improved Pluripotency of Murine Embryonal Carcinoma Cell-Derived Pluripotent Stem Cells

Somatic cell nuclear transfer (SCNT) has been proved capable of reprogramming various differentiated somatic cells into pluripotent stem cells. Recently, induced pluripotent stem cells (iPS) have been successfully derived from mouse and human somatic cells by the over-expression of a combination of transcription factors. However, the molecular mechanisms underlying the reprogramming mediated by either the SCNT or iPS approach are poorly understood. Increasing evidence indicates that many tumor pathways play roles in the derivation of iPS cells. Embryonal carcinoma (EC) cells have the characteristics of both stem cells and cancer cells and thus they might be the better candidates for elucidating the details of the reprogramming process. Although previous studies indicate that EC cells cannot be reprogrammed into real pluripotent stem cells, the reasons for this remain unclear. Here, nuclei from mouse EC cells (P19) were transplanted into enucleated oocytes and pluripotent stem cells (P19 NTES cells) were subsequently established. Interestingly, P19 NTES cells prolonged the development of tetraploid aggregated embryos compared to EC cells alone. More importantly, we found that the expression recovery of the imprinted H19 gene was dependent on the methylation state in the differential methylation region (DMR). The induction of Nanog expression, however, was independent of the promoter region DNA methylation state in P19 NTES cells. A whole-genome transcriptome analysis further demonstrated that P19 NTES cells were indeed the intermediates between P19 cells and ES cells and many interesting genes were uncovered that may be responsible for the failed reprogramming of P19 cells. To our knowledge, for the first time, we linked incomplete reprogramming to the improved pluripotency of EC cell-derived pluripotent stem cells. The candidate genes we discovered may be useful not only for understanding the mechanisms of reprogramming, but also for deciphering the transition between tumorigenesis and pluripotency.


Introduction
Various differentiated somatic cells can be reprogrammed into a totipotent, or at least pluripotent state by somatic cell nuclear transfer (SCNT), which includes fetus-derived epithelial cell lines [1], cumulus cells [2], mature B and T lymphocytes [3], olfactory sensory neurons [4,5] and natural killer T cells [6]. This reprogramming process requires the reversal of epigenetic modifications, such as DNA methylation, histone modifications, and the condensation state of chromatin [7]. Recently, induced pluripotent stem (iPS) cells were generated by the forced expression of four transcription factors in mouse fibroblasts, and the derived iPS cells are similar to ES cells [8]. However, the detailed mechanisms underlying these complicated reprogramming events are not well understood.
Epigenetic modifications play important roles during the development of embryos and the initiation of disease. The definition of cell fate always coincides with changes in its epigenetic modifications, such as DNA methylation and histone modifications. Aberrant epigenetic modifications could result in many kinds of diseases, such as cancer [9]. For instance, it is well known that the promoter regions of many important tumor suppressor-genes are always hypermethylated, which inactivates the supervisory roles of tumor suppressor genes, thus resulting in the neoplasia [10]. During the reprogramming process mediated by SCNT, genetic alternations cannot be corrected, whereas the epigenetic modifications can indeed be reset.
Embryonal carcinoma (EC) cells, derived from teratocarcinomas, are capable of unlimited self-renewal and can differentiate into several kinds of somatic cells. The isolation of EC cells also provides us with a preliminary framework for embryonic stem cells [11]. Thus, EC cells are widely used as the in vitro models for dissecting several fundamental questions related to development and pluripotency [12]. Moreover, the discovery of EC cells demonstrated the existence of the so-called ''cancer stem cells'' for the first time, predating the current interest in these by several decades. Although EC cells still have similar characteristics to other cancer cells, such as the genetic mutations, they differ from other kinds of cancer cells in the developmental potential, as assessed by the blastocyst injection assay [13].
Previous studies have indicated that some tumor cells are able to direct the development of early cloned embryos, producing morphologically normal blastocysts that give rise to NTES cell lines, but the cloned embryos are not able to develop into live pups after their transfer into the uterus [14,15,16]. The failed reprogramming of tumor cells may be due to characteristics of the donor cells, such as the profound genetic changes or the differentiation states of these cells. Accumulated evidence implies that cancer cells may have an inseparable connection with induced pluripotent cells [8,17,18]. Considering the special status of EC cells, which have the dual identities of both cancer cells and multipotent cells, we have designed our experiments relying on EC cells. The reprogramming of EC cells may provide an excellent model for understanding the maintenance of tumorigenic potential and pluripotency.
In the present study, we attempted to answer two questions by reprogramming P19 EC cells through SCNT: (I) Can EC cells be reprogrammed and become pluripotent, and to what extent can they be reprogrammed? (II) What molecular events occur during the reprogramming of EC cells?

The development of cloned embryos reconstructed with P19 cells and the establishment of P19 cell-derived pluripotent stem cells
In this study, two tumor cell lines (N2a and P19) with different development potentials were chosen as the donors for the nuclear transplantation. Compared to the N2a cell line, the P19 cell line was highly tumorigenic and in a low differentiation state, characteristics that were verified by the subcutaneous injection into immunodeficient mice (data not shown). P19 cells also always adhered to the culture dish and showed a typical morphology of cancer cells when cultured in vitro ( Figure 1A).
Nuclei from the two tumor cell lines were injected into enucleated B6D2F1 oocytes and the reconstructed oocytes were activated and cultured to evaluate their in vitro developmental capacity. We were unable to obtain cloned blastocysts from 53 reconstructed oocytes containing the nuclei of N2a cells, but we succeeded in producing 171 cloned blastocysts from 628 oocytes after the transfer of nuclei from P19 cells. The development of P19 cell-derived cloned embryos resembled the embryos derived from cumulus cells, which always acted as the control for the routine nuclear transfer assay ( Figure 1B, C, D, E). The proportion of cleaved oocytes that reached the blastocyst stage was nearly 30% (Table 1).
Because blastocysts reconstructed with P19 cells had a higher developmental rate, we were very interested in whether these cloned blastocysts could maintain the full-term development of embryos. Therefore, cloned blastocysts derived from P19 cells were transferred back into the uteri of pseudo-pregnant mice, but we failed to get full-term pups. Then we tried to obtain cloned pups through the ''two-step'' cloning method, which was efficient in overcoming the underlying handicaps that resulted from the direct cloning [3,4]. First, seven NTES cell lines were established from 21 cloned blastocysts derived from P19 cells with a rate of about 33.3%, which was comparable to the rate of cell lines derived from fertilized embryos (data not shown). These NTES cells from the P19 cells all had the typical morphology of normal ES cells and could be passaged in vitro long-term ( Figure 1F, G, H). As tumor cells always had karyotype problems, we eliminated the impact resulting from abnormal karyotypes by selecting normal ES cell lines. As seen from Figure 1I, the karyotypes of the P19 cells were relatively normal, 78.7% of which contained 40 chromosomes. Among the P19 NTES cells, however, only P19 NTES7 cells showed relatively normal karyotypes (73.3%) and this line was thus selected for subsequent experiments ( Figure 1J, K).
2. Pluripotent P19 NTES cells dramatically prolong the development of tetraploid aggregated embryos P19 NTES cells maintained the typical morphology of normal murine ES cells when cultured in vitro long-term. After this period of culturing, we performed teratoma formation and tetraploid blastocyst aggregation assays to verify their pluripotency. After being subcutaneously injected into SCID mice for three to four weeks, P19 NTES cells could form teratomas with a high frequency. The histology of the resulting tumors from P19 NTES cells exhibited differentiation into neural and glial cells (Figure 2A), gland and column-like epithelia ( Figure 2B), and skeletal muscle ( Figure 2C). Moreover, the expression of pluripotent genes, such as Zfp42, Nanog, Sox2 and Pou5f1, were also observed in P19 NTES cells, although P19 cells also had a high background expression of these pluripotent markers ( Figure 2D).
The EC cells joined in the formation of somatic tissues after they were injected into blastocysts [13]. Therefore, the detection of three germ layers in the teratoma tissues could just reflect the in vivo differentiation potential of P19 NTES cells, not their real pluripotency. It was also unknown whether pluripotent stem cells  derived from the EC cells could eliminate their tumorigenic potential and maintain the full-term development of the tetraploid chimera, which was the stringent standard for pluripotency. Therefore, tetraploid blastocyst aggregation assays were performed with P19 cells and P19 NTES cells separately. Among the 121 blastocysts injected with P19 cells, only four placentas under absorption were observed at embryonic day (E) 8.5, which was consistent with the results of previous studies, and no live fetuses could be isolated from the absorbing tissues after C-section ( Figure 2E and Table 2). In contrast to P19 cells, P19 NTES7 could prolong the development of tetraploid aggregated embryos dramatically. As shown in Figure 2F and Table 2, tetraploid aggregated embryos derived from P19 NTES7 always had bigger placentas than those of P19 cells and the implantation rate was very high, which was a direct indication of increased pluripotency. Although most of these implanted tetraploid complementary embryos derived from P19 NTES7 died before E8.5, we were able to obtain six live E8.5 fetuses of P19 NTES7 from the pseudopregnant mice via C-section ( Table 2). In contrast to P19 NTES7 cells, P19 NTES1 cells failed to maintain the early development of tetraploid aggregated embryos beyond E8.5 ( Figure 2G and Table 2). Therefore, although P19 NTES7 also failed to maintain the full-term development of tetraploid aggregated embryos, it did have an increased pluripotency compared to its progenitor P19 cells.
These data indicated that P19 NTES cells had an increased pluripotency compared to their progenitor cells and the improved pluripotency could extend the development of tetraploid aggregated embryos dramatically. However, not every P19 NTES cell line had the similar improved pluripotency of P19 NTES7, with P19 NTES1 as the typical example.

The results of molecular analyses indicate that P19 NTES cells are intermediate between EC cells and ES cells
Because P19 NTES cells showed an improved pluripotency as compared with progenitor cells, we hypothesized that undiscovered molecular events lead to this cell fate transition. We thus decided to characterize this phenomenon by thoroughly analyzing the molecular characteristics of P19 NTES cells. We used realtime PCR and microarrays to generate a global view of the transcriptional change during the reprogramming process of P19 cells. First, we analyzed the expression of many cancer or pluripotency-related genes in P19 NTES7 cells with several specific cell lines as controls. These cell lines included the normal ES cells (R1), which are the gold standard of true pluripotency, NTES cells derived from Sertoli cells (S16), iPS cells generated from fibroblasts (iPS) and mouse embryonic fibroblasts (MEF). As shown in Figure 3A-F, we classified these genes into three groups. The first group included the genes that could be induced during the reprogramming process of P19 cells by NT, with Nanog as the representative gene. The second group comprised the genes that could not be induced to the expression level of normal ES cells, with Trp53 as the example. The third group contained the genes with no expression change during this process, such as Pou5f1 and Sox2. Thus, we had direct evidence supporting our hypothesis that P19 NTES7 had a unique molecular code, which was worthy of a deeper analysis.
To test this hypothesis, we used microarray for analyzing the transcriptome of P19 NTES cells to help us understand the improved pluripotency detected (Accession number: GSE18691). The microarray results were first displayed as scatter plots, in which the differentially expressed probe sets were marked by red and green dots, while the unchanged probe sets were marked by blue dots. As shown in Figure 4A, P19 cells had a distinct expression pattern compared to R1 ES cells, and these differentially expressed probe sets represented the difference between EC cells and ES cells. In contrast to the distinct expression pattern in the P19 versus R1 group, the P19 NTES7 versus P19 group and the P19 NTES7 versus R1 group had fewer differentially expressed probe sets. In order to reduce the noise in our analysis, we increased the threshold level (fold change.2, oneway analysis of variance (ANOVA) p-value,0.01). Among the probe sets covered in the GeneChipH Gene 1.0 ST Array System for the mouse genome, we found 1641 differentially expressed probe sets between P19 cells and R1 representing 1514 genes (Dataset S1). The sorted genes included several important genes, such as BMP4, Fbxo15, Dppa5a, Esrrb, Trp53, Nanog, Klf4, H19, Tbx3 and c-Myc. Nearly all of these genes met the more stringent selection standard of significance (fold change.5, one-way ANOVA p-value,0.001), with c-Myc as an exception for its special role in both tumorigenesis and reprogramming. Interestingly, these genes always related to pluripotency (BMP4, Fbxo15, Dppa5a, Nanog) and tumorigenesis (Trp53, Klf4, H19, c-Myc). The expression patterns of these genes were verified by real-time PCR with P19 cells as the reference. As shown in Figure 4B and Dataset S2, the results of the quantitative PCR corresponded well to the microarray results (Max fold change.2.5, p-value,0.01) ( Table 3). Among these genes, Nanog and H19 in P19 NTES7 were the outstanding ones because they were successfully induced to the normal expression level, while Fbxo15, Dppa5a, Esrrb, Klf4 and c-Myc were only induced to an intermediate level. In striking contrast, BMP4, Trp53 and Tbx3 still maintained the expression patterns of EC cells, without any induction trend. As we know, these genes are all involved in critical signaling pathways, and the dysfunction of any of these genes will lead to tumorigeneis. Altogether, these results suggested that P19 NTES7 cells were really the intermediate between P19 cells and R1 cells and this intermediate had the identities of partial pluripotent stem cells, not only at the level of development potential, but also at the molecular level. Taking this into consideration, we decided to further explore the hierarchy among them.
Next, an unsupervised hierarchical clustering analysis was performed to assess the similarities in single gene expression profiles in an unbiased way among the P19, P19 NTES7 and R1 cells and all the differentially expressed genes were covered ( Figure 4C). In the resulting sample tree, P19 cells and R1 cells A principal components analysis (PCA), an exploratory multivariate statistical technique always used to simplify complex data sets, was used to evaluate the microarray data. Using this approach, similar samples should be close in distance [19]. As seen from Figure 4D, the three replicates of each cell line were all the closest in the PCA and the separation among the three kinds of cell lines was clear. Therefore, the PCA analysis indicated that the repeatability of microarray data was reliable and the main variable component came from the origin of every cell line.
In summary, these experiments demonstrated that P19 NTES cells had unique molecular characteristics and they were indeed the intermediates between EC cells and ES cells. The gain of function of many pivotal genes in P19 NTES cells coming from partial reprogramming might account for the extended develop-ment of tetraploid aggregated embryos observed in the earlier experiments.

Gene Ontology and pathways enrichment of differentially expressed genes among P19 cells, P19 NTES cells and normal ES cells
In order to understand functions of differentially expressed genes among the P19, P19 NTES7 and R1 cells, we pursued an alternative comparative approach and a gene ontology (GO) analysis was used to identify categories within which genes differed significantly in these cells. First, we screened the differentially expressed genes from P19 versus R1 and this set of genes was named as GeneSet A or A (Dataset S1) (fold change .2, one-way ANOVA p-value ,0.01). Then we further sub-classified GeneSet A into two sub-groups. The first was named as GeneSet B or B (Dataset S3), and genes in this set met the significance standard for GeneSet A and the one for P19 versus P19 NTES7 (fold change . 2, one-way ANOVA p-value ,0.01). Genes in GeneSet C or C (Dataset S4) met the significance standard for GeneSet A and the one for P19 NTES7 versus R1 (fold change .2, one-way ANOVA p-value ,0.01). Under this stringent classification standard, GeneSet A was regarded as the molecular differences between EC cells and normal ES cells, GeneSet B represented the genes that were successfully reprogrammed during EC cells reprogramming, and GeneSet C represented the genes which have not been successfully reprogrammed.
The reliability of this classification was confirmed by the overlapping results in Figure 4E and Dataset S5. Among the 1514 genes included in GeneSet A, GeneSet B contained 306 genes and GeneSet C contained 635 genes. GeneSet B and GeneSet C were also well separated with only 42 overlapping genes. Therefore, this sub-classification standard could provide us with more reliable information for further analyzing the functions of these genes.
Next, GO analysis was used to identify the functional categories overrepresented in the cluster genes from GeneSet B and GeneSet C. As shown in Figure 5A-C, the functional categories identified from the two groups of genes were very similar and the enriched categories included several biological processes, such as the developmental process, metabolic process, transcription, transport and cell differentiation. Intriguingly, these biological processes corresponded to several molecular functions, for instance, protein binding, catalytic activity and ion binding. When we reduced the scope by comparing the functional categories covered in GeneSet B and C, we found that the protein binding activity was the only different one  A pathways enrichment analysis was then performed to refine the differentially expressed genes into specific signaling pathways during the reprogramming process. As shown in Dataset S9, GeneSet A covered six conventional signaling pathways, including the MAPK, focal adhesion, P53, ECM-receptor interaction, tight junction and VEGF signaling pathways, and these pathways needed to be recovered when the EC cells were reprogrammed into pluripotent stem cells (p-value ,0.05). Regarding GeneSet B and C, we found the pathways covered in GeneSet B were reduced dramatically and only contained the MAPK, ECM-receptor interaction and focal adhesion signaling pathways. Nevertheless, GeneSet C could map to the tight junction, focal adhesion, VEGF and p53 signaling pathways, which was in contrast to GeneSet B (Dataset S10, Dataset S11) (p-value ,0.05). The recovered genes mainly fell into the MAPK, ECM-receptor interaction and focal adhesion signaling pathways. For instance, c-Myc, the downstream member of the MAPK pathway, was induced partially while the focal adhesion signal pathway could only be repaired partly. Most importance of all, the P53 pathway failed to be recovered, with many critical genes inactive, such as Trp53, Gadd45 and Cdkn1a ( Figure S1, Figure  S2). The role of P53 was complex for its non-protecting function in germ cell tumors and negative expression in germ cells and sertoli cells [20,21]. Therefore, P19-derived pluripotent stem cells still bear the similar attributes of cancer cells and a P53 independent mechanism may exist for protection. In brief, the pathways enrichment results suggested that the reprogramming of EC cells through NT could repair the MAPK, ECM-receptor interaction and focal adhesion signaling pathways, at least partially, but not the P53 pathway.

The expression recovery of H19 is dependent on methylation in the differential methylation region (DMR), while the induction of Nanog is independent of the promoter region methylation
As showed in the data thus far, P19 NTES cells resulted from an incomplete reprogramming, with many genes fully induced, such as H19 and Nanog, and many genes partially induced, such as Fbxo15, Dppa5a and Esrrb. However, the mechanisms underlying the induction of these genes were unclear, especially for the representatives of the fully induced genes, H19 and Nanog. Combined with the GO results, which indicated that GeneSet B always took part in the protein binding, we reasoned that specific epigenetic transformations may have been involved in the reprogramming of the EC cells.
In order to test this hypothesis, we analyzed the DNA methylation changes during the induction of H19 and Nanog using the bisulfite genomic sequencing method. Moreover, western blot was also performed to analyze the expression status of Nanog during the process of reprogramming, but not H19 due to its protein non-coding property [22]. As shown in Table 3 and Figure 6A, the expression of H19 could be recovered to a normal level. The expression of the other imprinted gene, Peg3, had no change, which has been verified by the microarray and real-time PCR ( Figure 6B). Next, the DMRs of H19 and Peg3 were analyzed for their DNA methylation patterns. Interestingly, the methylation in the DMR of H19 in P19 cells was very high (135/150), while the levels in P19 NTES7 and R1 were within the normal range (55/ 150, 71/135) ( Figure 6C, D, E, F). In contrast to H19, there was no change of the DNA methylation profiles of Peg3 among these cell lines and it acted as a better control for H19 ( Figure 6G, H, I, J). The observed hypermethylation of P19 cells corresponded to the repression of H19 and the induction of H19 was dependent on the methylation in the DMR.
Similar to H19, the mRNA level of Nanog was also induced to a relatively normal level in P19 NTES cells (Table 3 and Figure 3A). Western blot results further confirmed the microarray and realtime PCR data ( Figure 7B). We also analyzed the DNA methylation profiles in the promoter region of Nanog, with Pou5f1 as the control ( Figure 7A, G). As shown in Figure 7C, D, E, F, the methylation of Nanog in P19, P19 NTES7 and R1 were all very low (9/132; 7/132; 7/132) while the methylation level in MEF was very high (90/132). Nanog was one of the important factors in the generation of iPS cells and it also played major roles in maintaining pluripotency. The regulation of Nanog always relied on the methylation state in the promoter region, with hypomethylation resulting in activation and hypermethylation resulting in silence. Therefore, our results provided direct evidence for the existence of a specific regulatory circuit in controlling the expression of Nanog in EC cells and the recovery of Nanog was the remarkable indicator of the improved pluripotency in P19 NTES cells. When we analyzed the methylation patterns of Pou5f1, we found that the unchanged expression of Pou5f1 correlated with the ubiquitous hypomethylation of the promoter region in P19, P19 NTES7 and R1, with MEF as the negative control (Table 3 and Figure 7I, G, K, L). The DNA sequencing results were further confirmed by restriction enzyme digestion assays ( Figure 7H).
In summary, the induction of H19 during the partial reprogramming of EC cells was in a DMR methylation-dependent pattern while the induction of Nanog was independent of the promoter region methylation.

Discussion
In this study, we addressed two questions that remain regarding the reprogramming of embryonal carcinoma cells by using a ''two-  step'' cloning approach. Our results demonstrate that P19 NTES cells have an improved pluripotency compared to their progenitor cells, as we observed in the tetraploid aggregated embryos assays. The gain of function of several pluripotent genes identified by high-throughput methods may account for this phenomenon. However, the reprogramming of EC cells is indeed incomplete and the partial reprogramming of P19 cells mainly results from the characteristics of the donor nuclei, such as the specific epigenetic modification patterns, with no direct relationship to the reprogramming process itself. Strikingly, the induction of important genes in EC cells correlates with different regulatory mechanisms, with H19 in a DMR methylation-dependent pattern and Nanog in a promoter region methylation-independent pattern. Therefore, the successful reprogramming of EC cells is still challenging and further investigation of the regulatory circuits involved in the reprogramming of EC cells will help us to ultimately conquer this fundamental question.
Previous attempts at reprogramming tumor cells proved that it was difficult to reprogram tumor cells or get stable ES cell lines from cloned blastocysts [14,15,16]. Therefore, tumor cells might have different attributes that result in the failed reprogramming. It was reported that EC cells carry many genetic mutations and these unrecoverable changes may partially explain the failure in the cloning of EC cells [16]. EC cells have a limited pluripotency, and they may be easy to reprogram, as suggested by the accumulating evidence from ES cell cloning and the derivation of iPS cells [23,24]. We observed that early embryos derived from P19 cells had a similar developmental rate to those generated from cumulus cells and ES cell lines could easily be derived from the cloned blastocysts. Interestingly, one of the derived P19 NTES cell lines (P19 NTES7) had an improved pluripotency and it could prolong the development of tetraploid aggregated embryos. We succeeded in obtaining live E8.5 tetraploid aggregated embryos, which was our strongest evidence for the improved pluripotency. Compared to the previous report [16], we employed here tetraploid complementation assay to evaluate the improved pluripotency of P19 NTES cells because the previous study used chimera analysis through which they could not detect obvious improved pluripotency after NT of P19 cells. The improved pluripotency we observed was verified by the microarray analysis, by which many biomarkers during this transition process were detected. Although the improved pluripotency was not sufficient, it was the basis for the final success in transforming tumor cells into real pluripotent stem cells.
Recently, new types of pluripotent cells, such as iPS cells and germline-derived pluripotent stem (gPS) cells have been derived under specific induction conditions [8,25,26,27,28,29,30]. The emergence of so many pluripotent cells provided us with models not only for investigating the occurrence of many regenerative diseases, but also for investigating the reprogramming mechanisms. Compared to the derivation of iPS and other pluripotent cells, the ''two-step'' cloning that relies on the cytoplasm of the MII oocyte and further in vitro induction could effectively reprogram cells in a short time. Currently, the mechanism of reprogramming is still a mystery and we prefer the hypothesis proposed by Yamanaka [31]. Of note, based on previous work [17], he hypothesized that the transformation resulting from c-Myc and Klf4 was followed by a pluripotent cell induction process regulated by Pou5f1 and Sox2. Recently, the role of Trp53 in the generation of iPS cells strengthened this hypothesis and the emergence of cancer-like cells that act as an intermediate may be indispensable for getting pluripotent iPS cells successfully [18,32,33,34,35]. Similarly, we found that Trp53 and the members of P53 signaling pathways were still inactive in EC cell-derived pluripotent stem cells. In a sense, the repression of the P53 signaling pathway further indicated that EC-derived pluripotent stem cells had different identities from real pluripotent stem cells. Meanwhile, the incomplete or failed induction of c-Myc and Klf4 were also observed, which may due to the inactivity of Trp53. EC cells always had high levels of wildtype P53, which was in a repressed state modulating by an N-terminal repression domain; and the repression state would be activated under the induction of differentiation reagents, such as retinoic acid (RA) [36,37]. As seen from this, EC cell-derived pluripotent stem cells still carried the tumorigenic potential and they could escape apoptosis induced by the Trp53 pathway [38].
The integration of specific signaling pathways into the definitions of the identities of pluripotent stem cells provided insights into the features of their transcriptional regulatory networks [39]. The core of the pluripotency network consisted of Pou5f1, Sox2 and Nanog, while the LIF-STAT3 and BMP4-ID pathways could interact with the core members and the final balanced state among these genes maintained the self-renewal potential of the pluripotent stem cells [39,40,41,42]. Because the expression levels of Pou5f1 and Sox2 in the EC cells were comparable to those of the ES cells, we shifted to study the Nanog, LIF-STAT3 and BMP4-ID pathways. Intriguingly, we found that Nanog and BMP4 were both within the differentially expressed gene set, but not Stat3 or id. Therefore, the core of the pluripotency network was better-maintained in the P19 cellderived pluripotent stem cells, but the external signaling molecules, such as BMP4, were still somewhat aberrant. Nevertheless, P19 NTES cells underwent the pluripotency acquisition process, with Nanog as the typical indication [40]. Nanog has recently been found as the gateway keeper for the ground-state pluripotency and it played an essential role in the pluripotency acquisition during the generation of iPS cells and the conversion of epiblast-derived stem cells (EpiSCs) into ES cells [43]. Moreover, our results indicate that the induction of Nanog in P19 NTES cells was promoter region methylation-independent, which was different from its conventional regulatory mechanism. In contrast to Nanog, H19 was induced in a DMR methylationdependent pattern. H19, a paternally imprinted gene, played an important role in the developing embryo; and the dysfunction of H19 always related to many diseases [44,45]. In sum, these data provide strong evidence that diverse reprogramming mechanisms worked on different genes. The reprogramming of EC cells may be a complicated process and the experiences accumulated in this study may also provide clues for deciphering the reprogramming mechanisms.
There is increasing evidence that tumor tissues are hierarchically organized and sustained by a distinct subpopulation of cancer stem cells (CSCs). CSCs were first isolated from human acute myeloid leukemia [46], which provided an attractive model for investigating tumorigenesis. Similar paradigms suggested in solid tumors were further verified by isolating CSCs from these transformed tissues [47]. Different CSCs had specific growing environments, also called ''niches'', and the properties of CSCs appeared to be influenced by their specific genetic mutations in a given tumor tissue as well as the progression of disease [48]. Therefore, the most challenging problem facing this field is the variability of CSCs coming from their origins. A possible solution to this challenge would be the identification of an in vitro model of CSCs. In this study, pluripotent cells were derived from cancer cells and we named them according to their origins, not their identities. In order to know their real identities, high-throughput method was used to find the differences among cancer cells, cancer cell-derived pluripotent cells and normal ES cells. The final results demonstrated that pluripotent cells derived from cancer cells were intermediates between cancer cells and ES cells. Although these intermediates were not fully pluripotent, they still could maintain the self-renewal ability in vitro and had distinct molecular characteristics. We predict that when these counterparts of cancer cells are used in drug screening, they might provide a new path for obtaining information about killing cancer stem cells.
At the same time, two additional complicated questions remain to be resolved: What is the difference between pluripotent cells from cancer cells and CSCs, and what is the hierarchy between them? It is our hope that our work will pave the way to obtain the counterpart of CSCs in vitro and perhaps it may help us understand the occurrence of tumors and the clinical treatment of cancers more thoroughly.

NT and NTES Cell Derivation
Specific pathogen free (SPF) grade mice were housed in the animal facility of the National Institute of Biological Sciences. Animal care and handling were conducted in accordance with policies promulgated by the Ethics Committee of the National Institute of Biological Sciences.

Karyotype Analysis
Cells were cultured in ES culture medium containing 0.05 mg/ ml demecolcine (Sigma) for 5 h to arrest cells at M-phase. Then cells were trypsinized, pelleted by centrifugation and resuspended in hypotonic solution containing 0.4% NaCl and 0.4% citrate for 5 min. Freshly prepared methanol:acetic acid (3:1) were used to fix the cells at room temperature for 40 min. Finally, the resuspended cells were placed onto pre-cleaned slides for chromosomes counting.

RT-PCR
RT-PCR was performed to investigate the expression of pluripotency related genes. Trizol (Invitrogen, http://www. invitrogen.com/) was used to extract total RNA following the manufacturer's instructions. Then total RNA was treated with DNase before reversely transcribing to cDNA using the MLV reverse transcriptase system (Promega, http://www.promega.com. cn/). Primer sets for Nanog, Pou5f1, Sox2 and Zfp42 were designed as previously described [49].

Western Blot
Protein samples were loaded onto 12% SDS-PAGE gel, resolved and electroblotted onto PVDF membrane. The membrane was blocked in Tris-buffered saline containing 0.1% Tween 20 and 5% low fat dry milk for 1 hour at room temperature. Then, the PVDF membrane was incubated with diluted primary antibodies at 4uC overnight. The peroxidase based detection was performed with Super Signal West Pico Chemiluminescent Substrate Kits (Pierce, http://www.piercenet.com/). Actin was used as the inner control.

Teratoma Formation and Tetraploid Blastocyst Complementation
Teratoma formation was performed to evaluate the in vivo differentiation capacity of NTES cells. Cells distributed in 200 ml PBS (Invitrogen) were subcutaneously injected into the forelimb of SCID-beige mice. Approximately 4 weeks after injection, teratomas were examined histologically using standard protocols. Briefly, teratomas were dissected, weighed and fixed in 4% formaldehyde. The fixed samples were then embedded in paraffin and tissue sections were stained with hematoxylin and eosin for analysis (HE).
For the tetraploid aggregated assays, tetraploid blastocysts were first produced by electrofusion of 2-cell stage embryos of B6D2F1 mice and cultured in KSOM until blastocyst stage [50]. And ES cells or P19 cells were injected into tetraploid blastocysts using a piezo-actuated microinjection pipette. Then the tetraploid aggregated embryos were transfer back to KSOM medium for several hours and transplanted into the uterus of pseudo-pregnant mice.

Real-time PCR
Briefly, Trizol (Invitrogen) was used to extract total RNA. All the manipulations were according to the manufacturer's instructions. Total RNA was treated with DNase (Chemicon) for 30 min, and then an equal amount of RNA was reverse transcribed to cDNA using the MLV reverse transcriptase (Promega). Gapdh was used as the internal control. SYBR green PCR master mix (TaKaRa, http://www.takara-bio.com/) was employed as the detector of fluorescence signal, and ABI 7500 was used in accordance to the manufacturer's instructions. One independent experiment contained three replicates of both targeted genes and inner control. The results of three independent experiments in duplicate were averaged to get the mean value of every gene. Paired Student's t-tests were performed to assess the statistical difference. The significant standard was set as: fold change .2; p-value ,0.05. The primer pairs for real-time PCR were summarized in Table S1.

Bisulfite Genomic Sequencing Analysis of DNA Methylation
TIANamp Micro DNA Kit (Tiangen, http://www.tiangen. com/) was used to extract DNA from 1000 tumor cells or ES cells; and the extracted DNA was treated by EpiTect Bisulfite Kit (QIAGEN, http://www.qiagen.com/) according to the manufacturer's instructions. Bisulfite modification treatments were triplicated for each sample. Then the treated samples were used as the template of nest PCR by using Takara Tag HS (TaKaRa). Primer sets for nest PCR were showed in Table S1. PCR products were recovered using Wizard SV Gel and PCR Clean-Up System (Promega). Then the PCR products were cloned into pMD18-T vector (TaKaRa) and sequenced by using ABI PRISM 3100 Genetic Analyzer (Applied BioSystems, http:// www.appliedbiosystems.com/).

Methylation Sensitive Restrictive Enzyme Digestion Assay of DNA Methylation
Methylation sensitive restrictive enzyme digestion was performed to analyze the methylation profiles of Pou5f1. TaqI (New England Biolabs, http://www.neb.com/) was used in this assay to recognize the two sites at the promoter region of Pou5f1. The digested fragments were electrophoresed on 2.5% agarose gels containing ethidium bromide.

Microarray Data Processing and Statistical Analysis
In brief, we collected P19 cells and feeder free ES cells from 10 cm diameter culture dish. The collected cells were used for RNA extraction in Trizol (Invitrogen) reagent. Per sample contained 3 replicates. GeneChipH Gene 1.0 ST Array System (http://www.affymetrix.com) was used for microarray assays. For the data processing and analysis, R/BioConductor was first performed to analyze the raw data of microarray, and Robust Multichip Average (RMA) was used for the normalization (scale median = 500). Then, 1-way ANOVA (MATLAB 7.5) (Math-Works, Inc. http://www.mathworks.com/) was used to identity the differentially expressed genes. The significance standard was set as: fold change .2, p-value ,0.01. PCA was also performed to evaluate the repeatability of every sample.
For the hierarchical clustering analysis of significantly changed genes, raw signal values were log2-transformed, centered relative to the median, with the parameters that the distance was a Euclidean distance and the linkage was the average. Expression Analysis Systematic Explorer (EASE) was used to analyze gene ontology and KEGG pathways. Over representation of genes in a gene ontology node or a KEGG pathway is present if a larger fraction of genes within that gene ontology node/pathway is differentially expressed compared with all genes. An EASE score #0.05 was set as the cutoff [51].

Supporting Information
Table S1 Primer sets for Real-time PCR and nest PCR. Found at: doi:10.1371/journal.pone.0010320.s001 (0.05 MB DOC) Figure S1 Differentially expressed genes of GeneSet B in P53 signaling pathway. Genes in GeneSet B were mapped to the KEGG signaling pathway database; and the overrepresented genes were marked in yellow, with the non-corresponding ones marked in green. p-value , 0.05. Found at: doi:10.1371/journal.pone.0010320.s002 (1.07 MB TIF) Figure S2 Differentially expressed genes of GeneSet C in P53 signaling pathway. Genes in GeneSet C were mapped to the KEGG signaling pathway database; and the overrepresented genes were marked in blue, with the non-corresponding ones marked in green. p-value , 0.05.