Molecular mechanisms of pluripotency and reprogramming

Pluripotent stem cells are able to form any terminally differentiated cell. They have opened new doors for experimental and therapeutic studies to understand early development and to cure degenerative diseases in a way not previously possible. Nevertheless, it remains important to resolve and define the mechanisms underlying pluripotent stem cells, as that understanding will impact strongly on future medical applications. The capture of pluripotent stem cells in a dish is bound to several landmark discoveries, from the initial culture and phenotyping of pluripotent embryonal carcinoma cells to the recent induction of pluripotency in somatic cells. On this developmental time line, key transcription factors, such as Oct4, Sox2 or Nanog, have been revealed not only to regulate but also to functionally induce pluripotency. These early master regulators of development control developmental signalling pathways that affect the cell cycle, regulate gene expression, modulate the epigenetic state and repair DNA damage. Besides transcription factors, microRNAs have recently been shown to play important roles in gene expression and are embedded into the regulatory network orchestrating cellular development. However, there are species-specific differences in pluripotent cells, such as surface marker expression and growth factor requirements. Such differences and their underlying developmental pathways require clear definition and have major impacts on the preclinical test bed of pluripotent cells.

Pluripotency is the potential of stem cells to give rise to any cell of the embryo proper. Th e study of pluripotent stem cells from both mouse and human began with the study of teratocarcinomas, germ cell tumours that occur predominantly in the testis and constitute the most common cancer of young men. In 1954, Stevens and Little [1] found that males of the 129 mouse strain developed testicular teratocarcinomas at a signifi cant rate. Th is fi nding opened the way for detailed studies of these peculiar cancers, which may contain a haphazard array of almost any somatic cell type found in the developing embryo [2]. Th e stem cells of these tumours are embryonal carcinoma (EC) cells, which express charac teristics, including a developmental potential, similar to those of the inner cell mass (ICM) of the early embryo [3]. Experience with these pluripotent malignant EC cells from mouse teratocarcinomas provided the basis for the derivation of embryonic stem (ES) cells from explants of mouse blastocysts independently [4,5]. Indeed, mouse EC and ES cells closely resemble one another, expressing similar markers and, in some cases, similar developmental potentials, although a report of germ line derivation from mouse EC cells in chimeras [6] has never been confi rmed.

Pluripotent cells in the embryo -inner cell mass cells
At the morula stage, totipotent cells start to specialize as they form the blastocyst, comprising an outer layer of cells, the trophectoderm (TE), and a group of pluripotent cells, the ICM. While the TE will develop into placental tissues, the ICM gives rise to all cells of the embryo proper as well as several extraembryonic tissues. Th e earliest factors known to regulate the formation of pluripotent ICM cells are OCT4 and NANOG [7][8][9]. Without OCT4, epiblast cells fail to form and ES cells cannot be derived, while NANOG is required for the germline formation [7][8][9].
Recent studies in the laboratory mouse have provided insights into the molecular mechanisms and key factors regulating the specifi cation of ICM and TE lineages. At Abstract Pluripotent stem cells are able to form any terminally diff erentiated cell. They have opened new doors for experimental and therapeutic studies to understand early development and to cure degenerative diseases in a way not previously possible. Nevertheless, it remains important to resolve and defi ne the mechanisms underlying pluripotent stem cells, as that understanding will impact strongly on future medical applications. The capture of pluripotent stem cells in a dish is bound to several landmark discoveries, from the initial culture and phenotyping of pluripotent embryonal carcinoma cells to the recent induction of pluripotency in somatic cells. On this developmental time line, key transcription factors, such as Oct4, Sox2 or Nanog, have been revealed not only to regulate but also to functionally induce pluripotency. These early master regulators of development control developmental signalling pathways that aff ect the cell cycle, regulate gene expression, modulate the epigenetic state and repair DNA damage. Besides transcription factors, microRNAs have recently been shown to play important roles in gene expression and are embedded into the regulatory network orchestrating cellular development. However, there are species-specifi c diff erences in pluripotent cells, such as surface marker expression and growth factor requirements. Such diff erences and their underlying developmental pathways require clear defi nition and have major impacts on the preclinical test bed of pluripotent cells.

Molecular mechanisms of pluripotency and reprogramming
Jie Na 1 *, Jordan Plews 2 , Jianliang Li 2 , Patompon Wongtrakoongate 2 , Timo Tuuri 2 , Anis Feki 3 , Peter W Andrews 2 and Christian Unger 2 * the morula stage, cells choose their fate depending on their position and polarity [10]. In outside cells, Yap, the co-activator for transcription factor Tead4, localises in the nucleus and increases Tead4 activity. Tead4 subsequently activates the TE master factor Cdx2, which determines the cell fate [10]. Embryos lacking either Tead4 or Cdx2 fail to produce functional trophectodermal tissue but ICM cells remain intact and ES cells can be derived [11,12]. Th e dominance of Cdx2 suppresses Oct4 expression in the outer cells and restricts its expression in the inner cells, which become ICM cells at the blastocyst stage. Th us, the counter-activity between Oct4 and Cdx2 allows the segregation of the fi rst two embryonic lineages [13]. It is noteworthy that this mechanism might be specifi c to mouse as in both rhesus monkey and human, the expression of NANOG is reported to be restricted to the ICM, but OCT4 was detected in TE as well as ICM cells [14].

Embryonic stem cells and species diff erences
Although human ES cells were not derived until 1998 [15], studies of EC cells from human testicular cancers demonstrated signifi cant diff erences between mouse and human EC cells and, by implication, ES cells [16][17][18][19]. Most notably, the cell surface antigens SSEA1 and SSEA3 and 4 are expressed diff erently: mouse EC and ES cells are SSEA1(+)/SSEA3(-)/SSEA4(-), whereas human EC cells are SSEA1(-)/SSEA3(+)/SSEA4(+). Th is surface antigen phenotype of human EC cells is similar to that of human ES cells [15,20] and human ICM cells [21]. A large panel of surface antigen markers and characteristic gene expression patterns for human ES cells has now been identifi ed [20]. A further distinction between human and mouse ES cells, which was also evident in EC cells, is the capacity of human EC and ES cells to generate trophoblastic cells [16]. Th is does not usually occur in mouse EC and ES cells, except after genetic manipulation [13]. Especially in the mouse, a clear distinction between ES cells and epiblast stem cells is being made [22][23][24]. Recent work with human induced pluripotent stem (iPS) and ES cells has produced cells more similar to mouse ES cells by maintenance in low oxygen conditions, or overexpression of OCT4, KLF4 and KLF2 and inhibition of glycogen synthase kinase 3 and mitogen activated protein kinase [25,26]. Th ese culture conditions with physiological oxygen levels (5%) are able to maintain more naïve ES cells [26]. However, it remains to be seen if this reduction of oxidative stress is important for the use of pluripotent stem cells in therapeutic applications.

The history of reprogramming
Th rough early embryonic development and cellular diff er entiation, cells progressively lose developmental potency and choose a specifi c fate [27]. However, the seminal somatic cell nuclear transfer studies of Briggs and King [28] showed that blastula cell nuclei retain the genetic information required for pluripotency when injected into enucleated frog oocytes. Th is phenomenon was investigated further by Gurdon and Uehlinger [29], who demonstrated that even more diff erentiated intestinal cells were capable of directing development into adult frogs following somatic cell nuclear transfer, albeit at low effi ciency (approximately 1%). Th ese early cloning experiments proved that nuclei from terminally diff er entiated cells are capable of generating viable cloned animals, and formed the basis of later mammalian cloning experiments [30].
Th e creation of the fi rst cloned sheep, 'Dolly' , by Wilmut and colleagues [30] together with many other later successful mammalian cloning attempts convincingly demonstrated that the developmental restrictions established during diff erentiation are due to reversible changes in the epigenome, rather than to permanent modifi cations to the genome [31]. Fusing somatic cells with ES cells or exposing them to EC cell extracts can also generate cells with pluripotent phenotypes [32,33]. Th us, the cytoplasm of the oocyte and pluripotent stem cells must contain factors necessary for reprogramming. Th ese studies indicate that key factors that are important for pluripotency within germ cells, early embryos and ES cells may also have the reprogramming ability.
Studies with somatic cells demonstrated that one could redirect cell fate by forced expression of a single lineagespecifi c transcription factor. Weintraub and colleagues [34] found that overexpression of MyoD is suffi cient to convert fi broblasts into muscle cells, while mature B cells can be reprogrammed into macrophages by enforced expression of C/EBPα or C/EBPβ within 3 to 4 days [35]. Th ese studies highlighted the possibility that transdiff erentiation or even dediff erentiation may be mediated by a few defi ned factors.

Induced pluripotency with key factors
In 2006, the ground breaking work by Takahashi and Yamanaka [36] demonstrated that forced expression of four ES cell factors (Oct4, Sox2, cMyc, and Klf4) in fi broblast cells can reprogram them to a pluripotent state. Th e most effi cient method to make iPS cells is through viral transduction due to their high integration effi ciency [37,38]. In properly reprogrammed iPS cells, the transgene driven by the viral promoter should be completely silenced [39]. Failure of silencing indicates incomplete reprogramming and raises the danger of carcinogenesis by the oncogene cMyc [39]. To avoid insertional mutagenesis and transgene reactivation, asso ciated with the viral approach, other methods that do not alter the genome have been developed, such as non-integrating episomal vectors [40], minicircle vectors [41] and the PiggyBac transposon system [42,43]. Transgene-free iPS cells were successfully derived but with lower effi ciency. Th e most attractive approach may be using permeable recombinant proteins [44][45][46], as this elimi nates the possibility of genome alteration by introduced foreign DNA. Th is would also allow the dosage to be controlled and the exposure time of each factor optimised, although this method has not been widely successfully applied.

Molecular mechanisms of reprogramming
Re-establishing pluripotency in a somatic cell is a complicated process. Th e most important changes include the activation of an ES-cell-specifi c transcription network, re-setting the epigenetic landscape, alteration of the cell cycle signature and overcoming the DNA damage response triggered by these drastic changes.

ES-cell-specifi c transcription factors and transcription network
Th e four reprogramming factors discovered by Takahashi and Yamanaka, Oct4, Sox2, Klf4 and cMyc, all have vital roles in early embryogenesis and ES cells [36]. Th e POU domain transcription factor Oct4 is required for the pluripotency of ICM cells and ES cells and is an essential factor in most reprogramming experiments [7]. Although in one recent report the nuclear receptor Nr5a2 was able to replace Oct4, the underlying mechanism appeared to be that Nr5a2 activates Oct4 and Nanog by binding to their promoters and upregulating their expression [47]. A protein interaction study in mouse ES cells showed that Oct4 binds to as many as 92 proteins. Many of these are only expressed by ES cells, but some are ubiquitously expressed in all cells, such as the nucleosome remodelling and deacetylase (NuRD) complex [48]. Th e cellular protein environment can have a signifi cant infl uence on reprogramming. For example, when fusing a somatic cell with an ES cell, or transfering its nucleus into an oocyte, where many OCT4 binding partners naturally exist, reprogramming is much quicker and more effi cient [32,49]. By choosing adult cell types that express more OCT4 interacting proteins, such as neural stem cells and melanocytes where SOX2, a Sry-related high mobility group box transcription factor, is present, one can obtain iPS cells with higher effi ciency and in a shorter time frame [50,51].
In mouse ES cells, it has been shown that Sox2 closely works with Oct4 to regulate the transcription of key pluripotency genes, including Oct4, Sox2 and Nanog [52]. Without Sox2, ES cells cannot eff ectively activate the Oct-Sox enhancers. However, higher levels of Oct4 were able to compensate for the absence of Sox2 and maintain the ES cell phenotype [52]. During reprogramming of mouse fi broblast cells, Sox2 can be replaced by trans forming growth factor-β inhibitors, which have been shown to induce both Nanog and cMyc expression [53,54]. Th us, it appears that Oct4 could work with factors other than Sox2 to achieve cellular repro gramming.
Th e Krüppel-like zinc fi nger transcription factor Klf4 is highly expressed by mouse ES cells and can cooperate with the Oct4-Sox2 complex to activate certain ES-cellspecifi c genes such as Lefty1 [55]. It plays an important role in the pluripotency circuitry by regulating the expression of Sox2 and Nanog [56,57]. By overexpressing Klf4, mouse epiblast derived stem cells (epistem cells) can be returned to the naïve ES cell state [58]. Similarly, increasing the expression of KLF4 and OCT4 or KLF4 and KLF2 enabled human ES cells and iPS cells to exhibit mouse ES cell characteristics, including the ability to grow in leukemia inhibitory factor (LIF) and 2i (ERK1/2 and glycogen synthase kinase 3 inhibitors) as well as the activation of both X chromosomes [25]. iPS cells reprogrammed using Oct4, Sox2 and Klf4 but without cMyc showed lower tumorigenicity [59]. Th e above evidence suggests that Klf4 can prompt cells to acquire a more authentic and naïve ES cell phenotype.
cMyc is an oncogene and seems to act as a catalyst in the reprogramming process as it can signifi cantly increase the effi ciency of iPS cell generation [59]. In ES cells, cMYC was found to occupy promoters of active genes and ES-cell-specifi c microRNAs (miRNAs), including miR-291-3p, miR-294, miR-295, miR-141, miR-200, and miR-429 [57,60,61]. Overexpression of these miRNAs either promoted iPS cell generation or reduced mouse ES cell diff erentiation [60,61]. cMYC can recruit multiple chromatin remodellers, such as histone acetyltransferase GCN5 and histone demethylase Lid, to create an open chromatin state. Th is allows the ectopically expressed ES cell transcription factors to activate their target genes more easily [62]. Th e negative side of cMYC's action is tumorigenecity [59]. Th us, iPS cells created using cMyc need to be carefully scrutinised to ensure the silence of this oncogene.
NANOG and LIN28 can replace KLF4 and cMYC to reprogram human fi broblast cells to iPS cells [38]. Nanog is a core member of the pluripotency circuitry [57] and constitutive expression is suffi cient to support selfrenewal of mouse ES cells in the absence of LIF [8]. Nanog is also required for germline development [63]. Although not absolutely required for reprogramming, including Nanog increased the effi ciency of iPS generation [64].
LIN28 is an evolutionarily conserved RNA binding protein highly enriched in mouse and human ES cells [20,65]. Its function in reprogramming will be discussed in more detail later in the section on miRNAs.
Studies in mouse ES cells revealed that the promoter region of actively transcribed genes was often occupied by multiple key pluripotency transcription factors [57]. Moreover, these factors can self-regulate to reinforce the undiff erentiated state. Diff erent combinations of transcription factors may control distinct subgroups of genes [57]. Th us, to activate the entire ES cell transcriptome, not only is the cooperation of key factors (namely OCT4, SOX2, KLF4 and cMYC) required, but their levels and ratio of expression are also critical [66]. Two very recent publications reported that reprogramming factors can also orchestrate a mesenchymal to epithelial transition, which is important for the initiation stage of repro gramming [67,68]. Down-regulation of epithelialspecifi c factors, such as ECADHERIN, PAR3 and CRB3, suppressed the formation of iPS colonies [68], while suppression of transforming growth factor-β signalling, which is important for epithelial to mesenchymal transition, improved reprogramming effi ciency [67].

Epigenetics
Th e chromatin and DNA modifi cation machinery play critical roles during reprogramming as the epigenetic landscape of a somatic cell needs to be completely reshaped to ES-cell-like states. Th e epigenome of ES cells is characterised by the demethylation of the promoter regions of key pluripotency transcription factors, such as Oct4, Sox2 and Nanog, as well as bivalent chromatin modifi cations on developmentally important transcription factors [69]. Th is ensures a high level of expression of the core factors that maintain pluripotency. At the same time cells reside in a poised state, ready to diff erentiate in response to developmental signals [69].
During reprogramming, the genome loci occupied by histone H3 lysine 4 trimethylation (H3K4me3) and histone H3 lysine 27 trimethylation, which are commonly associated with active and repressive gene expression, respectively, appears to change in accordance with the dediff erentiation process. For example, H3K4me3 marking was lost from promoter regions of mouse embryonic fi broblast-specifi c genes, but increased signifi cantly on the promoters/enhancers of the ES-cellspecifi c genes Fgf4, Oct4 and Nanog [70]. Moreover, the DNA methylation was erased at promoters of pluripotency genes in fully reprogrammed cells but not in mouse embryonic fi broblasts or partially reprogrammed cells [70]. To date, many cell types, including some cancer cells, have been shown to be amenable for reprogramming, refl ecting the plasticity of the epigenome [51,64,71,72]. Diff erent cell types may possess diff erent degrees of plasticity; compared to skin fi broblast cells, epithelial cell types, such as keratinocytes, liver and stomach cells, can be converted to iPS cells with higher effi ciency [72,73]. In addition, a hierarchy of epigenetic states may correlate with a cell's diff erentiation stage. It was found that in the hematopoietic lineage, stem and progenitor cells give rise to iPS cells much more effi ciently than terminally diff erentiated B and T cells [74]. Manipulating the DNA and chromatin modifi cations can greatly facilitate iPS cell formation. Th e DNA methyltransferase inhibitor 5'-azacytidine and the histone deacetylase inhibitor valproic acid increased the reprogramming effi ciency 5-fold and more than 100-fold, respectively [75]. BIX-01294, an inhibitor of the G9a histone methyltransferase, was able to substitute cMyc to induce pluripotency from neural stem cells together with Oct4 and Klf4 [76].
Two recent studies observed that early passage iPS cells still retain some degree of somatic cell memory, which can infl uence the diff erentiation preference of these cells [77,78] . However, these remaining epigenetic memories appeared to attenuate after continuous in vitro culture [77,78]. In addition, some mouse iPS cell lines displayed aberrant silencing of imprinted genes such as the Dlk1-Dio3 cluster. Th ese lines showed poor contribution to chimeric animals and were not germline competent [79,80]. Th e ability of germline transmission is also infl uenced by the combination of reprogramming factors. Mouse iPS cells generated by Oct4, Sox2, Klf4 and Tbx3 were found to contribute to the germ tissue with higher effi ciency compared to iPS cells reprogrammed by Oct4, Sox2, and Klf4 or Oct4, Sox2, and Esrrb [81]. Imprinting abnormalities were found in human iPS cells, including the biallelic expression of H19 and KCNQ10T1 [82]. Th e Fragile X syndrome gene (FX) was active in ES cells derived from embryos with the FX mutation but remained silenced in iPS cells reprogrammed from FXfi broblast cells [83]. Th us, if iPS cells and their derivatives are to be used to model human diseases or in therapeutic applications, several aspects need to be carefully evaluated: the tissue origin and passage number; the reprogramming factors used; the status of imprinted genes; and the histone modifi cation of disease-related genome loci.

microRNAs and reprogramming
miRNAs are approximately 22-nucleotide RNAs that bind to complementary sequences in the 3' untranslated regions of protein coding mRNAs to regulate their degradation or translation [84]. As important modulators of developmental timing and stem cell diff erentiation, they have, unsurprisingly, also been implicated in reprogramming. Th e well-known Let-7 family miRNAs are ubiquitously expressed in somatic cells and upregulated upon ES cell diff erentiation. Th eir mRNA targets include those encoding cell cycle regulators such as K-RAS, cMYC, CDC25A, cyclinD1, and stem cell factors HMGA2, Mlin-41 and IMP-1 [82,85]. Lin28 is an ES-cell-specifi c factor whose major function is to keep let-7 miRNAs at low level by promoting their degradation [86,87]. Indeed, an insightful study by Hanna and colleagues [64] showed that overexpression of Lin28 shortened the cell cycle in monoclonal B cells and sped up iPS cell generation. In another report, the ES-cellspecifi c miRNA miR-294 increased the effi ciency of iPS cell generation by approximately tenfold when introduced together with Oct4, Sox2 and Klf4, but not when cMyc was present [60]. Th e authors then found that miR-294 was a downstream target of cMyc [60]. Interestingly, using a green fl uorescent protein (GFP) reporter driven by the Oct4 promoter, most colonies from the Oct4, Sox2, Klf4 and miR-294 group were positive for GFP expression, indicating that they are more homogenous iPS cell colonies. In contrast, cMyc signifi cantly increased the number of GFP-positive as well as GFP-negative colonies when added together with Oct4, Sox2, and Klf4 [60]. Th is study suggests that ES-cell-specifi c miRNAs are able to fi ne tune the reprogramming process and may be useful to reduce the heterogeneity in iPS cells.

DNA damage
Th e cellular stress imposed by reprogramming can trigger the DNA damage response and subsequently result in cell cycle arrest and senescence. Only a few cells were able to overcome this barrier and become iPS cells. Th is may be the reason why the effi ciency of reprogramming is extremely low. Several studies have demonstrated that when key components (such as p53 and p21) of the DNA damage machinery were deleted, the rate of iPS cell generation is signifi cantly increased [88][89][90][91][92][93][94]. Th e detailed mechanism will be discussed by a separate review in this issue. It has been shown recently that vitamin C supplementation can improve reprogramming effi ciency by alleviating p53-induced cell senescence and synergizing with epigenetic regulators [95]. However, it is dangerous to obtain rapid reprogramming at the cost of inappropriate suppression of DNA damage pathways. SV40 large T can disrupt the nuclear DNArepair foci [96]. When it was added together with the four factors to reprogramme human fi broblast cells, iPS cell colonies emerged after 8 days but many of those iPS cells gained chromosomal abnormalities [97]. Th us, the intricate balance between safeguarding genome integrity and changing cell fate must be carefully maintained during reprogramming.

Future challenges
By elucidating the mechanisms of how pluripotency factors interact with one another and with the genome, it should be possible to devise means to signifi cantly improve reprogramming effi ciency and speed. New interaction partners or pathways might provide explanations to species diff erences and provide the means to a defi ned in vitro culture of pluripotent cells. It remains to be seen whether human ES cells resemble an ICM or epiblast like stage, and whether or not that has any impact on their clinical applicability.
Several articles have reported that iPS cells are notably distinct from ES cells in terms of their gene expression, epigenetic profile, proliferative capacity and the susceptibility of their diff erentiated progeny to cellular senescence and apoptosis [82,83,[98][99][100]. Th ese diff er ences need to be clearly defi ned and may become of importance if developmental research should be translated into the clinic. Th e defi nition of the cells in terms of pluripotency markers and the reproducibility of cell culture conditions will have a major impact on possible future therapeutical applications. Th ere is a need for standardization in clinical protocols, which profi ts from fully defi ned media conditions allowing reproducible growth of pluripotent cells. Initial requirements, such as good manufacturing practice, are the same for human ES cells and human iPS cells [101]. However, iPS cells are certainly a step further away from clinical application than ES cells, as more parameters are yet to be characterized. First among these are issues of safety and effi cacy. Th e earliest methods for the derivation of iPS cells used viral vectors, which may induce insertional mutagenesis and transgene reactivation. Alternative methods for inducing pluripotency without the use of gene insertion have been reported, though their effi ciency needs improvement. Other safety criteria, such as long-term karyotypic stability, appropriate in situ localization, and potential diff erentiation of somatic cells derived from iPS cells, are to be investigated [102,103].

Conclusions
Ultimately, understanding of the underlying mechanisms of pluripotency will be able to guide the way to a safe and new cell-based medicine. Th e modelling of disease and normal development, if well understood, provides the chance to design completely new treatment modalities. Being autologous cells, iPS cells especially allow for a new individualised approach and are able to create a cell model as well as a cell source for each and every person.