Overexpression of NANOG in human ES cells enables feeder-free growth while inducing primitive ectoderm features

Human embryonic stem cells (HESCs) are pluripotent cells derived from the inner cell mass (ICM) of blastocyst-stage embryos(Reubinoff et al., 2000; Thomson et al., 1998). These cells have two distinctive properties: an unlimited capacity for self-renewal and pluripotency (the ability to differentiate to each of the three embryonic germ layers, endoderm, mesoderm and ectoderm). Pluripotency has been shown in vitro by inducing undifferentiated ES cells to form embryoid bodies (EBs)composed of cells from the three germ layers(Itskovitz-Eldor et al., 2000; Schuldiner et al., 2000), and in vivo by injecting undifferentiated cells to immunodeficient mice where they form tumors termed teratomas, composed of multiple differentiated cell types(Reubinoff et al., 2000; Thomson et al., 1998). Additionally, numerous studies have described differentiation of HESCs to specific cell types using growth factors (for reviews, see Schuldiner and Benvenisty,2003). Thus, HESCs hold the promise to serve as a source of cells in transplantation medicine. As they mimic in vitro the onset of early development they are also viewed as a promising model for early human development, during stages that are not accessible to research.

The propagation of HESCs requires the presence of feeder cells (e.g. mouse embryonic fibroblasts, MEFs) or addition of media conditioned by them(Xu et al., 2001). The factors secreted by the MEFs and required by HESCs are not fully characterized,although some factors have been suggested to inhibit the differentiation of cells (Amit et al., 2004; Sato et al., 2004; Xu et al., 2005). One such factor is bFGF, which has been suggested to enable HESCs propagation in the absence of conditioned media (Wang et al.,2005; Xu et al.,2005).

In addition, the intracellular factors that maintain pluripotency and self-renewal of HESCs remain elusive. One factor known to be involved in maintaining pluripotency is OCT4, the downregulation of which in both mouse and human ES cells causes cell differentiation(Matin et al., 2004; Niwa et al., 2000). However,since the derivation of HESCs, it has become apparent that knowledge accumulated on mouse embryonic stem cells (MESCs) cannot automatically be deduced for HESCs. One fundamental difference observed so far is the role of LIF, which in MESCs substitutes the need for support by feeder cells(Smith et al., 1988; Williams et al., 1988). In MESCs, LIF activates Stat3 signaling, which is sufficient to replace the requirement for feeder cells (Niwa et al.,1998). By contrast, in HESCs neither the addition of LIF nor the activation of STAT3 are able to release the cells from dependence on feeder cells (Daheron et al., 2004; Humphrey et al., 2004). BMP4 is another factor shown to be involved in maintenance of self-renewal in MESCs,by inhibiting neural differentiation in serum-free media(Ying et al., 2003). However,when added to HESCs, BMP4 was shown to actually promote differentiation to trophectoderm even in the presence of MEF-conditioned media(Xu et al., 2002). Additional differences between these cells exist, among them are differences in expression of cell surface markers (Ginis et al., 2004) and a different differentiation potential, because HESCs, as opposed to MESCs, are capable of differentiating to trophoblast(Xu et al., 2002).

A gene recently shown to be fundamental in maintaining pluripotency in MESCs is Nanog (Chambers et al.,2003; Mitsui et al.,2003). Overexpression of Nanog releases the cells from LIF dependency, and prevents differentiation upon LIF withdrawal. In addition,its downregulation leads to differentiation to extra-embryonic endoderm. Nanog does not function via the Stat3 pathway, but cooperates with it in maintaining ES cell identity. It was suggested that NANOG may have a key role in sustaining ES cell identity also in HESCs, and that it may have taken over the role of STAT3, which does not seem to be involved in HESC self-renewal.

In this study, we examined the role of NANOG in HESC self-renewal and pluripotency. NANOG is used as a marker for undifferentiated HESCs but its role in these cells is not yet fully characterized. Although it has been shown that its downregulation leads to differentiation of HESCs(Zaehres et al., 2005), the result of its overexpression has not yet been examined. We show that overexpression of NANOG enables the propagation of HESCs for multiple passages in the absence of feeder cells or conditioned media (CM). These cells grow as colonies derived from single cells even in the absence of CM, and lose this ability when the transgene is excised. Additionally, we show that NANOG expression in wild-type cells is upregulated during early differentiation, and that its overexpression in HESCs modifies the expression of marker genes to an expression pattern similar to that of primitive ectoderm cells. Using microarray analysis, we suggest new marker genes that may distinguish between the ICM and primitive ectoderm cells in human.

Human ES cells (H9 and H13 cell lines)(Thomson et al., 1998) were cultured on Mitomycin-C treated mouse embryonic fibroblast (MEF) feeder layer(obtained from 13.5 day embryos) in 85% KnockOut DMEM medium (GIBCO-BRL),supplemented with 15% KnockOut SR (a serum-free formulation) (GIBCO-BRL), 1 mM glutamine, 0.1 mM β-mercaptoethanol (Sigma), 1% nonessential amino acids stock (GIBCO-BRL), Penicillin (50 units/ml), Streptomycin (50 μg/ml),ITSX100 (insulin-transferrin-selenium) in a 1:200 dilution (GIBCO-Invirogen Corporation), and 4 ng/ml basic fibroblast growth factor (bFGF, PeproTech). To obtain a feeder-free culture, the cells were plated on laminin (1μg/cm², Sigma) or gelatin (0.1%, Merck) coated plates and grown in media conditioned for at least 24 hours by MEFs. Differentiation in vitro into embryoid bodies (EBs) was performed by withdrawal of bFGF from the growth media and allowing aggregation in petri dishes. Differentiation in vivo into teratomas was performed by injecting undifferentiated ES cells under the kidney capsule of immunodeficient mice and taken out for analysis 2 months later.

Wild-type ES cells were transfected using the calcium phosphate method as described previously (Chen and Okayama,1988). The cells were transfected with a plasmid described earlier(Chambers et al., 2003) that contains a human NANOG transgene followed by an IRES and a puromycin resistance gene. This transcriptional unit is located between two lox-P sites. Following CRE-recombinase excision there is removal of both the NANOGgene and puromycin resistance, and the transcriptional activation of a GFP gene. NANOG overexpressing clones were established by puromycin selection (0.3 μg/ml, Sigma) following transfection. Revertant clones were established by transfecting NANOG overexpressing clones with a CRE-recombinase plasmid. Following transfection, the cells were trypsinzed to single cells and seeded in low densities (1:1000 split). Cells expressing GFP were selected, expanded and verified to have lost puromycin resistance.

Cells were seeded in 96-well dishes coated with 0.1% gelatin in a density of 2×10⁴ cells per cm² and medium was changed daily. Final cell densities were determined by fixating the cells with 0.5%glutardialdehyde (Sigma) and staining with Methylene Blue (Sigma) dissolved in 0.1 M boric acid (pH 8.5). Color extraction was performed using 0.1 M hydrochloric acid, and staining (which is proportional to cell number) was quantified by measuring absorbance at 650 nm. Each experiment was performed in triplicate.

H9 and H13 Cells were trypsinized to a single cell suspension and seeded in 12-well dishes to a density of 500 cells per cm². After 8 days, the cells were fixed and assayed for alkaline phosphatase activity (86R kit,Sigma) according to the manufacturer's instructions. Each experiment was performed in triplicate and at the end of the experiments the positively stained colonies were counted.

RNA was extracted using TRI-reagent for total RNA isolation according to the manufacturer's instructions (Sigma). cDNA was synthesized using random hexamer primers. Ampification was performed on the cDNA using Takara Ex-Taq. PCR conditions include a first step of 3 minutes at 94°C, a second step of 25-30 cycles of 30 seconds at 94°C, 45 seconds annealing step at 58-64°C, 1 minute at 72°C and a final step of 7 minutes at 72°C. GAPDH was used as a housekeeping gene to evaluate and compare quality of different cDNA samples. Primers and product sizes are listed in Table S1 in the supplementary material. Final products were examined by gel electrophoresis on 2% agarose ethidium bromide-stained gels. Real-time RT-PCR analysis was performed using Rotor-gene 2000 (Corbett Research, Sydney). The reaction was carried out according to the manufacturer's protocol using Absolute SYBR Green Rox Mix (from ABgene, used according to the manufacturer's recommendations) with PCR program as follows: 95°C for 5 minutes; a second step of 35 cycles of 20 seconds at 95°C, 15 seconds annealing step at 60°C, 25 seconds at 72°C and 15 seconds at 82°C; and a final step of 1 minute at 72°C.

For immunostaining cells were washed once with PBS and fixed with 4%paraformaldehyde. Blocking and permeabilization were performed with 2% BSA,10% low-fat milk and 0.1% Triton-X in PBS. Staining with primary mouse anti-human OCT4 was performed for 1 hour (Santa Cruz Biotechnology, used at a 1:50 dilution). As a secondary antibody, Cy3-conjugated goat anti-mouse IgG(H+L; Jackson ImmunoResearch, dilution 1:200), was used. Nuclear staining was performed with Hoechst 33258 (Sigma). FACS analysis for TRA-1-60 expression was performed after trypsinization of the cells. The cells were washed with 3%BSA in PBS with 0.05% Sodium Azid, incubated with TRA-1-60 antibody (kind gift from Prof. Peter Andrews) for 1 hour, incubated with Cy3-conjugated rabbit anti-mouse IgM (Jackson Immunoresearch) and after washes analyzed using the FACSCalibur system (Becton Dickinson). Analysis was performed on CELLQUEST software (Becton Dickinson). Forward- and side-scatter plots were used to exclude dead cells and debris from the histogram analysis.

Western blot analysis was performed according to standard protocols. For NANOG detection a polyclonal rabbit antibody against human NANOG protein(abcam) in 1:1000 dilution was incubated for 18 hours. A secondary antibody(peroxidase conjugated Affinipure goat anti-rabbit IgG by Jackson Immunoresearch) was incubated for 1 hour in 1:10000 dilution. For loading control we used a mouse antibody against α-TUBULIN (Sigma).

Total RNA was extracted according to the manufacturers protocol(Affymetrix). When extracting RNA from undifferentiated ES cells, the cells were grown for one passage on gelatin-coated plates with conditioned media in order to avoid contamination by feeder cells. Hybridization to the U133A DNA microarray, washing and scanning were performed according to the manufacturer's protocol, and expression patterns were compared between samples. Signals were normalized by dividing each probe by the average value of the chip to avoid differences between different chips and experiments.

H9 HESCs were transfected with a plasmid harboring a human NANOGtransgene transcribed from a constitutive promoter and followed by a puromycin resistance gene (Chambers et al.,2003). These sequences are enclosed by lox-P sites, and upon CRE-recombinase excision they are removed, causing a GFP gene to be activated(Chambers et al., 2003). NANOG overexpressing stable clones were isolated following selection with puromycin and expression of the NANOG transgene was verified by RT-PCR analysis using specific primers to the transgene(Fig. 1A, part I). As the puromycin resistant gene lies on the same transcriptional unit as the NANOG transgene, constant selection with puromycin guaranteed stable expression of the transgene over time. To asses the overexpression level of the transgene in different clones, we performed real-time RT-PCR analysis and observed that the total level of NANOG mRNA in the clones was significantly higher than in wild-type cells (Fig. 1A,part II). Concurrently, elevated protein levels were observed in the clones by western blot (Fig. 1A, part III). It has previously been shown that Nanog overexpression releases MESCs from LIF dependency (Chambers et al.,2003; Mitsui et al.,2003). However, regulation of self-renewal in HESCs is not identical to that in MESCs. Thus, we set out to determine the role of NANOG in HESCs. HESCs are grown on mouse embryonic fibroblasts (MEFs)or in the presence of media conditioned by MEFs. Therefore, we first examined whether the requirement of cells for conditioned media (CM) could be replaced by overexpression of NANOG transgene. When wild-type and NANOG overexpressing cells were grown in the presence of CM, both cell types showed similar growth rates(Fig. 1B, upper panel). However, when CM was removed, wild-type cells ceased to proliferate, while NANOG overexpressing cells kept dividing(Fig. 1B, lower panel) even though their growth rate was markedly decreased. In the presence of feeder cells, overall colony morphology of the NANOG overexpressing clones was very similar to that of wild-type cells, albeit being slightly flatter (see Fig. S1 in the supplementary material). However, in the absence of CM, a striking difference is observed, as early as a few days after CM withdrawal. While wild-type cells created only very small and partially differentiated colonies,Nanog overexpressing clones created much larger colonies, a difference that was observed as early as 4 days after CM withdrawal (see Fig. S1 in the supplementary material).

To verify that NANOG overexpressing cells may indeed proliferate without CM, the cells were grown for multiple passages in a feeder-free culture without CM. Under these conditions, wild-type cells ceased proliferating after a few passages and showed completely differentiated morphology (Fig. 2A, left panel). However, NANOG clones continued proliferating for more than 20 passages without CM, with no apparent decline in growth rate or doubling time throughout the passages (Fig. 2A, right panel). These cells remained responsive to CM and grew faster in its presence than in its absence (data not shown). Verification of the undifferentiated identity was assessed using immunostaining for OCT4(Fig. 2B) and FACS analysis using TRA-1-60 antibody. TRA-1-60 staining showed that the percentage of undifferentiated cells (TRA-1-60 positive population) in the culture did not decrease after the withdrawal of feeder cells(Fig. 2C). To further examine whether the cells were still pluripotent, they were assayed for their capacity to form teratomas after injection to SCID mice. Clone-derived teratomas arose at a comparable rate to that observed in wild-type cells and cells grown on feeder cells, and were composed of different cell types from the three embryonic germ layers (Fig. 2D). Furthermore, cell type composition was very similar between the wild-type cell- and NANOG overexpressing cell-derived teratomas.

As NANOG overexpressing clones proliferate in the absence of CM,we examined whether the cells can also form colonies from single cells under these conditions. Cells were trypsinized to single cells, seeded at very low density, and the ability to form colonies was assayed. Resulting colonies were verified as undifferentiated ES cells by assaying the activity of the ES cell marker alkaline phosphatase (AP). When wild type and NANOGover-expressing cells were seeded in the presence of CM, both were able to create a large number of AP-positive colonies(Fig. 3A). When cells were seeded in medium not conditioned by MEFs and not containing bFGF (basal media), the number of colonies formed by the wild-type cells was negligible,while NANOG overexpressing cells were capable of forming a considerably larger number of AP-positive colonies(Fig. 3A,B, AP activity shown as red staining). Therefore, NANOG is capable of maintaining ES cells undifferentiated independently of CM, even though the frequency of colony formation is lower than with CM. To confirm that the effect observed upon NANOG overexpression was not unique to one cell line, NANOG overexpressing clones were established in another cell line, H13. As in H9 cells, H13 cells were able to create colonies in the absence of CM only when transfected with a NANOG overexpression vector (Fig. 3A,B).

To verify that the phenotype observed is dependent on overexpression of NANOG and not on a specific sub-clone, we established revertant clones from NANOG overexpressing cells. Through application of CRE-recombinase, we removed the NANOG transgene, and collected GFP positive colonies (see Materials and methods). The assay was then repeated for the parental clones overexpressing NANOG, and three revertant clones. The results for the revertant clones were similar to those of wild-type cells,and growth in basal media was virtually abolished(Fig. 2C,D). This confirms that the phenotype observed indeed results from the overexpression of NANOG.

In mouse, Nanog acts a stage later than Oct4 during normal embryonic development. Oct4 inhibits differentiation to trophoblast and its knockout in MESCs triggers differentiation to trophoblast(Niwa et al., 2000). Nanog acts to inhibit differentiation to primitive endoderm and its knockout in MESCs triggers differentiation to primitive endoderm(Mitsui et al., 2003). We examined whether this temporal difference is observed in a human system by assaying the expression pattern of both OCT4 and NANOG genes during the differentiation of HESCs. Four time points along their differentiation were assayed: ES cells, the pluripotent undifferentiated stage; 2-day-old embryoid bodies (EBs), the early stage of differentiation;10-day-old EBs, which represent mid differentiation; and 30-day-old EBs, which represent late differentiation (Dvash et al., 2004; Leahy et al.,1999). OCT4 expression is high in ES cells, 2- and 10-day-old EBs, and drops drastically in 30-day-old EBs(Fig. 4B). However, the expression pattern of NANOG is different. It is expressed in ES cells, but its level increases tenfold in 2-day-old EBs, remains high in 10-day-old EBs and drops in 30-day-old EBs(Fig. 4A,B). This implies that NANOG may actually be expressed at the highest level in early differentiating cells and not in undifferentiated cells.

Following its formation, the ICM differentiates to primitive endoderm and primitive ectoderm. We hypothesized that NANOG may be expressed at higher levels in primitive ectoderm than in the ICM in human. There are several known marker genes in the mouse that distinguish ICM from primitive ectoderm (Rathjen and Rathjen,2003). Rex1 (Zfp42 - Mouse Genome Informatics), Gbx2 and Crtr1(Tcfcp2l1 - Mouse Genome Informatics) are enriched in ICM, and Fgf5 and PRCE are absent from ES cells and ICM, and enriched in primitive ectoderm(Fig. 5A). When compared with wild-type cells, in NANOG overexpressing cells there is a significant downregulation of REX1 (fourfold) and GBX2 (twofold) and upregulation of FGF5(ninefold) (Fig. 5B,C; see Fig. S2 in the supplementary material). No change in the expression pattern of CRTR1 and PRCE was observed (data not shown). The magnitude of the change seemed to correlate with the level of overexpression of NANOG, with the highest overexpression (as assessed by the results shown in Fig. 1) leading to the most significant changes in mRNA levels (clone 9, see Fig. S2 in the supplementary material) and the clone with the lowest NANOG expression levels showing less significant changes, and only in two of the markers examined(clone 4, see Fig. S2 in the supplementary material). To examine if the same phenomenon exists in MESCs, we looked at the expression pattern of markers known to distinguish ICM from primitive ectoderm in wild-type and Nanog overexpressing MESCs. MESCs overexpressing Nanog were obtained from the laboratories that demonstrated the effect of Nanogon MESCs (Chambers et al.,2003; Mitsui et al.,2003). However, no change was observed upon overexpression of Nanog in any of the marker genes examined, in the various cell lines(Fig. 5D).

In mouse, it is known that primitive ectoderm like (EPL) cells are capable of reverting back to ES cells (as defined by marker gene expression pattern)(see Rathjen and Rathjen,2003). We therefore examined whether this is also the case in HESCs. The expression of the differential markers was examined in the revertant clones derived from NANOG overexpressing clones. However,the expression of the examined genes remained similar to that of the overexpressing clones, and no reversion to the expression pattern of wild-type cells was observed (Fig. 5C).

DNA microarray analysis was performed on RNA extracted from wild-type and NANOG overexpressing ES cells, and three stages of differentiating embryoid bodies (EBs). Dendogram analysis, which clusters samples according to their degree of similarity, shows that upon overexpression of NANOGin HESCs, the similarity to early differentiating cells increases(Fig. 6A). However, the cells do not appear differentiated as they cluster apart from the samples of 2- and 10-day-old EBs. Four groups of genes differentially expressed between NANOG overexpressing and wild-type cells were identified(Fig. 6B).

(1) Genes upregulated in both NANOG cells and in 2-day-old EBs. These may be new markers of primitive ectoderm. 2 genes belong to this group(Fig. 6C). ALPPL2(alkaline phosphatase, placental-like 2), a gene expressed in certain germ cell tumors and HSPA1A (heat shock protein 70 kDa protein 1A), which encodes a chaperone protein, also involved in the inhibition of apoptosis.

(2) Genes upregulated in NANOG overexpressing cells compared with ES cells and 2-day-old EBs. These may represent transcriptional targets of NANOG.

(3) Genes downregulated in NANOG overexpressing cells and 2-day-old EBs compared with wild-type ES cells(Fig. 6D). These may be ICM(and therefore ES cell) markers that are downregulated upon the transition to primitive ectoderm. Among these genes are LECTIN (LGALS1), a known marker of HESCs (Dvash et al.,2004) that is absent from all samples but ES cells, and HOXA1 (homeo box A1), a transcription factor involved in development. Among the genes downregulated in NANOG overexpressing cells but to a smaller extent are genes suggested to be downstream targets of HOXA1(Shen et al., 2000), such as GBX2 and BMP2.

(4) Genes downregulated in NANOG cells compared with both 2-day-old EBs and wild-type ES cells. These genes may also be targets of NANOG, which has been suggested to be both a transcriptional activator and repressor. In addition, genes reported to be upregulated upon knock-down of Nanog in MESCs(Mitsui et al., 2003) are downregulated upon overexpression of NANOG in HESCs. These include parietal endoderm markers (like LAMB1, downregulated by 3.5-fold) and visceral endoderm markers (like BMP2, downregulated by 14-fold). It has been suggested that NANOG represses expression of primitive endoderm genes in MESCs, and it seems that the same effect is observed in HESCs.

The transition to EPL cells from MESCs has been shown to follow treatment with MEDII medium (a medium conditioned by the human hepatocellularcarcinoma cell line, Hep G2). This medium has been shown in MESCs to lead to a change in the marker gene pattern expression from that of ICM to that of primitive ectoderm. The treatment of HESCs with MEDII(Calhoun et al., 2004) has also shown that the resultant cells are pluripotent in nature though with an early differentiated phenotype. However, unlike MESCs, which show transition to EPL cells, HESCs treated with MEDII displayed similar gene expression to primitive streak cells and nascent mesoderm cells, through activation of the TGFβ1/NODAL pathway. We therefore examined NANOG overexpressing cells to asses whether they show similar changes in gene expression as has been shown after treatment with MEDII. Our NANOG overexpressing cells did not fully mimic the effect of MEDII media, as reported for HESCs. Thus, three of the markers [CRIPTO (TDGF1 - Human Gene Nomenclature Database), FST and TBX1] examined in the report on MEDII treatment showed a similar response in our clones, whereas two showed no effect [GATA6 and ZNF1A1 (ZNFN1A1 - Human Gene Nomenclature Database)], and one had an opposite effect [LEFTYA(LEFTY2 - Human Gene Nomenclature Database), see Fig. S3 in the supplementary material].

The pathways underlying pluripotency and self-renewal in HESCs are largely unknown. Understanding these pathways is hampered by the fact that although much knowledge has been accumulated concerning them in MESCs, much of it does not prove valid for HESCs. In this work, we show that overexpression of NANOG in HESCs enables their feeder-free growth. Single cells over-expressing NANOG are capable of forming colonies in the absence of CM, a capability lost upon excision of the transgene. Although NANOG overexpressing HESCs are capable of growing for multiple generations in the absence of CM, their growth is slowed. Therefore, it would seem that an additional pathway is activated by the addition of CM. After the discovery of the role of Nanog in MESCs, it has been suggested that the lack of effect by LIF on HESCs results from high enough levels of NANOG, so that LIF signaling is dispensable. However, it seems that,as in mouse, human NANOG cannot fully compensate for lack of exogenous factors supplied by the CM.

We further show that NANOG is upregulated upon early differentiation of HESCs. This implies the possible involvement of NANOG in early differentiation. In mouse, Nanog seems to act one stage later in development than Oct4. Oct4 expression is essential for the development of the ICM(Nichols et al., 1998),whereas Nanog expression is essential for the development of primitive ectoderm. Mouse embryos mutated in the Oct4 gene develop up to the blastocyst stage, but their ICM is restricted to the trophoblast lineage (Nichols et al.,1998). Mouse embryos mutated in the Nanog gene also develop up to the blastocyst stage, but when cultured in vitro, their ICM creates only primitive endoderm cells and no primitive ectoderm(Mitsui et al., 2003). This indicates that in both human and mouse, NANOG may promote the transition from ICM to primitive ectoderm. However, the idea that NANOG can actively promote this transition, not only by preventing the transition to primitive endoderm but by actually driving the transition to primitive ectoderm, has not been suggested or shown yet, and this is the first report of such a function. Indeed, a change in marker gene expression occurs upon overexpression of NANOG in HESCs from that characteristic of ICM to that characteristic of primitive ectoderm. A protocol for differentiation to primitive ectoderm was established for MESCs(Rathjen et al., 1999). These cells, termed early primitive ectoderm-like (EPL) cells have been defined in the mouse by three characteristics: different expression of a subset of marker genes that distinguish ICM from epiblast; different cytokine dependency, which includes decreased dependency on LIF and creation of undifferentiated colonies in its absence; and failure to form chimeras upon injection into a blastocyst. We have shown the first two requirements in HESCs overexpressing NANOG, while the third cannot be examined in humans. Although we did not observe a change in all markers known to distinguish ES cells from EPL cells in mouse, this may result from inherent differences between the two species, known to exist for several markers. For example, PRCE, which in mouse is expressed mainly in epiblast and not in ICM or ES cells, is expressed in wild-type HESCs. As human primitive ectoderm cells are not available for research, nothing is known about the in vivo expression pattern of the examined genes, and our analysis was based on the data established in mouse. However, although the population of NANOG overexpressing cells is undoubtedly a pluripotent population, a divergence from the normal marker gene expression pattern of ES cells does occur. NANOG overexpressing cells also have different cytokine dependency, as observed by the ability to grow in the absence of CM, another characteristic of mouse primitive ectoderm cells.

When expression of ICM-specific marker genes was examined in wild-type and in Nanog overexpressing MESCs, no change was observed in either of the cell lines examined. One possible explanation is species differences between mouse and human. Another possibility is that this is a result of culture conditions or in vitro effects. For example, LIF has been shown to inhibit the transition from ICM to primitive ectoderm, and the appearance of primitive ectoderm markers, such as Fgf5(Shen and Leder, 1992).

Another feature of EPL cells in mouse is that the primitive ectoderm gene expression pattern can be reverted back to an ICM expression pattern. In MESCs converted to EPL cells, once MEDII is removed and LIF is returned to the medium, genes that were differentially expressed are reverted back to levels distinctive to ICM. However, in HESC revertant clones (after the removal of the NANOG transgene), although the growth characteristics are restored to those of wild-type cells, the expression of REX1, GBX2and FGF5 remains similar to that in the parental clones. This difference can be attributed to several differences between the experimental systems. One of these differences may be the difference between mouse and human ES cells, which could mean that although in mouse this transition is reversible, in human cells it is not. However, it could also be the result of the different time-frames used in each experiment. While in the mouse system the method to derive EPL cells used treatment with MEDII for only a few days(these cells have been reported to undergo crisis after 8 days, see Rathjen and Rathjen, 2003), in our hands the cells were passaged several times (at least five) before deriving the revertant clones. Therefore, the human mouse discrepancy may not originate from lack of the ability of human cells to revert from primitive ectoderm to ICM, but it may be that the longer time frame fixed the fate of the cells as primitive ectoderm cells, not permitting return to an ICM state.

When NANOG overexpression in HESCs is compared with what is published on HESCs treated with MEDII(Calhoun et al., 2004), two points seem noteworthy. First, in both cases the cells remain pluripotent and express markers of pluripotent cells; second, in both cases there seems to be an activation of the TGFβ1/NODAL pathway. Upregulation of CRIPTO(a co-receptor/ligand of NODAL) and downregulation of FST (which may function as an activin-binding inhibiting protein and therefore an inhibitior of this pathway) were observed in both cases. However, in contrast to treatment with MEDII, in our cells LEFTYA was actually upregulated. Still, LEFTYA is a transcriptional target of NODAL and therefore this is consistent with activation of the pathway. Following treatment with MEDII no change was reported regarding the primitive ectoderm markers, REX1,GBX2 and FGF5, and some mesoderm markers shown to be upregulated upon MEDII treatment, did not change upon overexpression of NANOG. Treatment with MEDII results in the addition of multiple factors to the cells,factors which may work together or apart to direct cell fate to more than one direction. By contrast, overexpression of NANOG probably creates a more unified cell culture, as only the expression of a single gene is altered compared with wild-type cells. Therefore, this report is the first to describe the transition from ICM to primitive ectoderm induced by the genetic manipulation of a specific gene.

Using microarray analysis, we searched for genes differentially expressed between wild-type and NANOG overexpressing HESCs. NANOG has been suggested both to repress differentiation to primitive endoderm and to actively maintain pluripotency. As parietal endoderm markers are among the genes downregulated upon overexpression of NANOG, it is likely that human NANOG also inhibits differentiation to primitive endoderm. The larger number of genes downregulated by NANOG overexpression than those upregulated might suggest that a significant portion of the effects of NANOG may come from repressing differentiation into primitive endoderm. Genes upregulated after NANOG overexpression may have importance in future research as they may contribute to the release from CM dependency. One pressing issue in current ES cell research is the establishment of feeder-free cultures. Therefore, pathways that govern HESC self-renewal and pluripotency have to be elucidated. Future research of the group of genes upregulated in NANOG over-expressing clones may facilitate and optimize this goal, as understanding how NANOG enables feeder free growth may improve the culture obtained and define the requirements of the cells. A clue to a possible mechanism of the release from conditioned media dependency comes from the fact that LECTIN1, one of the genes downregulated upon overexpression of NANOG, is involved in promotion of apoptosis (Yang and Liu,2003). Similarly, HSPA1A, one of the genes upregulated by NANOG, is known to inhibit apoptosis(Gabai et al., 2002). Therefore, the total effect observed upon NANOG overexpression may result from two activities: the inhibition of differentiation and an increase in cell survival that results from downregulation of mechanisms that promote apoptosis.

Finally, using microarrays, new genes expressed differentially between ICM and primitive ectoderm in humans are suggested. Genes expressed differentially between these two closely related populations can be used to identify the primitive ectoderm population. It is likely that at least some of these genes have a role in the transition from one cell population to the other. Most importantly, the recapitulation of the first differentiation step of the ICM in vitro shows that early stages of human development, not accessible otherwise to research, may indeed be mimicked in culture.

We thank Dr Ian Chambers and Prof. Austin Smith for their assistance with establishing part of the experiments described and for kindly supplying us with the NANOG expression vector. We also thank Prof. Shinya Yamanaka for kindly supplying us with mRNA from wild-type and Nanog overexpressing RF8 MESCs. This research was partially supported by funds from the Israel Science Foundation and by an NIH grant.