During early embryogenesis, Oct4 and Sox2 are found to be coexpressed in several pluripotent cells such as the morula, ICM, epiblast, and germ cells. Gene knockout studies revealed that the primary defect for both the Pou5f1-
-null animals is in the pluripotent epiblast, though there are slight differences between the two null phenotypes. There is no epiblast development in the Pou5f1
-null blastocyst, and the fate of all cell types is towards trophectoderm lineage (26
). On the other hand, Sox2
-null animals are capable of giving rise, at least transiently, to the epiblast, as epiblast-derived extraembryonic endoderm is detected (3
). This transient epiblast formation was suggested to result from maternally derived Sox2 protein. Both Pou5f1-
-null blastocysts are incapable of giving rise to pluripotent ESCs.
In this study, we use ESCs as a model for understanding the role of Oct4 and Sox2 in the genetic regulatory network of pluripotent cells. We first established the conditions under which Oct4 and Sox2 interact with the Pou5f1 and Sox2 enhancers in order to understand the function of these regulators at these cis-regulatory target sites. Using specific antibodies against Oct4 and Sox2, we showed by ChIP that these two transcription factors bind to both enhancer elements in mouse and human ESCs (Fig. , , and ). Hence, these enhancer elements are the direct targets of their respective gene products and are reciprocally bound by the other regulator. As we detected the highest level of binding in undifferentiated cells when these genes are known to be most transcriptionally active, we conclude that they are associated with the transcriptional activation of these genes and may play a role in positively regulating expression. This was further confirmed through a functional analysis of the interactions in genetic studies.
Using RNAi with reporter gene constructs, we showed that the silencing of Pou5f1 or Sox2 led to the down-regulation of Pou5f1 and Sox2 enhancer activities (Fig. , , and ). More importantly, the endogenous transcripts and proteins were also reduced by RNAi in a manner that was consistent with the reporter studies. The effects of RNAi on the reporters may have been indirect, as the silencing of endogenous Pou5f1 or Sox2 led to the differentiation of ESCs and further caused the level of endogenous Oct4 and Sox2 to decrease. However, coupled with the observation that Oct4 and Sox2 directly bind to these regulatory elements when the genes are active, our results indicate that Oct4 and Sox2 play positive roles in the expression of Pou5f1 and Sox2.
The interactions of this Sox2-Oct4 complex with the respective genes can be described as a transcriptional regulatory network consisting of autoregulatory and multicomponent loops (Fig. ) (21
). A network motif is a fundamental unit within a complex transcriptional regulatory network. In an autoregulation model, the gene product binds to its own regulatory element (Fig. ). This may allow for self-perpetuation and enhanced stability of gene expression. An additional relationship between Oct4 and Sox2 is depicted by a multicomponent loop motif whereby a regulator binds to the regulatory elements of another regulator in a closed loop (Fig. ). Such a closed-circuit loop can efficiently generate a bistable system with the ability to switch between two different states. For ESCs, the two states may be the decision to undergo self-renewal or to differentiate and exit from symmetrical division.
FIG. 8. Oct4-Sox2 regulatory circuitry in ESCs. Transcription factors are represented by ovals, and regulatory elements of the genes are represented by rectangles (gene names are printed in italics). (A) The relationships between Oct4 and Sox2 and their respective (more ...)
This model also requires that the concentrations of the two factors remain relatively constant, as any slight change in the abundance of one protein will destabilize the circuitry. Interestingly, it has been shown that mouse ESCs are exquisitely sensitive to the level of Oct4 (29
). Increasing the level of Oct4 by 50% is sufficient to induce differentiation of ESCs into primitive endoderm and mesoderm. We show that mutating either the oct or sox site in the distal enhancer does not completely abolish its enhancer activity, with 60% of the activity still being retained (Fig. ). It is conceivable that the modest reduction in Pou5f1
expression may have a significant cellular effect. It would also be of interest to determine whether the alteration of Sox2 levels gives a similar phenotype.
Differential occupancy of the Oct4/Sox2 complex on the various regulatory elements was suggested by the EMSA results. For instance, a direct comparison of ESC nuclear extracts binding to Nanog
at the sox-oct composite elements indicated clear differences in binary and ternary formation between the two (39
). In addition, the EMSA indicated that Oct4 was incapable of binding by itself on the Fbx15
composite element (46
). This differential binding is likely attributed to variations in the sox-oct composite element, given that the Fgf4
element contains an intervening 3 bp while the Fbx15
element contains an A rather than a C in the otherwise invariant fourth position in the octamer motif (Fig. ). Whether the differential binding (as observed in the EMSA) alters transcriptional activity of the Oct4/Sox2 complex itself remains to be seen, but gene-specific sequence conservation within these sox-oct composite elements suggests functional significance.
Although ESCs grown in culture may not mimic the physiological conditions of cells within the ICM of the blastocyst, they provide a good model for understanding the transcriptional regulatory networks in pluripotent cells. Oct4 and Sox2 are key regulators for pluripotency in ESCs. We have recently identified Nanog
as a downstream target of Oct4 and Sox2 (39
). This expands the list of genes (Fgf4
, and Fbx15
) which are potentially regulated by both Oct4 and Sox2. It is also important to note that direct binding of these regulators on the regulatory elements of Fgf4
, and Fbx15
has not been demonstrated by ChIP. Nevertheless, it is apparent that Oct4 and Sox2 are at the top of the hierarchy of transcriptional regulators in ESCs. The cooperative binding of Oct4 and Sox2 may thus be instrumental in recruiting various other interacting protein partners. It should be emphasized that not all Oct4 and Sox2 sites on the same regulatory region are synergistic in transcriptional activation. For example, in the Opn
intron, a sox site 39 bp away from an inverted pair of Oct4 sites acts antagonistically in transactivation by Oct4 (6
). Repression by Sox2 was shown to require DNA binding and a carboxy-terminal transactivation domain. For Fgf4
, and Fbx15
, the regulatory elements contain Oct4 and Sox2 sites in proximity (either 0- or 3-bp separation) and the Oct4/Sox2 complex is implicated in transactivation. This raises the interesting possibility that perhaps Oct4 and Sox2 collaborate to globally control ESC-specific gene expression through the sox-oct motifs. To address this possibility, the direct targets of both regulators have to be identified. One can then compare these two factors and identify the commonality between them. It is also plausible that the Oct4/Sox2 complex interacts with another factor(s) to activate the network of ESC-specific genes. For example, the Pou5f1
distal enhancer contains three distinct regions (CR4-A to -C) that contribute to expression. Does the Oct4/Sox2 complex communicate with the regulatory protein(s) bound to CR4-A and CR4-C? Are they ubiquitous or ESC-specific factors? These questions require further characterization of the regulators bound to CR4-A and CR4-C. It is also important to note that there are other key regulators for maintenance of the undifferentiated state of ESCs. The LIF/Stat3 pathway is essential for self-renewal of mouse ESCs (23
). The removal of LIF leads to the inactivation of Stat3 and induces differentiation. The other key regulator is Nanog. The removal of Nanog
via gene targeting or RNAi leads to differentiation of mouse ESCs (10
). Intriguingly, overexpression of Nanog is sufficient to bypass the LIF/Stat3 requirement. However, how Stat3 and Nanog interact with the Oct4/Sox2 pathway remains to be studied.
An autoregulation mode can generate a self-perpetuating cycle to maintain stable gene expression through a positive feedback loop. The question is, how can this cycle be broken? In the case of Pou5f1
, one key answer is the involvement of a repressor(s) that is induced in differentiated cells. The transcription of Pou5f1
is down-regulated when ESCs differentiate in vitro and when the epiblast differentiates during embryogenesis. There exists a mechanism to break out of the loop of Pou5f1
expression. The promoter of Pou5f1
contains negative regulatory elements which are required for repression when embryonal carcinoma cells differentiate (4
). Interestingly, germ cell nuclear factor (Gcnf) has been shown to mediate repression of the Pou5f1
proximal promoter (16
). The expression of Gcnf
is inversely correlated with the Pou5f1
expression in embryonal carcinoma cells. More importantly, in Gcnf
-knockout mouse embryos, the Pou5f1
expression is no longer confined to the germ cell lineage and novel Pou5f1
expression domains are detected. The loss of Pou5f1
expression may subsequently extinguish Sox2
transcription and the expression of other downstream target genes. It is not clear whether Sox2
is similarly subjected to repression when ESCs differentiate. Nevertheless, the finding highlights the importance of an active mechanism to shut down a key regulator(s) upon differentiation.