By integrating protein-protein interaction and protein-DNA interaction studies, we constructed a Myc-centered transcriptional regulatory network in an effort to complement the previously identified core regulatory, and Polycomb networks in ES cells. Our approach, analyzed together with data of others, delineates three major transcriptional regulatory subnetworks in ES cells. Based on the target co-occupancy of factors in each network, we defined three functionally separable regulatory modules ( and ), and showed that the overall ES cell gene transcription program can be subdivided largely into functionally independent regulatory units.
It is interesting to note that a previous RNAi-based screen revealed that members of the NuA4 HAT complex (or Tip40-Ep400 complex) are critical in ES cell identity (Fazzio et al., 2008
). Upon knockdown of some of NuA4 HAT complex proteins, as well as Myc, we also observed that ES cells display flattened morphology (Figure S1E
). Of note, knockdown of Ep400 or Dmap1 did not change the expression level of Oct4 and Nanog proteins, nor did knockdown of Nanog change the protein level of Ep400 and Dmap1 ((Fazzio et al., 2008
), also Figure S1D
). These data support the conclusion that the Core and Myc-centered subnetworks in ES cells are separable units with unique roles in maintaining ES cell self-renewal.
Previous studies have suggested that Myc is critical at an early stage in somatic cell reprogramming (Sridharan et al., 2009
). Our work suggests that, beyond Myc itself, reactivation of a larger module comprised of more than 500 genes is critical to achieve partially or fully reprogrammed stem cell-like cells. It is particularly interesting that the Core module, which is composed of more than 100 genes, remains inactive in piPS cells, again implying that the reactivation of an entire functional module by a limited set of factors is critical to achieving induced pluripotency. It will be of interest to determine whether specific small molecules or genes selectively modulate the activity of the ES cell modules in efforts to identify new chemicals or factors not only for replacing Myc or other factors in somatic cell reprogramming, but also for selection of putative therapeutic targets in cancer. Since Myc interacts with NuA4 complex proteins in ES cells, recruitment of the NuA4 HAT complex by Myc may be a critical step in somatic cell reprogramming.
The relationship between ES cell and cancer signatures has been a focus of attention given that self-renewal is a hallmark of both cell types. It has been proposed that the activation of an ESC-like gene expression program in adult cells may confer self-renewal to cancer cells or cancer stem cells (Ben-Porath et al., 2008
; Wong et al., 2008a
). It is noteworthy that we observed very similar patterns of module activity between our Myc module and the previously defined ESC-likes (Core ESC-like gene module and mouse ESC-like gene module) (Wong et al., 2008a
), but not with our Core module, in situations we tested. In accordance with this observation, approximately 60% of genes in the previously defined Core ESC-like module (Wong et al., 2008a
) are Myc targets that we identified (Kim et al., 2008
). Notably, 57% of genes in the Core ESC-like module (Wong et al., 2008a
) are common targets of at least 5 factors among 7 factors in the Myc cluster (). In contrast, less than 2% of genes in the previously defined ESC-like module are shared with the Core module. These findings argue that the previously described ESC-like module (Wong et al., 2008a
) conveys information largely contributed by the Myc module, and conversely that the ESC-like module is quite distinct from the Core module. The simple interpretation that the presence of ESC-like module activity in cancer reflects dedifferentiation or regression to an embryonic or ES-like state (Wong et al., 2008a
) is inconsistent with our analysis.
In their recent work, Ben-Porath et al. (2008)
compiled 13 partially overlapping gene sets belonging to four groups (ES-expressed, active NOS (Nanog, Oct4, and Sox2) targets, Polycomb targets, and Myc targets) which are similar to the modules utilized in our analysis. They showed that poorly differentiated tumors show preferential expression of ES cell specific genes, in addition to preferential repression of Polycomb target genes. Interestingly, their analysis revealed that ES-expressed and Polycomb-target sets show the most significant degree of enrichment in most tumors, while the other gene sets are not a major determinant of their ES cell-like gene expression signature. Of special note, we find that 38% and 52% genes in their ES-expressed gene sets (ES exp1 and ES exp2, respectively) contain the common targets of at least 5 factors among 7 factors in the Myc cluster, suggesting that a large portion of genes in their ES-expressed gene sets are in turn Myc-related genes. It is noteworthy that the PrC module defined in ES cells is also largely repressed in most cancers we tested, suggesting a role of Polycomb complex proteins and their targets in cancer initiation and/or progression.
Our analysis is conceptually different from prior approaches in that we have stringently defined regulatory modules based on common gene targets of multiple factors. By this strategy we have defined modules that serve as powerful analytical tools to interrogate different cellular states and the relatedness of gene expression signatures of ES cells and cancers. Reanalysis of prior datasets in this manner raises concern regarding the hypothesis that cancer cells, or cancer stem cells, recapitulate regulatory programs characteristic of embryonic stem cells. As a unifying view the hypothesis is attractive and has gained considerable attention in recent literature. Nonetheless, our findings should temper enthusiasm and stimulate further reassessment of these issues. Moreover, our findings reemphasize the critical nature of regulatory pathways controlled by Myc in cancer.