The characterization of protein-protein interactions is a very efficient strategy for understanding protein function and regulation. The development of high-affinity tags, including the TAP (Rigaut et al., 1999
) and in vivo biotinylation tag (de Boer et al., 2003
), in combination with advances in mass spectrometry that now allow protein identification with high sensitivity and accuracy, has recently produced several protein interaction network reports. However, most studies in the literature rely on cDNA overexpression driven by exogenous promoters or transgenic random integration approaches.
We report here an epitope-tagging strategy for the purification of protein complexes in mouse ESCs. We introduced the tag by recombineering into a full-length Oct4
-containing BAC and then integrated this in a precise location in the mouse genome. This approach has the advantage of maintaining the endogenous promoter and therefore natural transcriptional regulation. The technology is amenable to high-throughput delivery, as recently demonstrated by random integration of tagged BAC transgenes (Poser et al., 2008
), and should greatly facilitate systematic tagging of genes and analysis of protein complexes with roles in development in different contexts, be it in stem cells, differentiated cell types, or even mouse tissue (Fernández et al., 2009
The affinity purification method described here is rapid, with the goal of capturing weak or short-lived interactions. Previous proteomic studies of Oct4 protein complexes have relied on lengthy single or tandem purifications from nuclear extracts with streptavidin capture (Wang et al., 2006
) or anti-Oct4 antibodies (Liang et al., 2008
) and yielded small data sets, very similar to our tandem purification data set. We identified all of the partners reported by the Liang study except Hdac2, and only five Oct4 partners found in the Wang study were not detected in our data set, maybe because of our use of whole extracts. Indeed, our approach has produced by far the most extensive analysis of Oct4-associated proteins to date.
By using whole extracts, thereby not restricting the analysis to the nuclear environment, our data set encompasses diverse aspects of the life of Oct4, both nuclear and nonnuclear. The broad data set puts Oct4 at the center of diverse cellular processes that can have an impact on aspects of stem cell biology (), the most interesting of which are discussed below.
Schematic Model of the Oct4 Interactome
Oct4 can both activate and repress transcriptional targets in mouse and human ESCs (Babaie et al., 2007; Loh et al., 2006
). To date, Oct4 has been shown to be associated mainly with members of repressor complexes NuRD and SWI/SNF (Liang et al., 2008; Wang et al., 2006
). We found both among our data set of Oct4-copurifying proteins. NuRD, a histone deacetylase complex, was the most prominent, further confirming this link. Sall4, a well-known Oct4 partner, and other members of the Spalt-like family of transcriptional cofactors have been shown to associate to NuRD (Lauberth and Rauchman, 2006
), raising the possibility that they may bridge the interaction between Oct4 and NuRD. This hypothesis is also supported by the similar amounts in which they are detected in our experiments. We also found several subunits of the SWI/SNF nucleosome-remodeler complex, some of which have previously been linked to Nanog (Liang et al., 2008
), confirming the link to this chromatin remodeling complex.
Also among Oct4 binding proteins we found various molecules involved in positive regulation of transcription, including several activators and coactivators and chromatin-modifying enzymes such as Myst2, a histone H4 acetyltransferase (Doyon et al., 2006; Sterner and Berger, 2000
). In addition, we detected Ttf2, a component of the general transcription machinery, providing evidence of a physical link between pluripotency factors and basal transcription players. The Oct4 interactome included other basal DNA-process-related factors such as proteins involved in DNA replication, recombination, and repair. This could explain why many of the Oct4-interacting proteins are ubiquitously expressed in both differentiated and undifferentiated cells. Our experiments suggest that the interaction is not DNA mediated, given that copurification of DNA-binding proteins still occurs upon DNA elimination by benzonase.
Importantly, we have uncovered enzymes with a potential role in Oct4 regulation. Ogt is responsible for posttranslational addition of O-linked N-acetylglucosamine (O-GlcNAc), a regulatory protein modification similar to phosphorylation possibly working in concert with it (Kamemura and Hart, 2003
). Oct4 is modified by O-GlcNAc in human ESCs (Webster et al., 2009
), and Sp1, one of Oct4 partners, is too (Jackson and Tjian, 1988
). A thorough analysis of O-GlcNAc modification in the Oct4 interactome might yield important insight into dynamic modulation of stem cell factors. Posttranslational modification of transcription factors and cofactors is proving to be a critical component of the regulation of gene transcription in general, and important specifically in stem cell biology (Brill et al., 2009; Van Hoof et al., 2009
Half of Oct4-associated proteins seem to be directly regulated by transcription factors with key roles in stem cell pluripotency and/or reprogramming. This is also a characteristic of pluripotency networks derived from smaller data sets from different entry points (Orkin et al., 2008; Wang et al., 2006
). This indicates that even in the expanded and functionally diverse network, this attribute still holds true, supporting a previously unsuspected role in stem cell biology for some of the proteins we identify here.
Expression of Oct4 decreases in a switch-like fashion as ESCs differentiate into lineage-specific cell types, including progenitor cells. Our analysis has uncovered 33 physical interactors of Oct4 that share this trend. Among these are several transcription factors, such as the DNA methyltransferase 3-like regulatory protein Dnmt3l, which stimulates genomic imprinting in germ cells (Bourc'his et al., 2001; Gowher et al., 2005
). This is consistent with a recent report demonstrating that treatment with DNA methyltransferase inhibitors can improve the efficiency of the reprogramming process of differentiated cells (Mikkelsen et al., 2008
). Therefore, the 33 interactors upregulated in ESCs and the transcription factors that regulate them might be interesting candidates whose expression could be manipulated to facilitate reprogramming.
We find that loss of function of most Oct4-associated genes studied to date results in embryonic or perinatal lethality, suggesting that many serve crucial functions in development. Interestingly, most Oct4-binding proteins linked to a human hereditary disorder (13 of 14), mostly developmental or cancer predisposition, give rise to a related phenotype when absent in the mouse. We find cancer-associated genes, either causal or predisposing, to be transcriptional regulators involved in processes relating to the cell cycle, differentiation, and DNA repair, acting through chromatin remodeling, signaling, or transcription factor activity. These results implicate the orthologs of Oct4-interacting proteins in roles in human development and cancer, and therefore the data presented here should be useful in elucidating their part in human disease.
In summation, the extensive systems-level analyses described here compiling data sets of currently available genome-wide studies provide an integrated vision of the Oct4 interactome. Detailed investigation of this information should facilitate the choice of candidate factors to test for roles in ESC maintenance, differentiation, and reprogramming and provide great insight into the transcriptional regulation of ESC biology.