Prediction and Testing of Novel Transcriptional Networks Regulating Embryonic Stem Cell Self-Renewal and Commitment

Prediction and Testing of Novel Transcriptional Networks Regulating Embryonic Stem Cell Self-Renewal and Commitment

Cell Stem Cell Article Prediction and Testing of Novel Transcriptional Networks Regulating Embryonic Stem Cell Self-Renewal and Commitment Emily Walk...

3MB Sizes 0 Downloads 9 Views

Cell Stem Cell

Article Prediction and Testing of Novel Transcriptional Networks Regulating Embryonic Stem Cell Self-Renewal and Commitment Emily Walker,1 Minako Ohishi,1 Ryan E. Davey,1 Wen Zhang,2 Paul A. Cassar,3 Tetsuya S. Tanaka,1 Sandy D. Der,4 Quaid Morris,5 Timothy R. Hughes,2,5 Peter W. Zandstra,1,6 and William L. Stanford1,2,6,* 1

Institute of Biomaterials and Biomedical Engineering Department of Molecular and Medical Genetics 3 Institute of Medical Science 4 Laboratory Medicine and Pathobiology 5 Banting and Best Department of Medical Research 6 Department of Chemical Engineering and Applied Chemistry University of Toronto, 164 College Street, Toronto, ON, M5S 3G9, Canada *Correspondence: [email protected] DOI 10.1016/j.stem.2007.04.002 2


Stem cell fate is governed by the integration of intrinsic and extrinsic positive and negative signals upon inherent transcriptional networks. To identify novel embryonic stem cell (ESC) regulators and assemble transcriptional networks controlling ESC fate, we performed temporal expression microarray analyses of ESCs after the initiation of commitment and integrated these data with known genome-wide transcription factor binding. Effects of forced under- or overexpression of predicted novel regulators, defined as differentially expressed genes with potential binding sites for known regulators of pluripotency, demonstrated greater than 90% correspondence with predicted function, as assessed by functional and high-content assays of self-renewal. We next assembled 43 theoretical transcriptional networks in ESCs, 82% (23 out of 28 tested) of which were supported by analysis of genome-wide expression in Oct4 knockdown cells. By using this integrative approach, we have formulated novel networks describing gene repression of key developmental regulators in undifferentiated ESCs and successfully predicted the outcomes of genetic manipulation of these networks. INTRODUCTION Embryonic stem cells (ESCs) are unspecialized cells that have the ability to self-renew, producing daughter cells with equivalent developmental potential, or to differentiate into more specialized cells. They are derived from the inner cell mass of the preimplantation embryo and are pluripotent, as they are able to differentiate in vivo into all

cell types of the adult organism, but not into extraembryonic tissue. Exogenous control of the ESC state can be achieved by a limited number of factors. When grown in the presence of murine embryonic fibroblast feeder cells (Evans and Kaufman, 1981; Martin, 1981) or leukemia inhibitory factor (LIF) (Smith et al., 1988; Smith and Hooper, 1987; Williams et al., 1988), murine ESCs remain undifferentiated. Three transcription factors are known to be critical in the maintenance of ESC pluripotency: Oct4, Nanog, and Sox2. Oct4 (Pou5f1) has a highly conserved role in maintaining pluripotent cell populations (Nichols et al., 1998; Morrison and Brickman, 2006), and its expression level dictates ESC fate (Niwa et al., 2000). SOX2 forms a complex with OCT4 and is necessary to cooperatively activate target genes in ESCs (Yuan et al., 1995; Ambrosetti et al., 1997). These factors comprise one essential circuit regulating ESC pluripotency in which OCT4 regulates Sox2, and additionally, the OCT4-SOX2 complex activates Oct4 expression (Okumura-Nakanishi et al., 2005). Forced overexpression of Nanog maintains pluripotency and OCT4 levels in ESCs, even in the absence of LIF (Chambers et al., 2003; Mitsui et al., 2003), while it is itself regulated by OCT4 and SOX2 (Rodda et al., 2005). All three factors are downregulated during differentiation induced by LIF withdrawal or retinoic acid (RA) induction. Despite advancements in our understanding of these three critical transcription factors, there remains limited understanding of upstream and downstream regulators of stem cell fate and of how different regulatory networks are activated to guide lineage commitment. We hypothesized that networks important for the stability of the ESC state would be systematically perturbed at the initiation of commitment. Here, we present an analytical framework, which enabled us to capture the expression profile of ESCs at the transition from self-renewal to commitment. We combined our expression data with tissue-specific microarray studies and with available ChIP-chip analyses to predict novel regulatory networks controlling ESC fate. These networks are novel in both Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 71

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

the predicted cooperation of OCT4 and NANOG with additional pluripotency genes, as well as in the identified developmental targets. To confirm these predicted interactions, we perturbed expression of key transcription factors, Oct4 and Sox2 by shRNA knockdown, and performed microarray expression analyses. Finally, to confirm the importance of predicted pluripotency genes, we generated shRNA knockdowns of candidate genes and developed a unique, high-throughput screen quantifying self-renewal. We have used this integrated approach of combining multiple genomics platforms, genetic manipulation, and high-content screening to make the following discoveries regarding ESC fate: (1) highly expressed transcription factors and chromatin remodeling genes are downregulated during differentiation, whereas developmental genes are upregulated; (2) 281 genes are consistently downregulated regardless of the method of differentiation, whereas the identity of upregulated genes depends on the method of differentiation; (3) of all previously identified binding (but not functionally validated) targets of OCT4, NANOG, and SOX2, 39 are downregulated and 71 are upregulated both during differentiation and after Oct4 and Sox2 knockdown; (4) knockdown of downregulated genes impairs ESC self-renewal, whereas knockdown of upregulated genes does not; (5) enforced overexpression of the downregulated gene MKRN1 confers heightened ability to selfrenew in differentiating conditions; (6) in 43 novel regulatory networks, polycomb group (PcG) proteins EED and PHC1 are predicted to cooperate with OCT4 and NANOG to maintain repressive control of key developmental regulators; and (7) 82% of the targets of these networks were supported by genome-wide expression analysis of Oct4 knockdown cells. Together, these findings define the regulatory balance that maintains ESC state and suggest that lineage commitment is directly connected to the loss of self-renewal. RESULTS Time Course of ESC Commitment As a first step in understanding differentiation dynamics, we reasoned that genes critical for maintaining pluripotency would be downregulated during commitment, whereas genes critical for commitment would be upregulated. To identify these genes, we designed two time courses of differentiation. Mouse Oct4:eGFP R1 ESCs (Viswanathan et al., 2003) were differentiated in monolayer culture after either LIF withdrawal or LIF withdrawal supplemented by RA addition. Cells at each time point, including undifferentiated controls, expressed variable levels of eGFP, consistent with other data describing heterogeneous OCT4 expression within a population of ESCs (Davey and Zandstra, 2006). Despite this, the vast majority of the control population showed high eGFP expression, which decreased throughout both time courses (Figure 1A). eGFP expression decreased more rapidly in the RA time course, likely due to the direct repression of the Oct4 promoter, which contains 72 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

an RA response element (Schoorlemmer et al., 1994). At each time point, cells sorted by eGFP expression were designated as ‘‘high,’’ ‘‘medium,’’ and ‘‘low,’’ while recognizing that all three populations in fact exhibited very high eGFP expression (Figure 1A) and were thus likely still at the initial stages of differentiation (i.e., commitment). All cells below the ‘‘low’’ threshold were discarded to avoid confounding results with gene expression changes occurring in more differentiated cells. Microarray analysis was performed on each sorted population of cells. Microarray probes targeting Oct4 showed gradual and consistent downregulation throughout both time courses, as predicted by the downregulation in eGFP, as did Nanog and Sox2 probes (Figure 1B). Oct4 downregulation, as well as regulation of eight additional genes, including differentiation markers T and Gsc, was confirmed by quantitative real-time PCR (qPCR) (Figure 1C and Figure S1B, which can be found in the Supplemental Data available with this article online). Thus, by differentiating ESCs under carefully defined conditions, we created a model of the initial stages of commitment, the validity of which is supported by the observed incremental losses in OCT4, NANOG, and SOX2. Studying global gene expression changes at these time points will enable us to define the precise temporal relationships between genes involved in pluripotency and commitment. Prediction of a Repressive Model of Stem Cell Maintenance To ask whether regulated genes had a common function, all regulated genes were sorted into four categories: down or upregulated after either LIF withdrawal or RA addition (Figures 2A–2D and Figure S2). Unregulated probes either fluctuated or showed insignificant regulation (Figure S3A) and were used as a control group in later analyses. We compiled a list of Gene Ontology (GO) terms that showed statistically significant overrepresentation within each regulated group, compared to the representation of that GO term within the list of mouse genes in the MGI database (Beissbarth and Speed, 2004). Downregulated lists were enriched for transcription factors, transcriptional repressors, DNA binding proteins, and chromatin remodeling genes (Figures 2A and 2B). Upregulated genes were involved not only in transcription but also with cell differentiation, morphogenesis, pattern specification, and tissue, organ, and system development (Figures 2C and 2D). The specific developmental programs initiated by each time course were not the same. LIF withdrawal caused upregulation of genes involved in blood vessel, skeletal, and nervous system development (Figure 2C), whereas RA addition caused upregulation of genes involved in nervous system development and neurogenesis (Figure 2D), consistent with the practice of differentiating ESCs to neural precursors with RA. We predicted that unregulated genes would not be involved in self-renewal or differentiation pathways, and GO analysis confirmed that the unregulated list was enriched for terms related to normal cell function, including cell cycle, cellular biosynthesis, and cellular metabolism (Figure S3B).

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

Figure 1. Differentiation Time Courses Established a Model of ESC Commitment (A) FACS data showed decreasing Oct4:eGFP expression over 5 days of LIF differentiation and 2 days of +RA differentiation. At each time point, cells were sorted into three groups—high (H), medium (M), and low (L)—based on Oct4:eGFP expression, and a separate microarray hybridization was performed for each sorted population, for a total of three hybridizations for each time point. (B) Oct4, Nanog, and Sox2 expression correlated well with the observed decrease in eGFP, decreasing gradually and consistently between each sorted population of cells, and between the two modes of differentiation. Dotted lines are drawn when a hybridization was not performed for an individual sorted cell population. Specifically, there is no hybridization for Control (low) or RA, day 2 (high). (C) qPCR of nine probes confirmed expression profiles observed in the microarray experiment (Figure S1C).

Based on this analysis, we hypothesized that ESCs differentiate into cell types of all three germ layers after LIF withdrawal but preferentially differentiate toward ectoderm after exposure to RA. To further test this, we analyzed the expression of the upregulated genes in 55 adult tissues (Zhang et al., 2004) (Figure 3). After LIF withdrawal,

there was an upregulation of genes expressed in tissues of each germ layer (11% developing embryo, 4.7% ectoderm, 4.3% mesoderm, and 26% endoderm and mixed origin). In comparison, genes upregulated after RA exposure were predominately expressed in the developing embryo or ectoderm (14% developing embryo, 18% Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 73

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

74 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

ectoderm, and 11% endoderm and mixed origin). These tissue-specific genes are regulated very early in ESC commitment and potentially required for commitment to a specific lineage; thus, we suggest that overexpression of these genes may be a more efficient method of driving differentiation toward a specific lineage than using later markers of these cell types. We further selected only those genes exhibiting tight correlation between sorted cell populations and created a temporal cascade of genes regulated during early commitment by sorting according to the first day upon which they showed altered expression (Figure 4 and Figures S4 and S5). We predicted that genes critical in maintaining pluripotency must be commonly downregulated, and here we show that 74% of the genes downregulated after LIF withdrawal were also downregulated after RA addition (Figure 4B). These genes include Oct4, Nanog, and Sox2, previously identified targets of OCT4—Fgf4, Utfl, Fbxo15, Rex1 (Zfp42), and Foxd3 (Yuan et al., 1995, Nichols et al., 1998, Nishimoto et al., 2005, Tokuzawa et al., 2003)— transcriptional repressors, histone acetyltransferases, DNA and histone methyltransferases, and PcG genes. As expected, we did not observe as significant a number of commonly upregulated genes (Figures S4A, S4B, and S5). Next, we incorporated ChIP-chip data sets describing promoter occupancy by OCT4, NANOG, and SOX2 (Boyer et al., 2005; Loh et al., 2006). We considered genes identified by either group as valid potential targets. We found that the following were bound by some combination of OCT4, NANOG, and SOX2: (1) 20% of genes downregulated after LIF withdrawal, (2) 23% of genes downregulated after RA addition, (3) 34% of genes common to both time courses, (4) 28% of the genes upregulated after LIF withdrawal, (5) 29% of genes upregulated after RA addition, and (6) only 7.2% of the unregulated control group (Figure S3C). Thus, our screen, designed to identify key pluripotency genes, also systematically enriched for genes bound by, and thus potentially regulated by, OCT4, SOX2, and NANOG. These analyses support a model of stem cell maintenance in which highly expressed transcription factors and chromatin remodeling proteins maintain the stem cell in an undifferentiated state by repressing specific developmental programs. Upon initiation of commitment, these transcription factors are downregulated, activating specific developmental pathways and allowing the cell to commit to a particular fate. Oct4 and Sox2 shRNA Knockdowns Support New Predicted Targets of Pluripotency We postulated that true targets of OCT4, NANOG, and SOX2 would be regulated in both the time course and after

perturbation of Oct4 or Sox2 by shRNA knockdown. To test this, we performed microarray analyses on clonally derived shRNA-knockdown cell lines (Kunath et al., 2003) for both Oct4 and Sox2. Presumably because of the importance of these genes in maintaining the viability of ESCs, the best knockdowns maintained 41% of normal Sox2 and 55% of normal Oct4 mRNA expression (Figure 5A). We suggest that clones experiencing a greater knockdown either differentiated immediately or were not viable and thus did not survive expansion. As predicted, 95% of genes both commonly downregulated and OCT4 bound were downregulated in the Oct4 knockdown (compare Figures 4 and 5B). Ninety percent of genes both OCT4 bound and upregulated in the LIF time course were upregulated in the Oct4 knockdown, and 100% of genes commonly upregulated and OCT4 bound were upregulated in the Oct4 knockdown (compare Figure S4 and Figure 5B). Only 80% of OCT4-bound genes that were upregulated in the RA time course were upregulated in the Oct4 knockdown (data not shown), and we suggest that this inconsistency could be due to the effect of RA, to which the knockdowns were not exposed. In virtually all cases, genes regulated by OCT4 were identically affected by Sox2 knockdown (Figure 5B), supporting the model that OCT4 and SOX2 cooperate to control transcription and maintain stable levels of target genes. SOX2-bound genes were predictably regulated in the Sox2 knockdowns (compare Figure 4 and Figure S6), and because Nanog expression was reduced in both Oct4 and Sox2 knockdowns, we successfully predicted that NANOG-bound genes regulated in our time courses were similarly regulated in both the Oct4 and Sox2 knockdowns (Figure 5C). Thus, we have compiled a set of 110 predicted targets of pluripotency that are bound by OCT4, NANOG, or SOX2, regulated during commitment, and similarly regulated upon knockdown of Oct4 and Sox2. Functional Analysis of Candidate Regulators of Stem Cell Fate Confirms Their Influence on Pluripotency To test the effect of regulated genes on pluripotency, we generated shRNA cell lines for genes both down- and upregulated in the time courses and functionally analyzed them in a novel screen for self-renewal. We expected that following electroporation of the shRNA plasmid, severe knockdowns of genes essential for self-renewal would be nonviable or would simply differentiate, resulting in the formation of fewer healthy, undifferentiated colonies. We stained colonies with alkaline phosphatase (ALP), a marker of undifferentiated ESCs (Pease et al., 1990), and categorized them as undifferentiated, partially differentiated, or differentiated. Control colonies were generated by electroporation with the shRNA plasmid targeting

Figure 2. GO Analysis Revealed a Repressive Model of ESC Maintenance Down (A and B) and upregulated (C and D) gene clusters, determined by K means clustering of all regulated genes (Saeed et al., 2003) were entered separately into the GoStat analysis software (Beissbarth and Speed, 2004). Statistically overrepresented GO terms were determined by comparing the incidence of a GO term within the input cluster (observed, blue bar) to the incidence of that GO term among the entire mouse genome recorded in the MGI database (expected, yellow bar). Results are reported as the percentage of a particular GO term in the input cluster (blue) or the percentage of that GO term in the MGI database (yellow). Fisher’s Exact Test was used to determine a p value for each term.

Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 75

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

Figure 3. Differentiation Conditions Determine ESC Fate Decisions The expression of upregulated probes was examined in each of 55 adult mouse tissues (Zhang et al., 2004). It should be noted that expression data for all upregulated genes were not available. Tissue hybridizations were organized into the three germ layers, embryonic origin, and mixed source. Clustering analysis was performed to identify which genes from adult tissues, if any, were being upregulated under both differentiation conditions. The percentage of the genes in each of these clusters out of the total number of genes for which tissue data was available is reported in the text.

Rasgap, a gene not expressed in early development (Figures 6A–6C) (Kunath et al., 2003). Control colonies and knockdowns of upregulated genes were primarily undifferentiated (Figures 6B and 6C and Figure S7), whereas knockdowns of downregulated genes were partially or fully differentiated (Figures 6A and 6C and Figure S7). 76 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

ALP staining and morphology characterization offered a qualitative assessment of the state of differentiation of each knockdown. However, because these cell lines were successfully maintained, they clearly retained partial ability to self-renew. We developed a quantitative assay that allowed us to monitor self-renewal and commitment kinetics that could not be observed with ALP staining

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

Figure 4. Commonly Downregulated Genes Were Involved in Transcription and Chromatin Remodeling (A) Commonly downregulated genes were organized by general function and by promoter occupancy. Each gene found to be commonly regulated was compared to ChIP-chip data for Oct4, Nanog, and Sox2. Each gene is color coded according to the legend above the table; depending on which combination of OCT4/NANOG/SOX2 occupies its promoter. (B) Venn diagram illustrating the overlap of consistently downregulated genes from both time courses. See Figure S2 for a breakdown of consistently regulated genes.

Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 77

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

Figure 5. shRNA Knockdowns of Oct4 and Sox2 Confirmed Predicted Targets of OCT4, SOX2, and NANOG (A) qPCR was used to measure extent of knockdown in Oct4 and Sox2 shRNA cell lines. (B) Microarrays on Oct4 and Sox2 knockdown cell lines show that the majority of genes bound by OCT4 and regulated during differentiation are similarly regulated in both the Oct4 and Sox2 knockdowns.

78 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

alone. This novel assay employed automated fluorescence microscopy and single-cell, high-content imaging analysis to measure changes in OCT4 protein expression (Davey and Zandstra, 2006). We postulated that knockdowns of genes critical for self-renewal would differentiate and lose OCT4 protein expression, even in supportive (+LIF) conditions, and would lose OCT4 protein more rapidly than controls in LIF conditions, whereas knockdowns of genes upregulated during differentiation would display OCT4 expression identical to the control. As such, we used our assay to measure the increase in the percentage of cells that had lost OCT4 expression (percentage of differentiated cells) after 3, 24, 48, and 72 hr in +LIF or LIF conditions. We first analyzed Oct4 knockdowns with varying levels of mRNA reduction (Figure 6D). Oct4 knockdown resulted in a rapid loss of OCT4 protein in both +LIF and LIF conditions over 72 hr (Figure 6E) as well as decreased expression of ALP when cultured in +LIF conditions (Figure 6F). In general, knockdowns of additional downregulated genes also showed rapid loss of OCT4 expression over 72 hr, whereas knockdowns of upregulated genes did not (Figure 6G). Knockdowns of Fgf4 did not result in impaired self-renewal (Figures 6A and 6G), which was expected because FGF4 functions noncell autonomously in the inner cell mass to induce proliferation of trophectoderm cells (Tanaka et al., 1998). Because of the parameters of the screen, we only validated Oct4 and Fgf4 knockdowns by using more than one target siRNA sequence and thus cannot rule out off-target effects as contributing to the other observed phenotypes. We did, however, use three clonally derived knockdown cell lines for each gene in each of the described assays. Based on our preliminary shRNA analyses, we decided to further explore the role of one downregulated gene, Mkrn1, in the maintenance of pluripotentiality. MKRN1, an E3 ubiquitin ligase, has no known association with stem cell regulation and was targeted because its downregulation in our time courses displayed exceptionally close correlation with Oct4 downregulation. Loss of Mkrn1 mRNA led to a corresponding loss in Oct4 mRNA, whereas an increase in Mkrn1 led to an increase in Oct4 (Figure 6H), suggesting that Oct4 expression could be regulated by MKRN1, either directly or indirectly, through MKRN1 regulation of self-renewal machinery. Because loss of MKRN1 induced differentiation, we hypothesized that enforced overexpression would maintain the undifferentiated state, even in culture conditions favoring differentiation. Five days after LIF withdrawal, Mkrn1 overexpression cell lines maintained high ALP expression, whereas the control had lost ALP expression by the third day (Figures 6I and 6J), confirming that these ESCs had a heightened ability to remain undifferentiated, even in the absence of LIF. Finally, Mkrn1 overexpressing cell lines were able to maintain OCT4 protein levels after LIF withdrawal, as compared to Mkrn1 knockdown cell lines (Figure 6G).

These data revealed that depletion of genes identified by our methodology severely impairs self-renewal, and thus, each of these genes is critical for maintaining the undifferentiated state of the ESC. Furthermore, forced overexpression of Mkrn1 is sufficient to maintain pluripotentiality. Predicting Novel Regulatory Networks To construct potential ESC regulatory networks, we combined our time course expression data with published ChIP-chip experiments for Oct4, Nanog, Sox2, Eed, and Phc1 (Boyer et al., 2005, 2006; Loh et al., 2006). The proximal promoter of Phc1 can be bound by both NANOG and OCT4, whereas the Eed enhancer can be bound by NANOG (Loh et al., 2006). Further, the time course data was used to establish that both Eed and Phc1 are downregulated immediately after the downregulation of Oct4 and Nanog. Thus, we predicted that, in undifferentiated ESCs, expression of Eed and Phc1 is maintained by OCT4 and NANOG. Figure 7A illustrates five general feed-forward networks and the temporal expression of each proposed member of these networks in our LIF withdrawal time course. Any gene bound by OCT4, NANOG, EED, or PHC1 and upregulated after their downregulation was shown as repressed by Oct4, Nanog, Eed, or Phc1 in undifferentiated ESCs. To test the plausibility of these relationships, we examined the microarrays of the Oct4 shRNA knockdown. As predicted by our networks, we observed that PHC1 was downregulated in the Oct4 knockdown, whereas 82% (23 out of 28) of ‘‘gene a-e’’ tested showed upregulation (Figure 7B). This demonstrates that when Oct4, Nanog, and Phc1 expression is reduced, either during differentiation or directly by targeted inhibition of Oct4, the repression of the target genes is relaxed and their expression is upregulated. Finally, to further test the hypothesized networks, we perturbed Eed and Phc1 with shRNA and measured the expression of four hypothesized ‘‘gene a’’ targets: Hand1, Gata6, Eomes, and Hoxb1. The expression of these markers was measured by qPCR in two Eed knockdown clones expressing an average of 41% of wild-type Eed expression and two Phc1 clones expressing an average of 27% of wild-type Phc1 expression. As predicted by the hypothesized networks, expression of all four genes was upregulated in the Eed and Phc1 knockdown clones. In fact, a dramatic increase in mRNA copy number ranging from ten times Eomes levels to almost 80 times Hand1 levels was observed in both knockdowns. To ensure that the effect was due to the knockdown of Eed or Phc1 and not to the global differentiation of the cells, we measured Oct4 expression in the knockdowns and found that it was maintained at 80% of wild-type level (data not shown). In addition, as predicted by the hypothesized networks, microarray analysis of Nanog knockdown cells expressing 41% of wild-type Nanog measured by qPCR

(C) Nanog is downregulated in both the Oct4 and Sox2 knockdowns, and here we show that NANOG-bound genes are also predictably regulated in these knockdowns. Gene expression levels were in comparison to the control Rasgap shRNA cell line.

Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 79

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

Figure 6. High-Throughput siRNA Screen Confirmed Target Influence on Pluripotency (A and B) Fifteen million ESCs were electroporated with a linearized shRNA vector encoding Rasgap shRNA (control) or shRNA targeting the indicated gene. After 7 days of selection, colonies were stained for alkaline phosphatase (ALP) expression. All colonies were imaged on a light microscope at 103 magnification. (A) Colonies depleted of downregulated genes have decreased ALP expression. (B) Colonies depleted of upregulated genes show no phenotype.

80 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

(data not shown) showed upregulation of Hand1, Gata6, and Hoxb1 (Figure 7D). Upregulation of Eomes was not observed nor expected because Nanog knockdowns have been previously shown to preferentially differentiate toward primitive endoderm, not trophectoderm (Mitsui et al., 2003). Thus, by combining expression arrays and ChIP-chip data, we found that genes in Figure 7A are (1) bound by OCT4, NANOG, EED, or PHC1 and (2) upregulated immediately after changes in Oct4, Nanog, Eed, or Phc1 expression, either during differentiation or after knockdown. These networks implicate PcG proteins, EED, and PHC1 in cooperative repressive mechanisms with OCT4 and NANOG. These relationships suggest feed-forward loops that can function to reduce noise in a transcriptional network, in this case, by providing a repressive signal through multiple means and resulting in continuous repression of specific developmental regulators and stabilization of the ESC in the undifferentiated state, despite fluctuations in one of the repressive factors. The relatively few genes that fit these networks are primarily transcription factors recognized to be key regulators of ESC fate (Figure 7A). For example, Eomes is required for the development of the trophectoderm lineage (Russ et al., 2000), Gata6 for primitive endoderm development (Fujikura et al., 2002), Gsc, Bmp7, and Hand1 for mesoderm development (Tada et al., 2005; Zakin et al., 2005; McFadden et al., 2005), and Evx1, Hoxb1, and Dll1 are important for ectoderm development (Dush and Martin, 1992; Beckers et al., 1999) (Figure 7A). All above listed genes (gene a) are bound by OCT4, NANOG, EED, and PHC1, suggesting extremely tight restriction of expression that can only be completely released when all of these factors are downregulated. Thus, we have found that OCT4 and NANOG, in cooperation with Polycomb group proteins, EED, and PHC1, maintain direct repressive control of key regulators of all three germ layers, as well as the trophectoderm lineage, through interaction with specific target genes. Target

genes are mainly transcription factors known to be involved in early development, but we also predict the importance of several unannotated or previously unstudied genes. We suggest that upregulation of these specific target genes could be the impetus for the transcriptional cascades describing each germ layer. DISCUSSION Various genomic strategies have been previously used to identify key regulators of stem cells, including ESCs. Microarray studies were used to identify genes that are commonly expressed among different stem cell populations or that are dynamically expressed throughout differentiation (Ivanova et al., 2002, 2003, 2006; Ramalho-Santos et al., 2002; Sato et al., 2003; Fortunel et al., 2003). Genes expressed in ESCs or those dynamically regulated throughout differentiation represent possible targets of ESC transcription factors. However, physiologically irrelevant changes in gene expression can lead to false-positive target predictions and the fact that multiple factors influence changes in any given gene can lead to false-negative target predictions. Another strategy for target identification uses ChIP-chip analysis to discover possible genomic binding locations of known transcriptional regulators. In ESCs, this strategy was used to predict targets of OCT4, NANOG, and SOX2 (Boyer et al., 2005; Loh et al., 2006). However, ChIP-chip observed binding of promoters does not guarantee a regulatory relationship (Chua et al., 2004) and ChIP-chip alone fails to identify a number of functional targets of transcription factors (Chua et al., 2006). Though neither expression nor transcription factor binding studies in isolation are sufficient to establish a regulatory relationship between a transcription factor and its targets, integrating these methodologies provides two independent sources of evidence for predicting the regulation of a gene, greatly increasing confidence in a predicted interaction. This strategy has previously been successful

(C) The distribution of colonies on each plate that were undifferentiated, partially differentiated, or fully differentiated. The results are an average of at least two plates. Error bars represent the standard deviation of triplicate experiments. (D) Oct4 mRNA was measured in individual Oct4-shRNA clones by qPCR. Results are reported as the percentage of Oct4 expression measured in the control cell line. Error bars represent the standard deviation of triplicate experiments. (E) OCT4 protein was measured in 10,000 individual cells of three different Oct4-shRNA knockdown clones over 72 hr in both +LIF and LIF conditions. The percentage of cells that had lost OCT4 expression at each time point is reported. Error bars represent the standard deviation of triplicate experiments. (F) ALP expression of cultures of six different Oct4-shRNA clones in +LIF conditions. Error bars represent the standard deviation of triplicate experiments. (G) Two or three shRNA-knockdown clones per target gene were seeded at a low density, and OCT4 protein expression was measured in 10,000 individual cells at 3 hr and at 72 hr in both the presence and absence of LIF. Results are reported as the fold increase in the percentage of OCT4negative cells between 3 and 72 hr. Each experiment was performed with its own control, so to normalize between experiments, the fold increases were divided by the fold increase for each specific control. Mkrn1 overexpressing cell lines showed an ability to maintain OCT4 protein levels while Mkrn1 knockdowns lost OCT4 protein more rapidly. Error bars in the right panel represent the standard deviation of triplicate experiments; error bars in the left panel represent the standard deviation of three individual shRNA clones. (H) Mkrn1 shRNA-knockdown clones showed loss in Oct4 mRNA that closely tracks loss in Mkrn1 mRNA, as measured by qPCR. Likewise, Mkrn1 overexpressing cells showed an increase in Oct4 mRNA. (I) Mkrn1 overexpression clones were cultured after the withdrawal of LIF for 5 days alongside an empty-vector control and stained for ALP expression. Overexpression clones maintained high levels of ALP. (J) Mkrn1 overexpression maintains the majority of ESC colonies in the undifferentiated state, even in the absence of LIF, as measured by the percentage of the area of a culture of cells which maintained ALP expression.

Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 81

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

82 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

in the budding yeast Saccharomyces cerevisiae (Gao et al., 2004), and now, we are applying this integrative approach, pioneered in yeast systems biology, to a mammalian stem cell system. In this study, greater than 90% of the genes up- or downregulated in the time courses and bound by OCT4 were similarly up- or downregulated after Oct4 and Sox2 knockdowns, revealing the efficacy of our integrative analyses and the ability of our data to successfully predict functional regulation. In addition, 82% of the networks predicted by combining expression and ChIP-chip data were supported by genome-wide expression analysis of Oct4 knockdown cells. Genes identified as dynamically regulated during the loss of stem cell self-renewal were the foundation of these network studies. We argue that the method of differentiation and the time points chosen for analysis are critical factors to ensure the authenticity of the results and the ability to infer function in stem cells or differentiating cells. We studied ESCs in a monolayer culture, instead of forming traditional embryoid bodies (EBs), eliminating gene expression changes due to cell-cell interactions within the EB and other microenvironmental changes. Furthermore, monolayer culture differentiates much more slowly than EBs, thus we could focus on the initial changes in gene expression associated with commitment, rather than comparing undifferentiated ESCs to cells within EBs undergoing epiblast-like differentiation. Unlike a previous time course study (Ivanova et al., 2006), we analyzed homogenous cell populations by FACS sorting our cells for Oct4: eGFP and discarding cells expressing Oct4 message below a certain threshold. We eliminated genes directly regulated by RA by designing two separate time courses of ESC commitment and hypothesized that genes required to maintain pluripotency would be downregulated in both experiments. Our LIF time course also enabled us to observe subtle differences in timing of gene expression change missed by previous studies using only an accelerated RA differentiation (Ivanova et al., 2006). Instead of imposing an arbitrary fold-change cut-off, we analyzed all genes consistently regulated among the three samples arrayed at each time point. This is because Oct4, Sox2, and Nanog demonstrated slight, but very consistent, changes in expression, and we sought to identify all genes with similar expression kinetics. We argue that by precisely defining this biological event, we have successfully developed a model describing the transcriptional events occurring during ESC commitment. Establishing the precise timing of gene expres-

sion changes enabled us to distinguish potential primary targets of OCT4, NANOG, and SOX2, which responded immediately, from secondary targets, which responded at later time points. The temporal expression was also critical in deriving the proposed networks as a gene’s regulation could only be reported if it were (1) bound by some activating gene and (2) its expression changed after the change in expression of that activating gene. Genome-wide approaches have revealed that key developmental genes that are silenced in ESCs but are eventually expressed in differentiated cells are targeted by various core PcG proteins and are marked by a modified chromatin structure in ESCs (Bracken et al., 2006; Boyer et al., 2006; Bernstein et al., 2006). In addition, Farnham and colleagues postulated that OCT4 and the PcG protein SUZ12 cooperate to repress developmental targets, based on their observation that some targets of OCT4 were also bound by SUZ12 (Squazzo et al., 2006). Although the Suz12 enhancer was shown to be bound by OCT4 (Loh et al., 2006), we did not observe downregulation of Suz12 in our time courses. We suggest that PcG proteins that are dynamically regulated after the withdrawal of LIF are important mediators of differentiation in ESCs. We found that Eed and Phc1 are downregulated at the onset of commitment, are bound by OCT4 and NANOG, and that all four factors can target the promoters of a small set of genes, implicating them as partners in strengthening repression of developmentally associated genes. Our self-renewal screen is validated by the capture of Mybl2, which is required for inner cell mass formation (Tanaka et al., 1999), consistent with our observation that shRNA knockdown significantly impaired self-renewal. Our screen also identified that Mkrn1, previously unstudied in ESCs, plays a role in self-renewal. The Mkrn1 promoter was bound by both OCT4 and NANOG in undifferentiated ESCs, and Mkrn1 knockdowns exhibited decreased Oct4 and ALP expression, whereas Oct4 knockdowns exhibited decreased Mkrn1 expression. This reciprocal regulation suggests the presence of additional feedback loops controlling pluripotency in which targets of OCT4 act to regulate Oct4 expression. Several other genes captured in this screen have been implicated in controlling either pluripotency, self-renewal, or commitment, confirming the significance of our data for studying ESC transcriptional networks. Examples include (1) the identification of Fbxo15 as a target of OCT4 (Tokuzawa et al., 2003), (2) the requirement of Arid3b for the survival of neural crest cells during embryogenesis (Kobayashi et al., 2006), and (3) the role of Hmgb3 in

Figure 7. Polycomb Genes EED and PHC1 Cooperate with OCT4 and NANOG to Repress Activators of Key Developmental Pathways (A) ChIP-chip data were integrated with expression time course data to identify scenarios in which genes were both bound by OCT4, NANOG, PHC1, or EED and also regulated during commitment. Genes connected by an arrow were activated by the initiating gene, whereas genes connected by a line were repressed. The cluster diagram shows the temporal expression of each gene proposed to be involved in these pathways. (B) Microarray of Oct4-shRNA knockdown revealed that genes predicted to be activated by OCT4 were downregulated in the knockdown, whereas genes predicted to be repressed were upregulated. (C) qPCR revealed that all four ‘‘gene a’’ targets from (A) analyzed were upregulated in the knockdowns of both Eed and Phc1. Error bars represent the standard deviation of two individual shRNA clones. (D) Three gene a targets from (A) were upregulated in the Nanog knockdown.

Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 83

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

maintaining the self-renewal properties of hematopoietic stem cells (Nemeth et al., 2006). In each case, our temporal data support these findings and integration of binding data either supports the noted mechanism or suggests an unpredicted mechanism of regulation in ESCs (Figure 4 and Figures S4 and S5). The network modules presented here represent novel predicted nodes within the comprehensive transcriptional network governing ESC maintenance and differentiation. Our long-term goal is to draft an extensive network regulating ESC fate, which will be validated by various experimental systems. We also plan to test relevant network modules in the guidance of ESC commitment to target stem and progenitor lineages as well as test these networks in tissue-resident stem cell lineages. Therefore, to further validate the importance of the genes and networks identified in this study, we are utilizing the gene trap resources generated by our lab and others (To et al., 2004; Nord et al., 2006). Analysis of gene trap mutant ESCs and mice derived from them will enable us to test our predictions in ESCs, in the development of tissue-resident stem cells in the embryo, and in long-term maintenance of somatic stem cell populations (Stanford et al., 2001). Data obtained from ChIP-chip experiments alone provide insufficient evidence of a gene’s role in a process. However, by combining binding data with expression profiles of genes over a meticulously defined event, we have been able to predict the direct functional interactions of previously described binding scenarios. In this way, we have been able to construct a number of regulatory pathways controlling important developmental activators. In addition, we have formulated specific testable hypotheses regarding the response of candidate genes to genetic manipulation to confirm the relevance of our analyses. This study demonstrates the tremendous potential of integrating temporal expression experiments with ChIP-chip data to identify genes that are functionally interacting during a defined biological event. Our approach represents a powerful tool that will aid in unraveling the complex transcriptional cascades responsible for defining ESC fate, which can be used as a model to apply to other less tractable stem cell lineages.

Time Course of ES Cell Differentiation Oct4:eGFP ESCs were plated on 10 cm tissue culture-treated dishes (Falcon) coated with 0.1% gelatin in differentiation media at a density of 0.5 3 106 cells/dish for 5 days of differentiation, 106 cells/dish for 3 days of differentiation, 2 3 106 cells/dish for 2 days of differentiation, or 3.5 3 106 cells/dish for 1 day of differentiation. Control cells were harvested 2 days after plating in +LIF media. Differentiating cells were FACS sorted by eGFP expression after trypsinization and resuspension in 2% FBS in PBS at a dilution of 8 3 106 cells/mL. Microarray Hybridizations For the time courses of differentiation, total RNA was extracted with Trizol (Invitrogen) and used to perform two cycles of standard cDNA synthesis and in vitro transcription. The protocol was modified from the Affymetrix GeneChip technical note entitled ‘‘Eukaryotic Small Sample Target Labeling.’’ Five-hundred nanograms of cRNA was used for the second cycle of amplification and 1 mg of purified cDNA from the second cycle was used for biotin labeling. It was determined that this resulted in optimized linearity of amplification. Amplified and biotin-labeled cRNA was hybridized to both the Affymetrix GeneChips MG_U74av2 and MG_U74bv2. The normalized data obtained from MAS5.0 analysis of the Affymetrix MGU74a and MGU74b chips of the 16 time points described above were analyzed as outlined in Figure S2. For the shRNA-knockdown cell lines, microarray analysis was performed as described in Zhang et al. (2004). siRNA and Overexpression Vector Design Twenty-one base-pair siRNA sequences homologous to mRNA for each target gene were generated with online siRNA design tools (Qiagen, Dharmacon, and Ambion). Sequences were chosen that were further than 100 bases away from both the start and termination codons, had 50% GC content, no more than three successive G or Cs or four successive A or Ts, and were not homologous to any other murine gene, as determined by a BLAST search. Custom-designed siRNA oligonucleotides were synthesized by Invitrogen. Sequences and vector construction are described in the Supplemental Experimental Procedures. ESC Electroporation and Knockdown Clone Selection Twenty-five microliters of 1 mg/mL linearized plasmid DNA was added to 15 3 106 cells in one electroporation cuvette (VWR Scientific, 47727644). Cells were electroporated with 250 V using the GenePulser XCell (Biorad). Cells were put on ice for 10 min, into warmed media for 20 min, and plated onto two 0.1% gelatin-coated 10 cm TCP dishes. Selection media was added 24 hr after electroporation and was changed daily. After 7 days, 96 single colonies were picked and maintained by splitting 1:3 onto feeders in selection media. RNA was extracted from 24 single colonies, and qPCR was performed to determine the extent of knockdown of each clone. Successful knockdowns showing at least 60% knockdown (data not shown) were expanded individually, and at least three were used in each assay.

EXPERIMENTAL PROCEDURES ES Cell Culture R1 ESCs and Oct4:eGFP ESCs (Viswanathan et al., 2003) were cultured at 37 C and 5% CO2, on a layer of mitomycin-treated embryonic fibroblasts (MEFs) in ESC media consisting of Dulbecco’s modified Eagle medium (DMEM) supplemented with 15% FBS (North Bio, Lot SF30408), 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 2 mM L-glutamine (all from Gibco), 1000 U/mL leukemia inhibitory factor (LIF) (ESGRO, from Chemicon, batch 11061065), and 100 mM bmercaptoethanol (Sigma). Differentiation media consisted of either ESC media without LIF or ESC media without LIF and supplemented with 0.1 mM retinoic acid. Selection media consisted of ESC media supplemented with 150 mg/mL G418 (Gibco). ESCs were passaged every 2 days at a ratio of 1:5 by washing with PBS (Gibco), dissociating with 0.05% trypsin (Gibco) for 5 min at 37 C, and resuspending in ESC media. Media were changed daily.

84 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

ALP Staining Clones were fixed in neutral formalin buffer containing 3.8% formalin (Sigma) for 45 min and washed three times with PBS. Staining solution contained naphthol AS-MX phosphate (Sigma), N,N-dimethylformamide (Sigma), and fast red violet LB salt (Sigma) dissolved in 0.2 M Tris-HCl and was filtered through Whatman’s paper immediately before use. Clones were incubated with staining solution for 1 hr before being washed three times with PBS and stored at 4 C. Cells were imaged with a Leica DC200 light microscope and Leica IM50 V1.20 digital camera and software. Quantitative Real-Time PCR Total RNA was isolated by using the RNeasy Mini kit (Qiagen) and then treated with DNase (DNAfree kit, Ambion). RNA (1 mg) was reverse transcribed with SuperScript II RNase H Reverse Transcriptase (Invitrogen) with oligo(dT)23 primers (Sigma). Quantitative real-time PCR

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

was performed with genomic DNA as a universal external standard, as described in Yun et al. (2006). Measured transcript levels were normalized to both b-actin and elongation factor and compared to a control, untreated sample. Samples were run in triplicate. Primers are listed in the Supplemental Experimental Procedures. Quantitative Image Analysis of ESC Self-Renewal Cells were plated in a 96-well plate (6005182; Packard) coated with a fibronectin/gelatin mixture (12.5 mg/ml fibronectin; F1141; Sigma-Aldrich, 0.02% gelatin) at a density of 12,000 cells/well for the 3 hr time point and 6000 cells/well for the 24, 48, and 72 hr time points. Cells were cultured in DMEM with 15% knockout-serum replacement (10828-028; Invitrogen) in both LIF and +LIF (ESGRO, Chemicon; ESG1106) conditions. All cells were plated in +LIF conditions, and media were changed to LIF after 3 hr. Each cell line was plated in triplicate. Antibody staining for OCT4 was performed as described in Viswanathan et al. (2003). Cells were imaged by using the ArrayScan automated fluorescent microscope (Cellomics). Average pixel intensity of OCT4-Alexa Fluor 546 fluorophore within the nuclear area (as defined by Hoechst staining) of individual cells was determined. Ten-thousand individual cells were imaged, and the percentage of OCT4-negative cells was determined. Supplemental Data Supplemental Data include Supplemental Experimental Procedures and seven figures and can be found with this article online at http:// ACKNOWLEDGMENTS We thank Drs. Gary Bader, Jim Dennis, James Ellis, Caryn Ito, Josh Brickman, and Janet Rossant for critically reading the manuscript and Erin Frohwerk, Heidi Au, Sarah Kwon, Brett Runnalls, Qing Lan, and Tammy Reid for technical assistance. This work was funded by grants from the Canadian Institutes of Health Research to W.L.S., P.W.Z., and T.R.H.; Premier’s Research Excellence Award to W.L.S.; and Genome Canada to W.L.S. W.L.S., P.W.Z., and T.R.H. are supported by Canada Research Chairs. Received: December 6, 2006 Revised: March 16, 2007 Accepted: April 19, 2007 Published: June 6, 2007

comb complexes repress developmental regulators in murine embryonic stem cells. Nature 441, 349–353. Bracken, A.P., Dietrich, N., Pasini, D., Hansen, K.H., and Helin, K. (2006). Genome-wide mapping of polycomb target genes unravels their roles in cell fate transitions. Genes Dev. 20, 1123–1136. Chambers, I., Colby, D., Robertson, M., Nichols, J., Lee, S., Tweedie, S., and Smith, A. (2003). Functional expression cloning of Nanog, a pluripotency sustaining factor in embryonic stem cells. Cell 113, 643–655. Chua, G., Robinson, M.D., Morris, Q., and Hughes, T.R. (2004). Transcriptional networks: reverse-engineering gene regulation on a global scale. Curr. Opin. Microbiol. 7, 638–646. Chua, G., Morris, Q.D., Sopko, R., Robinson, M.D., Ryan, O., Chan, E.T., Frey, B.J., Andrews, B.J., Boone, C., and Hughes, T.R. (2006). Identifying transcription factor functions and targets by phenotypic activation. Proc. Natl. Acad. Sci. USA 103, 12045–12050. Davey, R.E., and Zandstra, P.W. (2006). Spatial organization of embryonic stem cell responsiveness to gp130 ligands reveals an autoregulatory stem cell niche. Stem Cells 24, 2538–2548. Dush, M.K., and Martin, G.R. (1992). Analysis of mouse Evx genes: Evx-1 displays graded expression in the primitive streak. Dev. Biol. 151, 273–287. Evans, M.J., and Kaufman, M.H. (1981). Establishment in culture of pluripotential cells from mouse embryos. Nature 292, 154–156. Fortunel, N.O., Out, H.H., Ng, H.H., Chen, J., Mu, X., Chevassut, T., Li, X., Joseph, M., Bailey, C., Hatzfeld, J.A., et al. (2003). Comment on ’’ ‘Stemness’: transcriptional profiling of embryonic and adult stem cells’’ and ‘‘a stem cell molecular signature’’. Science 302, 393. Fujikura, J., Yamato, E., Yonemura, S., Hosoda, K., Masui, S., Nakao, K., Miyazaki Ji, J., and Niwa, H. (2002). Differentiation of embryonic stem cells is induced by GATA factors. Genes Dev. 16, 784–789. Gao, F., Foat, B.C., and Bussemaker, H.J. (2004). Defining transcriptional networks through integrative modeling of mRNA expression and transcription factor binding data. BMC Bioinformatics 18, 5:31. Ivanova, N.B., Dimos, J.T., Schaniel, C., Hackney, J.A., Moore, K.A., and Lemischka, I.R. (2002). A stem cell molecular signature. Science 298, 601–604. Ivanova, N.B., Dimos, J.T., Schaniel, C., Hackney, J.A., Moore, K.A., Ramalho-Santos, M., Yoon, S., Matsuzaki, Y., Mulligan, R.C., Melton, D.A., et al. (2003). Response to comments on ‘‘‘Stemness’: transcriptional profiling of embryonic and adult stem cells’’ and ‘‘a stem cell molecular signature. Science 302, 393.


Ivanova, N., Dobrin, R., Lu, R., Kotenko, I., Levorse, J., DeCoste, C., Schafer, X., Lun, Y., and Lemischka, I.R. (2006). Dissecting selfrenewal in stem cells with RNA interference. Nature 442, 533–538.

Ambrosetti, D.C., Basilico, C., and Dailey, L. (1997). Synergistic activation of the fibroblast growth factor 4 enhancer by Sox2 and Oct-3 depends on protein-protein interactions facilitated by a specific spatial arrangement of factor binding sites. Mol. Cell. Biol. 17, 6321–6329.

Kobayashi, K., Era, T., Takebe, A., Jakt, L.M., and Nishikawa, S. (2006). ARID3B Induces malignant transformation of mouse embryonic fibroblasts and is strongly associated with malignant neuroblastoma. Cancer Res. 66, 8331–8336.

Beckers, J., Clark, A., Wunsch, K., Hrabe De Angelis, M., and Gossler, A. (1999). Expression of the mouse Delta1 gene during organogenesis and fetal development. Mech. Dev. 84, 165–168.

Kunath, T., Gish, G., Lickert, H., Jones, N., Pawson, T., and Rossant, J. (2003). Transgenic RNA interference in ES cell-derived embryos recapitulates a genetic null phenotype. Nat. Biotechnol. 21, 559–561.

Beissbarth, T., and Speed, T.P. (2004). GOstat: Find statistically overrepresented Gene Ontologies within a group of genes. Bioinformatics 20, 1464–1465.

Loh, Y.H., Wu, Q., Chew, J.L., Vega, V.B., Zhang, W., Chen, X., Bourque, G., George, J., Leong, B., Liu, J., et al. (2006). The Oct4 and Nanog transcription network regulates pluripotency in mouse embryonic stem cells. Nat. Genet. 38, 431–440.

Bernstein, B.E., Mikkelsen, T.S., Xie, X., Kamal, M., Huebert, D.J., Cuff, J., Fry, B., Meissner, A., Wernig, M., Plath, K., et al. (2006). A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell 125, 315–326.

Martin, G.R. (1981). Isolation of a pluripotent cell line from early mouse embryos cultured in medium conditioned by teratocarcinoma stem cells. Proc. Natl. Acad. Sci. USA 78, 7634–7638.

Boyer, L.A., Lee, T.I., Cole, M.F., Johnstone, S.E., Levine, S.S., Zucker, J.P., Guenther, M.G., Kumar, R.M., Murray, H.L., Jenner, R.G., et al. (2005). Core transcriptional regulatory circuitry in human embryonic stem cells. Cell 122, 947–956.

McFadden, D.G., Barbosa, A.C., Richardson, J.A., Schneider, M.D., Srivastava, D., and Olson, E.N. (2005). The Hand1 and Hand2 transcription factors regulate expansion of the embryonic cardiac ventricles in a gene dosage-dependent manner. Development 132, 189–201.

Boyer, L.A., Plath, K., Zeitlinger, J., Brambrink, T., Medeiros, L.A., Lee, T.I., Levine, S.S., Wernig, M., Tajonar, A., Ray, M.K., et al. (2006). Poly-

Mitsui, K., Tokuzawa, Y., Itoh, H., Segawa, K., Murakami, M., Takahashi, K., Maruyama, M., Maeda, M., and Yamanaka, S. (2003). The

Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc. 85

Cell Stem Cell Systems Biology Approach to Stem Cell Self-Renewal

homeoprotein Nanog is required for maintenance of pluripotency in mouse epiblast and ES cells. Cell 113, 631–642. Morrison, G.M., and Brickman, J.M. (2006). Conserved roles for Oct4 homologues in maintaining multipotency during early vertebrate development. Development 133, 2011–2022. Nemeth, M.J., Kirby, M.R., and Bodine, D.M. (2006). Hmgb3 regulates the balance between hematopoietic stem cell self-renewal and differentiation. Proc. Natl. Acad. Sci. USA 103, 13783–13788. Nichols, J., Zevnik, B., Anastassiadis, K., Niwa, H., Klewe-Nebenius, D., Chambers, I., Scholer, H., and Smith, A. (1998). Formation of pluripotent stem cells in the mammalian embryo depends on the POU transcription factor Oct4. Cell 95, 379–391. Nishimoto, M., Miyagi, S., Yamagishi, T., Sakaguchi, T., Niwa, H., Muramatsu, M., and Okuda, A. (2005). Oct-3/4 maintains the proliferative embryonic stem cell state via specific binding to a variant octamer sequence in the regulatory region of the UTF1 locus. Mol. Cell. Biol. 25, 5084–5094. Niwa, H., Miyazaki, J., and Smith, A.G. (2000). Quantitative expression of Oct-3/4 defines differentiation, dedifferentiation or self-renewal of ES cells. Nat. Genet. 24, 372–376. Nord, A.S., Chang, P.J., Conklin, B.R., Cox, A.V., Harper, C.A., Hicks, G.G., Huang, C.C., Johns, S.J., Kawamoto, M., Liu, S., et al. (2006). The International Gene Trap Consortium Website: a portal to all publicly available gene trap cell lines in mouse. Nucleic Acids Res. 34, D642–D648. Okumura-Nakanishi, S., Saito, M., Niwa, H., and Ishikawa, F. (2005). Oct-3/4 and Sox2 regulate Oct-3/4 gene in embryonic stem cells. J. Biol. Chem. 280, 5307–5317. Pease, S., Braghetta, P., Gearing, D., Grail, D., and Williams, R.L. (1990). Isolation of embryonic stem (ES) cells in media supplemented with recombinant leukemia inhibitory factor (LIF). Dev. Biol. 141, 344–352. Ramalho-Santos, M., Yoon, S., Matsuzaki, Y., Mulligan, R.C., and Melton, D.A. (2002). ‘‘Stemness’’: transcriptional profiling of embryonic and adult stem cells. Science 298, 597–600. Rodda, D.J., Chew, J.L., Lim, L.H., Loh, Y.H., Wang, B., Ng, H.H., and Robson, P. (2005). Transcriptional regulation of nanog by OCT4 and SOX2. J. Biol. Chem. 280, 24731–24737. Russ, A.P., Wattler, S., Colledge, W.H., Aparicio, S.A., Carlton, M.B., Pearce, J.J., Barton, S.C., Surani, M.A., Ryan, K., Nehls, M.C., et al. (2000). Eomesodermin is required for mouse trophoblast development and mesoderm formation. Nature 404, 95–99.

Smith, A.G., Heath, J.K., Donaldson, D.D., Wong, G.G., Moreau, J., Stahl, M., and Rogers, D. (1988). Inhibition of pluripotential embryonic stem cell differentiation by purified polypeptides. Nature 336, 688–690. Squazzo, S.L., O’Geen, H., Komashko, V.M., Krig, S.R., Jin, V.X., Jang, S.W., Margueron, R., Reinberg, D., Green, R., and Farnham, P.J. (2006). Suz12 binds to silenced regions of the genome in a cell-typespecific manner. Genome Res. 16, 890–900. Stanford, W.L., Cohn, J.B., and Cordes, S.P. (2001). Gene trap mutagenesis: past, present and beyond. Nat. Rev. Genet. 2, 756–768. Tada, S., Era, T., Furusawa, C., Sakurai, H., Nishikawa, S., Kinoshita, M., Nakao, K., Chiba, T., and Nishikawa, S. (2005). Characterization of mesendoderm: a diverging point of the definitive endoderm and mesoderm in embryonic stem cell differentiation culture. Development 132, 4363–4374. Tanaka, S., Kunath, T., Hadjantonakis, A.K., Nagy, A., and Rossant, J. (1998). Promotion of trophoblast stem cell proliferation by FGF4. Science 282, 2072–2075. Tanaka, Y., Patestos, N.P., Maekawa, T., and Ishii, S. (1999). B-myb is required for inner cell mass formation at an early stage of development. J. Biol. Chem. 274, 28067–28070. To, C., Epp, T., Reid, T., Lan, Q., Yu, M., Li, C.Y.J., Ohishi, M., Hant, P., Tsao, N., Casallo, G., et al. (2004). The centre for modeling human disease gene trap resource. Nucleic Acids Res. 32, D557–D559. Tokuzawa, Y., Kaiho, E., Maruyama, M., Takahashi, K., Mitsui, K., Maeda, M., Niwa, H., and Yamanaka, S. (2003). Fbx15 is a novel target of Oct3/4 but is dispensable for embryonic stem cell self-renewal and mouse development. Mol. Cell. Biol. 23, 2699–2708. Viswanathan, S., Benatar, T., Mileikovsky, M., Lauffenburger, D.A., Nagy, A., and Zandstra, P.W. (2003). Supplementation-dependent differences in the rates of embryonic stem cell self-renewal, differentiation, and apoptosis. Biotechnol. Bioeng. 84, 505–517. Williams, R.L., Hilton, D.J., Pease, S., Willson, T.A., Stewart, C.L., Gearing, D.P., Wagner, E.F., Metcalf, D., Nicola, N.A., and Gough, N.M. (1988). Myeloid leukaemia inhibitory factor maintains the developmental potential of embryonic stem cells. Nature 336, 684–687. Yuan, H., Corbi, N., Basilico, C., and Dailey, L. (1995). Developmentalspecific activity of the FGF-4 enhancer requires the synergistic action of Sox2 and Oct-3. Genes Dev. 9, 2635–2645. Yun, J.J., Heisler, L.E., Hwang, I.I., Wilkins, O., Lau, S.K., Hyrcza, M., Jayabalasingham, B., Jin, J., McLaurin, J., Tsao, M.S., et al. (2006). Genomic DNA functions as a universal external standard in quantitative real-time PCR. Nucleic Acids Res. 34, e85.

Saeed, A.I., Sharov, V., White, J., Li, J., Liang, W., Bhagabati, N., Braisted, J., Klapa, M., Currier, T., Thiagarajan, M., et al. (2003). TM4: a free, open-source system for microarray data management and analysis. Biotechniques 34, 374–378.

Zakin, L., Reversade, B., Kuroda, H., Lyons, K.M., and De Robertis, E.M. (2005). Sirenomelia in Bmp7 and Tsg compound mutant mice: requirement for Bmp signaling in the development of ventral posterior mesoderm. Development 132, 2489–2499.

Sato, N., Sanjuan, I.M., Heke, M., Uchida, M., Naef, F., and Brivanlou, A.H. (2003). Molecular signature of human embryonic stem cells and its comparison with the mouse. Dev. Biol. 260, 404–413.

Zhang, W., Morris, Q.D., Chang, R., Shai, O., Bakowski, M.A., Mitsakakis, N., Mohammad, N., Robinson, M.D., Zirngibl, R., Somogyi, E., et al. (2004). The functional landscape of mouse gene expression. J. Biol. 3, 21.

Schoorlemmer, J., van Puijenbroek, A., van Den Eijnden, M., Jonk, L., Pals, C., and Kruijer, W. (1994). Characterization of a negative retinoic acid response element in the murine Oct4 promoter. Mol. Cell. Biol. 14, 1122–1136. Smith, A.G., and Hooper, M.L. (1987). Buffalo rat liver cells produce a diffusible activity which inhibits the differentiation of murine embryonal carcinoma and embryonic stem cells. Dev. Biol. 121, 1–9.

86 Cell Stem Cell 1, 71–86, July 2007 ª2007 Elsevier Inc.

Accession Numbers The data discussed in this publication have been deposited in NCBIs Gene Expression Omnibus (GEO; and are accessible through GEO Series accession numbers GSE7506 (time course) and GSE7520 (Oct4 and Sox2 knockdowns).