SEARCH

SEARCH BY CITATION

Keywords:

  • Activin;
  • Smad2/3;
  • Embryonic stem cells;
  • Epiblast stem cells;
  • Induced pluripotent stem cells;
  • Pluripotency;
  • Endoderm;
  • Human;
  • Mouse

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Activin/Nodal signaling is necessary to maintain pluripotency of human embryonic stem cells (hESCs) and to induce their differentiation toward endoderm. However, the mechanisms by which Activin/Nodal signaling achieves these opposite functions remain unclear. To unravel these mechanisms, we examined the transcriptional network controlled in hESCs by Smad2 and Smad3, which represent the direct effectors of Activin/Nodal signaling. These analyses reveal that Smad2/3 participate in the control of the core transcriptional network characterizing pluripotency, which includes Oct-4, Nanog, FoxD3, Dppa4, Tert, Myc, and UTF1. In addition, similar experiments performed on endoderm cells confirm that a broad part of the transcriptional network directing differentiation is downstream of Smad2/3. Therefore, Activin/Nodal signaling appears to control divergent transcriptional networks in hESCs and in endoderm. Importantly, we observed an overlap between the transcriptional network downstream of Nanog and Smad2/3 in hESCs; whereas, functional studies showed that both factors cooperate to control the expression of pluripotency genes. Therefore, the effect of Activin/Nodal signaling on pluripotency and differentiation could be dictated by tissue specific Smad2/3 partners such as Nanog, explaining the mechanisms by which signaling pathways can orchestrate divergent cell fate decisions. STEM CELLS 2011;29:1176–1185


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Human pluripotent stem cells derived from embryos at the blastocyst stage (human embryonic stem cells or hESCs) have the capacity to differentiate into a broad number of cell types. This characteristic is maintained in hESCs [1, 2] by the Activin/Nodal signaling pathway that controls the expression of the pluripotency factor, Nanog, which in turn blocks the expression of neuroectoderm genes induced by FGF [1]. The apparent conservation of these mechanisms in pluripotent stem cells generated from reprogrammed somatic cells (human induced pluripotent stem cells or hIPSCs) [3] and in pluripotent stem cells derived from the epiblast of postimplantation mouse embryo (mouse epiblast stem cells or mEpiSCs) [4] suggests that hESCs, hIPSCs, and mEpiSCs could share the same pluripotent state. However, the transcriptional network downstream of Activin/Nodal signaling in pluripotent stem cells has never been described and thus the precise mechanisms by which this pathway maintains pluripotency remains unclear. Finally, Activin/Nodal signaling is also known to be required for definitive endoderm (DE) differentiation in vivo [5] and in vitro [6] and the mechanism by which the same signaling pathway can induce differentiation and maintain pluripotency requires further investigation.

Here, we address these questions by examining the transcriptional network controlled in hESCs and DE progenitors by Smad2 and Smad3 (Smad2/3), which represent the direct effectors of Activin/Nodal signaling [7]. These genome-wide analyses reveal for the first time that Smad2/3 could contribute to the control of a broad part of the transcriptional network characterizing the pluripotent state of hESCs. Similar genome-wide experiments performed in DE progenitors demonstrate that Smad2/3 proteins change genomic location upon differentiation while being necessary for the expression of a large number of endoderm markers, suggesting that Smad2/3 activity could be dictated by tissue specific partners. We subsequently identify Nanog as an essential Smad2/3 partner in hESCs by showing that Nanog and Smad2/3 cooperate to maintain the expression of essential pluripotency markers in hESCs. Taken together, these results represent a first step toward understanding the molecular mechanisms by which the Activin/Nodal signaling pathway interacts with pluripotency factors to control early cell fate decisions.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

hESCs, hIPSCs, mEpiSCs, and mESCs Culture Conditions

H9 hESCs (WiCell) were grown in chemically defined medium (CDM) as described previously [4, 8] and were induced to differentiate into endoderm using CDM-polyvinyl alcohol supplemented with Activin (100 ng/ml) + BMP4 (10 ng/ml) + FGF2 (20 ng/ml) + LY294002 (10 μM) [9]. Inhibition of Activin/Nodal signaling was achieved using CDM supplemented with SB-431542 (10 μM) + FGF2 12 ng/ml.

Luciferase Assays

The genomic regions corresponding to Smad2/3 binding sites in Sox17, Mixl1, GSC, and eomesodermin (Eomes) genes were isolated by polymerase chain reaction (PCR) (See Supporting Information Table 7 for primer sequence) and then sublcones into in the firefly luciferase reporter vector PGL3 (Promega, Madison, WI, http://www.promega.com). The resulting vectors were cotransfected with CMV-Renilla (Promega) into hESCs with Lipofectamine 2000 (Invitrogen, Grand Island, NY, http://www.invitrogen.com). Then, the cells were either grown in culture conditions compatible with pluripotency (CDM Activin + FGF2) or inductive for endoderm differentiation (CDM Activinhigh + BMP4 + FGF2 + Ly). The ratio between luciferase reporters and CMV-Renilla was 100:1. Medium was changed after 18 hours and cells were harvested every 24 hours for luciferase assay. Luciferase activity was measured with the dual luciferase assay kit following (Promega) manufacturer instructions. Firefly luciferase activity was normalized to Renilla luciferase activity for cell density and transfection efficiency.

ChIP and ChIP-XL, ChIP-Seq, and Bioinformatic Analyses

See Materials and Methods section in the Supporting Information.

Knockdown of Smad2/3 Expression in hESCs

H9 hESCs were transfected with expression vector for short hairpin RNAs (ShRNAs) specifically targeting Smad2/3 (Sigma, Saint Louis, www.sigmaaldrich.com, SHCLNG-NM_005901 and SHCLNG-NM_005902) using lipofectamine 2000 as described [10]. hESC colonies generated after 5 days of puromycin selection were immediately analyzed for the expression of Smad2/3 or pluripotency markers using Q-PCR. Alternatively, the resulting colonies were grown for 3 days in culture conditions inductive for endoderm differentiation and then the expression of endoderm markers was analyzed using Q-PCR.

FACS, Immunostaining, and Real Time-PCR

Methods for FACS, immunostaining analyses, and real time-PCR have been described elsewhere [9]. Q-PCR data are presented as the mean of three independent experiments and error bars indicate SD. Absence of statistically significant difference (p > .05, df = 4) in gene expression is marked by “X” in all the figures. Primers used for real time-PCR analyses are given in Supporting Information Table 7.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Specific Combinations of Chemical Cross-Linkers Improve the Efficiency of ChIP

To define the transcriptional network controlled by Smad2/3 in hESCs, we performed chromatin immunoprecipitation (ChIP) combined with real time-PCR (ChIP-Q-PCR) analyses to detect the presence of Smad2/3 proteins on known target genes of Activin/Nodal signaling such as Nanog, Lefty, Smad7, and Nodal. Despite testing seven antibodies, we systematically obtained weak enrichment (Supporting Information Fig. S1A and data not shown). Smad2/3 proteins are known to constantly shuttle between the cytoplasm and the nucleus [7], and therefore, Smad2/3 interactions with DNA could happen transiently, especially in cells with an unstable phenotype such as hESCs. Furthermore, Smad2 binds DNA through cofactors such as Smad4 [7] and formaldehyde, which is used to cross-link protein to DNA in conventional ChIP that might have a reduced capacity to capture such indirect interactions. To bypass these limitations, we used protein cross-linkers which are known to facilitate ChIP by stabilizing protein complexes [11] and thus more likely to capture Smad2/3 interactions with DNA and other cofactors. Importantly, the use of additional cross-linkers to improve ChIP has been validated previously on diverse transcription factors [11–13]. To define the effect of these chemicals on ChIP efficiency in hESCs, we performed Smad2/3 ChIP-Q-PCR analyses using a combination of diverse cross-linkers in hESCs (Supporting Information Fig. S1B). In sum, we observed that a combination of formaldehyde + dimethyl 3,3′-dithiopropionimidate dihydrochloride + 3,3′-dithiodipropionic acid di(N-hydroxysuccinimide ester) increased Smad2/3 bound DNA enrichment between 5- and 10-fold depending on the genomic region analyzed (Fig. 1A). Notably, the use of additional cross-linkers did not increase nonspecific background (Fig. 1A and Supporting Information Fig. S1B) and similar results were obtained with other transcription factors including Nanog and Eomes (Fig. 1A and data not shown). Collectively, these results demonstrate that this improved method, which we named ChIP-XL, increases the efficacy of conventional ChIP.

thumbnail image

Figure 1. Chromatin immunoprecipitation-XL coupled with deep sequencing (ChIP-XL-Seq) reveals the transcriptional network functioning downstream Activin/Nodal signaling in human embryonic stem cells (hESCs). (A): Protein cross-linkers significantly increase the efficiency of ChIP analysis in hESCs. (B): Relative position of the genomic regions bound by Smad2/3 to the transcriptional start site. (C): Smad2/3 binding regions in hESCs are evolutionarily conserved. The average conservation (phastCons, scaled 0–1) of the genomic regions bound by Smad2/3 was determined as described in Materials and Methods section. Matched nonspecific genomic regions were used as a negative control (red line). (D): Validation of Smad2/3 binding sites in hESCs. ChIP-Q-PCR were performed on hESCs to confirm the presence of Smad2/3 on the genomic regions identified using ChIP-XL-Seq. Results were normalized against control region (Smad73UTR) and are expressed as ±SD from three experiments. (E): Representative view of Smad2/3 binding sites in genomic regions containing pluripotency genes. Y-axis shows the number of sequence tags per 200 bp window. Arrows point to statistically significant enriched genomic regions. (F): Q-PCR analyses showing that inhibition of Activin/Nodal signaling in hESCs using SB-431542 blocks the expression of pluripotency genes bound by Smad2/3. Absence of statistically significant difference (p > .05, df = 4) in gene expression is marked by X. (G): Q-PCR analyses showing that knockdown of Smad2 and/or Smad3 expression decreases the expression of genes identified using ChIP-XL-Seq. Statistically significant differences (p < .05, df = 4) in gene expression is marked by *. Abbreviation: hESC, human embryonic stem cells.

Download figure to PowerPoint

ChIP-XL Combined with Deep Sequencing Enables the Identification of Smad2/3 Binding Sites in hESCs

Following these results, we proceeded to identify the target genes of Smad2/3 in hESCs using ChIP-XL coupled with deep sequencing (ChIP-XL-Seq). These experiments were performed on hESCs grown in a fully defined culture system [4], which is particularly advantageous for such analyses as it provides a homogenous population of pluripotent cells (Supporting Information Fig. S1C) in the absence of unknown factors that could interfere with experimental outcomes [8]. In sum, 14,143 genomic regions bound by Smad2/3 were detected using ChIP-XL (Supporting Information Table S1a) and 5,045 genes (as defined by unique HUGO gene nomenclature committee [HGNC] symbols) were located in the immediate proximity (between 10 kb upstream of transcriptional start site [TSS] and 2 kb downstream of TSS) of Smad2/3 binding sites (Supporting Information Table S1b). The number of Smad2/3 candidate target genes (5,045 over approximately 18,200 similarly annotated human genes) is within a normal range in the context of contemporary ChIP-Seq analyses (between 3,000 and 15,000, as shown in recent studies with different factors in diverse cell types [14–19]). Furthermore, several known target genes of Activin/Nodal signaling (Nodal, Lefty1/2, Smad7, Nanog, TDGF1, FoxH1, and SKIL) [7] were found to be bound by Smad2/3 on previously characterized binding sites (Supporting Information Fig. S1D and Table S1a–c) confirming the relevance of the binding events detected using ChIP-XL-Seq. Additional validations showed that binding events identified using genome-wide analysis could be detected by ChIP combined with Q-PCR carried out on genomic regions from across a range of fold enrichment (Supporting Information Table S1a and Fig. 1D). Importantly, such binding events were not detected in hESCs grown in the absence of Activin/Nodal signaling (Supporting Information Fig. S1E), further confirming that ChIP-XL only captures interactions between DNA and Smad2/3. Binding sites were also more frequently located in close proximity to the TSS of candidate target genes (Fig. 1B), while analysis of sequence conservation across species revealed that Smad2/3 bound regions were evolutionarily conserved above background (Fig. 1C). Finally, de novo motif inference analyses on genomic regions bound by Smad2/3 revealed binding motifs matching those of specific pluripotency factors, such as Oct-4 and KLF-4 (Supporting Information Fig. S1F). Together, these results exclude that Smad2/3 binding sites detected by ChIP-XL-Seq represent random interactions between Smad2/3 and genomic DNA provoked by the use of additional cross-linkers and demonstrate that this approach could be used efficiently to identify target genes of transcription factors, which are difficult to capture using conventional ChIP-Seq.

Activin/Nodal Signaling via Smad2/3 is Involved in the Control of the Core Transcriptional Network Characterizing Pluripotency in hESCs

Previous studies have established that binding events detected using ChIP-Seq experiments are not systematically associated with either transcriptional activation or repression [20]. Consequently, functional validations are necessary to demonstrate that the presence of a transcription factor on a specific genomic location modulates gene expression. To address this issue, we compared the gene expression profile of hESCs grown in CDM + Activin + FGF2 with that of hESCs grown for 36 hours in CDM + FGF2 + SB-431542, a chemical inhibitor of Activin receptors [21]. These analyses revealed that 346 genes bound by Smad2/3 in hESCs were exclusively down regulated in the absence of Activin/Nodal signaling and 447 genes were upregulated (Supporting Information Table S2a–c). Therefore, 15.7% (793 vs. 700 expected, p = 5.0 × 10−6) of Smad2/3 target genes identified by ChIP-XL-Seq appeared to rely on Activin/Nodal signaling to maintain or to repress their transcriptional activity. However, 59.9% of genes bound by Smad2/3 are expressed above background (Affymetrix signal detection call) in hESCs, suggesting that Activin/Nodal signaling could also participate in the transcriptional regulation of a large number of genes in hESCs without being necessary to maintain their expression. Importantly, similar analyses performed with Nanog, Oct-4, and Sox2 have shown that only 10%–25% of the genes bound by these pluripotency factors are expressed in mESCs [20], while only 5%–10% of their binding sites could be functional. In this context, these results suggest Smad2/3 involvement in the control of a large portion of the transcriptional network characterizing hESCs thus reinforcing the essential function of Activin/Nodal signaling in the mechanisms controlling pluripotency.

To gain further insights into the transcriptional network controlled by Activin/Nodal signaling in hESCs, we performed gene ontology (GO) analysis on candidate Smad2/3 target genes upregulated or downregulated in the absence of Activin/Nodal signaling in hESCs. These analyses showed a significant enrichment (p < .05) for genes associated with developmental processes, including stem cell maintenance and cell fate commitment, thereby reconfirming the important function of Activin/Nodal signaling in pluripotency (Supporting Information Fig. S1G). Further analyses identified key pluripotency factors, epigenetic regulators, self-renewal factors, and effectors of extracellular signaling pathways (Fig. 1E and Supporting Information Table S1c). We next examined the function of Smad2/3 in the transcriptional regulation of representative members for each category (Oct-4, Nanog, Dpp4, UTF1, FoxD3, Suz12, Myc, TERT, and Nodal) (Supporting Information Table S1c). We first validated the functionality of these binding events by inhibiting the Activin/Nodal signaling pathway in hESCs using SB-431542 (Fig. 1F). Absence of Activin/Nodal signaling for 24 hours resulted in a significant decrease in the expression of Oct-4, Nanog, Myc, TERT, Dppa4, UTF1, and Nodal. Prolonged treatment with SB-431542 completely abolished the expression of these genes, confirming that Activin/Nodal signaling is necessary to maintain their expression in hESCs. However, the expression of Suz12 and FoxD3 only decreased transiently upon inhibition of Activin/Nodal signaling, suggesting that Smad2/3 might only play a limited role in the transcriptional regulation of these two genes or that their expression is maintained in neuroectoderm cells generated upon SB-431542 treatment. To further validate these results, Smad2 and Smad3 were individually knocked down in hESCs using specific ShRNA. Q-PCR analyses revealed that reduced expression of Smad2 decreased the expression of Oct-4, Nanog, Dppa4, UTF1, FoxD3, Tert, and Nodal, whereas knockdown of Smad3 expression only decreased the expression of Suz12, Myc, and Nodal (Fig. 1G). These results imply that Activin/Nodal signaling could be involved in the control of a broad part of the transcriptional network characterizing the pluripotency state of hESCs and that Smad2 rather than Smad3 is necessary for the expression of pluripotency factors in hESCs.

Activin/Nodal Signaling is Involved in the Control of a Broad Part of the Transcriptional Network Directing Endoderm Differentiation

To further understand the mechanisms by which Activin/Nodal signaling controls early cell fate decisions in pluripotent stem cells, we identified the genes controlled by Activin/Nodal signaling during DE differentiation of hESCs using Smad2/3 ChIP-XL-Seq. For this, we took advantage of a fully defined culture system allowing the generation of a near homogenous population of DE cells expressing Eomes, GATA6, FoxA2, and Sox17 (Fig. 2A–2C and Supporting Information Fig. S2A) [6]. Smad2/3-ChIP-XL-Seq analyses were performed at an early stage of DE differentiation characterized by the upregulation of Sox17 (Fig. 2C). Using this approach, we were able to identify 15,486 Smad2/3 binding sites corresponding to 4,056 potential target genes for Activin/Nodal signaling in early DE cells (Supporting Information Table S3a, b). GO analyses showed a significant enrichment (p = .001) for genes associated with gastrulation (Supporting Information Fig. S2B). To further validate these data, ChIP-XL-Q-PCR analyses were performed to establish the presence of Smad2/3 on a representative panel of genomic regions (Supporting Information Fig. S2C). Analysis of sequence conservation between species revealed that Smad2/3 binding regions identified in endoderm cells were evolutionarily conserved above background (Supporting Information Fig. S2D). Furthermore, Smad2/3 binding sites were more frequently located in close proximity to the TSS (Supporting Information Fig. S2E). Notably, Smad2/3 binding sites were identified near known Primitive Streak (Eomes, Mixl1, FGF8, and Wnt3; Supporting Information Table S3) and DE (Sox17, CXCR4, GATA4, GATA6, FoxA2, Lhx1, and GSC; Supporting Information Table S3) genes [5]; whereas ChIP-Seq analyses did not reveal the presence of Smad2/3 on genes known to be expressed upon neuroectoderm (Sox1, Pax6, Sox3, Gbx2, and olig3) or mesoderm (Brachyury, Hand1, Mesp1, Mesp2, Tbx6, Scl, Hoxb1, MyoD, Myf5, and Flk-1). Finally, de novo motif inference analyses on genomic regions bound by Smad2/3 upon differentiation revealed binding motifs for transcription factors specifically expressed in endoderm cells such as FoxA1/2 and T/Eomes (Fig. 2D), reinforcing recent studies suggesting that Smad2/3 could cooperate with endodermal partners such as Eomes to promote differentiation [22].

thumbnail image

Figure 2. Activin/Nodal signaling pathway controls a broad part of the transcriptional network characterizing endoderm cells. (A): Expression of pluripotency markers (Oct-4, Nanog, and Sox2) and endoderm markers (Eomes, Sox17, GATA6, and FoxA2) in human embryonic stem cells (hESCs) and in their endodermal derivatives. Scale bar = 50 μM. (B): Heat map showing the expression of key endoderm genes bound by Smad2/3 in definitive endoderm (DE) cells grown in the presence (A) or in the absence of Activin signaling (SB-431542). (C): Q-PCR analyses showing that Activin/Nodal signaling controls the expression of the genes bound by Smad2/3 during endoderm differentiation. hESCs (H9) were grown for 3 days in culture conditions inductive for endoderm differentiation (day 1, day 2, and day 3). Alternatively, hESCs were grown in culture conditions inductive for endoderm differentiation for 2 days and then the resulting cells were grown for an additional day in the absence of Activin/Nodal signaling (D3 + SB-431542). Absence of statistically significant difference (p > .05, df = 4) in gene expression is marked by X. (D): Schematic representation of inferred transcription factor binding motifs in genomic regions bound by Smad2/3 in endoderm. Statistical analyses were performed as described in Supporting Information. (E): Q-PCR analyses showing that knockdown of Smad2 and/or Smad3 expression decreases the expression of endoderm markers in hESCs grown for 3 days in CDM-Acitivin, FGF2, BMP, Ly294002. Statistically significant differences (p < .05, df = 4) in gene expression are marked by *. (F): Luciferase assays showing the transcriptional activity associated with Smad2/3 binding sites located in key endoderm genes (Mixl1, goosecoid (GSC), and Sox17) in hESCs, upon differentiation and in differentiating DE cells grown for 24 hours in the presence of SB-431542 (DED3 + SB-431542). Top panels show schematic representation of the corresponding genes. Y-axis shows the number of sequence tags per 200 bp window. Arrows point to statistically significant enriched genomic regions. Luciferase data are presented as the mean of three independent experiments and error bars indicate SD. Statistically significant differences (p < .05, df = 4) are marked by *. Abbreviations: DED1, definitive endoderm day 1; DED2, definitive endoderm day 2; DED3, definitive endoderm day 3; hESC, human embryonic stem cells. Abbreviation: AFBLY: Acitivin, FGF2, BMP, Ly294002 GSC = goosecoid gene symbol.

Download figure to PowerPoint

Gene expression profile experiments showed that 9.1% (369 vs. 231 expected, p = 2.7 × 10−24) of 4,056 genes bound by Smad2/3 were downregulated or upregulated in the absence of Activin/Nodal signaling during endoderm differentiation (Fig. 2B, 2C and Supporting Information Table S4a–c). Notably, inhibition of Activin/Nodal signaling during endoderm differentiation decreases the expression of key endoderm markers bound by Smad2/3 (Sox17, HHex, Cer1, Eomes, Gata4/6, GSC, FZD8, Pitx2, Lhx1, Otx2, and TDGF1), while resulting in the upregulation of genes associated with mesoderm differentiation (Wnt3, Lefty2, PDGFRα, VWF, and PCDH19) (Fig. 2B, 2C, Supporting Information Table S4a–c, and data not shown). These functional regulations observations were confirmed by showing that knockdown in Smad2 expression decreased DE marker expression (Eomes, Mixl1, Lhx1, Sox17, GSC, FoxA2, and Cer), while similar knockdown in Smad3 expression has little effect (Fig. 2E). Thus, Activin/Nodal signaling could promote endoderm differentiation by inducing the expression of essential endoderm regulators and by limiting mesoderm differentiation. Furthermore, 36.7% (1,490 vs. 1,278 expected, p = 3.7 × 10−16) of 4,056 genes bound by Smad2/3 appeared to be upregulated or downregulated in endoderm cells, suggesting that Activin/Nodal could be involved in the organization of a broad part of the transcriptional network characterizing endoderm cells. Considered together, these data uncover for the first time the extent of the transcriptional network downstream of Smad2/3 and confirm the importance of Activin/Nodal signaling in the mechanisms controlling endoderm specification of hESCs.

Smad2/3 is Involved in the Organization of Divergent Transcriptional Networks in hESCs and in Their Endodermal Derivatives

Additional comparison of the ChIP-XL-Seq data generated in pluripotent cells and their endodermal derivatives revealed that 2,332 candidate target genes overlapped between the two cell types (Supporting Information Table S5). In other words, 46.2% (1,125 expected, p ∼ 0) of the transcriptional network identified by Smad2/3 binding in hESCs is maintained through differentiation. This retained network of genes bound by Smad2/3 includes pluripotency factors (Oct-4, Nanog, Tert, Myc, Dppa4, and Nodal), which are not expressed in endoderm cells (Fig. 2A and Supporting Information Fig. S2F), and key regulators of endoderm differentiation (Eomes, Mixl1, Wnt3, FGF8, GATA4, and FoxA2), which are not expressed in hESCs (Fig. 2A–2C). Therefore, Smad2/3 appears to bind key pluripotency genes in endoderm cells without maintaining their transcription. Conversely, Activin/Nodal signaling is necessary to activate the transcription of key endoderm genes upon differentiation without being sufficient to induce their expression in hESCs.

Interestingly, the location and/or the number of Smad2/3 binding sites on Eomes, Mixl1, GSC, GATA4, and FGF8 appeared to change between hESCs and endoderm (Supporting Information Table S5). This in turn suggests that the position of Smad2/3 on these genomic regions could be important in defining their transcriptional activity during the transition from pluripotency to endoderm. This shift in Smad2/3 binding site location and/or number was observed in 37.4% of candidate target genes in DE and 35.9% (545 vs. 478 expected, p = 6.3 × 10−5) of these genes were upregulated or downregulated upon endoderm differentiation. Finally, our analyses also revealed that Smad2/3 acquires the capacity to bind new promoters of key endoderm markers such as GATA6, Sox17, and N-cadherin. To confirm these observations, we decided to analyze the transcriptional activity of genomic regions bound by Smad2/3, specifically in key endoderm markers (Mixl-1, GSC, and Sox17; Fig. 2F) using luciferease reporter assay. These analyses showed that GSC1 and Sox17 binding sites display an increased transcriptional activity upon endoderm differentiation and inhibition of Activin/Nodal signaling limits the transcriptional activity of the same regions in DE. These results imply that these genomic regions contain functional Smad2/3 binding sites and also suggest that our genome-wide analyses could be used to identify enhancers directing expression of DE genes. Finally, Smad2/3 bound genomic regions located in the Mixl1 gene did not show any induction of transcriptional activity, suggesting that additional regulatory sequences are likely to be required for this gene to be expressed in endoderm. Together, these observations show that the location of Smad2/3 binding sites changes on a subset of target genes upon differentiation and that this relocation correlates with the induction of key regulators of endoderm differentiation (Eomes, Mixl1, GATA4, FGF8, GATA6, Sox17, and N-cadherin). However, only systematic studies focusing on separate promoters will uncover the detailed mechanisms controlling the transcriptional regulations of these DE markers.

Smad2/3 and Nanog Cooperate to Control the Transcriptional Network Maintaining Pluripotency in hESCs

Recent studies in our laboratory have shown that the capacity of Smad2/3 to quickly reverse their transcriptional activity upon differentiation is most likely explained by the existence of Smad2/3 partners specifically expressed in endoderm such as Eomes [22]. Conversely, Smad2/3 could interact with transcription factors specifically expressed in hESCs to maintain pluripotency. Interestingly, de novo motif searching and motif enrichment analyses showed that Oct-4 binding sites were over-represented in genomic regions bound by Smad2/3 (Supporting Information Fig. S1F) and thus that this essential pluripotency factor could work together with Smad2/3. Nevertheless, coimmunoprecipitation combined with Western blot analyses have failed to reveal any interactions between Smad2/3 and Oct-4 proteins (data not shown). Moreover, we identified several known Smad2/3 partners in hESCs using gene expression profiling (FOXH1, SKIL, SRF, NANOG, and SIP1; Supporting Information Tables S2 and S5). However, FOXH1 and SRF are expressed in both hESCs and DE, excluding a function for these factors uniquely in pluripotency, while SIP1 and SKIL limit the positive effect of Activin/Nodal signaling on endoderm differentiation without being necessary to maintain the expression of pluripotency markers in hESCs [23] (K. Quinlan, manuscript in preparation). These observations led us to consider Nanog as the most relevant Smad2/3 partner in hESCs [1]. Indeed, Nanog is the first pluripotency gene to disappear upon inhibition of Activin/Nodal signaling [1]. Comparing Smad2/3-ChIP-XL-Seq data and recently available Nanog-ChIP-Seq analyses [24] established that 34.5% (1,721 vs. 1,384 expected, p = 2.1 × 10−35) of 4,990 Nanog target genes were also bound by Smad2/3 (Fig. 3A). Further overlap of genomic locations bound by Nanog and Smad2/3, respectively, revealed a subset of 1,048 regions wherein Nanog and Smad2/3 binding centers overlap (See Oct-4, Nanog, FoxD3, Tert, UTF1, Dppa4, Nodal, Myc, and Cer; Supporting Information Table S6a–c). These genome-wide observations suggest that Smad2/3 and Nanog could be part of a protein complex that controls the transcriptional activity of pluripotency genes in hESCs. To further validate this hypothesis, we performed functional studies showing that Nanog knockdown (Fig. 3B and Supporting Information Fig. S3A) decreased the expression of Smad2/3-Nanog target genes in hESCs (Fig. 3C) with the notable exception of Myc. Thus, Nanog may be a necessary Smad2/3 cofactor that maintains transcriptional activity associated with Activin/Nodal signaling in hESCs. Furthermore, Nanog overexpression in hESCs (Supporting Information Fig. S3B) increased the expression of Smad2/3 target genes (Oct-4, UTF1, dppa4, Myc, Nodal, and Cerberus; Fig. 3D), while maintaining their expression (Oct-4, UTF1, FoxD3, Dppa4, and Tert; Fig. 3E) in the absence of Activin/Nodal signaling. However, Nanog was not sufficient to maintain the expression of Myc, Nodal, and Cer in hESCs thereby establishing that the function of Nanog could vary between different Smad2/3 target genes. ChIP-XL-Q-PCR revealed that Nanog could maintain the expression of Activin/Nodal signaling target genes in the absence of Smad2/3 on their promoter (Supporting Information Fig. S3C) suggesting that overexpression of Nanog can compensate for the absence of transcriptional activity associated with Smad2/3.

thumbnail image

Figure 3. Nanog and Smad2/3 cooperate to maintain the transcriptional network characterizing pluripotency in human embryonic stem cells (hESCs). (A): Number of genes located in the proximity of Smad2/3 binding sites and Nanog binding sites in hESCs. (B): Immunostaining showing the absence of Nanog protein in hESCs stably expressing short hairpin RNA directed against Nanog. (C): Q-PCR analyses showing that knockdown of Nanog decreases the expression of Smad2/3 target genes in hESCs. Statistically significant difference (p < .05, df = 4) in gene expression is marked by *. (D): Nanog overexpression increases the expression of Smad2/3 target genes in hESCs. (E): Nanog overexpression maintains the expression of Smad2/3 target genes in the absence of Activin/Nodal signaling. (F): Inhibition of Activin/Nodal signaling further decreases the expression of Smad2/3 target genes in hESCs in which Nanog expression is knocked down. (G): Luciferase assays showing the transcriptional activity associated with Smad2/3 binding site located in the Dppa4 promoter. hESCs were transiently transfected with different combinations of overexpression vectors for Smad2, Smad3, and Nanog in the presence of Activin or in the presence of SB-431542. Luciferase data are presented as the mean of three independent experiments and error bars indicate SD. Statistically significant differences (p < .05, df = 4) are marked by *. Abbreviation: hESC, human embryonic stem cells.

Download figure to PowerPoint

Further to the above, we also observed that hESCs knocked down for the expression of Nanog at a level equivalent to SB-431542 treatment (ShNanog-hESCs) (Fig. 3B and Supporting Information Fig. S3A) could be grown for more than 20 passages in the presence of Activin while maintaining a decreased but constant expression of pluripotency markers [1]. On the other hand, inhibition of Activin/Nodal signaling was sufficient to drive differentiation of ShNanog-hESCs into fully differentiated neuroectoderm progenitors [1]. Therefore, Activin/Nodal signaling could maintain self-renewal in the absence of Nanog. Subsequent Q-PCR analyses revealed that inhibition of Activin/Nodal signaling further decreased the expression of Smad2/3 target genes in ShNanog-hESCs (Pou5F1, C-Myc, Nodal, and Cer; Fig. 3F).

To confirm these observations, we analyzed the transcriptional activity of a genomic region cobound by Smad2/3 and Nanog located in the promoter of the pluripotency gene Dppa4 using luciferase reporter assays in hESCs. These analyses showed that increases in Smad2, Smad3, and Nanog expression augment the transcriptional activity associated with this genomic region, while combinations of Nanog and Smad2/3 further increases it (Fig. 3G), thereby confirming that Smad2/3 could work synergistically with Nanog to maintain the expression of pluripotency genes. Furthermore, Nanog was sufficient to rescue the transcriptional activity of the same genomic region in hESCs grown in the absence of Activin/Nodal signaling (Fig. 3G) implying that Nanog could substitute entirely for Smad2/3 activity. Considered together, these results imply that Nanog and Smad2/3 could cooperate in hESCs on a genome-wide level to maintain the transcriptional activity of essential pluripotency genes.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Combining ChIP-XL and deep sequencing, we have been able to show that Activin/Nodal signaling acts through Smad2/3 to participate in the control of a large part of the transcriptional networks responsible for maintaining the pluripotent state of hESCs and for inducing endoderm differentiation. Individual knockdown of Smad2/3 also suggests that Smad2 rather than Smad3 is the main effector of Activin/Nodal signaling in hESCs and in their endoderm derivatives. These results reinforce previous observation showing that Smad2 is required for DE specification in the mouse embryo [25]. Importantly, the presence of Smad2/3 on specific promoters does not preclude that the transcriptional activity of the corresponding genes could be regulated by additional factors with an equally essential function. Indeed, recent studies have shown that up to 12 transcription factors could be involved in the transcriptional regulation of pluripotency genes in mESCs [26, 27] and thus it is likely that the regulation of transcriptional networks functioning downstream of Activin/Nodal signaling will require the cooperation of multiple factors in hESCs.

Importantly, the results described in this manuscript can be combined with previous studies to establish a hypothetical model explaining the mechanisms allowing Activin/Nodal signaling to control early cell fate specifications (Fig. 4). In this model, Activin/Nodal signaling maintains the expression of Nanog in hESCs [1], which in turn interacts with Smad2/3 to maintain the expression of pluripotency genes. The same protein complex can be found on the promoter of primitive streak and DE genes without activating their expression in hESCs, suggesting the existence of additional factors repressing the transcriptional activity of these promoters [22, 28]. Nanog may be one of these inhibitors, because we recently showed that Nanog can limit the transcriptional activity associated with Smad2/3 and endoderm differentiation induced by Activin/Nodal [1]. However, the decrease in Nanog expression results in neuroectoderm differentiation and not endoderm differentiation [1]; whereas, data presented here demonstrate that Nanog is necessary to maintain the expression of Smad2/3 target genes in hESCs. Furthermore, recent studies have proposed an active function for Nanog in mesendoderm specification [22, 28]. Consequently, Nanog function may involve tight control of Smad2/3 to allow sufficient transcriptional activity for pluripotency genes to be expressed, while preventing any abnormal expression of DE markers. This hypothesis also suggests that additional inhibitors such as SIP1 [23] may be necessary to fully silence these DE genes in hESCs. Upon endoderm differentiation, these factors and Nanog expression would decrease, allowing Smad2/3 to interact with additional partners and to activate a broad part of the transcriptional network driving endoderm specification. FoxH1 and Eomes are the most likely candidates for such a role, since genetic studies in the mouse and in hESCs have shown that FoxH1 [29] and Eomes [22, 30, 31] represent essential Smad2/3 partners during DE specification. Importantly, this model does not exclude the possibility that Smad2/3 interact with several other partners and that this diversity of interaction might enable Activin/Nodal/TGFβ signaling to accomplish simultaneously diverse functions such as cell cycle regulation and cellular adhesion. Precise future studies of alternative Smad2/3 partners will be necessary to fully define the role of their interactions with Smad2/3 in pluripotency and endoderm differentiation.

thumbnail image

Figure 4. Model showing the mechanisms by which Activin/Nodal signaling controls early cell fate decisions of pluripotent stem cells. In pluripotent cells, Activin/Nodal signaling controls the expression of Nanog which in turn interacts with Smad2/3 to maintain the expression of pluripotency genes. In addition, Nanog and Smad2/3 can bind endoderm genes without inducing their expression, suggesting the action of other factors capable of blocking the expression of such genes in human embryonic stem cells. Decrease in Nanog expression upon differentiation frees Smad2/3 proteins, which can then interact with additional partners. These endoderm-specific partners can then direct Smad2/3 protein onto new regulatory sequences to induce the expression of key endoderm genes. Abbreviations: DE, definitive endoderm; hESC, human embryonic stem cells.

Download figure to PowerPoint

CONCLUSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

To conclude, our results provide new insights into the mechanisms by which Activin/Nodal signaling controls early cell fate decisions and our genome-wide analyses represent an important step toward the identification of the key factors controlling the transition between pluripotency and endoderm. The potential relevance of these insights to hIPSCs, which also rely on Activin/Nodal signaling to maintain their pluripotency [32], also promises a wider contribution which will facilitate the development of robust and universal methods to differentiate human pluripotent stem cells into endoderm cells.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

We thank the Genome Technology and Biology group at the Genome Institute of Singapore for the sequencing of ChIP samples. This work was supported by a MRC senior nonclinical fellowship (to S.B. and L.V.), by the Juvenile Diabetes Research Foundation (to R.P., M.T., and L.V.), by MRC Programme grant (to R.P.), by EU grant LivES (to N.H.), by ASTAR studentship (to A.T.), and by NIH Research Cambridge Biomedical Research Centre.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
STEM_666_sm_suppfigure1.tif3175KSupporting Figure 1. Chromatin immunoprecipitation performed in the presence of additional protein cross linkers and combined with deep sequencing reveal that Activin/Nodal signalling controls the core pluripotency transcriptional network in hESCs. (A) Lack of efficiency of conventional ChIP approach to detect the interactions of Smad2/3 proteins with genomic DNA in hESCs. Three separate ChIP-QPCR experiments were performed on hESCs to detect the presence of Smad2/3 on the promoters of Smad7 and Lefty, two known genes of Activin/Nodal signaling. Results were normalised against a control region (Smad7 3UTR) and are expressed as +/- standard derivation from three technical replicates. (B) Combination of diverse chemical cross linkers improves ChIP efficiency. ChIP was performed in the presence of different combination of cross linkers (Dimethyl dithiobispropionimidate or DTBP; Disuccinimidyl tartarate or DST, Ethylene glycol bis succinimidylsuccinate, or EGS) as described in the Supplementary Methods. Importantly, these experiments were performed simultaneously on the same batch of cells to avoid experimental variations. The immunoprecipitated DNA was then amplified using Q-PCR with specific primers to detect enrichment in the denoted genomic regions (Supplementary Table 7). Results were normalised against a control region (Smad7 UTR) and are expressed as +/- standard derivation from three replicates. (C) FACS analyses showing that hESCs grown in CDM+Activin+FGF2 express homogenously the pluripotency marker Nanog. hESCs immuno-stained with secondary antibody alone were used as negative control to define the gate for statistical analyses. (D) Representative view of Smad2/3 binding sites in genomic regions containing known Activin/Nodal target genes. Y axis shows the number of sequence tags per 200 bp window. Red Arrows point to statistically significant enriched genomic regions while blue arrow point to previously characterised binding sites. (E) Validation of Smad2/3 binding sites in hESCs. ChIP-QPCR were performed on hESCs grown in the presence or in the absence of Actvin/Nodal signalling to confirm the specific presence of Smad2/3 on the genomic regions identified using ChIP-XL-Seq. Results were normalised against control region (Smad73UR) and are expressed as +/- standard deviation. (F) Schematic representation of transcription factor binding motifs inferred from genomic regions bound by Smad2/3 in hESCs. Statistical analyses performed as described in supplementary materials. (G) Gene Ontology analysis of genes bound by Smad2/3 which are down or up regulated in the absence of Activin/Nodal signalling in hESCs. The Y axis shows the GO term and the X axis shows the p value for the significance of enrichment for the top 30 GO terms.
STEM_666_sm_suppfigure2.tif3175KSupporting Figure 2. Activin/Nodal signalling pathway controls a broad part of the transcriptional network characterising endoderm cells. (A) FACS analysis showing that endoderm cells generated from hESCs express homogenously Sox17. DE cells immuno-stained with secondary antibody alone were used as negative control to define the gate for statistical analyses. (B) Gene Ontology analysis of genes bound by Smad2/3 which are down or up regulated in the absence of Activin/Nodal signalling in endoderm cells. The Y axis shows the GO term and the X axis shows the p value for the significance of enrichment for the top 30 GO terms. (C) Validation of Smad2/3 binding sites in endoderm cells. ChIP-XL-QPCR was performed on endoderm cells to confirm the presence of Smad2/3 on the genomic regions identified using ChIP-XL-Seq. Results were normalised against control region (Smad73UTR) and are expressed as +/- standard deviation from three experiments. (D) Smad2/3 binding regions in endoderm cells are evolutionary conserved. The average conservation (phastCons, scaled 0-1) of the genomic regions bound by Smad2/3 was determined as described in Materials and Methods. Matched non-specific genomic regions were used as negative control (Red lines). (E) Relative position of the genomic regions bound by Smad2/3 in endoderm to the Transcriptional Start Site (TSS). (F) Expression of pluripotency markers is down regulated upon endoderm differentiation. hESCs (H9) were grown for 3 days in culture conditions inductive for endoderm differentiation (CDM + Activin + FGF2 + BMP4 +Ly) (Day1, Day2 and Day3). RNAs were extracted every day, and then Q-PCR analyses were performed to detect the genes denoted. Data represent the mean of three independent experiments and error bars indicate standard deviation.
STEM_666_sm_suppfigure3.tif3175KSupporting Figure 3. Nanog and Smad2/3 cooperate to maintain the transcriptional network characterizing pluripotency in hESCs. (A) Q-PCR analyses showing the absence of Nanog transcripts in hESCs stably expressing ShRNA targeting Nanog (ShNanog-hESCs). hESCs (H9) grown in CDM + Activin + FGF2 were used as positive control (hESCs) while hESCs grown for 7 days in the presence of SB431542 were used as negative control (hESCs + SB). (B) Q-PCR analyses showing the expression of Nanog in hESCs stably transfected with an expression vector for Nanog grown in CDM + Activin + FGF2 (OE Nanog) or grown for 7 days in the presence of SB431542 (OE Nanog + SB). hESCs (H9) grown in similar culture conditions were used as control (hESCs and hESCs +SB). (C) ChIP analyses showing that Smad2/3 does not bind candidate target genes in hESCs overexpressing Nanog grown in the presence of SB431542 for 24 hours.
STEM_666_sm_supptable1.XLSX5463KSupporting Table 1. Smad2/3 binding sites and associated genes in hESCs.
STEM_666_sm_supptable2.XLSX897KSupporting Table 2. Smad2/3 target genes up and down regulated in the absence of Activin/Nodal signalling in hESCs.
STEM_666_sm_supptable3.XLSX4400KSupporting Table 3. Smad2/3 binding sites and associated genes in Definitive Endoderm.
STEM_666_sm_supptable4.XLSX1443KSupporting Table 4. Smad2/3 target genes up and down regulated in the absence of Activin/Nodal signalling in Definitive endoderm.
STEM_666_sm_supptable5.XLSX315KSupporting Table 5. Candidate target genes bound in hESC and in Definitive Endoderm.
STEM_666_sm_supptable6.XLSX5198KSupporting Table 6. Genes bound by Smad2/3 and Nanog in hESCs
STEM_666_sm_supptable7.XLS32KSupporting Table 7. List of primers for ChIP-Q-PCR and luciferase assays.
STEM_666_sm_suppinformation.doc77KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.