|Home | About | Journals | Submit | Contact Us | Français|
Elucidating the underlying transcriptional control of pluripotent cells is necessary for the development of new methods of inducing and maintaining pluripotent cells in vitro. Three transcription factors, Nanog, Oct4, and Sox2, have been reported to form a feedforward circuit promoting pluripotent cell self renewal in embryonic stem cells (ESC). Previously, we found that a transcriptional repressor activity of Tcf3, a DNA-binding effector of Wnt signaling, reduced Nanog promoter activity and Nanog levels in mouse ESC (mESC). The objective of this study was to determine the scope of Tcf3 effects on gene expression and self renewal beyond the regulation of Nanog levels. We show that Tcf3 acts broadly on a genome-wide scale to reduce the levels of several promoters of self renewal (Nanog, Tcl1, Tbx3, Esrrb) while not affecting other ESC genes (Oct4, Sox2, Fgf4). Comparing effects of Tcf3 ablation with Oct4 or Nanog knockdown revealed that Tcf3 counteracted effects of both Nanog and Oct4. Interestingly, effects of Tcf3 were more strongly correlated with Oct4 than with Nanog, despite the normal levels of Oct4 in TCF3−/− mESC. The deranged gene expression allowed TCF3−/− mESC self renewal even in the absence of leukemia inhibitory factor (Lif) and delayed differentiation in embryoid bodies. These findings identify Tcf3 as a cell-intrinsic inhibitor of pluripotent cell self renewal that functions by limiting steady-state levels of self renewal factors.
Maintenance of pluripotency as stem cells proliferate (i.e. self renewal) is necessary for the function of the inner cell mass (ICM) during embryonic development and for the potential of embryonic stem cells (ESC) in cell-based therapies in humans. Several studies have identified core transcriptional components that stimulate self renewal. Knockout mouse experiments showed that the Oct4 transcription factor was necessary for ICM maintenance 1, and in vitro experiments demonstrated that tight regulation of Oct4 expression (within 0.5x–1.5x normal levels) was necessary to maintain mESC self renewal and prevent differentiation 2. Nanog was first identified in mESC for its restricted pattern of expression and its ability to stimulate self renewal in the absence of Leukemia inhibitory factor (Lif)/Jak/Stat3 stimulation 3, 4. Despite gene ablation experiments that showed Nanog was not absolutely necessary for mESC self renewal in vitro 5, reduction of Nanog levels substantially inhibited efficiency of mESC self renewal and stimulated differentiation 4–6. Sox2 physically bound to Oct4 on promoters of target genes 7, and was necessary for Oct4-mediated activation of mESC-expressed genes such as Utf1, Fgf4, and Nanog 8, 9. Like Oct4, genetic ablation of Nanog and Sox2 demonstrated the requirement for these factors for the maintenance of pluripotent cells in mouse embryos 4, 10.
In both mouse and human ESC, genome wide examination of Oct4 and Nanog DNA-binding by chromatin immunoprecipitation (ChIP) demonstrated that the two proteins bound to largely overlapping sets of genes 11–13. Oct4, Sox2, and Nanog have been found to regulate one another’s expression in stem cells 8, 9, 11, 14, 15. Taken together, these observations supported a feedforward circuit of Oct4, Sox2, and Nanog transcription factors that promote pluripotent cell self renewal 11. Interestingly, while the degree of overlapping promoter binding between Oct4 and Nanog was high within each species of ESC, relatively few transcriptional regulators (only 18 genes) were conserved between mouse and human ESC 11, 12.
Knowledge of the basic molecular underpinnings promoting pluripotency was essential for reprogramming of differentiated cells (dermal fibroblasts, hepatocytes, gastric epithelial cells) to an induced pluripotent stem cell (iPS) state through the forced expression combinations of Oct4, Nanog, Sox2, Klf4, c-myc, and Lin28 16–19. The combination of Oct4, Sox2, c-myc, and Klf4 need to be expressed only during the first two weeks of the iPS reprogramming process to initiate a cascade of currently unknown events leading to a stable ESC-like state 20. The iPS approach of reprogramming poses substantial technical and efficiency advantages over other reprogramming approaches, such as somatic cell nuclear transfer and cell fusion 21, 22. A primary barrier blocking therapeutic potential of iPS cells in humans is the risk of oncogenic events caused by random integration of gene expression vectors in the genome 23. Potentially, biochemical or siRNA-mediated inactivation of inhibitors of the pluripotent cell self renewal could provide an alternative, safer method of reprogramming to iPS cells. To accommodate these approaches, inhibitors of pluripotent cell self renewal must be identified as potential targets of manipulation.
The TCF3 (TCF7L1 in humans) gene was one of the 18 transcriptional regulator genes that was bound by Oct4 and Nanog in both mouse and human ESC 12. TCF3 encodes an HMG-domain containing transcription factor belonging to the Tcf/Lef family of Wnt-signaling effector molecules. Tcf3 possesses a β-catenin binding domain 24, and it has been suggested to function both as a β-catenin-dependent activator of target gene transcription and as a β-catenin-independent repressor of target genes 25, 26. Previously, we reported that Tcf3 bound to the promoter region of the Nanog gene and inhibited the levels of Nanog transcription, mRNA, and protein in mESC 27.
In this study, we tested the ability of Tcf3 to limit the effects of pro-self-renewal factors in mESC. We find that Tcf3 regulates a substantially overlapping set of Oct4- and Nanog-regulated genes; however, Tcf3 affected gene expression with a strongly opposite effect compared to Oct4 and Nanog. By using mESC in which both alleles of TCF3 were ablated, we found that absence of Tcf3 stimulated the long term self renewal of mESC. The self-renewal inhibitory effects of Tcf3 were cell autonomous and pronounced in the absence of Lif/Jak/Stat3 signaling. Thus, despite the relatively high levels of Tcf3 in mESC, Tcf3 functions as a limiter of the transcription circuits necessary for pluripotent cell self renewal.
Mouse ESC were maintained using standard culture conditions. All experiments were performed using feeder-free mESC lines which were propagated on gelatinized (30min at room temperature) 10 cm tissue culture dishes (Falcon 35–3003) in Knockout Dulbecco’s Modified Eagle Medium (Knockout-DMEM, GIBCO 10829) supplemented with 15% fetal bovine serum (Atlanta Biologicals), 2 mM L-glutamine (GIBCO), 10 mM HEPES (GIBCO), 100 U/ml penicillin (GIBCO), 100 µg/ml streptomiycin (GIBCO), 0.1 mM nonessential amino acids (GIBCO), 0.1 mM β-mercaptoethanol (GIBCO), and 1000 U/ml Lif (Chemicon). Media was replaced daily and every 2–3 days, single cell suspensions of mESC were created by trypsin treatment (0.25% trypsin- EDTA; GIBCO) and passaged onto new gelatinized plates.
Quantitative real-time PCR reactions were performed with the iTaq SYBR Green Supermix (BioRad) and an iCycler apparatus (BioRad). Amplification was achieved by the following protocol: 1X 95°C, 2min; 40X 95°C, 30 sec, 60°C, 30 sec; 1X 95°C, 1 min. To identify potential amplification of contaminating genomic DNA, control reactions using mock cDNA preparations lacking reverse transcriptase were run in parallel for each analysis. To ensure specificity of PCR, melt-curve analyses were performed at the end of all PCR reactions. The relative amount of target cDNA was determined from the appropriate standard curve and divided by the amount of GAPDH cDNA present in each sample for normalization. All PCRs had an efficiency of 80% or higher. Each sample was analyzed in duplicate, and results were expressed as means +/− standard deviations. Primer sequences and reaction conditions are available for each primer set upon request.
Total RNA was extracted from three independently prepared cultures of TCF3 +/+ and TCF3 −/− mESC using Trizol Reagent (Invitrogen) and purified using RNeasy Mini Kit (Qiagen). Quality of RNA was assessed by spectrophotometer (A260/A280 between 1.9 and 2.1) and by assessing degradation of rRNA on a denaturing Northern gel prepared with NorthernMax reagents (Ambion). 10 µg (1 µg/µl) of each sample was sent for microarray analysis. Labeling of the RNA, hybridization onto Affymetrix mouse genechip 430 2.0 microarrays, and scanning of samples were performed by the UIC Genomics Core Facility. Raw data were normalized by quantiles, and summarized by robust multiarray average (RMA). Pairwise comparisons between each group of three replicates were performed and t-test was used to rank each probeset based on its likely statistical significance using ComparativeMarkerSelection module of GenePattern software. Probesets displaying a significant difference as defined by a Benjamini-Hochberg adjusted t-test p value < 0.05 were utilized for subsequent analyses. CEL files containing Oct4 and Nanog shRNA were downloaded from GEO omnibus 12, and processed identically. Heirarchical clustering and Venn comparisons were completed using GenePattern software. Spearman rank order correlation and linear regression were performed using the combination of Microsoft Excel and the Handbook of Biological Statistics (http://udel.edu/~mcdonald/statspearman.html).
Previously, we reported that Tcf3 was the most highly expressed TCF family member in mESC, and Tcf3 reduced the level of Nanog by directly binding to and inhibiting Nanog promoter activity 27. Based on the partial rescue of differentiation defects in TCF3−/− mESC from reducing Nanog by RNAi 27, and the ability of Tcfs to regulate many target genes in other contexts 28, 29, we posited that the regulation of additional genes was also likely to be important for Tcf3 function. Since dynamic and uncertain levels of knockdown efficiency through RNAi-based approaches could introduce unwanted caveats, we examined determined steady state gene expression levels in TCF3−/− mESC lines 27, 30.
Gene expression profiles were compared for TCF3+/+ and TCF3−/− mESC grown under self renewal (Lif+) conditions. Using a cutoff value of a Benjamini-Hochberg adjusted p value < 0.05 31, 2508 probesets (corresponding to 1338 known genes identified by UCSC Genome Database) from the Affymetrix mouse genechip 430 2.0 microarray were significantly different in TCF3+/+ and TCF3−/− mESC (Supplementary Table 1). There were 1414 probesets (742 UCSC known genes) that were decreased and 1094 probesets (596 UCSC known genes) that were increased in TCF3−/− mESC. Quantitative RT-PCR analysis of downregulated genes (Id2, Akt1, Krt8, Gadd45b, Cyr61, Jun, Egfr, Cdkn1c, Fgf5) was used to validate the results of the microarray experiments, and all were found to be downregulated with this independent method (Figure 1). Similarly, 8 of 9 genes identified as upregulated in TCF3−/− and 3 of 4 genes not identified as being changed were validated by qPCR (Figure 1).
In addition to Nanog, which was a good internal control for the success of the microarray procedure, several other previously described “stem cell genes” were significantly increased in TCF3−/− mESC (Supplementary Table 1). Three genes necessary for mESC self renewal (Tcl1, Tbx3, Esrrb) 6 and the Klf4 gene necessary for reprogramming adult cells into a pluripotent state 16, 17 were all detected as increased in TCF3−/− mESC by microarrays and confirmed by qPCR (Figure 1). Interestingly, Oct4, Sox2, and Fgf4 were not found to be significantly different between TCF3+/+ and TCF3−/− samples, suggesting that not all mESC-specific genes were increased. This was an important finding because it ruled out the possibility that differential gene expression could be caused by simply increasing the ratio of undifferentiated to differentiated cells in the cultures of TCF3−/− compared to TCF3+/+ cells. The lack of statistically significant differences in Oct4, Sox2, and Fgf4 were confirmed by qPCR, whereas Zfp42 levels were found to be increased in TCF3−/− mESC when assessed by qPCR (Figure 1). Thus, Tcf3 inhibited the expression of several genes (Esrrb, Nanog, Tcl1, Tbx3) necessary for efficient stem cell self renewal, yet it did so in a specific manner because expression of other mESC-specific genes (Oct4, Fgf4, Sox2) was not significantly changed in TCF3−/− mESC.
Based largely on binding of Oct4, Nanog, and Tcf3 to one another’s gene determined by ChIP assays, Tcf3 has been suggested to be an integral member of the so called stem cell transcriptional network 12, 13. To determine if DNA binding of Oct4 and Nanog to the Tcf3 gene could have important effects on Tcf3 levels in mESC, Tcf3 mRNA and protein levels were measured after siRNA-mediated knockdown of either Oct4 or Nanog in TCF3+/+ mESC. Within 24 hours, transient transfection of gene-specific siRNA molecules reduced Oct4 or Nanog mRNA and protein levels to below 5% of their normal levels, which were unaffected by mock transfections using non-specific siRNA molecules (Figure 2). Levels of Tcf3 mRNA were reduced 26% by Nanog-knockdown and 52% by Oct4-knockdown (Figure 2A). Note that these levels were comparable to the reduction of Oct4 by Nanog-knockdown and the lack of reduction of Nanog by Oct4-knockdown at this early timepoint (Figure 2A). A decrease in Tcf3 protein by Oct4 siRNA was evident by Western blot analysis (Figure 2B), whereas only a slight decrease in Tcf3 was detected after Nanog siRNA at 24 hrs. These results suggested that Tcf3 expression was activated by Oct4, and to a lesser extent Nanog, in mESC. Moreover, decreased Tcf3 mRNA expression by Oct4 siRNA treatment occurred prior to effects on Nanog expression, which was consistent with previous reports examining the time course of Oct4 effects 32. The requirement for Oct4 and Nanog for normal Tcf3 expression levels in mESC supported the hypothesis that Tcf3 played an important role in regulating stem cell self renewal.
To examine the relatedness of effects caused by Tcf3, Nanog, and Oct4, signal intensity data from microarray experiments were processed identically. Two independent studies have previously reported genome-wide effects on mESC gene expression in response to Oct4 or Nanog reduction 6, 12. Data from the Loh et al study was chosen for direct comparison to the effects of Tcf3 because it utilized the same microarray platform as the Tcf3 experiments allowing results to be directly compared 12. Analysis using all probesets identified as significantly changed (p<0.05) for all treatments generated a similar trend; however the large number of genes changed by Oct4 shRNA obscured the potentially most important similarities (Supplementary Figure 1A). To provide a fair assessment of overlap, the same number of genes was used for each condition. The top 2500 ranked differentially expressed probesets for each comparison (TCF3+/+ vs. TCF3−/−, Mock vs. Nanog shRNA, and Mock vs. Oct4 shRNA) were compared. The list of Tcf3-regulated probesets shared 488 probesets with Nanog-regulated and 571 probesets with Oct4-regulated lists (Supplementary Figure 1B). As previously reported 6, 12, these analyses revealed a high degree of overlap between Nanog and Oct4 with 814 common probesets (Supplementary Figure 1B). Expression of 233 probesets was changed by each of the three conditions (Supplementary Figure 1B). Thus, approximately one fourth of the genes affected by both Oct4 and Nanog were also affected by Tcf3 in mESC.
Hierarchical clustering and statistical correlation analyses were used to determine if the effects of Tcf3, Nanog, and Oct4 on gene expression were related in terms of the direction (upregulated vs. downregulated) and the magnitude (fold change) of effects. Separately comparing the effects of Tcf3 with those of either Oct4 or Nanog was performed by hierarchical clustering using only unique gene identifiers within the top 2500 statistically significant changes to probesets for each treatment. For each analysis, two primary clusters emerged. Most genes that were increased in TCF3−/− were decreased by Oct4 shRNA or Nanog shRNA (87% and 77%, respectively). Most genes that were decreased in TCF3−/− were increased by Oct4 shRNA or Nanog shRNA (88% and 81%, respectively) (Figure 2C). Interestingly, despite the increased levels of Nanog and normal levels of Oct4 in TCF3−/− mESC, downstream effects of Tcf3 were more closely related to those of Oct4 (Figure 2D). This point was exemplified by a greater Spearman rank order coefficient for Oct4 shRNA versus TCF3−/− effects (ρ= −0.78; p-value = 1.2×10−98) than for Nanog shRNA versus TCF3−/− effects (ρ = −0.61; p-value = 1.7×10−42) (Figure 2D). Consistent with previously published analysis of the effects of Nanog and Oct4 shRNA, our analysis showed that Nanog and Oct4 shRNA treatments caused very similar effects on gene expression as 87% of genes were regulated in the same direction (ρ= 0.70) (Figure 2D). It should be pointed out that since siRNA-reduction of Oct4 reduced Tcf3 mRNA and protein, the opposing effects of Tcf3 and Oct4 are not caused by TCF3 gene being downstream of Oct4-mediated regulation. Similarly, since Oct4 mRNA and protein levels were not changed in TCF3−/− mESC, the opposing effects are not caused by Oct4-encoding gene being downstream of Tcf3-mediated regulation. Thus, these data suggest Tcf3 regulates a substantial proportion of Oct4-regulated genes and is doing so in an Oct4-independent manner.
Since Tcf3 caused effects with significant overlap with those caused by Oct4 or by Nanog (Figure 2C, D), and Nanog and Oct4 have been shown to regulate highly overlapping sets of target genes 11, 12, the possibility that Tcf3, Oct4, and Nanog coordinately regulate gene expression was examined. Hierarchical clustering was performed to classify the 199 unique genes that were found to be significantly affected by all three factors (Supplementary Figure 2A). Two primary groups and several smaller groups of genes were identified: 60 group 1 genes were increased in TCF3−/− and decreased by Oct4 shRNA and by Nanog shRNA, 104 group 2 genes were decreased in TCF3−/− and increased by Oct4 shRNA and by Nanog shRNA, 35 group 3 genes were all remaining genes that did not fall into groups 1 or 2 (Supplementary Figure 2). Comparing the fold change by which each unique gene was affected by Tcf3, Oct4, and Nanog revealed a strong correlation between the effects of Tcf3 and those of Oct4 and Nanog on gene expression (Figure 3). Tcf3 caused effects that directly counteracted the effects caused by Oct4, and to a lesser extent, Nanog.
Absence of Tcf3 was previously found to delay differentiation in short-term colony and embryoid body assays13, 27. The effects on induced differentiation could be caused by a requirement for Tcf3 in formation or survival of specific cell lineages, or they could reflect a more general role for Tcf3 in regulating mESC self renewal. Interestingly, despite Tcf3’s role in counteracting Oct4- and Nanog- programs of gene expression, Tcf3 expression was maintained at its high levels even after long term culture of mESC (Figure 4A). Therefore, if Tcf3 limited self renewal in mESC, it would suggest potential for artificially removing Tcf3 to stabilize the epigenetic programming of the genome to support long term ESC self renewal.
To determine if Tcf3 functioned to limit mESC self renewal, we closely examined the proliferation and self renewal characteristics of TCF3−/− mESC. Comparison of the plating efficiency of TCF3+/+ and TCF3−/− mESC revealed that Tcf3 inhibited the clonogenic ability of mESC in the presence of 1000U/ml Lif (Supplementary Figure 3A). In addition, the ratio of cells generating alkaline phosphatase (AP) -positive mESC colonies was consistently higher in TCF3−/− mESC compared to TCF3+/+ mESC (Supplementary Figures 3A). The difference between TCF3+/+ and TCF3−/− mESC was increased at suboptimal levels of Lif cytokine (Supplementary Figures 3C).
To test the potential for sustained effects of TCF3 ablation, mESC were subjected to long term culture in the absence of exogenous Lif. As expected, TCF3+/+ mESC lost Oct4 expression by the second passage and stopped proliferating at that point (Figure 4B, C). In contrast, TCF3−/− mESC continued to proliferate and continued to express Oct4 after Lif was removed from culture media (Figure 4). The proliferation rate of TCF3−/− mESC was reduced during days 3–15 (passages 2–5) before recovering to a level that was stable for at least 20 additional passages. Thus, Tcf3 was not required for proliferation of mESC; moreover, Tcf3 inhibited proliferation in the absence of Lif. The TCF3−/− mESC that have undergone at least ten consecutive passages in the absence of exogenous Lif are referred to “Lif-independent TCF3−/− mESC” throughout the rest of the manuscript.
To determine whether Lif-independent TCF3−/− mESC retained important mESC characteristics, expression of self renewal markers was examined. Immunofluorescent staining of colonies revealed that even after 30 days (10 passages) of culture without exogenous Lif, nuclear Oct4 and Nanog were still expressed by Lif-independent TCF3−/− mESC at levels similar to mESC passaged in the presence of Lif (Figure 4D). In Lif+ media, greater than 72% of TCF3+/+ colonies were AP positive after two days and 85% of TCF3−/− mESC were AP positive; the percent of AP positive mESC colonies remained high (78%) when Lif-independent TCF3−/− mESC were cultured in the absence of Lif (Supplementary Figure 4). Measurement of 22 genes by qPCR showed that levels of 19 genes were similar in TCF3−/− mESC and Lif-independent TCF3−/− mESC (Supplementary Figure 5).
To test the ability of Lif-independent TCF3−/− mESC to contribute to distinct cell lineages, cells were subjected to in vitro differentiation in embryoid body (EB) experiments. As reported previously, TCF3+/+ and TCF3−/− mESC cultured in Lif+ media were able to differentiate into EBs and express marker genes corresponding to cell types in each of the three primary germ layers, but the differentiation of TCF3−/− mESC was delayed (Figure 4E) 27. The Lif-independent TCF3−/− mESC were also able to form EBs and expressed markers for endoderm (Gata4/6), mesoderm (Goosecoid, Brachyury), and ectoderm (Fgf5) (Figure 4E). Note that the expression of each differentiation marker was decreased and/or further delayed in EBs from Lif-independent TCF3−/− mESC compared to the Tcf3+ and the Lif+ controls. In addition, Sox1 levels were substantially decreased in Lif-independent TCF3−/− mESC suggesting a potential defect in the differentiation of neural lineages (Figure 4E). Taken together, these data show the TCF3−/− mESC can continue to self renew in the absence of exogenous Lif.
Upon Lif withdrawal, wild-type mESC stop proliferating and become sensitive to apoptosis 33–35. Western blot analysis of protein from TCF3+/+ mESC showed that cleaved caspase 3 was increased by passage 1 and 2 after Lif withdrawal (Figure 5A); the few TCF3+/+ cells that remained attached to the tissue culture dish after 3 passages no longer expressed Oct4 and no longer proliferated (Figure 5A, B). In contrast, during the second and third passages, TCF3−/− mESC displayed only a slight increase in cleaved caspase 3, which was similar to that observed in TCF3+/+ mESC even in the presence of Lif (Figure 5A). As described previously, wild-type mESC divided rapidly (every 10–12 hrs) in Lif+ media and exhibited a unique cell division cycle where more than 75% of cells were in S-phase (Figure 5B, C, Supplementary Figure 6) 36, 37. Lif withdrawal inhibited cell division and increased the percentage of G1 phase TCF3+/+ cells from 8.9% in Lif+ to 29.1% by passage 2 (Figure 5B, C); analysis of TCF3+/+ cells after passage 2 was not performed due to extensive cell death. The percentage of TCF3−/− mESC in G1-phase increased from 8.6% in Lif+ media up to 15.4% by the third passage in Lif- media and then returned to 9.0% by the fourth passage (Figure 5B, C). At no point were there less than 73% of TCF3−/− mESC in S-phase (Figure 5B, C). The increased G1 phase occurred concomitantly with the slowing of cell proliferation during a brief period of delay before recovery (see Figure 4B). Thus, absence of Tcf3 inhibited the transition from the rapid ES cell cycle to a G1-enriched somatic cell cycle which was associated with a requirement for cytoprotective signals to prevent apoptosis 36.
Mouse ESC express low levels of Lif and other cytokines that have been suggested to be sufficient for self renewal at high mESC colony densities 38. Although TCF3+/+ control cells differentiated and died when subjected to identical culture conditions that allowed TCF3−/− mESC to self renew, it was possible that self renewal of TCF3−/− mESC occurred through non-cell autonomous effects. To determine if any factor secreted by TCF3−/− mESC could enable mESC to self renew in the absence of exogenous Lif, a simple co-culture experiment was performed. Equal number of TCF3+/+ and TCF3−/− mESC were mixed together in the same culture, and the mixture of cells was switched to Lif- media. At the end of each passage, the number of each cell type was determined by counting total cells (Figure 6A) and using PCR to determine the ratio of TCF3+ and TCF3− alleles in genomic DNA isolated (Figure 6B). TCF3+/+ cells failed to expand in the mixed culture between passages 2–3, and TCF3−/− mESC exhibited the typical delay in proliferation before regaining high proliferation rates after the fifth passage (Figure 6A, B). Thus, although Tcf3 could potentially affect several different extracellular signals through deregulated gene expression, these results demonstrated that those effects would be insufficient to promote Lif-independent self renewal. Instead, the self renewal effects of Tcf3 were cell autonomous.
Given Tcf3’s biochemical function as a transcription factor, the cell autonomous manner of Tcf3-mediated effects suggested that the effects could be epigenetic in nature. One characteristic that can distinguish an epigenetic from a genetic effect is reversibility in which transcriptional programs can be reset via culture conditions. We took advantage of the delayed proliferation exhibited by TCF3−/− after initial withdrawal of Lif (Figure 4B; passages 2–5) to use it as a distinguishing difference between Lif-independent TCF3−/− mESC and TCF3−/− mESC cultured only in Lif+ media (naïve TCF3−/− mESC). Lif-independent TCF3−/− mESC were returned to grow in the presence of 1000U/ml Lif for 10 passages (adapted Lif-independent TCF3−/− mESC) and then subjected to the Lif-withdrawal assays described for Figure 4B and Figure 6C. In these assays, the TCF3+/+ mESC die by the third passage, and the naïve TCF3−/− mESC never exposed to Lif- conditions displayed the typical slowing after passage 2 (Figure 6D). In contrast to the constant proliferation rate of Lif-independent TCF3−/− mESC, the adapted Lif-independent TCF3−/− mESC displayed reduced proliferation during passage 3 and 4, which was similar to, albeit less severe than, the reduced proliferation of naïve TCF3−/− mESC (Figure 6D). This result suggested that the effects of Lif-independent culture of TCF3−/− were reversible, which was consistent with the effect being epigenetic in nature.
By genetically ablating Tcf3 from mESC, we showed that Tcf3 broadly functioned to counteract the effects of pro-self-renewal factors Nanog and Oct4. The data presented here were consistent with our previously published finding that Tcf3 repressed Nanog expression, and they provide substantial new insights by identifying Nanog-independent effects of Tcf3. As opposed to transient RNAi-based approaches, the use of TCF3−/− mESC lines revealed a stable, steady-state inhibitory effect of Tcf3 on pluripotent cell self renewal. As a result of increased expression of Tcf3/Oct4/Nanog-regulated genes, TCF3−/− mESC were capable of indefinite self renewal in the absence of Lif-Jak-Stat3 stimulation. Taken together, these findings suggest that Tcf3 limits the ability of Nanog and Oct4 to activate target genes that promote self renewal.
While the analysis of gene expression data revealed a strong inverse correlation between the effects of Tcf3 and the effects of both Oct4 and Nanog, the observed effects are inconsistent with a simple hierarchical relationship between Tcf3 and the proposed Nanog-Oct4-Sox2 network of regulators. Since siRNA-mediated knockdown of Oct4 and Nanog each reduced Tcf3 levels (Figure 2A, B), the majority of effects of Nanog and Oct4 on gene expression cannot occur through regulation of Tcf3 levels. Since Oct4 and Sox2 levels were not significantly different in TCF3−/− mESC, the effects of Tcf3 cannot occur through regulation of Oct4 or Sox2 levels (Figure 1). Although Nanog was increased by ablation of Tcf3, the overlap between Tcf3 and Nanog was actually less than that between Tcf3 and Oct4 (Figure 2C, D). The closer relationship between Tcf3- and Oct4- mediated effects (ρ = −0.78) compared to Tcf3- and Nanog- mediated effects (ρ = −0.61), or Oct4- and Nanog- mediated effects (ρ = 0.70), further supports a model in which Tcf3 and Oct4 directly regulate the same target genes. The repressor activity of Tcf3 in mESC 27 and the activator activity of Oct4 9, 39 are consistent with our observation that Tcf3 and Oct4 caused essentially opposite effects on co-regulated genes. Therefore, these new data suggest that effects of Tcf3 occur in parallel to effects of Nanog and Oct4 and oppose the effects of Nanog and Oct4.
These conclusions are consistent with a recently published report identifying genome-wide chromatin occupancy for transcription factors in mESC. The study by Cole et al used the ChIP-CHIP method to reveal that Tcf3, Oct4, and Nanog bound to a significantly overlapping set of putative promoter regions, including those for one another’s genes13. They also showed that Tcf3-bound genes were more frequently increased upon shRNA reduction of Tcf3 than unbound genes 13. The co-occupancy of Tcf3, Oct4 and Nanog on putative target genes provides a good mechanism for our finding of the parallel effects of Tcf3 with Nanog and Oct4 on gene expression in mESC (Figure 2, Figure 3). Thus despite important differences between the two studies (discussed later), our results support the overall conclusion that Tcf3 is an integral component of mESC circuitry.
Differences between our study and the Cole et al study could provide the most interesting insights into Tcf3’s regulation of self renewal. Knockdown of Tcf3 by shRNA elevated Oct4, Sox2 and Nanog after 48 hours 13 whereas constitutive knockout of the TCF3 gene elevated only Nanog levels (Figure 1, Figure 7). Although the reason for the difference in steady state effects versus dynamic effects is not known, it is tempting to speculate that Tcf3 represses a mechanism of inhibiting Oct4 and Sox2 expression. Regardless of the underlying mechanism, the normal levels of Oct4 and Sox2 in TCF3−/− mESC are important. The increased levels of Oct4 and Sox2 after shRNA reduction of Tcf3 precluded conclusions that gene expression effects on co-occupied genes could be caused by increased Oct4, by increased Sox2, or by decreased Tcf3. In contrast, changes to the steady state levels of gene expression in TCF3−/− mESC can be confidently concluded to be Oct4-independent and Sox2-independent.
It is interesting to consider that Tcf3 inhibits self renewal yet it is stably expressed in self renewing pluripotent cells. Reconciling this apparent contradiction can be aided by combining classic and new perspectives on the role of pluripotency in embryonic development. Teratoma and lineage tracing experiments using epiblast cells from post-implantation staged mouse embryos showed that the majority of cells in pre-streak and early-streak embryos gave rise to cells in multiple germ layers40, 41,42, 43. Recently, cells cultures from post-implantation embryos (e5.5–e6.5) were clonally derived and maintained in vitro, and shown to be pluripotent by teratoma and embryoid body assays 44, 45. These findings demonstrated that the epiblast progeny of the ICM retained pluripotency until lineage commitment during gastrulation, at which time lineage commitment must occur rapidly. Molecular genetic experiments revealed that mESC lineage commitment was blocked by high levels of Nanog (all lineages) 3, 46 and perturbed by ectopic expression of Tcl1 (neurectoderm), Esrrb (meso/endoderm), Tbx3 (mesoderm), and Sox2 (mesoderm) 6. These observations support a requirement for a mechanism to limit the activities of stem cell self renewal circuitries for timely differentiation of pluripotent cells in embryos. Our data have revealed that Tcf3 provides a fundamental molecular mechanism that could accommodate this need, and Tcf3 protein expression in the epiblast is consistent with this model (unpublished observations). Thus, in contrast to the model proposed by Cole et al in which Tcf3-repressor stimulates differentiation 13, we propose Tcf3 function is analogous to that of a current limiter in electronic circuits as Tcf3 must prevent over-activation of transcriptional circuits promoting pluripotent cell self renewal (Figure 7).
The cellular consequences of loss of Tcf3 in mESC have been described here. Effects on embryonic development have been described by us previously 30; however, our previous studies did not examine effects on pluripotency of cells because the identity of pluripotency factors was largely unknown at the time. It will be interesting to examine the role of Tcf3 inhibition of self renewal circuits for normal embryonic development. In addition it will be important to determine if absence of Tcf3 will provide a benefit to the induction of pluripotent cells by reprogramming cells from adult mammals 16–18. Given its effects on Nanog and Oct4 downstream target genes, and the Lif-independent self renewal of TCF3−/− mESC, one would anticipate that absence of Tcf3 could provide a selective advantage to cells undergoing reprogramming. As such, inhibition of Tcf3 could promote more efficient reprogramming by stabilizing an epigenetic program of pluripotent cell proliferation.
We would like to thank members of the Merrill Lab (Jackson Hoffman, Chun-I Wu and Travis Leonard) for helpful discussion, and thank Jackson Hoffman and Danielle DeWaal for their special contributions to Lif dose response assays. Also deserving our gratitude are our colleagues in the Biochemistry and Molecular Genetics Department at UIC, most notably the late Dr. Rob Costa, for generously sharing his enthusiasm, reagents and equipment, and members of the UIC Core Genomics Facility and UIC Flow Cytometry Facility for their expert assistance with experimental methods. This work was supported by funds awarded to B.J.M. from the American Cancer Society (IL#06–45 and #RSG GGC 112994), the Stem Cell Research Foundation, and The Schweppe Foundation.
Fei Yi: Conception and Design, Collection of data, Data analysis and interpretation, Manuscript writing
Laura Pereira: Conception and design, Provision of study materials
Brad Merrill: Concept and design, Financial support, Data analysis and interpretation, Manuscript writing