|Home | About | Journals | Submit | Contact Us | Français|
The observation that Tcf3 (MGI name: Tcf7l1) bound the same genes as core stem cell transcription factors, Oct4 (MGI name:Pou5f1), Sox2 and Nanog, revealed a potentially important aspect of the poorly understood mechanism whereby Wnts stimulate self renewal of pluripotent mouse embryonic stem (ES) cells. Although the conventional view of Tcf proteins as the β-catenin-binding effectors of Wnt signaling suggested Tcf3-β-catenin mediated activation of target genes would stimulate ES cell self renewal, here we show that an antagonistic relationship between Wnt3a and Tcf3 on gene expression is important for regulating ES cell self renewal. Genetic ablation of Tcf3 replaced the requirement for exogenous Wnt3a or GSK3-inhibition for self renewal of ES cells, demonstrating that inhibition of Tcf3-repressor is the necessary downstream effect of Wnt signaling. Interestingly, the molecular mechanism underlying Wnt’s effects required both Tcf3-β-catenin and Tcf1-β-catenin interactions, as they each contributed to Wnt stimulation of self renewal and gene expression. Finally, the combination of Tcf3 and Tcf1 was necessary to recruit Wnt-stabilized β-catenin to Oct4 binding sites in ES cell chromatin. These results elucidate the molecular link between the effects of Wnt and the regulation of the Oct4/Sox2/Nanog network.
The Oct4/Sox2/Nanog core self renewal circuit was defined primarily by experiments showing physical interactions between proteins1, 2, substantially overlapping sites of chromatin occupancy3, 4, and similar effects in loss of function experiments5–9. Several data indicated that Tcf3 was an integral component of the Oct4/Sox2/Nanog self renewal circuit. ChIP-seq experiments showed that of the 1369 known genes identified as Tcf3-bound, 1173 were also bound by Oct4, and 942 were bound by all four factors (Tcf3, Oct4, Sox2, and Nanog)10. The co-occupancy of Tcf3 with Oct4, Sox2 and Nanog on a genome-wide scale was also independently determined with ChIP-chip experiments11, 12. Whereas knockdown of Nanog or Oct4 caused largely similar effects on gene expression4, the ablation of Tcf3 caused strongly opposite effects on gene expression compared to Nanog knockdown and Oct4 knockdown13.
One intriguing aspect of Tcf3’s function in stem cells has been its potential role as a downstream effector of the canonical Wnt signaling pathway. An important consequence of Wnt ligand binding its receptor is the inhibition of GSK3 phosphorylation of β-catenin14. This stabilizes β-catenin from ubiquitin-mediated degradation and promotes signaling by allowing β-catenin to form transcription-activating complexes with DNA-binding proteins, most notably Tcf factors15–17. Treating cells with Wnts18 or blocking GSK3 activity by genetic ablation19 or small molecule inhibitors 18, 20, 21 all effectively stimulated ES cell self renewal. Indeed, inhibition of GSK3 combined with inhibition of ERK activity in so-called two-inhibitor (2i) conditions was sufficient to promote ES cell self renewal in the absence of serum and extrinsic cellular signals20, 21.
The role of Tcf-β-catenin complexes as downstream mediators of Wnt and GSK3 effects has been complicated by the facts that GSK3 affects many cellular activities in addition to destabilizing β-catenin, and it is difficult to experimentally control for the broad effects of GSK3 inhibition. As such, it is not clear what biochemical activities downstream of Wnt and GSK3-inhibtion were most important for ES cell self renewal; however, previously published examination suggested three intriguing hypotheses: either, 1. the effects of Wnt-stimulated β-catenin needed to function through an Oct4-β-catenin interaction and Tcf-β-catenin interactions only secondarily affected self renewal22, 2. Tcf3-β-catenin activator complexes directly stimulated self renewal and the expression of self renewal genes11, or 3. Tcf3 transcriptional-repressor activity inhibited self renewal and was suppressed by Wnt13, 23.
To test these hypotheses, a series of cell-based experiments were completed. First, the ability of Wnt signaling to replace LIF and to support self renewal in Tcf3+/+ cells was confirmed by showing that Oct4 promoter activity and protein levels were maintained in ES cells by adding recombinant Wnt3a to LIF deficient cultures (Supplementary Fig. S1a,b). This effect required an intact Oct4/Sox2/Nanog circuit, as Wnt3a was unable to rescue LIF deficiency in cells genetically lacking Nanog (Supplementary Fig. S1c,d). Since a similar experiment using Tcf3−/− would not be informative because these cells self renew in the absence of exogenous LIF even without Wnt stimulation13, we tested the possibility that forced overexpression of Tcf3 could inhibit self renewal by using a doxycycline-regulated system to increase Tcf3 levels in Tcf3+/+ cells (Supplementary Fig. S2a)24. If the hypothesis of a Tcf3-β-catenin transactivator complex was correct, one would expect that ectopic expression of Tcf3 would enhance ES cell self renewal upon Wnt3a stimulation. Instead, a twofold increase in Tcf3 protein efficiently blocked self renewal, as no alkaline phosphatase (AP)+ ES colonies were present four days after induction of Tcf3 in +LIF cultures (Fig. 1a). Flow cytometry of propidium iodide-stained cells showed a conversion from the high S-phase, ES cell cycle profile to a G1-enriched, somatic cell cycle profile within 72 hours of Tcf3 overexpression (Fig. 1a). Wnt3a sharply inhibited the differentiation-promoting effects of Tcf3 by restoring AP+ colonies and an ES cell cycle profile to Tcf3-overexpressing cells in both +LIF and LIF-deficient conditions (Fig. 1a,b); however, colony formation of Wnt3a-treated, Tcf3-overexpressing cells was diminished relative to the non-overexpressing controls. These results best support the hypothesis that Tcf3 represses self renewal and Wnt3a antagonizes Tcf3-repression.
Recently, Tcf-β-catenin-independent effects were suggested to provide a mechanism whereby inhibition of GSK3 stimulated the self renewal of ES cells21, 22. These conclusions were based primarily on the inability of heterologously expressed Lef1ΔN21 or Tcf4ΔN22 proteins to rescue differentiation defects caused by stable β-catenin. These observations suggested that the self-renewal stimulating effects of Wnt or GSK3-inhibition were independent of Tcf protein function21, 22, which appeared to be inconsistent with the ability of full length Tcf3 to inhibit self-renewal when overexpressed (Fig. 1a,b). To provide a better test of the role of endogenous Tcf proteins in ES cells, we performed experiments using loss of function techniques. First, we tested whether ablation of Tcf3 could replace the requirement for GSK3 inhibition for self renewal in serum free culture conditions21. After a single passage in serum free conditions, Tcf3−/− cells generated a high percentage of AP+ colonies when only PD0325901 (ERK Inh) or both PD0325901 and CHIR99021 (2i) were added to media (Fig. 1c). In contrast, Tcf3+/+ cells generated a high percentage of AP+ colonies only in 2i conditions and differentiated in ERK Inh only (Fig. 1c). While both Tcf3+/+ and Tcf3−/− cells generated AP+ colonies in the presence of CHIR99021 alone (GSK3 Inh) for a single passage (Fig. 1c), both cell lines lost viability and differentiated in GSK3 Inh conditions after being passaged (Fig. 1d). Importantly, Tcf3−/− cells were serially passaged in ERK Inh conditions and maintained a high level of proliferation and AP+ morphology (Fig. 1d). Thus, ablation of Tcf3 effectively replaced the need for GSK3-inhibition in serum free culture conditions, suggesting that antagonism of Tcf3 is a critical downstream effect of GSK3-inhibitors in 2i culture conditions.
To define better how Tcf3 and Wnt signaling functioned together in ES cells, we compared effects of recombinant Wnt3a-treatment (W) or control media (C) on gene expression in Tcf3+/+ and Tcf3−/− ES cells. Principal component analysis (PCA) was used to explore correlations among gene expression microarray datasets for three biological replicates for each of the four different conditions (Tcf3+/+(C), Tcf3−/−(C), Tcf3+/+(W), Tcf3−/−(W)) (see Methods)25. Plotting each replicate according to its principal component (PC)1 and PC2 values effectively separated each of the four conditions while grouping replicates according to their condition (Fig. 2a). Interestingly, Tcf3−/−(C) and Tcf3+/+(W) conditions shared similar PC2 scores, suggesting these two conditions shared a common set of variations that differed in the other conditions. Tcf3−/−(C) and Tcf3+/+(W) had different PC1 scores, suggesting that these two conditions differed from each other based on the variations comprising PC1. Thus, Tcf3-ablation and Wnt3a-treatment cause one set of common changes and one set of different changes.
Since Tcf3-ablation and Wnt-activation both stimulated self renewal of ES cells11, 18, 23, common effects were likely to be mediators of self renewal. To identify these genes, we focused on the 831 genes that were significantly increased (Benjamini-Hochberg adjusted FDR< 0.05; Supplementary Fig. S3a) in any of three comparisons: 1. Increased in Tcf3+/+(W) vs Tcf3+/+(C), 2. Increased in Tcf3−/−(C) vs Tcf3+/+(C), and 3. Increased in Tcf3−/−(W) vs Tcf3−/−(C) (Supplementary Table S1). Each gene was given three scores corresponding to their ranked fold-increase for each of the three comparisons (Supplementary Table S2). To populate a group of common changes, a gene was assigned to group 1 (Red; Fig. 2b,c) when the effect (i.e. ranked fold-increase) caused by Wnt3a in Tcf3+/+ cells (score #1) was closer to the effect caused by Tcf3 ablation (score #2) than to the effect of Wnt3a in Tcf3−/− cells (score #3). For example, Tbx3 was placed in group 1 because its ranked fold-increases for comparisons 1–3 were 29|20|336, respectively; Cdx1 was placed in group 2 with ranked fold-increases of 10|257|17 (Supplementary Table S2). Performing this relatively simple calculation separated all genes into two strikingly distinct groups, which was evident when ranked-fold increase values were used to plot individual genes (Fig. 2b). For group 1 genes, Wnt3a effects on Tcf3+/+ cells displayed a strong correlation (Pearson correlation coefficient ρ= 0.90) when compared to effects of ablating Tcf3 (Fig. 2b, bottom left). Wnt3a had different effects on group 1 genes in Tcf3+/+ cells compared to Tcf3−/− cells (ρ = −0.23) (Fig. 2b, top left). In contrast, for group 2 genes (Blue; Fig. 2b,c), Wnt3a effects on Tcf3+/+ cells displayed a strong correlation (ρ = 0.91) when compared to Wnt3a effects on Tcf3−/− cells (Fig. 2b; top right). Thus, Wnt3a stimulated one set of genes similarly to the ablation of Tcf3 (group 1), and stimulated a second set of genes regardless of Tcf3’s presence (group 2).
When the effects of Wnt3a and Tcf3 were compared to the effects of Oct4 shRNA for the two groups of genes (Fig. 2c), a moderate to strong anti-correlation was observed only for the Tcf3-dependent group 1 genes (ρ = −0.65 for Wnt3a on Tcf3+/+ cells, and ρ= −0.72 for Tcf3 ablation). No relationship with Oct4 was apparent for the Tcf3-independent group 2 genes (ρ values from −0.40 to 0.32), or with the effects of Wnt3a on Tcf3-dependent group 1 genes in the absence of Tcf3 (i.e. in Tcf3−/− cells; ρ = 0.31). Thus, group 1 genes correlated well with regulation by Oct4, consistent with these genes being most relevant for self renewal. In addition, a separate, independent set of 311 genes was selected because they were decreased by Nanog and Oct4 shRNA4; most of these genes (81%) were increased in Tcf3−/− cells compared to Tcf3+/+ cells. Wnt3a stimulation of these genes frequently achieved a higher level (Fig. 2d, green bar, 44% of genes) or nearly the same level (Fig 2d, blue bar; 12% of genes) in Tcf3−/− cells compared to Tcf3+/+ cells. Wnt3a-stimulation of genes rarely achieved a higher level in Tcf3+/+ cells compared to Tcf3−/− cells (Fig. 2d, orange bar, 2% of genes; Supplementary Fig. S3b). The overlapping chromatin occupancy of Nanog and Tcf3 at common genes10, 11 and Wnt3a/Tcf3-regulation of Nanog levels (Fig. 3a,c; Supplementary Fig S2a,b)23 suggested the possibility that Wnt3a/Tcf3 effects could occur secondarily through deregulated Nanog. We tested this by reducing Nanog transiently in Tcf3+/+ cells by siRNA (Fig3a) or stably in Tcf3−/− cells by shRNA (Fig 3b) and by increasing Nanog in Tcf3+/+ cells by stable transfection of a Nanog cDNA (Fig. 3c). Measuring levels of group 1 gene products after either loss or gain of function experiments indicated that Wnt3a/Tcf3 effects did not correlate with increased Nanog levels (Fig. 3d), indicating that increased Nanog levels did not mediate Wnt3a/Tcf3 effects on gene expression. Taken together, these data show that Tcf3-inhibits whereas Wnt3a stimulates a common set of genes, and these common genes are most closely related to the effects of the core Oct4/Sox2/Nanog self renewal circuit.
We were intrigued by the observation that most Wnt3a-stimulated genes did not require Tcf3 for activation, and instead were stimulated to a higher level in Tcf3−/− cells than Tcf3+/+ cells. Consistent with results from Fig. 1, this observation suggested that Tcf3-mediated repression was only inhibited by Wnt3a, and that Tcf3 did not function as an activator of target genes. These observations suggested Tcf3-β-catenin may not be important for actively stimulating Nanog/Oct4/Sox2/Tcf3 co-occupied genes, as suggested by Cole et al 11. To test this possibility and to determine the role of Tcf3-β-catenin interaction in ES cells, we sought to block the Tcf3- β-catenin interaction in ES cells. Given the differentiation-inducing effects of twofold overexpression of Tcf3 (Fig. 1), we reasoned that experiments utilizing heterologous expression of Tcf3 mutants would be prone to neomorphic effects; therefore, we engineered Tcf3ΔN/ΔN ES cells in which nine residues necessary for β-catenin binding were deleted from the endogenous Tcf3 gene product by homologous recombination (Supplementary Fig. S4a,b). When grown in self renewal culture conditions (+LIF, +serum) without exogenous Wnt3a, Tcf3ΔN/ΔN ES cells were indistinguishable from Tcf3+/+ cells in terms of cell proliferation, cell cycle kinetics, and expression of Tcf3, Oct4, and Nanog (Supplementary Fig.S4c, S5c). Interestingly, Wnt3a stimulation of AP+ colony formation in the absence of LIF was reduced, but not eliminated, in Tcf3ΔN/ΔN cells (56% for Tcf3ΔN/ΔN compared to 77% for Tcf3+/+) (Fig. 4a) and, treating Tcf3−/− cells with Wnt3a also stimulated AP+ colony formation (from 70% to 86%) (Fig. 4a). Similar results were obtained with experiments performed using serum-free, 2i conditions (Supplementary Fig. S4d). Moreover, ablation of the Tcf3-β-catenin interaction diminished but did not eliminate the ability of cells to activate Tcf-β-catenin responsive promoter reporters, SuperTOPFlash, Nanog, and Nr5a2 (also called LRH-1) (Fig. 4b). These data are not consistent with a previously described mechanism11 suggesting a key role for Tcf3-β-catenin as an activator. Moreover, the only partial inhibition of Wnt3a stimulation of self renewal and target gene activation in Tcf3ΔN/ΔN cells revealed that Wnt3a stimulation of self renewal involves a combination of Tcf3-β-catenin dependent and previously uncharacterized Tcf3-independent mechanisms.
The possibility that another Tcf protein could contribute Tcf3-independent effects of Wnt3a was supported by several observations. First, the highly conserved HMG domain shared by Tcf proteins allows all Tcf proteins to bind to the same sequences as Tcf326, 27. The more severe phenotypes of double mutant mouse embryos (Tcf1−/−; Tcf4−/−, Tcf1−/−; Lef1−/− and Tcf3cKO, Tcf4−/−) compared to single mutants demonstrated that one Tcf protein can compensate for the absence of a family member, thus indicating the conserved HMG domains allow Tcf family members to affect an overlapping set of target genes28–30. Finally, since the SuperTOPFlash reporter is specific for Tcf proteins31, reporter activation in Tcf3−/− and Tcf3ΔN/ΔN cells showed that another endogenous Tcf protein was present in ES cells, and that this Tcf was capable of activating target genes.
To determine which Tcf protein was activating Tcf3-target genes, we first measured levels of Tcf gene products in ES cell lines. Tcf3 mRNA was the most abundant family member in Tcf3+/+ and Tcf3ΔN/ΔN cells (Fig. 5a), Tcf1 and Tcf4 were also detected at levels that we reasoned could have a biological effect (Fig. 5a). The ability of each Tcf protein to activate gene expression in response to stable β-catenin was tested by measuring the activation of SuperTOPFlash (Fig. 5b). Tcf1 and Lef1 displayed a similar activation of Tcf-β-catenin signaling, whereas Tcf3 and Tcf4 both repressed SuperTOPFlash and Nanog reporter activity in ES cells (Fig. 5b, Supplementary Fig. S6b). Since, Tcf1 was the best candidate to function endogenously in ES cells to activate Tcf3-repressed genes in response to Wnt-stabilized β-catenin, we used stable lentiviral shRNA knockdown to inhibit the Tcf1 gene product in each Tcf3+/+, Tcf3−/−, and Tcf3ΔN/ΔN cells. This approach was effective at reducing Tcf1 mRNA and protein levels (Fig. 5c), and it affected >90% of cells for each cell line as indicated by GFP reporter gene expressed from the lentiviral shRNA vector (Supplementary Fig. S5a). Control cell lines harboring a scrambled shRNA sequence (SCRM) were also generated for each Tcf3 line. In the absence of exogenous Wnt3a, each of the Tcf1 shRNA cell lines were similar to corresponding SCRM cell lines in terms of cell proliferation, cell cycle kinetics, and expression of Tcf3 and Oct4 (Fig. 5c, Supplementary Fig. S5b,c). Thus, Tcf1 knockdown did not cause an apparent cellular defect under normal self renewal conditions, without the addition of exogenous Wnt to the media. In contrast, Tcf1 shRNA substantially reduced transcriptional activation by stable β-catenin in each Tcf3+/+, Tcf3−/−, and Tcf3ΔN/ΔN ES cells (Fig. 5d). Note that promoter activity was higher in the absence of Tcf3 (in Tcf3−/− cells) (Fig. 5d) without an increase in Tcf1 or Lef1 mRNA or protein or β-catenin protein (Fig 5a, Supplementary Fig. S6a), which was consistent with Tcf3 functioning as a repressor and directly interfering with Tcf1-β-catenin mediated transcriptional activation. Interestingly, whereas Tcf3ΔN/ΔN; Tcf1 shRNA and Tcf3+/+; Tcf1 shRNA cells displayed very low levels of activity, the Tcf3−/−; Tcf1 shRNA cells displayed promoter activity comparable to Tcf3+/+; SCRM cells. These data suggest that even residual levels of Tcf1 remaining after shRNA knockdown (Fig. 5c), and possibly the low endogenous levels of Lef1 (Fig. 5a,b), were sufficient to activate SuperTOPFlash when Tcf3-repression is eliminated by genetic ablation.
A role for endogenous Tcf1 in Wnt-stimulation of self renewal was tested with colony formation in serum-containing media lacking LIF. Without exogenous Wnt3a, Tcf1 shRNA and SCRM cell lines differentiated similarly to form primarily AP-negative colonies in Tcf3+/+ and Tcf3ΔN/ΔN cells and primarily AP+ colonies in Tcf3−/− cells (Fig. 6a). When exogenous Wnt3a was added, Tcf1 shRNA decreased the generation of AP+ colonies by half for both Tcf3+/+ and Tcf3ΔN/ΔN cells (Fig. 6a). In contrast, Tcf1 shRNA had a less substantial effect on Wnt3a treated Tcf3−/− cells, which still generated AP+ colonies from 68% of Tcf1 shRNA-containing cells. To confirm that self renewal effects were caused by Tcf1 knockdown and not off target effects, transfection experiments using siRNA molecules targeting a different Tcf1 sequence were performed; Tcf1 siRNA caused a similar 50% reduction in AP+ colonies for Tcf3+/+ and Tcf3ΔN/ΔN cells compared to control (Supplementary Fig. S5d,e). Considering both the clonal nature of the assay using shRNA cell lines and that 9% of Tcf3ΔN/ΔN; Tcf1 shRNA cells plated were GFP-negative (Supplementary Fig. 5a), the 14% of Tcf3ΔN/ΔN; Tcf1 shRNA colonies displaying an AP+ ES colony morphology suggested that Wnt3a stimulation of self renewal primarily relies on the combination of Tcf1 and Tcf3-β-catenin interaction. These results show that endogenous Tcf1 is a key mediator of Wnt stimulation of self renewal in ES cells.
To determine if endogenous Tcf1 was important for Wnt-effects on Tcf3-regulated genes, we measured expression of several Wnt3a-stimulated and Tcf3-repressed genes by qPCR. Expression levels of endogenous genes were measured for 12 conditions (+/− Tcf1 shRNA and +/− Wnt3a treatment in each Tcf3+/+, Tcf3−/−, and Tcf3ΔN/ΔN ES cells). The relative level of each gene is shown by heat map (Fig. 6b); numerical values and standard deviations are shown in Supplementary Table S5. For previously characterized Wnt/Tcf target genes (T, Cdx1, Axin2), Wnt3a stimulated high expression levels in Tcf3+/+ and Tcf3−/− cells, and much lower levels in Tcf3ΔN/ΔN cells (Fig. 6b, top). In each Tcf3+/+, Tcf3−/− and Tcf3ΔN/ΔN cells, Tcf1 shRNA reduced the level of Wnt3a stimulation. These results were consistent with Axin2, T, and Cdx1 being predominantly regulated by Tcfs and Wnt signaling in ES cells. Moreover, they showed that Tcf1 was necessary for full Wnt3a-stimulation of gene expression. Genetically blocking Tcf3-β-catenin interaction (in Tcf3ΔN/ΔN ES cells) inhibited both Wnt3a-mediated suppression of Tcf3-repressor activity and Tcf1-activator activity.
To determine if Wnt3a-effects on self renewal (Fig. 1, ,4,4, ,6a)6a) were caused by the combined activities of Tcf1 and Tcf3 described above, expression of several genes previously implicated in ES cell self renewal was examined. These genes behaved in one of three manners with respect to the 12 conditions (Fig. 6b, bottom). First, Gpa33, Tcl1, Nanog, Prdm14, Jam2, Nr6a1, Nr5a2, Eomes, Kit, and Tnfrsf19 were increased both by Wnt3a and by Tcf3 ablation. Tcf1 shRNA decreased the Wnt3a-stimulation of these genes in each Tcf3+/+, Tcf3−/−, and Tcf3ΔN/ΔN cells, demonstrating a role for Tcf1 in Wnt3a activation of gene expression. Second, Esrrb, Tbx3, Klf4, Nodal, and Nfib were increased by Tcf3 ablation, but not increased by Wnt3a. Interestingly, Tcf1 shRNA reduced the levels of these genes, suggesting a Wnt3a-independent role for Tcf1. However, the effect of Tcf1 on expression was small and did not correspond to a cellular effect on self renewal (Fig. 6a). Third, Oct4, Sox2, Rif1 and Mycn gene products were only modestly (<1.5 fold) increased by any combination of treatments. As expected, Tcf1 shRNA did not substantially affect expression of these genes. Together, these data were consistent with Tcf3 repression of endogenous genes being inhibited by Wnt through a combination of Tcf3-β-catenin-interaction-dependent and Tcf1-dependent mechanisms, and full Wnt-stimulation of genes requires both mechanisms.
Finally, to determine if the mechanism for Wnt stimulation of self renewal involves β-catenin recruitment to sites of Tcf/Oct4/Sox2/Nanog co-occupancy in ES chromatin, a series of ChIP-qPCR experiments was performed. The genomic binding sites for analysis were selected from previously published datasets1, 4, 10–12. Criteria for choosing common sites (Tcfcp2l1, Mycn, Nodal, Oct4, Sox2, Tcf3, Zic5) included: multiple studies reporting a high ChIP signal for Oct4 and Nanog occupancy, a known relevance of the gene to ES cell self-renewal, and reproducible enrichment of the site in our ChIP-qPCR experiments. In addition, previously characterized Tcf-β-catenin responsive sites (Cdx1 and Axin2) were included as positive controls for Tcf3-binding1, 32. Tcf3 and Oct4 antibodies both enriched all common sites as well as the Tcf3-binding controls (Fig. 6c,d), confirming the high level of overlapping binding discovered with ChIP-chip and ChIP-seq techniques10–12. Interestingly, the relative amount of Oct4 occupancy on chromatin was increased for 8 of 10 sites in Tcf3−/− cells compared to Tcf3+/+ cells (Fig 6d), suggesting the possibility that Oct4 and Tcf3 compete for occupancy. This observation was important as it provided an excellent context to test the relative contribution of Oct4 and Tcf proteins to the recruitment of Wnt-stabilized β-catenin to target genes. If Oct4-β-catenin complexes mediate Wnt-stimulation of self renewal22, then the increased Oct4 occupancy should enhance β-catenin recruitment in Tcf3−/− cells. Instead, β-catenin occupancy at common sites was reduced in Tcf3−/− cells, and was further reduced by Tcf1 shRNA for all 10 sites (Fig 6e). Taken together, these ChIP data do not support a substantial role for Oct4 directly binding to and recruiting β-catenin to chromatin in Wnt-stimulated ES cells. Instead they show Tcf3-β-catenin and Tcf1-β-catenin interactions recruit Wnt-stabilized β-catenin to chromatin.
The data shown here are most consistent with Tcf proteins being the critical regulators of Wnt’s stimulation of pluripotent cell self renewal. That said, shRNA knockdown of Tcf1 did not completely eliminate Wnt3a-stimulation of AP+ colony formation in Tcf3ΔN/ΔN cells, as a small percentage of these Wnt3a treated cells formed AP+ colonies (Fig. 6a). This may have been caused by incomplete Tcf1 knockdown or activity of endogenous Lef1; however, the low levels of reporter activity (Fig. 5d) and endogenous gene expression (Fig. 6b) in Tcf3ΔN/ΔN; Tcf1 shRNA cells makes this possibility appear unlikely. Alternatively, a Tcf-independent effect of Wnt-stabilized β-catenin22 may contribute to stimulating self renewal. A primary role for Tcf-independent effects of β-catenin was suggested by Kelly et al., as overexpression a ΔNTcf4 protein failed to rescue differentiation defects in GSK3 double knockout ES cells 22. While our results do not rule out the possibility of a secondary effect of Tcf-independent mechanisms, they strongly suggest that Tcf proteins play the primary role because: Overexpressing full length, wild-type Tcf3 blocked ES cell self renewal (Fig 1a,b). Self renewal was restored in Tcf3-overexpressing cells with exogenous Wnt3a (Fig. 1 a,b). Conversely, ablating Tcf3 replaced the need to inhibit GSK3 for ES cell self renewal in serum free conditions (Fig. 1c,d). The common effects of Wnt3a and Tcf3 ablation of gene expression correlated with Nanog/Oct4 regulation, whereas Tcf3-independent effects of Wnt3a did not (Fig. 2). The effect of Wnt3a on Tcf3ΔN/ΔN; Tcf1 shRNA AP colony formation was small compared to Tcf3 ablation (Fig 6a). Finally, the Tcf-independent mechanism of β-catenin was suggested to work through an Oct4-β-catenin interaction22, but Oct4 did not recruit β-catenin to chromatin in the absence of Tcf3 and Tcf1 (Fig. 6c–e). We conclude that Wnt stabilized β-catenin stimulates ES cell self renewal primarily by inhibiting Tcf3-mediated repression of target genes.
The role for Tcf3 as a limiter of ES cell self renewal appears to be unique among Tcf factors. Moreover, endogenous Tcf1 opposed Tcf3-repression and was necessary for the complete Wnt3a-stimulation of target genes and self renewal (Fig 6a,b). In light of the need to suppress Tcf3-repression for LIF-independent self renewal, it is interesting to consider why overexpression of ΔNLef121, ΔNTcf122 and ΔNTcf422 each failed to inhibit self renewal or failed to rescue differentiation defects caused by β-catenin stabilization. We believe the simplest explanation is that Tcf3 possesses a biochemical activity not shared by other Tcf/Lefs, and this transcriptional repressor activity inhibits ES cell self renewal. Expression of another Tcf/Lef would effectively inhibit Tcf3-repression by competing for DNA binding via virtually identical HMG domains shared by all Tcf/Lefs. As such, overexpression of Tcf1, Tcf4 or Lef1 could have a similar effect as downregulating or inhibiting Tcf3. This mechanism is consistent with the observation that endogenous Tcf1 was necessary for maximal Wnt3a-suppression of Tcf3-repressed target genes (Fig 6b). Comparing biochemical properties of Tcf/Lef proteins in ES cells has potential to reveal a unique co-repressor that limits activity of the self renewal circuitry in pluripotent cells.
Fragments of the murine Nanog promoter and 5’ UTR encompassing genomic region chr6:122652598–122657800 were amplified by PCR from a mouse genomic BAC (clone RP23-406B15, BACPAC Resources, Oakland, CA) and cloned into a pGL3b luciferase reporter vector (Promega, Madison, WI).
One thousand ES cells were plated on each well of gelatinized 6-well plate and media were replaced daily. Colonies were stained for alkaline phosphatase (AP) activity after 4 days. Briefly, cells were washed once with PBS, fixed in citrate-acetate-formaldehyde (18mM NaCitrate, 9mM NaCl, pH 3.6, 65% acetone, 3.7% formaldehyde) for 45 seconds, washed three times with water, and stained with NBT/BCIP solution (Promega). For each well, 100 random colonies were scored as AP activity positive (AP+) or negative (AP−). Experiments were performed in duplicate and results were expressed as means +/− standard deviations.
Single cell suspensions were made by trypsin and resuspended in 100 µl PBS. Cells were fixed by drop-wise addition of 1ml cold (−20°C) 70% ethanol while vortexing. Cell cycle profile was checked by analyzing DNA content. Briefly, fixed cells were rehydrated and washed twice with 10 ml PBS before finally being suspended in 200 µl PBS. Cells were stained by incubating with of 5 µg/ml propidium iodide (Sigma) and 1 mg/ml RNase A for 20 min at room temperature and analyzed on Beckman-Counter Cell Lab Quanta SC MPL using Cell Lab Quanta Collection program (Beckman Coulter).
For each condition, three independent Tcf3+/+ and Tcf3−/− ESC cultures were treated with or without 50ng rmWnt3A (Recombinant mouse Wnt3A, #1324-WN, R&D System) for 24 hours. Total RNA were extracted by Trizol reagent (Invitrogen, Carlsbad, CA) and purified using RNeasy Mini Kit (Qiagen, Hilden, Germany). Quality of RNA sample was assessed by spectrophotometer (A260/A280 between 1.9 and 2.1). 10 µg (1 µg/µl) total RNA of each sample were used for microarray analysis. Labeling of the RNA, hybridization onto Affymetrix mouse GeneChip 430 2.0 microarrays (Affymetrix, Santa Clara, CA), and scanning of samples were performed by the Functional Genomics Facility, University of Chicago. Raw data were normalized by quantiles and summarized by robust multiarray average. Pairwise comparisons between groups of three replicates were performed, and the t test was used to rank each probe set on the basis of its likely statistical significance using the Comparative Marker Selection module of GenePattern software (Broad Institute, Cambridge, MA). Probe sets displaying a significant difference as defined by a Benjamini-Hochberg adjusted t test p value < 0.05 were used for subsequent analyses. Principal component analysis (PCA) was performed using the PCA module of GenePattern software. All microarray data points were subjected to PCA to create a set of principal experimental components to identify relevant patterns or variations across the four conditions: Tcf3+/+ (C), Tcf3−/− (C), Tcf3+/+ (W), Tcf3−/− (W). The analysis identifies patterns by finding principal components (PC) that account for variance in data points among all samples. Each PC thus represents a combination of gene expression values, and each PC is mutually uncorrelated and orthogonal. The first two PC scores (PC1 and PC2) resolved 99% of variance amongst all samples. A two dimension plot of PC1 and PC2 was generated using the PCAviewer module of GenePattern.
Total RNA was isolated with TRIZOL reagent (Invitrogen, Carlsbad, CA), and 3.0 µg total RNA of each sample was used as template for cDNA synthesis. Reverse transcription was carried out with oligo-dT primers (0.5 mg/ml) using SuperScript™ III First-Strand cDNA Synthesis kit (Invitrogen). 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: 1× 95°C for 2 min; 40× 95°C for 30 sec; 60°C for 30 sec. 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 85% or higher. Each sample was analyzed in duplicate, and results were expressed as means +/− standard deviations. Primer sequences used for qPCR are specified in Supplementary Table S5.
Transient knockdown of Nanog expression in wild-type ESC was achieved as previously described by transfection of siRNA targeting Nanog or SCRM siRNA23. Twelve hours after siRNA transfection, cells were treated with or without 50 ng/ml Wnt3a for 24 hours. Overexpression of Nanog in Tcf3+/+ ESC was achieved by transfection of the ppyCag-Nanog-IP expression vector8 or empty ppyCagIP vector control and G418 selection for transfectants. Nanog knockdown in Tcf3−/− ESC was performed by stable integration of a pSuper shRNA vector (OligoEngine) designed with a target sequence directed to the Nanog transcript (GAGACAGTGAGGTGCATAT) as previously described33 or empty pSuper vector control. The pSuper vectors were linearized and electroporated for stable insertion into the genome. After puromycin selection, colonies were picked and clonally expanded to generate individual cell lines. RNA isolation and quantitative real-time PCR were performed as described above. Verification of Nanog knockdown or overexpression was performed by SDS-Page gel electrophoresis and western blotting with Nanog specific antibodies (Bethyl Labs A300-397A, CosmoBio RCAB0002P-F, and Abcam ab21603) and by quantitative real-time PCR.
Cells were plated 24 hours before transfection on gelatinized 24-well plates at a density of 1×105 cells/ml. Transient transfection of DNA vectors were conducted by Lipofectamine 2000 (Invitrogen) following the manufacturer’s protocol. 24 hours after transfection, cells were lysed by 1X passive lysis buffer (Promega). Luciferase activities were measured by Clarity luminometer (Bio-Tek) with Dual-luciferase reporter assay system (Promega). Each transfection was carried out in duplicate. Relative activity was calculated as the ratio of the reporter plasmid Firefly luciferase activity to Renilla luciferase activity (pRL-CMV), which was co-transfected as control.
Tcf1 shRNA sequence was subcloned into lentiviral vector pLVTHM. Lentivirus were produced following previously established procedures. Briefly, 293T cells were plated 3 days before transfection. Tcf1 shRNA pLVTHM plasmid were mixed with envelop plasmid (pMD2G) and packaging plasmid (psPAX2) in desired ratio and transfected 293T cells by calcium phosphate. Fresh media were changed 14–16 hours post-transfection and supernatant containing lentivirus were collected 12 hours later. Supernatants can be harvested 3 times every 12 hours. Titers of virus were determined using 293T and ES cells by FACS analysis for percentage of GFP expressing cells. Three targeting sequences for Tcf1 were used to generate shRNA lentivirus (#1 TCTCCACTCTACGAACATT; #2 GGAGCTGCAGCCATATGAT; #3 GCCAAGGTCATTGCTGAGT) and tested for Tcf1 knockdown. The #3 sequence (GCCAAGGTCATTGCTGAGT) had the highest Tcf1 knockdown efficiency and was used to generate stable Tcf1 ES cell lines in this study. A scramble shRNA sequence (GTCACGATAAGACAATGAT) was used to generate control lentivirus (SCRM). Clonal Tcf1 knockdown ES cell lines and SCRM control lines were established by single colony picking and following expansion. Efficiency of Tcf1 knockdown was confirmed by qPCR and Western blot.
Approximately 2 × 107 ES cells were cultured and processed for each ChIP experiment. Cells were chemically crosslinked by the addition of one-tenth volume of fresh formaldehyde solution (11% formaldehyde, 0.1 M NaCl, 1 mM EDTA, 0.5 mM EGTA, 50 mM Hepes pH 8) for either 10 minutes at room temperature (Tcf3) or overnight at 4°C (Oct4, β-catenin). The reaction was inactivated by the addition of 2.5 M Glycine (0.125 M final concentration) and incubated at room temperature for 10 minutes. Cells were rinsed twice with cold 1xPBS, harvested using a silicon scraper, and flash frozen in liquid nitrogen. Cells were stored at −80°C until the day of use.
Cells were resuspended in lysis buffer LB1 containing protease inhibitors (50 mM Hepes pH 7.7, 140 mM NaCl, 1 mM EDTA, 10% Glycerol, 0.5% NP-40, 0.25% Trtiton-X-100), incubated for 10 minutes on ice and centrifuged at 4°C for 10 minutes at 2500 rpm. The cell pellet was resuspended in LB2 (200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 10 mM Tris pH 7.5), incubated for 10 minutes on ice and centrifuged at 4°C for 10 minutes at 2500 rpm. Finally the cell pellet was resuspend in sonication buffer LB3 (1 mM EDTA, 0.5 mM EGTA, 10 mM Tris pH 7.5, 100 mM NaCl, 0.1% Na-deoxycholate) and sonicated to solubilize and shear crosslinked DNA. Sonication was performed using a Branson sonicator for 20 µ 30 second pulses (60 second pause between pulses) at a 60% amplitude. Triton-X-100 was added to the lysates after sonication to a final concentration of 1%. Cellular debris was removed by centrifugation. Samples were kept on ice at all times.
The resulting chromatin extract containing DNA fragments with an average size of 500 bp were immunoprecipitated overnight at 4°C using magnetic beads (Protein G Dynabeads, Invitrogen) that had been pre-incubated with 5 µg of the appropriate antibody. The antibodies used for immunoprecipitation were Tcf3 (Merrill Lab), Oct4 (Santa Cruz sc8628) and β-catenin (Invitrogen 71–2700). Beads were washed 3 times with RIPA ChIP buffer (50 mM Hepes, 1mM EDTA, 0.7% Na-deoxycholate, 1% NP-40, 0.5 M LiCl) and 1 time with TBS (50 mM Tris, 150 mM NaCl, pH 7.6). Bound complexes were eluted from the beads by heating at 65°C with occasional vortexing, and crosslinks were reversed by overnight incubation at 65°C. Whole cell extract DNA (reserved from the sonication step) was also treated for crosslink reversal. Eluted protein-DNA complexes were treated with RNase A (0.2 µg/µl) and proteinase K (0.2 µg/µl). Following phenol-chloroform extraction and ethanol precipitation, DNA was dissolved in Tris-EDTA.
Stealth siRNA duplexes (Invitrogen) designed against mouse Tcf1 (AGAGACUGUCUACUCUGCCUUCAAU) or a control siRNA (GGAAGACUAGAGGCGGUCAUGAGUU) were used to specifically downregulate Tcf1 mRNA expression levels. For transfection, ES cells were trypsinized, pelleted by centrigugation and resuspended in media without antibiotics. For each phenotype, 100,000 cells per well were combined with Lipofectamine 2000 transfection complexes of control or Tcf1 siRNA (50 nM) following the manufacturer’s protocol (Invitrogen) and plated on gelatin on a 6-well dish. Twelve hours after transfection cells were split onto new plates and used in a colony differentiation assay.
The complete microarray dataset was submitted to NCBI GEO repository and is available as GSE27455.
We thank Kathrin Plath for her helpful discussions and thoughtful comments on the manuscript, and Dr. Minoru Ko and Dr. Akira Nishiyama for the inducible Tcf3 overexpression cell line. The work was funded by grants from the American Cancer Society (RSG GGC 112994, B.J.M.), the National Institutes of Health (R01-CA128571, B.J.M. and R01-GM065400, D.R.L.) and the Howard Hughes Medical Institute (D.R.L.).
AUTHOR CONTRIBUTIONSAll authors designed and analysed experiments. F.Y, L.P. J.A.H, B.S. and C.M.Y. performed experiments. F.Y., L.P. and B.J.M. wrote the manuscript. B.J.M and D.R.L. supervised the project.