|Home | About | Journals | Submit | Contact Us | Français|
Wnt signaling is intrinsic to mouse embryonic stem cell self-renewal. Therefore it is surprising that reprogramming of somatic cells to induced pluripotent stem cells (iPSCs) is not strongly enhanced by Wnt signaling. Here, we demonstrate that active Wnt signaling inhibits the early stage of reprogramming to iPSCs, while it is required and even stimulating during the late stage. Mechanistically, this biphasic effect of Wnt signaling is accompanied by a change in the requirement of all four of its transcriptional effectors: Tcf1, Lef1, Tcf3, and Tcf4. For example, Tcf3 and Tcf4 are stimulatory early but inhibitory late in the reprogramming process. Accordingly, ectopic expression of Tcf3 early in reprogramming combined with its loss-of-function late enables efficient reprogramming in the absence of ectopic Sox2. Together, our data indicate that the step-wise process of reprogramming to iPSCs is critically dependent on the stage-specific control and action of all four Tcfs and Wnt signaling.
The generation of induced pluripotent stem cells (iPSCs) from fibroblasts by ectopic expression of Oct4, Sox2, cMyc, and Klf4 established a major landmark in the field of stem cell biology as it allows the establishment of patient-specific pluripotent cells (Maherali et al., 2007; Okita et al., 2007; Takahashi and Yamanaka, 2006; Wernig et al., 2007). The reprogramming process is quite robust in that ectopic expression of the reprogramming factors works on a wide range of differentiated cells to produce iPSCs (Stadtfeld and Hochedlinger, 2010). However, reprogramming to iPSCs is inefficient in that only a few somatic cells of the starting population transition to pluripotency after a latency period of around two weeks (Papp and Plath, 2013). Thus, currently largely unknown events need to occur to achieve reprogramming to the pluripotent state. Indeed, starting cell type, the reprogramming factor combination used, the method of overexpression, and culture conditions all have major effects on the activation of the endogenous pluripotency gene regulatory network and even the epigenetic state of the reprogrammed cells (Papp and Plath, 2013). In this study, we are focusing on the role of Wnt signaling in reprogramming to iPSCs.
The Wnt/β-catenin signaling pathway is intricately linked to the pluripotent state (Clevers and Nusse, 2012). For instance, mouse ESCs secrete active Wnt ligands and autocrine Wnt activity is required to prevent their differentiation (ten Berge et al., 2011), indicating that Wnt signaling is both necessary and sufficient for the self-renewal of these cells. Mouse ESCs can even self-renew efficiently in the absence of serum and extrinsic signals as long as Wnt/β-catenin signaling is stimulated and ERK kinases are inhibited (“2i” culture condition) (Ying et al., 2008).
Canonical Wnt signaling is classically described to function in two states. In the absence of a Wnt ligand, a complex of proteins, including Axin, Apc, Ck1, and Gsk3 stimulates the ubiquitin-mediated destruction of β-catenin (Clevers and Nusse, 2012). In the absence of stable β-catenin, T-cell factor (Tcf) proteins (Tcf1, Lef1, Tcf3, and Tcf4 in mammals) transcriptionally repress Wnt target genes by interacting with co-repressor proteins, such as Groucho or the C-terminal binding protein (Ctbp), and recruiting them to their DNA recognition sites through the HMG domain, which is nearly identical in all Tcfs (Clevers and Nusse, 2012). When a Wnt ligand activates the pathway, the β-catenin destruction complex is inhibited, enabling β-catenin to translocate to the nucleus where it can bind to a conserved domain present near the amino terminal of all Tcfs (Clevers and Nusse, 2012). Upon binding to a Tcf, β-catenin can switch the activity of Tcfs from transcriptional repression to activation by recruiting co-activators, such as CBP (Takemaru and Moon, 2000). Although Tcfs share homologous HMG and β-catenin interaction domains, differences among individual Tcfs cause them to function uniquely within the Wnt pathway. For example, the effect of β-catenin binding can differ, either inducing the classic conversion from a repressor to transactivator for Tcf1 and Lef1, or only inactivating the repressor activity of Tcf3 (BJM, unpublished observation). Thus, individual Tcfs can cause overlapping or diverse effects, depending on how their conserved and unique elements are regulated.
Important understanding of how Wnt signaling affects ESCs has come through the appreciation of diverse effects of Tcfs. Together with core pluripotency transcription factors, Oct4, Sox2, and Nanog, Tcf3 co-occupies many pluripotency genes, including Nanog and Esrrb (Cole et al., 2008; Marson et al., 2008b; Martello et al., 2012; Tam et al., 2008; Yi et al., 2008). Ablation of Tcf3 stimulates Nanog and Esrrb expression, similar to the activation of Wnt/β-catenin signaling (Cole et al., 2008; Martello et al., 2012; Pereira et al., 2006; Yi et al., 2011), and allows self-renewal of ESCs in serum-free conditions without Wnt pathway stimulation (Yi et al., 2011). It is therefore thought that Tcf3 acts exclusively as a transcriptional repressor in ESCs, even in the presence of stable β-catenin. Tcf4 mainly displays similar transcriptional repressor activity as Tcf3, but it is expressed at low levels in ESCs (Pereira et al., 2006; Yi et al., 2011). By contrast, Tcf1 and Lef1 display β-catenin-dependent transcriptional activator activity in ESCs, and endogenous Tcf1 activity counteracts some, but not all, transcriptional repression by Tcf3 (Yi et al., 2011).
The central importance of Wnt signaling and inhibition of Tcf3 for self-renewal of mouse ESC has stimulated investigations into the effects of Wnt signaling on reprogramming to pluripotency. In experiments where somatic cell nuclei are reprogrammed to the pluripotent state upon fusion with ESCs, treating ESCs with exogenous Wnt3a, stabilized β-catenin, or downregulation of Tcf3, each stimulate the efficiency by which somatic cells are reprogrammed (Han et al., 2010; Lluis et al., 2011; Lluis et al., 2008). The effects of Wnt3a or Tcf3 ablation on fusion-mediated reprogramming are substantial, increasing reprogramming efficiency up to thousand-fold (Lluis et al., 2011; Lluis et al., 2008). By contrast, Wnt/β-catenin stimulation or Tcf3 depletion cause only a weak enhancement of reprogramming to iPSCs (Lluis et al., 2011; Marson et al., 2008a). In addition, β-catenin was among the original 24 factors screened by Takahashi and Yamanaka, and its overexpression was found to have no significant effect on iPSC formation (Takahashi and Yamanaka, 2006). Currently it remains unclear how Wnt signaling has such a substantial impact on the self-renewal of mouse ESC and reprogramming by fusion with ESCs, yet causes relatively minor effects on the outcome of iPSC-based reprogramming experiments.
To elucidate how Wnt/β-catenin signaling affects reprogramming to iPSCs, we determined the effects of inhibiting or stimulating Wnt signaling and the requirement for the four Tcfs during different stages of the reprogramming process. Our results demonstrate that early events in reprogramming are stimulated by inhibiting Wnt signaling, whereas late events are stimulated by activating of the pathway. These effects are mediated by differential activities of the four Tcfs, and dynamic manipulation of Tcf3 levels allow for the efficient formation of iPSCs without exogenous Sox2. Our findings showcase that the poor efficiency of reprogramming is at least partially caused by changing molecular requirements in the process, where events promoting one phase are inhibitory for a subsequent phase, calling for further optimization of iPSC technology.
We began elucidating the roles of Wnt signaling in reprogramming to iPSCs by determining whether endogenous Wnt signaling is necessary for the process. In these experiments, we employed mouse embryonic fibroblasts (MEFs) carrying a single tetracycline-inducible polycistronic cassette encoding Oct4, Sox2, cMyc, and Klf4 (inducible OSCK: iOSCK) and a reverse tetracycline transactivator (M2rtTA) transgene (Figure S1A), induced reprogramming by addition of doxycycline, and assessed reprogramming efficiency under various treatments by quantifying colonies positive for expression of the pluripotency factor Nanog. To inhibit endogenous Wnt signaling, iOSCK MEFs were transduced with a retrovirus expressing Dickkopf1 (Dkk1), a secreted ligand and natural antagonist to the Wnt co-receptor LRP5/6 (Mao et al., 2001). Reduction of transcript levels of the Wnt target gene Axin2 and of TOPflash luciferase reporter activity confirmed that ectopic Dkk1 expression efficiently inhibited Wnt signaling (Biechele et al., 2009) (Figure S1B and S1C). Notably, Dkk1 expression greatly reduced the numbers of Nanog positive colonies (Figure 1A), suggesting that endogenous Wnt signaling is essential for the formation of iPSCs. To confirm that the reduction of Nanog positive colonies was due to effects on Wnt signaling, IWP2, a potent small molecule inhibitor of Porcupine, which is necessary for the processing and secretion of all Wnt ligands (Chen et al., 2009), was continuously added throughout the reprogramming process. This independent method of inhibiting endogenous Wnt signaling strongly blocked the formation of Nanog positive colonies (Figure 1B), demonstrating that the production of active Wnt ligands is essential for reprogramming.
The substantial negative effects of inhibiting endogenous Wnt signaling (Figure 1A and 1B) contrasted with the minimal positive effects of exogenously stimulating Wnt reported previously, even no effect when cMyc was included in the reprogramming cocktail with Oct4, Sox2, and Klf4 (Marson et al., 2008a). To determine if an effect of exogenous Wnt signaling could be observed with our experimental set-up, Nanog expressing colonies were measured in reprogramming experiments continuously treated with purified recombinant Wnt3a protein, which strongly activated a TOPflash luciferase reporter (Figure S1C). Consistent with previous reports, we observed no increase in the reprogramming of iOSCK MEFs to Nanog expressing colonies upon Wnt3a treatment (Figure 1C). Moreover, recombinant Wnt3a reduced the number of Nanog expressing colonies by half, indicating that constitutive exogenous Wnt signaling is even inhibitory for the induction of Nanog when reprogramming is performed with our inducible OSCK polycistronic expression cassette.
Since both stimulating and blocking Wnt signaling resulted in an inhibition of iOSCK reprogramming, albeit to different extents, we examined the possibility of a different responses to Wnt signaling during the progression to pluripotency. To test this, we divided the reprogramming process into early, mid, and late stages and observed the effects of Wnt inhibition and stimulation on reprogramming (Figure 1D). Early treatment with IWP2 yielded a two-fold increase of Nanog positive colonies, whereas late treatment strongly reduced reprogramming (Figure 1E). Conversely, Wnt3a addition early resulted in a dramatic inhibition of the formation of Nanog positive colonies, while treatment late increased the number of Nanog positive colonies over two-fold (Figure 1F). The efficiency of reprogramming was not affected when either IWP2 or Wnt3a were added in the middle phase of the process (Figure 1E and 1F). Together, these data demonstrate a biphasic response of the iPSC reprogramming process to Wnt signaling.
The possibility that this biphasic response could be caused by different responses to graded levels of Wnt activity, as previously described by Cosma and colleagues for cell fusion experiments (Lluis et al., 2008), was ruled out by the dose-dependent manner by which IWP2 affected Nanog positive colony formation (Figure S1D–S1F). Moreover, effects on levels of well-established Wnt targets such as Axin2 and Tcf1 confirmed that IWP2 and Wnt3a stimulated and inhibited Wnt signaling, respectively, as expected (Figure S1E and S1F).
We conclude that our experimental approach resolves reprogramming effects that were previously overlooked by continuous treatment of reprogramming cultures with Wnt pathway effectors. Two distinct phases of Wnt response affect two stages of reprogramming, which can be temporally defined as early and late stages. The conclusion that the stimulation of Wnt signaling establishes a strong barrier for early reprogramming events is further supported by the findings that pre-incubation of MEFs with Wnt3a before the start of reprogramming dramatically impaired reprogramming and pre-incubation with IWP2 enhanced reprogramming (Figure S1G). In contrast, the late phase does not only depend on endogenously active Wnt signaling, but also benefits from the ectopic stimulation of the pathway.
Wnt ligands can elicit multiple downstream effects, some of which are independent of the canonical Tcf-β-catenin regulation of target genes. To determine whether the biphasic effects of Wnt signaling on reprogramming to iPSCs were mediated by Tcfs, we depleted each Tcf (Tcf1, Lef1, Tcf3, and Tcf4) during reprogramming by siRNA-mediated knockdown. We also tested the effect of depleting all possible pairs of Tcfs to address the potential for redundancy between family members. The requirement for the Tcfs was first examined during the early stage of reprogramming (Figure 2A, and S2A–S2D). The strongest effects due to loss-of-function of a single Tcf early in the process were seen for Lef1 and Tcf4, respectively (Figure 2B). Their knockdown had opposing effects on the induction Nanog positive colonies; depletion of Lef1 increased and Tcf4 knockdown decreased colony numbers. These effects were magnified when the knockdown of Lef1 was combined with that of Tcf1, or when Tcf4 was depleted together with Tcf3 (Figure 2B). Our findings reveal (i) redundancies among Tcf family members and (ii) an antagonistic effect between two groups of Tcfs early in reprogramming: endogenous Tcf1 and Lef1 are inhibitors while Tcf3 and Tcf4 are enhancers in this phase of reprogramming.
Based on the effects of Wnt signaling on the early reprogramming phase, one would predict that Tcf1 and Lef1 mediate Wnt effects whereas Tcf3 and Tcf4 counteract Wnt effects during this stage. Consistent with this hypothesis, the transcript levels of Wnt target genes Tcf1, Lef1, and Axin2 were significantly increased upon Tcf3 and Tcf4 knockdown early in reprogramming (Figure 2C). To test this hypothesis further, we combined IWP2 treatment with knockdown of Tcf1/Lef1 or Tcf3/Tcf4 early in reprogramming (Figure 2D, S2E, and S2F). The combined knockdown of Tcf1/Lef1 increased the number of Nanog positive colonies in the absence of IWP2, but failed to further increase colony numbers when endogenous Wnt signaling was blocked by IWP2 (Figure 2D). Thus, the effects of Tcf1 and Lef1 early in reprogramming overlap with those seen by inhibiting endogenous Wnt signaling, which is consistent with Wnt/β-catenin-dependent transcriptional activator activities for Tcf1 and Lef1 during this reprogramming phase. Conversely, knockdown of Tcf3/Tcf4 inhibited reprogramming regardless of the presence or absence of IWP2 (Figure 2D). The Wnt target genes Tcf1, Lef1, and Axin2 were upregulated upon Tcf3/Tcf4 knockdown even in the presence of IWP2, i.e. without active Wnt signaling (Figure S2G). Therefore, the inhibitory effect of Tcf3/4 depletion on early reprogramming does not require active Wnt signaling, which is most consistent with transcriptional repressor activities for Tcf3 and Tcf4. Notably, the simultaneous knockdown of all four Tcfs reduced of the number of Nanog positive colonies compared to control (Figure 2E, S2H, and S2I). This result indicates that the mediators of active Wnt signaling, Tcf1 and Lef1, are not the critical targets of Tcf3 and Tcf4 repression during the early stage of reprogramming, as reducing the aberrant Tcf1/Lef1 upregulation observed upon Tcf3/Tcf4 depletion did not rescue the reprogramming efficiency.
To test the role of Tcfs during the late phase of reprogramming, we transfected siRNAs once, on day 6 of reprogramming, and assessed the formation of Nanog positive colonies three days later (Figure 3A, S3A–S3D). Among all siRNA treatments, only the knockdown of Tcf3 and Tcf4, individually or together, consistently enhanced reprogramming (Figure 3B and S3E), indicating that Tcf3 and Tcf4 inhibit late reprogramming events. Because the effect of the Tcf3 and Tcf4 double knockdown was not additive compared to their respective single knockdowns, Tcf3 and Tcf4 likely act in the same pathway. Importantly, the reprogramming enhancement due to Tcf3 or Tcf4 knockdown was nullified when Tcf1 or Lef1 were concurrently depleted (Figure 3B and S3F). Since these data demonstrated that the loss of Tcf3 or Tcf4 requires Tcf1 of Lef1 for a positive effect late in reprogramming, which are typically the mediators of active Wnt signaling, we next tested the requirement of Wnt signaling in this context more directly by combining the Tcf3/Tcf4 knockdown with IWP2 treatment. Our results show that IWP2 prevented the enhancing effect of Tcf3 or Tcf4 depletion late in reprogramming (Figure 3C, green bars). This effect does not appear to be due to a dramatic change in Tcf1 and Lef1 levels (Figure S3G). Together, these findings indicate that depletion of Tcf3 and Tcf4 promotes the late stage of reprogramming through a mechanism that requires Tcf1 or Lef1 as mediators of active Wnt signaling.
While depletion of Tcf1 and/or Lef1 during the late phase did not inhibit reprogramming (Figure 3B), their depletion mitigated the inhibitory effect of Wnt inhibition, i.e. IWP2 treatment, during the late phase of reprogramming even when combined with depletion of Tcf3 and Tcf4 (Figure 3C, orange bars). These data are most consistent with the interpretation that the activity of endogenous Wnt signaling during the late phase is necessary to prevent Tcf1 and Lef1 from becoming potent inhibitors of reprogramming. We speculate that in the absence of active Wnt signaling late in reprogramming, Tcf1/Lef1 are transcriptional repressors at target genes that are essential for the induction of pluripotency. Depleting Tcf1/Lef1 under “no Wnt” conditions would relieve the repressive effect and allow other, alternative pathways to activate these critical genes. Such alternative pathways may also explain why depletion of Tcf1 and Lef1 alone did not inhibit reprogramming as seen in Figure 3B. Although we favor this explanation, it is also possible that residual activity of Tcf1 or Lef1 after siRNA knockdown is enough to fulfill a critical function, which could be addressed in the future by using genetic knockout models.
Taken together, our data uncover different requirements of the Tcfs early and late in reprogramming, which is consistent with the changing role of Wnt signaling between the early and late stages. Early in reprogramming, Tcf3 and Tcf4 stimulate reprogramming and are inhibited by active Wnt signaling mediated by Tcf1 and Lef1. Late in reprogramming, Tcf3 and Tcf4 are inhibitory and regulate the activity of the Wnt signaling pathway. We posit that the distinct activities of individual Tcf factors are responsible for the biphasic effects of Wnt signaling on iPSC reprogramming.
Given the fundamental role of Tcf3 in regulating pluripotency in ESC, we reasoned that elucidating how Tcf3 contributes to the biphasic Wnt signaling effect during reprogramming to iPSCs would provide the greatest insights into mechanisms of the process. First, we determined if overexpression of Tcf3 would affect the dynamics of reprogramming (Figure 4A). Constitutive Tcf3 expression throughout reprogramming reduced the number of cells positive for the surface marker SSEA1, which marks late reprogramming intermediates (Stadtfeld et al., 2008), and also decreased the formation of Nanog and Oct4-GFP positive colonies in a dose-dependent manner (Figure 4B, 4C, S4A, and S4B). Proliferation of the reprogramming culture was not affected by Tcf3 expression (data not shown), and qPCR confirmed the reduction of Nanog and Esrrb transcripts in Tcf3 expressing reprogramming cultures (Figure S4C–S4E), confirming that Tcf3 overexpression is incompatible with late stages of iPSC formation. However, the expression of an early marker of reprogramming, E-cadherin (Cdh1), which marks the mesenchymal-to-epithelial transition (Samavarchi-Tehrani et al., 2010), was increased when Tcf3 was overexpressed (Figure 4D). Similarly, Tcf3 overexpression resulted in a dramatic increase in alkaline phosphatase (AP) positive colonies, which normally arise at a mid-point of reprogramming (Figure 4E). Together, these data suggest that Tcf3 overexpression stimulates early reprogramming events and colony formation, but inhibits later events including pluripotency gene induction. These data are in agreement with the observation that depletion of endogenous Tcf3/Tcf4 early in reprogramming is inhibitory (Figure 2), while depletion late promotes reprogramming (Figure 3).
Tcf3 has been described to function in mice exclusively as a transcriptional repressor, whereas the other Tcfs have been shown to be able to switch between repressor and activator states (Wu et al., 2012) (BJM, unpublished observation). The effects of Tcf3 overexpression on reprogramming enabled mutational analysis of the domains of Tcf3 required to stimulate AP-positive colony formation using previously characterized mutants. Tcf3 mutants that lack the domain responsible for the interaction with β-catenin (ΔN) or Ctbp (ΔC) repress Tcf-β-catenin target genes similarly to wild-type Tcf3. Those that that lack the groucho-interaction region (ΔG) or carry point mutations in the HMG DNA binding domain (ΔH) do not repress Tcf-β-catenin target genes (Merrill et al., 2001) (Figure 4Fi). During reprogramming, all Tcf3 mutants were expressed at similar levels and localized to the nucleus, ruling out the possibility that differences between mutants could be due to lack of expression or different subcellular localization (Figure S4F and S4G). Similar to wild-type Tcf3, expression of ΔC and ΔN mutants increased the number of AP-positive colonies (Figure 4Fii). By contrast, the ΔG and ΔH mutants that lacked repressor activity also lacked the ability to stimulate AP colony formation (Figure 4Fii). Therefore, direct binding of Tcf3 to DNA and Tcf3’s repressor activity are important for stimulating the early phase of reprogramming.
To determine whether the effects of endogenous Tcf3 was modulated by the reprogramming factors, iPSC reprogramming was examined using all possible combinations of reprogramming factors. For these experiments, we established the genetic ablation of Tcf3 during reprogramming employing MEFs homozygous for a conditional Tcf3 allele (Merrill et al., 2004) that also carry an estrogen receptor-tagged Cre recombinase transgene. These MEFs were initially transduced with separate retroviruses to express the reprogramming factors Oct4, Sox2, cMyc, and Klf4, and after splitting, half of the reprogramming culture was treated with Tamoxifen (Tam) to induce Tcf3 ablation. Upon 24 hours of exposure to Tam, excision of the loxp-flanked cassette (Figure 5A) and elimination of Tcf3 protein occurred efficiently (Figure 5B). Deletion of Tcf3 increased the number of Nanog positive colonies consistently, but less than twofold, without enhancing the kinetics of the process (Figure 5Ci). A similar effect due to Tcf3 deletion was also observed when cMyc was omitted from the reprogramming factor cocktail (Figure 5Cii, 5D, and S5A). The enhancement of OSCK and OSK reprogramming by Tcf3 loss was observed in media containing fetal bovine serum or knockout serum replacement, which is known to enhance reprogramming (Esteban et al., 2010) (Figure S5B), and was not simply a consequence of an increased proliferation rate (Figure S5C and S5D). Deletion of Tcf3 at the very beginning of the reprogramming process reduced the enhancing effect and yielded only few more Nanog positive colonies than control (Figure S5E), indicating that consitutive ablation of Tcf3 throughout the entire reprogramming process causes only a minor increase in the overall efficiency. These data are consistent with our findings that the timing of Tcf3 activity is critical due to the biphasic nature of Wnt effects on iPSC reprogramming.
Of all the possible combinations of reprogramming factors, ablation of Tcf3 caused the strongest affect on OCK reprogramming. Previous studies have reported that reprogramming in the absence of ectopic Sox2 results in the generation of partially reprogrammed ESC-like colonies, in which the pluripotency network is not activated (Takahashi and Yamanaka, 2006). Initially, we found that a very small number of these ESC-like colonies obtained upon OCK-induced reprogramming expressed Nanog in the complete absence of Tcf3 (Figure 5D and S5F), indicating that constitutive Tcf3 deletion enabled OCK reprogramming but at an extremely low rate and with dramatically delayed kinetics compared to OSK or OSCK reprogramming. However, passaging-dependent mechanisms magnified this effect. Specifically, we observed that ESC-like colonies isolated and expanded from a Tcf3−/− OCK reprogramming culture at day 30, induced Nanog expression with high efficiency, while Nanog remained largely undetectable when clones from a parallel Tcf3+/+ OCK reprogramming culture were expanded (Figure S5G). Similarly, splitting Tcf3−/− OCK reprogramming cultures resulted in the induction of Nanog expression in many colonies (Figure 6C/D).
These Nanog positive Tcf3−/− OCK reprogrammed cell lines displayed silencing of retroviral reprogramming factor expression and lacked Tcf3 and retroviral Sox2 integration (Figure S5H and S5I). Hierarchical clustering and Pearson correlation of genome-wide gene expression data showed that OSK and OCK Tcf3−/− iPSC lines were similar to wild-type ESCs and iPSCs, and clearly different from MEFs and a line of partially reprogrammed OCK pre-iPSCs (Figure 5E, S5J, Table S1 and S2). Tcf3−/− reprogrammed lines also produced teratomas with three embryonic germ layers (Figure S5K), and upregulated markers of each germ layer during embryoid body differentiation, albeit with delayed kinetics relative to wild-type iPSCs (Figure 5F), which is a characteristic of Tcf3−/− ESCs compared to wild-type ESCs (Yi et al., 2008). Furthermore, Tcf3−/− iPSC lines bear similar expression differences as Tcf3−/− ESCs when compared to their wild-type counterparts (Figure S5L), further indicating that they closely resemble Tcf3−/− ESCs. Together, these data demonstrate that reprogramming in the absence of Tcf3 and ectopic Sox2 yields bona fide iPSCs.
Since Tcf3 deletion is advantageous for the late stage of OSCK reprogramming and enabled completion of OCK reprogramming, we tested whether partially reprogrammed colonies that normally are the end-product of OCK reprogramming (OCK pre-iPSCs), characterized by an ESC-like morphology and lack of pluripotency network expression (Sridharan et al., 2009; Takahashi and Yamanaka, 2006), are blocked from reaching pluripotency by Tcf repressor activity. Notably, knockdown of Tcf3 and/or Tcf4 yielded a large number of Nanog-GFP positive colonies as early as 72 hours post siRNA transduction, while Tcf1 knockdown did not induce Nanog-GFP expression (Figure 6A, 6B, S6A–S6C). Tcf3 and Tcf4 knockdown in OCK pre-iPSCs induced the Wnt signaling target genes Lef1, Tcf1, and Axin2 (Figure S6D), and the concurrent knockdown of Tcf1 dramatically inhibited the appearance of Nanog-GFP positive colonies without affecting overall colony morphology or cell number (Figure 6B, S6E, and S6F). Lef1 siRNA knockdown did not affect the OCK pre-iPSC to iPSC transition (data not shown). Together, these data indicate that Tcf3 and Tcf4 knockdown can rapidly trigger induction of pluripotency in OCK pre-iPSCs. Furthermore, the transition from OCK pre-iPSCs to pluripotency appears to require a similar mechanism as the late stage of OSCK reprogramming, i.e a Tcf1/Lef1-dependent pathway, likely requiring active Wnt signaling.
To determine downstream genes mediating the effects of Tcf3, we analyzed the gene expression changes in OCK reprogramming cultures in the presence and absence of Tcf3. Parallel Tcf3+/+ and Tcf3−/− reprogramming cultures were split at day 21 to enhance the Tcf3-mediated reprogramming effect, and RNA samples were collected at several time points throughout the reprogramming experiment (Figure 6C and 6D). qPCR confirmed the decrease of Tcf3 mRNA levels upon activation of Cre, and the increase in Nanog transcript levels in the Tcf3−/− OCK reprogramming culture at late time points (Figure S6G). None of the endogenous Sox family members were precociously upregulated in the absence of Tcf3 (Figure S6G), thereby discounting a simple mechanism by which Tcf3 ablation could enable the induction of pluripotency in the absence of ectopic Sox2 (Nakagawa et al., 2008).
We next combined our genome-wide gene expression data with unsupervised Short Time-series Expression Miner (STEM) analysis (Ernst and Bar-Joseph, 2006) to capture expression differences and groups of co-regulated genes between the Tcf3+/+ and Tcf3−/− OCK reprogramming cultures (Tables S1–S3). The three most significant groups of co-regulated genes are depicted in Figure 6E (Table S4). Group 1 genes are more highly expressed in ESCs than MEFs, initially (at day 15) expressed at lower levels in the Tcf3−/− reprogramming culture compared to the Tcf3+/+ culture, but slightly surpassed the levels of the Tcf3+/+ culture by day 22. Based on gene ontology (GO) analysis, these genes are function in the regulation of cell proliferation (Figure 6E). Group 2 genes are strongly induced in the Tcf3+/+ reprogramming culture but not in the Tcf3−/− culture at day 26, and are implicated in morphogenesis and neuronal differentiation. Group 3 genes are more highly expressed in the Tcf3−/− reprogramming culture at day 26 and include several pluripotency-related genes such as Zfp42, Dppa3, Esrrb, and Tcfcp2l1, consistent with the induction of faithful reprogramming specifically in the absence of Tcf3. These data indicate that OCK transduced MEFs progress faster into an intermediate reprogramming state in the presence of Tcf3, but then upregulate various lineage regulators later in the reprogramming process. Since the expression of developmental genes has been suggested to be a barrier to reprogramming (Mikkelsen et al., 2008), these genes could block the entry into pluripotency. In the absence of Tcf3, the upregulation of a large number of developmental genes appears to be efficiently suppressed, which could overcome the pluripotency blockade.
To confirm that the suppression of developmental genes late in reprogramming is not simply a consequence of expression changes that occurred earlier in the process due to continuous Tcf3 deletion, we determined direct expression changes upon Tcf3 depletion in a late reprogramming stage. Depletion of Tcf3 in OCK-pre-iPSCs led to the downregulation of a similar group of developmental genes as defined by Group 2 (Figure S6H and Table S5). Interestingly, active Wnt signaling is known as a negative regulator of neural genes (Aubert et al., 2002; Yoshikawa et al., 1997). Since Wnt target genes such as Tcf1, Lef1, Cdx1, and Brachyury were upregulated both in late Tcf3−/− reprogramming cultures and Tcf3-depleted OCK-pre-iPSCs (Figure S6D and S6I), and active Wnt signaling is required for the enhancing effects of Tcf3 deletion late in reprogramming (Figure 3, ,6B),6B), the induction of Wnt signaling upon Tcf3 deletion may therefore be directly responsible for the suppression of neural genes late in reprogramming.
Taken together, these data demonstrate that Tcf3 has different targets in the early and late stages of the process, which is consistent with the biphasic role of Wnt signaling during reprogramming.
The biphasic response to Wnt signaling and stage specific effects of Tcfs indicate that, to arrive at the pluripotent state, individual cells progress through a Wnt “low” (or Tcf3 high) state followed by progression through a Wnt “high” (or Tcf3 low) activity state. To test this idea directly, we established a system that allowed us to manipulate Tcf3 levels in a stage-dependent manner where each cell expressed high Tcf3 levels at the early stage and reduced Tcf3 levels at the late stage of reprogramming (Figure 7A). Based on our data, we reasoned that elevated Tcf3 should promote early reprogramming events and subsequent depletion of Tcf3 would then promote late events. This hypothesis was tested in the context of OCK reprogramming, the best system to observe reprogramming enhancement in a Tcf3-dependent manner. We expressed Tcf3 at a range of levels early in OCK reprogramming, from day 1 to day 8, taking advantage of a doxycycline-inducible expression system (Figure 7A). At 0 and 0.002μg/ml of dox, representing MEF and ESC-like levels of Tcf3, respectively (Figure 7B), OCK reprogramming cultures appeared similar at day 8 of reprogramming, displaying nascent colonies (Figure S7A). At much higher Tcf3 levels induced by 0.02μg/ml dox (Figure 7B), more and bigger colonies emerged (Figure S7A). On day 8, dox was withdrawn to stop Tcf3 overexpression and siRNAs targeting Tcf3 were added to ensure the reduction of Tcf3 in the late phase (Figure 7A). Reprogramming cultures were monitored daily for Oct4-GFP positive colonies, prompting the following conclusions (Figure 7C): i) Tcf3 overexpression early in OCK reprogramming is not sufficient for the induction of reprogrammed cells. ii) Tcf3 knockdown late, without prior overexpression of Tcf3, only yielded rare Oct4-GFP positive colonies similar to our findings described in Figure 5. iii) Induction of ESC-like transcript levels of Tcf3 early (0.002μg/ml dox) followed by Tcf3 knockdown late resulted in a large numbers of Oct4-GFP positive colonies. iv) Very high levels of Tcf3 (0.02μg/ml dox) early in reprogramming eventually gave rise to some Oct4-GFP positive colonies when combined with Tcf3 knockdown late, albeit with lower efficiency even though this condition resulted in the most promising induction of ESC-like colonies at day 8, indicating that the exact levels of Tcf3 early in reprogramming are critical.
Three Oct4-GFP positive OCK colonies, treated 0.002μg/ml dox and subsequent Tcf3 siRNA knockdown, were stably expanded from this experiment and confirmed to lack the Sox2 reprogramming vector (Figure S7B). These cell lines exhibited typical characteristics of pluripotent stem cells; in addition to their ESC-like morphology, they have silenced the retroviral expression of the reprogramming factors (Figure S7C), expressed the endogenous pluripotency genes Sox2 and Nanog, and displayed ESC-like Tcf3 transcript levels (Figure S7D and S7E). Upon blastocyst injection of two clones, we received pups with contribution of iPSCs to various tissues as tested by PCR for the retroviral Tcf3 transgene (Figure 7D and S7F).
Taken together, this experiment provides the proof-of-principle that, during reprogramming, cells transition through stages in which the activity of the Wnt/Tcf machinery dramatically differs, and where precise levels of Tcf3 are critical to achieve successful reprogramming.
Somatic cells en route to the pluripotent state undergo specific events starting with the loss of somatic cell identity and culminating in the expression of the full pluripotency network (Papp and Plath, 2013). In this study, we performed a comprehensive analysis of the role of Wnt signaling and the requirement of its transcriptional effectors Tcf1, Lef1, Tcf3, and Tcf4 in this process. Our work shows that reprogramming is biphasic with respect to its dependence on endogenous Wnt signaling, Tcf proteins, and the consequences of ectopic Wnt stimulation (summarized in Figure 7E).
Two phases of Wnt signaling could be temporally separated into early and late stages of reprogramming, which enabled us to study the molecular roles for Wnt and Tcfs during each phase. In the early stage, the activation of Wnt signaling leads to a reprogramming block via Tcf1 and Lef1, likely due to induction of Wnt-target genes that interfere with reprogramming events. In contrast, Tcf3 and Tcf4 promote early reprogramming events by repressing Wnt pathway target genes, including Tcf1 and Lef1, and likely other targets not stimulated by Tcf1/Lef1 and active Wnt signaling. The targets of Tcf3/Tcf4 repression interfere with efficient reprogramming when expressed during the early stage. In the late stage, Wnt signaling promotes reprogramming. Interestingly, Tcf1/Lef1 and Tcf3/Tcf4 have opposing roles as they did in the early stage; however, the relationship between Tcf1/Lef1 and Tcf3/Tcf4 is different compared to the early stage. Our data suggest that Tcf3 and Tcf4 repress the expression of Tcf1 and Lef1 late in reprogramming, thereby limiting the activity of Wnt signaling. Accordingly, deletion of Tcf3/Tcf4 late in reprogramming enhances iPSC formation through a mechanism that requires Tcf1 or Lef1 and active Wnt signaling. Thus, Tcf1 and Lef1 appear to be critical target genes of Tcf3 and Tcf4 late in reprogramming. We propose that Wnt3a addition stimulates the late stage of reprogramming primarily by making Tcf1/Lef1 strong activators of key target genes and preventing Tcf1/Lef1 from acting as transcriptional repressors. Although, we do not exclude a direct effect of Wnt3a on Tcf3 or Tcf4 activity or levels, our results suggest that Wnt stimulation acts upstream of Tcf1/Lef1 to enhance the late reprogramming phase reprogramming. The late stage of reprogramming is likely unaffected by Lef1 or Tcf1 depletion because alternative pathways are active that can act on a similar set of target genes. One such pathway may be the Leukemia inhibitory factor (Lif)/Jak/Stat signaling pathway. Notably, in the presence of Lif, there is no consequence on ESC self-renewal upon Tcf1 depletion (Yi et al., 2011). However, the ability of Wnt3a to sustain ESC self-renewal upon Lif-withdrawal is stimulated by Tcf1 (Yi et al., 2011), indicating a redundancy between distinct signaling pathways in maintaining the pluripotent state, which may extend to a redundancy in acquiring pluripotency.
Throughout reprogramming, we suggest that the grouping of Tcf1/Lef1 versus Tcf3/Tcf4 reflects predominant Wnt dependent activator functions of Tcf1/Lef1 and repressor functions of Tcf3/Tcf4. The observation that the four Tcfs fall into two distinct groups for their effect on reprogramming to iPSCs provides further insight into the roles of the factors as mediators of Wnt signaling. The grouping of the factors supports the diversification of the Tcf family into isoforms with specialized and distinct activities (Cadigan and Waterman, 2012). This contrasts the switch model pertaining to invertebrates, where a single Tcf gene product performs both activation and repression. The activator effect attributed to Tcf1/Lef1 during reprogramming is consistent with analysis of Lef1−/−;Tcf1−/− double mutant mice, which display a Wnt3a−/− like phenotype (Galceran et al., 1999). The repressor activity of Tcf3/Tcf4 is consistent with the β-catenin independent effects caused by conditional Tcf3 ablation in the skin of Tcf4−/− mice (Nguyen et al., 2009).
We made the striking discovery that solely manipulating the levels of Tcf3, from slight overexpression early to depletion late in the process, allows efficient and faithful reprogramming in the absence of ectopic Sox2. On a molecular level, this finding highlights a function of Sox2 that can be complemented by regulators of the Wnt pathway. The recently described competition between Tcf3 and Sox2 for binding at Oct-Sox DNA sites provides a possible mechanistic explanation for the effects of Tcf3 ablation during late reprogramming (Zhang et al., 2013). Notably, our data highlight that the degree to which Wnt signaling activation and inhibition affect the early and late stages of reprogramming is dependent on the reprogramming factor combination used.
The duality of effects of Wnt during reprogramming provides a strong example of a factor being necessary at one step but being a barrier at a different step in the long reprogramming process. A prior, it is likely that many factors could cause similar biphasic or context specific effects during reprogramming. Reprogramming methods that account for dynamic changes in signaling requirements, perhaps in other pathways, will more efficiently guide somatic cells into the desired pluripotent state. Moving forward, determining the reprogramming stage-specific target genes of Tcf3/4 and Tcf1/Lef1 under Wnt “on” and “off” conditions, along with different reprogramming factor combinations, will be a key question to answer to further understand the biphasic action of Wnt signaling in reprogramming to iPSCs.
For reprogramming with retroviral factors, Oct4, Sox2, Klf4, and cMyc were expressed from pMX retroviruses as previously described (Maherali et al., 2007). For overexpression, the cDNAs encoding full length Tcf3 or its domain mutants (Merrill et al., 2001), Dkk1, or Tomato fluorescent protein (used as control), were also cloned into the pMX vector. For inducible Tcf3 overexpression experiments, the Tcf3 cDNA was cloned into the pRetroX-Tight-Hyg vector, allowing doxycycline inducible expression in MEFs carrying the M2rtTA transgene in the Rosa26 (R26) locus. For reprogramming experiments utilizing tet-inducible OSCK (iOSCK) reprogramming factors, MEFs harboring the R26-M2rtTA and a single, doxycycline inducible, polycistronic cassette coding for OSCK (Sommer et al., 2009) in the Col1A locus were generated from mice similarly to a published report (Stadtfeld et al., 2010). Some of the MEFs used for reprogramming carried the Oct4-GFP transgene (Szabo et al., 2002) or the GFP knockin in the endogenous Nanog locus (Maherali et al., 2007) as indicated. Reprogramming experiments, various treatments of reprogramming cultures with siRNAs or biologicals, and the characterization of generated iPSC lines were performed as described in the Extended Experimental Procedures.
RNA expression profiling was performed on the Affymetrix Gene Chip Mouse Genome 430 2.0 arrays at the UCLA microarray core facility and a list of all data sets used in this study is given in Table S1. See Extended Experimental Procedures for the expression analysis performed.
Total RNA was isolated from cells using the RNeasy kit (Qiagen) and cDNA generated with Superscript III (Invitrogen). qPCR values were generated using the ddCT method normalized to U6, unless otherwise indicated. Primers used for detecting expression of pMX transgenes and in the embryoid body assays have been previously described (Maherali et al., 2007; Yi et al., 2008). Primer sequences used to measure expression by qPCR and for PCR genotyping are listed in Table S6.
See Extended Experimental Procedures for details and antibodies used.
KP is supported by the NIH (DP2OD001686 and P01 GM099134), and CIRM (RN1-00564), and the UCLA Broad Center of Regenerative Medicine and Stem Cell Research; RH by a NIH Training Grant (5T32AI060567-07) and the UCLA Graduate Division Dissertation Year Fellowship; and BJM by the NIH (R01-CA128571). We thank Mark Chin, Sanjeet Patel, Rupa Sridharan, Robin McKee, Serena Lee, Amander Clark, and Gustavo Mostoslavsky for constructs, assistance with experiments and data analysis, and members of the Plath lab for helpful discussions.
Data are available at GEO under GSE46532.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.