Barhl2 maintains T cell factors as repressors and thereby switches off the Wnt/β-Catenin response driving Spemann organizer formation

The Wnt/β-Catenin signaling pathway has many important roles in embryogenesis, stem cell biology and tumorigenesis (reviewed by Cadigan, 2012; Holland et al., 2013; Nusse and Clevers, 2017). The key step in the transcriptional activation mediated by the canonical Wnt pathway is the nuclear increase of β-Catenin levels that transiently converts T cell factors (Tcfs, also known as transcription factors)/Lymphoid enhancer binding factor 1 (Lef1), bound to Wnt-responsive enhancers (WREs) into a transcriptional activator (Behrens et al., 1996) (reviewed by Cadigan and Waterman, 2012; Clevers and Nusse, 2012). In the Wnt ‘off’ state, Tcfs interact with Groucho (Gro)/Transducin-like Enhancer-of-split (Tle) proteins and act as transcriptional repressors (Roose et al., 1998; Brantjes et al., 2001; Daniels and Weis, 2005) (reviewed by Cinnamon and Paroush, 2008; Jennings and Ish-Horowicz, 2008; Ramakrishnan et al., 2018). The net transcriptional readout of the canonical Wnt signaling pathway is dependent on both the cell and developmental context. Given that β-Catenin is the sole mediator of these Wnt transcriptional responses (Schuijers et al., 2014), the response diversity could be explained by either epigenetic mechanisms or developmental stage-specific Tcfs (Roël et al., 2002; Blythe et al., 2010; Hikasa et al., 2010) (reviewed by Sokol, 2011; Cadigan, 2012; Nakamura and Hoppler, 2017), both of which are poorly understood.

In vertebrate embryos, different Tcf isoforms have distinct activities (reviewed by Cadigan and Waterman, 2012). Some, such as Lef1 or Tcf7 (also known as Tcf1), are required for transcriptional activation (reviewed by Arce et al., 2006), whereas others, such as Tcf7-like 1 (Tcf7l1, also known as Tcf3) mostly act as transcriptional repressors (Houston et al., 2002; Liu et al., 2005). The Gro/Tle family of proteins is conserved in Metazoa. The Gro proteins do not have a DNA-binding domain and rely on their interaction with transcription factors for their specific recognition of promoter and/or enhancer DNA sequences (reviewed by Cinnamon and Paroush, 2008; Jennings and Ish-Horowicz, 2008). Gro proteins bind Tcfs via their glutamine-rich (Q) domain (Cadigan and Waterman, 2012). Gro proteins also contain a WD-repeat domain, which interacts with the Engrailed Homology 1 (EH1) motif. The EH1 motif mediates transcriptional repression in vivo and in vitro and is present in numerous homeodomain-containing transcription factors (Fig. S1) (Smith and Jaynes, 1996; Jimenez et al., 1997; Pickles et al., 2002) (reviewed by Copley, 2005). Both Tcf and Gro proteins interact with histone deacetylases (Hdacs), which establish regional repressive chromatin structures (Chen et al., 1999; Billin et al., 2000; Eshelman et al., 2017) (reviewed by Laugesen and Helin, 2014; Seto and Yoshida, 2014). Until now, there has been little information on developmental contexts in which Tcf and/or Gro proteins mediate transcriptional silencing via Hdac activity (Kaul et al., 2014; Chambers et al., 2017) (reviewed by Cinnamon and Paroush, 2008).

One of the earliest, well-documented and evolutionarily conserved functions of Wnt/β-Catenin signaling is the induction of the blastopore lip organizer (Kraus et al., 2016). In Xenopus embryos, before initiation of zygotic transcription (stage 7), maternal determinants stabilize β-Catenin on the dorsal side of the embryo (Schneider et al., 1996; Larabell et al., 1997). At the same stage, maternally encoded Tcf7l1 represses gene transcription throughout the embryo (Houston et al., 2002). In the presumptive organizer territory, the local increase in nuclear β-Catenin derepresses Tcf7l1 and initiates the dorsal development program that leads to Spemann organizer formation (Molenaar et al., 1996; Schneider et al., 1996; Roose et al., 1998; Houston et al., 2002; Xanthos et al., 2002; Tao et al., 2005; Blythe et al., 2010). A genome-wide study performed in Xenopus laevis (X. laevis) identified all the genes derepressed by β-Catenin in the dorsal part of the embryo (Ding et al., 2017). After initiation of the zygotic program, a second wave of Wnt signaling, driven by wnt family member 8a (wnt8a) expressed in the ventrolateral mesoderm, promotes development of the posterior part of the embryo. Using both RNA-sequencing (seq) and chromatin immunoprecipitation (ChIP)-seq approaches, direct and indirect target genes of zygotic Wnt8a signaling have been identified (Kjolby and Harland, 2017; Nakamura and Hoppler, 2017). Comparing the early and late Wnt signatures revealed that all but three of the early dorsal β-Catenin target genes are unresponsive to the second wave of Wnt signaling (Ding et al., 2017).

We previously investigated the embryological roles of the transcription factor BarH-like homeobox 2 (Barhl2), which carries two EH1 motifs (reviewed in Schuhmacher et al., 2011; Juraver-Geslin and Durand, 2015; Sena et al., 2016). In Xenopus embryos, barhl2 is expressed in the neural plate and the diencephalic primordium. Barhl2 promotes apoptosis in the neuroectoderm and mesoderm, thereby influencing the Bone morphogenetic protein (Bmp) and Sonic hedgehog (Shh) signaling activity gradients within the neural plate (Offner et al., 2005). On the basis of the patterns of apoptosis (Hensey and Gautier, 1998; Yeo and Gautier, 2003), we suggested that Barhl2 acts on the formation of the axial organizing center, that is, the prechordal plate and the notochord, which are two tissues derived from the Spemann organizer (Offner et al., 2005). We also demonstrated that, within the diencephalic primordium, Barhl2 limits β-Catenin transcriptional activation (Juraver-Geslin et al., 2011). RNA-seq databases report the presence of a small number of barhl2 transcripts at stages 8.5-9, when the Spemann organizer forms (Owens et al., 2016; Session et al., 2016; Ding et al., 2017), but developmental apoptosis only occurs after the onset of gastrulation (Hensey and Gautier, 1997). Taken together, these observations suggested that Barhl2 influences the establishment of the Spemann organizer through its activity as a brake on Wnt/β-Catenin-mediated transcriptional activation.

In the present study, we investigated the molecular mechanisms by which Barhl2 influences the Wnt/β-Catenin transcriptional response, using Xenopus embryos and mammalian cells as model systems. We show that Barhl2 limits both the size and the signaling activity of the Spemann organizer and that overexpressed Barhl2 abolishes Spemann organizer formation. In both mammalian cells and X. laevis, Barhl2 directly binds to Gro4 via its EH1 motifs, and interacts with Tcf7l1 independently of Gro. Using Wnt/Tcf reporter activity assays in vitro and in vivo, we demonstrate that Barhl2 prevents the β-Catenin-induced Tcf transcriptional activation switch and inhibits both Tcf7l1, and Tcf7 to activate the early β-Catenin response, probably by interacting with them when bound on DNA. Our data also indicate that Barhl2 promotes Hdac1-dependent repression of Spemann organizer formation and deacetylation of key organizer gene promoters. Taken together, our observations demonstrate that Barhl2 switches off the first wave of Wnt/β-Catenin transcriptional activity, therefore limiting in time and/or space the Spemann organizer induction process.

RNA-seq databases report a small peak of barhl2 expression (4000 transcripts) between stages 7.5 and 9.5; expression drops to low levels at stage 10 and then rises again at stage 12.5, with an average number of 10⁵ transcripts detected at stage 14 (Owens et al., 2016; Session et al., 2016). Using RT-quantitative PCR (RT-qPCR) on individual embryos, we confirmed the presence of barhl2 transcripts in blastula embryos (Fig. 1A). Using whole-mount in situ hybridization (ISH), we observed barhl2 transcripts in the presumptive organizer region at stage 8.5 (Fig. 1B). Using non-quantitative RT-PCR analysis on batches of embryos, we were previously unable to detect barhl2 expression at stage 8 (Offner et al., 2005). This discrepancy probably reflects sensitivity differences in the RNA detection methods used.

We next assessed whether Barhl2 acts in the establishment of the anteroposterior (AP) and dorsoventral (DV) axis of X. laevis embryos. To deplete Barhl2 activity, we used previously characterized morpholinos (MOs) (Offner et al., 2005) (see Materials and Methods). MO depletion of Barhl2 activity generated an anteriorization phenotype, with stage-25 embryos exhibiting enlarged cement glands (Fig. 1Ca,b). Conversely, both mice (mBarhl2) and Xenopus barhl2 (xbarhl2) mRNA injections generated ventralized embryos, lacking dorsoanterior structures (Fig. 1Cd,e). Injection of a barhl2-encoding DNA vector that requires zygotic transcription to be expressed (Christian and Moon, 1993) generated embryos with diminished dorsoanterior structures, particularly a diminished cement gland (Fig. 1Cf). For each condition, the penetrance of the phenotype was quantified (Fig. 1D). We previously established a specific dose-dependent inhibition of barhl2 mRNA translation by MObarhl2 (Offner et al., 2005). Here, we further confirmed that, at two different doses, injection of Barhl2, which does not carry a sequence complementary to that of MObarhl2, rescued the enlargement of the cement gland induced by depletion of endogenous Barhl2 activity (Fig. 1E).

Taken together, these findings provide evidence that barhl2 transcripts are present in the presumptive organizer territory, and that Barhl2 normally helps establish the embryo organization plan.

We next investigated the role of Barhl2 on Spemann organizer formation and signaling. We measured changes in chordin (chd) expression territory in stage-10 barhl2 gain-of-function (GOF) and loss-of-function (LOF) embryos using a new ImageJ macro (see Materials and Methods). In both GOF and LOF conditions, we observed significant differences between the injected and control sides (Fig. 2AB; Fig. S2): Barhl2 overexpression decreased the chd expression territory, whereas Barhl2 depletion increased it (Fig. 2A,B).

Using RT-qPCR on individual stage 8-10 embryos depleted of Barhl2 activity, we quantified expression changes in both organizer genes [barhl2, siamois1 (sia1), chd, goosecoid homeobox (gsc), orthodenticle homeobox 2 (otx2) and notochord homeobox (not)] and ventro/lateral mesoderm markers [vent homeobox 2 (vent2), wnt8a and brachyury] (Fig. 2C). In agreement with a role for Barhl2 after stage 8, we did not detect significant changes in the expression of sia1. However, we detected a 2-5-fold increase in the expression of all the organizer genes analyzed (Fig. 2C). By contrast, we found a decrease in all the ventro/lateral mesoderm markers analyzed (Fig. 2C). Using ISH on stage-10 embryos, we confirmed these observations on Barhl2-depleted embryos (Fig. S3). By contrast, stage-10 xbarhl2-overexpressing embryos exhibited a decrease in otx2 and gsc territories, and an increase in wnt8a and vent2 territories (Fig. S3). Therefore, Barhl2 LOF increased the expression of key organizer genes, whereas Barhl2 GOF decreased it.

During gastrulation, the Spemann organizer and its derived structures, the prechordal plate and notochord, produce planar and vertical signals (including Bmp inhibitors and Shh) that induce and pattern the mesoderm and the emerging nervous system (reviewed in De Robertis and Kuroda, 2004; Niehrs, 2004; Stern, 2006). We tested whether Barhl2 acts on this organizer signaling activity and thereby also influences patterning of the neuroepithelium.

Our previous analysis of the effects of Barhl2 on early neurula phenotypes demonstrated that Barhl2 influences the Shh and Bmp signaling gradients within the neural plate (Offner et al., 2005; Juraver-Geslin et al., 2011). Here, we confirmed that Barhl2 depletion produced a widening of the neuroepithelium and associated defects in AP and DV patterning of the neural plate (Fig. S4A). We hypothesized that the impact of Barhl2 GOF and LOF on neuroepithelium patterning resulted from Barhl2 acting on Spemann organizer formation. However, to exclude the possibility that Barhl2, which is undetectable in the dorsal ectoderm, nonetheless acted by changing the response of these ectodermal cells to extracellular signals, we targeted Barhl2 depletion to different parts of the dorsal tissues of the embryo. Based on the expression of a co-injected tracer, we assessed retrospectively whether MObarhl2 was injected in the dorsal mesoderm and axial dorsal ectoderm or in the lateral mesoderm and lateral ectoderm (Fig. 2Da). We observed the characteristic increase in sry box 3 (sox3) territory only when Barhl2 depletion was targeted to the presumptive organizer territory that arises from the dorsal mesoderm (Fig. 2Db; Fig. S4B,C).

Thus, we conclude that Barhl2 normally limits the extent of the Spemann organizer territory and, consequently, the signaling activity of the organizer.

We tested whether Barhl2 activity on organizer formation requires its DNA-binding domain, its EH1 domains, or both (Fig. 3A).

In full-length barhl2 (barhl2fl), we mutated amino acids (AAs) in the homeodomain (HD) known to be instrumental for the correct binding to target DNA, to generate barhl2flHoxM (Fig. 3A) (Bürglin and Affolter, 2016). We also removed the EH1 motifs to generate barhl2ΔEHs and mutated three AAs in each EH1 motif to generate barhl2flEHsM (Fig. 3A). As discussed earlier, Barhl2 activity on the neural plate size is at least partly an indirect consequence of its activity on the Spemann organizer. At stage 14, Barhl2 GOF caused a decrease in the size of the neural plate (Fig. 3Ba,C) (Offner et al., 2005). By contrast, overexpression of mutant forms of barhl2 cDNA did not affect the size of the neural plate nor the expression of organizer and ventro/lateral mesoderm markers (Fig. 3Bb-d,C; Fig. S5). Therefore, Barhl2 activity on the organizer requires both its DNA-binding domain and its two EH1 motifs.

We next investigated interactions between Barhl2fl, Gro4 and Tcf7l1, performing co-immunoprecipitation (Co-IP) experiments on protein extracts from transfected HEK293T cells (Fig. 3D-F). We observed that Barhl2fl co-immunoprecipitated with Gro4 (Fig. 3D). Using Tcf7l1 as bait, we observed that, in the absence of Barhl2fl, the interaction between Tcf7l1 and Gro4 could be detected with overnight Co-IP (Fig. 3E). We next tested whether Barhl2 could bind Tcf7l1, and whether it acts on formation of the Tcf7l1-Gro4 complex. We observed that Tcf7l1 co-immunoprecipitated with Barhl2fl, both in the absence and in the presence of Gro4 (Fig. 3F; lanes 2, 8). Tcf7l1 also co-immunoprecipitated with Barhl2flEHsM, which has mutated EH1 motifs (Fig. 3F; lanes 3 and 9). Thus, Barhl2 binds to Tcf7l1 independently of Gro4. Moreover, in the presence of Barhl2fl, the ability of Tcf7l1 to co-immunoprecipitate with Gro4 was increased (Fig. 3F; compare lanes 7 and 8). The effect was dependent on the EH1 motifs because Barhl2flEHsM did not promote the ability of Tcf7l1 to precipitate with Gro4 (Fig. 3F; lane 9).

We then investigated the functional consequences of the ability of Barhl2 to stabilize the Tcf7l1-Gro4 interaction. We followed the transcriptional output of the switch from repressive to active forms of Tcf induced by β-Catenin. Mean luciferase activities of pTop-Flash (Korinek et al., 1997) were measured in HEK293 T cells transfected with increasing amounts of a constitutively active form of β-Catenin (Activated-β-Cat) (Wei et al., 2003), together with a vector encoding Barhl2fl. When pTop-Flash and Activated-β-Cat were co-transfected with Barhl2fl, luciferase activity decreased at all doses of Activated-β-Cat tested, and the decrease was Barhl2-dose dependent (Fig. 3G).

Therefore, the ability of Barhl2 to limit the size of the Spemann organizer requires it to bind to DNA and an EH1-mediated protein–protein interaction. In mammalian cells, Barhl2fl interacts with Gro4 through its EH1 motifs, binds Tcf7l1 independently of Gro4, and prevents β-Catenin-driven activation of synthetic Tcf target genes.

We tested whether the ability of Barhl2 to limit β-Catenin-driven activation of Tcf target genes was conserved in blastula embryos.

We first performed Co-IP experiments using protein extracts from Xenopus embryos (Fig. 4A-C; Fig. S1). In Xenopus extracts, Gro4 co-immunoprecipitated with Barhl2fl (Fig. 4A), and Tcf7l1 co-immunoprecipitated with Barhl2 and Gro4 when injected either separately or together (Fig. 4B,C). Therefore, Barhl2 interacts with both Gro4 and Tcf7l1 in embryos.

We next investigated the effects of Barhl2 on Tcf activity during Spemann organizer development. We used a Xenopus tropicalis transgenic Wnt reporter line called pbin7LefdGFP (Tran et al., 2010), which contains one copy of a wnt reporter gene (Fig. S6). From stage 8 to stage 9.5, we observed localized Tcf activity in an area corresponding to the presumptive organizer territory (Fig. 4Da-c). Barhl2 depletion increased the number of Wnt-responsive cells in this territory (Fig. 4Ea-c). RT-qPCR quantification of the Barhl2 depletion effect demonstrated a significant increase in gfp mRNA levels in Barhl2-depleted embryos compared with the control at stage 8.5 (Fig. 4F). Therefore, between stage 8 and stage 9.5, Barhl2 limits the ability of Tcf to activate transcription in the presumptive organizer territory.

Whereas at stage 8.5, Tcf activity marked the presumptive organizer territory (Fig. 4Da-c), from stage 10 onwards, the activity of Tcf was undetectable within the dorsal blastopore lip (Fig. 4Dd,Ga). To validate our observations, we assessed the development of the dorsal blastopore lip (i.e. the organizer) upon Barhl2 LOF and GOF in stage-10.5 X. tropicalis transgenic Wnt reporter embryos. In agreement with a role for Barhl2 in limiting formation of the organizer, at stages 10 and 10.5, depletion of Barhl2 enlarged this area (Fig. 4Ed,Gb), whereas at stage 10.5, Barhl2 GOF dramatically decreased it (Fig. 4Gc).

Thus, we conclude that, in Xenopus embryos, Barhl2 interacts with both Gro4 and Tcf7l1 and, within the presumptive organizer territory, Barhl2 normally limits the activation of Tcf target genes.

We next investigated whether Barhl2 prevents the β-Catenin-induced Tcf transcriptional switch and, consequently, the initiation of the dorsal development program.

Ventral injection of wnt8a RNA, which locally increases β-Catenin activity, induces duplication of the embryonic axis (McMahon and Moon, 1989; Sokol et al., 1991). Similarly, wnt-8b, a close paralog of wnt8a, can induce a secondary axis (Cui et al., 1995). We investigated the ability of Barhl2 to block this wnt8b induction potential (Fig. 5A,B; Fig. S7). As expected, ventral injection of wnt8b induced the formation of a double axis in 85% of embryos, whereas ventral injection of xbarhl2fl or mBarhl2fl did not (Fig. 5A,B). However, when either xbarhl2fl or mBarhl2fl was co-injected with wnt8b, the number of embryos developing secondary axes decreased dramatically (Fig. 5A,B). When MObarhl2 was ventrally injected alone or with wnt8b, no significant impact was observed on the wnt8b-induced axis duplication (Fig. 5A,B). However, when co-injected with Xbarhl2fl, MObarhl2 completely abolished its inhibitory effect on wnt8b double-axis induction. In similar conditions, and in agreement with the absence of a sequence complementary to that of MObarhl2 in the mBarhl2 RNA construct, MObarhl2 had no influence on the ability of mBarhl2 to block the inducing effect of wnt8b (Fig. 5A,B). Therefore, in Xenopus embryos, barhl2 prevents Wnt/β-Catenin-driven activation of the Spemann organizer program.

We next investigated whether Barhl2 acts solely on Tcf7l1, or on different Tcfs, including Tcf7. In Xenopus embryos, between stage 7 and stage 10, transcripts encoding both tcf7l1 and tcf7 are detected, whereas those encoding tcf7 like 2 (tcf7l2, also known as tcf4) and lef1 are not (Owens et al., 2016; Session et al., 2016). Ventral injection of RNA encoding Tcf7l1 had little developmental impact (Fig. 5C). We ventrally injected RNA encoding tcf7l1-vp16-gr, an inducible constitutively activating form of Tcf7l1 that carries the transactivation domain of VP16 and the ligand-binding domain of the human glucocorticoid receptor (de Croze et al., 2011) (Fig. 5C). We exposed tcf7l1-vp16-gr-injected embryos to dexamethasone from stage 5 to stage 10 and observed formation of double axis in 89% of embryos, 16% of which had a complete double axis (Fig. 5C). Similarly, ventral injection of RNA encoding tcf7 generated 85% of embryos with a partial double axis (Fig. 5C). Co-injection of xbarhl2fl or mBarhl2fl with either tcf7, or tcf7l1-vp16-gr significantly abolished the development of secondary axes (Fig. 5C). Therefore, Barhl2 inhibits the ability of both Tcf7 and Tcf7l1-Vp16 to induce the dorsal developmental program.

Whereas depletion of Barhl2 induced the formation of embryos with large heads and cement glands (Fig. 1; Fig. 5Da), MO-mediated inhibition of Tcf7l1 activity generated the opposite phenotype, in that Tcf7l1-depleted embryos were microcephalic (Hikasa et al., 2010) (Fig. 5Dc). To assess whether depletion of Barhl2 compensates for the depletion of Tcf7l1, we measured the cement gland width in embryos injected with either MObarhl2 or MOtcf7l1 or both (Fig. 5D). Depletion of both transcription factors rescued the phenotypes of both heads (Fig. 5Db,d). Using a previously validated MO for Tcf7 (Liu et al., 2005), we depleted Tcf7 either alone or together with Tcf7l1. Whereas Tcf7 depletion induced a small decrease in head and cement gland size, its co-depletion with Tcf7l1 generated an additive effect on cement gland development (Fig. 5E). Therefore, Barhl2, Tcf7 and Tcf7l1 activities participate in the induction of the dorsal program.

Finally, we investigated whether Tcf7l1 and/or Barhl2 bind to Tcf DNA-binding motifs. A 0.2-kb fragment of the gsc promoter contains two typical consensus Tcf cis-regulatory motifs (CRM), (CTTTG(A/T)(A/T)), and no Barhl2 CRM (CAATTA) (Berger et al., 2008) (Fig. 5Fa). We performed ChIP-qPCR experiments on sonicated Xenopus extracts injected with RNA encoding tcf7l1-flag together with barhl2-myc, using their sibling wild-type (WT) embryos as controls (Blythe et al., 2009). We performed ChIP using Myc and Flag antibodies and observed an increase in the gsc promoter fragment containing Tcf-CRM, compared with a fragment of the gapdh gene, which contains neither Tcf-CRM nor Barhl2-CRM. Therefore, both Tcf7l1 and Barhl2 can interact with the Tcf-CRM of the gsc promoter (Fig. 5Fb).

In conclusion, during induction of the dorsal development program, Barhl2 limits the Tcf switch from an inhibiting to an activating form, regulating both Tcf7l1 and Tcf7, probably when Tcfs are bound on DNA.

Taken together, our results suggested that Barhl2 switches off the early dorsal Wnt/β-Catenin response after its initial induction by sia1. In the blastula and early gastrula, there is an 86% correlation of CRM bound by β-Catenin and Gro co-repressors (Nakamura and Hoppler, 2017). Thus, we tested whether, at these developmental stages, Barhl2 inhibited transcription of the organizer program through Gro-mediated transcriptional repression.

To produce molecular tools that compete with Barhl2fl for Gro binding, we generated barhl2EHs and barhl2EHsM (Fig. 6A). Gro4 co-immunoprecipitated with Barhl2EHs, but not with Barhl2EHsM (Fig. S8A). Moreover, Barhl2EHs co-immunoprecipitated with Gro4 with a higher efficiency than with Gro1, arguing that Barhl2EHs interacts preferentially with Gro4 (Fig. S8B). The ability of Gro4 to co-immunoprecipitate with Barhl2EHs was also observed in embryos (Fig. S8C).

We investigated whether Barhl2EHs competes with Barhl2fl for binding to Gro4 (Fig. 5B). Using a limiting amount of Gro4 as Co-IP bait, we observed that, in the presence of Barhl2fl, increasing the amount of Barhl2EHs diminished the interaction of Gro4 with Barhl2fl (Fig. 6B, compare lane 3 to lane 5). These findings indicate that Barhl2EHs competes with Barhl2fl for Gro4 binding.

We assessed the functional consequences of Barhl2EHs interaction with Gro4 on Gro co-repressor activity. In transient transfection assays, Gro4 inhibits the activation of synthetic Tcf reporter gene transcription (Roose et al., 1998). We observed a strong decrease in pTop-Flash activity when Activated-β-Cat was co-transfected with Gro4 (Fig. 6C). This Gro-induced decrease of pTop-Flash activity was significantly alleviated in the presence of Barhl2EHs, but not in the presence of Barhl2EHsM (Fig. 6C). Using the same tools, we observed that co-transfection of Barhl2EHs reverted the inhibitory activity of Barhl2fl of Activated-β-Cat on pTop-Flash (Fig. 6D). Therefore, Barhl2EHs alleviates Gro4 and Barhl2 inhibitory activities on Tcf transcriptional activation.

We next investigated whether Barhl2 activity on organizer formation is efficiently antagonized by overexpression of its N-terminal part. At stage 14, Barhl2 depletion caused an enlargement of the neural plate (Fig. 6Ea,F). We observed that overexpression of the N-terminal part of Barhl2 (barhl2EHs) mimicked the Barhl2 depletion phenotype (Fig. 6Eb,F). This activity was dependent on the presence of functional EH1 motifs, because overexpression of barhl2EHsM did not cause any change in neural plate size (Fig. 6Ec,F). We next quantified the neural plate enlargement phenotype in embryos injected with RNAs encoding Barhl2 and/or Barhl2EHs, either separately or together. We observed that Barhl2EHs overexpression rescued the Barhl2 overexpression phenotype (Fig. 6G). Therefore, Barhl2 physiological activity is efficiently antagonized by overexpression of its N-terminal part.

Finally, we assessed Barhl2EHs activity on Spemann organizer formation. At stage 10, Barhl2EHs expanded the expression domain of the organizer markers chd, gsc and otx2, and shrank the expression domain of the ventrolateral mesodermal marker vent homeobox 1 (vent1) in the dorsal blastopore lip area (Fig. 6H). Using chd and not as markers, we also observed that Barhl2EHs increased notochord (a derivative of the Spemann organizer) size, whereas Barhl2EHsM did not (Fig. S9). Therefore, similar to the effect of MO-mediated depletion of Barhl2 activity, Barhl2EHs overexpression increased Spemann organizer size.

Taken together, these results indicate that Barhl2 limits organizer formation through the recruitment of Gro4, which mediates transcriptional repression of the early dorsal β-Catenin signature.

Both Tcf and Gro interact with Hdac1 (Chen et al., 1999; Billin et al., 2000). Thus, we next tested whether Hdac1 activity participates in the Barhl2-mediated repression of organizer genes in blastula embryos.

We first investigated whether Barhl2fl influences the ability of Tcf7l1 to recruit Hdac1 and/or influences the ability of Tcf7l1 to recruit Gro4 in the presence of Hdac1. In HEK293T cells, Tcf7l1 co-immunoprecipitates with Hdac1 (Billin et al., 2000), irrespective of the presence or absence of Gro4 (Fig. 7A, lanes 4 and 5). In the presence of Hdac1, Barhl2fl increased the ability of Tcf7l1 to recruit Gro4, and Tcf7l1 co-immunoprecipitated with Barhl2fl, Gro4 and Hdac1 (Fig. 7A, lane 6). The interaction of Tcf7l1 and Barhl2fl with Hdac1 appeared to be independent of either Gro4 or the EH1 motifs, because Tcf7l1 similarly co-immunoprecipitated with Barhl2flEHsM and Hdac1 (Fig. 7A, lane 8).

These Co-IP experiments were done using an excess of Tcf7l1 as bait. Consequently, they did not indicate whether Tcf7l1 interacts with Hdac1 in a complex also containing Barhl2fl and Gro4 or, alternatively, with Hdac1 and separately with Barhl2 and Gro. In the latter case, Tcf7l1 would co-immunoprecipitate with two different protein complexes. We performed Co-IP experiments using Gro4 in limiting amounts as bait in the presence of Hdac1, Tcf7l1 and increasing amounts of Barhl2fl (Fig. 7B). When Barhl2fl was absent, Gro interacted preferentially with Hdac1 (Fig. 7B, lane 4). In the presence of Tcf7l1 and Hdac1, interaction between Gro and Hdac1 was detected, but not between Gro and Tcf7l1 (Fig. 7B, lane 4). When the amount of Barhl2fl was increased, Gro co-immunoprecipitated with Hdac1 together with Barhl2fl and Tcf7l1 (Fig. 7B, lanes 5-7). Taken together, these observations indicate that Gro4 is part of a protein complex containing Tcf7l1, Barhl2fl and Hdac1.

We next tested whether Hdac1 activity participates in the repression of Tcf target genes in blastula embryos. We depleted Hdac1 activity using previously characterized MOhdac1 (Tao et al., 2015), injected into pbin7LefdGFP embryos. At stage 8.5, we observed an increase in the number of Wnt-responsive cells in the Hdac1-depleted territory (Fig. 7C) as well as in the levels of gfp mRNA in embryos depleted for Hdac1 activity (Fig. 7D). Using RT-qPCR in individual stage-10 embryos, we observed an increase in both chd and gsc mRNA levels in embryos depleted for Hdac1 activity (Fig. 7E). Therefore, Hdac1 activity participates in transcriptional repression of organizer genes (Gao et al., 2016).

We next investigated whether Hdac1 activity contributes to Barhl2-driven transcriptional repression of organizer genes. We injected barhl2 RNAs and MOhdac1 either separately or together into X. laevis embryos and assessed organizer size by following the expression of chd. As previously described (Fig. 2A,B), we observed a decrease in chd expression in the barhl2-injected side of stage-10 embryos (Fig. 7Fa). By contrast, depletion of Hdac1 activity produced an enlargement of the chd expression territory (Fig. 7Fb). When both MOhdac1 and barhl2 RNA were injected together, we observed a rescue of the Barhl2 overexpression phenotype (Fig. 7Fc). We found that co-injection of barhl2, together with MOhdac1, significantly rescued the changes in organizer size (Fig. 7G; Fig. S10).

Hdac1 catalyzes the removal of acetyl groups from selected lysines in the conserved tails of core histone proteins. One of these acetylated marks is on Lysine 9 of Histone H3 (H3K9Ac) and is associated with promoters of actively transcribed genes. We tested whether Barhl2 depletion affected Hdac activity on the promoters of the organizer genes gsc, chd, otx2 and sia1. We performed ChIP-qPCR experiments on sonicated Xenopus extracts, either depleted of Barhl2 or injected with RNA encoding barhl2, using their sibling WT embryos as controls (Blythe et al., 2009). Depletion of Barhl2 induced a significant increase in the H3K9Ac marks at the promoter regions of gsc, chd and otx2, whereas barhl2 overexpression did not (Fig. 7H). No changes were observed in the presence of H3K9Ac marks at the promoter of sia1, ef1a and gapdh (Fig. 7H). Therefore, Barhl2 depletion appears to increase the promoter activity of key organizer genes via Hdac activity.

In conclusion, at the blastula and early gastrula stages, Hdac1 activity probably participates in Barhl2-driven transcriptional repression of organizer genes.

In this study, we showed that, in both mammalian cells and Xenopus embryos, the transcription factor Barhl2 enhanced Tcf-mediated transcriptional repression. At the blastula and early gastrula stages, Barhl2 silenced the early, dorsal, β-Catenin transcriptional response and, consequently, limited Spemann organizer formation in time and/or space (Fig. 8). Barhl2 repression activity was mediated via Hdac1, which is known to have a global and long-term repressive effect on transcription.

Whereas it had previously been proposed that β-Catenin activates Wnt-responsive genes by simply displacing Gro (Roose et al., 1998; Daniels and Weis, 2005), recent studies revealed that it promotes gene activation by displacing the whole Tcf7l1-Gro repressor complex and replacing it with an activator complex containing β-Catenin in association with Tcf7 (Roël et al., 2002; Hikasa et al., 2010) (Fig. 8). Here, we showed that Barhl2 blocked the ability of both Tcf7 and Tcf7l1 to induce axis duplication; we also provide evidence that Tcf7l1, and Tcf7 activities participated with Barhl2 in organizer formation. Thus, we propose that, during Spemann organizer formation, stabilization of β-Catenin first derepresses Tcf7l1 and then initiates the dorsal development program through activating either Tcf7 or Tcf7l1, or both. Once Barhl2 is present, it locks Tcf7l1 and/or Tcf7 in an inhibitory state (Fig. 8) and, consequently, limits induction of the dorsal development program.

In all cases, this switch from repression to activation of transcription for Wnt/β-Catenin target genes depends on the stability of the Tcf7l1-Gro repressor complex, which promotes compaction of chromatin when the canonical Wnt pathway is switched off (Shy et al., 2013). Gro co-repressors can not only attenuate gene transcription (Kaul et al., 2015), but also silence target genes by tetramerizing on a Tcf7l1-Gro complex, thereby promoting structural transitions of chromatin and transcriptional repression (Chen et al., 1998; Chodaparambil et al., 2014). In cells, Barhl2 dramatically increased Gro stoichiometry in a protein complex containing Tcf7l1 alone, or with Hdac1. Therefore, Barhl2 might promote Gro tetramerization in a complex containing Tcf7l1 and Hdac1, thereby enhancing the binding of the complex to histones and the long-term silencing of Tcf target genes.

The developmental contexts in which Tcf and/or Gro proteins mediate transcriptional silencing via Hdac activity are poorly characterized (Kaul et al., 2014; Chambers et al., 2017) (reviewed by Cinnamon and Paroush, 2008). Here, we provide evidence that Hdac1 activity contributes to Barhl2-mediated repression of canonical Wnt target genes in late blastula embryos. Our observations should help identify other developmental contexts in which Gro, together with Hdac1, act over long distances to silence gene transcription.

Post-transcriptional modifications, including phosphorylation and/or ubiquitination of both Tcf7l1 and Gro, can influence increase or decrease Gro-Tcf interactions (Hikasa et al., 2010; Hanson et al., 2012) (reviewed by Turki-Judeh and Courey, 2012). Although Barhl2 interacted with both Gro4 and Tcf7l1 in whole X. laevis embryos, it did not stabilize Gro/Tcf7l1 complexes in Co-IP experiments. It is possible that the affinity of Barhl2 for Tcfs and/or Gro is modified by post-translational modifications, which could explain the differences we observed between cells and embryos.

Before stage 7.5, the dorsal accumulation of β-Catenin establishes transcriptionally poised chromatin architecture at the promoter of sia1, via the recruitment of the arginine methyltransferase Prmt2 (Blythe et al., 2010). Consequently, barhl2 has no influence on the early expression of sia1. A study in early X. tropicalis gastrula revealed the presence of β-Catenin-binding sites on sia1, chd, gsc, otx2, not and barhl2 loci (Nakamura et al., 2016). Whereas all the organizer genes are activated by maternal Wnt/β-Catenin signals, barhl2 is neither part of the early dorsal β-Catenin signature nor induced by overexpression of RNA encoding wnt8b (Ding et al., 2017). However, injection of RNA encoding sia1 increases barhl2 expression (Ding et al., 2017). Given that the time window between barhl2 and sia1 expression is short, these findings suggest that barhl2 is directly under the influence of sia1. In any case, barhl2 starts to be expressed when the transcription of early, dorsal, β-Catenin target genes has already been initiated (Fig. 8) (Ding et al., 2017).

In this study, we demonstrated that Barhl2 limited Spemann organizer formation. Barhl2 depletion expanded both the organizer territory and its signaling activity. It also enhanced the acetylation status of promoters of key organizer genes, thereby increasing their activity. Taken together, our results indicate that Barhl2 participates in the progression of the blastula developmental program and suggest that its physiological function is to extinguish the early dorsal Wnt/β-Catenin transcriptional wave after its initial induction by sia1 (Fig. 8). Our results do not indicate whether barhl2 turns off the organizer program within the entire organizer territory, or whether it acts to define the organizer boundaries.

Our results argue that Barhl2EHs acts by sequestering Gro proteins and that the interaction of Barhl2 with Gro, probably Gro4, is strictly necessary for the normal activity of Barhl2. Nakamura and Hoppler demonstrated that, at these early developmental stages, Gro binds to the same CRMs as β-Catenin (Nakamura and Hoppler, 2017). A large-scale analysis identified the tissue-specific CRMs responsible for expression of head and trunk organizer genes in Xenopus gastrula embryos and showed that Gro represses a large battery of these genes, at least partly in association with Gsc and Otx2 (Yasuoka et al., 2014). Further studies, including RNA-seq and ChIP-seq approaches, will be necessary to establish whether Barhl2 limits the expression of organizer genes solely by enhancing Tcf repressor activity, or also by interacting with Gro in association with other transcription factors. However, our findings indicate that Barhl2 turns off the expression of most early dorsal β-Catenin target genes through its interaction with Gro.

Our results suggest that, in embryos, Barhl2 acts while bound on distant enhancers and participates in gene silencing via the recruitment of Gro, together with chromatin modifiers. The gene promoters of both the pTop-Flash and the X. tropicalis pbin7LefdGFP Wnt reporter line do not contain Barhl2 DNA-binding motifs (Berger et al., 2008). In transfected cells, Barhl2 limited the β-Catenin activation of pTop-Flash but did not extinguish it. We suggest that, by stabilizing the Tcf7l1-Gro interaction, Barhl2 increases the pausing time of RNA polymerase II (Pol II) on the proximal promoter, a mechanism described for Gro-mediated attenuation of transcription (Kaul et al., 2015). In the X. tropicalis pbin7LefdGFP Wnt reporter line, Barhl2 could similarly act locally, without DNA binding. However, it could also act while bound on distant enhancers. Our ChIP-qPCR observations on the gsc promoter indicate that both Barhl2 and Tcf7l1 can interact with the same Tcf-CRM in the absence of an adjacent Barhl2-CRM. A form of Barhl2fl mutated for DNA binding was inactive in embryos. Further studies will be necessary to determine the contribution of the interaction of Barhl2 with DNA in the transcriptional repression process controlled by Barhl2 and Tcf. However, taken together, our observations are consistent with the long-range activity of Barhl2 via its specific binding on DNA, perhaps in superenhancers, as previously described (Lin et al., 2016).

Here, we described a novel mechanism by which the transcription factor Barhl2 holds the Tcf-Gro complexes in an inhibitory state. Barhl2 contributes to the Gro-mediated regulation of the chromatin state, helping to repress the transcriptional response to a given amount of Wnt signaling.

X. laevis embryos were obtained and staged using standard procedures (Nieuwkoop and Faber, 1994; Sive et al., 2010). X. tropicalis transgenic embryos were obtained by conventional methods of hormone-induced egg laying and in vitro fertilization between a WT female and a transgenic male carrying the Wnt reporter pbin7LefdGFP, as previously described (Tran et al., 2010; Tran and Vleminckx, 2014). The male was selected beforehand as having a single insertion site of the transgene (as inferred by mendelian ratios in its progeny) to insure homogeneous levels of green fluorescent protein (gfp) expression in the offspring (Borday et al., 2018). Experimental procedures were specifically approved by the ethics committee of the Institut Curie (Authorization 2016-013 given by CEEA-IC #118) and follow international guidelines (2010/63/UE). B.D. carries the Authorization for Vertebrates’ Experimental Use N°75-1548.

All mutated and/or truncated forms of Barhl2 were generated from mouse Barhl2 full-length cDNA in pCS2+ (Offner et al., 2005). The Barhl2 AA sequence is partly indicated, specifically the homeodomain (HD-60AAs) with its three helices (Fig. 3A; AAs in blue), a nuclear localization sequence (NLS) and its two EH1 domains comprising phenylalanine (F), isoleucine (I) and leucine (L). In Barhl2flHoxM, we mutated four AAs in the third helix of the homeodomain (HD-60AAs) known to interact with DNA-Q44E, T47G, Q51E and N52R. In Barhl2ΔEHs, the first 182 AAs of Barhl2fl were deleted. In Barhl2flEHsM, the F, I and L residues in position 1, 3 and 6 of each EH1 motif were mutated in C, K, and N, respectively. Barhl2EHs contained the first 177 AAs of mouse Barhl2 full-length cDNA (Barhl2fl). In Barhl2EHsM, the AA residues in red of each EH1 motif were mutated as in Barhl2flEHsM. Constructs were generated by sequential PCR amplification followed by subcloning and sequencing in pTOPO (TOPO TA cloning kit, Thermo Fisher Scientific) and subcloned in pCS2+. The full-length and mutated cDNA tagged constructs were generated by subcloning in either pCS2-Myc (C-terminal Myc tag for Barhl2EHs-Myc and Barhl2fl-Myc), in pEGFP-N1 (N-terminal GFP tag for Barhl2fl-GFP) or in pCS2-GFP (C-terminal GFP tag for Barhl2EHs-GFP and Barhl2EHsM-GFP). HA tag (YPYDVPDYATSTIV) and AM-Tag (CQDPQRKGNVILSQAYGCQDPQRKGNVILSQAY)-encoding oligonucleotides were subcloned in the C termini of pCS2-Barhl2fl and pCS2-Barhl2flEHsM. HA tag-encoding oligonucleotides were subcloned in the C termini of pCS2-groucho4 (a gift from Phil Jones, Hutchison-MRC Research Centre, Cambridge, UK). A vector carrying a Flag tag at the N-terminal end and HA (YPYDVPDYATSTIV) at the C-terminal part of Tcf7l1 was a gift from S. Sokol (Icahn School of Medicine, New York, NY, USA; Hikasa et al., 2010). pCS2+ Myc-Tcf7l1 (Xenopus) was a gift from Marc Kirschner (Harvard Medical School, Boston, MA, USA; Addgene plasmid #13439). p181 pK7-Hdac1 (GFP) was a gift from Ramesh Shivdasani (Dana-Farber Cancer Institute, Boston, MA, USA; Addgene plasmid #11054). pCS2+ Flag-Groucho4 and pCS2+ Myc-Groucho4 (X. laevis) were also gifts from Phil Jones. pCS2-XGroucho1-Flag was a gift from Nancy Papalopulu (Faculty of Life Sciences, Manchester, UK). Tc7l1p16gr was a gift from AH Monsoro Burq (de Croze et al., 2011). All necessary sequences were obtained from Xenbase (www.xenbase.org/, RRID:SCR_003280).

Synthetic capped RNAs were prepared from pCS2+ derivatives with the mMessage mMachine kit (Ambion). Antisense oligonucleotides either with no 5′ capping or coupled to fluorescein were made by Gene Tools, as previously published: MObarhl2 1 (15 ng) (previously Xbarhl2ASI), MObarhl2 2 (30 ng) (previously Xbarhl2ASII), a control MO [MObarhl2 ct (30 ng)] had the same sequence as MObarhl2 2, but with five mismatches (previously Xbarhl2ASIII) (Offner et al., 2005), MOhdac1 (0.3 ng) (Tao et al., 2015) and control MO MOct (3 ng). The effects of MObarhl2 1 and MObarhl2 2 were similar, but MObarhl2 1 generated stronger developmental disruptions. We further used MObarhl2 2 and referred to it as MObarhl2. The specificity of the MObarhl2 effect has been previously demonstrated (Offner et al., 2005) with all necessary controls (reviewed in Blum et al., 2015). The MO doses we used are standard in the Xenopus community (Heasman et al., 2000; Liu et al., 2005; Sander et al., 2007; Hikasa et al., 2010). The overall barhl2 mRNA sequence was conserved at 100% between X. laevis and X. tropicalis in the first 25 nucleotides on which the MObarhl2 hybridized. We used MObarhl2 to deplete Barhl2 activity in X. tropicalis embryos. We used a previously validated translation-blocking MO against the X. laevis Hdac1.L (Tao et al., 2015). Tao et al. performed all the necessary controls to validate the specificity and effectiveness of their MO, including the rescue experiment (Tao et al., 2015). The Hdac1 LOF phenotype we described is similar to that described by Gao et al. (2016), with two different MOs against Hdac1. Before using MOHdac1, we performed a dose-response phenotype analysis and observed toxicity effects, including gastrulation defects. We consequently chose 0.5 ng/blastomere for MOHdac1, a dose that did not induce any sign of toxicity. The hybridization site for the X. laevis Hdac1.L MO has only two mismatches with that of Hdac1.S and hybridizes perfectly on its last 23 nucleotides with the X. tropicalis Hdac1 translation initiation site sequence. Therefore, MOHdac1 binds all Xenopus Hdac1 transcripts with good affinity. All sequences were obtained from Xenbase (RRID:SCR_003280). MOs were heated for 10 min at 65°C before use. Except when otherwise specified, MOs or mRNAs were co-injected with β Galactosidase (b gal) mRNA (100 pg) or gfp mRNA (100 pg) or MOct (MO standard, Gene Tools) coupled to fluorescein for lineage tracing. β Gal activity was revealed in Red-Gal (Research Organics) staining solution (Offner et al., 2005). Embryos were injected unilaterally for ISH/morphology and on both sides for RT-qPCR and ChIP-seq experiments. For all ISH analyses, three independent experiments were performed and the results were pooled. A phenotype exhibited by at least 60% of the embryos was considered significant. Except when indicated otherwise, after stage 10, the embryos are shown in dorsal view, anterior-side up. At stage 10 and below, embryos are shown in ventral view, dorsal-side up. For all rescue experiments, we tested a range of mRNA doses (50-500 pg), and selected the minimal mRNA quantity that both induced the specific phenotype and displayed no toxicity.

For each time point, an RNA extract was obtained from five individual embryos obtained from the same fertilization and the experiment was repeated at least twice. Embryos were dissociated by pipetting up and down with a 26-G needle in Trizol (Invitrogen) and extracted according to the manufacturer's instructions. RNA samples were reverse-transcribed twice, using MMLV-RT enzyme (Promega). RT-qPCR was carried out using IQ SYBR green Supermix (BioRad) on a CFX96 touch Real-Time PCR Detection System (BioRad), using standard procedures and technical triplicates for each time point. Primers were designed using Primer 3 (www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi/). Primer oligomers used in qPCR are detailed in Table S1. elongation factor 1 (ef1), ornithine decarboxylase (odc) and glyceraldehyde-3-phosphate dehydrogenase (gapdh) primers were used as controls when indicated.

ChIP was carried out with Xenopus embryos at stage 10 as described by Blythe et al. (2009). Briefly, lysates of 100 injected or control embryos were crosslinked with 1% formaldehyde for 30 min at room temperature (RT). The crosslinked samples were sonicated in RIPA buffer containing 50 mM Tris-HCl pH 7.4, 1% NP-40, 0.25% Na-Deoxycholate, 150 mM NaCl, 1 mM EDTA and 0.1% SDS with a Branson sonifier 450 cell disruptor (three rounds of 20 s, intensity of 1 with a 50% duty cycle). Protein A Sepharose beads (Sigma) were used to precipitate Tcf7l1-Flag (SIGMA F3165), Barhl2-Myc (Santa Cruz 9E10), or Histone H3 acetylated K9 (Abcam 4441). After washing, protein complexes were reverse crosslinked at 65°C overnight and treated with proteinase K. DNA fragments were purified with a DNA Purification Kit (Qiagen). qPCR for the gsc, chd, otx2, sia1, ef1 and gapdh promoters was performed and normalized to uninjected embryo controls according to the Applied Biosystems Protocol using SYBR Green (Thermo Fisher Scientific 659). For primers, see Table S2. Relative quantification was performed using the ΔΔC(t) method according to Blythe et al. (2009).

Control or injected embryos were collected at the indicated stages. Single ISHs were performed using digoxigenin (DIG)-labeled probes. Antisense probes were generated for barhl2, sox3, vent2, chd, not2, otx2, gsc, wnt8a, vent1 and gfp according to the manufacturer's instructions (RNA Labeling Mix, Roche). pCS2-Gfp was a gift from David Turner (University of Michigan, Ann Arbor, MI, USA). ISH was processed following the protocol described by Juraver-Geslin et al. (2014). Briefly, following overnight incubation with the probes and then with alkaline phosphatase-conjugated anti-DIG and anti-Fluorescein (Roche) antibody, enzymatic activity was revealed using NBT/BCIP substrates (Roche). Then, bleaching treatment was performed after the postfixation of the embryos. Whole-mount images were captured using a stereomicroscope (Lumar V12). Sections (40-μm thick) were cut using a Leica VT1000 vibratome after gelatin-albumin embedding. Pictures were captured using a digital Axiocam 506 color camera on a Zeiss microscope at the PICT-IBiSA@Orsay Histology facility of Institut Curie and processed with the Zen program (version2.3), ImageJ and Adobe Photoshop CS6 software.

To perform luciferase assays, HEK293T cells grown at ∼60% confluence in 24-well plates were transfected using lipofectamine (GIBCOBRL, 11668-027) with 50 ng of pTop-Flash or pFop-Flash (firefly luciferase) reporter plasmid and pCS2-β Gal as a transfection control. pTop-Flash activity was measured relative to the pFop-Flash reporter, an altered version of pTop-Flash carrying point mutations in the Tcf-binding sites to render it unresponsive to β-Catenin activity (Korinek et al., 1997). Proteins were extracted and a luciferase assay was performed according to the manufacturer’s instructions (Luciferase Assay System, Promega) using a bio-luminometer TriStar LB 941 (Berthold). Firefly luciferase activity was normalized to β Gal activity (Beta-Glo Promega). The assay was performed in triplicate.

Cells were transfected with 0.5-2 µg of expression vectors encoding tagged proteins as indicated using the Phosphate Calcium method. Plasmids coding for pCS2+ or pSK+ were used as a supplement to ensure that in the same experiment, cells in different wells were transfected with the same quantity of expression vectors and plasmids. Thirty-six hours after transfection, cells were harvested and lysed on ice in lysis buffer containing 20 mM Tris pH7.6, 150 mM NaCl, 1% Triton, 1 mM EDTA and c0mplete™ protease inhibitor (Roche). Whole-cell lysates were cleared via centrifugation at 14,000 rpm (18,400 g) for 15 min. Alternatively, 40 stage-10 embryos injected on both sides with 0.7 ng of synthetic RNA were homogenized on ice in lysis buffer containing 0.5% NP40, 5 mM EDTA, 10 mM Tris pH 7.5, 100 mM NaCl, c0mplete protease inhibitor and PhosSTOP (Sigma-Aldrich). Extracts were cleared by centrifugation for 30 min at 14,000 rpm (18,400 g). In both cases, protein complexes were precipitated from the whole-cell lysates with anti-c-Myc Tag antibody (clone 9E10). Protein complexes were then precipitated with protein A-Sepharose beads (Sigma). Beads were pelleted by centrifugation and washed four times with lysis buffer. Boiling the beads in loading buffer eluted immunoprecipitates. Western blotting with cell lysates or immunoprecipitated protein elutes was performed using conventional methods. Blots were blocked with 5% milk and were detected with an ECL Western Blotting Detection Reagent (Amersham). Antibodies used are detailed in Tables S3 and S4.

We performed the analysis on ISH of chd-stained X laevis stage-10 embryos. Separation between the control and the injected sides was based on tracer staining. To perform an objective comparison of the different expression levels of RNA by ISH, the RGB images were analyzed using an ImageJ macro (http://rsb.info.nih.gov/ij/) (Abramoff et al., 2004; Schneider et al., 2012). In such a way that the expression level is proportional to the signal recorded on the blue channel, RGB images were split and only pixel values corresponding to the blue channel were analyzed. With the goal of removing unspecific signals, background value, estimated as the average intensity of pixels outside the region of interest (ROI), were subtracted from each image. The ROI was determined by automatic segmentation using IsoData autothresholding-based determination (Ridler and Calvard, 1978). This prevents any subjectivity in ROI determination. Surface and mean pixel density on ROIs were computed for each background-corrected image. Pictures are shown corresponding to original data after histogram equalization using ImageJ with the percentage of saturated pixel 0.4. The macro is available from the authors upon request and will be available as a plug-in in ImageJ.

Except otherwise specified, all experiments were repeated at least three times. All statistical analyses were performed using the Real statistics plugin from Microsoft Excel. For both firefly luciferase activity and images analyses, the Gaussian distribution of the data between embryos for the data sets and sections of one embryo, respectively, were verified using the Shapiro Wilk statistic test. The means between two different data sets were compared using the paired sample t-test using the Real statistics plugin from Microsoft Excel, considering that samples were independent and variances unequal. All tests were two-tailed. N is the number of embryos. Significant differences are indicated in the figures with asterisks as follows: *P≤0,05; **P≤0,01; ***P≤0,001. For all rescue experiments, the means between different data sets were compared using the Wilcoxon signed-rank test.

We thank Muriel Umbhauer, Phil Jones, Nancy Papalopulu, Muriel Perron, David Turner for gifts of materials; Sergio Marco for his advice on image processing and ImageJ macro writing; Simon Saule, Muriel Perron, Geneviève Almouzni for help and advice. The authors greatly acknowledge the PICT-IBiSA@Orsay Histology and Imaging Facilities of the Institut Curie supported by the French National Research Agency through the “Investments for the Future” program (Fran-BioImaging, ANR-10-INSB-04), the “plateau technique de biologie moléculaire” of I2BC (Véronique Henriot) for construction of plasmids, Elodie Belloir, Virginie Dangles-Marie and Isabelle Grandjean for animal care, Marion Wassef, Clemence Carron-Homo and Martin Raff for their comments on the manuscript, and Paul K. Johnson and Martin Raff for their editing work on the manuscript.