The definition of embryonic potency and induction of specific cell fates are intimately linked to the tight control over TGFβ signaling. Although extracellular regulation of ligand availability has received considerable attention in recent years, surprisingly little is known about the intracellular factors that negatively control Smad activity in mammalian tissues. By means of genetic ablation, we show that the Smad4 inhibitor ectodermin (Ecto, also known as Trim33 or Tif1γ) is required to limit Nodal responsiveness in vivo. New phenotypes, which are linked to excessive Nodal activity, emerge from such a modified landscape of Smad responsiveness in both embryonic and extra-embryonic territories. In extra-embryonic endoderm, Ecto is required to confine expression of Nodal antagonists to the anterior visceral endoderm. In trophoblast cells, Ecto precisely doses Nodal activity, balancing stem cell self-renewal and differentiation. Epiblast-specific Ecto deficiency shifts mesoderm fates towards node/organizer fates, revealing the requirement of Smad inhibition for the precise allocation of cells along the primitive streak. This study unveils that intracellular negative control of Smad function by ectodermin/Tif1γ is a crucial element in the cellular response to TGFβ signals in mammalian tissues.
The TGFβ cascade is a fundamental player in mammalian development and adult tissue homeostasis. TGFβ signals through cognate serine/threonine receptors and leads, intracellularly, to the activation of the R-Smad/Smad4 transcriptional complex (Moustakas and Heldin, 2009). Although TGFβ ligands are widely expressed in tissues, they can elicit their effects in a strict temporally and spatially controlled manner. For the signal to reach only the appropriate cells and with the correct intensity, mechanisms must be in place to determine where and when cells must not respond to TGFβ. This layer of regulation is just as likely to play a key role in defining cell fate as the signal itself, as suggested by phenotypes emerging from inactivation of extracellular TGFβ antagonists (Bachiller et al., 2000; Perea-Gomez et al., 2002; Zacchigna et al., 2006).
In addition to regulation in the extracellular space, intracellular control mechanisms also exist. One example is the phosphorylation/dephosphorylation cycle of receptor-Smads (Itoh and ten Dijke, 2007; Lin et al., 2006). Recently, we have proposed a parallel layer of control of Smad activity centered on a cycle of monoubiquitylation and deubiquitylation of Smad4, which is mediated, respectively, by ectodermin (Ecto, also known as Tif1γ or Trim33) and FAM/Usp9x (Dupont et al., 2009). Through these inhibitory systems, Smad transcriptional complexes are disassembled and R-Smads are forced to exit the nucleus and check the activity status of the receptors. It has been proposed that this mechanism avoids saturation of the signaling cascade, maintaining Smad activity proportional to – and finely tunable by – variations in extracellular ligand concentrations (Moustakas and Heldin, 2009). Despite these speculations, it remains unclear to what extent these negative regulatory steps impact on TGFβ responsiveness in vivo. Here, we used the mouse embryo as a model system to tackle this issue.
During early vertebrate embryogenesis, the graded activity of the TGFβ-related factor Nodal orchestrates the maintenance or restriction of embryonic pluripotency and establishes the body plan. In the mouse, Nodal induces and patterns the anterior visceral endoderm (AVE), and sustains trophoblast development. These tissues then provide fundamental instructive signals to the epiblast, cooperating with Nodal itself to induce the mesoderm and endoderm germ layers and to pattern them along the anteroposterior axis during gastrulation (Arnold and Robertson, 2009; Tam and Loebel, 2007). In recent years, systematic inactivation of positive transducers of the Nodal pathway indicated that cells of the embryo are able to interpret very subtle variations in Nodal signaling, as indicated, for example, by the phenotype of Nodal hypomorphic alleles (Norris et al., 2002) or by the requirement of Smad4 only for high-threshold responses (Chu et al., 2004). Yet, what generates such graded Nodal signaling activity in vivo is less clear. In this paper, we provide evidence that a negative intracellular Smad regulator, ectodermin, plays an essential role in how cells read TGFβ signals.
MATERIALS AND METHODS
Generation of Ecto knockout and conditional alleles
To generate the Ecto/Tif1g targeting vector, a genomic clone spanning exons 2, 3 and 4 was used (Yan et al., 2004). Briefly, a loxP flanked (floxed) PGK-Neo cassette was inserted within the first intron, and a third loxP site was inserted within the fourth intron (see Fig. S1 in the supplementary material). The targeting fragment was electroporated into 129/Sv H1 ES cells as described previously (Cammas et al., 2000). After selection, neomycin-resistant ES clones were expanded, and their genomic DNA was screened by PCR. Positive clones were further validated with Southern blotting analysis with two independent probes (not shown). ES cells bearing the correctly targeted allele were injected into C57BL/6 blastocysts to produce chimeric offspring. These were backcrossed with C57BL/6 mice, and their offspring was genotyped by PCR. Mice heterozygous for the targeted allele were then crossed with CMV-Cre transgenic mice (Dupe et al., 1997), and the offspring was analyzed by PCR to identify animals with either complete recombination of the loxP sites (null allele, Ecto–) or lacking of the PGK-Neo cassette owing to recombination of the first and second loxP sites (conditional allele, Ecto fl). Cre-negative Ecto+/– and Ecto fl/fl mice were subsequently kept on a C57BL/6 background for phenotypic analyses. Animal care was in accordance with our institutional guidelines.
Generation of Ecto-EpiKO and compound Ecto–/–; Smad4–/–, Ecto–/–; NodalΔ600/– embryos
To obtain epiblast-specific Ecto knockout embryos, Sox2-Cre; Ecto+/– males were crossed with Ecto fl/fl females. In this setup, the Sox2-Cre transgene selectively deletes the floxed alleles in ICM/epiblast cells (Hayashi et al., 2002). Embryos were genotyped after in situ hybridization for Ecto fl, Ecto+, Ecto– and Cre alleles. Embryos were scored as mutants in the presence of Cre, Ecto fl, Ecto– and absence of Ecto+ alleles.
To obtain embryos homozygous null for both Ecto and Smad4, Ecto fl/fl; Smad4 fl/fl (Bardeesy et al., 2006) males were crossed with CAG-Cre; Ecto+/–; Smad4+/– females. In this setup, the Cre protein supplied by the mother within the oocyte completely recombinates the paternal floxed alleles after fertilization, irrespective of transgene transmission (Sakai and Miyazaki, 1997), raising the expected frequency of compound null embryos to 25%. Embryos were genotyped after in situ hybridization for Ecto fl (recognizing also the Ecto+ allele), Ecto–, Smad4 fl (recognizing also the Smad4+ allele) and Smad4– alleles. Embryos were scored as compound mutants in the presence of Ecto– and Smad4–, and in the absence of Ecto fl and Smad4 fl alleles.
To obtain Ecto–/– embryos with reduced Nodal signaling, Ecto+/–; Nodal+/– [lacZ allele (Collignon et al., 1996)] mice were crossed with Ecto+/–; Nodal+/Δ600 (Norris et al., 2002) mice. Embryos were genotyped after in situ hybridization for Ecto+, Ecto–, lacZ and NodalΔ600 alleles.
Mouse embryos were staged based on their morphology, considering the morning of the vaginal plug as E0.5. Embryos were manually dissected in ice-cold DEPC-treated phosphate-buffered saline (PBS) and fixed overnight in PBS 4% PFA at 4°C, dehydrated (for storage) and rehydrated through methanol series. Whole-mount in situ hybridizations were performed according to http://www.hhmi.ucla.edu/derobertis/ (Xenopus ISH protocol), with minor modifications to ensure efficient genotyping after staining: day 1, post-fixing after proteinase K treatment was carried out with 4% PFA only, 1 hour at 4°C; day 3, washes were carried out with PBS 0.5% goat serum (GS, Invitrogen), without AP1 incubation before BM-Purple staining (Roche), and without post-fixation. Embryos were mounted in 80% glycerol and photographed with a Leica DMR microscope equipped with a Leica DC500 camera. Unless otherwise indicated, embryos of different genotypes stained with the same marker are shown at the same magnification. For each experiment, at least five embryos of every genotype were analyzed with consistent results.
For immunostaining, embryos were fixed overnight in PBS 4% PFA supplemented of phosphatase inhibitors (Sigma) at 4°C, dehydrated and rehydrated through methanol series. Embryos were permeabilized with two washes in PBS 0.5% NP40 for 20 minutes at 4°C, followed by one wash in PBS 0.3% Triton X-100 for 20 minutes at room temperature. After two washes in PBS 0.1% Triton X-100 (PBT) for 15 minutes at room temperature, embryos were blocked with two washes in PBT 10% GS for 1 hour at room temperature, and incubated overnight with rabbit anti-Ecto primary antibody (Sigma HPA004345, 1:75) in PBT 10% GS or in rabbit mAb anti-phospho-Smad2 (CST-3108, 1:50) in PBT 3% BSA. The following day, embryos were washed twice in PBT 2% GS for 15 minutes at 4°C, and five more times in PBT 2% GS for 1 hour at 4°C. Secondary Alexa555 goat anti-rabbit antibody (1:200) was incubated overnight in PBT 5% GS. The third day, embryos were washed five times in PBT for 15 minutes at room temperature, mounted in 80% glycerol and photographed with a Nikon Eclipse E600 confocal microscope equipped with a Bio-Rad Radiance2000 camera/laser scanning system. Nuclear localizations were confirmed by colocalization with YOYO1 staining (Invitrogen). Specificity of the phospho-Smad2 signal was confirmed by incubating E6.0 wild-type embryos for 8 hours in 10 μM SB431542 TGF-β receptor inhibitor, causing disappearance of the signal (not shown).
For histological analysis, deciduae were collected in PBS, fixed in Bouin's overnight, dehydrated and embedded in paraffin. Serial sections were cut at 6 μm and stained with Hematoxylin and Eosin according to standard procedures. Similar procedures were applied to obtain sections of embryos after in situ.
Offspring were genotyped by PCR on genomic tail DNA extracted by standard procedures. After in situ, individual embryos were manually dissected with a tungsten wire (FineScienceTools) to eliminate the EXE and ectoplacental cone, thus avoiding maternal DNA contaminations. Epiblast/VE tissues were lysed overnight at 55°C with mild agitation in 10 mM Tris/HCl (pH 8.0), 50 mM KCl, 2 mM MgCl2, 0.3% Tween-20, 0.5% NP-40 supplemented with fresh proteinase K (Invitrogen, 1:40). Lysis volume was adjusted according to the stage: E5.5, 20 μl; E6.5, 40 μl. After vortexing, proteinase K was inactivated for 10 minutes at 95°C, quenched on ice, and samples were centrifuged for 10 minutes at 4°C at 10,000 g. 4ul of the fresh supernatants were used for each PCR reaction using EX-Taq polymerase (Takara). For detection of the Ecto– allele in embryos of early stages, nested PCR was employed if necessary.
TS cell culture and RT-PCR analysis
TS cells were cultivated and passaged in feeder-free conditions as indicated previously (Oda et al., 2006). pLKO lentiviral shRNA targeting mouse Ecto was purchased from Sigma (5′-CCGGCGTGTGATAGATTGACGTGTACTCGAGTACACGTCAATCTATCACACGTTTTTG-3′). Control shGFP sequence was as described previously (Adorno et al., 2009). Lentivirally infected populations were established by puromycin selection as indicated previously (Moffat et al., 2006). For differentiation assays, TS cells were seeded and grown for 2 days in stem-cell medium; undifferentiated samples were allowed to differentiate further in the same conditions for 2 days; differentiated samples were changed to DMEM 10% FCS (t=0) and cultivated for the indicated times, renewing the culture medium every 2 days. TGFβ stimulation was provided by adding every day 100 ng/ml Activin-A (Peprotech) directly to the medium. Cultures were harvested in Trizol (Invitrogen) for RNA extraction, and contaminant DNA was removed by DNAse treatment. Real-time qPCR analyses were carried out on triplicate samplings of retrotranscribed cDNAs with RG3000 Corbett Research thermal cycler and analyzed with Rotor-Gene Analysis6.1 software. Experiments were performed at least twice, with duplicate biological replicates.
Ectodermin is required for early mouse embryonic patterning
To investigate the role of Ecto in vivo, we generated Ecto conditional and germline knockout alleles (see Fig. S1 in the supplementary material for details on the targeting procedure and validation of effective loss-of-Ecto). Mice heterozygous for the Ecto-null mutation (Ecto+/–) were viable and fertile; however, homozygosity resulted in embryonic lethality. Indeed, embryos from heterozygote intercrosses were collected at different stages of gestation and Ecto mutants could be recovered at the expected Mendelian ratios at E5.5 to E7.5, but not at later stages.
Morphological and histological analyses demonstrated that Ecto mutants display striking defects in embryonic polarity and tissue patterning. When compared with control littermates, E6.5 Ecto mutants were smaller and lacked a clear distinction between epiblast and extra-embryonic ectoderm (EXE). Wild-type embryos formed mesoderm as a consequence of gastrulation; by contrast, Ecto mutants could readily be identified by the undivided proamniotic cavity and the lack of a primitive streak (Fig. 1A,B). Defective mesoderm formation was confirmed by in situ hybridization at early streak stage examining the expression of markers, such as T, Eomes and Wnt3 (Fig. 1C,D; see Fig. S2 in the supplementary material). For these analyses and throughout the study, we analyzed at least five embryos of each genotype with consistent results; figures show representative phenotypes.
At first, lack of mesoderm in Ecto mutants came as a surprise, as this phenotype was opposite to the excessive mesoderm differentiation displayed by Ecto-depleted Xenopus embryos (Dupont et al., 2005). However, in contrast to amphibians, mesoderm formation in the mammalian embryo is a late event, requiring inputs from the EXE and AVE extra-embryonic lineages. As the development of such tissues relies on the activity of early-acting Nodal/Smad4 signaling (Arnold and Robertson, 2009), we tested whether defects in Ecto mutants initiated with abnormal extra-embryonic development. Expression of AVE markers at E5.5 was strikingly upregulated in Ecto mutants: when these markers were barely detectable in wild-type littermates, signals of the Nodal targets Cerberus-like (Cerl; Cer1 – Mouse Genome Informatics), Lefty1 and Lim1 (Lhx1 – Mouse Genome Informatics) mRNAs were already strong in knockout embryos, becoming rapidly saturated in an abnormally broad AVE domain (Fig. 1E-H and not shown). Although in E6.5 wild-type embryos AVE markers are usually restricted to an anterior narrow stripe of cells, in Ecto mutants, robust Cerl and Lim1 expression was expanded around the epiblast (Fig. 1I-L; Fig. S2G,H in the supplementary material).
Ectodermin is expressed ubiquitously in early mouse embryos by immunofluorescence (see Fig. S1E,G in the supplementary material and data not shown). Genetic evidence indicates that AVE responds to Nodal ligands emanating from the epiblast (Lu and Robertson, 2004). Thus, we next tested the possibility that AVE expansion in Ecto mutants is caused by a cell-autonomous enhanced Smad responsiveness, as opposed to being secondary to increased ligand expression/availability in the epiblast. To achieve this, we made use of the paternally inherited Sox2-Cre transgene, recombining the Ecto conditional allele in the epiblast lineage specifically (Sox2-Cre; Ecto fl/- embryos, hereafter Ecto-EpiKO; see Fig. S1G,H in the supplementary material for epiblast-specific protein depletion) (Hayashi et al., 2002; Di-Gregorio et al., 2007). In EpiKO mutants, a genetically wild-type AVE did not display any of the abnormalities characterizing the Ecto germline mutants, as Cerl and Lefty1 mRNAs were comparable in localization and intensity with wild-type embryos (Fig. 1M,N and not shown). In line with a cell-autonomous role for Ecto in AVE cells, at E5.5, Nodal is expressed normally in Ecto mutant embryos (Fig. 1O,P) and, by immunofluorescence, Smad2 phosphorylation is comparable between wild type and Ecto mutants (Fig. 1Q,R). This is in agreement with our previous observations indicating that Ecto inhibits TGFβ signaling acting specifically on Smad4 availability and not on R-Smads (Dupont et al., 2009). Together, these findings suggest that Ecto is required cell-autonomously to restrain Nodal responsiveness in AVE cells.
Defective AVE patterning of Ecto mutants is caused by unrestrained Nodal/Smad4 signaling
The phenotype of Ecto–/– embryos is opposite to those reported for Nodal, Smad2 and Smad4 knockouts (Brennan et al., 2001; Waldrip et al., 1998; Yang et al., 1998). Hence, we investigated the genetic relationships between Ecto and its biochemical target Smad4 (Dupont et al., 2009). We analyzed embryos from crosses of mice carrying the floxed alleles for the two genes (Ecto fl/– and Smad4 fl/–) that were undergoing zygotic deletion in the CAG-Cre maternal background (Sakai and Miyazaki, 1997) (see Materials and methods for details). Ecto fl/–;CAG-Cre embryos lacked of endogenous Ecto protein (not shown) and were phenotypically indistinguishable from Ecto germline homozygous mutants (compare Fig. 2B,F with Fig. 1F,H); Smad4 fl/–;CAG-Cre phenocopied morphologically the previously reported defects of the null allele (Yang et al., 1998). Extending these studies, we found that Smad4 is dispensable for VE specification (as revealed by the detection of the Afp marker, see Fig. S3 in the supplementary material), but required for Cerl and Lim1 induction (Fig. 2C,G). In line with previous biochemical findings, double mutants for Smad4 and Ecto were indistinguishable from Smad4 mutants (Fig. 2C,D,G,H). Thus, Ecto acts as inhibitor of Smad4-dependent signaling, and does not regulate AVE formation through an alternative Smad4-independent pathway.
Data presented so far suggest that disruption of the Ecto/Smad4 inhibitory axis leads to excessive Nodal responsiveness in AVE. If so, this should be rebalanced by a concomitant reduction of the Nodal dose. To this end, we combined Ecto mutant with a strongly attenuated Nodal mutant (NodalΔ600/–) (Norris et al., 2002), leading to a rescue of AVE patterning (Fig. 2I-K). Taking into account the cell-autonomous role of Ecto shown above, these results collectively suggest that the net activity of Nodal signaling, at least for AVE induction, is the result of two components: extracellular ligand availability and negative control of Smad responsiveness. Loss of the latter in Ecto mutants is sufficient to profoundly alter embryonic patterning.
Ecto maintains EXE self-renewal by opposing Nodal signaling
Next, we characterized molecularly the development in Ecto mutants of the other extra-embryonic tissue, the trophoblast lineage. As shown in Fig. 3, the trophoblast stem (TS) cells and EXE markers Eomes, Cdx2 and Bmp4 were undetectable in E5.5 Ecto–/– embryos (Fig. 3A-F). This represents a cell-autonomous requirement as Ecto-EpiKO embryos displayed normal EXE development (Fig. 3G,H). Lack of EXE in Ecto mutants is paradoxically similar to the phenotype of Nodal mutants (Brennan et al., 2001); however, in the case of Nodal, this is secondary to defective epiblast patterning where Nodal sustains Oct4 and Fgf4 transcription, which, in turn, maintains TS self-renewal (Guzman-Ayala et al., 2004; Lu and Robertson, 2004; Mesnard et al., 2006). By contrast, Fgf4 and Oct4 are normally expressed in Ecto mutants (Fig. 3I-L). Strikingly, Nodal attenuation rescued the EXE phenotype of Ecto mutants, as Eomes and Bmp4 transcripts were invariably rescued in combined Ecto–/–; NodalΔ600/– or Ecto–/–; Nodal+/– embryos (Fig. 3M-P for Bmp4 expression, see Fig. 5A-D for Eomes). Taken together, these data suggest that Ecto protects the TS lineage from excessive Nodal signaling.
To understand the nature of Ecto function in EXE, we monitored TS induction from earlier developmental stages. At earlier stages, Cdx2 was expressed in Ecto mutants (Fig. 4A,B), indicating that excessive Nodal responsiveness affects later events. We then monitored cell viability, and found comparable apoptosis and proliferation rates in wild-type and mutant embryos (see Fig. S4 in the supplementary material; and data not shown). As development proceeds, we found that Ecto mutants do retain expression of Spc4 (Pcsk6 – Mouse Genome Informatics) and Pem (Rhox5 – Mouse Genome Informatics) identifying the presence of more differentiated cells of the ectoplacental cone (Constam and Robertson, 2000; Lin et al., 1994) (Fig. 4C-F), but lose expression of Mash2 (Ascl2 – Mouse Genome Informatics), a marker for transit-amplifying trophoblast progenitors (Guillemot et al., 1995) (Fig. 4G,H). These data suggest that Nodal signaling also plays a direct role on trophoblast cells, promoting their differentiation.
To validate this hypothesis, we established control (shGFP) and Ecto-depleted (shEcto) mouse TS populations by lentiviral infection, and compared them for the expression of stem and differentiation markers (Fig. 4I,J). When cultured in stemness/proliferating medium, Control and shEcto TS cells were comparable in terms of marker expressions and cell cycle profiles (Fig. 4J and not shown), reinforcing the notion that Ecto is not required for TS cells induction or self-renewal. However, once TS cells were induced to differentiate, in the presence of the Nodal-related ligand Activin shEcto cells specifically displayed a robust increase in the expression of differentiation markers 4311 (Tanaka et al., 1998) and Gcm1 (Anson-Cartwright et al., 2000) (Fig. 4J), recapitulating in vitro our observations on Ecto mutants. Comparable results were obtained with an independent shRNA targeting Ecto (not shown). Tight control over Nodal activity is thus crucial for balancing stem cells renewal and differentiation in the trophoblast lineage; in Ecto mutants, uncontrolled Nodal signaling causes wholesale exhaustion of the stem cell pool (see model in Fig. 4K).
Nodal attenuation rescues mesoderm formation in Ecto mutants
By losing the EXE, Ecto mutants are deprived of an essential source of mesoderm inducing and patterning signals, including BMP4 (Arnold and Robertson, 2009); at the same time, they display enhanced expression of Nodal antagonists, such as Cerl and Lefty1. This raises questions about the primary cause of defective mesoderm in Ecto mutants. Remarkably, attenuation of Nodal signaling in compound Ecto/Nodal mutants rescues mesoderm development, as revealed by transcription of the pan-mesodermal markers Eomes and T at the early gastrula stage (Fig. 5A-H). Interestingly, although the combination Ecto–/–; NodalΔ600/– rescues EXE, mesoderm and AVE (Fig. 5D,H and Fig. 2K), compound Ecto–/–; Nodal +/– could rescue defective EXE and mesoderm but not AVE expansion (compare Fig. 5C,G with Fig. 5K). This suggests that lack of mesoderm in Ecto mutants is primarily due to lack of EXE; molecularly, this can be explained by the failure to induce BMP4 expression (Fig. 3) that represents an important mediator of a feed-forward loop between the EXE and epiblast feeding on Nodal expression and mesoderm induction (Arnold and Robertson, 2009).
A further complicating issue is the fact that AVE and EXE development might be linked, as the EXE has also been proposed to secrete AVE inhibiting factors (Rodriguez et al., 2005; Yamamoto et al., 2009). Is then the AVE expansion observed in Ecto mutants due to loss of EXE? Our results suggest this is not the case, because in Ecto–/–; Nodal +/– embryos these events are uncoupled (compare Fig. 3O and Fig. 5K): these compound mutants display rescued EXE in the presence of a still expanded AVE. Thus, data support the view that expanded AVE in Ecto mutants is primarily due to enhanced Nodal responsiveness of the visceral endoderm. Clearly, the loss of BMP expression in the EXE might amplify, to some extent, the enlarged AVE domain of the Ecto mutants.
A role for Ecto in restraining anterior meso-endoderm formation
The Sox2-Cre; Ecto fl/– embryos (Ecto-EpiKO) allow the more direct study of the role of Ecto in the epiblast, bypassing its early requirements in extra-embryonic tissues. Previous work established that a gradient of Nodal/Smad activity patterns the primitive streak (Dunn et al., 2004; Lowe et al., 2001; Vincent et al., 2003); in this context, Smad4 is required for peak signaling levels, namely, for the formation of the anterior primitive streak and node, marked by Foxa2 expression (Chu et al., 2004). Strikingly, we found that approximately one-third (4/13) of the Ecto-EpiKO embryos displayed an expanded Foxa2 expression at streak stages (Fig. 6A-6B′). These embryos appeared smaller, lacked an overtly elongated streak and probably failed to undergo proper gastrulation. At later stages, surviving Ecto-EpiKO embryos showed expansion of the Node (marked by Foxa2 staining, Fig. 6C,D), an almost radial expansion of the definitive endoderm marker Cerl (Fig. 6E,F), as well as duplications of Node and anterior axial mesendoderm tissues (T, Shh and Chordin in situs, Fig. 6G-J and not shown). Together, the data suggest that Ecto is essential for orchestrating the intensity of Nodal/Smad4 responses for proper primitive streak development. These early defects of Ecto-EpiKO are such that loss of Ecto in epiblast cells is incompatible with subsequent development. Indeed, we could identify only few Ecto-EpiKO embryos at E10.0, displaying defective brain development and open neural folds (not shown).
In this paper, we show that cell-autonomous Smad regulation operated by the Smad4 ubiquitin-ligase ectodermin is essential to dose Nodal responsiveness in mouse embryos. Analysis of Ecto mutants showed that loss of Ecto `upgrades' Nodal responses in extra-embryonic and embryonic lineages.
In the visceral endoderm, unrestrained Nodal responsiveness causes a massive expansion of the Cerl/Lefty1 expressing AVE territory; this is Smad4-dependent and can be rescued by reducing the dosage of Nodal. Thus, it appears that, in vivo, the net activity of Nodal/TGFβ is the result of a combination of two elements: extracellular ligand availability, which is defined by the expression of Nodal and its antagonists, and translates into receptor activation and Smad2/3 phosphorylation; and negative control over Smad4 availability. In Ecto mutants, loss of this second layer of control is sufficient to profoundly alter embryonic patterning.
Novel functions of Nodal are revealed by this analysis. In the EXE, excess of Nodal responsiveness in Ecto mutants leads to the wholesale differentiation and exhaustion of the trophoblast stem (TS) cell compartment. This had not previously inferred by Nodal or Smad loss-of-function analyses. Indeed, in Nodal mutants, EXE is induced but not maintained, a phenotype that superficially overlaps with that of Ecto mutants. Marker analysis in fact revealed a profound difference: although in Nodal mutants the trophoblast transient amplifying progenitors become expanded (Guzman-Ayala et al., 2004), in Ecto mutants this cellular pool is instead depleted, in favor of more differentiated cellular progenies (Fig. 4). This suggests that Nodal signaling drives – and Ecto inhibits – trophoblast differentiation; the ensuing equilibrium allows the homeostatic expansion and differentiation of trophoblast progenitors.
Secondary to deficiencies in extra-embryonic tissues, Ecto mutants ultimately lack primitive streak and mesoderm induction; notably, this defect in the embryo proper can be paradoxically rescued by a reduction of Nodal dose, because this normalizes extra-embryonic development. Mesoderm induction requires inputs from the EXE but is inhibited by Nodal antagonists emanating from the AVE; these are respectively missing and enhanced in Ecto mutants. So, what is the nature of their mesodermal defect? Our data suggest a primary role in the defective EXE, as in combined Ecto–/–; Nodal+/– embryos the AVE remains expanded but EXE and mesoderm formation is rescued.
In the epiblast, the generation of different mesoderm and endoderm derivatives appears to be a response to exposure to different intensities of Nodal signaling (Dunn et al., 2004; Lowe et al., 2001; Vincent et al., 2003). Although complete loss of Nodal prevents germ layer formation, smaller reductions primarily affect the anterior derivatives of the primitive streak. Similarly, Smad4 appears required for anterior but not posterior primitive streak derivatives (Chu et al., 2004). However, it remains unclear whether an anteroposterior extracellular Nodal gradient exists, also considering that Nodal is evenly expressed along the primitive streak (Tam and Loebel, 2007). Our data suggest that intracellular control of Smad activity by Ecto plays a crucial role in these morphogenetic effects. In contrast to Smad4 deficiencies, loss of Ecto leads to an expansion of anterior primitive streak and its derivatives, including Node and definitive endoderm.
This work also contributes towards resolving an issue regarding the function of ectodermin/Tif1γ in TGFβ signal transduction. We have discovered Ecto as TGFβ antagonist in an unbiased expression screen for determinants of germ-layer identity in the frog embryo (Dupont et al., 2005); others independently isolated the same molecule biochemically, as a Smad-interacting factor, and suggested that Ecto may act as Smad2/3 partner to mediate an alternative Smad4-independent TGFβ pathway (He et al., 2006). However, our genetic evidence supports the view of Ecto as inhibitor of canonical Nodal/TGFβ signaling, as defects of Ecto mutants depend on Smad4 and are rescued by reducing Nodal dose.
In summary, this study reveals how cell-autonomous negative modulation of Smads signaling endows embryonic cells with distinct interpretational keys to Nodal signals. This orchestrates the development of embryonic cells into distinct pluripotent cell lineages of the early mouse embryo and, probably, adult tissue homeostasis. An interesting possibility for future studies will be to determine whether Ecto activities are themselves patterned in vivo and whether this crucial regulatory layer can be exploited therapeutically in diseases characterized by excess of TGFβ activity, such as fibrosis or metastasis.
This paper is dedicated to Regine Losson who recently passed away. We thank J. Collignon, T. Rodriguez, J. Rossant, J. A. Belo, D. Constam, M. Pfeffer, S. Thilgam, G. Liguori for gifts of plasmids and cell lines. We are particularly grateful to E. J. Robertson for providing us Nodal-lacZ, NodalΔ600 and Sox2-Cre mice, to R. DePinho for Smad4-floxed mice and to J. Miyazaki for the CAG-Cre line. M. Cordenonsi, O. Wessely, G. Minchiotti, D. Volpin and all members of the Piccolo group offered invaluable discussions and comments on the manuscript. This work is supported by grants from the Centre National de la Recherche Scientifique (CNRS) to R.L. and from Fondazione Telethon, Italian Association on Cancer Research (AIRC) and Fondazione Cariparo (Excellence-grant) to S.P. and S.D. E.E. is recipient of a Fondazione Cariparo PhD fellowship.
Competing interests statement
The authors declare no competing financial interests.
L.M., S.D. and S.P. designed experiments and analyzed data. L.M. performed in situs and post in situ embryo genotyping; E.E. performed breedings, immunofluorescence staining and helped with in situs; M.A. performed real-time PCR and TS cell experiments with S.D., and colony genotyping with S.S.; K.Y., O.W., M.M., K.K., P.C. and R.L. designed and generated the Ecto knockout alleles; S.P. and S.D. wrote the paper.
Supplementary material for this article is available at http://dev.biologists.org/lookup/suppl/doi:10.1242/dev.053801/-/DC1
- Accepted May 26, 2010.
- © 2010.