Embryonic mesoderm and endoderm induction requires the actions of non-embryonic Nodal-related ligands and Mxtx2

Vertebrate mesoderm and endoderm (mesendoderm) formation requires Nodal proteins, a subclass of TGFβ superfamily ligands (Schier and Shen, 2000). Altered NODAL expression in humans is associated with heterotaxy (MIM ID: 601265), congenital heart disease and holoprosencephaly (Roessler et al., 2009) and melanoma (Topczewska et al., 2006). The mechanisms underlying the zygotic activation of Nodal family gene transcription are incompletely understood. In Xenopus, transcriptional initiation of Nodal-related gene expression requires the maternal transcription factor VegT (Kofron et al., 1999), but an analogous mechanism not been found in other species. In zebrafish, maternal β-catenin signaling accounts for Nodal-related (ndr1/2) gene activation in the dorsal organizer (Kelly et al., 2000), but the factors initiating zygotic Nodal-related gene transcription at more ventral positions are unknown. Some of these factors appear to reside in an extraembryonic domain termed the yolk syncytial layer (YSL), which harbors mRNA-encoded signals that induce adjacent cells of the embryonic margin to differentiate as mesendoderm (Carvalho and Heisenberg, 2010; Chen and Kimelman, 2000; Mizuno et al., 1996). However, assessing the role of the YSL in initiating Nodal-related gene expression or mesendoderm differentiation is complicated by the presence of Nodal-related gene transcripts in both the YSL and mesendoderm precursors.

Indeed, in zebrafish, as in mice, the relative roles of Nodal-like activities from embryonic, extraembryonic and maternal sources are only partially understood. Whereas positive feedback of Nodal signaling occurs in embryonic cells, this mechanism appears to be excluded from the YSL (Fan et al., 2007; Harvey and Smith, 2009). Expression of the zebrafish Nodal-related genes ndr1 (sqt) and ndr2 (cyc) in the YSL is required for a degree of mesoderm and endoderm induction (Fan et al., 2007), but the complete genetic disruption of ndr1 and ndr2 causes more severe mesendoderm deficits, as characterized by the absence of endoderm and anterior (head and trunk) mesoderm in ndr1;ndr2 mutants (Feldman et al., 1998). Finally, a requirement of maternal Ndr1 for dorsal and anterior axis formation is observed in certain contexts (Gore et al., 2005; Hagos et al., 2007), but not in others (Bennett et al., 2007b; Pei et al., 2007).

We have investigated the function of genes expressed in the YSL and report here that the transcription factor Mxtx2 is required for the extraembryonic expression of ndr2 and sox32 (cas), a pan-endodermal marker (Kikuchi et al., 2001), as well as for the Nodal signaling-independent expression of the pan-mesodermal marker no tail a (ntla) (Schulte-Merker et al., 1994) in embryonic cells. We further find that co-disruption of extraembryonic Ndr2, extraembryonic Ndr1 and maternal Ndr1 prevents the differentiation of endoderm and anterior mesoderm, thereby recapitulating the phenotype of embryos completely deficient for Nodal signaling via removal of exclusively non-embryonic factors.

The AB strain was used throughout and the oep^tz57 mutant allele was used to generate the MZoep line, as previously described (Gritsman et al., 1999). For ndr1 mutants, the ndr1^hi975 allele (Amsterdam et al., 2004; Pei et al., 2007), a presumed null allele, was used throughout. Wild-type (WT) males were crossed with ndr1^–/– females to produce embryos lacking maternal Ndr1 (i.e. lacking maternally derived Ndr1; also termed Mndr1 embryos). Embryos lacking zygotic Ndr1 (also termed Zndr1 embryos) and their control siblings were generated in equal proportions by crossing ndr1^–/– males with ndr1^+/– females. Embryos lacking maternal and zygotic Ndr1 (also termed MZndr1 embryos) were generated by crossing ndr1^–/– males with ndr1^–/– females. For embryo production from parents of identical genotypes, up to three males and three females were placed together in the morning and embryos were collected immediately after spawning. Incubations were at 28.5°C or room temperature, and staging was according to Kimmel (Kimmel et al., 1995). Animal handling was consistent with international guidelines.

For the mxtx2 mRNA expression construct, cDNA was PCR amplified (mxtx2-F, 5′-ATGAAAGACATGTGGACTGACTGC-3′; mxtx2-R, 5′-AAGATCCATCTGGCCATTGAAGC-3′) from single-stranded cDNA prepared from late blastula stage zebrafish embryos using the SuperScript III First-Strand Synthesis System (Invitrogen), then cloned into the pCS2+ expression vector. The forward primer had an additional 15 bases at the 5′ end so as to include a restriction enzyme site followed by a generic Kozak sequence (GCCACC) in the amplified product. To generate Eng-Mxtx2 and VP16-Mxtx2 constructs, the homeobox domain sequence of mxtx2 (corresponding to amino acids 13-81) was amplified (F, 5′-GGAAACTCACAGGCCAGTAAGATTG-3′; R, 5′-GTTGTCCAGCTGTGGGGAGCTCTG-3′) and subcloned into the XhoI-XbaI sites of pCS2+-Eng and pCS2+-VP16 vectors. To generate Flag-tagged mxtx2, full-length and homeobox domain-deleted mxtx2 were subcloned into pCS2+-Flag vectors.

All mRNAs were synthesized using the mMESSAGE mMACHINE SP6 Kit (Ambion). Fertilized embryos were injected with 10 pg mRNA for mxtx2 overexpression, or 2 pg for rescue, 25 pg for eng-mxtx2 and vp16-mxtx2 mRNA or 50 pg gfp mRNA as a control. The cycloheximide treatment assay shown in Fig. S2 in the supplementary material was performed as previously described (Poulain and Lepage, 2002). Briefly, eng-mxtx2, vp16-mxtx2 mRNA or gfp mRNA was injected as above, but a portion of the embryos were then treated with 50 μg/ml cycloheximide (Sigma) beginning at the 64- to 128-cell stage, and all embryos were fixed when untreated embryos reached the 30% epiboly stage and stained for ndr2 expression.

All MOs were purchased from Gene Tools (Philomath, OR, USA). Bold indicates nucleotides complementary to ATG start codons and lowercase indicates mismatches. Single or combined MO injections were at the following doses: 2 ng (1 ng for Fig. S1 in the supplementary material) Mxtx2 MO (ZFIN name MO1-mxtx2, 5′-CATTGAGTATTTTGCAGCTCTCTTG-3′) (Bruce et al., 2005) or 8 ng Mxtx2 mismatch MO (5′-CATTcAGTAaTTTcCAGgTCTgTTG-3′; the Ctrl MO for Fig. 1, Fig. 4E-P, Fig. 6 and Fig. 7), 8 ng standard Gene Tools control MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′; the Ctrl MO for Fig. 3, Fig. 4A-D and Fig. S1 in the supplementary material), 10 ng Ndr2 MO (5′-GCGACTCCGAGCGtGTGCATGATG-3′) (Karlen and Rebagliati, 2001), 10 ng Ndr1 MO (5′-ATCTGAGAGATTCTTACCTGCATGT-3′; used for Fig. 4N,P, Fig. 6C,F, Fig. 7E-H) (Gore et al., 2005), 10 ng Ndr1 MO (5′-ATGTCAAATCAAGGTAATAATCCAC-3′; used for Fig. 6B,E) (Feldman and Stemple, 2001).

Except for Fig. S1 in the supplementary material and Fig. 4F, in which a standard 1-cell injection was used, all MO injections were into the YSL as follows. Needles were targeted to the region of yolk just below the margin in 512-cell (2.75 hpf) to 1000-cell (3 hpf) embryos and, to assist with targeting, certain MOs had 3′ fluorescein tags: the Mxtx2 MO, the Mxtx2 mismatch MO, the Ndr2 MO and the Ndr1 MO in Fig. 6B,E. YSL injected embryos were viewed under a fluorescence dissection scope and only those embryos with fluorescence confined to the YSL and evenly distributed were used (this quality assurance was not employed for Fig. 4F because fluorescent Ndr2 MO was injected throughout the embryo, causing ubiquitous fluorescence). For the rescue shown in Fig. S2 in the supplementary material, 0.5 pg of ndr1 RNA was co-injected into the YSL of Mndr1 embryos along with 10 ng each of the Ndr1 MO (Gore et al., 2005) and the Ndr2 MO.

Digoxigenin-labeled RNA probes for ndr1 (Feldman et al., 1998), sox32 (Kikuchi et al., 2001), ndr2 (Rebagliati et al., 1998), gata5 (Reiter et al., 1999), ntla (ntl) (Schulte-Merker et al., 1992), bmp2b (Nikaido et al., 1997), foxa2 (axial) (Strahle et al., 1993), desma (Chen and Tsai, 2002) and egr2b (krox20) (Nieto et al., 1991) were prepared from linearized template DNAs using an RNA labeling kit (Roche). Whole-mount in situ hybridization was performed as described, except that post-hybridization washing was at 65°C (Toyama et al., 1995). For histology, stained whole-mount embryos were fixed in 4% paraformaldehyde (Sigma) overnight at room temperature, then gradually dehydrated through a methanol series and embedded in JB-4 plastic resin (Polysciences) according to the manufacturer's instructions. Sectioning (7 μm) was performed on a Leica RM2165 microtome (Deerfield, IL, USA). Whole-mount high-resolution double fluorescence in situ hybridizations for Fig. 5 were performed as described (Brend and Holley, 2009), using initial fluorescein labeling and a final fluorescein substrate to visualize the location of mxtx2 or sox32 transcripts in green or, alternatively, initial digoxigenin labeling and a final Cy3 substrate to visualize the location of ntla or sox32 transcripts in red.

Chromatin immunoprecipitation (ChIP) (Li et al., 2003) experiments were carried out using the protocol provided by NimbleGen with certain modifications, most of which have been described previously (Jang et al., 2009). Twenty-five pg of control (pCS2+-Flag) DNA, mxtx2-flag mRNA (full length) and delHD-mxtx2 mRNA (homeodomain deletion) were injected into 1- to 2-cell stage embryos and harvested at the 30% epiboly stage. Approximately 1000 injected embryos for each injection condition were fixed in 1.85% formaldehyde (Sigma) for 10 minutes at room temperature and the fixation was stopped with glycine. The embryos were then washed and lysed, and the chromatin DNA was harvested and sheared to an average size of 500 bp. A portion of this sheared DNA was kept (the pre-IP DNA) and the rest was subjected to ChIP. For ChIP, sheared DNA was incubated overnight with anti-FLAG M2 bound to Dynabeads protein G (Invitrogen). Negative control ChIPs were performed using standard mouse IgG instead of anti-FLAG M2. The IPs were washed extensively, chromatin DNA was eluted, reverse cross-linked with proteinase K, purified by organic extraction and precipitated with ethanol. Immunoprecipitated DNA was analyzed by real-time PCR using the primers listed in Table S1 in the supplementary material. For real-time PCR, the SYBR Green Master Mix (Roche) was used to monitor amplification of sample cDNA using an ABI Prism 7900HT Sequence Detection System. For each primer pair, cycle thresholds (Cts) were determined in amplifications of two duplicate IP DNA samples and a dilution series of pre-IP DNA. Percentage inputs for each primer pair were calculated by dividing the average IP sample DNA concentration by the pre-IP DNA concentration needed to achieve a Ct value equal to the average Ct value for the two IP samples. Standard deviations are a function of the variation between the Ct values of the two IP samples.

Luciferase constructs were generated in the pGL3-Promoter Vector. A 1.0 kb stretch of the first intron of ndr2, including the region enriched in our ChIP qPCR studies, was amplified with the following primers: F, 5′-CCCAACGAAATCTCATGAAATTGTG-3′; R, 5′-GTGTGATGAAAAGTCATGTGGTATCG-3′. As a negative control, 1.6 kb spanning the second intron of ndr2 was also amplified, using the following primers: F, 5′-CCAGATGTGCTCAACAATGACAAC-3′; R, 5′-GTGTCTCAGAATCATCTCTCCGTTC-3′. The amplified PCR fragments were inserted into KpnI-MluI sites of the pGL3-Promoter Vector.

Transient transfections were performed in 6-well dishes. COS-7 cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) with 10% fetal calf serum. Cells were transfected with the following DNAs: 0.5 μg firefly luciferase linked to the first or second intron of ndr2, as well as 0.5 μg of either mxtx2-Flag, delHD-Flag or Flag vector alone, and 1 ng Renilla luciferase, using the FuGENE 6 (Roche) transfection reagent. Luciferase activity was measured with the Dual-Luciferase Reporter Assay System (Promega) after 24 hours of transfection. All assays were performed at least three times.

To identify mesendoderm-inducing genes from the YSL, we injected zebrafish embryos with synthetic mRNA for YSL genes that we had previously identified through a microdissection-based screen (Hong et al., 2010) and assessed their ability to induce markers of various cell fates. We uncovered a novel and potent mesendoderm-inducing and ventralizing activity for mxtx2 (Fig. 1A-L), which encodes a Mix/Bix family transcription factor that is co-expressed in the YSL with the related family member mxtx1 (Hirata et al., 2000).

To examine the requirement of extraembryonic Mxtx2 for mesendoderm gene expression, we injected the YSL of wild-type (WT) embryos with an antisense morpholino oligonucleotide (MO) designed to disrupt translation of mxtx2 mRNA. Confirming previous reports (Bruce et al., 2005; Wilkins et al., 2008), Mxtx2-depleted embryos displayed a lethal defect in epiboly, which is a canonical gastrulation movement, that was rescued by co-expressing mxtx2 mRNA (see Fig. S1 in the supplementary material). Mxtx2-depleted embryos displayed normal mesendoderm gene expression at the onset of gastrulation (data not shown), but earlier, during the time of mesendoderm specification, they showed substantial reductions in mesendoderm gene expression that were not previously reported. Specifically, blastula stage expression of ndr1, ndr2, ntla, sox32 and gata5 was reduced, whereas expression of the ventral marker bmp2b was unaffected (Fig. 1M-X). Thus, extraembryonic Mxtx2 is essential for early mesendoderm specification, but not for ventral identity.

We noted particularly strong changes in ndr2 expression in Mxtx2 gain- and loss-of-function studies (Fig. 1F,R), leading us to hypothesize that Mxtx2 endogenously regulates ndr2 expression. We confirmed this in several ways. First, we verified that mxtx2 is co-expressed at the time of initial ndr2 expression. mxtx2 was initially expressed throughout the margin and vegetal embryo (Fig. 2A) and we observed scattered ndr2-expressing cells within the same domain (Fig. 2B). Later, mxtx2 expression became restricted to the YSL, and the YSL was also the first site of robust ndr2 transcription (Fig. 2C,D).

Second, to distinguish whether Mxtx2 upregulation of ndr2 is via repression or activation, we tested the effects of Engrailed (Eng) repressor (Conlon et al., 1996) and VP16 activator (Sadowski et al., 1988) forms of Mxtx2 on ndr2 expression (Fig. 2E-H). Overexpression of Eng-Mxtx2 quenched ndr2 expression (Fig. 2G), whereas overexpression of VP16-Mxtx2 increased ndr2 expression (Fig. 2H), demonstrating that ndr2 is upregulated via activation.

Third, to determine whether activation and repression of ndr2 expression by the Mxtx2 constructs is direct, we tested whether they could function when protein synthesis is blocked after 2 hours post-fertilization (hpf) with cycloheximide. This inhibition of protein synthesis during the time that the presynthesized constructs presumably act upon their targets did not block the respective induction or repression of ndr2 expression, indicating that Mxtx2 activation of ndr2 is independent of additional protein synthesis (see Fig. S2 in the supplementary material).

Fourth, to determine whether and where Mxtx2 binds to the ndr2 locus, we performed chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR). Embryos were injected with mRNA encoding Flag-tagged Mxtx2 or controls (Fig. 2I) and harvested at the 30% epiboly stage (Fig. 2J). Analysis of 18 kb of the ndr2 locus revealed enriched binding to the first intron of ndr2 (Fig. 2K). The first intron is a common location for transcriptional enhancers, as demonstrated for other Nodal-related genes including ndr1 (Fan et al., 2007; Osada et al., 2000). However, ChIP-qPCR across a similar span of the ndr1 locus, including its first intron, did not demonstrate Mxtx2 binding (Fig. 2L). Consistent with this, we observed no changes in ndr1 expression after injection of the Eng-Mxtx2 or VP16-Mxtx2 constructs (see Fig. S2 in the supplementary material).

Finally, to determine whether the first intron of ndr2 contains a genuine Mxtx2-response element, luciferase plasmids carrying either ndr2 intron 1, or ndr2 intron 2 as a control, were transfected into COS-7 cells along with an mxtx2 expression vector. An ndr2 intron 1-dependent enhancement of luciferase activity was observed (Fig. 2M,N). Thus, Mxtx2 activates ndr2 transcription via direct binding to its first intron. We presume that the modest alterations in ndr1 expression that we observed in response to Mxtx2 alterations (Fig. 1B,N) reflect an increased or decreased contribution of Ndr2 to the Ndr positive-feedback loop.

To better understand the recovery of downregulated genes in Mxtx2-depleted embryos, we focused on the spatiotemporal recovery of ndr2 and sox32 expression. Between 4.7 hpf (30% epiboly stage, Fig. 1) and 5.3 hpf (50% epiboly stage, Fig. 3), expression of these two genes had substantially recovered (compare Fig. 3F with Fig. 1P and Fig. 3J with Fig. 1R). Sections of sox32- and ndr2-stained embryos revealed that this recovery occurs in the embryonic margin, but not the YSL (Fig. 3A-D). Disruption of Nodal signaling causes a reciprocal effect on sox32 expression, eliminating it from the margin but not the YSL (Dickmeis et al., 2001; Kikuchi et al., 2001; Sakaguchi et al., 2001). We confirmed this by visualizing persistent sox32 staining in the YSL of MZoep embryos (Fig. 3G), which lack both maternal and zygotic sources of an essential Nodal co-receptor (Gritsman et al., 1999), and we observed the same YSL persistence for ndr2 (Fig. 3K). Consistent with the above, we were able to entirely eliminate ndr2 and sox32 expression by injecting Mxtx2 MOs into the YSL of MZoep embryos (Fig. 3H,L). Thus, extraembryonic Mxtx2 is essential for the extraembryonic expression of ndr2 and sox32.

We found that combining Mxtx2 and Nodal disruptions also affects the expression of ntla, the human ortholog of which, BRACHURY (T), is implicated in spina bifida and chordomas (MIM ID: 601397) (Fig. 4). Unlike the expression of most margin-specific genes, expression of ntla persists in the absence of Nodal signaling in all but the dorsal-most aspect (Fig. 4C, arrowhead) (Gritsman et al., 1999). Disruption of Mxtx2 in the YSL of WT embryos had no lasting effect on ntla expression (compare Fig. 4B with Fig. 1V), but co-disruption of Oep and extraembryonic Mxtx2 nearly eliminated expression (Fig. 4D). Thus, extraembryonic Mxtx2 and Nodal signaling are collectively required, but individually sufficient, to activate ventrolateral ntla expression in the margin, explaining the resilience of ntla expression. Persistence of ntla expression in MZoep embryos accounts for their residual posterior mesoderm formation (Harvey et al., 2010); therefore, our findings identify a combinatorial requirement for Mxtx2 and Nodal signaling that underlies the specification of posterior mesoderm.

To examine the possibility that Mxtx2 directly activates ntla transcription, we searched for a domain of mxtx2 and ntla co-expression. We were able to visualize the co-expression of sox32 with mxtx2 (Fig. 5C,G) or ntla (Fig. 5D,H) using double fluorescence in situ hybridization, but mxtx2 and ntla transcripts were respectively confined to the YSL and embryonic margin, with no overlap in expression (Fig. 5B,F). This, and the theoretical inability of a non-secreted transcription factor to cross the cellular membranes that separate the YSL cytoplasm from the cytoplasm of neighboring embryonic cells, argue that Mxtx2 influences ntla via activation of an intercellular signaling intermediary. Candidate intermediary factors include BMP and Wnt ligands, based on the demonstrated synergy of these pathways with the Nodal pathway in activating ntla expression (Harvey et al., 2010).

Both Ndr1 and Ndr2 are expressed during pre-gastrula stages and Oep mediates the signaling of both ligands. To determine whether the Mxtx2/Nodal synergy driving ntla expression is specific to Ndr1 or Ndr2, we tested several combinatorial disruptions. Co-depletion of Mxtx2 and Ndr2 from the YSL did not recapitulate the Mxtx2/Oep synergy (Fig. 4F), as might have been predicted because ndr2 expression in the YSL should also be eliminated in embryos injected with Mxtx2 MO (Fig. 4B). However, disruption of maternal and zygotic sources of Ndr1 synergized with Mxtx2 disruption to nearly eliminate ntla expression in the same manner as Oep disruptions (Fig. 4H). Thus, ntla expression, and therefore posterior mesoderm specification, depends on the combined actions of Mxtx2 and Ndr1.

To further dissect the genetic interaction of Mxtx2 and Ndr1, we looked at the effects of Mxtx2 removal on ntla expression in embryos lacking zygotic, extraembryonic and maternal sources of Ndr1. We uncovered two scenarios of combinatorial disruption with strong effects on ntla expression. First, ntla expression was disrupted by removal of Mxtx2 from the YSL of embryos that lack zygotic sources of Ndr1 (Fig. 4J). In the second scenario, ntla expression was disrupted by simultaneous removal of Mxtx2 from the YSL, Ndr1 from the YSL and maternal Ndr1 (Fig. 4P). This latter scenario was achieved by co-injecting MOs for Mxtx2 and Ndr1 into the YSL of Mndr1 embryos generated by crossing an ndr1^–/– female and a WT male. Thus, a combination of non-embryonic factors, namely maternal Ndr1, extraembryonic Ndr1 and extraembryonic Mxtx2, is synergistically required and individually sufficient for posterior mesoderm specification.

Although there is a clear function for embryonic expression of ntla in posterior mesoderm specification, the developmental function of extraembryonic Ndr2 downstream of Mxtx2 is less apparent. We found that extraembryonic Ndr2 synergizes with Ndr1 in an analogous fashion to Mxtx2 with Ndr1. Here we were able to examine the effects of our perturbations on the second day of embryonic development, during the so-called pharyngula stage when zebrafish embryos acquire the phylotypic vertebrate body plan (Ballard, 1981). This had not been possible for the Mxtx2-depletion studies owing to early lethality.

It was previously shown, by injecting the YSL with MOs that block Ndr1 and Ndr2 translation, that endoderm specification and a degree of dorsal mesoderm specification require an essential synergy of extraembryonic Ndr2 and extraembryonic Ndr1 (Fan et al., 2007). We revisited this experiment, obtaining very similar results (Fig. 6B,E), but found that additional mesendoderm deficits arose when we used an alternate Ndr1 MO (Gore et al., 2005) (Fig. 6C,F). Thus, the synergistic requirements for extraembryonic Ndr1 and Ndr2 in mesendoderm induction are stronger than previously believed.

Working with embryos carrying ndr1 mutations, we uncovered an even more dramatic requirement for extraembryonic Ndr2. Specifically, removal of extraembryonic Ndr2 from ndr1 mutant embryos (Fig. 6H,K) recapitulated the Ndr signaling-deficient phenotype of ndr1;ndr2 double-mutant embryos (Feldman et al., 1998), MZoep mutant embryos (Gritsman et al., 1999) and embryos in which lefty1 (antivin) is overexpressed (Thisse and Thisse, 1999). This phenotype is characterized by a complete lack of endoderm and anterior (head and trunk) mesoderm, with persistence of some mesoderm in the tail. This recapitulation was evidenced by the indistinguishable morphology of these embryos from MZoep embryos (compare Fig. 6H and 6I) and was further supported by the loss of heart and trunk somite labeling in both classes of embryo (compare Fig. 6K and 6L). Thus, in the absence of all sources of zygotic Ndr1, extraembryonic expression of ndr2 is required to support endoderm and anterior mesoderm formation.

Finally, similar to the second scenario that we delineated for Mxtx2-Ndr1 interaction, we found that losses of extraembryonic Ndr2 and extraembryonic Ndr1 synergized with a loss of maternal Ndr1 (Fig. 7). Embryos lacking maternal Ndr1 displayed normal morphologies and normal labeling of heart muscle, somite muscles and hindbrain rhombomeres 3 and 5 on the second day of development (Fig. 7B). At the gastrula stage, these embryos exhibited normal endoderm and dorsal mesoderm labeling (Fig. 7D). By contrast, MO-based perturbation of extraembryonic Ndr2 and extraembryonic Ndr1 in embryos lacking maternal Ndr1 recapitulated the mesoderm and endoderm deficiencies of Ndr signaling-deficient embryos. This was apparent from their morphology, from their lack of heart and trunk somite marker expression at the pharyngula stage (Fig. 7F) and from their lack of dorsal mesoderm and endoderm marker expression at the gastrula stage (Fig. 7H). In support of the specificity of this effect, we were able to substantially rescue this Ndr deficiency-like phenotype by co-injecting ndr1 RNA into the YSL (see Fig. S3 in the supplementary material) and milder morphological and molecular marker expression alterations were also seen in WT embryos in which extraembryonic Ndr2 and extraembryonic Ndr1 were depleted (Fig. 7E,G). Thus, the specification of endoderm and anterior mesoderm is determined by the combined actions of Nodal-related ligands from non-embryonic (extraembryonic and maternal) sources.

We propose the following model of mesendoderm specification (Fig. 7I). Induction of endoderm and anterior mesoderm is executed by a failsafe combination of three non-embryonic sources of Nodal-related ligands with overlapping activities. Nodal signaling is then amplified in the embryonic margin through autoregulatory positive feedback requiring embryonic Ndr1 and Ndr2 (Feldman et al., 2002; Osada et al., 2000), with a particular reliance on Ndr1 when the stimulus of non-embryonic Nodal factors is low. Recent reports indicate that positive Nodal feedback cannot occur in the YSL (Fan et al., 2007; Harvey and Smith, 2009), and expression of mxtx2 in the YSL is Nodal independent (Hirata et al., 2000), suggesting that the extraembryonic to embryonic signaling that we have uncovered is unidirectional and that Mxtx2 has a primary role upstream of Nodal signaling. This is distinct from other Mix/Bix factors, including the zebrafish family members Bon (Mixer) and Mezzo (Og9x – Zebrafish Information Network), the expression of which is downstream to, and dependent upon, Nodal signaling (Kikuchi et al., 2000; Poulain and Lepage, 2002). Mxtx2 also activates a Nodal-independent signal that is sufficient for the establishment of posterior mesoderm.

Our demonstration of a functional requirement for maternal Ndr1 is significant because the function of maternal Ndr1 is controversial (Bennett et al., 2007a; Gore et al., 2005). This is the first demonstration of an invariant requirement for maternal ndr1 using a mutant maternal allele. Our findings support the proposal that maternal Ndr1 functions in dorsal mesoderm specification (Gore et al., 2005), but we find that it is only essential in a combinatorial context and we did not recapitulate the requirement for maternal Ndr1 in anterior axis formation (Gore et al., 2005; Hagos et al., 2007).

The signaling mechanisms that we have discovered in zebrafish might prove to be relevant to higher vertebrates. Low levels of Nodal mRNA are detected throughout the preimplantation inner cell mass of mouse embryos (Takaoka et al., 2006) and no transcription factor driving this expression has been identified. In light of our finding that Mxtx2 regulates ndr2, it will be interesting to learn whether early Nodal expression in mouse is regulated by any of several co-expressed paired homeodomain proteins similar to Mxtx2.

Our evidence for a failsafe switch comprising diversely sourced Nodal-related ligands acting upstream of the Nodal autoregulatory loop might also be relevant to the early mouse embryo, in which other ligands with Nodal-like activities, particularly Gdf3 (Chen et al., 2006) and maternally derived activins (Jones et al., 2006), are detected alongside Nodal. Perhaps a combination of these TGFβ family ligands acts in an analogous fashion to ensure the initiation of Nodal autoregulation in mice.

In conclusion, this study elucidates essential molecular inputs underlying three central embryonic patterning events: the early activation of a Nodal-related gene (ndr2), which requires Mxtx2 binding to the first intron of ndr2; posterior mesoderm specification, which depends on the combined actions of extraembryonic Mxtx2, maternal Ndr1 and extraembryonic zygotic Ndr1; and the induction of endoderm and anterior mesoderm, which is implemented by an essential trio of non-embryonic sources of Nodal-related ligands.

We thank Alexander Schier, Igor Dawid, Karuna Sampath and Jérôme Collignon for helpful suggestions. This research was supported by the Intramural Research Programs of the National Human Genome Research Institute and the National Institute of Allergy and Infectious Diseases, National Institutes of Health. Deposited in PMC for release after 12 months.