The zebrafish Tuebingen strain was used as wildtype fish and for generating mutants. The eomesa ^tsu007 mutant line that carries a 353-bp deletion was generated by the CRISPR-Cas9 system with a gRNA (5′-ggc​gga​aag​tgg​gtg​acc​tgc​gg-3′) targeting the second exon of eomesa (Figure 1A). For genotyping the eomesa ^tsu007 mutant allele, the upper primer (5′- CTC​AGC​TCG​ATG​CCC​ATT​C-3′) and the lower primer (5′- ATA​CAG​TCT​TTG​TCG​GAG​ATG-3′) were used for PCR. Like the previously reported eomesa ^fh105 mutant line (Du et al., 2012), zygotic eomesa ^tsu007 (Zeomesa ^tsu007 ) homozygous embryos were able to grow up to adulthood with loss of the dorsal fin (Figure 1A). The hwa ^tsu01sm mutant line was used and its genotyping was described before (Yan et al., 2018). The eomesa ^tsu007/+ ;hwa ^tsu01sm/+ double heterozygotes were obtained by crossing eomesa ^tsu007/+ female to hwa ^tsu01sm/+ male. Then, Zeomesa ^{tsu007/tsu007} ;Zhwa ^{tsu01sm/tsu01sm} double homozygous fish (i.e., Zeomesa;Zhwa double homozygotes) were obtained by intercrossing the double heterozygotes (Figure 1B). Like Zeomesa mutant female that were unable to naturally ovulate (Du et al., 2012; Xu P. et al., 2014), Zeomesa;Zhwa double homozygous female were unable to naturally ovulate and in vitro fertilization using their squeezed eggs and wildtype male-derived sperms were performed to obtain maternal double mutant (Meomesa;Mhwa) embryos (Figure 1B).

Embryos were maintained in Holtfreter’s water at 28.5°C. Developmental stages of WT embryos were determined as described before (Kimmel et al., 1995) while those of mutants were indirectly judged by developmental time matching to WT embryos in the same conditions. Embryo treatment with SB431542 (SB) were performed as described before (Sun et al., 2006). Briefly, One-cell stage embryos of different mutant or WT lines were incubated in Holtfreter’s water with addition of freshly made SB to a final concentration of 50 μM in a dish and then harvested for observation or assays at desired stages. All experiments were approved by Tsinghua University Animal Care and Use Committee.

The plasmids pCS2-eomesa-Myc (Bruce et al., 2003) and pCS2-hwa-HA (Yan et al., 2018) were used to in vitro synthesize capped mRNAs using mMESSAGE mMACHINE Kit (Ambion). To knock down zebrafish β-catenin2, β-cat2-MO and its control morpholino (cMO) were used as described before (Zhang et al., 2012). mRNA or MO was injected into embryos at the one-cell stage.

Embryos or eggs (15 per sample) were harvested at desired stages, and used to extract total RNA by RNeasy Mini Kit (Qiagen) as previously described (Jia et al., 2009). cDNA was synthesized using the M-MLV reverse transcriptase (Promega) and qRT-PCR was performed with TransStart Top Green qPCR SuperMix (TransGen Biotech) as described (Sun et al., 2018). Expression levels were normalized to the reference gene eif4g2a unless otherwise stated. The Student’s t-test (two-tailed, unequal variance) was used to determine p-values. Primers and sequences for qRT-PCR analysis were as follows: ndr1-F (5′-TTG​GAT​ATG​CTC​CTT​GAC​CC-3′), ndr1-R (5′-ACA​GAT​AAG​GCA​AAC​ACG​CAA​A-3′); ndr2-F (5′-GAA​ATA​TCA​TCA​CCC​CAG​TCG​T-3′), ndr2-R (5′-CTC​CAC​CTG​CAT​GTC​CTC​GT-3′); tbxta-F (5′-TTG​GAA​CAA​CTT​GAG​GGT​GA-3′), tbxta-R (5′-CGG​TCA​CTT​TTC​AAA​GCG​TAT-3′); sox32-F (5′-TCT​GCC​ACG​GTC​TGC​TTA​C-3′), sox32-R (5′-CAG​AGA​AGG​TCC​ACC​CAA​AC-3′); gata2a-F (5′-CTC​CTC​AGC​GGA​TCC​GCT​TCC​AGC-3′), gata2a-R (5′-GGT​CGT​GGT​TGT​CTG​GCA​GTT​CGC-3′); gsc-F (5′-GAG​ACG​ACA​CCG​AAC​CAT​TT-3′), gsc-R (5′-CCT​CTG​ACG​ACG​ACC​TTT​TC-3′).

The graphs and t-test were finished with GraphPad Prism 7. Error bars represent mean ± SD. p values are two-sided. Significance levels were indicated by nonsignificant (ns); *, p < 0.05; **, p < 0.01; ***, p < 0.001.

The eomesa ^tsu007 mutant allele harbors a 353-bp deletion between the first and the second exon, resulting in a premature stop codon upstream of the T-box coding region (Figure 1A). Maternal eomesa mutants (Meomesa ^tsu007 ) showed delayed epibolic process; the majority of Meomesa ^tsu007 mutants died before 24 hpf and survivors at 24 hpf had a normal head with thin posterior trunk (posterior defect) or had thin anterior trunk with a bulged yolk extension (yolk defect) (Figures 1C-D’). These defects are similar to those observed in Meomesa ^fh105 mutants (Du et al., 2012).

Using eomesa ^tsu007 and hwa ^tsu01sm lines (Yan et al., 2018), we managed to obtain eomesa;hwa double homozygotes (Zeomesa;Zhwa) female fish, which were fertile and used to produce Meomesa;Mhwa embryos by in vitro fertilization using sperms squeezed from WT males. Generally, the Meomesa;Mhwa double mutants exhibited more severe phenotype than either of single mutants (Figures 1D–F). An average of 91% maternal double mutant embryos were arrested and deformed during gastrulation and the remaining embryos at 24 hpf had a degenerating tail-like structure with missing of other tissues such as head and anterior trunk (Figures 1F, F’), indicating cooperative roles of maternal eomesa and hwa in embryonic survival.

Then, we examined expression patterns of the endodermal marker sox32, the mesodermal marker tbxta (previously named ntla) and the epidermal marker gata2a in the single and double mutants at 4.7 hpf (30% epiboly stage) and 6 hpf (shield stage) by whole mount in situ hybridization (WISH) (Figure 2A). Compared to WT embryos, sox32 expression in either of single mutants was weaker with some missing domains in the blastodermal margin, whereas it was completely abolished in Meomesa;Mhwa double mutants. Meomesa mutants showed missing of tbxta expression in some portions of the margin, which is consistent with the pattern observed in Meomesa ^fh105 embryos (Xu P. et al., 2014), and Mhwa mutants appeared to express tbxta in the whole margin; in contrast, 73% of Meomesa;Mhwa double mutants lacked tbxta expression and the remaining proportion had a few small tbxta-expressing patches that might be caused by evoked genetic compensation through unknown mechanisms. The expression domain of gata2a was dorsally expanded in either of single mutants but further expanded throughout the blastoderm in the double mutants. qRT-PCR analysis using specific primers revealed a drastic decrease of sox32 and tbxta expression levels with a concomitant increase of gata2a levels in the double mutants (Figure 2B). These results indicate that maternal eomesa and hwa are two essential genes for mesendoderm induction and the whole blastoderm with their simultaneous loss may acquire the epidermis fate.

Given that Nodal signaling is critical for mesendoderm induction, we wondered how ndr1 and ndr2 expression were altered in Meomesa;Mhwa double mutants. WISH results showed that either ndr1 or ndr2 expression was undetectable in the double mutants at 4.3 hpf and 4.7 hpf while detected in either of the single mutants (Figure 2C). qRT-PCR analyses using embryo pools disclosed that, compared to those in WT embryos, ndr1 and ndr2 levels were decreased significantly in Meomesa mutant embryos and further dropped in Meomesa;Mhwa double mutants whereas their expression levels were not changed significantly in Mhwa embryos (Figure 2D). These results indicate that ndr1 and ndr2 expression is initiated in the absence of either maternal eomesa or hwa but fail to initiate in the absence of both maternal factors.

Next, we tested the capability of exogenous eomesa and hwa to induce ndr1 and ndr2 in the absence of both endogenous Eomesa and Hwa. We injected myc-eomesa, hwa or both mRNAs into Meomesa;Mhwa double mutant embryos at the one-cell stage. Morphological observation at 6 hpf indicated that eomesa but not hwa overexpression could largely rescue the Meomesa;Mhwa phenotype of slow epiboly, and co-overexpression of eomesa and hwa could restore the embryonic shield (Figure 3A). The injected embryos were then examined for ndr1 and ndr2 expression at 4.3 hpf and 4.7 hpf by WISH. Results disclosed that myc-eomesa overexpression induced ndr1 and ndr2 expression in the whole blastodermal margin, hwa overexpression activated their expression only in one side of the blastoderm (presumably dorsal side), and co-overexpression induced their expression at higher levels (Figure 3B). The induction of ndr1 and ndr2 by hwa mRNA was abolished when β-cat2 was knocked down with an antisense morpholino, which corroborates that hwa mainly exerts its effect through activation of β-catenin signaling (Yan et al., 2018). Besides, hwa showed a stronger induction activity for ndr1 than for ndr2 while eomesa had a stronger induction activity for ndr2 than for ndr1. These observations were confirmed by qRT-PCR data (Figure 3C). These results imply that either Eomesa or Hwa is sufficient to activate nodal genes expression, however, the former may be a more general activator while the latter may act as a regional activator.

We then investigated mesendoderm induction capacity of eomesa and hwa in maternal double mutants. We found that overexpression of eomesa or hwa alone or together in Meomesa;Mhwa embryos could induce expression of sox32, gsc and tbxta but reduce gata2 expression as examined by WISH and qRT-PCR (Figures 3D,E). Notably, overexpression effect of hwa was eliminated (on sox32, gsc and tbxta) or reduced (on tbxta and gata2a) when β-catenin2 was knocked down at the same time, which confirmed dependence of hwa function on β-catenin2 (Yan et al., 2018). Besides, compared to hwa, ectopic eomesa exhibited a stronger inductive effect on sox32 and tbxta but weaker effect on the dorsal mesodermal marker gsc, supporting the idea that eomesa plays a more general role in mesendoderm induction.

We extended our observation to morphological changes in WT, Mhwa, Meomesa, or Meomesa;Mhwa embryos at 24 hpf after overexpression of eomesa, hwa or together. Generally, hwa overexpression in WT embryos led to strong embryonic dorsalization with missing posterior structures as reported before (Yan et al., 2018), whereas eomesa overexpression caused relatively weak dorsalization with thinner posterior structures (Figure 4A). Notably, 90.6% (n = 96) of Mhwa mutants, which lack the head and anterior trunk structures (Yan et al., 2018), were able to form the head and the whole trunk following eomesa overexpression (Figure 4C), while hwa overexpression in Meomesa mutants still caused strong dorsalized phenotype (Figure 4B). As described above, most of Meomesa;Mhwa double mutants died before 24 hpf and the survivors all had a degenerating tail-like structure without a head; however, overexpression of eomesa or hwa alone or together appeared unable to evidently reduce the mortality (Figure 4D). Nevertheless, eomesa overexpression alone or co-overexpression with hwa allowed 8–11% of embryos to form the head and the trunk, whereas hwa overexpression alone allowed only 6.3% of embryos to form an abnormal head and anterior trunk with missing of posterior trunk structures (Figure 4D). Although the above overexpression effects should be investigated further by titrating dosages of ectopic mRNA species, our observations support an idea that the role of eomesa in development of ventrolateral mesendoderm-derived tissues may not be replaced by hwa.

Previous studies have demonstrated that the TGFβ signaling inhibitor SB431542 (SB) or SB505124 can efficiently block Nodal signaling, resulting in loss of mesendodermal tissues in zebrafish embryos (Sun et al., 2006; Hagos and Dougan, 2007). We set out to look into effect of Nodal signaling inhibition on mesendodermal induction in Meomesa, Mhwa or Meomesa;Mhwa mutant embryos. One-cell stage embryos of different mutant or WT lines were incubated in the presence of 50 μM SB until harvested for observation or assays. As shown before (Sun et al., 2006), SB treatment caused loss of the embryonic shield at the shield stage (6 hpf) and most mesendodermal tissues in WT embryos at 24 hpf (Figure 5A). SB-treated Mhwa or Meomesa mutant embryos at 24 hpf also showed more severe defects compared to the untreated control mutants (Figure 5A). The complete loss of the Nodal target genes lefty1 and lefty2 in SB-treated WT embryos confirmed the effectiveness of SB treatment (Figure 5B). Then, we examined expression of sox32 and tbxta at 4.7 hpf and 6 hpf by WISH and qRT-PCR analysis (Figures 5C,D). WISH results showed that sox32 and tbxta expression became very weak at 4.7 hpf and largely recovered at 6 hpf in SB-treated WT and Mhwa embryos; in contrast, their expression appeared undetectable at both stages in SB-treated Meomesa embryos (Figure 5C). qRT-PCR results also confirmed that SB treatment significantly inhibited sox32 and tbxta expression in WT and Mhwa embryos but caused loss of sox32 and tbxta expression in Meomesa embryos (Figure 5D). Taken together, these results suggest that Nodal signaling may contribute to mesendoderm induction at variable levels in different genetic backgrounds.

Based on the above data, we hypothesize that the mesendodermal fate in the zebrafish embryo is induced via Ndr1 and Ndr2 by three factors, i.e., maternal eomesa, maternal hwa-activated β-catenin signaling and Nodal autoregulation. We assume that zygotic ndr1 and ndr2 expression in Meomesa mutants depends on maternal hwa and Nodal autoregulation while their expression in Mhwa mutants relies on maternal eomesa and Nodal autoregulation. To assess contributions of individual factors, we examined ndr1 and ndr2 expression patterns by WISH as well as their total levels by qRT-PCR analysis in WT, Meomesa and Mhwa embryos from 3.7 hpf to 6 hpf without or with SB treatment (Figures 6, 7).

We first investigated ndr1 expression in detail. In SB-treated WT embryos, the expression pattern of ndr1 was unaltered as indicated by WISH (Figure 6A), but its expression level decreased from 3.7 hpf to 6 hpf as examined by qRT-PCR (Figure 6B), which suggest a role of Nodal autoregulation in maintaining ndr1 expression. Interestingly, the degree of ndr1 reduction due to SB treatment gradually increased from 3.7 hpf to 4.7 hpf, implying that Nodal autoregulation contributes to ndr1 expression more and more during that period.

In Meomesa mutants without SB treatment, ndr1 expression was activated in the dorsal margin with a smaller area than in WT embryos at 3.7 hpf, then propagated ventrally with a slower rate than in WT embryos, and occurred in the whole blastodermal margin at 5 hpf (Figure 6C, upper panel), which were consistent with previous observation (Xu P. et al., 2014). In SB-treated Meomesa mutants, in contrast, ndr1 was still activated in the dorsal margin at a reduced level at 3.7 hpf, its expression domain expanded ventrally but never occupied the whole margin (Figure 6C, lower panel). qRT-PCR results showed that SB-treatment caused a significant reduction (by 31%–62%) of the ndr1 expression level from 3.7 hpf to 6 hpf (Figure 6D). Thus, in the absence of maternal eomesa, maternal hwa alone can activate ndr1 expression in the dorsal margin; and the ventral expansion of ndr1 expression domain as well as increments of ndr1 expression heavily depend on positive feedback of Nodal signaling.

In Mhwa mutants without SB treatment, ndr1 expression was not prominent in the dorsal margin but similar to WT in the other marginal areas at 3.7 hpf, and its expression pattern later on was comparable to WT embryos (Figure 6E, upper panel; and also see Figure 2C). In SB-treated Mhwa mutants, ndr1 expression pattern was not obviously altered at all examined stages as detected by WISH (Figure 6E, lower panel); however, qRT-PCR results showed a significant reduction of ndr1 expression level at 4.3 hpf, 4.7 hpf and 5 hpf while changes at other stages were not significant (Figure 6F). Apparently, maternal Eomesa alone is capable of activating ndr1 expression in the whole blastoderm margin but its enhancement during late blastulation requires the contribution of Nodal autoregulation.

Taken together, these results suggest that maternal eomesa can activate ndr1 expression in the whole blastodermal margin, maternal hwa activates ndr1 only in the dorsal margin, and Nodal autoregulation contributes to enhancement of ndr1 expression.

We similarly investigated implication of maternal eomesa, maternal hwa and Nodal autoregulation in ndr2 expression. In SB-treated WT embryos, the ndr2 expression pattern was not obviously altered from 3.7 hpf to 4.7 hpf (Figure 7A). However, unlike in untreated WT embryos, ndr2 expression in SB-treated WT embryos at 5 hpf and 6 hpf was not prominently enriched in the dorsal margin, instead it appeared enhanced in the ventrolateral margin, for which we did not find an explanation at the moment. The total expression level of ndr2 in SB-treated embryos, as revealed by qRT-PCR analysis, was not significantly decreased at all examined stages (Figure 7B). It appears that Nodal autoregulation is less important for ndr2 expression than for ndr1 expression in WT embryos.

In Meomesa mutants without SB treatment (Figure 7C, upper panel), ndr2 expression occurred in the dorsal margin from 4 hpf to 5 hpf and extended to the whole margin at 6 hpf as reported before (Xu P. et al., 2014). In SB-treated Meomesa mutants, however, ndr2 expression was hardly detectable by WISH (Figure 7C, lower panel). qRT-PCR results showed that, compared to untreated Meomesa mutants, the ndr2 expression level in SB-treated Meomesa mutants was almost abolished (Figure 7D). This observation implies that, in the absence of maternal eomesa, maternal hwa might initiate ndr1 expression in the dorsal margin at low levels and existing Ndr1 thereof promotes ndr2 expression through Nodal signaling feedback.

As previously shown in Figure 2D, overall level of ndr2 expression in Mhwa was comparable to that in WT embryos. Surprisingly, we observed that the expression pattern and overall level of ndr2 in Mhwa embryos was unchanged by SB treatment (Figures 7E,F). Taken together, these data strongly suggest that maternal eomesa plays a major role in activation and maintenance of ndr2 expression in WT background.