Gdf3 is required for robust Nodal signaling during germ layer formation and left-right patterning

Essential revisions:

1) The authors find residual Nodal activity in MZgdf3 mutants and therefore conclude that gdf3 facilitates Nodal signaling. However the fact that Nodal signaling is at all possible without gdf3 could be shown more clearly. The authors could, for example, over-express nodal ligands in the MZgdf3 mutants and show to what extent signaling can be recovered compared to over-expression in the WT.

We have performed the experiment the reviewer suggested. By overexpressing ndr1 (squint) via mRNA injections at the one-cell stage, we determined that higher amounts of Nodal could activate the Nodal target gene goosecoid (gsc) in MZgdf3 mutants. Although 1 pg of ndr1 led to a modest increase in gsc expression in WT embryos it did not activate gsc expression in MZgdf3 embryos. However, 10pg of injected ndr1 mRNA did lead to activation of gsc expression in MZgdf3 mutant embryos. This new data has been included in Figure 3 and suggests that Ndr1 can signal to activate genes independently of Gdf3 when it is present at very high levels.

2) The ubiquitous expression of gdf3 at early stages of development suggests that gdf3 may act as a facilitator of Nodal signaling, while the patterning of mesendoderm tissues is achieved by localized expression of cyc and sqt. This point could be better shown by more precise rescue experiments. Is it possible to obtain fully patterned embryos by simple over-expression of gdf3 in MZgdf3 mutants? The experiment in Figure 2S suggests that gsc expression can be recovered but that gsc localization remains aberrant.

We have addressed this comment by performing various rescue experiments in M/MZgdf3 embryos. mRNA generated from the previously published gdf3 plasmid construct we received as a gift (Peterson et al.2013) did not rescue mutant embryos. This was likely due to the presence of an HA tag at the C-terminus of the protein. Therefore, we generated a new untagged gdf3 plasmid construct from cDNA and performed rescue experiments using mRNA from this. We were able to obtain fully patterned embryos by overexpressing this untagged Gdf3 in MZgdf3 embryos and included this new data in Figure 1. This mRNA was also able to rescue Nodal target gene expression, data now included in Figure 2 and in Figure 2—figure supplement 1. In each case, there was correct spatial expression of each target gene. This suggests that Gdf3 is a permissive factor and is not required in any particular spatial domain for proper Nodal target gene induction.

3) The experiment on KV morphology and LR asymmetry are interesting as they show that gdf3 acts as a facilitator of all known Nodal ligands in zebrafish. However the authors should confirm their findings with a rescue of the MO phenotype they show or, alternatively, by obtaining similar results by partial rescue of MZgdf3 mutant with gdf3 mRNA.

We have addressed this comment by performing gdf3 mRNA overexpression experiments in Mgdf3 and MZgdf3 embryos in which early mesendoderm defects are rescued. Defects in left-right patterning (a later event) were still present to varying degrees. Increasing amounts of injected gdf3 mRNA led to increasing rescue of L-R defects in MZgdf3 mutants, as assessed by the direction of cardiac jogging and the laterality of spaw and dand5 expression. Moreover, Mgdf3 mutants, in which zygotically-derived Gdf3 is unperturbed, could be rescued for L-R patterning to a greater extent than MZgdf3 mutants. These new results are presented in Figure 5, and confirm our initial conclusion that Gdf3 acts as a facilitator of Nodal signaling during L-R patterning events. Moreover, the difference in "rescuability" of MZgdf3 and Mgdf3 mutants suggests some role for zygotic Gdf3 in L-R patterning that is not revealed in Zgdf3 mutants likely owing to compensation by maternal Gdf3. This is a curious result, and we added some thoughts on this in our Discussion.

4) The authors state in the paper several times that perhaps Nodals and Vg1 dimerize. Could this be shown directly (e.g., dimerization of Gdf3 and southpaw)? If so, this would boost the overall impact of the manuscript. However, similar findings have been previously shown for, for example, BMPs.

We appreciate this suggestion and agree that testing for direct interaction between Gdf3 and zebrafish Nodal orthologs would be informative. Unfortunately, we are unable to supply these data in sufficient quality given the time constraints, and we hope the reviewers agree that this does not detract significantly from our manuscript. We also note that Nodal-Gdf3 heterodimerization has been demonstrated in other systems (Fuerer et al., 2014; Tanaka et al.2007), something we have emphasized in our manuscript.