Nodal patterning without Lefty inhibitory feedback is functional but fragile

Developmental signaling pathways often activate their own inhibitors. Such inhibitory feedback has been suggested to restrict the spatial and temporal extent of signaling or mitigate signaling fluctuations, but these models are difficult to rigorously test. Here, we determine whether the ability of the mesendoderm inducer Nodal to activate its inhibitor Lefty is required for development. We find that zebrafish lefty mutants exhibit excess Nodal signaling and increased specification of mesendoderm, resulting in embryonic lethality. Strikingly, development can be fully restored without feedback: Lethal patterning defects in lefty mutants can be rescued by ectopic expression of lefty far from its normal expression domain or by spatially and temporally uniform exposure to a Nodal inhibitor drug. While drug-treated mutants are less tolerant of mild perturbations to Nodal signaling levels than wild type embryos, they can develop into healthy adults. These results indicate that patterning without inhibitory feedback is functional but fragile.


Introduction
Feedback inhibition is a common feature of developmental pathways across phyla (Freeman, 2000;Freeman and Gurdon, 2002;Meinhardt, 2009;Piddini and Vincent, 2009;Ribes and Briscoe, 2009;Rogers and Schier, 2011). Feedback inhibitors contribute to patterning processes in tissues ranging from the mouse neural tube (Ribes and Briscoe, 2009) and the zebrafish hindbrain (White and Schilling, 2008), to the Drosophila wing (Gerlitz and Basler, 2002;Piddini and Vincent, 2009;Zeng et al., 2000) and eye (Freeman, 1997). Consistent with a general requirement for feedback control in development, inactivation of feedback inhibitors often results in disastrous patterning defects. However, directly testing the role of feedback per se has remained challenging: Eliminating inhibitors removes feedback, but it also increases signaling levels. Experiments that decouple inhibitor activation and inhibition while maintaining near-normal signaling levels are therefore required to unambiguously test the role of feedback in developmental patterning. Inhibitory feedback has been invoked to explain the sophisticated spatiotemporal control and robust performance observed in patterning circuits (Ribes and Briscoe, 2009). Inhibitory feedback has been suggested to turn off pathway activity when it is no longer needed, and to regulate spatial profiles of pathway activation (Barkai and Shilo, 2009;Ben-Zvi et al., 2008;Dessaud et al., 2007;Freeman, 2000;Gerlitz and Basler, 2002;Golembo et al., 1996;Lecuit and Cohen, 1998;Piddini and Vincent, 2009;Ribes and Briscoe, 2009;Schilling et al., 2012;Shiratori and Hamada, 2006;van Boxtel et al., 2015). Additionally, theoretical considerations suggest that negative feedback could enable the embryo to adjust signaling levels in response to unexpected perturbations or biochemical fluctuations (Barkai and Shilo, 2009;Eldar et al., 2003;Lander et al., 2009).
The Nodal/Lefty system has become a paradigm for feedback inhibition in development (Chen and Schier, 2002;Duboc et al., 2008;Freeman, 2000;Hamada, 2012;Kondo and Miura, 2010;Meinhardt, 2009;Meno et al., 1999;Nakamura et al., 2006;Rogers and Schier, 2011;Schier, 2009;Shen, 2007). Nodal is a TGFb superfamily ligand that induces the phosphorylation and subsequent nuclear localization of the transcription factor Smad2 in target cells. In early embryos, Nodal signaling induces mesendodermal fates through graded activation of target genes, with higher signaling intensities biasing cells toward endoderm. Nodal signaling intensity is shaped by the interplay between Nodal ligands and Leftys, secreted signaling inhibitors that prevent Nodal from binding to its receptors (Chen and Shen, 2004;Cheng et al., 2004). Studies of mouse lefty mutants and zebrafish lefty morphants reveal that loss of inhibition results in expanded domains of Nodal target expression, increased mesendodermal specification, and embryonic lethality (Agathon et al., 2001;Chen and Schier, 2002;Feldman et al., 2002;Meno et al., 1999;van Boxtel et al., 2015). Antagonism by Lefty is thus important for preventing overactive Nodal signaling during mesendodermal patterning. eLife digest During animal development, a single fertilized cell gives rise to different tissues and organs. This 'patterning' process depends on signaling molecules that instruct cells in different positions in the embryo to acquire different identities. To avoid mistakes during patterning, each cell must receive the correct amount of signal at the appropriate time.
In a process called 'inhibitory feedback', a signaling molecule instructs cells to produce molecules that block its own signaling. Although inhibitory feedback is widely used during patterning in organisms ranging from sea urchins to mammals, its exact purpose is often not clear. In part this is because feedback is challenging to experimentally manipulate. Removing the inhibitor disrupts feedback, but also increases signaling. Since the effects of broken feedback and increased signaling are intertwined, any resulting developmental defects do not provide information about what feedback specifically does. In order to examine the role of feedback, it is therefore necessary to disconnect the production of the inhibitor from the signaling process.
In developing embryos, a well-known signaling molecule called Nodal instructs cells to become specific types -for example, a heart or gut cell. Nodal also promotes the production of its inhibitor, Lefty. To understand how this feedback system works, Rogers, Lord et al. first removed Lefty from zebrafish embryos. These embryos had excessive levels of Nodal signaling, did not develop correctly, and could not survive. Bathing the embryos in a drug that inhibits Nodal reduced excess signaling and allowed them to develop successfully. In these drug-treated embryos, inhibitor production is disconnected from the signaling process, allowing the role of feedback to be examined. Drug-treated embryos were less able to tolerate fluctuations in Nodal signaling than normal zebrafish embryos, which could compensate for such disturbances by adjusting Lefty levels.
Overall, it appears that inhibitory feedback in this patterning system is important to compensate for alterations in Nodal signaling, but is not essential for development. Understanding the role of inhibitory feedback will be useful for efforts to grow tissues and organs in the laboratory for clinical use. The results presented by Rogers, Lord et al. also suggest the possibility that drug treatments could be developed to help correct birth defects in the womb.
Several roles for Lefty feedback have been suggested. First, Nodal and Lefty were proposed to form a reaction-diffusion patterning system that regulates both mesendoderm formation and left/ right patterning in zebrafish (Chen and Schier, 2002;Kondo and Miura, 2010;Meinhardt, 2009;Müller et al., 2012;Schier, 2009;Shen, 2007;Shiratori and Hamada, 2006). Notably, Nodal and Lefty fulfill the key biophysical requirements of a classical reaction-diffusion system: Both Nodal ligands (Cyclops and Squint) act as short-range (Cyclops) and mid-range (Squint) activators that induce their own expression as well as that of Lefty1 and Lefty2, which act as long-range, highly mobile inhibitors (Chen and Schier, 2002;Feldman et al., 2002;Meno et al., 1999;Müller et al., 2012). Second, Lefty was argued to temporally restrict Nodal signaling by creating a 'window' of signaling competence (van Boxtel et al., 2015). In this model, Nodal signaling proceeds until sufficient Lefty accumulates to shut down further signaling. Third, theoretical studies suggest that inhibitory feedback has the potential to mitigate fluctuations in signaling (Lander et al., 2009). Deleterious increases or decreases in Nodal signaling could therefore be offset by adjustments in Lefty-mediated inhibition, ensuring robust development in the face of variation in the external environment or expression of pathway components.
To understand the role of inhibitory feedback in the Nodal/Lefty patterning system, we created embryos in which Nodal inhibition was decoupled from Nodal signaling. We found that inhibitory feedback mitigates signaling perturbations but is dispensable for development.

Complete lefty loss causes lethal expansion of Nodal signaling and mesendoderm
To determine the consequences of removing feedback inhibition, we first used TALENs to generate null lefty1 and lefty2 alleles in zebrafish ( Figure 1C-T) (Bedell et al., 2012;Reyon et al., 2012;Sander et al., 2011;Sanjana et al., 2012). lefty1 a145 contains a 13-base-pair deletion that removes the translational start site and part of the predicted signal sequence ( Figure 1C), and lefty2 a146 contains an 11-base-pair deletion that results in a stop codon after amino acid 18 ( Figure 1D). In contrast to wild type lefty mRNA, mutant lefty mRNAs were unable to induce Nodal loss-of-function phenotypes when injected into zebrafish embryos ( Figure 1E-J"). lefty1 -/and lefty2 -/single mutants were viable and exhibited no or only minor increases in mesendodermal gene expression ( Figure 1N (Bakkers et al., 2009) and reflecting their distinct spatial expression patterns during left-right patterning (Bisgrove et al., 1999).
Mutations in nodal genes partially rescue lefty1 -/-;lefty2 -/mutants The patterning defects in lefty double mutants show a requirement for reduction of Nodal activity but do not establish a requirement specifically for inhibitory feedback. It is possible that a reduction in Nodal signaling by means other than inhibitory feedback could support patterning. For example, reducing nodal gene dosage could suppress lefty double mutant defects (Chen and Schier, 2002;Feldman et al., 2002;Meno et al., 1999). In support of this hypothesis, mutations in the Nodal genes squint or cyclops suppressed multiple aspects of the lefty1 -/-;lefty2 -/mutant phenotype (Figure 4). cyclops -/-;lefty1 -/-;lefty2 -/-( Figure 4C'') and squint -/-;lefty1 -/-;lefty2 -/-( Figure 4F'') mutants formed eyes and full-length tails, structures missing in lefty1 -/-;lefty2 -/mutants ( Figure 4A'', D'', M, N). Moreover, upregulation of mesendodermal gene expression was suppressed in squint -/-;lefty1 -/-; lefty2 -/mutants compared to lefty1 -/-;lefty2 -/mutants ( Figure 4G-L'''). Although these triple mutants are not viable, two functional nodal alleles are sufficient to generate mesendodermal gene expression patterns that are similar to those observed in wild type embryos with four nodal alleles and four lefty alleles. In previous studies, removal of squint, but not cyclops, partially suppressed defects in lefty double morphants (Chen and Schier, 2002;Feldman et al., 2002), but our results indicate that Lefty inhibits both Squint and Cyclops. The failure to fully rescue development may reflect an inability to precisely modulate Nodal dosage with this genetic approach, but reduction of Nodal dosage can partially rescue development in the complete absence of Lefty-mediated inhibition.
To determine the mechanism by which Nodal inhibitor-treated mutants are rescued, we analyzed Smad2 phosphorylation and mesendodermal gene expression. Interestingly, despite their eventual rescue, lefty1 -/-;lefty2 -/mutants that were exposed to inhibitor before the onset of Nodal signaling  . The mechanism by which premature inhibitor exposure rescues development therefore involves a delay in full activation of Nodal signaling. Highlighting the sensitivity to Nodal inhibitor drug concentration, lefty double mutants exposed to excess or sub-rescuing doses exhibited diminished or expanded pSmad2 activity gradients, respectively, with corresponding Nodal loss-or gain-of-function phenotypes ( Figure 6-figure supplements 1 and 4). Mutants exposed to inhibitor drug after the onset of Nodal signaling exhibited excess mesendodermal gene expression until~5 hr post-exposure ( Figure 6P-U'', Figure 6-figure supplement 3E,F), but they ultimately developed normally ( Figure 6F,I, Figure 6- figure supplements 1 and 2). Together, these results demonstrate that the precise spatial and temporal coupling of lefty expression and Nodal activity is not essential for normal development.

Inhibitory feedback can mitigate signaling perturbations
The rescue of lefty mutants with a Nodal inhibitor drug shows that inhibitory feedback is not a requirement for mesendoderm patterning. This result leaves open the possibility that inhibitory feedback enhances developmental robustness by mitigating signaling fluctuations (Lander et al., 2009). Perturbations that decrease Nodal signaling might be corrected by a compensatory decrease in activation of Lefty, thus restoring the activator/inhibitor balance. To test this model, we challenged wild type embryos with exogenous Nodal pathway activation or inhibition. Injecting low levels of lefty1 mRNA into wild type embryos dramatically reduced lefty2 transcript abundance ( Figure 7A,C,B,D), but left expression of the mesoderm marker noto relatively intact ( Figure 7A', C', B', D'). Conversely, increasing signaling by injecting mRNA encoding constitutively-active smad2 (CA-smad2, (Baker and Harland, 1996;Dick et al., 2000;Gritsman et al., 1999;Müller et al., 1999)) resulted in a marked increase in lefty2 expression ( Figure 7E,G,F,H), but unchanged noto expression ( Figure 7E', G', F' and H'). Wild type embryos thus appear to compensate for perturbed Nodal signaling by sensitively adjusting lefty levels. This model makes a key prediction: If Lefty feedback allows the embryo to correct signaling perturbations, embryos with compromised feedback should be more sensitive to Nodal signaling challenges. To test this hypothesis, we assessed developmental outcomes after manipulating Nodal signaling levels in wild type embryos and in drug-treated lefty1 -/-;lefty2 -/mutants, in which inhibition is decoupled from signaling ( Figure 6A). Wild type embryos challenged with a small dose of lefty1 mRNA exhibited normal phenotypes at 24 hpf ( Figure 8A,B) as well as normal mesendodermal gene expression at 50% epiboly and shield stages ( Figure 8H-I'). In contrast, challenging feedback-compromised embryos with the same dose of lefty1 mRNA led to markedly decreased noto expression ( Figure 8I,K,M,I', K', M'), and 24 hpf phenotypes resembling partial loss-of-function Nodal mutants ( Figure 8D,E) (Feldman et al., 1998;Gritsman et al., 1999). Challenging with a modest increase in Nodal signaling revealed a similar difference in sensitivity. Wild type embryos tolerated a small dose of CA-smad2 mRNA ( Figure 8C), while drug-rescued lefty double mutants developed severe phenotypic defects in response to the same treatment ( Figure 8F), although changes in early mesendodermal gene expression were relatively minor ( Figure 8N-S'). These results reveal that Lefty-mediated inhibitory feedback can mitigate aberrant fluctuations in Nodal signaling levels ( Figure 8G).

Discussion
The results in this study show that inhibitory feedback in the Nodal/Lefty system stabilizes Nodal signaling but is not essential for mesendoderm patterning and viability. The rescue of lefty mutants by ectopic lefty expression ( (Müller et al., 2012), but is surprising in light of the functions assigned to inhibitory feedback. Specifically, inhibitory feedback has been implicated in (1) shutting down pathway activity at the appropriate time to generate a pulse or window of signaling, (2) shaping spatial signaling profiles, (3) acting as part of self-organizing reaction-diffusion systems, and (4) mitigating fluctuations in signaling activity. Below, we discuss our results in the context of these models.
First, inhibitory feedback has been suggested to turn off signaling activity when it is no longer needed (Dessaud et al., 2007;Freeman, 2000;Golembo et al., 1996;Ribes and Briscoe, 2009;Shiratori and Hamada, 2006). For Nodal-mediated patterning, it has been proposed that progressively increasing Lefty levels shut down Nodal signaling at the onset of gastrulation (van Boxtel et al., 2015). However, we find that Nodal signaling is already increased in lefty double mutants by sphere stage, suggesting an earlier role for Lefty ( Figure 2). Moreover, Lefty can be replaced by an inhibitor drug added as early as the 8 cell stage (Figure 6, Figure 6-figure supplements 1-4, Figure 2-figure supplement 1), demonstrating that mesendoderm patterning can proceed without progressively increasing inhibition and without temporally precise feedback inhibition. Our results do not rule out that Lefty accumulation shuts down Nodal signaling during normal development, but they do indicate that this is not an absolute requirement for patterning. Second, inhibitory feedback has been implicated in shaping the spatial profile of pathway activity (Barkai and Shilo, 2009;Ben-Zvi et al., 2008;Freeman, 2000;Gerlitz and Basler, 2002;Lecuit and Cohen, 1998;Piddini and Vincent, 2009;Ribes and Briscoe, 2009;Schilling et al., 2012;van Boxtel et al., 2015). In drug-rescued lefty1 -/-;lefty2 -/mutants, Nodal activity gradients and mesendodermal gene expression patterns initially differed from wild type ( Figure 6J Figure 8. Correction of Nodal signaling challenges requires Lefty feedback. (A-F) Wild type embryos with intact feedback circuits (A-C) and feedbackdecoupled drug-treated lft1 -/-;lft2 -/mutants (D-F) were challenged by decreasing Nodal signaling with injection of 0.5 pg lft1 mRNA (B,E) or by increasing Nodal signaling with injection of 7.5 pg CA-smad2 (C,F). lft1 -/-;lft2 -/mutant embryos were exposed to 2 mM SB-505124 (Nodal inhibitor drug) at the 8 cell stage and phenotypes were assessed at 1 dpf. 0.5 pg lft1 causes mild Nodal loss-of-function phenotypes in wild type embryos (B), but strong Nodal loss-of-function phenotypes such as cyclopia and loss of head and trunk mesendoderm in feedback-decoupled embryos (E). CA-smad2 is well-tolerated in wild type embryos (C), but causes reduction of eyes, axis curvature, and accumulation of blood precursors in feedback-decoupled embryos (F). (G) Percentages of embryos with indicated phenotypes. Total embryos analyzed, lft challenge experiment: wild type + 0.5 pg gfp mRNA = 10, wild type +0.5 pg lft1 mRNA = 16, drug-treated lft1 -/-;lft2 -/mutants + 0.5 pg gfp mRNA = 16, drug-treated lft1 -/-;lft2 -/mutants + 0.5 pg lft1 mRNA = 12. Total embryos analyzed, CA-smad2 challenge experiment: wild type +7.5 pg gfp mRNA = 18, wild type +7.5 pg CA-smad2 mRNA = 20, drug-treated lft1 -/-;lft2 -/mutants + 7.5 pg gfp mRNA = 19, drug-treated lft1 -/-;lft2 -/mutants + 7.5 pg CA-smad2 mRNA = 20. (H-M') Wild type (H-I') and lft1 -/-;lft2 -/mutant embryos (J-M') were injected with 0.5 pg gfp or lft1 mRNA at the one-cell stage. Mutants were exposed to 2 mM Nodal inhibitor drug SB-505124 starting at the 8 cell stage, and expression of the endoderm marker sox32 (H, H', J, J', L, L') and the mesoderm marker noto (I, I', K, K', M, M') were assessed at 50% epiboly and shield stage as indicated via in situ hybridization. Note the decreased expression of noto in drug-treated double mutants injected with lft1 mRNA (K, K') compared to wild type embryos injected with lft1 mRNA (I,I'). (N-S') Wild type (N-O') and lft1 -/-;lft2 -/mutant embryos (P-S') were injected with 7.5 pg gfp or CA-smad2 mRNA at the one-cell stage. Mutants were exposed to 2 mM Nodal inhibitor drug SB-505124 starting at the 8 cell stage, and expression of sox32 (N,N', P, P', R, R') and noto (O, O', Q, Q',S, S') were assessed at 50% epiboly and shield stage as indicated via in situ hybridization. DOI: https://doi.org/10.7554/eLife.28785.021 The following figure supplement is available for figure 8:   Figure 6-figure supplements 3 and 4). However, rescued mutants developed into fertile adults ( Figure 6H, Figure 6-figure supplement 2), demonstrating that precise wild type activity gradients and mesendodermal gene expression patterns during early embryogenesis are not essential for germ layer development. Inhibitory feedback may subtly shape gene expression patterns, but the patterns achieved without feedback are a suitable starting point for successful development.
Third, the dispensability of inhibitory feedback in the Nodal/Lefty system raises questions about the role of Nodal and Lefty as an activator/inhibitor pair in self-organizing reaction-diffusion models. Although Nodal and Lefty fulfill the key regulatory and biophysical requirements of a short-to-midrange autocatalytic activator and a long-range feedback inhibitor (Hamada, 2012;Meno et al., 1999;Müller et al., 2012;Schier, 2009), our finding that development can be normal without inhibitory feedback indicates that this system does not require the ability to form self-organizing reaction-diffusion patterns. Instead, it is conceivable that the pre-patterning of the early embryo by maternal factors and the local activation of Nodal eliminate the need for self-organizing pattern generation by the Nodal/Lefty circuit. In this scenario, Nodal and Lefty may have constituted a reactiondiffusion activator/inhibitor pair in ancestral organisms but, through the addition of other regulatory layers, mesendoderm patterning lost the requirement for inhibitory feedback.
Finally, feedback inhibition has been implicated in buffering fluctuations in pathway activity (Barkai and Shilo, 2009;Eldar et al., 2003;Lander et al., 2009). Feedback may be required to optimize inhibitor levels, as suggested by the narrow range (~2 fold) of inhibitor concentrations that rescue Lefty loss ( Figure 6I, Figure 6-figure supplements 1 and 2). Indeed, the adjustment of lefty expression in response to slight alterations in Nodal signaling ( Figure 7) and the failure of feedbackdecoupled embryos to cope with perturbations in Nodal signaling ( Figure 8) support this idea. The ability to dynamically adjust pathway activity may allow the embryo to create reliable patterns in the face of endogenous signaling fluctuations and uncertain environmental conditions. We note, however, that Lefty feedback does not protect the embryo against all perturbations: Drug-rescued lefty mutants actually fared better than wild type embryos when challenged with injection of squint mRNA (Figure 8-figure supplement 1). Dissecting why Lefty feedback corrects some perturbations but not others will provide a window into the mechanisms and limits of robust patterning.
Our results have implications not only for the roles of feedback inhibition during development, but also demonstrate the feasibility of preventing patterning defects with small molecule drug exposure. Although suggested applications to human embryos might currently seem fanciful and would be challenging and fraught with ethical concerns, embryos bearing compromised patterning circuits could be identified by sequencing a single embryonic cell, and birth defects could be prevented by exposure to the appropriate small molecule. More generally, our study adds a new facet to recent revisions of classical patterning models (Alexandre et al., 2014;Chen et al., 2012;Dominici et al., 2017;Dubrulle et al., 2015;Varadarajan et al., 2017). For example, a tethered form of Wingless can replace endogenous Wingless, challenging models in which a gradient of diffusing Wingless is indispensable for tissue patterning (Alexandre et al., 2014). In the same vein, our observations challenge models in which inhibitory feedback is an absolute requirement for patterning and viability, but support the idea that inhibitory feedback enhances robustness by stabilizing signaling during development.
lefty TALENs target sites: lefty1 TALEN L: tcctgcaccttgaaaaga lefty1 TALEN R: tgcgcaaaggaggcacgc lefty2 TALEN L: ttcatccagctgttcatttt lefty2 TALEN R: tgctggaatccctgtgtgag Embryos from an incross of the TLAB wild type strain were injected at the one-cell stage with 300-450 pg mRNA encoding each TALEN pair. Injected fish were grown using standard fish husbandry protocols and fin clipped as adults. Genomic DNA was generated from fin material using the Hot Shot method (Meeker et al., 2007). To identify animals carrying mutations, PCR using primers flanking the target sites was carried out and the resulting amplicons were re-annealed and digested with mismatch-cleaving T7 endonuclease I (NEB) (Mussolino et al., 2011). PCR products from positive animals were cloned using a TOPO TA kit (Life Technologies) and sequenced. Positive animals were outcrossed and progeny were sequenced and tested for germline transmission.

Genotyping
Genomic DNA was generated using the Hot Shot method (Meeker et al., 2007). lefty1: The 13-base-pair deletion in lefty1 a145 destroys a PshAI restriction site. To genotype the lefty1 locus, PCR amplicons were generated using primers flanking the deletion and subsequently digested with PshAI endonuclease (NEB). Genotyping primers were identical to the lefty1 forward/ reverse primers described above. Complete digestion by PshAI indicates that both alleles are wild type, partial digestion indicates heterozygosity, and failure to digest indicates homozygosity for the lefty1 a145 mutation.
lefty2: The 11-base-pair deletion in lefty2 a146 was detected using a mutant-specific forward primer that spans the deletion. A forward primer specific to the wild type allele was also designed, as well as a reverse primer that is fully complementary to both alleles. To genotype the lefty2 locus, PCR was carried out using either the wild type-or mutant-specific forward primer and the common reverse primer. A band with the wild type-but not mutant-specific primer indicates that both alleles are wild type, bands with both primer sets indicate heterozygosity, and a band with the mutant-but not wild type-specific primers indicates homozygosity for the lefty2 a146 mutation. Optimal PCR conditions: Taq polymerase, 25 cycles, 57˚C annealing temperature.
lefty2 wild type genotyping forward primer: cattttgaccacagcgat lefty2 mutant genotyping forward primer: gttcattttgaccactcac The common reverse primer was identical to lefty2 reverse primer described above. squint: The squint cz35 allele has a~1.9 kb insertion in exon 1, and was detected as in (Feldman et al., 1998).
cyclops: The cyclops m294 mutation destroys an AgeI restriction site, and was detected as in (Sampath et al., 1998). lefty and squint expression constructs To generate mRNA from all constructs, plasmids were linearized with NotI-HF endonuclease (NEB) and purified using a Qiagen PCR clean-up kit. Capped mRNA was generated from linearized plasmid using an SP6 mMessage mMachine kit (Ambion) and purified with a Qiagen RNeasy kit.
After purification, mRNA was quantified using a NanoDrop spectrophotometer (ThermoFisher) and diluted to the appropriate concentration. For microinjections, a micrometer was used to adjust the drop volume to 0.5 nl. Depending on the concentration of the injection mix, a total volume of 1-2 nl was injected per embryo.
Lefty1-GFP, Lefty2-GFP, and untagged Lefty1 and Lefty2 constructs used in Figures 5 and 7 were identical to those in (Müller et al., 2012). These constructs lack endogenous UTRs and contain the consensus Kozak sequence gccacc immediately preceding the start codon.
Allele activity experiments: To determine whether mutant lefty alleles retain Nodal inhibitory activity, wild type and mutant lefty mRNA was injected into wild type embryos and Nodal loss-offunction phenotypes were assessed (Figure 1).
The 13 bp lefty1 a145 mutation removes part of the endogenous Kozak sequence (gaaaag). Therefore, lefty1 constructs containing this endogenous Kozak were generated, rather than the consensus Kozak sequence gccacc as in (Müller et al., 2012). Primers with either the endogenous Kozak sequence (for the wild type construct) or the truncated endogenous sequence and deleted region of coding sequence (for the mutant construct) were designed, and the Lefty1 construct from (Müller et al., 2012) was used as a PCR template. The resulting fragments were cloned into BamHI and XhoI sites in pCS2(+). Both constructs lack endogenous UTRs.
The lefty2 wild type construct was the same used in (Müller et al., 2012) and in the transplantation experiments in Figure 5. The lefty2 a146 construct was made by generating cDNA from lefty2 homozygous embryos, amplifying the mutant lefty2 coding sequence, and cloning the resulting fragment into ClaI and XhoI in the pCS2(+) vector. In addition to the 11 bp deletion, the mutant construct contains three silent SNPs at position 184 (T->C), 932 (A->C), and 943 (T->A). Both constructs lack endogenous UTRs and contain the consensus Kozak sequence gccacc immediately preceding the start codon.
Nodal overexpression experiment: The construct used to generate squint mRNA in the Nodal overexpression experiment (Figure 8-figure supplement 1) was identical to that used in (Müller et al., 2012). This construct lacks endogenous UTRs and contains the consensus Kozak sequence gccacc immediately preceding the start codon.

a-pSmad2 imaging
Embryos were mounted in agarose and cleared with 2:1 benzyl benzoate:benzyl alcohol (BBBA) (Yokomizo et al., 2012). Briefly, a dehydrated embryo was dropped into molten low-melting point agarose (1% (w/v) in H 2 O), transferred onto a coverglass and oriented manually. All embryos (other than sphere stage) were mounted 'margin down' (i.e. with the animal/vegetal axis parallel to the coverglass, see below for rationale). Shield stage embryos were rolled to ensure that the dorsal/ventral axis was parallel to the coverglass. Sphere stage embryos were mounted with the animal/vegetal axis perpendicular to the coverglass (animal pole facing up). The agarose drop was then dehydrated with three washes of 100% MeOH, two washes of 50:50 (v/v) BBBA:MeOH, and 3 washes of BBBA. Cleared embryos were then sealed to a microscope slide using fast wells reagent reservoirs (Grace Bio-Labs). Imaging was performed on Sytox Green and Alexa 647 channels using an LSM 700 confocal microscope (20x air objective, 0.5 NA). Image stacks extended from the embryo margin (adjacent to the coverglass) to beyond the center of the embryo. Z-planes were spaced at 2 mm intervals.

a-pSmad2 image segmentation and quantification
Quantification of pSmad2 and Sytox green staining intensity in laterally-mounted embryos was performed using the ten z-slices surrounding the center of the embryo axis (i.e. five slices above and five slices below the embryo center). This region was chosen to minimize artifacts due to light scattering, which causes a decrease in apparent fluorescence intensity in deeper tissue planes. Our procedure-looking only at slices close to the 'central plane' of the embryo-allows the entire gradient to be sampled within each slice, and ensured that all data were taken from planes within a narrow range of imaging depths, effectively controlling for signal drop-off with imaging depth.
Nuclei were segmented from the Sytox Green channel images using a custom pipeline implemented in MATLAB (Source code 1). Briefly, out-of-plane background signal was approximated by blurring adjacent z-slices (i.e. the slice above and below the plane being segmented) with a Gaussian smoothing kernel and summing. This background was subtracted from the segmentation image, and preliminary boundaries for nuclei were identified by adaptive thresholding (http://homepages.inf.ed. ac.uk/rbf/HIPR2/adpthrsh.htm) of the resulting image. Spurious objects were discarded by morphological filtering (based on object size). Final segmentation boundaries were defined after manual checking and correction with a custom MATLAB script (Source code 1). The fluorescence intensity of each segmented nucleus was defined as the mean intensity of its constituent pixels.
The distance of each nucleus from the margin was defined along a curved embryo contour (Figure 2-figure supplement 1). This contour was defined by 1) projecting all segmented nuclei centroids onto a single z-plane, 2) creating a full embryo 'mask' by filling a convex hull containing all of these points, 3) identifying the left and right margin boundaries as the points of maximum curvature on the convex hull, 4) taking the distance transform of the embryo mask, and 5) stepping along the 'valley' of the distance transform that connects the left and right margins (as defined above). This rough contour was then smoothed using a Savitsky-Golay filter to yield the final contour. The position of each nucleus was then projected onto the contour, and the distance from the margin (as plotted in Figure 2C and Figure 2-figure supplement 1) was determined as the distance to the closest margin along this curve.

In situ hybridization
Embryos were fixed overnight at 4˚C with 4% formaldehyde in PBS. In situ hybridization was carried out as in (Thisse and Thisse, 2008) and representative embryos were imaged in 2:1 benzyl benzoate:benzyl alcohol (BBBA) with a Zeiss Axio Imager.Z1 microscope. When genotyping was necessary, genomic DNA was generated after imaging as in (Meeker et al., 2007) and used in the described genotyping assays.

Nodal inhibitor drug
The Nodal inhibitor SB-505124 (S4696, Sigma-Aldrich) (DaCosta Byfield et al., 2004;Fan et al., 2007;van Boxtel et al., 2015;Vogt et al., 2011) was dissolved in DMSO to a generate a stock at 10 mM and stored at 4˚C. Fresh dilutions were made the same day experiments were carried out. Some batch-to-batch variability occurred, as well as a slight decrease in efficacy of the stock over time. The precise rescuing drug concentration must therefore be empirically determined for each stock of SB-505124. accordance with their conflict of interest policies. The other authors declare that no competing interests exist.