Conserved regulation of Nodal-mediated left-right patterning in zebrafish and mouse

On its surface, the vertebrate body plan appears bilaterally symmetric, but the morphology and positioning of internal organs reveal an underlying left-right asymmetry: the heart jogs to the left, the liver is positioned on the right and the gut undergoes asymmetric rotation. Nodal, a ligand in the TGFβ protein family, is a major regulator of left-right axis establishment (Blum and Ott, 2018; Blum et al., 2014; Grimes and Burdine, 2017). After its role in mesendoderm patterning, Nodal is expressed in the embryonic node and in the left lateral plate mesoderm (LPM) in mouse (Collignon et al., 1996; Lowe et al., 1996; Zhou et al., 1993), zebrafish (Long et al., 2003), chick (Levin et al., 1995; Pagán-Westphal and Tabin, 1998) and Xenopus (Lowe et al., 1996), and it is expressed asymmetrically in Amphioxus (Li et al., 2017), snails (Grande and Patel, 2009; Kuroda et al., 2009), echinoderms (Duboc et al., 2005) and ascidians (Morokuma et al., 2002).

The mechanisms underlying left-right patterning are well understood in mice, but less established in zebrafish, owing to a reliance on morpholino knockdown phenotypes and an absence of null mutants. Knockdown of southpaw (spaw, a zebrafish Nodal ortholog) results in an absence of spaw in the LPM, and subsequent heart jogging and looping phenotypes (Long et al., 2003), similar to Nodal mutant mice (Brennan et al., 2002; Kumar et al., 2008; Saijoh et al., 2003). In contrast, spaw mutants with a C-terminal point mutation still initiate (but do not propagate) spaw expression in the LPM, and then develop heart jogging, but not looping, phenotypes (Noël et al., 2013). This has led to the conclusion that heart chirality in zebrafish is controlled by a Nodal-independent mechanism (Noël et al., 2013). However, it is unclear whether the existing spaw allele is a null mutant. Thus, a loss-of-function mutation in spaw is needed to directly test its function in zebrafish left-right patterning, and to determine whether Nodal function is conserved between fish and mouse.

A second TGFβ family member, Gdf1, is also expressed in the node and LPM (Rankin et al., 2000). Nodal-Gdf1 heterodimers pattern the mouse left-right axis (Tanaka et al., 2007). Consistent with Gdf1 mutant phenotypes, knockdown of zebrafish vg1/gdf3 causes a reduction or absence of spaw in the LPM and heart looping phenotypes (Peterson et al., 2013), and rescue experiments of maternal-zygotic vg1 mutants suggest that Spaw requires Vg1 to pattern the left-right axis (Pelliccia et al., 2017). However, in contrast to mouse Nodal-Gdf1, Spaw-Vg1 heterodimers have not been detected in zebrafish (Peterson et al., 2013). Thus, it remains unclear whether the role of Vg1/Gdf1 in left-right patterning is conserved between fish and mouse.

Asymmetric Nodal expression in the mouse LPM is initiated and maintained through two means of inhibition. First, Nodal is inhibited on the right side of the node by Cerl2/Dand5, a member of the Cerberus/Dan family of secreted TGFβ antagonists. Like Cerl2/Dand5 mouse mutants (Marques et al., 2004), dand5/charon zebrafish morphants develop bilateral spaw expression, and heart jogging and looping defects (Hashimoto et al., 2004). Lowering dand5 levels by morpholino knockdown causes premature induction of spaw expression in the LPM, but no change in the rate of propagation was observed (Wang and Yost, 2008). Dand5 mutants have not been reported in zebrafish.

Second, Lefty1 at the midline is thought to restrict Nodal activity to the left. Zebrafish lefty1 morphants (Wang and Yost, 2008) and mutants (Rogers et al., 2017) have defects in heart jogging. lefty1 morphants also develop bilateral spaw expression (Wang and Yost, 2008), and similar to mouse lefty1 mutants (Meno et al., 1998), left LPM spaw expression precedes the right (Wang and Yost, 2008). These data have led to the suggestion that Lefty1 functions as a midline barrier during left-right patterning (Lenhart et al., 2011; Meno et al., 1998; Wang and Yost, 2008), but Lefty1 has also been proposed to act through a self-enhancement and lateral-inhibition (SELI) feedback system (Nakamura et al., 2006). Surprisingly, during mesendoderm patterning in zebrafish, Lefty1 and Lefty2 can be replaced with uniform inhibition via a Nodal inhibitor drug (Rogers et al., 2017). This raises the question of whether (1) lefty1 localization, (2) the timing of lefty1 expression and (3) dynamic feedback between Nodal and lefty1 are required for left-right patterning.

In this study, we assessed to what extent the mechanisms driving left-right patterning are conserved between mouse and zebrafish. We genetically tested the role of the secreted factors spaw, dand5 and lefty1 in zebrafish by generating single and double null mutants. Our results reveal that Spaw is required to transfer spaw expression to the LPM, Dand5 and Lefty1 restrict Spaw to the left side, and Spaw and Vg1 form heterodimers. Additionally, we show that Dand5 and Lefty1 control the timing and speed of Spaw propagation, and that the localization and timing of lefty1 expression are not always crucial for left-right patterning. Collectively, these results clarify and unify the regulatory mechanisms of left-right patterning.

To test whether the mechanisms of left-right axis establishment are conserved between mouse and zebrafish, we analyzed single and double mutants for spaw, dand5 and lefty1 (Fig. 1A,B). We used CRISPR-Cas9 to generate frameshifting mutants for dand5 and spaw, and used previously generated lefty1 mutants (Rogers et al., 2017) (Table S1). All six mutant combinations were homozygous viable and lacked gross phenotypes (Fig. 1C) but displayed randomized or symmetric heart looping and jogging (Fig. 1D,E). In particular, spaw mutants displayed randomization of heart jogging, and failed to produce a normal heart loop, contrary to spaw C-terminal point mutants (Noël et al., 2013) but similar to spaw morphants (Long et al., 2003) and mouse Nodal mutants (Brennan et al., 2002; Kumar et al., 2008; Saijoh et al., 2003).

To understand the effect of the mutations on gene expression, we analyzed lefty1, lefty2 and spaw expression during somitogenesis (Fig. 1F). In spaw, dand5;spaw and lefty1;spaw mutants, spaw expression was reduced around Kupffer's vesicle (the zebrafish equivalent to the mouse node in left-right patterning), and lefty1 and lefty2 expression was lost. In dand5, lefty1 and dand5;lefty1 mutants, spaw was expressed in both the left and right LPM, lefty1 expression was higher in the midline, and lefty2 was expressed in both heart fields (Fig. 1F). In addition, in lefty1 and dand5;lefty1 mutants, spaw was expressed in the midline, a phenotype not observed in zebrafish morphants (Wang and Yost, 2008) and only weakly visible in mouse mutants (Meno et al., 1998). However, spaw midline expression is consistent with the endogenous expression of the Nodal co-receptor one-eyed pinhead (oep)/tdgf1 and the Nodal pathway transcription factor foxh1 in the midline (Pogoda et al., 2000; Sirotkin et al., 2000; Zhang et al., 1998). Thus, Spaw is an upstream activator of spaw, lefty1 and lefty2, and Dand5 and Lefty1 restrict the localization of spaw. Collectively, these results reveal that the roles of Nodal, Dand5 and Lefty1 are conserved from mouse to zebrafish.

Vg1 is not detectably processed or secreted on its own, but upon co-expression with its heterodimeric Nodal partners cyclops or squint, it is cleaved to its mature form and secreted (Montague and Schier, 2017). To determine whether Spaw and Vg1 also form functional heterodimers like Cyclops-Vg1, Squint-Vg1 and mouse Nodal-Gdf1, we generated superfolderGFP (sfGFP) and epitope-tagged versions of Spaw and Vg1, and tested the activity, processing, interaction and localization of Spaw and Vg1. Expression of vg1-sfGFP in the presence and absence of spaw revealed that, while Spaw-sfGFP alone was cleaved to its mature form and secreted, Vg1-sfGFP was cleaved and secreted only in the presence of Spaw (Fig. 2A-C).

If Vg1 and Spaw form heterodimers, a biochemical interaction should be detectable. We performed co-immunoprecipitation experiments by co-expressing vg1-Flag with spaw-HA, or squint-HA as a positive control (Montague and Schier, 2017). In contrast to previous studies (Peterson et al., 2013), an interaction was detected between Vg1 and Spaw (Fig. 2D), indicating that Vg1 and Spaw form heterodimers.

To test whether Spaw activity is dependent on Vg1, we expressed Spaw in the presence or absence of Vg1. Expression of 0.25-5 pg of spaw mRNA in blastula embryos induced ectopic Nodal target gene expression in wild-type embryos, but failed to induce ectopic gene expression in maternal vg1 (Mvg1) mutants (Fig. 2E), indicating that Spaw requires Vg1 for full activity. Nodal target gene expression was induced in embryos injected with 20 pg or more of spaw mRNA in the absence of vg1, suggesting that, at high concentrations, Spaw might be able to form active homodimers. Collectively, these results suggest that, after Spaw is translated, it heterodimerizes with inactive, unprocessed Vg1. Vg1-Spaw dimerization facilitates the cleavage and secretion of Vg1, producing a mature active heterodimer.

Analysis of spaw expression in the single and double mutants revealed that spaw propagated through the LPM of dand5 and lefty1 mutants before it had initiated LPM expression in wild-type embryos (Fig. 1F). To understand whether this was due to a difference in the onset of spaw expression, the rate of spaw propagation, or both, we analyzed spaw expression (co-stained with the somite marker myod1) across multiple stages of somitogenesis in the single and double mutants. spaw expression never initiated in the LPM of spaw, dand5;spaw or lefty1;spaw mutants (Fig. 3A,B), contrary to previous results (Noël et al., 2013), indicating that Spaw activity is required to initiate spaw expression in the LPM. spaw expression initiated prematurely in the LPM of dand5 and lefty1 mutants, consistent with previous results using morpholino knockdown (Wang and Yost, 2008), and it initiated earliest in dand5;lefty1 mutants (Fig. 3A,B). To understand whether the premature LPM spaw expression was caused by earlier initiation of spaw around Kupffer's vesicle, we analyzed the onset of spaw expression in wild type, dand5, lefty1 and dand5;lefty1 mutants. Indeed, spaw expression was already present around Kupffer's vesicle two to four somite stages (1-2 h) earlier in dand5, lefty1 and dand5;lefty1 mutants than in wild-type embryos (Fig. S1). Thus, Dand5 and Lefty1 both spatially and temporally regulate Spaw activity.

To test whether the rate of spaw propagation is affected by the loss of its inhibitors, we measured the distance of spaw expression from the tailbud as a function of somite number (as a proxy for time) in the dand5, lefty1 and dand5;lefty1 mutants. spaw propagated at an approximate rate of 108 µm per somite generation in wild-type embryos and ∼112 µm/somite in dand5 mutants. Contrary to previous morpholino studies (Wang and Yost, 2008), the rate of spaw propagation was slightly higher in lefty1 mutants (∼140 µm/somite) and dand5;lefty1 double mutants (∼125 µm/somite) (Fig. 3B, Fig. S2). Thus, Lefty1 may reduce the speed of Spaw propagation during normal left-right patterning.

Collectively, these results suggest that peri-nodal Spaw induces the expression of spaw in the LPM. Dand5 and Lefty1 inhibit Spaw activity both temporally, which delays the onset of spaw expression, and spatially, which restricts spaw expression to the left LPM. In addition, Lefty1 may reduce the speed at which spaw propagates through the LPM. Although the timing of Nodal propagation in the presence and absence of inhibitors has not been assessed in mouse, the spatial restriction of Nodal by its inhibitors is conserved from fish to mouse.

Our results indicate that Lefty1 is required to restrict and/or maintain Spaw activity in the left LPM. To test whether the localization or timing of lefty1 expression is necessary for correct patterning, we replaced lefty1 expression with uniform Nodal inhibition by soaking lefty1 embryos in 1 µM of a Nodal signaling inhibitor drug, SB-505124, from 90% epiboly onwards (prior to the onset of lefty1 expression) (Rogers et al., 2017). Surprisingly, leftward heart jogging was rescued in the majority of drug-treated lefty1 embryos (Fig. 4A,D). Analysis of spaw expression in drug-treated or untreated lefty1 mutants revealed that treatment with SB-505124 was also able to rescue left LPM spaw expression in a subset of embryos (Fig. 4B-D). To dissect why there was a difference between the proportion of embryos with left spaw expression versus left-biased hearts, we analyzed lefty2 expression in the heart field at an intermediate stage between spaw LPM expression and heart positioning at two different concentrations of Nodal inhibitor. lefty2 expression and leftward heart jogging were rescued in a similar proportion of drug-treated lefty1 mutant embryos (Fig. 4D). This suggests that in situ hybridization is not sensitive enough to detect low levels of spaw expression that can lead to asymmetric lefty2 expression, or that leftward heart jogging is not entirely dependent on the presence of an asymmetric Nodal signal.

These results reveal that left-right patterning can occur in the absence of timely expression and midline localization of lefty1. In addition, although Lefty1 and Spaw normally function in a feedback system (Fig. 1B), Spaw-Lefty1 feedback can be removed for normal left-right patterning. Thus, lefty1-mediated inhibition may primarily function to dampen Nodal signaling to prevent its amplification and extension to the right LPM.

Our mutant analyses clarify four roles of Spaw and its inhibitors in zebrafish left-right patterning. First, spaw-null mutants show disruptions to both heart jogging and looping, indicating that Spaw is required for proper heart asymmetry. In addition, spaw, spaw;dand5 and spaw;lefty1 mutants fail to initiate spaw expression in the LPM, revealing that Spaw around Kupffer's vesicle is required to transfer spaw expression to the LPM. This is consistent with mouse data: mouse mutants that lack Nodal activity in the node fail to express Nodal in the LPM, and later develop left-right patterning defects, including heart jogging and looping phenotypes (Brennan et al., 2002; Kumar et al., 2008; Saijoh et al., 2003). Thus, these results reveal conservation between the roles of zebrafish Spaw and mouse Nodal in left-right patterning.

Second, our results show that in blastoderm assays, Vg1 is required for Spaw activity, Spaw allows the secretion and processing of Vg1, and Spaw and Vg1 interact. These results are consistent with genetic and biochemical data showing Gdf1 is a heterodimeric partner of Nodal during mouse left-right patterning (Tanaka et al., 2007).

Third, mutations in the Nodal inhibitors dand5 and lefty1 result in bilateral spaw expression, demonstrating functional conservation with the mouse orthologs Cerl2/Dand5 and Lefty1, the absence of which leads to bilateral Nodal expression and organ asymmetry defects (Marques et al., 2004; Meno et al., 1998). In addition, our results reveal that the loss of Dand5 and Lefty1 leads to precocious spaw expression, and the loss of Lefty1 leads to accelerated spaw progression, suggesting these inhibitors normally control the timing, speed and localization of Nodal expression.

Fourth, application of a Nodal inhibitor drug to lefty1 mutant embryos is sufficient to rescue normal heart jogging and left-sided expression of spaw in most animals, suggesting that the localization of Lefty1 activity is not always crucial. Rather, Lefty1 dampens Nodal activity to prevent its amplification and extension to the right side of the embryo.

Collectively, our results show that the mechanisms of left-right patterning are highly conserved from mouse to zebrafish and suggest a refined model of zebrafish left-right patterning: (1) rotation of cilia in Kupffer's vesicle causes leftward flow of fluid; (2) Dand5 creates an initial bias in Spaw-Vg1 heterodimer activity at the node; (3) Spaw-Vg1 induces the expression of spaw in the LPM; and (4) Lefty1-mediated inhibition maintains the left-right bias of Spaw activity during auto-induction (Fig. 5). Given that the basic functions of Nodal, Dand5, Lefty and cilia/the left-right organizer are also conserved in Xenopus left-right patterning (Blum et al., 2014; Branford et al., 2000; Schweickert et al., 2010, 2007; Vonica and Brivanlou, 2007), these results suggest that the roles of the Nodal cascade and left-right organizer may be conserved across all vertebrates, with the exception of the sauropsida (Blum and Ott, 2018). These studies lay the foundation for future biochemical and biophysical analyses needed to determine how fluid flow is converted into asymmetric gene expression, and how Nodal pathway components interact to generate precise timing and localization during left-right patterning.

All vertebrate animal work was performed at the facilities of Harvard University, Faculty of Arts and Sciences (HU/FAS). The HU/FAS animal care and use program maintains full AAALAC accreditation, is assured with OLAW (A3593-01) and is currently registered with the USDA. This study was approved by the Harvard University/Faculty of Arts and Sciences Standing Committee on the Use of Animals in Research and Teaching under protocol number 25-08.

Zebrafish embryos were grown at 28°C and staged according to Kimmel et al. (1995). For somite stages (except drug treatments), embryos were grown at 28°C until sphere stage, and then grown overnight at room temperature. Embryos were cultured in blue water (250 mg/l Instant Ocean salt, 1 mg/l Methylene Blue in reverse osmosis water adjusted to pH 7 with NaHCO₃). Embryos for immunoblot and live imaging experiments were chemically dechorionated with 1 mg/ml pronase (protease type XIV from Streptomyces griseus, Sigma) prior to injection, and cultured in agarose-coated dishes. Embryos were injected at the one-cell stage unless otherwise stated.

sgRNAs targeting the dand5/charon and spaw genes were designed using CHOPCHOP (Labun et al., 2016; Montague et al., 2014) (Table S1) and synthesized as previously described (Gagnon et al., 2014). TLAB wild-type zebrafish embryos were injected at the one-cell stage with sgRNAs and ∼0.5 nl of 50 µM Cas9 protein. Injected embryos were raised to adulthood and outcrossed to TLAB adults. The resulting clutches were genotyped to identify potential founders with germline mutations in dand5 or spaw by MiSeq sequencing. Animals were recovered with a 4 bp deletion in the first exon of dand5 (dand5^a204), causing a frameshift and truncation of Dand5 from 243 amino acids to 54 amino acids, and a 49 bp deletion in the first exon of spaw (spaw^a205), resulting in a frameshift and truncation of Spaw from 404 amino acids to 49 amino acids (Table S1). The offspring of confirmed founders were raised to adulthood and genotyped by fin-clipping to determine heterozygous individuals. Homozygous animals were generated by intercrossing heterozygous fish.

dand5;lefty1, dand5;spaw and lefty1;spaw mutants were generated by intercrossing homozygous dand5, spaw and lefty1 (Rogers et al., 2017) mutants to generate double heterozygous mutants. Double heterozygous mutants were intercrossed and genotyped to identify double homozygous mutants.

Genomic DNA was extracted from embryos and fin-clips using the HotSHOT method (Meeker et al., 2007), and PCR was performed using standard methods. spaw and vg1/gdf3 mutants (vg1^a165) (Montague and Schier, 2017) were genotyped by gel electrophoresis, dand5 mutants were genotyped by Sanger sequencing, and lefty1 mutants (lefty1^a145) were genotyped by PshAI digestion (Rogers et al., 2017). Primer sequences were as follows: dand5_genotype_F, TCACAACAATTGGCGCTTTC; dand5_genotype_R, GCCAAGAACGCGGGAAAC; lefty1_genotype_F, CGTGGCTTTCATGTATCACCTTC; lefty1_genotype_R, GGATGCCGGCCAAACTG; spaw_genotype_F, CGCGTTATTTGTTTTACGCGTTG; spaw_genotype_R, TGCTTACCTTGTGCAATCAAGC; vg1_genotype_F, CTTGCAGATGTGGATTTCTGGCC; and vg1_genotype_R, CATGATGCGATGGTTTGGGTCG.

The spaw CDS sequence was PCR amplified from a somite-stage cDNA library and cloned into the pCS2(+) vector using Gibson cloning (Gibson et al., 2009) to generate pCS2(+)-spaw. spaw-sfGFP was generated by inserting the superfolder GFP (sfGFP) sequence (Pédelacq et al., 2006) downstream of the spaw cleavage site (RHKR) in pCS2(+)-spaw with Gibson cloning. mCherry-CAAX was generated by inserting a farnesylation sequence at the 3′ end of the mCherry CDS in a pCS2(+) vector. vg1, vg1-sfGFP, vg1-Flag and sqt-HA have been previously reported (Montague and Schier, 2017). spaw-HA was generated by inserting the HA tag (YPYDVPDYA) sequence downstream of the cleavage site by site-directed mutagenesis of pCS2(+)-spaw. For all fusion constructs, the epitope or sfGFP tag was flanked by a GSTGTT linker at the 5′ end of the tag and by a GS linker at the 3′ end of the tag. In situ probes targeting lefty1, spaw, lefty2, myl7 and myod1 were generated by inserting the full CDS, or up to 1000 bp of the CDS of the gene into the pSC vector using Strataclone (Agilent).

Expression constructs were linearized by digestion with NotI, followed by mRNA synthesis using the SP6 mMessage Machine Kit (ThermoFisher). In situ antisense probes were synthesized using a DIG Probe Synthesis Kit (Roche).

Embryos were fixed overnight at 4°C in 4% formaldehyde, followed by whole-mount in situ hybridization according to standard protocols (Thisse and Thisse, 2008). NBT/BCIP/alkaline phosphatase-stained embryos were imaged using a Zeiss Axio Imager.Z1 after dehydration with methanol and clearing with benzyl benzoate:benzyl alcohol (BBBA).

Embryos at 24 h post-fertilization (hpf) were anesthetized using Tricaine (Sigma) and imaged in 2% methylcellulose. spaw-sfGFP-injected embryos were mounted at sphere stage in 1% low gelling temperature agarose (Sigma) and live imaged using a Zeiss LSM 700 confocal microscope.

Embryos were injected with 50 pg of mRNA (unless otherwise stated), and grown until 50% epiboly, then flash frozen (eight embryos per sample) in liquid nitrogen after manual deyolking using forceps. The samples were boiled for 5 min at 95°C with 2× SDS loading buffer and DTT, and then loaded onto Any kD protein gels (Bio-Rad). The samples were transferred to polyvinylidene fluoride (PVDF) membranes (GE Healthcare) before blocking in 5% non-fat milk (Bio-Rad) in TBST at 4°C overnight with primary antibodies (rabbit anti-GFP, 1:5000, ThermoFisher A11122; mouse anti-α-Tubulin, 1:5000, ICN Biomedicals 691251; rabbit anti-Flag, 1:2000, Sigma F7425). HRP-coupled secondary antibody was applied to the membranes (goat anti-rabbit, 1:15,000, Jackson ImmunoResearch Labs 111-035-144) and Amersham ECL reagent (GE Healthcare) was used for chemiluminescence detection. For co-immunoprecipitation experiments, 80 embryos per sample were homogenized in 400 µl of ice-cold lysis buffer (50 mM Tris at pH 7.5, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 1% Triton X-100 and protease inhibitors, Sigma 11836170001) and vortexed every 5 min during a 30-min incubation on ice. Samples were spun at 4°C for 30 min, and then the supernatant was transferred to a tube with 50 µl of anti-HA affinity matrix (Roche 11815016001). Samples were rocked at 4°C overnight. The samples were spun for 2 min and washed in ice-cold wash buffer (50 mM Tris at pH 7.5, 150 mM NaCl, 1% Triton X-100 and protease inhibitors) five times before adding 2×SDS loading buffer and DTT. Immunoblots were performed as above.

Images were processed in ImageJ/FIJI (Schindelin et al., 2012). Brightness, contrast and color balance were applied uniformly to images.

Embryos were grown at 28°C until 75% epiboly, then dechorionated using pronase. At 90% epiboly, the embryos were placed in agarose-coated dishes containing pre-warmed Nodal inhibitor SB-505124 (S4696, Sigma) diluted in blue water. After development to 18 somites, 22 somites or 28 hpf at 28°C, the embryos were fixed for in situ hybridization in 4% formaldehyde. SB-505124 was stored at 4°C in a 10 mM stock.

Embryos were co-stained using a mix of the spaw and myod1 probes. To calculate spaw propagation in the LPM, a measurement was made from the lowest point of myoD staining (base of the notochord) to the highest point of spaw staining on the left side of the embryo. For embryos in which spaw had propagated a substantial distance up the LPM, the embryo was imaged from multiple orientations, and the distances summed. Embryos were also imaged from the side to ensure that the curved surface of the embryo did not affect the final measurement. For spaw propagation calculations, a linear regression line was fit to the scatter plot of spaw distance at time points after initiation of expression in the LPM. This ensured that the timing of spaw propagation did not affect the slope of spaw propagation.

We are grateful to Andrea Pauli and Nate Lord for helpful comments on the manuscript, Fred Rubino for immunoblot advice, Andrea Pauli for generating the mCherry-CAAX plasmid, and Kathryn Berg and Joo Won Choi for assistance with genotyping.