To understand the relationship between somitogenesis and LR axis formation in the Xenopus embryo, dmrt2 was chosen for analysis because it is expressed in the fish LRO (Kupffer’s vesicle), suggesting a specific role during symmetry breakage (Saúde et al., 2005; Liu et al., 2009; Lourenço et al., 2010). In early tadpole stages, dmrt2 is expressed in somitic tissue as demonstrated by whole-mount in situ hybridization (WMISH; data provided by Soeren S Lienkamp @ Xenbase; Bowes et al., 2009), indicating a conserved activity within vertebrates. Using WMISH, we detected strong dmrt2 expression in the Xenopus LRO at neurula stages, resembling expression in the fish LRO. However, dmrt2 was restricted to flow-generating cLRO cells, while lateral sLRO cells did not express dmrt2 (Supplemental Figures S1A, A′).

A unique feature of the frog system is the ability to restrict experimental manipulations in the early embryo on the left or right side, making it particularly suited to analyze LR axis development. Unilateral injections of synthetic mRNAs or antisense morpholino oligos (MO) into four to eight cell embryos allow to perform site-directed gain- or loss-of-function experiments and analyzing their impact on LR axis formation. To analyze the potential role of Dmrt2 during LR development, a translation-blocking morpholino oligo (dmrt2 MO) was designed. dmrt2 MO was injected in a site-specific manner and laterality was determined by pitx2 expression. Untreated controls and right-sided dmrt2 knockdown showed wildtype (WT) pitx2 asymmetry (Figure 1A and not shown). Left dmrt2 MO injections, however, resulted in the loss of left pitx2 transcription in about 60% of cases (Figures 1B, C). Importantly, asymmetry was statistically significantly restored by co-injecting full-length dmrt2 mRNA, which was insensitive to the dmrt2 MO, indicating the specificity of the observed phenotype (Figure 1C). Next, we analyzed the effect of dmrt2 loss of function on leftward flow. dmrt2 MO was bilaterally injected into four to eight cell embryos, targeting the central LRO lineage. The dorsal explants of neurula embryos were dissected and morphants, as well as untreated controls, were processed for flow analysis by adding fluorescent microbeads and subsequent recording of bead motion. While controls showed WT leftward movement of beads (Figures 1D, F, G), flow velocity and directionality were statistically significantly diminished in dmrt2 morphants (Figures 1E, F, G), demonstrating that Dmrt2 is required for cilia-driven symmetry breakage. Next, flow-generating LRO cilia of controls and unilaterally injected morphants were analyzed by immunofluorescence (IF) using an anti-acetylated tubulin antibody. F-actin staining using fluorescently tagged phalloidin visualized cell borders. WT cLRO cells were ciliated and cilia length was around 6 µm on average (Figures 1H, J), matching our previous findings (Schweickert et al., 2007). Although the pattern of ciliation was unaltered in dmrt2 morphants (Figure 1I), cilia were substantially shortened to about 2–2.5 µm (Figures 1I, J), providing an explanation for flow deficiency. Recently, the master regulator of motile cilia, foxj1, was shown to be a transcriptional target of Dmrt2 in fish (Pinto et al., 2018). The loss of flow and shortened cilia in dmrt2 morphants could therefore reflect impaired foxj1 expression in the cLRO precursor cells at gastrula stages. Indeed, 80% of unilaterally injected dmrt2 morphants showed diminished foxj1 expression in the SM on the targeted embryo half (Figures 1K–M), which correlated with defective ciliogenesis at the LRO. Next, we asked whether dmrt2 was differentially expressed in SM and the underlying deep mesoderm (DM). To address this question, dorsal mesodermal explants of early gastrula embryos were dissected and further bisected into SM and DM. Using RT-PCR, dmrt2 mRNA was detected in both tissues (Supplemental Figures S1B, C). However, dmrt2 knockdown did not diminish DM expression of the organizer genes goosecoid and chordin (Supplemental Figures S1D–G), which excludes an impact on organizer formation. We conclude that at gastrula stages, Dmrt2 activity is required for cLRO morphogenesis and thus for correct LR development.

During the analysis of LRO cilia, we noted that in untreated specimens, F-actin staining was more intense in sLRO than cLRO cells (Figure 1H; data not shown). Enhanced actin signals can be the consequence of apical constriction, a cell shape change observed in sLRO cells (Shook et al., 2004). Surprisingly, this sLRO-specific actin staining was not detected in dmrt2 morphants (Figure 1I). Based on our recent work on Fgf function during LRO morphogenesis (Schneider et al., 2019), we proposed that either apical constriction failed or sLRO cells were entirely absent. To analyze this sLRO phenotype in more detail, the expression of nodal1 was assessed. The morphogen nodal1 is specifically expressed in sLRO cells and is required to transfer left identity to the LPM (Figure 2A; Blum et al., 2014b). Targeting left sLRO cells (c.f. Tingler et al., 2014) with dmrt2 MO diminished nodal1 signals (Figures 2B, D), suggesting that this effect contributed to the failure of Nodal cascade induction in the left LPM. Co-injecting full-length dmrt2 mRNA restored nodal1 expression in morphants, although domains were generally smaller in size compared to WT embryos (Figure 2C). Similar results were obtained when right-sided knockdown was performed or dand5 was analyzed (data not shown). In addition, myod1 expression was lost upon knockdown of dmrt2 in sLRO cells. Histological sections revealed that the endodermal layer, which is located at a distance to sLRO cells in control embryos (Supplemental Figure S2A), is shifted and located next to notochordal cLRO cells in dmrt2 morphants (Supplemental Figure S2B). Taken together, these data strongly suggest that the presence, but not apical constriction, of sLRO cells depends on dmrt2 activity.

Next, we asked whether Dmrt2 function during paraxial patterning or myogenesis caused the absence of nodal1-expressing sLRO cells in neurula stages. In order to address such a connection during Xenopus LR development, we first analyzed the two myogenic marker genes tbx6 and myf5 in dmrt2 morphants. The genes were chosen as 1) tbx6 knockout mice display LRO defects and 2) murine myf5 is a direct transcriptional target of Dmrt2 (Hadjantonakis et al., 2008; Sato et al., 2010; Concepcion et al., 2018). At neurula stages, both genes were expressed in presomitic mesoderm and importantly in sLRO cells (Supplemental Figures S3A, B), strongly suggesting a connection between myogenic pathways and sLRO morphogenesis. We thus analyzed tbx6 and myf5 expression in gastrula embryos, which had been unilaterally injected with dmrt2 MO. Both genes were strongly downregulated when dmrt2 function was inhibited (Figures 2E–L). The reintroduction of dmrt2 mRNA statistically significantly restored tbx6 expression (Figure 2H), further underscoring MO specificity. These results showed that Dmrt2 acts upstream of the myogenic transcription factors tbx6 and myf5, the latter being in accordance with published data in mice (Sato et al., 2010).

Together, the above results showed that Dmrt2 regulates both somitogenic and sLRO genes, suggesting a novel functional link between somitogenesis and the processing of LR cues. If patterning of the paraxial mesoderm is linked to LR asymmetry, the loss of function of myogenic key genes should impact laterality. We, therefore, turned to manipulate myf5 (Pownall et al., 2002), which, unlike tbx6, has not been implicated in LR axis formation. Tbx6 was shown to transcriptionally activate myf5 (Li et al., 2006), rendering myf5 an ideal downstream target for the loss-of-function experiments. A translation blocking MO was used to analyze the role of myf5 during LR axis formation. At st. 31, control embryos expressed pitx2 on the left side (Figures 3A, D). Right-sided myf5 MO injections had no effect on pitx2 asymmetry (data not shown). However, applying myf5 MO to the left sLRO lineage prevented pitx2 induction (Figures 3B, D). The loss of pitx2 asymmetry was specific, as left pitx2 expression was restored in morphants co-injected with myf5 rescue mRNA (Figures 3C, D). Next, we analyzed sLRO specification by detecting nodal1 mRNA (Figures 3E–H). Left-sided myf5 knockdown either impeded nodal1 expression entirely or substantially reduced its domain (Figures 3F, H). In addition, sLRO cells were not present in myf5 morphants as visualized by the lack of myod1 expression (Supplemental Figure S2). The reintroduction of myf5 mRNA reduced the severity of the loss-of-function phenotype, suggesting MO specificity (Figures 3G, H). This demonstrates a crucial role for the myogenic transcription factor myf5 in LR axis determination, as it specifies the sensory cells of a functional LRO.

The loss of function of myf5 ultimately phenocopied the dmrt2 loss of function, strongly suggesting that both act in the same pathway. To test whether both genes co-operate, suboptimal dmrt2 MO and myf5 MO doses were injected either individually or together into the sLRO lineage. Compared to control embryos, which expressed pitx2 exclusively on the left side (Figure 4A), the individual injection of each MO at a low dose affected LR development in 20–30% of embryos (Figure 4C). However, in embryos that received a combination of both MOs at a low dose, pitx2 expression was altered in the 75% of cases (Figures 4B, C), suggesting functional cooperation of myf5 and dmrt2 in LR determination. Formally, the genes could interfere with LR development individually at different stages or in different tissues. To demonstrate that both, dmrt2 and myf5, act together in the same process, i.e. the specification of sLRO cells, nodal1 transcription was analyzed at neurula stages using the same experimental setup as described above. Compared to controls (Figures 4D, G), individual injection of suboptimal doses of dmrt2 MO or myf5 MO mildly reduced the nodal1 expression domain in about 60% of specimens (Figures 4E, G). In contrast, co-injecting low concentrations of dmrt2 MO and myf5 MO entirely prevented nodal1 expression in 80% of specimens (Figures 4F, G). These results strongly argue that Dmrt2 and Myf5 jointly specify sLRO tissue. Next, we investigated whether this functional relationship was epistatic. The strong effect of dmrt2 loss of function on myf5 expression during gastrula stages suggests that myf5 acts downstream of dmrt2. Indeed, the loss of nodal1 expression in dmrt2 morphants was very efficiently rescued by co-injecting myf5 mRNA (Figures 4H–K). The frequency of restored nodal1 transcription almost reached WT levels, demonstrating the sequential order of gene activities. Together, these results show that dmrt2 governs both cLRO morphogenesis in the axial midline as well as paraxial mesoderm patterning. We identify the joint specification of sLRO and somitic cells as a prerequisite for LR axis specification.

The aim of this study was to reveal a functional interaction between LR axis specification and paraxial patterning/somitogenesis. At first glance, it appears that these two processes are mutually exclusive in vertebrate embryos and must thus occur independently of each other: the perfect symmetry of somitogenesis which lays the ground for the symmetric formation of vertebrae and ribs, must not be disturbed by the asymmetry created along the LR axis.

During somitogenesis, a complex gene regulatory network that includes oscillating gene expression is orchestrated to ensure perfectly bilaterally symmetric development. Asymmetries in this context could result in nonfunctional musculature and skeletal defects, threatening the survival of the embryo. On the other hand, a highly complex mode of symmetry breakage, generated by a cilia-driven flow of extracellular fluids, is translated into the asymmetric release of the very potent morphogen Nodal. Nodal transfers leftness into the LPM and therefore could broadly impact various neighboring tissues along the left anterior-posterior axis. Indeed, several reports showed that Nodal interferes with the left somitic clock, i.e., the oscillatory gene expression module, in mice and chicks. Retinoic acid (RA) is thought to prevent such interference by shielding left-sided somites from Nodal-induced signal transduction (Vermot and Pourquié, 2005; Sirbu and Duester, 2006; Brend and Holley, 2009; Grimes, 2019). However, RA-mediated protection acts much later than the factors that we identify here, showing that both reflect distinct processes. Interestingly, the loss of dmrt2 in fish desynchronized the somitic clock and led to LR defects, underscoring a molecular link between both processes (Saúde et al., 2005; Liu et al., 2009). Although we have not analyzed somite segmentation in Xenopus, we have identified a potential mechanism in which genes required for somitogenesis also act on LRO specification and morphogenesis.

In Xenopus, cell labeling experiments have demonstrated that sLRO cells are fated to become somitic tissue. More specifically, after flow sensing, sLRO cells ingress into the somites and differentiate into the horizontal myoseptum which divides the somite into dorsal and ventral regions (Shook et al., 2004). On the molecular level, we confirmed these observations by showing that the myogenic marker genes myod1, tbx6, and myf5 are expressed in the sensory part of the LRO (Schweickert et al., 2010; Schneider et al., 2019 and this work). Functionally, we demonstrated a requirement of Dmrt2 and Myf5 for sLRO specification/morphogenesis. The role of myf5 during LR development was particularly unanticipated because the involvement of a bona fide myogenic transcription factor in LR axis formation was not reported so far. Therefore, we conclude that in Xenopus, paraxial patterning, i.e. somitogenesis, is functionally linked to symmetry breakage.

We found that Dmrt2 is required for SM specification and paraxial patterning at gastrula stages. These early events are thus essential to establish a leftward flow driven by motile cilia on the cLRO and its subsequent left-sided sensing in somitic sLRO cells. How flow is perceived remains an open question. But is Dmrt2 acting in a cell-autonomous or non-cell-autonomous manner, i.e. in the SM or in the underlying DM? We detected dmrt2 transcripts in both cell layers (Supplemental Figures S1A, B) which did not allow for a differentiation between both modes. However, since foxj1 is exclusively expressed in SM cells (Stubbs et al., 2008; Beyer et al., 2012) and foxj1 is a transcriptional target of Dmrt2 in fish (Pinto et al., 2018), a cell-autonomous Dmrt2 activity to induce foxj1 expression seems plausible. Importantly, this likely applies to the axial part of the SM, the cLRO precursor cells, but not to lateral SM cells which are fated to develop into sLRO tissue. Our dissection approach at gastrula stages did not discriminate between axial and lateral SM or between axial (notochordal) and lateral (presomitic) DM. As myf5 expression is restricted to the lateral deep mesodermal layer (cf. Supplemental Figure S4) and because Myf5 acts epistatic to Dmrt2 during specification of nodal1-positive sLRO cells (Figure 4), a non-cell-autonomous activity for Dmrt2 seems to be plausible, too.

We have recently reported that Fgf signaling also plays a dual role during LRO formation (Schneider et al., 2019). Blocking Fgf signaling prior to gastrulation diminished foxj1 expression in gastrula embryos (Schneider et al., 2019), which is indicative of impaired SM specification and consequently loss of LRO cilia and loss of leftward flow (Stubbs et al., 2008; Beyer et al., 2012). This function is probably conserved, as LRO morphogenesis in mice and fish depends on Fgf signaling as well (Hong and Dawid, 2009; Sudheer et al., 2016). When Fgf signaling was blocked from mid-gastrula stages onward, foxj1 expression and LRO ciliation were not affected, but induction of the left-sided Nodal cascade failed due to a loss of sLRO cells (Schneider et al., 2019).

dmrt2 LOF also leads to a loss of sLRO cells. However, Dmrt2 functions in early gastrulae, i.e. substantially earlier than the time point at which the inhibition of Fgf signaling induces loss of sLRO cells. Interestingly, myf5, which is absent in dmrt2 morphants, induces the somitic expression of Fgf4 and Fgf6 in mice. Via this route, it may provide a secondary Fgf signal for sLRO morphogenesis (Grass et al., 1996; Fraidenraich et al., 2000). As only SM cells develop into the sLRO and since myf5 mRNA is only present in the DM that does not contribute to the LRO, it is still unclear how Myf5 is able to regulate sLRO formation. Together with published data, our observations suggest that Myf5 in the DM influences specification of the SM in a non-cell-autonomous manner, potentially via secreted Fgf ligands. The existence of two temporally distinct Fgf pathways is in agreement with published work on the role of Fgf during gastrulation. Early Fgf signaling is transduced by the MAPK pathway, whereas the late Fgf signal uses calcium as a second messenger (Nutt et al., 2001; Sivak et al., 2005). It remains to be seen whether Fgf ligands induced by Myf5 trigger the Fgf/Ca2⁺ pathway for sLRO specification and/or morphogenesis. In a hierarchical model, Dmrt2, potentially induced by an early Fgf signal, induces foxj1 in the LRO precursor tissue, which is required for ciliogenesis and for setting up a leftward flow. In parallel, Dmrt2 induces the myogenic genes tbx6 and myf5. Myf5, possibly via a second phase of Fgf signaling, induces sLRO specification and morphogenesis. In this dual setting, Dmrt2 represents a crucial factor for LR determination in Xenopus laevis (Figure 5).

To what extent is the role of dmrt2 evolutionarily conserved and thus transferable to other vertebrate species? It is most likely that Dmrt2-dependent LRO morphogenesis is conserved in fish. This notion is supported by three recent publications: 1) zebrafish with diminished dmrt2 levels develop LR and somitogenesis defects (Saúde et al., 2005; Liu et al., 2009); 2) the zebrafish LRO expresses dmrt2 (Lourenço et al., 2010), and 3) in zebrafish, the master control gene for the biogenesis of motile cilia foxj1, is a transcriptional target of Dmrt2 (Pinto et al., 2018). Surprisingly, dmrt2 knockout mice do not exhibit LR defects, suggesting that mouse LRO morphogenesis is independent of Dmrt2 (Lourenço et al., 2010). A potential interpretation of this finding is that mammalian Dmrt2 has lost its LR function during evolution, which is therefore found in lower vertebrates only (see below). Cell lineage analysis showed that murine LRO (node, posterior notochord) cells, have notochord identity (Wilson and Beddington, 1996; Kinder et al., 2001; Yamanaka et al., 2007; Wang and Ware, 2009; Babu and Roy, 2013). In addition, sLRO (crown cell)-specific transgenes, which are widely used (e.g. NDE-lacZ; Brennan et al., 2002; Krebs et al., 2003), have not been reported to mark somitic cells. Together, this renders evolutionary conservation of the cell lineage of the mouse and Xenopus LRO implausible. This might explain the lack of an LR phenotype in the dmrt2 knockout mouse (Saúde et al., 2005).

However, a link of LR asymmetry with somitogenesis appears to be conserved in other vertebrates. The human Klippel-Feil syndrome (KFS) is characterized by segmentation defects of the vertebrae, pointing to impaired embryonic somitogenesis. Intriguingly, several KFS case reports describe the concomitant occurrence of laterality defects, suggesting that somitogenesis and LR are linked in humans. Interestingly, mutations in the human GDF3 gene have been found to be causative of KFS (Jalil et al., 2008; Chacón-Camacho et al., 2012; Futane and Salunke, 2013; Karaca et al., 2015; Abdali et al., 2021). Therefore, GDF3 could directly connect KFS clinical pictures to its well-established function during laterality determination. Unfortunately, the genetic basis of KFS patients showing situs inversus or heterotaxia has not been mapped in most cases and needs further experimental validations.

In contrast to mice, labeling of the fish LRO precursor cells showed a notochordal and a somitic cell fate (Melby et al., 1996), indicating homology to Xenopus. However, the specific whereabouts of dand5/nodal positive sLRO cells have not been addressed so far. In Medaka, nodal was detected in presomitic mesoderm at the early LRO, prior to flow and dand5 asymmetry (Hojo et al., 2007). Based on the functional similarities of frog and fish dmrt2, a somitic fate in both species seems plausible. This argument is strongly supported by a recent report. In zebrafish, it was demonstrated that a dand5 promotor-driven EGFP transgene marked LRO cells, which at later stages were found to be integrated into the axial and presomitic mesoderm (Ikeda et al., 2022). Interestingly, studies in sauropsida such as turtles, geckos, and the chick identified bilateral nodal expression domains at the embryonic midline that have somitic cell fates (Otto et al., 2014; Kajikawa et al., 2020). This is notable because this vertebrate clade induces asymmetry by an as yet unidentified mechanism. This unknown process triggers downregulation of right-sided paraxial nodal, resulting in Nodal cascade induction only in the left LPM. In chick embryos, two extracellular inhibitors of the Cerberus family, caronte and cerberus itself are initially expressed in the presomitic mesoderm and were shown to be required for chick LR development (King and Brown, 1999; Esteban et al., 1999; Yokouchi et al., 1999; Yu et al., 2008; Katsu et al., 2012). From an evolutionary point of view, co-expression of Nodal and a Cerberus-related inhibitor seems to be a module that is conserved and active in LR determination of all vertebrates. This notion is further underscored by the development of the cephalochordate Branchiostoma, an animal that exhibits LR asymmetries of all organs and tissues, including the somites. Like in the frog, a cilia-driven leftward flow downregulates dand5 in Branchiostoma, which allows activation of a left-sided Nodal cascade. Unlike vertebrates, both processes, flow-dependent dand5 inhibition and Nodal cascade propagation are restricted to only one tissue, the presomitic mesoderm. In consequence, asymmetric gene expression induces asymmetric differentiation of somites and other tissues during embryogenesis (Blum et al., 2014a; Soukup et al., 2015; Li et al., 2017; Soukup, 2017; Zhu et al., 2019). We, therefore, postulate that a “Nodal/Cerberus-like inhibitor” module is conserved among vertebrates, although the modes of symmetry breakage change during evolution. It remains an open question whether our findings in the frog also apply to other vertebrate species. Taken together, we showed that paraxial mesodermal patterning specifies the sensory part of the LRO, thereby conjoining two embryonic processes that appear mutually exclusive at first glance.

Xenopus laevis were obtained from Nasco (901 Janesville Avenue PO Box 901 Fort Atkinson) and were treated in accordance with German Regulations and laws approved by the Regional Government Stuttgart (A379/12 Zo, “Molekulare Embryologie”, V340/17 ZO and V349/18 ZO, “Xenopus Embryonen in der Forschung”).

A dmrt2 probe (1,481 bp) for WMISH was amplified by RT-PCR using a 5′UTR forward primer 5′TCC​CAC​CAC​TAA​GGG​AAC​TG3′ and fourth exon reverse primer 5′TTTTCAAGATG TGCCTGCTG3′ and cloned into the pGEMT-easy vector. For rescue experiments, full-length dmrt2 (corresponding to NM_001096256.1) was amplified by RT-PCR and cloned into the pCS2+ vector. The following primers were used: Forward 5′ATCGGGATCCTTAGAAATGTATGAAATGAAAGCGCCTGCTGCCCCATCCTCTTCCTCGT3'; Reverse 5′ATCCATCGATGTTACTGACTAGAACGCTTGACTGTTGT TGAGGG3'.

Full-length myf5 in pBSK+ was a gift from V. Gawantka and C. Niehrs (corresponds to NM_001101779.2). For the gain of function experiments, myf5 was cloned into pCS²⁺ by restriction digest using EcoRI. A myf5 rescue construct was generated by PCR using forward 5′ATATCGATAT GGA​AAT​GGT​TGA​CAG​TTG​TCA​CTT​C3′ and reverse 5′ATG​GAA​ATG​GTT​GAC​AGT​TGT CACTTC3′ oligonucleotides.

For mRNAs synthesis, pCS²⁺ expression vectors were linearized by SacII (dmrt2) or NotI (myf5) and transcribed using the Invitrogen mMessage sp6 kit according to user instructions.

A volume of 4 ml was microinjected into the left dorsal marginal region of 4 and 8-cell stage embryos. Bilateral injections were performed for flow analysis. Antisense morpholinos were provided by GeneTools. dmrt2 MO 5′ TGC​CTT​CAT​CTC​GTA​CAT​CTC​CAG​C 3′ and myf5 MO 5′ ACC​ATC​TCC​ATT​CTG​AAT​AGT​GCT​G 3′were injected at a concentration of 1pMol/embryo. dmrt2 and myf5 mRNAs were applied at a concentration of 50–100 ng/μl or 50–60 ng/μl, respectively.

Superficial and deep mesodermal tissue of stage 10.5 embryos were manually dissected and separated in CMFM buffer (Calcium Magnesium Free Medium, 88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO₃, 7.5 mM Tris (pH 7,6); (Sargent et al., 1986). RNA was isolated by phenol-chloroform extraction. For cDNA synthesis and qPCR analysis, the Promega Kit GoTaq 2-Step RT-qPCR System (A6010) was used according to user instructions. Real-time quantitative PCR (RT-qPCR) was carried out in a 96-well plate on the Roche LightCycler System 96. Each sample was conducted in triplicates (technical replicates) and relative expression was calculated by ΔΔCT-method.

Primers used for conventional RT-PCR: dmrt2L_x1 forward 5′TGG​ACT​TTT​CTT​ACC​TAA​CCG​C3′ and dmrt2L_x1 reverse 5′TGA​CTC​CTT​TCC​TAA​GAA​GCA​GT3'. The primers odcL forward 5′TGC​AGA​GCC​TGG​GAG​ATA​CT3′ and odcL reverse: 5′GGC​AGC​AGT​ACA​GAC​AGC​AG3′ served as positive control. Primers for qPCR: dmrt2L_x1 forward 5′CAAAGCCCAGCATC ACAGAG3′ and dmrt2L_x1 reverse 5′TGG​TCC​CCA​GGT​AAG​AAT​CAG3'. Reference genes for qPCR (Mughal et al., 2018): sub1L forward 5′AGC​AGG​AGA​AAT​GAA​GCC​AGG3′, sub1L reverse 5′CCG​ACA​TCT​GCT​CCT​TCA​GT3′ and slc35b1L forward 5′CGC​ATT​TCC​AAA​CAG​GCT​CC3′, slc35b1L reverse 5′CAA​GAA​GTC​CCA​GAG​CTC​GC3'.

SP6 or T7 RNA polymerase (Promega) was used to synthesize Digoxigenin-labeled (Roche) RNA probes from linearized plasmids. tbx6 probe was kindly provided by Hideho Uchiyama. MEMFA was used to fix embryos and processed them following standard protocols. Whole-mount in situ hybridization (WMISH) was carried out according to Belo et al., 1997.

Flow analysis was carried out as described (Schweickert et al., 2007; Tingler et al., 2018).

For immunofluorescence, the monoclonal mouse anti-acetylated α-tubulin antibody (1:700; T6798 Sigma) and a secondary anti-mouse antibody (1:1,000; c2181 Sigma) were used and conducted as described (Tingler et al., 2018).

Comparisons of altered marker gene expression (pitx2, foxj1, myf5, tbx6) were statistically analyzed using one-sided Pearson’s chi-square test in statistical R. Statistical relevance of flow directionality and velocity as well as cilia length was calculated by the Wilcoxon-Match-Pair test (statistical R-3.0.1).