Sequential action of JNK genes establishes the embryonic left-right axis

Vertebrates exhibit external symmetry; however, many internal organs have asymmetric positioning that requires the establishment of a midline left-right axis in early development (Blum and Ott, 2018; Grimes and Burdine, 2017). Disruption to this axis can result in abnormal organ positioning and, when associated with congenital heart malformations, is known as heterotaxy (Kennedy et al., 2007; Shiraishi and Ichikawa, 2012). The embryonic node, a transient ciliated cavity or pit is believed to be essential for the establishment of left-right identity. Within the node, rotation of nodal cilia generates right-to-left fluid flow (nodal flow) (Amack, 2014; Nonaka et al., 1998), leading to asymmetric expression of the TGFβ family member Nodal in the left lateral plate mesoderm (LPM) (Tabin, 2005). Nodal expression is self-promoting (Yamamoto et al., 2003) and concomitantly activates its own inhibitors, in particular Lefty1 in the midline, which prevents Nodal signalling propagating into the right LPM (Meno et al., 1998; Nakamura et al., 2006). Downstream of Nodal, the highly conserved transcription factor Pitx2 (Shiratori et al., 2001) is expressed in the left LPM overlapping multiple organ analgen that undergo asymmetric morphogenesis (Campione et al., 1999). The Pitx2 gene locus generates two distinct mRNAs, Pitx2a and Pitx2c that perform subtly different roles in development (Essner et al., 2000). Left-right axis establishment is well conserved in zebrafish, where the function of the node, or Kupffer's vesicle (KV) (Essner et al., 2005; Kupffer, 1868), is proposed to promote left-sided expression of the Nodal homolog southpaw (spaw) (Long et al., 2003). Zebrafish abdominal organs are asymmetrically positioned, with left-sided positioning of the liver and right-sided positioning of the stomach, pancreas and spleen (Horne-Badovinac et al., 2003). In the heart, the first evidence of morphological left-right identity is manifested as jogging: the extension of the linear heart tube under the left eye (Chen et al., 1997; Smith and Uribe, 2021). Dextral heart looping establishes the asymmetric placement of the single atrium and ventricle.

Within the node, motile cilia are positioned distally on each cell (Nonaka et al., 2005), a patterning governed by the highly conserved Wnt planar cell polarity (PCP) pathway (Hashimoto et al., 2010; Minegishi et al., 2017). This pathway also plays crucial roles in regulating polarised cellular behaviours that drive convergent-extension movements during gastrulation (Roszko et al., 2009). Thus, mutations in classical PCP components result in a striking convergent-extension phenotype, seen in zebrafish as shortening of the antero-posterior axis (Jessen et al., 2002; Park and Moon, 2002). Loss of key components of the PCP pathway also impact node development. Vangl2 plays a role in defining the size of the node and positioning nodal cilia, whereas rock2b establishes their asymmetric antero-posterior arrangement (Borovina et al., 2010; Wang et al., 2011). Proposed downstream effectors of PCP signalling are the highly conserved Jun N-terminal kinases (JNKs) (Boutros et al., 1998; Moriguchi et al., 1999; Riesgo-Escovar et al., 1996): Ser/Thr kinases that are members of the mitogen activated protein kinase (MAPK) superfamily activated through a MAPKKK, MAPKK and MAPK phosphorylation cascade (Davis, 2000). Vertebrates have three JNK genes: Jnk1, Jnk2 and Jnk3 (Gupta et al., 1996), Jnk1 and Jnk2 are ubiquitously expressed, whilst Jnk3 expression is restricted to specific structures (Davis, 2000; Santos-Ledo et al., 2020). Double Jnk1/Jnk2 knockout mouse mutants are embryonic lethal between embryonic day (E) 11 and E12, whereas Jnk1/Jnk3 or Jnk2/Jnk3 double mutants are reported to be healthy, demonstrating redundancy and suggesting that genetic compensation could be active in the JNK gene family, but also questioning a direct role for JNK in the PCP pathway (Kuan et al., 1999). Similarly, our recent studies using zebrafish mutants do not implicate the duplicated jnk1a and jnk1b genes (mapk8a and mapk8b – ZFIN) in either the PCP pathway or left-right axis specification (Santos-Ledo et al., 2020). In contrast, other studies using antisense morpholino oligonucleotides have proposed that jnk1 is required for specifying left-right axis through correct nodal cilia length and suggested jnk1 morphants display defects in heart jogging (Gao et al., 2017). More generally, shorter nodal cilia are widely reported to impact on left-right asymmetry (Gao et al., 2017; Heigwer et al., 2020; Jacinto et al., 2021; Lopes et al., 2010; Neugebauer et al., 2009; Schottenfeld et al., 2007; Yamauchi et al., 2009). Separately, experiments in Xenopus have suggested a cooperative role for JNK in the PCP pathway (Kim and Han, 2005; Yamanaka et al., 2002).

Using zebrafish to investigate the link between node function and organ asymmetry (Smith and Uribe, 2021), we set out to address whether a role exists for members of the JNK family in regulating left-right axis development and whether genetic compensation between JNK family members may obscure PCP functions. Generating stable mutants, we characterised the impact of loss of the four zebrafish JNK genes on nodal cilia development and the subsequent result on organ asymmetry. We identify that jnk1a, jnk1b and jnk2 (mapk9 – ZFIN) function non-redundantly in the embryonic node to specify nodal cilia length and are required for directional nodal flow. We show that compromised KV function following loss of jnk1a, jnk1b and jnk2, although disrupting lateralised expression of spaw, does not result in abnormal organ asymmetry, defining an early, yet dispensable, role for jnk1 and jnk2 in left-right axis establishment. We also identify a novel, later requirement for jnk3 (mapk10 – ZFIN) in restricting pitx2c expression and promoting lateralised endodermal organ placement.

We have previously reported that maternal zygotic (MZ) jnk1a (MZjnk1a), MZjnk1b and MZjnk1a;MZjnk1b zebrafish mutants display no evidence of left-right disturbance with regard to heart development (Santos-Ledo et al., 2020). However, a morpholino-based study suggested a role for jnk1 in regulating nodal cilia length (Gao et al., 2017). Therefore, we set out to characterise the impact of loss of JNK genes on left-right axis development in zebrafish.

We first examined the impact of loss of jnk1 activity on KV size using whole-mount mRNA in situ hybridisation for dand5 (DAN domain family member 5, formerly charon) at the 8-somite stage (ss) (Fig. S1A), showing comparable KV size between wild type, MZjnk1a, MZjnk1b and MZjnk1a;MZjnk1b mutants (Fig. S1B). We next characterised number, distribution and length of cilia within KV at 10 ss using immunohistochemistry (Fig. 1A-E). Although total cilia number and antero-posterior distribution were unaffected in any MZjnk1-null mutants (Fig. 1F,G), both MZjnk1a and MZjnk1b mutants displayed a subtle decrease in the length of nodal cilia (2.6% and 3.1%, respectively) (Fig. 1H). However, in MZjnk1a;MZjnk1b double mutants, there was a 17.6% reduction in cilia length, demonstrating redundancy of these jnk1 paralogues in regulating KV cilial length (Fig. 1H). Surprisingly, despite a relatively minor reduction in cilial length, there were significant reductions in the counter-clockwise nodal flow in both MZjnk1a and MZjnk1b mutant embryos (Fig. 2C,D,F, Movies 1-3), and a greater reduction in KV flow in MZjnk1a;MZjnk1b mutants (Fig. 2E,F, Movie 4), potentially indicating reduction in cilia motility as well as length.

Leftward flow within KV is proposed to drive left-sided expression of spaw (Long et al., 2003; Yamamoto et al., 2003). We therefore examined spaw expression at 12-14 ss in our MZjnk1 mutants (Fig. 3A,B) to determine the effect of altered nodal flow. Despite a marked reduction in KV flow (Fig. 2F), there was no significant increase in isolated right-sided spaw expression (Fig. 3B) but there were increases in the penetrance of bilateral spaw expression in MZjnk1b and MZjnk1a;MZjnk1b mutants, suggesting that loss of jnk1b disrupts KV function (Fig. 3B). We then investigated whether abnormal spaw expression would translate into abnormal organ positioning. Left-sided expression of spaw is required for leftward heart jogging at 1 dpf (day post-fertilisation) (Lenhart et al., 2013; Noël et al., 2013; Smith et al., 2008; Veerkamp et al., 2013) (Fig. 3C) and for the asymmetric movements of the LPM, positioning the liver on the left and the pancreas on the right of the midline (Horne-Badovinac et al., 2003; Yin et al., 2010) (Fig. 3E). In keeping with correct lateralised spaw expression, heart jogging and liver/pancreas positioning was normal in MZjnk1a mutants (Fig. 3D,F). However, despite the increase in bilateral spaw expression in MZjnk1b and MZjnk1a;MZjnk1b mutants (Fig. 3B), there was no disturbance of heart jogging or abdominal organ placement (Fig. 3D,F). To investigate the uncoupling between spaw laterality and organ asymmetry in jnk1 mutants, we examined the expression of the highly conserved Nodal-target gene paired-liked homeodomain 2 (pitx2) in the LPM at 18-19 ss (Fig. 3G). Although we observed predominantly left-sided expression of the pitx2c splice-form in wild type, we were surprised to see high frequencies of absent and bilateral expression (Fig. 3H). Expression of pitx2c in MZjnk1a, MZjnk1b and MZjnk1a;MZjnk1b was comparable with wild-type controls and no increase in bilateral pitx2c expression was observed (Fig. 3H).

Taken together, these data show that jnk1 is required for the specification of cilial length, for nodal flow in KV and for lateralised spaw expression. However, bilateral spaw expression in jnk1 mutants does not translate into bilateral pitx2c expression or into abnormal organ placement, suggesting other factors acting parallel to the KV axis could be functioning to drive asymmetric organ positioning or that other genes function redundantly with jnk1 to establish laterality.

Members of the JNK gene family, particularly Jnk1 and Jnk2 have been suggested to compensate for one another (Kuan et al., 1999). We therefore generated jnk2 mutants using CRISPR-Cas9 genome editing to investigate whether compensation was occurring within KV (Fig. S2A,B). As with MZjnk1 mutants, MZjnk2 mutants were fertile and appear morphologically normal (Fig. 4A). We first characterised the impact of loss of jnk2 on KV structure and function, noting a small increase in KV size associated with increased number of nodal cilia (Fig. S2C, Fig. 4C). Although normally distributed across the KV (Fig. S2D), MZjnk2 mutants had significantly shorter nodal cilia (17.3%) (Fig. 4D), similar to MZjnk1a;MZjnk1b mutants (Fig. 1H). Although MZjnk2 mutants maintain a counter-clockwise nodal flow that is reduced (Fig. 4E,E′, Movie 5), this was small and not in keeping with dramatic changes seen in the MZjnk1 mutants, suggesting cilial motility might be additionally impaired in the MZjnk1 mutants. Confirming a minimal impact on KV function, spaw expression was normal (Fig. 4F), but, surprisingly, there was a significant increase in the proportion of MZjnk2 embryos without pitx2c expression in the LPM (Fig. 4G). However, despite this, heart and abdominal organ placement was normal (Fig. 4H,I).

Having identified a partially overlapping role for jnk2 with jnk1a and jnk1b in KV, we generated MZjnk1a;MZjnk1b;Zjnk2 (Z, zygotic) mutants to examine potential redundancy between jnk1 and jnk2. MZjnk1a;MZjnk1b;Zjnk2 mutants did not display any overt morphological abnormalities, KV size was normal, as was the number and distribution of cilia (Fig. 5A-E). However, there was a 47% reduction in nodal cilial length and a 59% reduction in speed of nodal flow (Fig. 5F-G′) in which injected beads showed no directional movement, demonstrating a loss of KV function (Movie 6). The expression pattern of spaw was greatly disturbed in MZjnk1a;MZjnk1b;Zjnk2 mutants, with 40% exhibiting either right-sided or bilateral expression (Fig. 6A). Despite these severe disturbances, there was no increase in the proportion of embryos that displayed bilateral or right-sided pitx2c expression, but instead there was an increase in the proportion of MZjnk1a;MZjnk1b;Zjnk2 mutants without pitx2c expression in the LPM (Fig. 6B).

Having established that loss of jnk1 and jnk2 has a dramatic impact on KV function, resulting in disrupted spaw expression, we characterised the result of these early defects on organ asymmetry at 72 hpf. Using a combination of myosin, light chain 7, regulatory (myl7) and forkhead box A3 (foxa3) antisense mRNA probes for whole-mount in situ hybridisation, we examined the directionality of heart (Fig. 6C-C″) and gut (Fig. 3E-E″) looping in the same embryo. We did not observe any differences in organ asymmetry in MZjnk1a;MZjnk1b;Zjnk2 mutants compared with their siblings (heterozygotes or wild type for the jnk2 mutation) or a wild-type population (Fig. 6D,E). In summary, significant disruption to the lateralised expression of spaw and pitx2c in MZjnk1a;MZjnk1b;Zjnk2 mutants does not translate to loss of stereotypical asymmetric organ placement. This suggests that other mechanisms, functioning in parallel to nodal flow, can correctly establish organ asymmetry during early development.

To complete our analysis, we also generated jnk3 mutants by CRISPR-Cas9 mutagenesis (Fig. S3A,B). As with all our generated JNK mutants, MZjnk3 mutants are morphologically normal and fertile (Fig. 7A). Similar to MZjnk2 mutants, there was an increase in the number of nodal cilia present in KV of MZjnk3 mutants, but KV size was unaffected (Fig. 7C, Fig. S3C). Cilia length and distribution was normal in MZjnk3 mutants (Fig. 7D, Fig. S3D). Despite normal cilia length, there was a 25% reduction in speed of nodal flow but, similar to other single JNK mutants, directionality was not affected (Fig. 7E,E′, Movie 7), and the function of KV was sufficient to ensure normal expression of spaw in the LPM (Fig. 7F). Strikingly, and in contrast to normal spaw expression, 25% of MZjnk3 mutants showed bilateral pitx2c expression (Fig. 7G). This did not impact on heart jogging (Fig. 7H), but did correlate with the abnormal presence of bilateral liver and pancreatic anlagen in 20% of MZjnk3 embryos (Fig. 7I).

We next generated MZjnk1a;MZjnk1b;Zjnk3 mutants (Fig. S4A) to further examine potential compensation in the JNK gene family. Loss of zygotic jnk3 had no impact on the severity of the MZjnk1a;MZjnk1b phenotype with respect to cilia length or distribution, and the reduction in KV flow was comparable with MZjnk1a;MZjnk1b mutants (Fig. S4B-F′, Movie 8). These findings suggest that jnk3 may act downstream of jnk1 in an epistatic mechanism. We also characterised organ looping in MZjnk1a;MZjnk1b;Zjnk3 mutants and did not observe any significant disturbances to laterality (Fig. S4G,H).

Having established that all JNK genes play a role in generating nodal flow, but to seemingly different extents, we analysed flow patterns within anterior and posterior compartments of KV (Fig. S5). This identified that the main contributing factor to a reduction in flow is disruption to flow in the anterior compartment of KV (Fig. S5A) whereas posterior flow was only significantly affected when both jnk1a and jnk1b were absent (Fig S5B). This analysis also confirmed that MZjnk1a;MZjnk1b;Zjnk2 mutants display the most compromised KV function, particularly in the anterior compartment (Fig. S5A).

Although we wished to examine KV structure and function in MZjnk1a;MZjnk1b;Zjnk2;Zjnk3 mutant embryos, this proved difficult as both jnk2 and jnk3 lie on chromosome 21, and breeding between these alleles could only produce heterozygotes in trans. Meiotic recombination events that would bring jnk2 and jnk3 alleles into cis were extremely rare (<1%) and we were not successful in establishing the line.

Analysis of left-right specification in JNK mutants identified that bilateral expression of spaw is the principal abnormality seen in MZjnk1a;MZjnk1b;Zjnk2 mutants (Fig. 6A), whereas bilateral pitx2c expression is observed in MZjnk3 mutants (Fig. 7H). This suggested that the midline barrier might be dependent on JNK activity. In wild type at 13-14 ss, lefty1 (lft1) expression extends anteriorly from the caudal region of the embryo, setting up a molecular barrier to maintain left-sided spaw expression (Fig. 8A,B). We identified a failure in the propagation of lft1 in MZjnk1a, MZjnk1b and MZjnk1a;MZjnk1b null embryos (Fig. 8C-E,I), with a less severe reduction in the length of the lft1 domain in MZjnk1a mutants, which correlates with the reduced penetrance of bilateral spaw expression (Figs 3B and 8C,I). Surprisingly and despite normal spaw expression in the left LPM (Fig. 4G), MZjnk2 mutants also displayed a failure in anterior propagation of lft1 along the midline, similar to MZjnk1b and MZjnk1a;MZjnk1b mutants (Fig. 8F,I). lft1 propagation is also compromised in MZjnk1a;MZjnk1b;Zjnk2 mutants, comparably to single MZjnk1 or MZjnk2 mutants (Fig. 8G,I). In MZjnk3 mutants, lft1 expression is unaffected (Fig. 8H,I), correlating with normal spaw expression (Fig. 7G). However, this does not account for the bilateral expression of pitx2c in MZjnk3 (Fig. 7H) and may suggest multiple roles for JNK genes, potentially outside KV in the establishment of laterality.

In conclusion, all JNK family members play overlapping and specific roles in establishing the embryonic left-right axis. jnk1a, jnk1b and jnk2 are required for normal nodal cilial length and it appears they may also be important in cilial motility, as jnk1a, jnk1b and jnk3 mutants appear to have much greater reductions in KV flow relative to cilial shortening. For jnk1b, this translates to abnormal bilateral expression of spaw, with an additional effect when jnk2 is also inactivated. This appears to be related to abnormal midline barrier function, as lft1 is caudally restricted in these mutants. There is an increase in the proportion of embryos lacking pitx2c expression in MZjnk1a;MZjnk1b;Zjnk2 mutants, but it seems that the mechanisms that maintain left-sided pitx2c are intact in these mutants as the frequency of bilateral pitx2c expression that might be expected based on spaw expression is not seen in these mutants (compare Fig. 6B with Fig. 6A) and asymmetrical organ placement is maintained (Fig. 6C,D). In contrast jnk3 plays a minor role in KV development and function. Although loss of jnk3 does not affect cilial length, there is a reduction in KV flow. However, despite this, normal spaw expression is established but, surprisingly, there is a marked increase in bilateral pitx2c expression (Fig. 7F,G), even though midline barrier function appears normal. Although development of the heart is unaffected, abdominal organ development is strongly affected, with bilateral liver and pancreatic anlagen in MZjnk3 mutants occurring at a frequency comparable with that of bilateral pitx2c expression (Fig. 7H,I).

We have characterised the roles of the JNK genes in establishing left-right asymmetry and show, as previously suggested, that jnk1 promotes nodal cilial length (Gao et al., 2017), but that this role is shared by jnk2; together, both jnk1 and jnk2 are required for normal KV function (Fig. 5, Fig. S5). How JNK genes co-ordinate cilia length remains an unresolved issue, as numerous proteins regulate cilia length at the level of post-translational modification. One interesting candidate, the JNK-interacting protein 1 (JIP1), is a regulator of JNK signalling and a cargo protein for the microtubular motor kinesin 1, which is required for axon elongation (Dajas-Bailador et al., 2008). Alternatively, jnk1 and/or jnk2 may regulate transcription factors involved in ciliogenesis, such as rfx3 or foxj1 (Alten et al., 2012; Bonnafe et al., 2004) in response to developmental signals. In zebrafish, upstream of foxj1a, Notch signalling is crucial in regulating nodal cilia length: overactivity increases cilia length, whereas deltaD mutants have shorter cilia (40% reduction), a dramatic reduction in nodal flow, and disruption to spaw and pitx2 expression (Lopes et al., 2010). foxj1 is also regulated by FGF signalling, with fgfr4 morphants displaying a 26% reduction in cilia length (Gao et al., 2017; Neugebauer et al., 2009). The micro-RNA miR103/107 appears to regulate ciliogenesis downstream of foxj1a, acting on a cohort of genes, including those involved in cilia assembly (Heigwer et al., 2020). miR103/107 morphants also display shorter cilia (20% reduction) and disrupted asymmetric organ positioning (Heigwer et al., 2020). Importantly, where nodal cilia are either excessively long or significantly shorter, expression of spaw and pitx2 are disrupted, suggesting a window of optimum nodal cilial length (Lopes et al., 2010). Despite a comparably dramatic reduction in nodal cilia length in MZjnk1a;MZjnk1b;Zjnk2 mutants (47%, Fig. 5) and loss of KV function (Movie 6), organ looping is normal, suggesting that nodal cilia may function redundantly to establish organ laterality (Fig. S6). Of note, whereas fgfr1 morphants have shortened nodal cilia, they also display a curved body axis, which is a more overt readout of cilial defects, that we did not observe in any JNK mutant (Figs 4, 5 and 7, Fig. S4) (Neugebauer et al., 2009; Santos-Ledo et al., 2020). The absence of such obvious morphological defects in JNK mutants suggests different programmes may generate distinct classes of cilia and that the role of JNK genes may be specific to nodal or, more generally, motile cilia.

Nodal cilia length, orientation and position in the node are crucial for the generation of nodal flow: the proposed symmetry-breaking event that establishes left-sided Nodal (Amack, 2014; Antic et al., 2010; Borovina et al., 2010; Minegishi et al., 2017; Nonaka et al., 1998, 2005; Song et al., 2010; Tabin, 2005). Although we have shown that MZjnk1a;MZjnk1b;Zjnk2 mutants have no directional flow (Movie 6), our other mutants display reductions in nodal flow, yet directionality is maintained (Movies 1-5, 7-8). This contrasts with MZvangl2 zebrafish and Vangl2Vangl1 mutant mice in which normal cilia are incorrectly positioned, resulting in irregular flow and laterality disturbances (Borovina et al., 2010; Song et al., 2010). MZmyo1d zebrafish mutants display similar disordered nodal flow patterns, together with more gross KV defects and ultimately disruptions to left-right asymmetry (Juan et al., 2018). Loss of a single copy of vangl2 in MZmyo1d mutants partially rescues nodal flow, spaw expression and cardiac morphogenesis in a model whereby myo1d and vangl2 interact to position nodal cilia (Juan et al., 2018). This KV-specific interaction may further support the possibility that different programmes exist to generate and position nodal cilia, and that the JNK gene family is required for specification of cilia length.

Two possibilities have been suggested for the nodal flow-generated signal: morphogen flow or the two-cilia model, where nodal flow activates mechanosensory cilia (Tabin and Vogan, 2003). In support of the two-cilia model, the calcium channel pkd2 is proposed to facilitate left-sided calcium signalling that is at least partly required for heart jogging (Yuan et al., 2015). Importantly, pkd2 functions both autonomously and non-autonomously in KV development (Jacinto et al., 2021). This highlights candidly that, although the most overt phenotype of loss of jnk1 and/or jnk2 activity may be shorter nodal cilia, JNK genes may act at multiple stages of KV function, such as generation of cilia motility or sensation of nodal flow, or may also function outside KV in other tissues. Interestingly, our analysis in MZjnk3 mutants (and also MZjnk2 mutants), has shown that reduction of KV flow to less than 25% of wild-type levels is still sufficient to maintain left-sided spaw expression (Fig. 7). This indicates that either there is a threshold of KV flow required for KV function to establish the left signal and that this threshold is much lower than that generated in wild-type embryos, or that left-right axis establishment has a high degree of redundancy with other mechanisms, which may lie outside KV (Fig. S6). There is some evidence for a minimal flow threshold within the Node: as few as two motile cilia have been shown to be sufficient to generate a left-sided Nodal signal in mice (Shinohara et al., 2012), whereas the minimal number of functional cilia in zebrafish required for correct organ placement has been suggested to be 29 out of ∼200 (Sampaio et al., 2014). The shortening of the lft1 expression in jnk1 and jnk2 mutants further supports that these genes may act not only in the KV, but also potentially downstream of spaw (Smith et al., 2011), possibly through a cilia-dependent mechanism at the midline (Shylo et al., 2018 preprint).

Our work supports an emerging viewpoint that certain organs possess intrinsic mechanisms that drive asymmetric morphogenesis. Loss of spaw does not lead to a loss of heart looping and a dextral bias is still maintained (Noël et al., 2013). Similarly, in mice, Nodal is dispensable for the morphogenesis of the heart tube itself (Desgrange et al., 2020; Le Garrec et al., 2017). This study, together with others, supports potential organ-specific mechanisms by demonstrating a disconnection between mesoderm and endodermal laterality (Fig. 7, Fig. S6) (Hochgreb-Hägele et al., 2013; Lopes et al., 2010; Noël et al., 2013; Sampaio et al., 2014). We have shown that MZjnk3 embryos initially establish a left-right axis and a robust midline, and that heart jogging is normal; yet a significant proportion of embryos develop bilateral guts (Figs 7 and 8), which is a similar organ laterality phenotype to that observed in deltaD and lamb1a mutants (Hochgreb-Hägele et al., 2013; Lopes et al., 2010).

Together with left-right asymmetry, deposition of ECM components, such as laminins, and their turnover by matrix metalloproteinases, which is regulated by the transcription factor hand2, are necessary for asymmetric gut looping (Hochgreb-Hägele et al., 2013; Yin et al., 2010). hand2 mutants have bilateral guts arising from failure in the necessary asymmetric cell rearrangements of the LPM (Yin et al., 2010). It is therefore tempting to speculate that jnk3 may have a role in regulating hand2 expression, which, coupled with our previous report of a role for jnk1a in regulating hand2 expression in cardiac progenitors (Santos-Ledo et al., 2020), may suggest JNK family members have organ-specific or potentially germ layer-specific roles in regulating hand2 activity. Additionally, the single JNK gene in Drosophila (basket, bsk) is required for anterior midgut looping (Taniguchi et al., 2007), which may suggest a partially conserved role. A further explanation for the MZjnk3 phenotype is suggested by the interesting observation that the severity in reduction of nodal flow in the KV appears to be linked to the impact on endodermal laterality (Sampaio et al., 2014). This would suggest that a KV-dependent spaw-independent mechanism functions to promote gut looping morphogenesis.

Bilateral pitx2c expression in MZjnk3 mutants, despite normal spaw and lft1 expression, is reminiscent of exposure of embryos to retinoic acid (Tsukui et al., 1999), possibly supporting crosstalk between left-right and antero-posterior axis establishment (Kawakami et al., 2005). Our data could also suggest that there may be an alternative, late-acting midline barrier that is defective in MZjnk3 mutants. Another possibility may be that in MZjnk3 mutants, the right LPM is receptive to nkx2.5 binding of the pitx2 left side enhancer (ASE) (Shiratori et al., 2001) independently of right-sided spaw activity, allowing maintenance of bilateral pitx2c. However, the specific role that pitx2 plays in zebrafish is unclear, as loss of pitx2 does not impact heart or gut looping (Ji et al., 2016). Furthermore, Pitx2 mutant mice have been reported to have initially normal organ looping but develop more-complex heart defects that may be independent of global left-right asymmetry (Ai et al., 2006; Gage et al., 1999; Lin et al., 1999; Lu et al., 1999). These observations may suggest that the endodermal phenotype in MZjnk3 is independent of pitx2c. A further characterisation of brain asymmetry may also shed light on whether there is a potentially ectoderm-specific mechanism that promotes asymmetry of the habenulae (Gamse et al., 2003) or whether this is tightly coupled to asymmetric spaw expression in the embryo.

From these observations, it is clear that multiple pathways function during left-right axis establishment, although their hierarchy remains unresolved, exemplified in zebrafish by a pitx2-independent KV function-dependent cascade, for which elovl6 is a known gene (Ji et al., 2016). Several other mechanisms, independent of nodal flow, have been proposed to establish left-right asymmetry (Levin, 2003; Tabin, 2005), possibly as early as the first cell division. Injection of dextran into one cell at the two-cell stage in zebrafish results in a lineage labelling of one side of the embryo (Noël et al., 2013), whereas, in Xenopus, abolishment of asymmetric localisation of 14-3-3E during the first cell division results in laterality disturbances (Bunney et al., 2003). Second, a mutation in atp1a1a.1, a component of the Na⁺/K⁺ transporter, presents with laterality defects, but KV structure and function appear normal (Ellertsdottir et al., 2006). Inhibition of the Na⁺/K⁺ pump with the chemical inhibitor Ouabain between 3 and 11 hpf in zebrafish also results in laterality defects (Ellertsdottir et al., 2006), and may define a critical time window because later treatment, between 10 and 13 hpf, results in KV morphogenesis defects (Juan et al., 2018), possibly suggesting that early activity of this ion pump is important for left-right asymmetry at least functioning in parallel with Nodal signalling (Fig. S6, green). Finally, JNK genes may also function outside KV to repress a right-sided factor, possibly of the Snail family (Isaac et al., 1997). However, the existence and role of such a factor remains controversial (Castroviejo et al., 2020; Ocaña et al., 2017; Tessadori et al., 2020). Organ asymmetry remains highly stereotypical in MZjnk1a;MZjnk1b;Zjnk2 mutants, despite a loss of nodal flow (Movie 6) and disruption to spaw expression (Fig. 6A), suggesting that nodal flow is not the crucial symmetry-breaking event because other mechanisms recover initially defective cues in the early embryo (Fig. S6). Further characterisation of left-right patterning in the heart prior to jogging using markers such as lft2 and bmp4 (Chocron et al., 2007; Smith et al., 2008; Veerkamp et al., 2013) may uncover whether disruption to asymmetric spaw is recovered.

Examination of a repertoire of vertebrate model organisms highlights that nodal- and cilia-independent mechanisms may be commonplace but not yet fully elucidated (Hamada and Tam, 2020). Whereas zebrafish, Xenopus and mouse possess a ciliated LRO (left-right organiser), there are no luminally positioned motile cilia in the chick LRO (Hensen's node) and loss of C2Cd3, a gene essential for ciliogenesis, does not result in disruption to laterality (Chang et al., 2014), instead asymmetric cell movements have been proposed to break bilateral symmetry (Cui et al., 2009; Gros et al., 2009). Furthermore, the observations regarding nodal cilia have been extended to pig embryos, with the suggestion that the node is not large enough to generate fluid flow or that it may not even exist (Gros et al., 2009). Comparisons with non-vertebrate model organisms may prove to be insightful, exemplified by the role of myo1D in establishment of laterality in both Drosophila and zebrafish (Juan et al., 2018).

Both genetic and cell-based studies have suggested that JNK family members are key components of the PCP pathway (Boutros et al., 1998; Kim and Han, 2005). However, many of these have examined only a single member of the JNK family, used pharmacological methods or injected antisense-morpholino oligonucleotides, rather characterising stable mutant lines. Although unable to examine total JNK nulls due to our jnk2 and jnk3 alleles being generated in trans, this and our previous study (Santos-Ledo et al., 2020) have failed to reveal even a mild convergent-extension phenotype in JNK mutants, the hallmark of PCP signalling disturbances (Jessen et al., 2002; Park and Moon, 2002; Solnica-Krezel et al., 1996). Furthermore, we observed no PCP-related node abnormalities, such as mal-positioning or mal-distribution of cilia (Borovina et al., 2010; Wang et al., 2011). Thus, it is possible that JNK activity is not a definitive requirement for PCP signalling. Supporting this, combinatorial JNK mouse mutants, or treatment of embryos heterozygous for PCP mutations with a JNK inhibitor, do not display defects in convergent extension (Kuan et al., 1999; Ybot-Gonzalez et al., 2007). Furthermore, although suppression of JNK activity in Drosophila is able to rescue defective PCP signalling, loss of JNK activity alone produces only a subtle PCP phenotype in the eye (Boutros et al., 1998; Paricio et al., 1999) and in Xenopus explants, JNK1 is not sufficient to regulate Wnt5a-driven convergent-extension (Yamanaka et al., 2002). However, in many of these experimental contexts, removal of kinase activity upstream of JNK signalling, such as misshapen/TNIK (Köppen et al., 2006; Paricio et al., 1999) or MKK7 (Yamanaka et al., 2002), does result in classical PCP phenotypes. In summary, this suggests that, although JNK activity is a readout of active PCP signalling (Boutros et al., 1998; Li et al., 1999; Moriguchi et al., 1999), the activity of JNK itself may not be crucial for the generation of planar cell polarity and that upstream factors may be more important (Paricio et al., 1999; Yamanaka et al., 2002) or might show redundancy with JNK genes.

Establishment of laterality is proposed to be a sequential process, through an initial symmetry-breaking event commonly thought to be at the node, which is amplified to a Nodal homolog during early development and relayed subsequently through genes such as Pitx2 to the organ anlagen to drive asymmetric morphogenesis. In this study, we have characterised the role of the JNK family members in the formation, function and downstream effects of nodal cilia in zebrafish. We have shown that jnk1/jnk2 function to specify nodal cilia length, promote nodal flow and establish the lft1 midline barrier, but that this is dispensable for the correct establishment of organ asymmetry and suggests other node-independent mechanisms are able to recover this early phenotype (Fig. S6). We have also identified a novel later requirement for jnk3 to maintain lateralised pitx2c expression that may either directly or indirectly promote asymmetric morphogenesis of the endoderm. Together, this demonstrates that multiple mechanisms, both JNK dependent and independent act redundantly at different stages during vertebrate axis establishment, ensuring robust asymmetric organ morphogenesis.

The following previously described zebrafish lines used in this study were wild type (AB), jnk1aⁿ² (mapk8a) and jnk1bⁿ³ (mapk8b) (Santos-Ledo et al., 2020). All procedures and experimental protocols were carried out in accordance with UK Home Office and Newcastle University (Project Licence P25F4F0F4). Embryos were obtained from natural pairwise mating and reared in standard conditions. Embryos were raised in Embryo Medium (E3) at 28.5°C and staged according to Kimmel et al. (1995).

jnk2 (mapk9, ENSDARG00000077364, ZDB-GENE-091117-28) and jnk3 (mapk10, ENSDARG00000102730, ZDB-GENE-051120-117) zebrafish mutants were generated using CRISPR-Cas9-mediated mutagenesis (Hwang et al., 2013). gRNAs were identified using CrisprScan (Moreno-Mateos et al., 2015). gRNAs and Cas9 RNA were synthesized according to previously published protocols (Gagnon et al., 2014) and injected at the one-cell stage into embryos obtained from an in-cross of wild-type (AB) adults (F0 generation). The 20 bp deletion in Exon 3 of jnk2 (allele designation mapk9ⁿ⁴) was generated using the single gRNA 5′-ATTTAGGTGACACTATAGCAATCTTCACATCCAGGACGTTTTAGAGCTAGAAATAGCAAG-3′ and genotyped using forward (5′-TTAAAGGGGATTGAGGAACAAA-3′) and reverse (5′-GTTAAGGGGACGTACGTTCTTG-3′) primers in a standard GoTaq G2 (Promega M784B) PCR with an annealing temperature of 58°C and 34 cycles. The mutation destroys a DdeI restriction enzyme site (New England Biolabs R1075). The 4 bp deletion in exon 5 of jnk3 (allele designation mapk10ⁿ⁵) was generated using the single gRNA 5′-ATTTAGGTGACACTATAGGCCACATTTCTGTCCAGGAGTTTTAGAGCTAGAAATAGCAAG-3′ and genotyped using forward (5′-AATTCCCATCTTGTGTTTCAGG-3′) and reverse (5′-TTTTGGGGAAAACCTGACTCTAA-3′) primers in a standard GoTaq G2 PCR with an annealing temperature of 58°C and 34 cycles. The mutation destroys a BstNI restriction enzyme site (New England Biolabs R1068). F0 adults identified as carrying desired mutations were outcrossed to wild-type animals prior to generation of homozygous mutant lines. Where lines were derived from multiple rounds of in-crossing, a minimum of six different breeding pairs were used to establish the next generation. Reduction in mRNA levels in MZjnk2 or MZjnk3 mutants was confirmed by RT-PCR using primers and protocols previously described (Santos-Ledo et al., 2020).

Fixed embryos were serially rehydrated from 100% methanol into PBST [0.2% Tween-20 (Sigma P2287) in 1× PBS (Oxoid BR0014G)], washed three times in 0.2% Triton-X (Sigma T8787) in 1× PBS (PBSTx) and then incubated in blocking buffer for 1 h at room temperature [5% sheep serum (Gibco, 16070-096), 10 mg/ml bovine serum albumin (A2153, Sigma) and 1% DMSO (D4540, Sigma) in PBSTx] for 1 h at room temperature before an overnight incubation at 4°C with gentle agitation with the following primary antibodies: mouse anti-acetylated tubulin (T6793, Sigma, 1:500) and rabbit anti-PKC ζ (aPKC) (sc-216, Santa Cruz, 1:400) in blocking buffer (Amack et al., 2007). The next day, embryos were rinsed extensively in PBSTx, before a further overnight incubation at 4°C with gentle agitation with the following secondaries: Alexa488-conjugated donkey anti-mouse (Invitrogen A21202, 1:200) and Alexa647-conjugated donkey anti-rabbit (Invitrogen A31573, 1:200) in blocking buffer. Embryos were washed extensively on day three before dissection or embedding. Representative images of Kupffer's vesicle were taken using a Nikon A1 inverted confocal using a 40× objective.

Embryos were embedded in 1.5% low melting agarose (ThermoFisher Scientific R0801) in PBS and imaged in an Axiotome microscope using a 40× objective. An image stack was recorded and used for quantification. Stacks were opened in Fiji where the number of cilia and their position relative to the anterior and posterior parts of Kupffer's vesicle were quantified (Wang et al., 2011). The xy coordinates of all cilia were recorded in every slice. The total length of the cilia was obtained as previously described (Dummer et al., 2016). The xy coordinates were translated to µm assuming the following equivalencies: 1 pixel=0.1623 µm, 1 slice=0.4 µm.

Analysis of nodal flow was carried out as previously described (Wang et al., 2013), using TransFluoSpheres beads of 1 µm diameter (Invitrogen T8880). High-speed movies were acquired using an Axiotome microscope with either a 40× or 60× objective attached to an IMPERX high speed camera with Video Savant 4.0 software (Multipix, UK). The frame rate of acquired movies was 207 frames per second. Stacks were imported into Fiji (Schindelin et al., 2012), the notochord oriented to the top of the image and the quadrants indicated. Every fifth frame representing ∼0.0125 s were used to track bead speed by the Manual Tracking plug-in. Beads were tracked that remained in the focal plane of the movie for ≥50 frames. Average speeds of beads were then calculated for the duration of tracking (Wang et al., 2011, 2013).

Embryos older than 24 hpf for use in in situ hybridisation were transferred into E3 medium containing 0.003% 1-phenyl 2-thiourea (PTU, Sigma P7629) to inhibit pigment formation and aid imaging. Embryos were fixed overnight in 4% paraformaldehyde (PFA, P6148, Sigma) in 1× phosphate-buffered saline, washed three times in PBST for 5 min at room temperature and then serially washed into 100% methanol for long-term storage at −20°C. Whole-mount in situ hybridisation was carried out according to standard protocols (Thisse and Thisse, 2008). The following, previously published probes were used: dand5 (Hashimoto et al., 2004), spaw (Long et al., 2003), myl7 (Yelon et al., 1999), foxa3 (Odenthal and Nüsslein-Volhard, 1998), lft1 (Bisgrove et al., 1999) and pitx2c (Yan et al., 1999).

In situ hybridisation images were imaged laterally using a Zeiss AxioPlan. Images were pooled and made unidentifiable using Image J Blind_Analysis plug-in as previously described (Derrick et al., 2021) and imported into Fiji (Schindelin et al., 2012). Using the Freehand Line tool, the length of the domain with continuous expression was measured from the caudal tip to the anterior extreme. Embryos without any visible lft1 expression were discounted from analysis.

No statistical tests were used to formally predetermine sample size, but the number of biological replicates were based on published studies and defined a priori. For all experiments where n was based on the individual embryo (e.g. KV flow), they were derived from at least three different clutches from distinct breeding pairs. For population analysis of laterality markers or organ asymmetry, n was represented by the clutch, and each clutch was derived from three distinct breeding pairs. Statistical tests were carried out in Prism (Graphpad).

The authors gratefully acknowledge the BioImaging Unit and Aquarium Technical team at Newcastle University for their support and assistance in this work.