Notch signalling regulates left-right asymmetry through ciliary length control

Internal body left-right (LR) asymmetry is a common trait among vertebrates. In contrast to the symmetric external body plan, the heart, digestive organs and parts of the brain display highly conserved asymmetric orientations that are essential for their correct functions. When the normal asymmetric distribution of the internal organs is compromised, human laterality disorders arise such as heterotaxy and situs inversus (Fliegauf et al., 2007; Gerdes et al., 2009).

The establishment of LR asymmetries in vertebrates occurs during early stages of embryonic development through a complex process involving epigenetic and genetic mechanisms. This culminates in the activation of a conserved Nodal signalling cascade that starts at the node and is then transferred to the left lateral plate mesoderm (LPM) (Hamada, 2008). Although the timing and mechanism responsible for breaking the initial embryonic symmetry might differ between species, it is clear that a directional fluid flow generated by motile cilia present in the node plays a fundamental role in LR determination, most likely by creating a LR information bias in the node (Levin and Palmer, 2007). Several lines of evidence support this idea, namely absent cilia in the mouse (Nonaka et al., 1998), shorter cilia in zebrafish mutants and morphants (Kramer-Zucker et al., 2005; Neugebauer et al., 2009), and immotile cilia in mouse, zebrafish and medaka (Essner et al., 2005; Hojo et al., 2007; Supp et al., 1999). All these cilia defects result in altered fluid flow in the node/Kupffer's vesicle (KV) and ultimately in LR defects. This implies that proper cilia length regulation is crucial for the establishment of LR asymmetries.

Expression of Nodal and of its antagonist Cerl2 (Dand5 – Mouse Genome Informatics) in the mouse node starts symmetrically (Collignon et al., 1996; Marques et al., 2004). However, the mechanism that produces asymmetrical expression of Nodal and Cerl2 on the left and right sides of the node, and its potential link to the leftward nodal flow, is not fully understood. In zebrafish, there is no evidence for asymmetric expression of spaw and of its negative regulator charon in KV. Nevertheless, it has been suggested that an asymmetric Charon function inhibits the transfer of spaw to the right LPM (Hashimoto et al., 2004). To date, the relationship between the fluid flow inside KV and the asymmetric function of Charon in medaka and its homologue Coco in Xenopus have been studied (Hojo et al., 2007; Schweickert et al., 2010), but this relationship has not been explored in zebrafish.

Notch signalling has been implicated in LR patterning in vertebrates, largely through the activation of Nodal expression in the murine node (Krebs et al., 2003; Przemeck et al., 2003; Raya et al., 2003; Takeuchi et al., 2007). In zebrafish, Notch signalling seems to activate charon expression in KV (Gourronc et al., 2007).

Apart from providing the first genetic evidence in zebrafish for the activation of charon expression in KV, we demonstrate that canonical Notch signalling, through the DeltaD ligand, is a key regulator of ciliary length at the KV. We show that reduction of Notch signalling produces shorter cilia, whereas hyperactivation of Notch produces longer cilia; thus, providing new insights into how cilia length is modulated. We find that the correct regulation of cilia length has a strong impact on the directionality and speed of fluid flow inside KV, and we were able to show that this biomechanical event correlates with asymmetric transcription of charon. Altogether, our results reveal a new and multistep role for the Notch signalling pathway that couples two fundamental processes in left-right determination: transcription of charon and ciliary length regulation that then culminates in breaking the symmetry of charon expression.

AB wild type and mutant (notch1a/des^Tp37, deltaC/bea^tit446, deltaD/aei^tr233) embryos were staged according to Kimmel et al. (Kimmel et al., 1995).

deltaD-MO as described previously (Holley et al., 2000) was injected from 0.5-1.0 mM into 512-cell stage wild-type embryos. The MO that blocks translation of Su(H)1 and Su(H)2 was used as previously described (Echeverri and Oates, 2007). Notch-intracellular and full-length deltaD constructs (Takke and Campos-Ortega, 1999; Takke et al., 1999) were injected at the one-cell stage at a concentration of 100 pg and 350 pg, respectively. foxj1a and GFP full-length cDNAs were independently cloned into pCS2⁺ and in vitro transcribed using mMESSAGE mMACHINE kit (Ambion). The foxj1a and GFP mRNAs were injected at a concentration of 100 pg. Embryos were left to develop at 28°C until the desired stage and then fixed in 4% PFA at 4°C overnight.

Whole-mount in situ hybridization was performed as described previously (Thisse and Thisse, 2008). Digoxigenin RNA probes were synthesized from DNA templates of spaw (Hashimoto et al., 2004), pitx2 (Essner et al., 2005), lefty2 (Bisgrove et al., 2005), foxa3 (Monteiro et al., 2008), ntl (Amack and Yost, 2004), shh (Essner et al., 2005), charon (Hashimoto et al., 2004), deltaD (Takke and Campos-Ortega, 1999), sox17 (Amack and Yost, 2004) and foxj1a (Neugebauer et al., 2009; Yu et al., 2008).

We followed the methods described previously (Neugebauer et al., 2009) for immunostaining. In some cases, we also used TOPRO-3 (1:1000; Invitrogen) diluted in blocking solution.

Flat-mounted embryos were examined with a Zeiss LSM 510 Meta confocal microscope. Three-colour confocal z-series images were acquired using sequential laser excitation, converted into a single plane projection and analyzed using ImageJ software (LSM Reader). The length of all cilia in each KV was measured using ImageJ Segmented Line Tool.

We used the confocal microscope Leica SP5 under the resonance mode, which allows imaging at 8000 lines per second, to acquire images of ciliary beating and debris movements inside KV of live embryos mounted in 1% low-melting agarose (Sigma). High-speed films of the debris were analyzed using ImageJ and the velocity was measured by tracing the debris frame by frame. The orientation and speed of the debris were calculated using Excel VBA, and vectors were drawn using the Gnuplot software and Gimp.

The total number of cilia and their distribution within KV were determined from confocal z-stacks as previously described (Kreiling et al. 2007).

In order to study the impact of Notch signalling pathway on crucial steps for LR patterning, we used several mutants and reagents. We started by evaluating the asymmetric information in the LPM using the left-identity molecular markers spaw and pitx2. The nodal-related gene spaw is the earliest molecule to be asymmetrically detected in the left LPM and is known to induce the expression of the transcription factor pitx2 (Hamada et al., 2002; Long et al., 2003).

We show that mutants in the Notch transmembrane receptor, des/notch1a^–/–, lose their laterality identity by becoming largely bilateral (Fig. 1A,B). Upon binding to its ligand, the intracellular domain of Notch (NICD) is cleaved and translocates into the nucleus. In the nucleus, NICD binds to Suppressor-of-Hairless [Su(H)] thereby changing its repressor status to an activator (Fior and Henrique, 2009). In zebrafish, two Su(H) gene paralogs (rbpja and rbpjb – Zebrafish Information Network) can be knocked down using a single antisense morpholino that blocks the translation of both [Su(H)1+2-MO] (Echeverri and Oates, 2007). Consistent with our results, Su(H)1+2 morphants also showed lack of laterality identity (Fig. 1A,B). To directly induce the Notch pathway, we injected the mRNA encoding NICD and, similar to what has been described previously (Raya et al., 2003), we again found a loss in laterality (Fig. 1A,B). These results indicate that canonical Notch signalling is required to define laterality in the LPM.

In order to assess which Notch ligand was the most relevant in LR establishment, the expression of the left-sided markers was analysed in the null mutants aei/deltaD^–/– and bea/deltaC^–/–. In clear contrast to bea^–/–, aei^–/– mutants have a strong pitx2 laterality phenotype that can be rescued by NICD (Fig. 1B). Thus, we identified DeltaD as the key Notch ligand for LR asymmetry in zebrafish and we focus our analysis on this null mutant. We find that the bilateral expression of pitx2 does not always follow the expression of spaw (compare Fig. 1A with 1B). A lack of a precise correlation between the expression of nodal and pitx2 in the LPM was already described in the context of delta-like 1 (Dll1) mouse mutants (Krebs et al., 2003; Przemeck et al., 2003; Raya et al., 2003). Altogether, these results show that Notch signalling regulates pitx2 expression in a Nodal/Spaw independent manner.

The axial midline is required to maintain LR asymmetry by acting as a midline barrier (Chin et al., 2000; Hamada et al., 2002). We checked whether the LR defects that arise upon Notch signalling perturbation were due to lack of a midline. We found that the expression pattern of both shh and ntl markers was normal in all the experimental conditions (see Fig. S1A-N in the supplementary material), with the exception of ntl downregulation in NICD-injected embryos (see Fig. S1N in the supplementary material), in agreement with what has been described previously (Latimer and Appel, 2006). Therefore, LR defects arising from reduced Notch signalling are not likely to be due to an absent midline.

In aei^–/– mutants, spaw expression is only mildly randomized (Fig. 1A), but the expression scored as being left-sided (Fig. 1A) shows three distinct patterns that have been quantified in detail in Fig. 1C. Namely, we could detect spaw expression in a left anterior domain (Fig. 1C,D,D′), in a left anterior and posterior domains (Fig. 1C,F,F′), and in a left anterior domain with bilateral posterior domains with competition (Fig. 1C,G). Although, a posterior downregulation of spaw occurs at the 23-somite stage as part of its normal anteroposterior (AP) expression pattern (Long et al., 2003), our analysis using an aei heterozygous cross, clearly shows that this spaw downregulation occurs prematurely at the 16-somite stage in 67% of the homozygous embryos (Fig. 1C,D,D′). In agreement with this posterior downregulation of spaw, we observe that lefty1, which is known to be induced by Spaw (Long et al., 2003; Wang and Yost, 2008), has also an abnormal expression gap in the corresponding midline domain (Fig. 1E). These results suggest that the maintenance of spaw expression in the posterior left LPM and possibly lefty1 in the midline domain require DeltaD function.

Next, we evaluated the impact that the abnormal expression of spaw and pitx2 seen in aei^–/– mutants has on heart and gut positioning. We scored aei^–/– mutants for heart asymmetry in vivo and made use of the prospective heart marker lefty2 and of the gut marker foxa3 in fixed embryos. We show that in aei^–/– mutants, gut looping is randomized (62.3% loops to the left, 18.8% failed loops and 18.8% loops to the right; n=85) (Fig. 1H-L). By contrast, heart jogging is not significantly affected when evaluated by lefty2 expression at 26-somite stage (99% on the left; n=112) or when judged by direct observation at 30 hpf (90.7% on the left; n=65) (Fig. 1M,N). These results indicate the uncoupling of laterality of an anterior organ, such as the heart, and of more posterior organs, such as the liver and pancreas. The heart is the first organ to be asymmetrically positioned in zebrafish, and it seems that its asymmetry may arise by a mechanism different from that governing other viscera, as suggested by others (Chin et al., 2000; Lin and Xu, 2009).

Our data indicate that the prospective heart domain is unaffected in the aei^–/– mutants and thus provide an explanation for why heart laterality is not disturbed. Namely, the anterior left-sided LPM expression of spaw is unaffected and pitx2 does not reach the anterior-most domain when expressed on the right LPM. We propose that the premature absence of spaw expression in the posterior LPM, coupled with a bilateral pitx2 expression in this region could underlie the gut laterality phenotype observed in 40% of aei^–/– mutants.

Increasing evidence shows that a properly formed node/KV with a directional strong fluid flow driven by the beating of motile cilia is essential for LR determination (Amack et al., 2007; Hamada, 2008). We therefore investigated whether the laterality problems observed in aei^–/– mutants were due to KV morphogenesis and/or ciliogenesis defects.

To test KV integrity, we performed immunostaining with ZO-1, a tight junction marker and aPKC an apical epithelial marker (Neugebauer et al., 2009) and observed live embryos with high resolution DIC microscopy. The three approaches revealed that KV structure is normal both in aei^–/– mutants and in deltaD overexpressing embryos (Fig. 2A-I). We thus conclude that KV morphogenesis and polarity are not affected by downregulation or upregulation of Notch signalling.

To investigate the role of DeltaD in cilia formation, we labelled the KV cilia with an anti-acetylated α-tubulin antibody. We found that cilia length is significantly reduced in aei^–/– mutants [Fig. 2J,K,M; 2.25±0.62 μm (s.d.m.); n=20 embryos; P<4.3×10^–9) when compared with cilia length in wild-type embryos (Fig. 2J,K,L; 3.73±0.76 μm (s.d.m.); n=27 embryos). Strikingly, we also found that hyperactivation of Notch signalling significantly increased the length of cilia in KV either by injection of NICD [Fig. 2J; 5.28±1.23 μm (s.d.m.); n=19 embryos; P<3.5×10⁴] or deltaD [Fig. 2J,K,N; 6.78±1.18 μm (s.d.m.); n=11 embryos; P<1.5×10^–4]. Additionally, we asked whether cilia motility was also affected in aei^–/– mutants. We filmed the KV cilia in live aei^–/– embryos and showed that, despite being shorter, cilia are still motile (see Movie 2 in the supplementary material). These results strongly suggest that Notch signalling, through DeltaD, controls ciliary length in KV.

deltaD is expressed in the KV (Fig. 2O,P) and also in its precursors at the bud stage (10 hpf), in the somites, presomitic mesoderm and tail bud throughout segmentation stages (Hsiao et al., 2007). Therefore, we decided to test whether the short cilia phenotype observed in aei^–/– mutants is caused by the lack of DeltaD function specifically in the KV precursors or in KV cells. We targeted the dorsal forerunner cells (DFC), which are the KV precursors (Essner et al., 2005), by co-injecting the lineage tracer fluorescein and a deltaD-MO at the midblastula stage. In KV cells where fluorescein (and therefore deltaD-MO) was successfully detected, we observed a reduction in cilia length comparable with the one quantified in aei^–/– mutants [Fig. 2R; 2.25 ±0.6 μm (s.d.m.); n=29 cells; P=0.3) and different from the one quantified in KV cells that were not successfully targeted [Fig. 2Q; 3.6±0.6 μm (s.d.m.); n=21 cells; P<1.6×10^–7]. This experiment indicates that DeltaD autonomously regulates cilia length at the level of the KV precursors or the KV cells and not in the somites, presomitic mesoderm and tail bud. In agreement, we found that in embryos where KV precursors were successfully targeted with deltaD-MO the trend in spaw (84% on the left; 7% on the right; 8% bilateral; n=57) and pitx2 (33% on the left; 41% bilateral; 25% absent; n=12) expression is similar to the one observed in aei^–/– mutants. Moreover, embryos injected with deltaD-MO at the midblastula stage, showed gut looping defects (74% loops to the left, 18% does not loop and 8% loops to the right; n=123) and had no heart laterality problems similar to what we described for the aei^–/– mutants.

In order to investigate why the aei^–/– mutant cilia are shorter, we analysed the expression pattern of foxj1a, a marker of KV motile cilia (Amack and Yost, 2004; Stubbs et al., 2008; Yu et al., 2008). At the bud stage, the DFC marker sox17 was detected in both aei^–/– and wild-type embryos (Fig. 3A,B), confirming presence of DFCs. By contrast, foxj1a expression was downregulated in aei^–/– mutants at the same stage compared with wild-type embryos (Fig. 3C,D). We further confirmed that foxj1a expression was downregulated at bud stages in 25% of the embryos that resulted from an aei heterozygous cross (n=157). In addition, overexpression of both deltaD and NICD mRNA resulted in ectopic expression of foxj1a in the DFCs region at the bud stage (Fig. 3E,F). We thus propose that DeltaD controls KV cilia length via the regulation of foxj1a. Consistently, foxj1a morphants also show shorter or absent cilia in the KV (Stubbs et al., 2008; Yu et al., 2008).

In order to test whether Notch signalling is working via Foxj1a, we performed an epistatic test by co-injecting foxj1a mRNA together with GFP mRNA, as a lineage tracer, into one-cell stage aei^–/– mutant embryos. We found that cilia length was rescued in the KV cells that expressed GFP (Fig. 3H). Although in the aei^–/– mutant KV cells the average cilia length was 2.25 μm, in the rescued aei^–/– mutant cells, the average was 4±0.37 μm (s.d.m.) (n=53 cilia/5 embryos; P<1.0×10^–5). This experiment confirmed that foxj1a functions downstream of DeltaD.

To evaluate the impact that the shorter cilia of aei^–/– mutants might have on the fluid flow inside the KV, we measured the velocity of the flow in these mutants and compared it with wild-type embryos. For this, we filmed KVs of live embryos and tracked naturally occurring particles that move with the fluid flow (Fig. 4A,A′,B,B′; see Movies 1 and 2 in the supplementary material). This non-invasive method safeguards the epithelium and natural osmotic pressure inside the KV. We showed that in aei^–/– mutants, velocity of the fluid flow is less than half than that of its wild-type siblings (Fig. 4C; 6.74 μm/s and 16.46 μm/s, respectively; P<1.5×10^–4). Importantly, directionality of the flow is only lost in aei^–/– mutants (Fig. 4B-B′, see Movie 2 in the supplementary material). Loss of directionality could be due to a perturbation in the normal distribution of cilia inside the KV. However, a detailed 3D reconstruction of the spatial distribution of cilia in the KV of aei^–/– mutants revealed no significant differences when compared with wild-type embryos (Fig. 4D). We thus conclude that the reduced cilia length documented here constitute the most probable cause for the slow and chaotic fluid flow observed in aei^–/– mutants.

Charon is a Cerberus/Dan family secreted factor expressed specifically in the cells that line the posterior domain of the KV in a horseshoe shape (Fig. 4H). This protein negatively regulates Nodal signalling by blocking the transfer of Spaw from the KV to the right side of the LPM in zebrafish and medaka (Hashimoto et al., 2004; Hojo et al., 2007). In addition, it has been suggested that charon might be a potential Notch signalling target in zebrafish (Gourronc et al., 2007). We confirmed this possibility by injecting deltaD or NICD mRNA into one-cell stage wild-type embryos, which resulted in clear ectopic expression of charon in the KV region (Fig. 4F,G). By contrast, the expression of charon in aei^–/– mutants was severely reduced in 35.4% of the embryos from eight- to 12-somite stage (Fig. 4O,P), whereas in bea^–/– mutants it was unaffected. Together, these data confirmed that charon is a target of the Notch signalling pathway in zebrafish.

In contrast to what has been described (Hashimoto et al., 2004), we found that charon expression in a wild-type zebrafish KV is initially symmetric from the five- to seven-somite stage and then becomes clearly asymmetric on the right side from eight-somite stage onwards (Fig. 4H-K). It is still unclear whether the switch to higher levels on the right side corresponds to a downregulation of charon levels on the left side or to an upregulation on the right side of the KV. Our results identify charon as the first asymmetric gene expressed in the zebrafish KV and provide a better explanation for the mechanism underlying the asymmetric inhibition of Spaw transfer to the right LPM.

As we show that charon expression becomes asymmetric after being symmetrically induced (Fig. 4H-K), we asked whether fluid flow in the KV could be the mechanism controlling this switch, as suggested in medaka and in Xenopus (Hojo et al., 2007; Schweickert et al., 2010). In contrast to wild-type embryos where there is a clear shift from symmetric to an asymmetric charon expression at the eight-somite stage (Fig. 4K), in the aei^–/– mutants, this shift does not occur (Fig. 4P). In fact, in aei^–/– mutant embryos where the expression of charon was not severely reduced, we could score it as being symmetric in 43.3% and asymmetric in only 21.3% from 8-12 somite stages (Fig. 4L-P). This suggests that the slower fluid flow in aei^–/– mutants characterized above is inefficient to promote the asymmetry switch of charon expression in KV cells.

Having shown that DeltaD regulates charon transcription (Fig. 4L-P), we could hypothesize that the inability to shift to the asymmetric charon expression observed in aei^–/– mutants may be caused by a transcriptional defect and not by the slow flow. To investigate the impact of KV fluid flow on charon expression in a transcription competent context, we repeated the experiment in left-right dynein1 (lrd1) morphants. These morphants have a very reduced or absent flow (Essner et al., 2005) and show no problems in charon transcription (Gourronc et al., 2007). We injected lrd1-MO into wild-type embryos and scored charon expression in eight- to 12-somite stage embryos. We observed that in lrd1 morphants the asymmetric shift in charon expression does not occur (Fig. 4R), strengthening the idea that the fluid flow inside the KV is crucial for inducing asymmetric charon expression. Our results, together with those obtained in other species, clearly show that charon transcription is sensitive to fluid flow. Therefore, we suggest that charon and its homologues may be candidates for flow-sensing genes in the KV/GRP/node environment.

The involvement of Notch signalling in the establishment of the left-right axis is evolutionary conserved among vertebrates. However, there are differences in the mode of action of this molecular signalling pathway between mammals and fish. In the mouse, the analysis of Notch pathway mutants reveals that nodal induction in the murine node is impaired (Krebs et al., 2003; Przemeck et al., 2003; Raya et al., 2003; Takeuchi et al., 2007), whereas the expression of the secreted Nodal antagonist cerl2 does not seem to be affected (Krebs et al., 2003; Przemeck et al., 2003; Raya et al., 2003; Takeuchi et al., 2007). By contrast, the expression of spaw is normal in zebrafish embryos treated with DAPT (a Notch signalling inhibitor) but the expression of the secreted Nodal antagonist charon is severely affected in the KV (Gourronc et al., 2007). Using a genetic approach, we were able to confirm and extend these pharmacological results to conclude that the relevant Notch ligand for charon transcription in zebrafish is DeltaD. Therefore, Notch signalling in these two organisms targets the transcription of different genes: nodal in the mouse and charon in the zebrafish.

In the mouse node, the symmetric expression of nodal, induced by Notch signalling, and that of cerl2, induced by an unknown mechanism, later becomes asymmetric. nodal levels become higher on the left side, whereas cerl2 levels are higher on the right side of the node (Marques et al., 2004). In contrast to what happens in the mouse, there was no evidence for an asymmetric spaw or charon expression in the zebrafish KV. In our study, we discovered a sharp transition from a symmetric to an asymmetric expression of charon at the eight-somite stage in the KV, with higher levels of expression on the right side. This finding strengthens the evolutionary conserved feature of charon expression because this gene, and its homologues, are asymmetrically expressed on the right side in medaka KV (Hojo et al., 2007), Xenopus gastrocoel roof plate (GRP) (Schweickert et al., 2010) and mouse node (Marques et al., 2004). The mechanism that triggers the left bias of nodal expression in the murine node and the right bias of charon and coco expression on the right side in fish KV and Xenopus GRP remains poorly understood.

There is increasing evidence in these organisms that the extracellular fluid flow generated by motile cilia present in the node/KV/GRP is essential to establish the LR asymmetries in the LPM. Therefore, this directional fluid flow is a good candidate to trigger the initial asymmetric expression at the node/KV/GRP (Hamada, 2008). We discovered that in aei^–/– mutants, the switch in charon expression does not occur, possibly as a consequence of the slower fluid flow generated by the short motile cilia. In agreement with the possibility that the dynamics of the fluid flow promote a switch in gene expression in the node/KV/GRP are the phenotypes described for the axonemal dynein morphants in fish and Xenopus and the mouse mutants iv and inv. In medaka and as shown here in zebrafish embryos, the right bias in charon expression is no longer detected when fluid flow is abolished by the knockdown of lrd1 (Hojo et al., 2007) (Fig. 4Q,R). In Xenopus, coco is no longer downregulated on the left side of the GRP in the absence of flow (Schweickert et al., 2010). In the iv mouse mutant, where cilia are immotile and no nodal flow is produced, the expression of nodal in the node becomes randomized (Okada et al., 1999). In addition, in the inv mouse mutant, where the flow is slow and turbulent, the expression of nodal is mostly symmetric in the node (Lowe et al., 1996; Okada et al., 1999; Oki et al., 2009).

Although no asymmetric charon expression had been detected before in zebrafish, asymmetric Charon function was suggested to be responsible for inhibiting the transference of Spaw from the KV domain to the right LPM (Hashimoto et al., 2004). However, it should be emphasized that in wild-type embryos, charon is also expressed on the left side of the KV and spaw can still be transferred to the left LPM. This suggests that it is not the absence of charon transcripts that allow Spaw transfer to the left LPM but the relative levels of charon expression between left and right sides. Thus, we reason that Spaw is transferred to the left LPM because the levels of charon transcripts are lower on the left side of the KV in a wild-type embryo.

Applying the same reasoning to the aei^–/– mutants, we propose that the severely reduced charon expression observed in 35.4% of the embryos (Fig. 4O,P) could correspond to: (1) no differences in the relative levels of charon [and therefore spaw is bilateraly transferred to the LPM as observed in 20% of aei^–/– embryos (Fig. 1A)]; or (2) undetected small differences in charon expression that lead to a weak bilateral situation described as a spaw left anterior domain with bilateral posterior domains as observed in 15% of aei^–/– embryos (Fig. 1A). In 43.3% of aei^–/– mutants, charon expression is largely symmetric as visualized by in situ hybridization (Fig. 4L,P). In this situation, we cannot rule out that a biased difference between the left and the right side of charon transcripts in aei^–/– mutants is still present as most embryos express left-sided spaw in the LPM (Fig. 1A). A clear asymmetric charon expression pattern on the right side of the KV (Fig. 4M) was observed in 18.3% of aei^–/– embryos, resulting in left-sided spaw expression in the LPM (Fig. 1A). A few embryos (3%) expressed charon more strongly on the left side of the KV (Fig. 4N), matching the equivalent percentage of right-sided spaw in the LPM (Fig. 1A).

In summary, we showed that shorter cilia present in aei^–/– mutants produce a slower fluid flow that leads to abnormal LR charon levels at the KV, which compromises the maintenance of spaw expression in the posterior domain of the left LPM. We show that in the same embryos where we observed a premature downregulation of spaw, we also see a reduced lefty1 expression in the posterior midline domain. We thus propose that this gap in lefty1 expression contributes, together with abnormal posterior pitx2 expression, to the gut laterality defects without affecting heart laterality. This phenotype is clinically relevant as there are reports on individuals with primary ciliary dyskinesia that develop situs inversus abdominalis with no heart laterality defects (Fliegauf et al., 2007).

To our knowledge, Notch signalling is the first pathway that can both increase or decrease ciliary length, upon upregulation or downregulation, respectively. By contrast, FGF signalling has only been involved in shortening cilia length, (Hong and Dawid, 2009; Yamauchi et al., 2009). Our experiments identify foxj1a, the master motile ciliogenic transcription factor (Yu et al., 2008) as being downstream of DeltaD because it successfully rescued cilia length in aei^–/– mutants in a cell-autonomous manner. These observations perhaps indicate that the motile cilia programme controlled by foxj1a, involves not only the transcription of motility genes such as dnah9 and wdr78 (Yu et al., 2008) but may also transcribe genes important for cilia growth yet to be discovered.

A systematic analysis of the experimental conditions where short and long cilia are generated by modulating the Notch activity will provide a framework to determine relevant ciliary length control genes. This will certainly bring new insights into how cilia/flagella length is regulated and may provide ways to increase ciliary length that could ultimately be used to treat ciliopathies often characterized by short cilia (Wemmer and Marshall, 2007).

We thank J. Lewis and S. Holley for fixed aei^tr233 and bea^tit446 mutant embryos, respectively; A. Oates for Su(H)1+2-MO; and S. Roy, W. Norton, S. Wilson, J. Yost, C.-P. Heisenberg, M. Hibi and J. Lewis for in situ probes. We are grateful to A. T. Tavares, J. Regan, D. Henrique, A. Jacinto, M. Godinho Ferreira, M. Bettencourt-Dias, J. Lewis, R. Monteiro, S. Marques and the Saúde labaoratory for comments on the manuscript; and to Lara Carvalho and the zebrafish facility team for excellent fish husbandry. S.S.L. and R.L. were supported by FCT fellowships (SFRH/BPD/34822/2007 and SFRH/BD/24861/2005). L.S. was supported by two FCT grants (PTDC/SAU-OBD/71596/2006 and PTDC/SAU-OBD/64628/2006).