To explore the signals downstream the Reissner fiber, we performed a transcriptomic analysis of scospondin mutants (Figure 1).

Figure 1 with 1 supplement see all

Download asset Open asset

Figure 1—source data 1

Data for Figure 1B–C.

https://cdn.elifesciences.org/articles/59469/elife-59469-fig1-data1-v3.xlsx

Download elife-59469-fig1-data1-v3.xlsx

Figure 1—source data 2

Up- and down-regulated transcripts in curled-down scospondin mutants compared to their control siblings.

After filtering for low expression, we kept a list of genes that were differentially expressed in null scospondin^icm13/icm13 and hypomorphic scospondin^icm15/icm15 mutant embryos at 48 hpf. The table shows the average fold changes in the two alleles and their mean. Genes are ranked based on their p-value that compare the reproducibility of the change inside and between the two alleles using a GLM framework. We adjusted the false discovery rate due to multiple comparisons using the Benjamini-Hochberg procedure (see Materials and methods for details). The raw RNA-seq data have been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-9615.

https://cdn.elifesciences.org/articles/59469/elife-59469-fig1-data2-v3.xlsx

Download elife-59469-fig1-data2-v3.xlsx

We took advantage of two previously-generated scospondin alleles (Cantaut-Belarif et al., 2018) to evaluate transcriptional modifications associated with the curled-down phenotype due to the loss of the Reissner fiber in the CSF. While the scospondin^icm13 null allele leads to a dual loss of the fiber in the central canal and of the SCO-spondin protein detection in secretory structures, the scospondin^icm15 hypomorphic allele retains protein expression but solely precludes the Reissner fiber formation (Figure 1A). Both homozygous mutants lack the Reissner fiber and exhibit a typical curled-down phenotype, which arises from 28 to 30 hr post-fertilization (hpf) onwards and was undetectable beforehand (Cantaut-Belarif et al., 2018, Figure 1A, Figure 1—figure supplement 1). We performed pairwise comparisons of the transcriptomes of homozygous mutants versus control siblings at 48 hpf when the body axis curvature defect is fully penetrant (Cantaut-Belarif et al., 2018). The resulting lists of up- and down-regulated transcripts that were common to the two scospondin alleles are presented in Figure 1—source data 2. Very few genes exhibited an important change in expression, and only a handful changed more than two folds. Noticeably, we observed a strong reduction of urp2 gene expression in curled-down mutant embryos lacking the Reissner fiber compared to their control siblings (mean ± SEM fold decrease: 4.07 ± 1.2 in scospondin^icm13/icm13 and 4.69 ± 1.77 in scospondin^icm15/icm15, n = 3 replicates each, p-value<0.00005, GLM test; see Materials and methods and Figure 1—source data 2 for details). There was no significant decrease in urp1 transcript levels (mean ± SEM fold decrease: 0.97 ± 0.06 in scospondin^icm13/icm13 and 1.22 ± 0.25 in scospondin^icm15/icm15, n = 3 replicates each, p-value=p = 0.191, GLM test).

The gene urp2 encodes for a secreted neuropeptide belonging to the Urotensin-II-related-peptide family (Tostivint et al., 2014). Together with urp1, these transcripts have recently been identified as both strongly downregulated in curled-down mutants (Lu et al., 2020; Zhang et al., 2018). To ascertain our RNAseq results and confirm the difference we observed with previous results, we carried out qRT-PCR analysis of urp1 and urp2 expression levels in the hypomorphic scospondin^icm15 allele. Interestingly, we observed that urp1 expression level is not significantly decreased in scospondin homozygous mutants compared to their control siblings, neither at 30 hpf nor at 48 hpf (Figure 1B). Consistently with transcriptomic results, urp2 expression level shows a strong decrease at 48 hpf in mutant embryos compared to their control siblings (3.6 ± 0.2 fold decrease; mean ± SEM; Figure 1C). This is also true at the onset of the curled-down phenotype (30 hpf: 4.2 ± 0.5 fold decrease; mean ± SEM; Figure 1C) indicating that urp2 gene expression level is affected when embryos start to develop an abnormal morphogenesis of the posterior axis. Taken together, these data show that in zebrafish embryo the presence of the Reissner fiber in the CSF is required for the normal expression level of urp2, but not urp1.

The expression of Urotensin-II-related peptides is restricted to the ventral population of CSF-cNs (Quan et al., 2015), known to exhibit spontaneous intracellular calcium variations around 30 hpf (Sternberg et al., 2018) when the curled-down phenotype becomes visible in scospondin mutants (Cantaut-Belarif et al., 2018 and Figure 1—figure supplement 1). Curled-down cilia-defective embryos lack these early calcium transients (Sternberg et al., 2018) and do not form a proper Reissner fiber (Cantaut-Belarif et al., 2018). We therefore hypothesized that the Reissner fiber may functionally interact with ventral CSF-cNs that are expressing urp2.

To address this question, we performed in vivo population calcium imaging at 28–30 hpf using the Tg(pkd2l1:GCaMP5G) line labeling the dorso-lateral and ventral CSF-cNs in the spinal cord (Figure 2A, Figure 2—video 1). As previously described, we observed that ventral CSF-cNs exhibit spontaneous calcium transients in wild-type embryos (scospondin^+/+, Figure 2B, Figure 2—video 1). Quantification of the integrated fluorescence variations over time showed that in heterozygous scospondin^icm15/+ embryos, which display a straight body axis and form a proper Reissner fiber (Cantaut-Belarif et al., 2018), ventral CSF-cNs retained the same level of activity than in wild-type (Figure 2C). On the contrary, homozygous scospondin^icm15/icm15 embryos exhibited a 52.8% decrease of calcium activity compared to wild-type (Figure 2C). A similar reduction of 39% occurred in the null scospondin^icm13/icm13 mutant compared to wild-type siblings (Figure 2—figure supplement 1A,B). These data show that the loss of the Reissner fiber reduces spontaneous calcium variations of ventral CSF-cNs during the critical time window for body axis straightening. On the contrary, the sparse spontaneous calcium transients in dorsal CSF-cNs were not affected by the absence of the Reissner fiber in both scospondin alleles (Figure 2—figure supplement 1C,D).

Figure 2 with 2 supplements see all

Download asset Open asset

Figure 2—source data 1

Data for Figure 2C and Figure 2—figure supplement 1B–C.

https://cdn.elifesciences.org/articles/59469/elife-59469-fig2-data1-v3.xlsx

Download elife-59469-fig2-data1-v3.xlsx

Early calcium transients in CSF-cNs require the non-selective cationic channels permeable Pkd2l1 channel in vivo (Sternberg et al., 2018) that is permeable to calcium ions. The reduction in calcium activity when the Reissner fiber is lacking could therefore be due to a defect in the Pkd2l1 channel localization or opening probability. In toto immunohistochemistry for Pkd2l1 protein showed that this channel is enriched at the differentiating apical extension of CSF-cNs in curled-down scospondin^icm15/icm15 mutants as well as in control embryos (Figure 2D). In order to assess Pkd2l1 channel properties, we performed in vivo whole-cell voltage-clamp recordings in double transgenic Tg(pkd2l1:GAL4;UAS:mCherry) embryos. In the absence of a Reissner fiber in scospondin^icm15/icm15 mutants, we observed spontaneous Pkd2l1 channel openings that were similar to control embryos (Figure 2E). Thus, the loss of the Reissner fiber decreases CSF-cNs intracellular calcium variations without preventing the opening nor the localization of Pkd2l1 channels in the differentiating apical extension of the cells. Altogether, these data suggest that the Reissner fiber is required for a signal acting on ventral CSF-cNs that controls both the spontaneous calcium variations and the expression of urp2 in the embryo.

As the Reissner fiber is required for both ventral CSF-cNs spontaneous calcium transients and urp2 gene expression, we tested whether intracellular calcium variations are necessary for a normal urp2 expression and axis straightness. We took advantage of the pkd2l1^icm02/icm02 mutant where the embryonic activity of CSF-cNs is abolished (Sternberg et al., 2018).

Thanks to the viability of pkd2l1^icm02/icm02 zygotic mutants, we generated genetically-related clutches that were either fully wild-type or fully maternal and zygotic (MZ) homozygous mutant (Figure 3A). We observed that pkd2l1^icm02/icm02 MZ mutant embryos did not display any defect in body axis morphogenesis and were morphologically indistinguishable from wild-type embryos (Figure 3B). Using immunohistochemistry, we observed that pkd2l1^icm02/icm02 mutants form a normal Reissner fiber in the central canal of the spinal cord at 30 hpf (Figure 3C). Next, using qRT-PCR, we tested whether urp1 and urp2 gene expression levels were diminished by the absence of calcium transients in CSF-cNs (Figure 3D). As the mutation in the icm02 allele generates a premature stop codon in the pkd2l1 gene (Böhm et al., 2016), one can predict that the pkd2l1 mRNA would be degraded by nonsense-mediated decay. Indeed, mutant clutches displayed a 4.9 fold decrease in pkd2l1 transcripts level compared to wild-type counterparts (Figure 3D). Interestingly, we observed that urp2 and urp1 gene expression were not decreased in pkd2l1^icm02/icm02 embryos compared to wild-types (Figure 3D). Altogether, these data show that the loss of Pkd2l1-driven calcium transients in CSF-cNs does not lead to a decreased expression of urp1 and urp2 transcripts, rejecting the hypothesis of a direct requirement for urp2 expression of Pkd2l1-dependent intracellular calcium variations. Instead, our results show a strict correlation between the presence of the Reissner fiber, a normal urp2 expression level and the proper morphogenesis of the embryonic body axis. Altogether, our results prompted us to hypothesize that the aberrant posterior axis curvature in absence of the Reissner fiber is a consequence of the decreased urp2 expression.

Figure 3 with 1 supplement see all

Download asset Open asset

Figure 3—source data 1

Data for Figure 3D.

https://cdn.elifesciences.org/articles/59469/elife-59469-fig3-data1-v3.xlsx

Download elife-59469-fig3-data1-v3.xlsx

Given the strict correlation between urp2 expression and body axis straightening, we hypothesized that restoring urp2 levels would rescue the posterior axis curvature developed by scospondin^icm15/icm15 mutants. To test this hypothesis, we performed one cell stage urp2 mRNA injections on clutches obtained from scospondin^icm15/+ parents. We sorted the injected embryos at 48 hpf into two morphological categories: curled-down and non-curled-down (Figure 3E and F). In the four independent experiments conducted, injections with a control mRNA led to a proportion of curled-down embryos close to 25%, as expected (Figure 3F). On the contrary, embryos overexpressing urp2 showed much lower proportions of curled-down embryos (10.7 ± 1.5%; mean ± SEM; Figure 3F), suggesting that some homozygous mutants are rescued in this condition. To confirm this rescue, we genotyped curled-down and non-curled-down 48 hpf embryos injected with a control mRNA or urp2 mRNA (Figure 3E and G, Figure 3—figure supplement 1). As expected, wild-type and heterozygous embryos injected with the control mRNA displayed a straight body axis, while scospondin^icm15/icm15 showed a downward curvature of the posterior axis. Instead, urp2 overexpression lead to detect straight or slightly curled-up scospondin^icm15/icm15 embryos at several instances (Figure 3E and G, Figure 3—figure supplement 1). Thus, restoring higher urp2 levels is sufficient to prevent the embryonic posterior axis defects in scospondin mutants. These observations confirm that urp2 neuropeptide expressed in CSF-cNs may signal at long-range, as suggested from the expression of the receptor in dorsal somites (Zhang et al., 2018), to ensure a proper axis morphogenesis.

Based on our observations, we assumed that the Reissner fiber is necessary for the activity of at least one signaling pathway regulating urp2 gene expression together with body axis morphogenesis. Epinephrine and norepinephrine belong to the monoamine neurotransmitter family and are known to bind the Reissner fiber in rats (Caprile et al., 2003) and frogs (Diederen et al., 1983). Recently, systemic bath applications of monoamines have been described to rescue body axis defects in curled-down cilia-defective mutants (Lu et al., 2020), (Zhang et al., 2018). We investigated the possible role of epinephrine and norepinephrine in the regulation of the RF signaling pathways.

First, we tested if epinephrine and norepinephrine could influence the curled-down phenotype developed by scospondin^icm15/icm15 mutants. We compared sibling animals generated from incrosses of scospondin^icm15/+. We first analyzed the effect of 2.5 hr bath applications of epinephrine and norepinephrine on control embryos at 30 hpf (Figure 4A). Epinephrine and norepinephrine have a moderate impact on the shape of the head-to-tail axis of initially straight embryos, which display a slight curled-up phenotype after exposure to monoamines (Figure 4B). To estimate the straightness of the posterior axis, we quantified the angle formed between the ear, the caudal limit of the yolk extension and the tip of the tail (Figure 4C, top panel). This angle is distributed around 190.6° in control embryos exposed to a vehicle solution (median value; Figure 4C, bottom left graph). Comparatively, control embryos exposed to monoamines exhibit a distribution of ear-to-tail angles shifted toward slightly higher values (median values: 220.4° and 213.6° for epinephrine and norepinephrine, respectively; Figure 4C, bottom right graph). Next, we analyzed the effect of epinephrine and norepinephrine on scospondin^icm15/icm15 embryos. While mutant animals exposed to a vehicle solution display a typical curled-down body axis, scospondin^icm15/icm15 mutants treated with monoamines exhibit a reduction in the downward curvature of the posterior axis (Figure 4B). Quantifications of ear-to-tail angles in scospondin^icm15/icm15 mutants show a large increase of the median angle value after exposure to monoamines, compared to embryos treated with a vehicle solution (106.5° for vehicle, 167.1° for epinephrine and 158.5° for norepinephrine; median values; Figure 4C). Altogether, these data show that epinephrine and norepinephrine can partially restore the posterior axis geometry of scospondin^icm15/icm15 mutants, suggesting that monoamines can rescue the Reissner fiber-dependent signal required for a straight embryonic body axis.

Figure 4

Download asset Open asset

Figure 4—source data 1

Data for Figure 4C–E.

https://cdn.elifesciences.org/articles/59469/elife-59469-fig4-data1-v3.xlsx

Download elife-59469-fig4-data1-v3.xlsx

Next, we analyzed the effect of epinephrine and norepinephrine on urp1 and urp2 gene expression using qRT-PCR. In control siblings, urp1 expression remains comparable after vehicle, epinephrine or norepinephrine exposure (Figure 4D). Consistently with our previous results, urp1 expression is not significantly modified in curled-down scospondin^icm15/icm15 embryos compared to their control siblings receiving the same treatment (1.01 ± 0.17-fold change; mean ± SEM; Figure 4D). However, we observed a slight increase of urp1 expression in curled-down mutant embryos treated with epinephrine and norepinephrine compared to vehicle treatment (1.58 ± 0.43 and 1.28 ± 0.23 fold changes for epinephrine and norepinephrine respectively; mean ± SEM). As expected, urp2 expression level is significantly decreased in curled-down scospondin^icm15/icm15 mutants compared to straight siblings (0.22 ± 0.06-fold change; mean ± SEM; Figure 4E). Epinephrine and norepinephrine treatments also do not change urp2 expression in control siblings, but significantly increase it in curled-down homozygous mutant embryos (7.42 ± 4,23 and 4.67 ± 1.15 fold changes for epinephrine and norepinephrine respectively; mean ± SEM; Figure 4E).

These observations show that the rescue of the posterior axis curvature of scospondin homozygous mutant embryos by monoamines is associated with an increase of the expression of urp1 and urp2 neuropeptides. Epinephrine and norepinephrine compensate the loss of the Reissner fiber both on Urotensin-II-related neuropeptides expression and on posterior axis curvature, suggesting that these compounds act on the Reissner fiber-dependent signaling pathway in the embryonic CSF.

The Reissner fiber is required for three concomitant events: intracellular calcium variations and urp2 expression in ventral CSF-cNs, and the straightening of the posterior axis. We therefore asked whether the delivery of monoamines in CSF could also restore calcium variations in ventral CSF-cNs. To address this question, we performed hindbrain ventricle injections of epinephrine or norepinephrine in the Tg(pkd2l1:GCaMP5G); scospondin^icm15 embryos at 30 hpf, and recorded calcium variations in ventral CSF-cNs within the spinal cord 20 to 60 min after the injections (Figure 5A). In order to assess how the monoamines used diffuse in the central canal of the spinal cord and whether they are stable enough in the CSF, we performed immunostainings against norepinephrine 30 and 60 min after hindbrain ventricle injections (Figure 5—figure supplement 1). We observed an abundant norepinephrine-positive signal in the central canal of the rostral most region of the spinal cord 30 min post-injection, and the median and caudal most-regions 60 min post-injection, which was absent in control injections (Figure 5—figure supplement 1). This observation indicates that norepinephrine diffuses down the central canal when injected in the same conditions used for in vivo calcium imaging and that it is stable in the CSF after an hour, in agreement with former results in the mammalian ventricular system of dogs (Maas and Landis, 1965), sheeps (Forbes and Baile, 1974) and rats (Fuxe and Ungerstedt, 1966; Levitt et al., 1983).

Figure 5 with 2 supplements see all

Download asset Open asset

Figure 5—source data 1

Data for Figure 5D–E.

https://cdn.elifesciences.org/articles/59469/elife-59469-fig5-data1-v3.xlsx

Download elife-59469-fig5-data1-v3.xlsx

We first analyzed the effect of epinephrine and norepinephrine injections on control embryos displaying a straight body axis at 30 hpf and observed that an exogenous delivery of monoamines did not influence the basal calcium variations of ventral CSF-cNs (Figure 5B and D). Next, we analyzed the effect of epinephrine and norepinephrine in scospondin^icm15/icm15 mutants (Figure 5C and E, Figure 5-video 1). As described previously, ventral CSF-cNs displayed a 44.1% decrease of calcium variations in scospondin^icm15/icm15 embryos compared to control siblings in control condition with vehicle injections (median value; Figure 5D and E), confirming the observations reported in Figure 3. Interestingly, injections of epinephrine or norepinephrine in scospondin^icm15/icm15 mutants increased the median ∫ΔF/F min⁻¹ by 70.02 and 88.6%, respectively compared to vehicle injections (Figure 5E). Moreover, the rescue of the spontaneous activity of CSF-cNs by monoamines in mutant embryos reached comparable levels to those observed in control siblings (Figure 5D–E). These data show that epinephrine and norepinephrine restore the Reissner fiber-dependent calcium transients in ventral CSF-cNs in scospondin mutants.

Altogether, our results are compatible with the existence of a signal that links the Reissner fiber to both urp2 expression and calcium variations in ventral CSF-cNs that can be modulated by monoaminergic activation. These observations suggest that endogenous epinephrine and norepinephrine may act locally to tune calcium signaling in CSF-cNs.

To assess that this modulation takes place at the interface between CSF and the cells lining the central canal, we performed immunostainings against endogenous norepinephrine at 30 hpf in scospondin^ut24Tg embryos labeling in vivo the Reissner fiber with GFP (Troutwine et al., 2020). As previously described, SCO-spondin positive material in the CSF contributes mainly to form the fiber, but is also present as punctated material in close vicinity the fiber, referred to here as extrafibrillar material, (Troutwine et al., 2020, Figure 6A). We observed norepinephrine positive puncta colocalized with the Reissner fiber as well as endogenous norepinephrine positive signals following patterns similar to extrafibrillar material labeled by the SCO-spondin-GFP fusion protein. This can also be observed closely at the level of the massa caudalis, formed by the accumulation of SCO-spondin at the caudal end of the central canal (Figure 6A). These observations suggest that norepinephrine is endogenously present in the embryonic CSF where it is associated with the Reissner-positive material in the central canal.

Figure 6 with 1 supplement see all

Download asset Open asset

To address the question of the receptor associated to this signaling, we performed immunostainings against the adrenergic receptor Adrb2, binding both epinephrine and norepinephrine, which is described to be transiently expressed in the zebrafish nervous system at early stages of embryonic life (Wang et al., 2009). We observed that Adrb2 is distributed ventrally in the neural tube, at the interface with the central canal, in a pattern that suggests a membrane localization in both control siblings and scospondin^icm15/icm15 mutants (Figure 6B). Thus, Adrb2 localization is suitable for binding endogenous ligands present in the CSF.

Next, we addressed the question of whether Adrb2 is expressed in CSF-contacting neurons, which could explain the rescue of calcium transients observed in scospondin^icm15/icm15 mutants after epinephrine and norepinephrine injections. Double immunostainings against Adrb2 and GFP in Tg(pkd2l1:GCaMP5G) control siblings show that both signals cover different domains in the ventral most region of the spinal cord (Figure 6C), where the Adrb2 positive domain remains in the ventral midline, inserted between two rows of ventral CSF-contacting neurons (Figure 6—video 1). As ventral CSF-contacting neurons are known to derive from lateral floor plate progenitors (Park et al., 2004; Yang et al., 2010), the medial cells expressing Adrb2 most probably correspond to the medial floor plate. Importantly, the same distribution is observed in scospondin^icm15/icm15 mutants (Figure 6C and Figure 6—video 1). This observation suggests that the loss of calcium signaling in scospondin^icm15/icm15 mutants is unlikely due to a defect in the Adrb2 distribution in the spinal cord.

Altogether, our results indicate that endogenous adrenergic signals could modulate the Reissner fiber-dependent signaling pathway that instructs body axis straightening during embryonic development, and suggests that monoamines act on CSF-contacting neurons through an indirect mechanism.

Previous results in both cilia-defective and scospondin mutants suggested that mutants that fail to form a Reissner fiber display a strong downregulation of both Urotensin-II-related neuropeptides 1 and 2 in the embryo (Lu et al., 2020; Rose et al., 2020; Zhang et al., 2018). Our new results contradict the observation that urp1 is downregulated in scospondin mutants (Lu et al., 2020). In the present study, our transcriptomic analysis of scospondin mutants yielded surprisingly few candidates commonly misregulated in the two alleles, and urp2 was the only gene with a strong downregulation, while urp1 showed no significant change. We confirmed this result by qRT-PCR and showed again that only urp2 is strongly downregulated, while urp1 transcripts levels were not significantly affected. In line with our results, the investigation of a newly generated scospondin hypomorphic allele (Rose et al., 2020) also revealed urp2 as the major downregulated gene in homozygous mutant and did not report a urp1 downregulation.

We further show that urp2 overexpression in scospondin mutants decreases the frequency of curled-down phenotypes, confirming that urp2 expression level is involved in the control of embryonic axis curvature (Lu et al., 2020; Zhang et al., 2018). However, urp2 may not be the only determinant of embryonic axis morphogenesis downstream of the Reissner fiber, as single urp2 morpholino knockdown does not show a defective axis curvature (Zhang et al., 2018). Parallel signaling pathways that would likely act post-transcriptionally and would therefore not be detected using our transcriptomic strategy may be involved.

The discovery of a new hypomorphic allele for scospondin recently revealed an inflammatory signature induced by the loss of the Reissner fiber at the embryonic stage (Rose et al., 2020). In our work, we did not detect such a signature in our transcriptomic analysis (see Figure 1—figure supplement 1). We can speculate that these differences are due to difference in the fish genetic background or husbandry conditions. Our work therefore begs for further studies in order to decipher the molecular pathways downstream and/or in parallel to Urp2 that regulate the morphogenesis of the embryonic axis.

In the embryo, ventral CSF-cNs are spontaneously active via the opening of the Pkd2l1 calcium channel enriched at the level of the developing apical extension of the cells (Sternberg et al., 2018). In addition to controlling urp2 gene expression in CSF-cNs, we report here that the Reissner fiber is also required for calcium signaling in urp2-expressing cells. By investigating the pkd2l1 mutant deprived of calcium signaling in embryonic CSF-cNs, our work rejects the simple explanation that the Reissner fiber controls urp2 mRNA level and axis straightness by increasing calcium intracellular concentrations in CSF-cNs. Indeed, the absence of urp genes downregulation observed in the pkd2l1 mutant compared to wild-type goes against this hypothesis.

One question remaining is how the Reissner Fiber controls calcium variations in ventral CSF-cNs. We recently showed that the Reissner fiber is functionally coupled to the mechano-sensory function of these interoceptive neurons in the larva (Orts-Del'Immagine et al., 2020). However, embryonic CSF-cNs are not fully differentiated, as they do not harbor a fully developed apical extension known to tune their mechanosensory function at larval stage (Desban et al., 2019). We therefore favor the hypothesis that the Reissner fiber acts via the modulation of the CSF content, which is supported by the restoration of calcium signaling in CSF-cNs upon monoamine injections in the brain ventricles.

Earlier reports suggest that the Reissner fiber can interact with different neuro- modulators, including monoamines (Caprile et al., 2003; Diederen et al., 1983). Monoamines globally supplied in the fish water were also found to rescue body curvature in Reissner fiber-defective mutants (Lu et al., 2020; Zhang et al., 2018). They are therefore good candidates to influence the Reissner fiber signal towards ventral CSF-cNs. We added to previous report (Lu et al., 2020) that the action of both epinephrine and norepinephrine can compensate locally, directly in the ventricular cavities, for the loss of the Reissner fiber on calcium signaling. We also used quantitative PCR to confirm the regulation of urp neuropeptides expression by both epinephrine and norepinephrine in scospondin mutants, extending the previous result obtained using in situ hybridization after epinephrine treatments (Lu et al., 2020). These two monoamines rescue the scospondin phenotype for three features: spontaneous calcium variations and Urotensin-II-related peptides expression in CSF-cNs, and body axis curvature.

One possible interpretation of our results is that the Reissner fiber is essential for the control of endogenous epinephrine and norepinephrine distribution in the embryonic CSF. This is consistent with our immunodetection of norepinephrine in close vicinity with the Reissner fiber in the embryonic central canal. This hypothesis is supported by the presence of noradrenergic neurons in the embryonic hindbrain as early as 24 hpf (Holzschuh et al., 2003), providing a potential source of monoamines that would need to be transported caudally to the central canal of the spinal cord. This hypothesis is reinforced by our original result showing the localization of the Adrb2 receptor at the interface with CSF in the ventral part of the neural tube, ideally located to bind norepinephrine ligands that we found distributed close to Reissner-positive material. Interestingly Adrb2 belongs to the G protein-coupled receptors family and was identified to trigger cytoplasmic calcium raise in vitro (Galaz-Montoya et al., 2017). In zebrafish, the adrenergic system plays important roles in the control of wakefulness (Singh et al., 2015) and of cardiac contractions (Steele et al., 2011). However, no morphological defects have been reported in animal missing the rate-limiting enzyme for the synthesis of epinephrine and norepinephrine (Singh et al., 2015). This might reflect the masking of the phenotype due to transcriptional adaptation to the genetic mutation (Rossi et al., 2015). Alternatively, redundant signaling pathways might mask the role of monoamines in axis curvature control. Nonetheless, our results support the idea that the cross-talk between the Reissner fiber and undifferentiated CSF-cNs is likely to be of chemical nature, possibly through monoamines themselves. They further support the idea that monoamines act on CSF-cNs in an indirect manner. Future investigations will allow to fully delineate the contribution and mechanism of action of endogenous monoamines on ventral CSF-cNs signaling and body axis curvature.

Altogether, our study unravels a signal from the Reissner fiber to the developing CSF-contacting neurons. We also show that adrenergic activation can modulate this signal during embryonic body axis morphogenesis. Interestingly, a temporally controlled inactivation of cilia motility leads to spine curves reminiscent of adolescent idiopathic scoliosis (Grimes et al., 2016). Recent results suggest that, as in the embryo, the Reissner fiber and Urp reception in slow muscles are also implicated in the maintenance of a straight spine during postembryonic development (Lu et al., 2020; Rose et al., 2020; Troutwine et al., 2020; Zhang et al., 2018). Our work will pave the way for future investigations to identify the potential of the interplay between CSF-cNs and the adrenergic system to modulate the Reissner fiber-dependent morphogenesis of the spine in the juvenile.

All procedures were performed on zebrafish embryos between 30 and 48 hpf in accordance with the European Communities Council Directive (2010/63/EU) and French law (87/848) and approved by the Paris Brain Institute (Institut du Cerveau) and the French Ministery for research (APAFIS agreement 2018071217081175). All experiments were performed on Danio rerio embryos of AB, Tüpfel long fin (TL) and nacre background. Animals were raised at 28.5°C under a 14/10 light/dark cycle until the start of the experiment. Genotyping was performed as previously described for scospondin^icm13 and scospondin^icm15 (Cantaut-Belarif et al., 2018) and pkd2l1^icm02 (Böhm et al., 2016).

For each experiment, condition (developmental time point and pharmacological treatment) and phenotype (either straight or curled-down), 30 sibling embryos were sorted in independent vials. After euthanasia in 0.2% MS 222 (Sigma, E10521), embryos were resuspended in 1 mL of Trizol (ThermoFischer Scientific, 15596026) and dissociated by multiple aspirations through the needle of a syringe. 200 µL chloroform were added prior to centrifugation and extraction of the aqueous phase. Nucleic acids were precipitated using 700 µL isopropanol, and the pellet was resuspended in 200 µL water. RNAs were purified using the RNeasy Micro Kit (Qiagen, 74004), following the provider's instruction. We performed the optional on-column DNAse treatment to improve RNAs purity. RNAs were eluted in 30 µL water to ensure a high concentration, and the quality and quantity of the extract were evaluated on a TapeStation System (Agilent).

RNA sequencing analysis mRNA library preparation was realized following manufacturer’s recommendations (KAPA mRNA HyperPrep Kit, Roche). Final samples pooled library preparations were sequenced on Nextseq 500 ILLUMINA, corresponding to 2 × 30 millions of 75 base pair reads per sample after demultiplexing. Quality of raw data was evaluated with FastQC (Andrews, 2010). Poor quality sequences were trimmed or removed using Trimmomatic software (Bolger et al., 2014) to retain only good quality paired reads. Star v2.5.3a (Dobin et al., 2013) was used to align reads on GRCz11 reference genome using standard options. Quantification of gene and isoform abundances was achieved using with Rsem 1.2.28, prior to normalization with the edgeR bioconductor package (Robinson et al., 2010). Finally, differential analysis was conducted with the GLM framework likelihood ratio test from edgeR.

The quality control and PCA analysis (not shown) indicated that one sample was not reaching the same reproducibility as the two others, even after trimming and normalization. We therefore used the average value of the two most reliable replicates to calculate the average expression level (count per million, cpm) of all the 28,214 genes either in straight controls or in curled-down embryos (average of the expression in the two alleles icm13 and icm15). We then filtered out low-expression genes that had an expression level below 1 c.p.m. in the average of the straight control replicates. We then kept genes that had an average fold change between curled embryos and controls of 0.75 for down-regulated genes (120 genes), and 1.45 for up-regulated genes (94 genes). We turned back to the original raw cpm values of the three replicates in the two alleles for these two short-lists. We tested the consistency of the fold change across replicates and across alleles with a general linear model (GLM) (Nelder and Wedderburn, 1972). The design matrix included 2 regressors of interest (encoding whether the sample is a straight control or curled embryos) and two confounding variables (the unwanted variability that might be associated with the two different genetic environments of the two families of fish carrying the icm13 and the icm15 allele). Statistical significance of the effect of interest (above and beyond confounding factors) was tested using a t-test. Because of the potential false discovery associated with multiple testing, we then used the Benjamini–Hochberg procedure (Benjamini and Hochberg, 1995).

Potential genes of interest were sorted according to their increasing p-values. To shorten the list of potential genes of interest, we determined at which rank i the p-value became higher than the Benjamini–Hochberg criterion using (i/total nb)× 0.2. This method allows for up to 20% of false-positives, but avoids the rejection of true-positives that was manifest with a more stringent correction as the Bonferroni correction. For example, the scospondin mRNA level in the icm13 allele is highly decreased, presumably through non-sense mediated decay: the Bonferroni correction would have rejected this result as not significant, whereas the Benjamini–Hochberg procedure keeps it in the list of potentially interesting genes (see Figure 1—source data 1 for the complete lists). However, this procedure requires a post-hoc validation with an independent technique for each potential gene of interest, as we did by qRT-PCR (see below).

1.2 µg of each RNA sample was retro-transcribed using the Verso cDNA Synthesis Kit (ThermoFischer Scientific, AB1453B) following the provider's instructions. A 1:3 ratio of random hexamer:polydT primers was used to favor mRNA amplification without a strong bias for the 3' end of messengers. qPCR experiments were performed using the LightCycler 480 SYBR Green I Master kit (Roche, 04707516001) on a LightCycler 96 machine (Roche). Each pair of primers (see Key Resource Table) was tested beforehand on a given cDNA stock that was diluted in a series of 4 points. Linearity of CT variation with the cDNA dilution as well as single peaks in the melting curves corresponding to single amplicons were assessed. For each new RNA extraction, we tested that no amplification was detectable when the PCR was performed directly on the RNA stock (-RT). We used the housekeeping gene lsm12b as an internal reference in each experiment. qPCRs were repeated two to three times for each cDNA stock and the average CT was used for further calculations. The relative abundance of the gene of interest was evaluated using the CT comparison formula: 2e^-ΔCT. All results were obtained on at least three biological replicates (except for pkd2l1 mutant extracts), each originating from a single mating.

To characterize the exact evolution of the geometry of the posterior axis in scospondin^icm15/icm15 mutants during the early stages on embryogenesis (Figure 1—figure supplement 1), we performed incrosses of scospondin^icm15/+ parents. Eggs were collected 30 min after mating and maintained at 28.5°C until 19.5 hpf. Embryos were staged according to Kimmel et al., 1995 at 19.5 hpf. Individual embryos were mounted laterally at 20 hpf in 1.5% low-melting point agarose, imaged using a AZ100M macroscope (Nikon), unmounted and kept at 28.5°C until they were remounted for the next stages (22, 24, 28, and 30 hpf). The time window necessary for mounting-imaging-unmounting at each step was short enough (10–15 min) to be neglected.

30 hpf embryos from scospondin^icm15/+ incrosses were treated during 2.5 hr at 28.5°C with E3 medium alone (vehicle), epinephrine hydrochloride (Sigma, E4642) or norepinephrine hydrochloride (Sigma, A7256) both diluted to 3 mM in E3 medium (Figure 4). To avoid light-induced oxidation of epinephrine and norepinephrine, dishes were covered with foil paper during the incubation time. To ensure a proper quantification of ear-to-tail angles, embryos were then fixed over-night in 4% PFA 6-well plates, rinsed 3 times during 45 min in 1X PBS, mounted laterally in 1.5% low-melting point agarose and imaged using a AZ100M macroscope (Nikon). For each embryo, the angle between the ear, the caudal limit of the yolk extension and the tip of the tail (Figure 4C) was quantified using Fiji (Schindelin et al., 2012). Representative traces of the global morphology of the embryos after treatment were drawn for all experimental conditions on one representative clutch (Figure 4B) by linking the center of the eye to the ear and following the dorsal line linking somite boundaries until the tip of the tail.

Tg(pkd2l1:GCaMP5G) embryos were manually dechorionated at 30 hpf, mounted laterally in 1.5% low-melting point agarose, and paralyzed by injecting 1–2 nL of 500 µM alpha-bungarotoxin (Tocris, 2133) in the caudal muscles of the trunk. When required (Figure 5), hindbrain ventricle injections were performed using artificial CSF (aCSF, containing in mM: 134 NaCl, 2.9 KCl, 1.2 MgCl₂, 10 HEPES, 10 glucose, and 2.1 CaCl₂; 290 mOsm.kg⁻¹, adjusted to pH 7.7–7.8 with NaOH) as a vehicle solution. Epinephrine hydrochloride (Sigma, E4642) and norepinephrine hydrochloride (Sigma, A7256) were diluted to 3 mM in aCSF before hindbrain ventricle injections, using a sharp funnel shape needle with an approximate tip diameter of 1–3 µm. Embryos were imaged 20 to 60 min after hindbrain ventricle injections. Calcium imaging was performed at 4 Hz using a spinning disk confocal microscope (Intelligent Imaging Systems, Denver) for 4 min. Imaging was restricted to the region of the spinal cord located above the yolk extension. Regions of interest were manually selected based on an average projection of the time-lapse and according to the dorso-ventral position of the cells. The integrals of normalized ΔF/F signals were calculated using a custom script on MATLAB (see full description in Sternberg et al., 2018). The MATLAB codes used for data analysis are available at Github https://github.com/wyartlab/Cantaut-Belarif-et-al.-2020 (Cantaut-Belarif et al., 2020; copy archived at swh:1:rev:3396034de4726cb8c895a6e43bbc3d774b726fcb).

Whole-cell recordings were performed in aCSF on 30 hpf double transgenic Tg(pkd2l1:GAL4;UAS:mCherry) embryos carrying the scospondin^icm15 mutation and their respective control siblings. Embryos were pinned through the notochord with 0.025 mm diameter tungsten pins. Skin and muscle from two to three segments around segment 10 were dissected out using a glass suction pipette. A MultiClamp 700B amplifier, a Digidata series 1440 A Digitizer, and pClamp 10.3 software (Axon Instruments, Molecular Devices, San Jose, California, USA) were used for acquisition. Raw signals were acquired at 50 kHz and low-pass filtered at 10 kHz. Patch pipettes (1B150F-4, WPI) with a tip resistance of 6–8 MΩ were filled with internal solution containing in (mM: K-gluconate 115, KCl 15, MgCl2 2, Mg-ATP 4, HEPES-free acid 10, EGTA 5 or 10, 290 mOsm/L, pH adjusted to 7.2 with KOH with Alexa 488 at 40 μM final concentration). Holding potential was − 85 mV, away from the calculated chloride reversal potential (E_Cl = - 51 mV).

Embryos were manually dechorionated and euthanized using 0.2% MS 222 (Sigma, E10521) prior to fixation.

To detect the Pkd2l1 channel together with GFP (Figure 2D), 30 hpf embryos were fixed 4 hr at 4°C in 4% PFA and then washed 3 times during 30 min in 1X PBS. Embryos were blocked overnight at 4°C in a solution containing 0.7% Triton, 1% DMSO, and 10% NGS. The primary antibodies were incubated overnight at 4°C in a solution containing 0.5% Triton, 1% DMSO, and 1% NGS and subsequently washed four times during 1 hr in a solution containing 0.5% Triton, and 1% DMSO (washing solution). Secondary antibodies were incubated 3 hr at room temperature in a solution containing 0.5% Triton, 1% DMSO, and 1% NGS. The mix of secondary antibodies was centrifuged 10 min at 10 000 rpm prior to the incubation to pellet unconjugated dyes (supernatant only was used for the incubation). Embryos were then washed 4 times during 2 hr using the washing solution and thereafter overnight at 4°C in 1X PBS.

To detect the RF (Figure 3C), 30 hpf embryos were fixed 4 hr at 4°C in 4% PFA and washed three times during 30 min in 1X PBS. Embryos were then blocked overnight in a solution containing 0.5% Triton, 1% DMSO, and 10% NGS. The primary antibody was diluted in a solution containing 0.5% Triton, 1% DMSO, and 1% NGS and incubated overnight at 4°C. Embryos were subsequently washed 4 times during 1 hr in a solution containing 0.5% Triton, and 1% DMSO (washing solution). Secondary antibodies were incubated 3 hr at room temperature in a solution containing 0.5% Triton, 1% DMSO and 1% NGS. The mix of secondary antibodies was centrifuged 10 min at 10 000 rpm prior to the incubation to pellet unconjugated dyes (supernatant only was used for the incubation). Embryos were subsequently washed 4 times during 2 hr using the washing solution and thereafter overnight at 4°C in 1X PBS.

To detect exogenous (Figure 5—figure supplement 1) and endogenous (Figure 5A) norepinephrine alone (Figure 5—figure supplement 1) or together with GFP (Figure 5A), we used an antibody designed against a Glutaraldehyde-conjugated form of norepinephrine (Millipore, AB120). Thus, 30 hpf embryos were fixed 1.5 hr at 4°C under agitation in a 4% PFA-0.125% Glutaraldehyde-3% sucrose solution to allow reactivity. Embryos were then washed 3 times in 1X PBS during 30 min and the yolks and skin from rostral parts of the trunk were removed. Embryos were blocked overnight at 4°C in a solution containing 0.6% Triton, 1.2% DMSO, and 10% normal goat serum (NGS). An additional blocking step was performed during 3 hr at room temperature the day after. Primary antibodies were incubated in a solution containing 1% Triton, 1% DMSO, and 1% NGS during one day at room temperature plus one night at 4°C. Embryos were then washed extensively 4 times during 2 hr in a solution containing 0.5% Triton and 1% DMSO (washing solution). Secondary antibodies were incubated 3 hr at room temperature in a solution containing 1% Triton, 1% DMSO, and 1% NGS. The mix of secondary antibodies was centrifuged 10 min at 10 000 rpm prior to the incubation to pellet unconjugated dyes (supernatant only was used for the incubation). Embryos were subsequently washed 4 times during 2 hr using the washing solution, 4 times during 1 hr using a solution of 50% washing solution-50% 1X PBS and thereafter overnight at 4°C in 1X PBS.

To detect Adrb2 alone (Figure 6B) or together with GFP (Figure 6C), embryos were fixed 2 hr at 4°C in a 4% PFA-3% sucrose solution and washed 3 times in 1X PBS during 30 min. The yolks and the skin from the rostral part of the trunk were removed, and embryos were blocked overnight in a solution containing 0.6% Triton, 1.2% DMSO, and 10% normal goat serum (NGS). Primary antibodies were incubated at 4°C during 2 days in a solution containing 1% Triton, 1% DMSO, and 1% NGS. Embryos were washed extensively 4 times during 2 hr in a solution containing 0.5% Triton and 1% DMSO (washing solution). Secondary antibodies were incubated 3 hr at room temperature in a solution containing 1% Triton, 1% DMSO, and 1% NGS. The mix of secondary antibodies was centrifuged 10 min at 10 000 rpm prior to the incubation to pellet unconjugated dyes (supernatant only was used for the incubation). Embryos were subsequently washed using the washing solution 4 times during 2 hr, 4 times during 1 hr using a solution of 50% washing solution-50% 1X PBS and thereafter washed in 1X PBS overnight at 4°C.

The following dilutions of primary antibodies were used: rabbit anti-Reissner fiber 1:200 (Didier et al., 1995), rabbit anti-Pkd2l1 1:200 (Sternberg et al., 2018), chicken anti-GFP 1:500 (Abcam ab139170), rabbit anti-norepinephrine 1:100 (Millipore, AB120), rabbit anti-Adrb2 1:200 (ThermoFischer Scientific, PA5-80323). All secondary antibodies were from Molecular Probes and used at 1:500. Systematic omission of the primary antibody confirmed the specificity of the immunostaining results. Zebrafish embryos were mounted laterally in Vectashield Antifade Mounting Medium (Clinisciences, H1000) and imaged on an inverted SP8 DLS confocal microscope (Leica). Images were then processed using Fiji (Schindelin et al., 2012). Maximal Z-projections of 6–9 microns in depth are represented in Figure 2D and Figures 3C, 9-12 microns in depth in Figure 6A and B and Figure Supplement 4B and 18-22 microns in Figure 6C. 3D views shown in Figure 6C were obtained using Imaris.

To produce mRNA, urp2 CDS was amplified from cDNA by PCR and cloned (BamHI - XbaI) into pCS2+. Messenger RNA were produced with the mMESSAGE mMACHINE kit (Ambion). 1 nL RNA-containing solution was injected into 1- to 2 cell stage embryos obtained from scospondin^icm15/+ incrosses. Each clutch was separated into three groups: uninjected and injected either with a control mRNA (100 ng / µL, ras-eGFP encoding for a membrane tagged GFP, [Ségalen et al., 2010]) or with a mix containing control mRNA and urp2 mRNA (100 ng / µL total, 1:1 ratio). To assess for injection quality, GFP-positive embryos were first sorted out at 1 dpf and then scored at 48 hpf for body axis curvature defects.

All values are represented as boxplots (median ± interquartile range) or mean ± SEM (stated for each in the figure legend). All statistics were performed using MATLAB and Excel. In the figure panels, asterisks denote the statistical significance calculated using the appropriate test (stated for each test in the legends): *p<0.05; **p<0.01; ***p<0.001; ns, p>0.05.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We gratefully thank Prof. Ryan S Gray for providing us with Tg(scospondin-GFP) embryos, Prof. Sylvie Schneider-Maunoury and Dr. Christine Vesque for providing us the Ras-eGFP plasmid, Dr. Stephane Gobron for the antibody directed against the Reissner fiber. We thank Monica Dicu and Antoine Arneau for fish care, the ICM.Quant imaging facility for instrument use, scientific and technical assistance and François-Xavier Lejeune for advice on statistics. This work benefited from equipment and services from the iGenSeq (RNA sequencing) and iCONICS (RNAseq analysis) core facilities at the ICM (Institut du Cerveau, Hôpital Pitié-Salpétrière, PARIS, France). We thank for critical feedback all members of the Wyart lab (www.wyartlab.com). This work was supported by, the HFSP Program Grants #RGP0063/2018, the Fondation Schlumberger pour l’Education et la Recherche (FSER/2017) for CW, and the Fondation des Treilles for YCB. The research leading to these results has received funding from the program ‘Investissements d’avenir’ ANR-10- IAIHU-06 (Big Brain Theory ICM Program) and ANR-11-INBS-0011 – NeurATRIS: Translational Research Infrastructure for Biotherapies in Neurosciences.

Animal experimentation: All procedures were performed on zebrafish embryos in accordance with the European Communities Council Directive (2010/63/EU) and French law (87/848). This project is included the APAFIS project #16469-2018071217081175 approved by the French Ministry for Research for the Paris Brain Institute (ICM).

© 2020, Cantaut-Belarif et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.