Notch signalling maintains Hedgehog responsiveness via a Gli-dependent mechanism during spinal cord patterning in zebrafish

Spinal cord patterning is orchestrated by multiple cell signalling pathways. Neural progenitors are maintained by Notch signalling, whereas ventral neural fates are specified by Hedgehog (Hh) signalling. However, how dynamic interactions between Notch and Hh signalling drive the precise pattern formation is still unknown. We applied the PHRESH (PHotoconvertible REporter of Signalling History) technique to analyse cell signalling dynamics in vivo during zebrafish spinal cord development. This approach reveals that Notch and Hh signalling display similar spatiotemporal kinetics throughout spinal cord patterning. Notch signalling functions upstream to control Hh response of neural progenitor cells. Using gain- and loss-of-function tools, we demonstrate that this regulation occurs not at the level of upstream regulators or primary cilia, but rather at the level of Gli transcription factors. Our results indicate that Notch signalling maintains Hh responsiveness of neural progenitors via a Gli-dependent mechanism in the spinal cord.


INTRODUCTION
Patterning of the spinal cord relies on the action of multiple cell signalling pathways with precise spatial and temporal dynamics (Briscoe and Novitch, 2008). Neural progenitors in the spinal cord are organised into discrete dorsoventral (DV) domains that can be identified by the combinatorial expression of conserved transcription factors (Alaynick et al., 2011;Dessaud et al., 2008;Jessell, 2000). Differentiated post-mitotic neurons migrate from the medial neural progenitor domain to occupy more lateral regions of the spinal cord.
To achieve precise patterning, the developing spinal cord employs anti-parallel signalling gradients of Bone Morphogenic Protein (BMP) and Hedgehog (Hh) to specify dorsal and ventral cell fates, respectively (Le Dréau and Martí, 2012). Cells acquire their fates via sensing both graded inputs. This dual signal interpretation mechanism provides more refined positional information than separate signal interpretation (Zagorski et al., 2017). The action of Sonic Hedgehog (Shh) in the ventral spinal cord is one of the most well studied examples of graded morphogen signalling (Briscoe and Therond, 2013;Cohen et al., 2013). In vertebrates, Hh signalling requires the integrity of primary cilia, microtubule-based organelles present on the surface of most cells (Eggenschwiler and Anderson, 2007). In the absence of the Shh ligand, the transmembrane receptor Patched  (Corbit et al., 2005;Rohatgi et al., 2007). This leads to the activation of the Gli family of transcription factors, resulting in expression of downstream target genes such as ptc. Shh thus controls the balance between full-length Gli activators and proteolytically processed Gli repressors (Huangfu and Anderson, 2006;Humke et al., 2010). In mouse, Gli2 is the main activator and its expression does not require active Hh signalling (Bai et al., 2002;Bai and Joyner, 2001).
In zebrafish, Gli1 is the main activator (Karlstrom et al., 2003). Although gli1 is a direct target of Hh signalling, low-level gli1 expression is maintained in the absence of Hh signalling via an unknown mechanism (Karlstrom et al., 2003). It is thought that Hhindependent gli expression allows cells to respond to Hh signals. In the ventral spinal cord, it has been shown that both the level and duration of Hh signalling is critical to the correct formation of the discrete neural progenitor domains along the dorsoventral axis 4 (Dessaud et al., , 2007. However, the temporal dynamics of Hh signalling has been challenging to visualize in vivo due to the lack of appropriate tools. In addition to BMP and Hh signalling, Notch signalling has also been implicated in spinal cord development (Louvi and Artavanis-Tsakonas, 2006;Pierfelice et al., 2011). In contrast to long-range Hh signalling, the Notch signalling pathway requires direct cell-cell interaction, as both receptor and ligand are membrane bound proteins (Kopan and Ilagan, 2009). The Notch receptor, present at the "receiving" cell membrane, is activated by the Delta and Jagged/Serrate family of ligands, present at the membrane of the neighbouring "sending" cell. This leads to multiple cleavage events of Notch, the last of which is mediated by a γ-secretase complex that releases the Notch intracellular domain (NICD).
NICD then translocates to the nucleus and forms a ternary transcription activation complex with the mastermind-like (MAML) coactivator and the DNA binding protein RBPJ.
This activation complex is essential for the transcription of downstream targets, such as the Hes/Hey family of transcription factors (Artavanis-Tsakonas and Simpson, 1991;Pierfelice et al., 2011). Two major roles of Notch signalling in neural development are to generate binary cell fate decisions through lateral inhibition and to maintain neural progenitor state (Formosa-Jordan et al., 2013;Kageyama et al., 2008). However, how Notch signalling interacts with Hh signalling during spinal cord patterning is not clear.
Using zebrafish lateral floor plate (LFP) development as a model, we previously demonstrated that Notch signalling maintains Hh responsiveness in LFP progenitor cells, while Hh signalling functions to induce cell fate identity (Huang et al., 2012). Thus, differentiation of Kolmer-Agduhr" (KA") interneurons from LFP progenitors requires the downregulation of both Notch and Hh signalling. Recent reports provide additional support for cross-talk between these pathways during spinal cord patterning in both chick and mouse (Kong et al., 2015;Stasiulewicz et al., 2015). Notch activation causes the Shhindependent accumulation of Smo to the primary cilia, whereas Notch inhibition results in ciliary enrichment of Ptc1. Accordingly, activation of Notch signalling enhances the response of neural progenitor cells to Shh, while inactivation of Notch signalling compromises Hh-dependent ventral fate specification. These results suggest that Notch signalling regulates Hh response by modulating the localisation of key Hh pathway components at primary cilia.
Here, we determine the interaction between Notch and Hh signalling during spinal cord patterning in zebrafish. Using the photoconversion based PHRESH technique, we show 5 that Notch and Hh response display similar spatiotemporal kinetics. Gain-and loss-of function experiments confirm that Notch signalling is required to maintain Hh response in neural progenitors. Surprisingly, Notch signalling doesn't regulate the Hh pathway at the level of Smo or primary cilia, but rather at the level of Gli transcription factors. Together, our data reveal that Notch signalling functions to control the Hh responsiveness of neural progenitors in a primary cilium-independent mechanism.

Generation of a Notch signalling reporter
Spinal cord patterning is a dynamic process with complex interactions of cell signalling pathways in both space and time. To visualise the signalling events in a spatiotemporal manner, we have previously developed the PHRESH (PHotoconvertible REporter of Signalling History) technique (Huang et al., 2012). This analysis takes advantage of the photoconvertible properties of the Kaede fluorescent protein to visualise the dynamics of cell signalling response at high temporal and spatial resolution. We have utilised the PHRESH technique to visualise Hh signalling dynamics during spinal cord patterning (Huang et al., 2012). To apply the same technique to Notch signalling, we generated a reporter line for her12, a target gene of Notch signalling (Bae et al., 2005). This target was chosen because among other Notch target genes co-expressed with her12, such as her2, her4, and hes5, her12 had the highest level of expression throughout the spinal cord ( Figure S1). By BAC (bacteria artificial chromosome) recombineering, we generated a her12:Kaede reporter by replacing the first coding exon of her12 with the coding sequence for the photoconvertible fluorescent protein Kaede ( Figure 1A). The resulting her12:Kaede BAC contains 135 kb upstream and 63 kb downstream regulatory sequences. The her12:Kaede reporter line faithfully recapitulated endogenous her12 expression ( Figure   1B). This reporter also responded to different Notch pathway manipulations ( Figure 1C-E).
The zebrafish mindbomb mutant is unable to activate Notch signalling due to an inability to endocytose the Delta ligand (Itoh et al., 2003). As expected, the expression of her12:Kaede was completely absent in the spinal cord of mindbomb mutants ( Figure 1C).
Similarly, inhibition of Notch signalling with the small molecule γ-secretase inhibitor LY-411575 (Fauq et al., 2007) completely abolished her12 expression within 4 hours ( Figure   1D and Figure S2). By contrast, ectopic expression of NICD (Notch intracellular domain) using the hsp:Gal4; UAS:NICD line (Scheer and Campos-Ortega, 1999) resulted in upregulation and expansion of the her12:Kaede expression domain ( Figure 1E). These results demonstrate that her12:Kaede is a sensitive reporter for Notch pathway activity in the spinal cord. The combination of small molecule inhibitors and the her12:Kaede reporter allows us to manipulate and monitor Notch signalling dynamics in a tightly controlled temporal manner. 7 Notch and Hh signalling display similar dynamics during spinal cord patterning Using the her12:Kaede reporter of Notch response ( Figure 1) in parallel with the previously described ptc2:Kaede reporter of Hh response (Huang et al, 2012), we can observe the timing and duration of both pathway activities in vivo ( Figure 2A). All responding cells are initially labelled by green-fluorescent Kaede (Kaede green ), which can be photoconverted to red-fluorescent Kaede (Kaede red ) at any specific time (t 0 ). If the cell has finished its signalling response prior to t 0 , only perduring Kaede red will be detected.
Conversely, if the cell begins its response after t 0 , only newly synthesised, unconverted Kaede green will be present. Finally, if the cell continuously responds to the signalling both before and after t 0 , a combination of newly-synthesised Kaede green and perduring Kaede red can be observed and the cell will appear yellow. Thus, Kaede red represents "past response" before t 0 , Kaede green indicates "new response" after t 0 , whereas Kaede red+green corresponds to "continued response" through t 0 ( Based on spatiotemporal maps of Notch and Hh response, we divided the signalling dynamics of spinal cord development into three general phases: "signalling activation" phase from 24 to 42 hpf, "signalling consolidation" phase from 42 to 66 hpf, and "signalling termination" phase from 66 to 78 hpf. In the first "signalling activation" phase, active Notch response occurred along the entire dorsoventral axis of the spinal cord ( Figure 2B, right), 8 while active Hh response constituted roughly the ventral 75% of the spinal cord in a graded manner ( Figure 2B, left). This pattern is consistent with the model that Notch signalling maintains neural progenitor domains, whereas Hh signalling patterns the ventral spinal cord. Interestingly, we found that the signalling response was not entirely homogeneous. In her12:Kaede embryos, the majority of cells showed continued Notch response throughout the "signalling activation" phase, but there were some isolated cells in which Kaede expression was completely absent. In ptc2:Kaede embryos, some cells had terminated their Hh response (marked by Kaede red ), while the majority of cells with the same dorsoventral positioning had continued Hh response. The differential Hh response at the same dorsoventral axis is reminiscent of the differentiation of the lateral floor plate domain (Huang et al., 2012).
In the "signalling consolidation" phase ( Figure 2C), we observed a dramatic remodelling of the response profiles of both pathways. First, there was an extensive increase in Kaede red domains for both reporters, indicating the termination of signalling response in these cells. This loss of response was localised to the ventral and lateral regions for Hh signalling ( Figure 2C, left) and the ventral, lateral and dorsal regions for Notch signalling ( Figure 2C, right). Second, the number of Kaede red cells increased as the "signalling consolidation" phase progressed. Finally, the active signalling domain consolidated into a tight medial region, which sharpened further to encompass 1 -2 cell tiers directly dorsal to the spinal canal ( Figure 2C, 54hpf + 6h).
Finally, during the "signalling termination" phase ( Figure 2D), both active Notch and active Hh responses (Kaede green ) slowly reduced to the basal level, and most of the spinal cord was marked by Kaede red . The active Notch response was notably weaker and restricted to a small medial domain above the spinal canal by 66 hpf before returning to a basal level by 72 hpf. Similarly, active Hh response marked a small medial domain at 66 hpf and 72 hpf, and reduced to a basal level by 78 hpf. After the end of the "signalling termination" phase, both pathways remained at the basal level as spinal cord development progressed ( Figure S3).
Comparison of spatiotemporal signalling profiles reveals that Hh and Notch signalling share similar responsive domains. To examine this directly in the same embryo, we performed double fluorescent in situ hybridisation to visualise her12 and ptc2 expression together during all three signalling phases of spinal cord development ( Figure S4A).
During the signalling activation phase ( Figure S4A, 24 hpf), ptc2 expression constituted 9 the ventral portion of the her12 expression domain, while during the signalling consolidation and termination phases ( Figure S4A, 48 hpf and 72 hpf, respectively), ptc2 and her12 expression was present within the same restricted medial domain. These results confirm that Notch and Hh response is active within the same cells of the spinal cord. Indeed, double labelling with neural progenitor cell marker sox2 showed that the medial domain with continued Notch and Hh response corresponded to the sox2 + neural progenitor domain ( Figure S4B-C).
Together, our analysis reveals that Hh signalling response follows similar spatiotemporal kinetics as Notch signalling response during spinal cord patterning, suggesting that Notch signalling plays a role in maintaining Hh response in neural progenitor cells.

Notch signalling maintains Hh response
To explore the mechanism of interaction between the Notch and Hh signalling pathways, we first performed loss-of-function experiments combining small molecule inhibitors with PHRESH analysis ( Figure 3A). We used the Smoothened antagonist cyclopamine (Chen et al., 2002) Figure 3A). These results suggest that Notch signalling is required for maintaining Hh response, but not vice versa. Interestingly, despite the inhibition of Notch signalling, cells outside of the spinal cord in ptc2:Kaede embryos maintained their normal Hh response, indicated by Kaede green expression (Arrowheads in Figure 3A). This result suggests that regulation of Hh response by Notch signalling is tissue specific.
To confirm these observations from our PHRESH analysis, we performed RNA in situ hybridisation following small molecule inhibition ( Figure 3B). Wild-type embryos were treated with cyclopamine or LY-411575 from 20 to 30 hpf. When Hh signalling was inhibited by cyclopamine, ptc2 expression in the spinal cord was significantly reduced but not abolished, while her12 expression in the spinal cord remained unchanged. By contrast, blocking Notch signalling by LY-411575 resulted in complete loss of both her12 and ptc2 expression in the spinal cord. As seen in the PHRESH analysis, ptc2 expression in cells outside of the spinal cord, such as the somites, was largely intact even after Notch inhibition. Similar results were observed in mindbomb mutants, where ptc2 expression was completely abolished in the spinal cord but not in somites ( Figure S5), confirming the effect of LY-411575 treatment. Together, these results are consistent with our model that Notch signalling regulates Hh response specifically in the spinal cord. It is also interesting to note that cyclopamine treated embryos showed residual levels of ptc2 expression in the spinal cord, whereas LY-411575 treatment completely eliminated ptc2 expression ( Figure   3B). It has been shown that zebrafish smoothened mutants maintain low-level gli1 expression in the spinal cord independent of Hh signalling, similar to cyclopamine treated embryos (Karlstrom et al., 2003). The complete loss of ptc2 expression after Notch inhibition suggests that in contrast to cyclopamine, Notch signalling controls Hh response via a different mechanism, likely downstream of Smo.
In converse experiments, we utilised gain-of-function tools to test the interactions between Notch and Hh signalling (  Figure 4B). Combining our results from the loss-and gain-of-function experiments, we conclude that Notch signalling is required to maintain Hh response, specifically in the spinal cord.

Notch signalling regulates Hh response downstream of Smo
Our results suggest that Notch signalling regulates Hh response during spinal cord patterning. To explore the molecular mechanisms by which Notch controls Hh response, Interestingly, induction of rSmoM2 did cause an expansion of ptc2 expression in the surrounding somites despite Notch inhibition ( Figure 5C). This observation is consistent with our previous experiments and suggests that this Smo independent mechanism of control is specific to the spinal cord.

Notch signalling regulates Hh response independent of primary cilia
Vertebrate canonical Hh signalling requires the integrity of primary cilia (Eggenschwiler and Anderson, 2007). To test whether Notch signalling feeds into the Hh pathway via primary cilia, we utilised the iguana mutant which lacks primary cilia due to a mutation in the centrosomal gene dzip1 (Glazer et al., 2010;Huang and Schier, 2009;Kim et al., 12 2010;Sekimizu et al., 2004;Tay et al., 2010;Wolff et al., 2004). In zebrafish, the complete loss of primary cilia, such as in iguana mutants, results in reduction of high-level Hh response concomitant with expansion of low-level Hh pathway activity (Ben et al., 2011;Huang and Schier, 2009). This expanded Hh pathway activation is dependent on low level activation of endogenous Gli1, but does not require upstream regulators of the Hh pathway, such as Shh, Ptc and Smo (Huang and Schier, 2009) (Figure 6A). Thus, the iguana mutant also allows us to determine whether low level activation of the endogenous Gli1 transcription factor is able to restore Hh response in Notch off spinal cords.

Ectopic expression of Gli1 partially rescues Hh response in Notch off spinal cords
Since low level activation of endogenous Gli1 in iguana mutants is not sufficient to restore Hh response in Notch off spinal cords, we hypothesised that Notch signalling regulates Hh response by maintaining gli1 expression. To test this possibility, we treated wild-type embryos with DMSO, cyclopamine, or LY-411575 at 20 hpf for 10 hours, then assayed for gli1 gene expression at 30 hpf ( Figure 7A). In DMSO treated controls, gli1 expression was present throughout the ventral spinal cord and in the somites. In 13 cyclopamine treated embryos, gli1 expression was dramatically reduced, but a low level remained in the spinal cord, corresponding to Hh-independent gli1 transcription (Karlstrom et al., 2003). domain was also larger ( Figure 7D), a similar phenotype to rSmoM2 induction in DMSO treated embryos. Strikingly, when EGFP-Gli1 induction was followed by LY-411575 treatment, we observed significant ptc2 expression in the spinal cord, although at a slightly lower level compared to DMSO treated hsp:EGFP-Gli1 embryos ( Figure 7D). Critically, the restored Hh pathway activation in LY-411575 treated hsp:EGFP-Gli1 embryos was able to activate olig2 expression, although substantially weaker than the wild-type level ( Figure   7D). Together, these results suggest that, in the spinal cord, Notch signalling regulates Hh response by modulating the Gli1 transcription factor, as ectopic Gli1 can partially rescue the Hh response in Notch off spinal cords. This regulation is partly through transcriptional control of gli1 expression. However, since the ectopic EGFP-Gli1 was unable to rescue the highest level of Hh response and cannot fully restore olig2 expression in Notch off spinal cords, it is likely that Notch signalling also plays additional roles in regulating Gli1 protein -such as its post-translational modification or stability.

DISCUSSION
We provide in vivo evidence for cross-talk between two conserved developmental signalling pathways, Notch and Hh signalling, in the zebrafish spinal cord. Through the PHRESH technique, we observe shared spatiotemporal dynamics of pathway activity throughout spinal cord patterning, highlighting a role for Notch and Hh interaction in neural progenitor maintenance and specification. Using both gain-and loss-of function techniques, we establish a primary cilia-independent mechanism by which Notch signalling permits neural progenitors to respond to Hh signalling via gli maintenance.

Studying cell signalling dynamics using PHRESH
We have previously developed the PHRESH technique to study the dynamics of Hh signalling in vivo (Huang et al., 2012). In this study, we demonstrate the versatility of the PHRESH method by correlating the dynamics of Hh and Notch signalling in vivo using the ptc2:Kaede reporter and a new her12:Kaede reporter. Traditional transcriptional GFP reporters fail to provide temporal information due to GFP perdurance, whereas destabilised fluorescent protein reporters can only provide current activity at the expense of signalling history. By contrast, the PHRESH technique utilises Kaede photoconversion to delineate the cell signalling history in any given time window by comparing newly synthesised Kaede green (new signalling) with photoconverted Kaede red (past signalling).
We envision that PHRESH analysis could be combined with cell transplantation and timelapse imaging to simultaneously analyse cell lineage and signalling dynamics at single cell resolution. Similar approaches can easily be adapted to study other dynamic events by using photoconvertible fluorescent reporters.

Spatiotemporal dynamics of Hh and Notch signalling
Using the PHRESH technique, we created a spatiotemporal map of signalling dynamics for the Hh and Notch pathways during spinal cord patterning. Strikingly, Notch and Hh signalling display similar activity profiles. We have characterised these profiles into three general phases: "signalling activation", "signalling consolidation", and "signalling termination". In the early "signalling activation" phase, Notch signalling is active throughout the spinal cord, while active Hh response occurs in the ventral ~75% of the spinal cord.
During "signal consolidation", the responsive domain of both pathways sharpens into a small medial domain dorsal to the spinal canal; in "signalling termination" the response to both pathways returns to a basal level. Our detailed time course reveals three key features of Notch and Hh signalling dynamics. First, early active Hh signalling shows a graded response with the highest level in the ventral domain, as predicted by the classical morphogen model (Briscoe and Small, 2015). By contrast, active Notch response does not appear to be graded along the ventral-dorsal axis. Second, despite showing the highest level of Hh response early, the ventral spinal cord terminates Hh response earlier than the more dorsal domains. Therefore, the ventral domain shows higher level Hh response for a shorter duration, whereas the dorsal domain shows lower level response for a longer duration. Our observation is reminiscent of the floor plate induction in chick and mouse embryos, where the specification of the floor plate requires an early high level of Hh signalling and subsequent termination of Hh response . Our result suggests that Hh signalling dynamics is also evolutionarily conserved. Third, lateral regions of the spinal cord lose both Notch and Hh response before the medial domains.
As the active signalling response consolidates into the medial domain, so does the expression of sox2, a neural progenitor marker, suggesting that neural differentiation is accompanied by the attenuation of Notch and Hh response. Our observation is consistent with the notion that neural progenitors occupy the medial domain of the spinal cord and that as they differentiate they move laterally.

Notch signalling regulates Hh response
The loss of Hh response is a necessary step for fate specification, as shown in the chick studying floor plate induction  and post-mitotic motor neuron precursors (Ericson et al., 1996). We have previously shown that the time at which cells attenuate their Hh response is crucial for fate specification in the zebrafish ventral spinal cord (Huang et al., 2012). How do neural progenitor cells in the spinal cord maintain their Hh responsiveness until the correct time in order to achieve their specific fates? Three  (Kong et al., 2015;Stasiulewicz et al., 2015), thereby modulating cellular responsiveness to Hh signals. By contrast, we show that in the absence of primary cilia in iguana mutants, the low level Hh response remaining due to constitutive Gli1 activation can be completely blocked by Notch inhibition. This result suggests that Notch signalling can regulate Hh response in a primary cilium independent manner. In zebrafish, Gli1 functions as the main activator downstream of Hh signalling, although Gli2a, Gli2b and Gli3 also contribute to the activator function (Karlstrom et al., 2003;Ke et al., 2008;Tyurina et al., 2005;Vanderlaan et al., 2005;Wang et al., 2013). Indeed, inhibition of Notch signalling abolishes both Hh-dependent and Hh-independent gli1 expression in the spinal cord. Similarly, gli2a, gli2b and gli3 expression in the spinal cord is largely eliminated in Notch off spinal cords. These results demonstrate that Notch signalling controls the transcription or mRNA stability of all members of the Gli family in the spinal cord. It is possible that gli genes are direct targets of Notch signalling, as shown in mouse cortical neural stem cells where N1ICD/RBPJ-κ binding regulates Gli2 and Gli3 expression (Li et al., 2012). Importantly, while ectopic expression of Gli1 from the hsp:EGFP-Gli1 transgene can re-establish Hh response as indicated by ptc2 expression in the Notch off spinal cord, it is unable to fully restore the olig2 motor neuron precursor domain. This finding argues that Notch signalling likely plays additional roles in regulating Gli1 protein level or activity. A similar mechanism has been suggested in Müller glia of the mouse retina, where Notch signalling controls Gli2 protein levels and therefore Hh response (Ringuette et al., 2016). Interestingly, the study by In summary, we demonstrate that Notch and Hh signalling share similar spatiotemporal kinetics during spinal cord patterning and that this dynamic interaction is likely required to maintain the neural progenitor zone. We also provide evidence for a primary ciliumindependent and Gli-dependent mechanism in which Notch signalling permits these neural progenitors to respond to Hh signalling.

Generation of the her12:Kaede BAC transgenic line
To generate the her12:Kaede transgenic line, BAC clone zK5I17 from the DanioKey library that contains the her12 locus and surrounding regulatory elements was selected for bacteria-mediated homologous recombination following standard protocol (Bussmann and Schulte-Merker, 2011). zK5I17 contains 135 kb upstream and 63 kb downstream regulatory sequences of her12. First, an iTol2-amp cassette containing two Tol2 arms in opposite directions flanking an ampicillin resistance gene was recombined into the vector backbone of zK5I17. Next, a cassette containing the Kaede open reading frame and the kanamycin resistance gene was recombined into the zK5I17-iTol2-amp to replace the first exon of the her12 gene. Successful recombinants were confirmed by PCR analysis. The resulting her12:Kaede BAC was co-injected with tol2 transposase mRNA into wild-type embryos and stable transgenic lines were established through screening for Kaede expression.

Photoconversion for PHRESH analysis
All fluorescent imaging was carried out using an Olympus FV1200 confocal microscope and Fluoview software. Photoconversion was carried out using the 405nm laser with a 20x objective. ptc2:Kaede and her12:Kaede embryos at the appropriate stages were anaesthetised with 0.4% tricaine and then embedded in 0.8% low melting agarose. To achieve complete conversion over a large area, a rectangular area of 1000 by 300 pixels was converted by scanning the area twice with 50% 405nm laser at 200 μs per pixel. Following confirmation of Kaede red expression, embryos were recovered in E3 water with phenylthiourea for 6 hours post-conversion before imaging. Appropriate imaging parameters were established using the unconverted region as a reference to avoid over or under exposure of the Kaede green signal. Cross-sections were generated using Fiji-ImageJ software (Schindelin et al., 2012) to create a 3D reconstruction of the image, then "resliced" to yield transverse views of the spinal cord.

Drug treatment
Embryos at the appropriate stage were treated with cyclopamine (Toronto Chemical, 100 μM), LY-411575 (Sigma, 50 μM), or DMSO control in E3 fish water. For PHRESH analysis, embryos were treated from 4 hours prior to the point of conversion until 6 hours post-conversion. To match this, all other drug treatments took place between 20 hpf and 30 hpf.

Heat shock experiments
To induce expression from the heat shock promoter, embryos at the relevant stage were placed in a 2 ml micro-centrifuge tube in a heat block set to 37°C for 30 minutes.
After heat shock, embryos were transferred back into E3 water in a petri dish and recovered at 28.5°C. For drug treatment after heat shock, embryos were transferred directly from the heat shock to E3 water containing the appropriate drug.

Cryosectioning
To obtain transverse sections after whole-mount in situ hybridisation, embryos were cryoprotected with 30% sucrose at 4°C before being embedded in OCT compound (VWR) and frozen in the -80°C freezer. Sections were cut between 10-16μm using a Leica cryostat. Sections were taken from the region of the trunk dorsal to the yolk extension.            her12:Kaede embryos were photoconverted at 48 hpf and imaged 6 hours after. Individual transverse sections generated by 3D reconstruction were prepared into a video. The first frame is the most anterior slice and each subsequent frame moves further posterior through the embryo. The merge (left) and Kaede green (right) channels are shown. The spinal cord is denoted by solid lines and the active signalling domain (Kaede green ) above 27 the spinal canal is indicated by an arrowhead. Note that Figure 2C shows one single slice in the middle of the converted region. Scale bar: 20 μm.