GPR17 is an essential regulator for the temporal adaptation of sonic hedgehog signalling in neural tube development

The neural tube, the embryonic organ of the central nervous system, consists of neural progenitors and post-mitotic neurons arrayed in an orderly manner (Le Dréau and Martí, 2012; Ribes and Briscoe, 2009). During neural tube development, extracellular signalling molecules (‘morphogens’) produced in the signalling centres of the organ provide positional cues to uncommitted neural progenitor cells, and thereby assign the cellular fates in a concentration-dependent manner (Dessaud et al., 2008). Identifying the molecular mechanism by which cells convert positional information into fate determination is one of the major goals of developmental biology (Dessaud et al., 2008).

Sonic hedgehog (Shh) is a signalling molecule expressed in the floor plate (FP) of the neural tube and its underlying mesodermal tissue, the notochord. Shh plays essential roles in the assignment of ventral identities (Dessaud et al., 2008). At the cellular level, the Shh ligand binds to the 12-domain transmembrane protein patched 1 (Ptch1) at the primary cilia. The binding activates another membrane protein, smoothened (Smo). Smo further activates the transcription factors Gli2 and Gli3, which are transported into the nucleus to induce the expression of their target genes (Dessaud et al., 2008). Gli2 and Gli3 have dual activities; in the absence of Shh, the precursor Gli2 and Gli3 proteins associate with the scaffold protein SuFu to form complexes with cullin 3 and protein kinase A (PKA), and are phosphorylated and proteolytically processed into their repressor forms (Niewiadomski et al., 2014; Tempe et al., 2006; Tuson et al., 2011; Wang and Li, 2006). In response to the Shh ligand, these complexes are dissociated, ubiquitylation is perturbed and Gli2 and Gli3 are converted into their active forms (Niewiadomski et al., 2014; Pan et al., 2006; Wang et al., 2000).

In the spinal cord, Shh forms a ventral-to-dorsal gradient with the highest level in the FP, and is required for the assignment of at least five ventral neural progenitor domains in addition to FP (Dessaud et al., 2008; Ribes and Briscoe, 2009). The ventral domains include p0-p2, pMN and p3, which are arranged in this order from dorsal to ventral, and can be defined by transcription factors expressed in a domain-specific manner. For example, two transcription factors, Olig2 and Nkx2.2, are expressed in the pMN and p3 domains, respectively, and are essential for these identities (Briscoe et al., 1999; Holz et al., 2010; Lu et al., 2000). During the neural tube development, Olig2 responds to Shh and is expressed first; the expression of Nkx2.2, which is also regulated by Shh, begins later (Dessaud et al., 2007). The higher concentration of Shh allows the quicker switch from an Olig2- to Nkx2.2-expressing state in the progenitor cells (Dessaud et al., 2010). In this manner, the progenitor cells exposed to a higher Shh concentration form the Nkx2.2-positive p3 domain at a more ventral position than the Olig2/pMN domain. Thus, the Shh gradient is reflected in the dynamic expression of the transcription factors, and various cell types are formed in the spinal cord (Balaskas et al., 2012).

The Gli activity partially reflects the concentration of Shh at the initial step of neural differentiation (Dessaud et al., 2010). After it is initially elevated, however, Gli activity decreases over time (Balaskas et al., 2012; Dessaud et al., 2010). Moreover, the half-life of the signal intensity determines the kinetics of the conversion of the downstream transcription factors that characterise the neural domains. Therefore, the dynamic change of Gli activity, which is termed temporal adaptation, is important for correct pattern formation of the neural tube (Balaskas et al., 2012; Dessaud et al., 2010, 2007; Stamataki et al., 2005).

Multiple mechanisms have been proposed to explain this adaptation (Cohen et al., 2015). One model involves upregulation of the Ptch1 gene. As Ptch1 is a negative regulator of the Shh signal, accumulation of Ptch1 depletes Shh and consequently decreases its intracellular signalling activity (Dessaud et al., 2007). In addition, the expression level of the transcription factor Gli2 decreases over time, and the transduction of the Shh signal is thus reduced as the neural development proceeds (Cohen et al., 2015). Importantly, this adaptation takes place only in the context of the neural differentiation, and not in cultured cells, such as NIH3T3 fibroblast cells (Cohen et al., 2015; Humke et al., 2010; Ribes and Briscoe, 2009), suggesting that some key genes are specifically induced in a developmental context. However, such regulators have yet to be identified.

In parallel with Shh activity, intracellular cyclic AMP (cAMP) level and the activity of the downstream mediator PKA have crucial roles in neural tube pattern formation (Epstein et al., 1996; Tuson et al., 2011). Absence of PKA activity affects the subcellular localisation and processing of Gli proteins at the cellular level, and results in the expansion of the ventral domains of the neural tube (Tuson et al., 2011). Likewise, the activity of adenylyl cyclases 5 and 6 (AC5 and AC6, respectively), which encourage the production of cAMP from adenosine triphosphate (ATP), affects the Gli activity and determination of the ventral cell fate (Moore et al., 2016; Vuolo et al., 2015). Moreover, G-protein-coupled receptors (GPCRs) have been suggested to control the neural tube pattern formation through regulating the intracellular cAMP level (Humke et al., 2010; Pusapati et al., 2018). GPCRs couple with G proteins comprising Gα, Gβ and Gγ subunits. G-proteins are categorised into several subclasses based on the type of the Gα protein; Gα_s and Gα_q can potentiate the cAMP levels, whereas Gα_i decreases this level (Rosenbaum et al., 2009). With respect to the neural tube development and Shh signal, GPR161 has been suggested to interact with Gα_s and negatively regulates the ventral neural identities by elevating cAMP levels (Mukhopadhyay et al., 2013). Conversely, GPR175 interacts with Gα_i protein and enhances the Shh signal (Singh et al., 2015). Therefore, the GPCR/cAMP/PKA axis is a critical regulatory pathway in the neural tube pattern formation. However, the molecular mechanisms by which cAMP levels and PKA activity are involved in the temporal changes of Gli activity remain to be revealed.

In this study, we found that the PKA activity changes over time in the ventral neural progenitor cells, and thus hypothesised the existence of the regulatory molecule(s) for this change. As cAMP and PKA levels are often regulated by GPCRs, we searched for the GPCRs induced by Shh, and performed a quantitative RT-PCR (RT-qPCR)-based screen in chick neural explants. We consequently identified Gpr17 as a candidate gene. Gpr17 is induced by Shh, but negatively regulates the Shh signal. GPR17 also functions as a negative regulator in neural differentiation of mouse embryonic stem (ES) cells. Taken together, our findings demonstrate that GPR17 is a negative regulator of Shh signal in multiple cellular contexts.

For understanding the mechanisms by which the temporal adaptation (Cohen et al., 2015; Dessaud et al., 2010, 2007) against the Shh signal takes place in the neural progenitor cells, we reasoned that the PKA activity changes over time (Epstein et al., 1996; Tuson et al., 2011; Vuolo et al., 2015). To address this hypothesis, we performed an ex vivo analysis using chick intermediate neural explants. We isolated neural progenitor cells from the preneural tube area (Delfino-Machin et al., 2005) of HH stage 9 (Hamburger and Hamilton, 1992) chick embryos and cultured the cells for 24 h. In the explants cultured with the low Shh concentration (Shh^L; see Fig. S1B-H and Materials and Methods), the majority of the cells expressed Olig2 with a small subset of Nkx2.2-positive cells (Fig. 1A,C,E,G). However, in the explants cultured with the PKA inhibitor H-89 together with Shh^L, the Nkx2.2-positive cells dominated the explants (Fig. 1B,D,F,G). As Nkx2.2 is assigned in the more ventral region in the neural tube (Ribes and Briscoe, 2009), this finding suggests that the neural progenitor cells acquired a more ventral identity by blockade of PKA activity. Although it has been shown that the forced downregulation of the PKA activity ventralises the neural identities (Epstein et al., 1996; Hammerschmidt et al., 1996; Vuolo et al., 2015), the explants treated with the sole H-89 did not show any expression of Olig2 or Nkx2.2 (A.Y., N.S., unpublished), suggesting that H-89 does not have the function to mimic the Shh activity at least at this concentration.

Based on this observation, we investigated the relationship between the PKA activity and the temporal change of the intracellular Shh signalling activity, and measured the Gli activity at different time points. We prepared pools of explants transfected with the GBS-Luc reporter construct (Sasaki et al., 1999), which harboured the luciferase gene driven by the Gli binding sequence (GBS). We incubated the explants with Shh^L or Shh^L with H-89 for 6 or 24 h, and measured the Gli activities.

Consistent with the previous reports (Dessaud et al., 2010, 2007), we found upregulated Gli activity 6 h after the starting of the culture (Fig. 1H, lane 1). Importantly, the activity level was comparable with the explants treated with Shh^L and H-89 (Fig. 1H, lane 2), suggesting that the effect of H-89 was not evident at this time point. By contrast, at 24 h, the Gli activity was kept high in the explants with Shh^L and H-89, while the activation was almost undetectable in the explants treated only with Shh^L (Fig. 1H, lanes 3 and 4). This observation suggests that the Gli activity in the neural explants is sensitive to the PKA inhibitor only when cells are at the late (24 h) time point.

It has been suggested that Gli3, a transcription factor that mediates Shh signalling (Litingtung and Chiang, 2000), is destabilised, and the total amount of the full-length and repressor forms of Gli3 is therefore decreased by treatment with Shh (Humke et al., 2010). Therefore, we determined whether the Gli3 levels were affected by the treatment with Shh^L or by the culture period in the explants. We cultured the explants either in the absence or in the presence of Shh^L for different time periods, and examined them by western blots using an anti-Gli3 antibody (Fig. 1I). When cells were treated with Shh^L for 6 h, both the full-length (Gli3FL) and repressor (Gli3R) forms of Gli3 became less abundant, which is consistent with the previous report that Gli3 is destabilised by Shh (Humke et al., 2010) (Fig. 1I, lanes 1 and 2, Fig. 1J). The same trend was found when H-89 was added with Shh^L (Fig. 1I, lanes 2 and 3, Fig. 1J). By contrast, in the explants cultured with Shh^L for 24 h, the processing and the total amount of Gli3 were more similar to those of the untreated explants (Fig. 1I, lanes 4 and 5, Fig. 1J), suggesting that the temporal adaptation corresponds to the dynamic change of the Gli3 processing (Fig. 1H, lane 3). However, this reversion of Gli3 processing by 24 h-Shh^L was perturbed by H-89, suggesting that the temporal Gli3 processing is sensitive to the PKA activity. The Gli3 processing did not significantly change at 6 and 24 h in the untreated explants (Fig. 1I, lanes 3 and 6, Fig. 1J), suggesting that the PKA activity changes only when neural progenitor cells are exposed to Shh. Taken together, the PKA activity changes over time and confers the temporal adaptation of the Shh signal during ventral neural differentiation.

Based on the observation in Fig. 1, we supposed the existence of the PKA activity-upregulating molecule(s) the expression of which is dependent on Shh. To identify such molecules, we focused on GPCRs, as many GPCRs regulate the activity of PKA and further modify the activities of Gli transcription factors (Moore et al., 2016). With this in mind, we selected genes encoding GPCRs that can bind to Gα_s or Gα_q, and performed a screen that combined RT-qPCR and in situ hybridisation (Fig. S2) to isolate the GPCR that responds to Shh. Based on its expression pattern and responsiveness to Shh, we identified the purinergic-like GPCR (a group of GPCRs that take purine as ligand) Gpr17 as a candidate.

We first investigated the spatio-temporal expression of Gpr17 in the chick spinal cord, and compared it with that of Olig2 and Nkx2.2 (Fig. 2A-F). In situ hybridisation analysis revealed that Gpr17 was not expressed at HH stage 10, immediately after the neural tube was formed and Olig2 started to be expressed (Fig. 2A,B). At HH stage 14, GPR17 was expressed broadly in the ventral region covering the area where Olig2 and Nkx2.2 are strongly expressed (Fig. 2C,D). At HH stage 24, the Gpr17 expression formed a dorsal-to-ventral gradient in the progenitor region of the neural tube, with the lowest level at the dorsal region and highest level at Olig2- and Nkx2.2-positive regions, (Fig. 2E,F). In addition, although the expression of Gpr17 was found in a wide area in the ventral region at the beginning of the neural tube development (Fig. 2C), the floor-plate expression halted as development progressed (Fig. 2E,F).

In the mouse spinal cord, GPR17 expression was evident in the ventral region, including pMN and p3 areas at embryonic day 10.5 (E10.5) when neural tube pattern formation is taking place, as confirmed both by in situ hybridisation and immunofluorescence (Fig. S3A,B). This expression pattern was essentially the same as in the developing spinal cord of chicks.

To investigate the relationship between Shh and Gpr17 expression, we cultured the intermediate neural explants in the presence of Shh^L or Shh^H (see Fig. S1 and Materials and Methods). Expression levels of Gpr17 were measured by RT-qPCR, every 12 h up to 48 h (Fig. 2G). At 12 h, Gpr17 expression was higher in explants treated with Shh^L or Shh^H than in untreated explants (Fig. 2G), and this trend continued at 24 and 36 h. By contrast, at 48 h, expression was downregulated in explants treated with Shh^H, but remained high in explants treated with Shh^L. This finding is consistent with the in vivo expression pattern of abolished expression in the floor plate (Fig. 2E,F).

Next, we explored the upstream transcription factors that induce Gpr17 expression. As Gpr17 requires Olig1 for its expression (Chen et al., 2009) and is one of the target genes of Olig2 during oligodendrocyte development (Ou et al., 2016), we assessed whether the overexpression of an Olig-type transcription factor could induce the Gpr17 transcript. To address this, we prepared intermediate neural explants that were electroporated with the Olig2 expression plasmid, and determined the Gpr17 expression level by RT-qPCR. Gpr17 expression was elevated at 24 h, suggesting that Olig2 alone is sufficient to induce it (Fig. 2H, lanes 1 and 2). By contrast, FoxA2, which characterises the floor plate, downregulated Gpr17 (Fig. 2E, lanes 1 and 3), which was consistent with the observation that the Gpr17 expression was lower in the floor-plate region. Znf488 (also known as Zfp488), which is expressed in the same regions as Olig2 (Rehimi et al., 2016; Wang et al., 2006), did not have a significant effect on GPR17 expression (Fig. 2H, lanes 1 and 4). These data supported the idea that Gpr17 expression is targeted by Olig2 (Ou et al., 2016), and becomes exclusive to the floor plate as development progresses.

We further investigated the requirement of Olig2 for the expression of Gpr17. We prepared the neural explants electroporated with the construct encoding the dominant-negative version of Olig2 (Zhou and Anderson, 2002), and cultured with Shh^L for 24 h (Fig. 2I). As a result, the GPR17 expression was downregulated, suggesting that the GPR17 expression requires Olig2 (Fig. 2I).

The collective results suggested that GPR17 expression is mainly regulated by Shh signal and its downstream transcription factor Olig2. This intriguing regulation of Gpr17 expression prompted us to further investigate its role in neural tube development.

We sought to characterise the molecular role of GPR17 in the Shh signalling pathway, and first analysed the subcellular localisation of GPR17. As Shh signal is mediated the membrane proteins localised to the cilia (Sasai and Briscoe, 2012), we particularly focused on the correlation with the cilia. However, in NIH3T3 cells, GPR17 was not accumulated at the cilia in the normal culture condition (Fig. S4A-C). This localisation did not change upon treatment with Shh (Fig. S4C). Moreover, in the mouse neural tube, GPR17 was not localised to the cilia of the apical surface. (Fig. S3C-G). Therefore the subcellular localisation of GPR17 does not have any correlation with the cilia.

Next, we attempted to examine the effect of GPR17 in the Shh intracellular signalling pathway. For this purpose, we prepared control or Gpr17-transfected NIH3T3 cells to assay the expression levels of Ptch1 and Gli1, which are primary target genes of the Shh signal (Cohen et al., 2015) (Fig. 3A). The mere transfection of GPR17 did not affect the expression levels of these genes (Fig. 3A). However, when the cells were treated with purmorphamine (Pur), an agonist of Smo, for 24 h, the extent of the upregulation in Gpr17-transfected cells was significantly lower than in the control cells (Fig. 3A). Given that Pur targets the Smo protein (Briscoe, 2006), this result suggests that GPR17 perturbs the intracellular Shh signal at the downstream level of Smo.

We next sought to determine whether the amount of Gli3FL and Gli3R was affected by the overexpression of Gpr17. We treated the control or Gpr17-transfected cells with Pur, and examined them by western blots using the anti-Gli3 antibody (Fig. 3B). Without Pur treatment, the ratio of the full-length over the repressor form of Gli3 was not changed by the transfection of GPR17 (Fig. 3B, lanes 1 and 2). By contrast, when cells were treated with Pur, Gli3FL became less abundant (Humke et al., 2010) (Fig. 3B, lane 3, Fig. 3C). However, in the GPR17-expressing cells, the change in Gli3FL was less than that of the control cells with Pur (Fig. 3B, lane 4, Fig. 3C), suggesting that the effect of Pur on Gli3 was perturbed by the presence of GPR17.

We further measured the cAMP levels in the cells, because modification of Gli3 proteins depends on the intracellular cAMP levels (Wang and Li, 2006). As a result, we found a higher cAMP level in GPR17-overexpressed cells than in control cells (Fig. 3D). Furthermore, a reporter assay measuring the activity of the cAMP-responsive element-binding region (CREB) revealed that the transfection of the Gpr17 expression plasmid raised the CREB activity, confirming that the intracellular cAMP level was raised upon the overexpression of Gpr17 (Fig. 3E). In cAMP-related assays (Fig. 3D,E), Gpr161 was used as a positive control (Mukhopadhyay et al., 2013). Taken together, these findings demonstrated that GPR17 functions as a negative regulator of the Shh signalling pathway by upregulating the cAMP level.

To investigate the activity of GPR17 in neural tube development and pattern formation, we overexpressed Gpr17 by in ovo electroporation in the neural tube of the chick embryos, and monitored its effect on dorsal-ventral pattern formation. However, pattern formation was not significantly altered (Fig. S5A-C″). This could be because a trigger (i.e. a ligand) is necessary to activate GPR17-mediated signalling, and we speculated that the combination of GPR17 with its agonist, MDL29951 (Hennen et al., 2013), would embody the phenotype of the Gpr17 overexpression.

To confirm that GPR17 and MDL29951 act synergistically, we administered a submaximal level of MDL29951 onto the embryos, and found that the neural tube patterning was not changed (Fig. S6A-C). By contrast, when MDL29951 was administered to the embryos in which GPR17 was overexpressed, the expression of Olig2 and Nkx2.2 was decreased in the electroporated cells, with the expansion of the dorsal marker Pax7 (Fig. S6D-F″). This finding suggests that GPR17 acts synergistically with MDL29951 in a cell-autonomous manner.

To find the effect of the GPR17-related signal in the dorsal-ventral patterning of the whole neural tube, we employed a new culture system, in which embryos were cultured ex ovo (Psychoyos and Finnell, 2008). The embryos were cultured in the absence or in the presence of an effective level of MDL29951. As a result, the ventral neural domains characterised by Olig2 and Nkx2.2 were significantly reduced (Fig. 4A-D), while the Pax7-expressing area was expanded (Fig. 4E,F), with the decreasing expression of Ptch1 (Fig. 4G,H), suggesting that the ventral neural differentiation was abrogated. By contrast, FoxA2 and Shh expression was intact (Fig. 4I-L), suggesting that perturbation of the dorsal-ventral pattern has arisen by a change in the characteristics of the neural progenitor cells, rather than by alteration of the expression level of Shh.

Next, to reveal the relationship between the intracellular Shh signal and GPR17 more directly, we conducted an explant analysis. Because the expression of Gpr17 is induced by Shh and GPR17 negatively regulates the Shh signal activity at the intracellular level (Figs 2G and 3A), we reasoned that GPR17 is involved in temporal regulation of Gli activity (Balaskas et al., 2012; Dessaud et al., 2010, 2007). To test this hypothesis, we measured the Gli activity over time. At 6 h, the Gli activity was comparable in the explants treated with Shh^H in the absence and presence of 30 µM of MDL29951 (Fig. 4M, lanes 1 and 2). However, at the 24 h time point, the explants added with MDL29951 along with Shh^H displayed a significantly reduced level of Gli activity than those without MDL29951 (Fig. 4M). These data suggest that the temporal Gli activity was perturbed by GPR17 and its related signals.

As the Gli activity was decreased in the Shh^H explants by MDL29951, we expected that the neural identities were also altered. Shh^H treatment of the explants for 24 h simulated the differentiation into ventral neural progenitor cells; Nkx2.2 expression was detectable in the majority of cells, and Olig2 in a smaller subset (Fig. 4N,P,R,T). By contrast, when the explants were treated with MDL29951 along with Shh^H, the number of cells expressing Nkx2.2 decreased, whereas Olig2-positive cells increased (Fig. 4O,Q,S,T). Given that the Olig2-expressing cells can be induced by a lower concentration of Shh (Dessaud et al., 2007), the results are consistent with those obtained from the measurement of Gli activity, and Shh was partially inhibited by the GPR17-associated signalling. This tendency was confirmed by using the Smo agonist Pur (500 nM), which induced a large number of Nkx2.2-positive cells. When used in combination with MDL29951, the numbers of Nkx2.2-positive cells decreased and Olig2-expressing cells instead increased (Fig. S6G-M). This result supports the idea that GPR17-related signal negatively regulates the effect of Shh.

Next, the same analyses were performed with the Shh^L-treated explants. As in the case of Shh^H, the Gli activity was comparable at 6 h (Fig. S6N). In addition, at 24 h, the activity in the MDL29951-treated Shh^L explants was at a similar level to those without MDL29951. Consistently, the assignment of the neural identities was not altered; Olig2-expressing cells dominated regardless of the MDL29951 treatment (Fig. S6O-U). Given that the Gli activity is downregulated quickly in Shh^L conditions (Dessaud et al., 2010, 2007), this result suggests that MDL29951 did not exert a significant blocking effect on the Shh^L, and the collective results above suggest that GPR17-related signal perturbed the Gli activity in a time-dependent manner.

It is unlikely that the alteration of the pattern formation is mediated by the programmed cell death, as a TdT-mediated dUTP nick end-labelling (TUNEL) assay also failed to detect a significantly increased number of positive signals in the explants (Fig. S7A-C).

We attempted to further investigate whether the activation of the GPR17-related signal correlated with the elevation of the intracellular cAMP level in the intermediate neural explants. Consistent with the previous observation (Hammerschmidt et al., 1996; Humke et al., 2010; Moore et al., 2016), the cAMP levels in the explants treated with Shh for 24 h showed a lower cAMP level than in the control explants (Fig. 4U, lanes 1 and 2). However, in the explants treated with MDL29951 in combination with Shh, this level was restored (Fig. 4U, lanes 1 and 3), suggesting that MDL29951 perturbed the Shh-mediated decrease in cAMP levels.

The regulation of cAMP levels by the GPR17-related signal occurred at upstream of PKA, because the ventralising effect induced by the dominant-negative PKA (dnPKA) could not be reverted by the coelectroporation with GPR17 followed by the administration of MDL29951, as confirmed by the expression of Nkx2.2 (Fig. S7D-E″). Together, these findings suggested that the GPR17-mediated signalling pathway negatively regulates Shh activity in the context of neural tube pattern formation through the upregulation of intracellular cAMP levels.

To further investigate the functions of GPR17 in the development of neural tube pattern formation, we employed a loss-of-function approach. For this purpose, we designed two siRNAs targeting GPR17 (si-Gpr17-1 and si-Gpr17-2), and carried out in ovo electroporation into the neural tube.

We first electroporated si-Gpr17 into the neural tube at HH stage 12. At 48 h post-treatment, we observed aberrant expansion of the ventral neural domains characterised by Olig2, Nkx2.2 and FoxA2 expression and of the motor neuron domain, identified by Islet1 expression (Fig. 5A-H″). Moreover, probably because the low level expression of Gpr17 was perturbed by si-Gpr17, the area positive for Pax7 expressed in the dorsal progenitor domains was diminished (Fig. 5I-J″). This finding suggests that GPR17 per se is a negative regulator of Shh signal. This aberrant expression was abolished by co-electroporation of mouse Gpr17, which is not targeted by si-Gpr17 (Fig. S5D-F″), suggesting that the phenotype observed in si-Gpr17 transfection was due to downregulation of Gpr17 expression. Furthermore, co-electroporation of Gpr161, another negative regulator of the Shh signalling pathway (Mukhopadhyay et al., 2013), did not abolish the phenotypes caused by si-Gpr17, confirming the specificity of si-Gpr17 (Fig. S5G-I″). The expansion of ventral identity was also observed when si-Gpr17-2 was electroporated into neural tubes (Fig. S8), validating the phenotype. Thus, GPR17 was demonstrated to be a negative regulator of the intracellular Shh signalling, and is essential for proper pattern formation in the neural tube.

Finally, we investigated whether GPR17 plays an essential role for the temporal regulation of Gli activity (Balaskas et al., 2012; Dessaud et al., 2010, 2007). First, we asked whether the intracellular Shh signal activity was aberrantly upregulated by preparing intermediate neural explants electroporated with si-Gpr17 then treating them with Shh^L. Although Olig2-expressing cells were predominant in the si-control electroporated explants (Fig. 6A-D,I), a significantly higher population of Nkx2.2-positive cells appeared at 24 h in the si-Gpr17-electroporated explants (Fig. 6E-H,I), suggesting that the intracellular Shh signal activity was upregulated when Gpr17 expression was reduced. This observation was supported by the effect of treatment with pranlukast, a chemical antagonist of GPR17 (Hennen et al., 2013; Parravicini et al., 2010); explants treated with 10 µM pranlukast in combination with Shh^L tended to contain larger proportions of Nkx2.2-positive cells than controls, suggesting that perturbation of GPR17 caused the more ventral identity (Fig. S9A-G).

We next sought to analyse dynamic control of the intracellular Shh signal activity during ventral neural differentiation. To achieve this, we performed luciferase reporter assays using GBS-Luc at various time points. We prepared explants electroporated with GBS-Luc with either si-control or si-Gpr17, and then measured Gli activity at a series of time points from 6 to 48 h after the initiation of the culture. In si-control-electroporated explants treated with Shh^L, Shh activity gradually decreased over time, peaking at 6 h and becoming undetectable by the 48 h time point (Fig. 6J, blue line), consistent with previous reports (Dessaud et al., 2010, 2007). On the other hand, in explants with si-Gpr17, Gli activity at 6 h was comparable with that of the si-control explants, but the activity was significantly higher at 24 h and still detectable at 36 h (Fig. 6J, red line).

We next attempted to confirm that the dynamic Gli activity reflects the gene expression. We therefore cultured the explants electroporated with si-control or si-Gpr17 in the presence of Shh^L for different time periods, and measured the expression levels of Ptch1 and Gli1 by RT-qPCR. As the result, the expression peaked at 12 h and then decreased quickly at 24 h in si-control-electroporated explants (Fig. 6K, Fig. S9H, blue lines). However, in the si-Gpr17-electroporated explants, although the expression at 12 h was comparable, the decrease was delayed and the expression level was significantly higher at 24 h (Fig. 6K, Fig. S9H, red lines). The successful perturbation of Gpr17 expression was confirmed by RT-qPCR against the GPR17 primers (Fig. S9I). This result supports the idea that the dynamic Gli activity was affected by the perturbation of GPR17 expression.

We then asked whether the dynamic Gli activity corresponds to the processing of the Gli3 protein. For this, we performed western blots with the anti-Gli3 antibody in the chick explants electroporated with either si-control or si-Gpr17. At 6 h, the Gli3 protein processing was significantly affected by the treatment with Shh both in si-control- and si-Gpr17-transfected cells (Fig. 6L, left panel, Fig. 6M). By contrast, at 24 h, although the si-Gpr17-electroporated explants did not change the Gli3 processing in the absence of Shh (Fig. 6L, lanes 5 and 6, Fig. 6M), they contained significantly more destabilised Gli3 than did the si-control explants upon the treatment with Shh (Fig. 6L, lanes 7 and 8, Fig. 6M). This suggests that GPR17 affects the temporal processing of Gli3. Together, these findings indicate that GPR17 is an essential upstream factor that controls the dynamic change between the full-length and repressor forms of Gli3, and thus regulates the temporal change in intracellular Shh signalling activity.

We finally assessed whether GPR17 functions are conserved in different organisms by investigating its functions in aspects of mouse development. For this analysis, we examined the directed neural differentiation of the ES cells, because the effects of treatments on gene expression are easily evaluated in this system. Mouse ES cells were differentiated into neural subtypes of pMN, p3 and floor plate using different combinations of retinoic acid (RA) and SAG, as described previously (Fig. S10A) (Kutejova et al., 2016). First, we evaluated GPR17 expression by RT-qPCR in each neural subtype. The results revealed significantly higher GPR17 expression in pMN and p3 cells than in FP 5 days after the start of differentiation (day 5) (Fig. 7A), suggesting that expression can be recapitulated in neural cells differentiated from ES cells.

By using this differentiation protocol, we analysed the effect of GPR17-related signal in the neural fate determination. During the differentiation into pMN, MDL29951 was added at day 3.5 and gene expression was evaluated at day 5.

In the pMN condition, the majority of cells expressed Olig2, with a small subset of Pax6-positive cells, as demonstrated by immunofluorescence (Fig. 7B,D,F). By contrast, the cells treated with MDL29951 displayed significantly larger numbers of Pax6-positive cells with reduced number of Olig2-positive cells (Fig. 7C,E,F), suggesting that the Shh signal was partially perturbed and cells obtained a more dorsal identity.

Conversely, we sought to analyse the essential function of GPR17 in neural differentiation and subtype determination, and transfected two different si-RNAs (si-Gpr17-1 and si-Gpr17-2) targeting Gpr17 in ES cells. The efficient knockdown of Gpr17 expression was validated by RT-qPCR (three experiments; Fig. S10B).

Under these conditions, ES cells were differentiated into pMN cells, and gene expression was evaluated on day 5. In the pMN condition, the majority of cells expressed Olig2, with a small subset of Nkx2.2-positive cells, as demonstrated by immunofluorescence (Fig. 7G,J,M,P). By contrast, treatment with si-Gpr17 increased the Nkx2.2-expressing cells (Fig. 7H,I,K,L,N-P). This observation suggests that si-Gpr17-treated cells became more susceptible to the Shh ligand and tended to differentiate into the more ventral identity of p3 (Fig. S10C-E), for which a higher level of Gli activity was required (Kutejova et al., 2016).

We next verified this tendency by RT-qPCR. In cells treated with si-Gpr17s, the Nkx2.2 expression was higher than in si-control-transfected cells at day 5, confirming the results obtained by the immunofluorescence (Fig. 7Q). Furthermore, the expression level of Ptch1, a target gene of the Shh signal, was higher in si-Gpr17 cells than in the controls (Fig. 7Q). These results were consistent with the findings obtained in the analyses of chick embryos. These findings suggest that mouse GPR17 is expressed in the pMN and p3 cells differentiated from ES cells, as in the chick embryos, and functions as a modulator for the ventral identities of the neural cells.

In this study, we have demonstrated that the temporal change of the PKA activity in the neural progenitor cells confers the temporal adaptation against Shh (Fig. 1), and we isolated GPR17 as a modulator for this temporal activity (Fig. 3D,E,U). Although its expression is induced by Shh, GPR17 negatively regulates the Shh signal, thereby affecting the dynamicity of the Gli activity (Dessaud et al., 2010, 2007). Negative regulation of the intracellular Shh signal by GPR17 is conserved in the mouse, as demonstrated by a system involving neural differentiation of ES cells (Fig. 7).

GPR17 was initially recognised as one of the genes that responds to neural tube injury, brain damage or pathological situations (Ceruti et al., 2009; Ciana et al., 2006; Lecca et al., 2008). A previous study using the mutant mice revealed that the deficiency of the Gpr17 gene caused the earlier differentiation and excessive production of oligodendrocyte cells, suggesting that GPR17 determined the timing of oligodendrocyte differentiation (Chen et al., 2009). Subsequent studies have established the idea that GPR17 regulates the proliferation and differentiation of oligodendrocyte precursor cells (Lu et al., 2018; Merten et al., 2018). Given that oligodendrocyte differentiation is supported by Shh (Oh et al., 2005), the phenotypes caused by the alteration of the GPR17 expression may be mediated by the modulation of the Gli activity (Chen et al., 2009).

Although GPR17 has been suggested to be a receptor for the uracil nucleotides and cysteinyl leukotrienes (cysLTs) (Ciana et al., 2006; Fumagalli et al., 2016; Lecca et al., 2008), the actual ligand(s) that bind to GPR17 during development have not been identified. In this study, we used MDL29951 to experimentally activate GPR17 (Figs 4B,D,F,H,J,L,O,Q,S,U, S6A-F″,H,J,L,N,P,R,T, S7B). Although MDL29951 is recognised to be an agonist of GPR17 (Hennen et al., 2013) and has been used to analyse the function of GPR17 (Ou et al., 2016), it is also recognised as an antagonist of the N-methyl-D-aspartate (NMDA) receptor (Heppenstall and Fleetwood-Walker, 1997; Jansen and Dannhardt, 2003). However, as we demonstrated, MDL29951 synergistically acted with GPR17 at least in the context of the dorsal-ventral patterning (Fig. S4).

The in vivo functions of GPR17 in the mouse CNS development remain unknown but it seems that GPR17 is not crucial for the embryonic development (Chen et al., 2009; Lu et al., 2018; Merten et al., 2018), as Gpr17-deficient mice are viable. This is probably due to the redundant roles of the multiple factors expressed in the mouse neural tube. One such candidate is adenylyl cyclase 5 (AC5; Acdy5), which catalyses the dissociation of ATP to make cAMP (Moore et al., 2016; Vuolo et al., 2015). As we found that Acdy5 expression was induced by Shh both in chick and mouse neural explants (Fig. S11), the regulatory mechanism of AC5 (and its related factor AC6) is similar to that of GPR17. It is possible that GPR17 and other factors work in concert with each other to regulate the temporal adaptation of Shh, and the dependency on each gene is species specific.

GPCRs can bind to different types of Gα proteins, including Gα_i, Gα_s, Gα_q and Gα_12,13 (Rosenbaum et al., 2009). Among these Gα proteins, Gα_q and Gα_s can increase the intracellular cAMP level, whereas Gα_i proteins mostly exert an inhibitory effect, and GPR17 can bind all types of G protein (Hennen et al., 2013). During the remyelination, GPR17 binds to Gα_i and decreases the cAMP level (Buccioni et al., 2011; Ou et al., 2016; Parravicini et al., 2008; Simon et al., 2016). By contrast, in our experiments GPR17 instead upregulated cAMP levels (Fig. 2G), presumably by binding to Gα_q and/or Gα_s, suggesting that it has diverse and cell type-specific functions.

A historical model suggested that a morphogen can produce a number of cell types depending on the thresholds of the signal concentration (Wolpert, 1969). This hypothesis relies on the supposition that signal concentration completely corresponds to the intracellular signal intensity. However, recent studies suggest that the temporal change in signal activity, so-called temporal adaptation, is crucial for cell fate determination and correct pattern formation (Dessaud et al., 2010, 2007; Ribes and Briscoe, 2009). The results of this study reveal a novel negative-feedback loop constructed by a GPCR through regulation of intracellular cAMP levels and processing of Gli transcription factors (Fig. 1). Recent mathematical modelling suggests the existence of negative regulators induced by Shh (Cohen et al., 2015), and our findings regarding GPR17 are in good agreement with this hypothesis.

The function of GPR17 is to modulate intracellular cAMP levels (Figs 3D and 4U). In this study, we analysed the processing of Gli3 to evaluate the Gli activity (Figs 1I, 3B and 6L) but Gli2 has been shown to play essential roles in the dorsal-ventral patterning of the neural tube (Pan et al., 2009). Therefore, the analysis on the dynamics of Gli2 processing may also be intriguing in a future study.

Although Shh induces the Gpr17 expression (Fig. 2G), Gpr17 does not seem to be a direct target gene of Gli transcription factors, as GBS has not been identified in the flanking regions of the Gpr17 gene locus (Kutejova et al., 2016). Instead, Olig2 is the direct regulator of Gpr17 expression (Ou et al., 2016). Consistent with this finding, overexpression of Olig2 in neural explants induces Gpr17 expression (Fig. 2H). This could explain why the negative-feedback regulation is cell type-specific and does not take place in the NIH3T3 cells treated with the Shh ligand (Cohen et al., 2015); Olig2 expression is induced only in the neural progenitor cells, but not in NIH3T3 cells (A.Y., A.H. and N.S., unpublished).

In a developmental context, negative-feedback regulation confers diversity of cell types and robustness of pattern formation (Dessaud et al., 2010). To date, multiple negative-feedback systems for Gli activity have been identified (Cohen et al., 2015). For example, the Ptch1 gene, the product of which negatively regulates the intracellular Shh signal activity, is a direct target of Shh (Fig. 8) (Dessaud et al., 2007). Moreover, expression of the transcriptional mediator Gli2 decreases over time during neural development, although the underlying regulatory mechanisms remain elusive (Cohen et al., 2015). In combination with the existence of GPR17, as this study demonstrated (Fig. 8), multiple negative-feedback systems could cause differences in the half-life of the signal in each progenitor cell, allowing diverse cell types to be produced by a single morphogen, Shh. Moreover, during formation of the morphogen gradient, uncertainties and fluctuations may arise. This noise in the signal is constantly modulated by negative feedback, and consequently the reliability and reproducibility of pattern formation can be established (Lander, 2011; Perrimon et al., 2012). Therefore, the negative-feedback loop formed by Shh and GPR17 is important not only for overall neural tube pattern formation, but also for fine tuning of signal intensity; consequently, it establishes the quantitative balance among different cell types during neural and neuronal differentiation. Further studies, including analysis of the temporal change of intracellular cAMP levels, will reveal the significance of GPR17 in modulating the Shh signal using negative feedback.

GPCRs that can couple with Gα_q or Gα_s were identified by referring to GPCRdb (gpcrdb.org); and qPCR primers were designed against the corresponding genes (Table S1). Relative expression levels were analysed by RT-qPCR in neural explants treated or untreated with Shh^H (Fig. S1). NCBI gene IDs for chicken and mouse Gpr17 are 769024 and 574402, respectively.

All animal experiments were performed with the approval of the Animal Welfare and Ethical Review Panel of Nara Institute of Science and Technology (approval numbers 1636 and 1810 for chicken and mouse experiments, respectively). Chicken eggs were purchased from the Shiroyama Kei-en farm (Kanagawa prefecture, Japan) and the Yamagishi farm (Wakayama prefecture, Japan), and the developmental stages were evaluated using Hamburger and Hamilton's criteria (Hamburger and Hamilton, 1992). In ovo electroporation of chick embryos was carried out with an ECM830 electroporator (BTX), and embryos were incubated at 38°C for the indicated times. Overexpression in the chick embryos was performed with the constructs subcloned into the pCIG expression plasmid, which contains an IRES (internal ribosomal entry site) -GFP gene downstream of the gene of interest (Megason and McMahon, 2002). pCIG-Olig2^DBD-VP16 was constructed by fusing the DNA-binding region of Olig2 with the transactivation domain of the viral protein VP16, as described previously (Zhou and Anderson, 2002; Zhou et al., 2001). Embryos were fixed with 4% paraformaldehyde, subsequently incubated with 15% sucrose for cryoprotection and embedded in 7.5% gelatine (Sigma). Cryosections were cut at 14 µm increments and analysed using immunofluorescence or in situ hybridisation. The sections from at least seven independent embryos were analysed, and the number of the embryos analysed is indicated in Table S2. For loss-of-function experiments, si-cGpr17-1 [GGAACAGAGUGGAGAAACACCUG(dA)(dA) (sense) and UUCAGGUGUUUCUCCACUCUGUUCCCA (antisense)] or its negative control si-control [AUGCUUCUCCGAACGUGUCACGU(dT)(dT) (sense) and UUUCGUGACACGUUCGGAGAAGCAUCA (antisense); Eurofin] and si-cGpr17-2 (CAGACUGUGGAAGUGAACAACACAA) or si-control-2 (siRNA Negative Control, Med GC; Invitrogen) was electroporated with the pCIG vector, which encodes GFP as a tracer. Unless mentioned otherwise, loss-of-function experiments were performed with si-cGpr17-1. DIG-labelled (Roche) complementary RNA probe was synthesised using RNA polymerase (Promega). In situ hybridisation on slides was performed as described previously (Sasai et al., 2014).

Immunofluorescence was performed as described previously (Sasai et al., 2014). Primary antibodies against the following proteins were used in this study: Olig2 (rabbit, Millipore AB9610), Nkx2.2 (mouse, DSHB 74.5A5; rabbit, Abcam ab19077), Islet1/2 (DSHB 39.4D5), Pax7 (DSHB), FoxA2 (goat, Santa Cruz sc-6554X), Sox2 (rabbit, Millipore AB5603), GPR17 (goat, Abcam ab106781 for mouse embryos; rabbit, Sigma SAB4501250 for cells), GFP (sheep, AbD Serotec 4745-1051), Gli3 (goat, R&D AF3690), β-Actin (rabbit, Abcam ab8227) and γ-acetylated tubulin (mouse, Sigma T7451). For GPR17 immunofluorescence in mouse embryonic sections, antigenic retrieval was required; slides were boiled for 1 min in 10 mM citric acid (pH 6.0) solution prior to the incubation with the primary antibody. Fluorophore- and HRP-conjugated secondary antibodies were purchased from Jackson Laboratory and Cell Signaling Technology, respectively.

New culture was performed as described previously (Psychoyos and Finnell, 2008). Culture plates containing albumen and agarose were prepared with the final concentration of 100 µM of MDL29951 or with dimethylsulfoxide (DMSO) (control). Chick embryos were taken from the eggs at HH stage 12 with a ring of filter paper, and put on the culture plates with the ventral side up. Embryos were cultured in the 38°C incubator for 36 h, and the plates were kept humid with Hanks' Balanced Salt solution (1×HBSS; Sigma-Aldrich).

Chick intermediate neural explants were prepared as described previously (Sasai et al., 2014). HH stage 9 embryos were isolated from the eggs and maintained in L-15 medium (Thermo Fisher Scientific). After treatment with Dispase II (Sigma), the intermediate region of the neural plate at the preneural tube level (Delfino-Machin et al., 2005) was manually excised from the embryos and cultured in DMEM/F-12 medium (Thermo Fisher Scientific) supplemented with Mito+Serum Extender (Sigma) and penicillin/streptomycin/glutamine (Wako). The explants were cultured in the pH-adjusted collagen gel (Sigma; for immunofluorescence and RT-qPCR) or on the glass plates coated with gelatin (Sigma; for western blots). Quantitation of positive cells was performed in at least five areas, each of which contained ∼200-250 cells, randomly chosen from independent explants. Shh^L and Shh^H were defined as the concentrations of Shh that differentiate explants into motor neuron and floor plate at 48 h, respectively (Fig. S1). Shh^H is four times as high as Shh^L. H-89 (Wako) and MDL29951 were used at the indicated concentrations. For RT-qPCR analyses, RNA was extracted from the pools of 15 chick intermediate neural explants by using PicoPure RNA extraction kit (Thermo Fisher Scientific). Each extraction produced a range of 500 ng (12-h culture) to 1.5 µg (48-h culture) of total RNA, as measured by Nanodrop (Thermo Fisher Scientific). Complementary DNAs (cDNAs) were synthesised by PrimeScript reverse transcriptase (TaKaRa), and qPCR was performed by using LightCycler 96 (Roche). At least two independent pools of explants were analysed in each experiment, and each gene expression level was normalised to the Gapdh expression. Mouse neural explants (Fig. S11) were prepared from the E8.5 embryos and were cultured in the same way as in the chick explants (Balaskas et al., 2012), except that the culture medium contained N2 and B27 supplements (Thermo Fisher Scientific).

NIH3T3 fibroblast cells were maintained in Dulbecco's modified Eagle's medium /F12 medium (Wako) containing 10% New-Born Calf Serum (MP Biomedicals). Lipofectamine 3000 (Invitrogen) was used for transfection. To obtain explicit evidence of the effects on the overexpression of the genes in Fig. 3E-I, the transfected cells were selected for 48 h with 500 µg/ml of G418 (Invitrogen) from 1 day after the pcDNA3-based expression plasmids were transfected and used for assays.

The GBS-Nano-Luc reporter construct containing the GBS was constructed by subcloning of the octameric GBS (Sasaki et al., 1999) to the pNL3.2 vector (Promega) and used to monitor Gli activity. pCS2-firefly luciferase was used as an internal control. For the luciferase assays on chick neural explants, GBS-Nano-Luc and pCS2-firefly-luciferase were electroporated at the preneural tube region of the caudal part of the chick embryos at HH stage 9− (Delfino-Machin et al., 2005). si-RNAs were co-electroporated if necessary. The explants were then prepared and cultured under the indicated conditions. The relative luciferase activities were compared with those cultured without Shh at every time point. Seven or eight explants were grouped for each measurement, and four to 10 groups of explants were measured per condition.

CREB-Luc (luciferase gene driven by the cAMP-responsive element-binding region) was a kind gift from Prof. Itoh (Nara Institute of Science and Technology, Ikoma, Japan) (Jenie et al., 2013) and pRL-CMV (Promega) was used for an internal control. The measurement of the chemiluminescence was performed by using the plate reader Tristar2 (Berthold Technologies).

For the cAMP assay in the NIH3T3 cells (Fig. 3D), the cells were transfected with the indicated plasmids and the transfected cells were selected for 24 h. 3-Isobutyl-1-methylxanthine (IBMX), a non-competitive selective phosphodiesterase inhibitor, was added at 1 µM in the final 30 min before cells were harvested to raise the basal level of intracellular cAMP. For the cAMP assay in the neural explants (Fig. 4U), 15 explants were prepared for each condition and cultured for 24 h. IBMX was added in the last 1 h. The cAMP assay was performed using DetectX high sensitivity direct cAMP Chemiluminescent Immunoassay Kit (Arbor Assays). The protein concentrations of the cell lysates were measured by CBB protein assay solution (Nacalai, Japan) and the measurement values were normalised.

TUNEL assay was performed to detect the apoptotic cells. The explants were incubated with TdT transferase (Roche) and DIG-labelled dUTP (Merck Millipore), and the signals were visualised using anti-digoxigenin-rhodamine antibody (Sigma). Staurosporine (Belmokhtar et al., 2001) was used as a positive control at 1 nM.

The Sox1-GFP ES cell line was kindly provided by Prof. Smith (University of Cambridge, UK), and maintenance and neural differentiation as a monolayer culture were performed as described previously (Kamiya et al., 2011; Kutejova et al., 2016; Ying et al., 2003). Briefly, on the fibronectin/collagen-coated plates, ∼1.5×10⁴ ES cells were seeded and cultured as monolayers in the differentiation medium (Kutejova et al., 2016) for the initial 3 days. For pMN differentiation, 300 nM RA (Sigma) was added on day 3.0 and 50 nM SAG was added on day 3.5, and the cells were incubated for a total of 5 days. For p3 differentiation, 30 nM RA and 500 nM SAG (Sigma) were added on day 3.0 and day 3.5, respectively. For FP differentiation, 500 nM SAG was added on day 3.0. Before differentiation was initiated, Stealth siRNAs (Invitrogen) (si-mGpr17-1, UCGCCUGCUUCUACCUUCUGGACUU; si-mGpr17-2, ACCGUUCAGUCUAUGUGCUUCACUA) or si-control (Invitrogen) were transfected twice at 24 h intervals using Lipofectamine RNAiMAX (Invitrogen). For RT-qPCR analysis, RNA was extracted by RNeasy RNA extraction kit (QIAGEN). cDNA synthesis and qPCR were performed as in the chick neural explants.

Images were collected by using an AxioVision2 microscope (for in situ hybridisation images; Zeiss), a LSM 710 confocal microscope (for immunofluorescence data; Zeiss) and a LAS4000 ImageQuant (for western blots; GE Healthcare). Signal intensities of the western blots were calculated by ImageJ. Images were processed using Photoshop (Adobe) and figures were arranged on Illustrator (Adobe). Statistical analysis was performed using Prism (GraphPad) with two-tailed t-test (Figs 1J, 2H, 3A,C,D, 4M,T, 6I-K and 7F,P,Q) or with one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test (Figs 1H,J, 2I, 3E, 4U, 6M and 7A). Data are presented as mean±s.e.m. and significance (*P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; n.s., not significant) is indicated in each graph.

The authors are grateful to Profs Austin Smith and Hiroshi Itoh for materials; to Dr James Briscoe for comments on the manuscript; to Prof. Naoyuki Inagaki, and Drs Yuichi Sakumura and Takunori Minegishi for advice; to Prof. Takashi Toda for encouragement; and to all laboratory members for support and valuable discussion.