Ciliary adenylyl cyclases control the Hedgehog pathway

Sonic hedgehog (Shh) belongs to the Hedgehog (Hh) family of secreted signalling proteins. Shh activity is crucial during the development of the nervous system to control the proliferation and the dorso-ventral patterning of the neural tube (Dessaud et al., 2008; Le Dréau and Martí, 2012), and for the clonal expansion of cerebellar granular cell precursors (CGNPs; Dahmane and Ruiz i Altaba, 1999; Wechsler-Reya and Scott, 1999). In vertebrates, most of the Hh signalling pathway components are located at or in the proximity of the primary cilium, and the integrity of this organelle is an absolute requirement for correct Hh signalling (Briscoe and Thérond, 2013). Indeed, in the absence of Hh, the 12-pass transmembrane protein that receives Hh ligands, Patched 1 (Ptch1), is concentrated in and around the primary cilium (Rohatgi et al., 2007). However, upon Hh binding, Ptch1 is excluded from the cilium and the G-protein-coupled receptor (GPCR) Smoothened (Smo) concentrates there (Corbit et al., 2005). Gli proteins are the only known transcriptional effectors of the Hh pathway and the activity of the pathway is fully dependent on the balance between activator and repressor forms of Gli2 and Gli3. In addition, protein kinase A (PKA), the main repressor of the Hh pathway, also accumulates at the base of the cilium (Barzi et al., 2010), where activation by cAMP induces repression driven by Gli2 and Gli3 (Wang et al., 2000; Tempe et al., 2006; Tuson et al., 2011). Although the activity of PKA is basically regulated by cAMP, it remains unclear to what extent the adenylyl cyclases, which produce cAMP from ATP, influence the Hh pathway.

Adenylyl cyclases are currently grouped into six classes that can be distinguished on the basis of their sequence. All adenylyl cyclases found in metazoans belong to the class III adenylyl cyclases, and in mammals, class III adenylyl cyclases can be further divided into membrane-bound adenylyl cyclases (type 1 to type 9, AC1–AC9, also known as ADCY1–ADCY9) and soluble adenylyl cyclases (sACs; Linder, 2006). The nine membrane-bound adenylyl cyclases are controlled by a complex set of factors, including heterotrimeric G-proteins. Stimulatory G-proteins (Gα_s) activate all nine adenylyl cyclase isoforms to some extent, whereas inhibitory G-proteins (Gα_i) only inhibit AC1, AC5 and AC6 (Linder, 2006; Sadana and Dessauer, 2009). Smo is a fully competent activator of Gα_i proteins (Shen et al., 2013) and in Drosophila, Smo acts as a canonical GPCR that signals through Gα_i in order to regulate the Hh pathway (Ogden et al., 2008). Moreover, in rat CGNPs, simultaneous knockdown of Gα_i2 and Gα_i3 (also known as GNAI2 and GNAI3, respectively) dramatically hinders the proliferation of these cells induced by Shh (Barzi et al., 2011). By contrast, GPR161, an orphan GPCR that engages Gα_s, exerts negative effects on the Hh pathway (Mukhopadhyay et al., 2013). Furthermore, targeted knockout of GNAS, the gene that encodes Gα_s, causes medulloblastoma-like tumours in mouse cerebellum (He et al., 2014).

In this work, we show that AC5 and AC6 are the two adenylyl cyclase isoforms most strongly expressed in two well established Shh-target tissues: CGNPs and the embryonic neural tube. We also show that overexpression of AC5 and AC6 represses the Hh pathway, whereas their knockdown results in consistent activation of the Hh pathway, both in CGNPs and the neural tube. In addition, we show that AC5 and AC6 concentrate in the primary cilium and that the mutations of a previously undescribed motif in AC5 suppress its targeting to the cilium and its capacity to inhibit the Hh pathway. Finally, we demonstrate that, as in CGNPs, the Hh pathway is positively and negatively regulated by Gα_i and Gα_s in the neural tube, respectively. Therefore, we propose that the activity of different ciliary GPCRs converges on AC5 and AC6 to control PKA activity and, hence, the Hh pathway.

The proliferation of CGNPs in the external germinative layer (EGL) of the cerebellum (Fig. 1A) depends entirely on Shh activity and consequently, this tissue has proved to be an excellent model to study the regulation of the Hh pathway. Conveniently, CGNPs can be easily purified and cultured (Rios et al., 2004). When we studied the expression of the nine membrane-bound adenylyl cyclase isoforms in purified mouse CGNPs by quantitative real-time PCR (q-PCR), we found that AC5 and AC6 were the most strongly expressed isoforms, followed by AC3 and AC1, whereas AC2, AC4, AC7, AC8 and AC9 were only weakly expressed (Fig. 1B). In fact, AC3, AC5 and AC6 were also the most strongly expressed isoforms in the cerebellar EGL when studied by in situ hybridization (ISH). It is worth noting that AC6 was the only adenylyl cyclase isoform that in cerebellum was almost exclusively expressed in the EGL.

Forskolin is an allosteric activator of adenylyl cyclases and dibutyril-cAMP (DBA) is a cAMP analogue, both of which have been reported to inhibit the Hh pathway (Dahmane and Ruiz i Altaba, 1999; Wechsler-Reya and Scott, 1999; Rios et al., 2004). Given that the effect of adenylyl cyclase expression on the Hh pathway had not been assessed previously, we measured the proliferation of CGNPs expressing AC1, AC3, AC5 or AC6 that was induced by Shh. Notably, we found that Shh-induced proliferation was significantly reduced in CGNPs ectopically expressing AC3, AC5 and AC6 but not in those expressing AC1, as reflected by both [³H]thymidine (Fig. 1D) and BrdU (Fig. 1E) incorporation. Indeed, similar results were obtained when the Hh pathway was activated by transfecting CGNPs with a constitutively active form of Smo, SmoM2 (Fig. 1F,G). In addition, the endogenous expression of Gli1 (a direct target of the pathway) was significantly diminished in cultured CGNPs grown in the presence of Shh when they expressed AC3, AC5 and AC6 but not AC1 (Fig. 1H). In fact, the effect of AC5 expression could be reverted by a dominant-negative form of PKA but not by SmoM2, confirming that the activity of adenylyl cyclases lies between Smo and PKA in the Hh pathway (Fig. 1I).

To further understand how adenylyl cyclases regulate the Hh pathway we used the GBS-Luc construct, a Luciferase reporter driven by a Gli-binding site (GBS). None of the adenylyl cyclases, nor an active form of PKAα catalytic subunit (PKACα^QR), used as control, reduced GBS-Luc activity when the pathway was activated by Gli1 or Gli2 transfection (Fig. 1J,K). Rather, and perhaps paradoxically, AC5, AC6 and PKACα^QR increased Gli2-stimulated GBS-Luc activity (Fig. 1K). However, although Gli3 only mildly activated luciferase expression when compared to Gli1 or Gli2, this activation was significantly dampened by AC3, AC5, AC6 and PKACα^QR (Fig. 1L). Moreover, if Gli3 was co-expressed with Gli1 the suppression of GBS-Luc activity induced by adenylyl cyclases, and especially by PKACα^QR, was significantly stronger (Fig. 1M). In addition we confirmed that the formation of the repressor form of Gli3 was reduced by SmoM2 and increased by AC5 or PKACα^QR (supplementary material Fig. S1A). These data are in agreement with the elevated capacity of Gli3 to drive Hh pathway inhibition compared to Gli1 or Gli2 (Pan et al., 2006). Thus, together these results indicate that AC3, AC5 and AC6 but not AC1 can antagonize Hh signalling in CGNPs, possibly by modulating Gli activity.

Although cAMP is a fully diffusible metabolite, its activity is largely constricted by phosphodiesterases. Thus, because PKA regulates the Hh pathway from the base of the cilium (Barzi et al., 2010; Tuson et al., 2011), a ciliary or peri-ciliary distribution of the adenylyl cyclases regulating PKA would also be expected. AC3 has previously been shown to accumulate at the primary cilium in CGNPs (Barzi et al., 2010); however, whether it regulates the Hh pathway was not known. To address this issue, we transfected CGNPs with HA-tagged constructs encoding the different adenylyl cyclase isoforms endogenously expressed in these cells (AC1, AC3, AC5 and AC6), and we found AC3, AC5 and AC6 but not AC1 at the cilium (Fig. 2A). Additionally, using an antibody that detects AC5 and AC6 with similar affinity but that does not cross-react with AC3 (supplementary material Fig. S1B), we confirmed that endogenous AC5 and AC6 can be found at the cilia of CGNPs and of NIH-3T cells (Fig. 2B). AC1 was the only adenylyl cyclase isoform that did not diminish Shh-induced proliferation of CGNPs (Fig. 1D,E), therefore a direct relationship between a ciliary localization and the capacity to repress the Hh pathway was plausible. To study this issue further, we evaluated the possibility that ciliary localization motifs could exist in AC5 by aligning this molecule with other ciliary proteins. Notably, we observed that mutating a WR motif that is also present in Smo (Corbit et al., 2005), involving amino acids 76 and 77 of the rabbit sequence (Fig. 2C), disrupted the distribution of AC5 at the cilium (Fig. 2D). Moreover, although the catalytic activity of the AC5^WR76-77AA mutant was preserved (supplementary material Fig. S1C,D), it no longer repressed Shh-induced proliferation of CGNPs (Fig. 2E,F).

It has been reported that activators of the Hh pathway like GliA or dominant-negative PKA induce overgrowth of neuroepithelial cells and a ventralization of the developing neural tube, whereas repressors like Gli3R reduced cell proliferation and induced dorsalization (Epstein et al., 1996; Stamataki et al., 2005; Cayuso et al., 2006; Dessaud et al., 2008). By using in ovo electroporation and fluorescent cell sorting, we studied the expression of endogenous AC1, AC3, AC5 and AC6 in the neuroblasts of HH-12 chicken neural tube at 12 hours post-electroporation (hpe) by q-PCR, and found that, as in CGNPs, AC5 and AC6 were the two isoforms most highly expressed in the neuroblasts of chicken neural tube (Fig. 3A). Thus we studied the effect of AC3, AC5 and AC6 overexpression on the different effects of Shh in the developing chicken neural tube. First, we observed that AC5 or AC6 expression but not that of AC3 significantly reduced cell proliferation in HH-12 stage chicken embryo neural tubes at 16 hpe with this decrease being most prominent in the ventral part of the tube (Fig. 3B–D). It is worth noting the strong anti-proliferative and pro-differentiative effect developed by the active form of PKA (PKACα^QR), used as a positive control, prevents any proliferative activity between transfection and fixation, enormously reducing the total number of GFP-positive cells (Barzi et al., 2010). In accordance, fluorescence-activated cell sorting (FACS) cell cycle analysis revealed a reduction in the number of cells in the S and M phases of the cell cycle in neural tubes transfected with AC5, and it was noteworthy that the distribution of cell cycle phases in neural tubes expressing AC5 was very similar to neural tubes transfected with Gli3R (a Gli3 mutant lacking the transactivation domain) plus Bcl2, this construct combination has been previously reported to block most Shh-dependent signals without inducing apoptosis in the neural tube (Cayuso et al., 2006; Pedersen and Rosenbaum, 2008) (Fig. 3E). Shh also induces the differentiation of motoneurons in the developing neural tube (Marti et al., 1995), stimulating the proliferation and survival of motoneuron precursors (which are Olig2-positive; Cayuso et al., 2006). AC5 or AC6 expression but not that of AC3, significantly reduced the number of Olig2-positive cells at 16 hpe (Fig. 3F,G). Moreover, co-electroporation of adenylyl cyclases with a reporter of Hh pathway activation (8×GBS–RFP), demonstrated that the Hh pathway was activated in fewer of the cells expressing AC5 and AC6 but not in fewer of those expressing AC3 (Fig. 3H,I), a similar result was obtained using the original luciferase expressing vector GBS-LUC (supplementary material Fig. S1E). These results demonstrate that, like in CGNPs, the activity of AC5 and AC6 can suppress the Hh pathway in the neural tube, whereas AC3 expression has some mild effects on occasions but these effects were never significant in the neural tube.

We next wondered whether suppression of adenylyl cyclases was sufficient to stimulate the Hh pathway and hence, we cloned inhibitory short hairpin RNAs (shRNAs) that target mouse AC3, AC5 or AC6 into the pSHIN vector and probed their effects in CGNP proliferation assays (the sequences targeted and shRNA efficiency are shown in supplementary material Table S1 and Fig. S1F, respectively). Most of the shRNAs designed against AC5 and AC6, but not those against AC3, significantly increased the proliferation of CGNPs both in the presence or absence of low doses of Shh (0.3 µg/ml: Fig. 4A), an effect that was not evident at saturating concentrations of Shh (data not shown). We also studied how adenylyl cyclase knockdown affected the Hh pathway during neural tube development using in ovo electroporation of luciferase (Fig. 4B) and RFP (Fig. 4C,D) reporters of Gli activity to test selected chicken shRNAs targeting AC3, AC5 or AC6 (targeted sequences and shRNA efficiency are shown in supplementary material Table S1 and Fig. S1G, respectively). Notably we found that the Hh pathway was activated significantly by AC5 and AC6 knockdown but not by the knockdown of AC3. In addition, the combined suppression of AC5 and AC6 induced further activation of this pathway over and above the knockdown of each enzyme alone. By contrast, combining the knockdown of AC3 and AC5 activated the Hh pathway to a similar extent as the knockdown of AC5 alone (Fig. 4B–D). As a further control to the empty vector, the effects of an shRNA with very mild activity (shAC6.2) was also evaluated in these experiments. It is worth noting that the stimulatory effect of AC5 and AC6 shRNAs was in all cases reversed by the overexpression AC5 and/or AC6 (data not shown). Thus, these results further support the hypothesis that repression of the Hh pathway in CGNPs and neural tube is mediated by the activity of the ciliary located AC5 and AC6.

The proliferation of CGNPs carrying GTPase-deficient mutants of Gα_i2 and Gα_i3 (Gα_i2^QL and Gα_i3^QL, note that Gα_i3^QL remains active longer) is significantly enhanced when they are grown in the presence of sub-proliferative doses of Shh. Moreover, the proliferation of CGNPs induced by either Shh or SmoM2 is strongly inhibited by the simultaneous knockdown of Gα_i2 and Gα_i3 (Barzi et al., 2011). Conversely, Gα_s acts as a tumour suppressor for medulloblastoma and its expression strongly represses CGNP growth (He et al., 2014). Here, we have shown that AC5 and AC6 are the most abundant adenylyl cyclase isoforms in CGNPs, and that in both CGNPs and the chicken neural tube, these adenylyl cyclases exert a relevant effect on the Hh pathway. Additionally, AC5 and AC6 are among the few adenylyl cyclases isoforms that can be simultaneously regulated by Gα_s and Gα_i proteins (Linder, 2006). However, it was not clear whether Gα proteins regulate the Hh pathway during neural tube development. We found that like SmoM2, Gα_i3^QL expression increased the number of Olig2-positive cells, as well as inducing the expression of Nkx2.2 and Olig2 in cells lying dorsally to their normal domains (Fig. 5A,B,E). Conversely, expression of Gα_s^RC (a GTPase-deficient mutant of Gα_s) in HH-12 chicken neural tubes greatly reduced the number of neuroblasts that expressed the Nkx2.2 and Olig2 ventral markers, with Pax7-positive cells (a dorsal marker) occupying most of the neural tube (Fig. 5A,B,E). Notably, the dorsalization of the neural tube induced by Gα_s^RC could be reversed by SmoM2 or Gα_i3^QL expression, with the maximal ventralization of this tissue observed upon co-expression of SmoM2 and Gα_i3^QL (Fig. 5C–E).

To obtain more quantitative data on Hh pathway activation in each situation, we used the GBS-Luc reporter and we found that the effects caused by the expression of Gα_s^RC and Gα_i3^QL on neural tube patterning were correlated with the levels of Hh pathway activation in the transfected cells (Fig. 5F). Thus, the results presented here demonstrate that, in CGNPs and in the chicken neural tube, the Hh pathway is controlled by ciliary AC5 and AC6, and we present strong evidence suggesting that its activation depends on the associated Gα_s–Gα_i equilibrium.

In this study, we have investigated the relationship between the PKA and Hh signalling pathways in two tissues in which the latter plays a fundamental role. It has previously been shown that there is an expansion of Shh-dependent ventral cell types in the developing neural tube of PKA-deficient mouse embryos [PKACα (Prkaca)^−/− PKACβ (Prkacb)^−/+ or PKACα^−/+ PKACβ^−/−; Huang et al., 2002], whereas the neural tube remains open in mice without any catalytic PKA activity (PKACα^−/−, PKACβ^−/−) and the strong expansion of ventral progenitors is coupled to an absence of dorsal interneuron progenitors (Tuson et al., 2011). Indeed, this phenotype of PKA-null mice is very similar to that of pthc1- and Sufu-null mice. Thus, the total absence of PKA activity fully emulates maximal activation of the Hh pathway. PKA appears to repress the Hh pathway through two main mechanisms: by preventing Gli2 activation and promoting Gli3 repressor formation. The relevance of each of these mechanisms for Hh activity varies in the different Hh target tissues, whereby PKA repression of ventral progenitors in the neural tube seems to be mediated by the inhibition of Gli2 activation (Tuson et al., 2011), yet in tissues, like the limb-bud, Hh repression is mainly achieved by PKA-induced production of the Gli3 repressor (te Welscher et al., 2002).

In recent years, compelling evidence has accumulated showing that the vertebrate Hh pathway fully depends on the integrity of the primary cilium, where activator and repressor forms of Gli transcription factors are generated. Although neither the chronology nor the morphological aspects of the process are as yet clear, a common argument in all the models proposed is the central role played by the pool of PKA located at the cilium base (Tuson et al., 2011; Nozawa et al., 2013; Mukhopadhyay and Rohatgi, 2014). Hence, it would seem that the ciliary components of the Hh pathway exert a tight control over the activation status of this PKA pool. Although the location of PKA at the cilium base is controlled by the A-kinase anchoring proteins (AKAPs; McConnachie et al., 2006; Barzi et al., 2010), its activity is fundamentally controlled by cAMP levels. Although cAMP is fully diffusible, the range at which it effectively regulates PKA activity is largely narrowed down by dilution and phosphodiesterase activity. However, the volume of the cilium is negligible compared to the total cell volume and thus, cAMP can be produced at very high molar concentrations in the cilium where it can control the PKA pool at its base without producing any measurable effects on other PKA pools. Accordingly, we hypothesized that the adenylyl cyclases regulating PKA activity at the cilium base, and therefore controlling the Hh pathway, should be located at the cilium.

Based on pharmacological data, it has long been assumed that adenylyl cyclases worked as inhibitors of the Hh pathway, although no direct evidence has been available to date. Here, we have shown that among those adenylyl cyclase isoforms expressed at a reasonable level in CGNPs (AC1, AC3, AC5 and AC6), only the isoforms that concentrate in the cilium (AC3, AC5 and AC6) inhibit the proliferation of these cells induced by Shh. Although not demonstrative, these results encouraged us to further confirm our hypothesis and thus, by comparing the sequences of AC5 and AC6 with those of other ciliary proteins, including Smo, we found several putative ciliary location motifs in the AC5 sequence. Notably, we found a WR motive in AC5, that although it was not located close to a transmembrane domain as in Smo or ODR-10, the mutation of which not only inhibited its ciliary localization but also, it abrogated its anti-proliferative effect. Given that the mutant AC5 still induced the phosphorylation of cytoplasmic and nuclear PKA substrates to a similar extent as the wild-type AC5, adenylyl cyclases apparently need to be confined to the cilium to regulate the Hh pathway.

To expand the relevance of our observations, we extended our study to the developing chicken neural tube. Notably, in this structure the expression of AC5 and AC6 significantly reduced different Shh-dependent parameters, such as cell proliferation, Olig2 expression and the activity of the endogenous Hh pathway (as measured with a GBS-RFP reporter). These results demonstrated that ciliary adenylyl cyclases could inhibit the Hh pathway in different Hh target tissues. Moreover, knockdown experiments demonstrated that AC5 and AC6, but not AC3, are active repressors of the Hh pathway in both CGNPs and the neural tube, supporting an inhibitory role of ciliary adenylyl cyclases on the Hh pathway. We measured cell proliferation in CGNPs, as we believe it to be the best measure of sustained Hh pathway activation in these cells, ensuring their health and viability. Alternatively, we used reporters of Gli activity to study the effect of adenylyl cyclase suppression in the neural tube. Although the results obtained by these two methods were very similar, the RFP reporter demonstrated that AC5 and AC6 knockdown (like the overexpression of a dominant-negative form of PKA) not only increased the activity of this pathway in the endogenous (ventral) domain but it was also ectopically activated at any dorso-ventral level. These results indicate that AC5 and AC6 actively repress Hh pathway activation in both CGNPs and the neural tube.

It is worth noting that the AC5 and AC6 double knockdown activated the pathway more strongly than the individual knockdowns alone, almost to the level of the dominant-negative PKA. Hence, it would appear that both AC5 and AC6 contribute to Hh pathway silencing. Although we cannot exclude the contribution of other adenylyl cyclases to this effect, AC5 and AC6 seem to be the principal effectors involved. In this regard, although AC3 overexpression effectively suppressed Shh-induced proliferation in CGNPs, the differences observed when AC3 was overexpressed in the neural tube or when it was knocked-down were never significant. We know that AC3 is endogenously expressed and located in the cilia in both CGNPs and neural tube, yet it is more weakly expressed than AC5 and AC6 in both tissues (data not shown), which could explain the milder effect of AC3 knockdown but not the lack of effect when AC3 is overexpressed in the neural tube. Thus, further studies will be necessary to determine whether AC3 is in fact less effective than AC5 and AC6 in repressing the Hh pathway.

Although adenylyl cyclases have been reported to be modulated by an extensive list of signalling molecules (Linder, 2006), regulation by stimulatory (Gα_s) and inhibitory (Gα_i) subunits of heterotrimeric G proteins occupies a prominent position. Notably, the two adenylyl cyclase isoforms that proved to exert the most relevant effect on the Hh pathway, AC5 and AC6, are also among the few that can be regulated by both stimulatory and inhibitory G proteins. In fact, Gα_i inhibits AC5 and AC6 but only when these molecules have been previously activated by Gα_s (Sadana and Dessauer, 2009). Given that the initial studies showing that pertussis toxin can inhibit Gα_i and induce a phenotype resembling Hh deficiency in zebrafish (Hammerschmidt and McMahon, 1998), there is much more evidence that Gα_i proteins influence the activity of the Hh pathway (Riobo et al., 2006; Ogden et al., 2008; Barzi et al., 2011; Shen et al., 2013). Nevertheless, it must be recognized that other studies have failed to implicate Gα_i in Hh signalling (Lum et al., 2003; Low et al., 2008). Moreover, Gpr161, a GPCR that signals through Gα_s, has recently been shown to be excluded from the cilium upon activation of the Hh pathway, and there are signs of increased Hh activity in Gpr161-null mice (Mukhopadhyay et al., 2013). In addition, Gα_s suppression has been reported to induce medulloblastoma-like tumours in mice (He et al., 2014). However, it is noteworthy that the phenotype associated with Hh activation in Gpr161-null mice was very mild compared to that in Gα_s null mice, in which the neural tube undergoes very intense ventralization, similar to that in PKA-null mice (Hwang and Mukhopadhyay, 2015). Thus, as the activity of Gα_s seems to be crucial to prevent the pathway from signalling in the absence of Hh, additional mechanisms besides the exclusion of Gpr161 from the cilium seem to be required to counteract Gα_s upon Hh pathway activation.

We have previously reported that Gα_i3^QL stimulated the proliferation of CGNPs cooperatively with Shh and indeed, combined knockdown of Gα_i2 and Gα_i3, the two Gα_i subunits located at the cilium, significantly reduce Shh-induced proliferation in these cells (Barzi et al., 2011). Here, we show that in contrast to Gα_s^RC, which induces the complete dorsalization of the chicken neural tube, Gα_i3^QL expression not only increases the number of Olig2-positive cells but that it also extends the Olig2 domain dorsally with a concomitant reduction in the Pax7 territory, similar to the effects of SmoM2. In addition, the extent to which Gα_i3^QL and SmoM2 reversed the effect of Gα_s^RC was very similar. Notably, the effects on neural tube patterning were in all cases correlated to the different extent of Hh activation evident through the GBS-Luc reporter. It is worth noting that like the results obtained previously in CGNPs (Barzi et al., 2011), we observed significant cooperation between Gα_i3^QL and SmoM2 in terms of GBS activation in the neural tube, which is consistent with the fact that Gα_i3^QL must still be engaged by a GPCR to be active. By contrast, no effect on neural tube pattering was associated with Gα_i2^QL expression (Low et al., 2008), although it must be acknowledged that Gα_i plays important roles in cell polarity, and in mitotic spindle orientation and cell division (Siller and Doe, 2009; McCaffrey and Macara, 2011). In fact, there are evident signs of cell depolarization and cell death in Gα_i2^QL-transfected neural tubes (Low et al., 2008). Therefore, the lack of effects on patterning reported upon Gα_i2^QL expression should be interpreted with caution. So far we are not able to confirm whether Gα_i is part of the Hh pathway in the neural tube, as observed in CGNPs, nor if it is directly activated by Smo or by another GPCR activated by it. However, our results clearly support the plausibility of considering Gα_i to be being part of the Hh pathway in the neural tube.

The different antibodies used in this study were: mouse monoclonal antibodies against Gli1 (Cell Signaling, MAB2643), Nkx2.2 (DSHB, 74.5A5), Pax7 (DSHB), α-tubulin (Sigma, T6199), acetylated tubulin (Sigma, 6-11B-1), FoxA2 (DSHB, 4c7); a rat monoclonal antibody against BrdU (AbD-Serotec, BU1/75); rabbit polyclonal antisera against phosphorylated (at serine and threonine residues) PKA substrates (PKA-PS; Cell Signaling 9621), Olig2 (Millipore, AB9610), AC3 (Santa Cruz, Sc588), both AC5 and AC6 (Santa Cruz SC-590), and GFP, RFP, FLAG and HA (produced in-house).

Alexa-Fluor- and peroxidase-labelled secondary antibodies were obtained from Molecular Probes and Pierce, respectively, whereas BrdU was purchased from Sigma and Rhodamine–Phalloidin from Invitrogen.

The bovine ADCY1 cDNA was obtained from Xavier Nicol (Institute de la vision, France), rat ADCY3 from Naoyuki Taniguchi (Department of Biochemistry, Osaka University, Japan), rabbit ADCY5 from Joachim Schultz (Pharmazeutisches Institut, University of Tübingen, Germany), human ADCY6 from Rennolds Ostrom (Department of Pharmacology, University of Tennessee, USA), human SmoM2 from Frederic J. Sauvage (Department of Molecular Oncology at Genentech, Inc., USA), and rat Gnas and Gnas^R201C from Natalia Riobo (Department of Biochemistry, Thomas Jefferson University, USA). Gli3R is a deleted form of human Gli3 encoding amino acids 1–768 (Persson et al., 2002). The luciferase reporter was provided by Yongbin Chen (Department of Developmental Biology, University of Texas, USA) and the RFP Gli-binding site (GBS) reporter was produced in our laboratory. All the constructs were cloned into pCIG, an EGFP-expressing bicistronic vector. Site-directed mutagenesis, and the addition of HA and FLAG tags was achieved by PCR. By contrast, inhibitory short hairpin RNAs (shRNA) were cloned into pSHIN (a GFP-expressing evolution of pSUPER). The shRNA constructs where electroporated in CGNPs with the Microporator MP-100 (see below) using 10 µg of DNA per 3×10⁶ cells in 100 µl of transfection solution. In the neural tube, the shRNA constructs were diluted at 1 µg/µl and ∼1 µl was used for each chicken neural tube electroporation. The list of sequences targeted by the shRNAs and all the primers used for q-PCR are shown in the supplementary material Table S1.

Cerebellar cultures were prepared using a modification of the Papain method, as described previously (Rios et al., 2004). For transient transfections, freshly isolated cells were electroporated in suspension and plated on dishes coated in poly-L-Lysine plus laminin in Neurobasal plus B-27 media (Invitrogen) supplemented with KCl (25 mM), glutamine (1 mM) and Shh (3 µg/ml). After 24 h, the medium was replaced by fresh medium containing the corresponding treatments, taking this moment as time 0. Electroporation was performed using the Microporator MP-100 (Digital Bio, Seoul, Korea) according to the manufacturer's instructions, with a single pulse of 1700 V for 20 ms. The NIH-3T3 and HEK293 cell lines were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS).

White-Leghorn chick embryos were staged according to Hamburger and Hamilton (1992), and electroporated with column purified plasmid DNA (1–2 µg/ml) in water containing Fast Green (50 ng/ml). Transfected embryos were allowed to develop to the stages specified before they were processed as indicated. All animal experiments were performed according to approved guidelines.

The cerebellum from P7 mice was fixed overnight in 4% paraformaldehyde and washed three times for 10 min with PBT (1× DEPC-PBS, 0.1% Triton X-100). The tissue was embedded in 10% sacarose and 5% agarose, and vibratome sections (50 μm, Leica) were obtained and dehydrated with increasing concentrations of methanol in PBT (25%, 50%, 75% and 100%). Hybridization was performed at 70°C using standard procedures and the probe distribution was detected with alkaline-phosphatase-coupled anti-digoxigenin Fab fragments (Boehringer Mannheim). RNA probes were produced from the corresponding cDNA sequences cloned in Bluescript-II-KS with the T7 polymerase.

Electroporated cells were plated for 24 h and then given a 4-h pulse of either [³H]thymidine [1 µCi (3.7×10⁴ Bq) each well, from Amersham, Buckinghamshire, UK] or BrdU (5 µg/ml). The cells that received [³H]thymidine were then lysed in SDS (0.04%) and proliferation was measured in a Wallac scintillation counter (Wallac-Perkin Elmer, Quebec, Canada). The BrdU-labelled cells were fixed and immunostained with an antibody against BrdU, and cell proliferation was calculated as the proportion of the transfected population (GFP expressing cells) that incorporated BrdU.

HH-12 chicken embryos were electroporated with the different constructs cloned in pCIG and at 24 hpe, a single-cell suspension was obtained by digestion for 10–15 min with Trypsin-EDTA (Sigma). At least three independent experiments were analysed by FACS (six embryos in each experimental condition). Hoechst and GFP fluorescence was determined by flow cytometry in a MoFlo cytometer (DakoCytomation, Fort Collins, CO), and the cellular DNA content was analysed (ploidy analysis) in single fluorescence histograms using Multicycle software (Phoenix Flow Systems, San Diego, CA).

Embryos were fixed in 4% paraformaldehyde for 2 h at 4°C and immunostaining was performed on vibratome (40 µm) sections following standard procedures. BrdU (∼1 µl at 0.5 µg/µl) was injected into the neural tube lumen 4 h prior to fixation and transverse sections of the neural tube were then stained with antibodies against BrdU (proliferation, red) and GFP (transfection, green), and DAPI (nucleus, blue). The total number of GFP-positive (transfected cells) and the number of GFP and BrdU double-labelled cells (transfected that incorporated BrdU) were counted for each slice and the percentage BrdU positive among the transfected population calculated. The percentage was calculated separately for the dorsal and ventral neural tube. To detect BrdU, sections were incubated in 2 N HCl for 30 min and then rinsed with 0.1 M Na₂B₄O₇ (pH 8.5). After washing in PBS with 0.1% Triton X-100, the sections were incubated with the appropriate primary antibodies, the binding of which was detected with Alexa-Fluor-conjugated secondary antibodies. After staining, the sections were mounted and examined on a Leica SP5 confocal microscope.

For immunobloting, transfected cells were dissolved in 1× SDS Laemmli sample buffer (five embryos per condition) and after sonication, the insoluble material was removed by centrifugation. Samples were resolved by 8% SDS-PAGE, transferred to nitrocellulose membranes, blocked with 8% non-fat dried milk powder in TTBS (150 mM NaCl, 0.05% Tween-20 and 20 mM Tris-HCl pH 7.4) and the membranes were probed with the primary antibodies. Antibody binding to the blots was detected using protein A/G-coupled peroxidase that was visualized with the ECL system and captured with a Versadoc Imaging System (Bio-Rad), or with Alexa-Fluor-labelled secondary antibodies that were visualized with the Odyssey Infrared Imaging System (LI-COR). The expression was quantified with Quantity One software (Bio-Rad) and the molecular masses of the proteins estimated using Bio-Rad Precision Molecular Weight Markers.

CGNP cultures or HH-12 chicken embryos were electroporated with the DNAs indicated together with a GBS-Luciferase reporter construct containing eight synthetic Gli-binding sites (GBS-Luc) and a CMV driven Renilla construct (Promega) for electroporation normalization. Total CGNP cultures or GFP-positive neural tubes were dissected out at 48 hpe and homogenized in Passive Lysis Buffer (Promega). Firefly and Renilla luciferase activity was measured with the Dual Luciferase Reporter Assay System (Promega).

PKA activity was evaluated using an antibody raised against phosphorylated PKA targets, as reported previously (Barzi et al., 2010; He et al., 2014). The relative amount of phosphorylated (at serine and threonine residues) PKA substrates (PKA-PS) was compared between HEK-293 cells transfected with either wild-type AC5 or the mutant AC5^WR76-77AA. Confocal images were acquired with a laser-scanning confocal microscope (Sp5, Leica). Maximum projections of three consecutive confocal sections (2 µm total Z) were quantified with ImageJ software (National Institutes of Health), converting all images into 32 bits and subtracting the background using the original thresholds set with the background as ‘NaN’ (‘not a number’). Finally, the mean grey value was quantified from an area containing the entire cell. The ratio between GFP (transfection level) and PKA-PS was calculated for each cell.

Total RNA extracts were obtained following the TRIzol protocol (Invitrogen). Reverse transcription and real-time PCR (q-PCR) were performed according to the manufacturer's instructions (Roche) using a lightcycler (LC 480; Roche).

Monoclonal antibodies were obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the NICHD and maintained by The University of Iowa (Department of Biological Sciences, Iowa City, IA 52242).