Distinct Activities of Gli1 and Gli2 in the Absence of Ift88 and the Primary Cilia

The primary cilia play essential roles in Hh-dependent Gli2 activation and Gli3 proteolytic processing in mammals. However, the roles of the cilia in Gli1 activation remain unresolved due to the loss of Gli1 transcription in cilia mutant embryos, and the inability to address this question by overexpression in cultured cells. Here, we address the roles of the cilia in Gli1 activation by expressing Gli1 from the Gli2 locus in mouse embryos. We find that the maximal activation of Gli1 depends on the cilia, but partial activation of Gli1 by Smo-mediated Hh signaling exists in the absence of the cilia. Combined with reduced Gli3 repressors, this partial activation of Gli1 leads to dorsal expansion of V3 interneuron and motor neuron domains in the absence of the cilia. Moreover, expressing Gli1 from the Gli2 locus in the presence of reduced Sufu has no recognizable impact on neural tube patterning, suggesting an imbalance between the dosages of Gli and Sufu does not explain the extra Gli1 activity. Finally, a non-ciliary Gli2 variant present at a higher level than Gli1 when expressed from the Gli2 locus fails to activate Hh pathway ectopically in the absence of the cilia, suggesting that increased protein level is unlikely the major factor underlying the ectopic activation of Hh signaling by Gli1 in the absence of the cilia.


Introduction
The Hedgehog (Hh) family of signaling molecules underlies numerous developmental processes and malignancies in humans and mice [1]. Hh signaling in mammals requires the primary cilia, a cell surface organelle present in almost all post-mitotic cells in the mammalian body [2]. The glioma-associated oncogene homolog (Gli) family of transcription factors mediates the transcriptional response of Hh signaling, and all three members of the family are localized to the tips of the cilia upon Hh stimulation [3,4]. Given the importance of Hh signaling and the primary cilia in development and diseases, it is critical to understand the roles of the cilia in the activation of Gli proteins.
Sonic hedgehog (Shh), one of the Hh family members, is secreted from cells of the notochord and the floor plate, a group of glia at the ventral midline of the neural tube, and induces various cell fates in the ventral neural tube [1]. Loss of Shh results in the complete loss of ventral cell types including the floor plate, V1, V2 and V3 interneurons and motor neurons [5]. Gli2 is the primary activator downstream of Shh and is essential for the fates of the floor plate and V3 interneurons [6,7].
Gli3 plays a predominantly negative role in Hh signaling, and removing Gli3 restores motor neurons in Shh;Gli3 double mutant neural tube [8]. Gli1 expression is dependent on Gli2 and Gli3, and loss of Gli1 does not disrupt mouse development [9][10][11]. However, loss of Gli1 leads to defects in Shh pathway activation and ventral neural tube development when one copy of Gli2 is removed, suggesting that it contributes to a threshold of Gli activator activity required for the full activation of the Shh pathway [9]. More importantly, Gli1 appears to be critical in pathogenesis of multiple types of malignancies, hence understanding the mechanism of its activation is clinically important [12][13][14][15].
The requirement for the cilia in Hh signaling was first revealed by the loss of floor plate and V3 interneurons, as well as reduced Hh target gene expression, in a few mutants that fail to grow cilia [16]. Specifically, both the activation of full-length Gli2 and the generation of Gli3 repressor through proteolytic processing are dependent on the cilia (e.g., [17][18][19]). We recently showed that removing Gli2 from the tips of the cilia prevents its Hh-dependent activation, further confirming the critical role of cilia in Gli2 activation [20].
Suppressor of fused (Sufu) is an essential negative regulator of Hh signaling in mammals, loss of which results in severe disruption of embryonic development including extreme ventralization of the neural tube [21,22]. Our previous double and triple mutant analyses indicated that all three Gli proteins underlie the extreme Hh pathway activation in Sufu mutants [23]. Biochemical analyses suggested that Sufu acts through direct physical interaction with Gli proteins, both in the cytoplasm and inside the nucleus [24][25][26][27]. Interestingly, loss of Sufu in the absence of the cilia leads to the over activation of Hh pathway, suggesting that the roles of the cilia in Hh signaling is to mediate the Hh-induced alleviation of repression on Gli proteins by Sufu [28,29]. Subsequent biochemical studies showed that separation between Sufu and Gli proteins was indeed dependent on the cilia [30,31].
Although the roles of the primary cilia in Gli2 activation and Gli3 processing have been elucidated, whether the activation of Gli1 is dependent on the cilia remains enigmatic. Gli1 transcription is severely reduced in cilia mutants, precluding the direct analysis of the roles of the cilia in Gli1 activation with these mutants [16,18]. The roles of cilia in Gli1 activation cannot be revealed by overexpressing Gli1 in cultured cilia mutant cells either, as insufficient Sufu is present in the cells to antagonize the activity of overexpressed Gli1, making it constitutively active independent of Hh signaling input and the primary cilia [28,29]. In the current study, we test the roles of the cilia in Gli1 activation by expressing Gli1 at a physiological level from the Gli2 locus (Gli2 1ki ) in cilia mutants. We find that loss of cilia prevented the maximal activation of Gli1 and the formation of the floor plate. Surprisingly, Gli1 was partially activated in the absence of the cilia, resulting in drastic dorsal expansion of the V3 interneuron and motor neuron domains. We show that expressing Gli1 from the Gli2 locus leads to increased motor neuron formation with reduced Gli3 dosage, suggesting that compromised Gli3 repressor production in the absence of cilia may contribute to the partial activation of Hh signaling in the neural tube when Gli1 is expressed from the Gli2 locus in the absence of the cilia. This cilia-independent activation of Gli1 is dependent on Hh signaling because expressing Gli1 from the Gli2 locus does not change neural tube patterning in the absence of Smo. Furthermore, Gli1 expression from the Gli2 locus did not alter neural tube patterning with reduced dosage of Sufu, suggesting it did not activate Hh pathway by changing the balance between dosages of Gli and Sufu proteins. Finally, replacing endogenous Gli2 with a stable, non-ciliary form of Gli2, Gli2 ∆CLR , failed to induce ectopic V3 interneurons and motor neurons in the absence of the cilia, suggesting that the cilia-independent Gli1 activity was not simply the result of higher protein level. In summary, we show different degrees of dependence of Gli1 and Gli2 on the cilia for their activation, suggesting that blocking ciliogenesis may not inhibit malignancies caused by aberrant Gli1 activation.

Immunofluorescence Assay on Cryosections
Mouse embryos were fixed in 4% paraformaldehyde (PFA) in phosphate-balanced saline (PBS) for up to 1 h, washed with PBS briefly and left in 30% sucrose overnight, embedded in O.C.T freezing media and frozen at −80 • C. Cryosections at 10 µm thickness were cut with a Leica LM1900 Cryostat. For immunofluorescence assays, sections were allowed to dry at room temperature for 1 h, blocked in blocking buffer (PBS plus 0.1% Triton X-100 and 1% goat serum). They were then incubated in blocking buffer with appropriate primary antibodies at 4 • C overnight, washed in blocking buffer three times and incubated in blocking buffer with Cy3-conjugated secondary antibodies, wash three more times and mounted with DABCO (Sigma-Aldrich, Saint Louis, MO, USA). Antibodies used: Foxa2, Nk2.2, Nkx6.1, Pax6, Pax7, Shh (DSHB) and Olig2 (Millipore, AB9610). Photos were taken on a Nikon E600 microscope with a Micropublisher CCD camera (QImaging, Surrey, BC, Canada).

Immunoblot Analyses
Whole-cell protein lysates were prepared, separated on SDS polyacrylamide gel and transferred to nitrocellulose membrane according to a previously described protocol [36]. After primary antibody incubation, membranes were incubated with IRD680-and IRD800-conjugated secondary antibodies (LI-COR), and scanned on a LICOR Odyssey CLx imaging system. Antibodies against FLAG and β-tubulin were purchased from Sigma-Aldrich. Quantitative analyses were performed using NIH Image J.

RNA In Situ Hybridization on Cryosections
Embryos were fixed in 4% PFA at 4 • C overnight, washed in DEPC-treated PBS and processed for cryosection. RNA in situ hybridization with Digoxigenin-labeled riboprobes against Gli1 was performed on the transverse sections through the neural tube according to the protocol described in [37]. Photos were taken on a Nikon E600 microscope with a Micropublisher CCD camera.

Gli1 Expression Was Uncoupled from Hh Signaling in Gli2 1ki Embryos
More than a decade of research has started to reveal the essential roles of the primary cilia in the activation of Gli2 [16,20,28,29] and proteolytic processing of Gli3 [17][18][19] in mammals. However, how the Gli1 protein responds to the loss of cilia remains an open question because Gli1 expression is dependent on Hh signaling and is greatly reduced in the absence of the cilia [18]. We and others also showed that an overexpression approach was not appropriate for addressing this question as it rendered both Gli1 and Gli2 cilia-independent due to the disruption of the stoichiometry between Gli proteins and their direct inhibitor Sufu [28,29]. Therefore, the only proper way to address this question is to express Gli1 at a physiological level independent of Hh pathway activation.
To achieve such a goal, we took advantage of a Gli2 1ki mouse strain in which the Gli1 open reading frame was inserted into the first coding exon of Gli2 [32]. Bai and Joyner (2001) showed that Gli1 was expressed in the same pattern as Gli2 in this knock-in animal [32]. To determine whether Gli1 expression became cilia-independent in Gli2 1ki embryos, we performed RNA in situ hybridization analyses. As reported, Gli1 was expressed in a ventral-to-dorsal gradient in the wild type neural tube ( Figure 1A) [38,39]. As a control, Sufu mutant embryos exhibited widespread high levels of Gli1 expression ( Figure 1B) [22]. Consistent with compromised Hh signaling activity in the absence of the cilia, Gli1 expression was greatly reduced in the Ift88 mutant neural tube that failed to grow cilia ( Figure 1C) [16,33]. In contrast, Gli1 was expressed throughout the progenitors of the neural tube in Gli2 1ki/+ ;Ift88 −/− double mutants, suggesting that Gli1 expression from the Gli2 locus was independent of the cilia ( Figure 1D). These data indicated that Gli1 expression from the Gli2 locus was independent of Hh signaling and the presence of cilia, making it possible to analyze the roles of cilia in Gli1 activation.

The Cilia Are Essential for Maximal Activation of Gli1
Bai and Joyner (2001) showed that expressing Gli1 from the Gli2 locus rescued most aspects of embryonic development in the absence of Gli2, suggesting functional redundancy between these two proteins [32]. We hypothesized that if Gli1 activation was independent of the cilia, expressing Gli1 from the Gli2 locus should rescue embryonic development in the absence of the cilia. To test this hypothesis, we crossed Gli2 1ki mice to Ift88 mutants. At E10.5, Gli2 1ki/+ embryos looked normal, consistent with a previous report ( Figure 2A,A'; [32]). Ift88 null mutants exhibited frequent exencephaly and twisted body (Figure 2A"). Gli2 1ki/+ ;Ift88 -/-double mutants also exhibited frequent exencephaly and twisted body (Figure 2A"'). The failure to rescue the Ift88 mutant phenotype appears to suggest that Gli1 activation is under the influence of the primary cilia.
To better determine the roles of cilia in Gli1 activation, we examined the dorsal/ventral (D/V) patterning of the neural tube at both the thoracic (anterior) and lumbar (posterior) levels. As similar neural tube patterning changes were present at both levels, we will focus on the results at the thoracic level. The floor plate at the ventral midline of the wild type neural tube expresses Foxa2 at E10.5 ( Figure 2B). Confirming a previous report [32], we found that Foxa2 was expressed properly in the Gli2 1ki/+ neural tube ( Figure 2B'). Foxa2 expression was absent in the Ift88 mutant neural tube due to compromised Gli2 activation in the absence of the cilia ( Figure 2B"; [16]). Foxa2 expression was also absent in Gli2 1ki/+ ;Ift88 -/-double mutants, suggesting that the maximal activation of Gli1 was dependent on the cilia ( Figure 2B"'). Shh was produced in the notochord and floor plate in the wild type and Gli2 1ki/+ neural tubes ( Figure 2C

The Cilia Are Essential for Maximal Activation of Gli1
Bai and Joyner (2001) showed that expressing Gli1 from the Gli2 locus rescued most aspects of embryonic development in the absence of Gli2, suggesting functional redundancy between these two proteins [32]. We hypothesized that if Gli1 activation was independent of the cilia, expressing Gli1 from the Gli2 locus should rescue embryonic development in the absence of the cilia. To test this hypothesis, we crossed Gli2 1ki mice to Ift88 mutants. At E10.5, Gli2 1ki/+ embryos looked normal, consistent with a previous report ( Figure 2A,A'; [32]). Ift88 null mutants exhibited frequent exencephaly and twisted body (Figure 2A"). Gli2 1ki/+ ;Ift88 −/− double mutants also exhibited frequent exencephaly and twisted body (Figure 2A"'). The failure to rescue the Ift88 mutant phenotype appears to suggest that Gli1 activation is under the influence of the primary cilia.
To better determine the roles of cilia in Gli1 activation, we examined the dorsal/ventral (D/V) patterning of the neural tube at both the thoracic (anterior) and lumbar (posterior) levels. As similar neural tube patterning changes were present at both levels, we will focus on the results at the thoracic level. The floor plate at the ventral midline of the wild type neural tube expresses Foxa2 at E10.5 ( Figure 2B). Confirming a previous report [32], we found that Foxa2 was expressed properly in the Gli2 1ki/+ neural tube ( Figure 2B'). Foxa2 expression was absent in the Ift88 mutant neural tube due to compromised Gli2 activation in the absence of the cilia ( Figure 2B"; [16]). Foxa2 expression was also absent in Gli2 1ki/+ ;Ift88 −/− double mutants, suggesting that the maximal activation of Gli1 was dependent on the cilia ( Figure 2B"'). Shh was produced in the notochord and floor plate in the wild type and Gli2 1ki/+ neural tubes ( Figure 2C

Gli1 Was Partially Activated in the Absence of Cilia
Nkx2.2-expressing V3 interneurons and their progenitors were adjacent to the floor plate and required lower levels of Shh signaling than those required for floor plate ( Figure 2D). The number and location of V3 interneurons were not altered in Gli2 1ki/+ mutants as previously reported ( Figure 2D') [32]. Consistent with an essential role of cilia in Gli2 activation, these neurons were completely absent in Ift88 null mutants ( Figure 2D"). To our surprise, the domain of V3 interneurons was not only present, but also greatly expanded dorsally in the Gli2 1ki/+ ;Ift88 −/− double mutant neural tube, implying an increase in Hh pathway activity ( Figure 2D"'). Motor neuron progenitors expressing Olig2 were dorsal to V3 interneuron progenitors in both the wild type and Gli2 1ki/+ neural tubes, and required moderate activation of the Hh pathway ( Figure 2E Figure 2G"'). In summary, we found that Gli1 could be partially activated in the absence of the cilia.

Ectopic Gli1 Partially Ventralized the Neural Tube in the Presence of Reduced Gli3
Paradoxically, the neural tube patterning indicated that more cells in the Gli2 1ki/+ ;Ift88 −/− neural tube experienced intermediate levels of Hh pathway activation than those of the Gli2 1ki/+ neural tube, suggesting a negative role of the cilia in Hh pathway activation. As it was known that the cilia were essential for the proteolytic processing of Gli3 [18,19,30], we hypothesized that the reduction in Gli3 repressor activity contributed to the increase in Hh pathway activation in lateral regions of the Gli2 1ki/+ ;Ift88 −/− neural tube. To test this hypothesis, we crossed Gli2 1ki/+ mice to Gli3 +/− mice to generate Gli2 1ki/+ ;Gli3 +/− double mutant embryos. At E10.5, Gli2 1ki/+ and Gli3 +/− embryos were indistinguishable from wild type embryos ( Figure 3A,B and data not shown). Gli3 −/− embryos exhibited smaller telencephalon (compare the brackets in Figure 3A,C; n = 4/6) and occasional exencephaly (data not shown; n = 2/6). Interestingly, Gli2 1ki/+ ;Gli3 +/− double mutants exhibited frequent exencephaly ( Figure 3D; n = 4/7), suggesting strong genetic interaction between ectopic Gli1 expression and reduced Gli3 dosage in the patterning and/or proliferation of the brain. We further analyzed the neural tube patterning and found that the size and location of the floor plate and V3 interneuron domains in the Gli2 1ki/+ ;Gli3 +/− double mutant neural tube were similar to those in the wild type, Gli2 1ki/+ and Gli3 −/− neural tubes ( Figure 3F-I,K-N). However, we observed a moderate dorsal expansion of the motor neuron progenitor domain in the Gli2 1ki/+ ;Gli3 +/− double mutant neural tube ( Figure 3S, compared to Figure 3P-R), suggesting that reducing Gli3 repressor activity did allow ectopically expressed Gli1 to moderately ventralize the lateral part of the neural tube through activating lower levels of Hh signaling.   Gli2 1ki/+ ;Gli3 +/− pups were sickly and died shortly after weaning, preventing further breeding. Gli2 1kie was similar to Gli2 1ki , but contained a floxed neo cassette interfering with the expression of Gli1 from the Gli2 locus [32]. Gli2 1kie/+ ;Gli3 +/− male were viable and fertile. Therefore, we crossed these mice with Gli3 +/− ;EIIaCre female mice to obtain Gli2 1kie/+ ;Gli3 −/− ;EIIaCre embryos. As EIIaCre was expressed ubiquitously in the embryos [35], removing the floxed neo cassette and restoring the full expression of Gli1 from the Gli2 locus, these embryos were equivalent to Gli2 1ki/+ ;Gli3 −/− . We found that these embryos also exhibited frequent exencephaly ( Figure 3E; n = 5/5). Although the floor plate and V3 interneurons were defined properly in these mutants ( Figure 3J,O), the motor neuron progenitor domain was drastically expanded dorsally in the neural tube of these embryos ( Figure 3T). These data suggest that Gli3 repressor prevents abnormal neural tube patterning when Gli1 is ectopically expressed from the Gli2 locus, and the reduction in Gli3 repressor likely contributes to the dorsal expansion of the ventral neural progenitor domains of the Gli2 1ki/+ ;Ift88 −/− double mutant neural tube.

Gli1 Expression from the Gli2 Locus Failed to Alter the Smo Mutant Phenotype
The ectopic moderate activation of the Hh pathway in the Gli2 1ki/+ ;Ift88 −/− neural tube was in striking contrast to the Ift88 −/− mutant neural tube where both high and intermediate levels of Hh pathway activities were compromised [16], suggesting a functional difference between Gli1 and Gli2. It is possible that Hh signaling may activate Gli1 independent of the cilia. Alternatively, Gli1 may exhibit Hh-independent basal activity that is sufficient to drive ventral neural tube cell fates in more dorsal regions with the reduction in Gli3 repressor. We hypothesized that if Gli1 exhibited a Hh-independent activity, expressing Gli1 from the Gli2 locus should allow some ventral neural cell fates in the absence of Smo. To test this hypothesis, we analyzed the patterning of the Gli2 1ki ;Smo double mutant neural tube.
Smo −/− mutants were significantly smaller than control littermates at E9.5 and died shortly after [40].  Figure 4O). These observations indicated that Gli1 was not activated independent of Hh signaling, suggesting that the cilia-independent partial activation of Gli1 was dependent on Hh signaling.

Increased Protein Level Was not the Major Reason for the Cilia-Independent Gli1 Activation
Gli1 is resistant to both Cul1-mediated proteolytic processing and Cul3-mediated degradation, making it a more stable protein than Gli2 and Gli3 [41][42][43][44]. It is possible that replacing Gli2 with the more stable Gli1 protein in Gli2 1ki mutant embryos brings a challenge to the Sufu-based negative regulation of Hh signaling, lowering the threshold for Hh pathway activation. If this is true, reducing the dosage of Sufu in the presence of ectopic Gli1 expression, as in the Gli2 1ki/+ ;Sufu +/-and Gli2 1ki/1ki ;Sufu +/-neural tubes, should lead to excess Hh pathway activation and dorsal expansion of some ventral progenitor domains. Our results showed that Gli2 1ki/+ ;Sufu +/-embryos were morphologically normal at E10.5 ( Figure 5A,B, n = 6), whereas one out of 4 Gli2 1ki/1ki ;Sufu +/-embryos exhibit midbrain exencephaly ( Figure 5C). In the neural tube, the floor plate ( Figure 5D-F

Increased Protein Level Was Not the Major Reason for the Cilia-Independent Gli1 Activation
Gli1 is resistant to both Cul1-mediated proteolytic processing and Cul3-mediated degradation, making it a more stable protein than Gli2 and Gli3 [41][42][43][44]. It is possible that replacing Gli2 with the more stable Gli1 protein in Gli2 1ki mutant embryos brings a challenge to the Sufu-based negative regulation of Hh signaling, lowering the threshold for Hh pathway activation. If this is true, reducing the dosage of Sufu in the presence of ectopic Gli1 expression, as in the Gli2 1ki/+ ;Sufu +/− and Gli2 1ki/1ki ;Sufu +/− neural tubes, should lead to excess Hh pathway activation and dorsal expansion of some ventral progenitor domains. Our results showed that Gli2 1ki/+ ;Sufu +/− embryos were morphologically normal at E10.5 ( Figure 5A,B, n = 6), whereas one out of 4 Gli2 1ki/1ki ;Sufu +/− embryos exhibit midbrain exencephaly ( Figure 5C). In the neural tube, the floor plate ( Figure 5D-F  We recently generated a mouse strain in which Gli2 was replaced by Gli2 ∆CLR , a variant that did not enter the cilia but retained its transcriptional activity in the absence of Sufu [20]. Interestingly, the level of Gli2 ∆CLR was higher than full-length Gli2 in E10.5 embryos, enabling us to directly test whether moderately increasing the level of Gli proteins was sufficient for a moderate activation of Hh signaling in the lateral regions of the neural tube in the absence of cilia. As shown in Figure 6, E10.5 Gli2 ∆CLRki/∆CLRki mutants were similar to wild type littermates morphologically ( Figure 6A,B), whereas Gli2 ∆CLRki/∆CLRki ;Ift88 −/− double mutant embryos exhibited exencephaly, twisted body and other morphological defects reminiscent of Ift88 −/− mutants ( Figure 6C, compare to Figure 2C). In the neural tube, Gli2 ∆CLRki/∆CLRki mutants exhibited a reduction of the floor plate ( Figure 6D,E) and V3 interneurons ( Figure 6G,H). On the other hand, Olig2-expressing motor neuron progenitors expanded ventrally in these embryos ( Figure 6J,K), whereas Pax6 expression remained unchanged ( Figure 6M,N). Different from Gli2 1ki/+ ;Ift88 −/− mutants in which Nkx2.2 and Olig2 expression domains were expanded dorsally, the expression of Foxa2 and Nkx2.2 was absent in Gli2 ∆CLRki/∆CLRki ;Ift88 −/− double mutant neural tube ( Figure 6F,I), and the Olig2 expression was expanded ventrally ( Figure 6L). The only sign of moderate rescue of the ventral neural fate in the Gli2 ∆CLRki/∆CLRki ;Ift88 −/− neural tube was the absence of Pax6 expression in the ventral-most part of the neural tube, in contrast to the expression of Pax6 throughout the Ift88 −/− mutant neural tube ( Figure 6O, compared to Figure 2G) [16,18]. These results indicated that Gli2 ∆CLR , albeit exhibiting higher stability than wild type Gli2, did not activate Hh pathway ectopically as did Gli1 in the absence of cilia.
One potential explanation for the difference between the Gli2 1ki/+ ;Ift88 −/− and Gli2 ∆CLRki/∆CLRki ;Ift88 −/− neural tubes was that the level of Gli1 in the former was higher than that of Gli2 ∆CLR in the latter. As a FLAG tag was introduced to the N-termini of both proteins in the knock-in embryos [32], we directly compared the expression levels of these two proteins in E10.5 embryos through immunoblot analyses. We found that the level of Gli2 ∆CLR in Gli2 ∆CLRki/+ embryos was nearly three fold of that of Gli1 in Gli2 1ki/+ embryos (Figure 7, n = 4 embryos for each strain). This result suggested that the surprising activation of the Hh signaling in the lateral neural tube of Gli2 1ki/+ ;Ift88 −/− was not simply the result of higher Gli1 protein level. a FLAG tag was introduced to the N-termini of both proteins in the knock-in embryos [32], we directly compared the expression levels of these two proteins in E10.5 embryos through immunoblot analyses. We found that the level of Gli2 ∆CLR in Gli2 ∆CLRki/+ embryos was nearly three fold of that of Gli1 in Gli2 1ki/+ embryos (Figure 7, n = 4 embryos for each strain). This result suggested that the surprising activation of the Hh signaling in the lateral neural tube of Gli2 1ki/+ ;Ift88 -/-was not simply the result of higher Gli1 protein level. Figure 7. The level of Gli1 in Gli2 1ki/+ embryos was lower than that of Gli2 ∆CLR in Gli2 ∆CLRki/+ embryos. Immunoblots of E10.5 whole embryo lysates with antibodies against FLAG and β-tubulin. **: p = 0.0018 in two-tailed student t-test.

Discussion
In this study, we compare the relationships between the cilia and two Gli family proteins, Gli1 and Gli2, in mouse neural tube patterning. Previous studies have shown that these relationships could not be addressed properly by in vitro overexpression due to the override of Sufu-mediated negative regulation [28,29]. To address this question properly, we compared neural tube patterning between Ift88 mutant and Gli2 1ki ;Ift88 double mutants. We found that although the cilia were required for the full activation of both Gli1 and Gli2, and the formation of the floor plate, they were not required for V3 interneuron and motor neuron progenitor formation when Gli1 was expressed from the Gli2 locus, suggesting that Gli1 activation was not fully dependent on the cilia (Figure 8).
Genetic analyses using Ift88 or Kif3a mutants have been widely used to test the roles of cilia in numerous biological processes (e.g., [45][46][47]). Although Ift88 has been shown to regulate mitotic spindle orientation, immune synapse, and cell migration independent of the cilia, to our best knowledge, no solid evidence exists to support a cilia-independent role for Ift88 in Shh-mediated neural tube patterning [48][49][50][51]. Therefore, we strongly believe that the neural tube patterning changes in the absence of Ift88 likely indicate the roles of the primary cilia in the activation of various Gli variants, rather than a cilia-independent function of Ift88.
One possible explanation for this cilia-independent partial Gli1 activation was that Gli1 was more stable than Gli2, leading to either Hh-independent activation of target gene expression or a lower threshold of cellular response to Hh signaling. Our results appeared to counter this explanation. First, we did not see any difference between Smo and Gli2 1ki ;Smo double mutant neural tube, suggesting that Gli1 was not activated in the absence of upstream Hh pathway input ( Figure   Figure 7. The level of Gli1 in Gli2 1ki/+ embryos was lower than that of Gli2 ∆CLR in Gli2 ∆CLRki/+ embryos. Immunoblots of E10.5 whole embryo lysates with antibodies against FLAG and β-tubulin. **: p = 0.0018 in two-tailed student t-test.

Discussion
In this study, we compare the relationships between the cilia and two Gli family proteins, Gli1 and Gli2, in mouse neural tube patterning. Previous studies have shown that these relationships could not be addressed properly by in vitro overexpression due to the override of Sufu-mediated negative regulation [28,29]. To address this question properly, we compared neural tube patterning between Ift88 mutant and Gli2 1ki ;Ift88 double mutants. We found that although the cilia were required for the full activation of both Gli1 and Gli2, and the formation of the floor plate, they were not required for V3 interneuron and motor neuron progenitor formation when Gli1 was expressed from the Gli2 locus, suggesting that Gli1 activation was not fully dependent on the cilia (Figure 8).
Genetic analyses using Ift88 or Kif3a mutants have been widely used to test the roles of cilia in numerous biological processes (e.g., [45][46][47]). Although Ift88 has been shown to regulate mitotic spindle orientation, immune synapse, and cell migration independent of the cilia, to our best knowledge, no solid evidence exists to support a cilia-independent role for Ift88 in Shh-mediated neural tube patterning [48][49][50][51]. Therefore, we strongly believe that the neural tube patterning changes in the absence of Ift88 likely indicate the roles of the primary cilia in the activation of various Gli variants, rather than a cilia-independent function of Ift88.
One possible explanation for this cilia-independent partial Gli1 activation was that Gli1 was more stable than Gli2, leading to either Hh-independent activation of target gene expression or a lower threshold of cellular response to Hh signaling. Our results appeared to counter this explanation. First, we did not see any difference between Smo and Gli2 1ki ;Smo double mutant neural tube, suggesting that Gli1 was not activated in the absence of upstream Hh pathway input ( Figure 8). Second, reducing the dosage of Sufu in Gli2 1ki embryos did not result in an increase in Hh pathway activity and a change in neural tube patterning, suggesting that it was unlikely that increased Gli1 level led to a lower threshold to Hh response through overriding Sufu function. Finally, we show that Gli2 ∆CLR , with a protein level significantly higher than that of Gli1 in their respective knock-in embryos, did not support ectopic V3 and motor neuron progenitor formation in the absence of cilia, suggesting that the phenotype in Gli2 1ki/+ ;Ift88 −/− double mutants did not result from a simple increase in Gli protein level.
Interestingly, Gli2 1ki neural tube exhibits normal pattern along its D/V axis whereas V3 and motor neuron domains were expanded dorsally in Gli2 1ki ;Ift88 double mutants, suggesting a negative role of the cilia in Hh signaling. We show that reducing Gli3 dosage in Gli2 1ki neural tube similarly resulted in dorsal expansion of these ventral neuronal domains, suggesting that reduced Gli3 repressor activity in the absence of cilia contributed to this negative role of the cilia. Similar negative roles of the cilia have been reported in previous studies of skin and brain tumors caused by activating mutations in Gli proteins [45,46]. activity and a change in neural tube patterning, suggesting that it was unlikely that increased Gli1 level led to a lower threshold to Hh response through overriding Sufu function. Finally, we show that Gli2 ∆CLR , with a protein level significantly higher than that of Gli1 in their respective knock-in embryos, did not support ectopic V3 and motor neuron progenitor formation in the absence of cilia, suggesting that the phenotype in Gli2 1ki/+ ;Ift88 -/-double mutants did not result from a simple increase in Gli protein level. Interestingly, Gli2 1ki neural tube exhibits normal pattern along its D/V axis whereas V3 and motor neuron domains were expanded dorsally in Gli2 1ki ;Ift88 double mutants, suggesting a negative role of the cilia in Hh signaling. We show that reducing Gli3 dosage in Gli2 1ki neural tube similarly resulted in dorsal expansion of these ventral neuronal domains, suggesting that reduced Gli3 repressor activity in the absence of cilia contributed to this negative role of the cilia. Similar negative roles of the cilia have been reported in previous studies of skin and brain tumors caused by activating mutations in Gli proteins [45,46].
One remaining question is why Gli1, but not Gli2, appears to be partially activated in the absence of cilia if the elevated protein level is not likely the major contributing factor. It is possible that Gli1 has a unique response to Hh signaling outside of the cilia. Alternatively, the difference may be quantitative as Gli1 does not appear to have a repressor domain at its N-terminus, making it a much stronger transcriptional activator [3]. Therefore, if a cilia-independent activation mechanism exists to activate Gli proteins at a very low level, the effect should be more detectable with Gli1 ( Figure 8). Although such a cilia-independent pathway for Gli-mediated transcription has not been detected, a non-canonical, transcription-independent Hh/Smo pathway appears to be independent of the cilia [52][53][54]. Further investigation will be needed to reveal the detailed molecular mechanism of the ciliaindependent Gli1 activity. There is no direct evidence for cilia-independent Gli2 activation, but this possibility cannot be completely ruled out yet.
One remaining question is why Gli1, but not Gli2, appears to be partially activated in the absence of cilia if the elevated protein level is not likely the major contributing factor. It is possible that Gli1 has a unique response to Hh signaling outside of the cilia. Alternatively, the difference may be quantitative as Gli1 does not appear to have a repressor domain at its N-terminus, making it a much stronger transcriptional activator [3]. Therefore, if a cilia-independent activation mechanism exists to activate Gli proteins at a very low level, the effect should be more detectable with Gli1 ( Figure 8). Although such a cilia-independent pathway for Gli-mediated transcription has not been detected, a non-canonical, transcription-independent Hh/Smo pathway appears to be independent of the cilia [52][53][54]. Further investigation will be needed to reveal the detailed molecular mechanism of the cilia-independent Gli1 activity.