Mouse intraflagellar transport proteins regulate both the activator and repressor functions of Gli transcription factors

The Hedgehog (Hh) family of secreted proteins play important roles in regulating growth and patterning in both invertebrate and vertebrate animals(Ingham and McMahon, 2001). In mammalian embryos, Hh proteins are required for normal development of numerous organ systems. Loss of Hh function results in defects in almost every aspect of mouse embryogenesis, including loss of left-right asymmetry, dorsalization of the central nervous system (CNS) and a great reduction in digit number(Chiang et al., 1996). Hh proteins are also important survival factors and/or mitogens for multiple cell types, and unregulated activation of Hh signaling is implicated in the formation of various tumors (Bak et al.,2003).

Despite extensive studies of the Hh pathway, the transduction and regulation of Hh signaling in vertebrate has yet to be fully understood(Lum and Beachy, 2004). In Drosophila, Hh protein binds and inactivates its receptor Patched(Ptc), allowing the activation of Smoothened (Smo), another transmembrane protein. The principal target of Hh signaling is the transcription factor Cubitus interruptus (Ci). In the absence of Hh signaling, Ci is proteolytically processed into a transcriptional repressor, whereas Hh signaling inhibits such processing and allows the unprocessed Ci to act as a transcriptional activator. In the mouse, homologues for some major components of Hh signaling have been isolated and exhibit conserved functions in Hh signaling. However, important divergence exists between the Hh signaling pathway in insects and in vertebrates. First, owing to gene duplication, there are three Hh proteins (Shh, Ihh and Dhh) and three Ci homologues (Gli1, Gli2 and Gli3) in the mouse. Among the three Gli proteins, Gli1 does not contain a repressor domain and cannot be proteolytically processed, and thus appears to be an obligate activator (Dai et al.,1999). Gli2 and Gli3 contain both repressor and activator domains(Dai et al., 1999; Sasaki et al., 1999). Gli3 can be processed into transcriptional repressors in vivo(Wang et al., 2000), although it remains to be determined whether Gli2 is processed in vivo. However, both biochemical and genetic evidence indicates that Gli3 predominantly acts as a repressor and Gli2 predominantly as an activator(Ding et al., 1998; Hui and Joyner, 1993; Matise et al., 1998). Furthermore, novel regulators of Hh signaling that appear to be specific to vertebrates have been isolated, suggesting that vertebrates have evolved different mechanisms to regulate Hh activity(Chuang and McMahon, 1999; Huangfu et al., 2003; Izraeli et al., 1999; Lee et al., 2001; McCarthy et al., 2002). This paper focuses on two novel regulators, both IFT proteins, to elucidate the molecular mechanisms underlying their regulation of Hh signal transduction.

The pattern of vertebrate limbs along the anteroposterior axis and formation of digits are regulated by the interaction between sonic hedgehog(Shh) and Gli3 (Niswander,2003). Shh is expressed in a small group of mesenchymal cells in the posterior distal region of the early limb bud (the zone of polarizing activity, ZPA). Application of either ZPA cells or Shh protein to the anterior region of the limb bud leads to the formation of extra digits(polydactyly) in chick and mouse (Liu et al., 1998; Riddle et al.,1993). By contrast, all digits but one are lost in the absence of Shh protein function, reflecting the requirement for Shh function in digit formation (Chiang et al.,1996). Conversely, loss of Gli3 function as found in the mouse mutant extra-toes results in the formation of multiple digits and there is ectopic Shh expression in the anterior region of the limb bud (Buscher et al., 1997; Hui and Joyner, 1993). Ectopic Hh activity, or loss of Gli3 repressor function, has also been observed in several other mouse and chicken mutants with polydactyly(Wang et al., 2000; Yang et al., 1998). Interestingly, in the absence of Shh and Gli3 function, skeletal formation is rescued and multiple digits are formed although they are unpatterned(Litingtung et al., 2002; te Welscher et al., 2002b). From this, it has been concluded that Shh-Gli3 interactions serve to control the number and pattern of the digits.

Intraflagellar transport (IFT) proteins are required for the biogenesis of flagella and cilia in multiple organisms, including the green alga Chlamydomonas reinhardtii, C. elegans, insects and mouse(Rosenbaum and Witman, 2002). We have previously shown that in the mouse spinal cord, mutations in two IFT proteins, IFT88 (TTC10 - Mouse Genome Informatics) and IFT172, lead to loss of several ventral cell types and reduced expression of the Hh downstream gene Ptch1 (Huangfu et al.,2003). However, the ventral defects in these mutants are not as severe as those in Shh or Smo mutants, raising the issue of whether these IFT proteins are essential components of the Hh pathway. In addition, these IFT mutants die around E10.5, preventing more thorough analysis of Hh signaling in other organs, especially in the limb where requirement for the repressor function of Gli proteins can be better addressed. Orpk, an existing hypomorphic allele of mouse Ift88, survives to young adulthood with multiple defects, including an extra thumb and polycystic kidneys(Moyer et al., 1994). However,it has been reported that Hh signaling and anteroposterior patterning gene expression is normal in Orpk limbs except for an anterior expansion of Fgf4 expression in the apical ectodermal ridge (AER)(Zhang et al., 2003).

In this report, we show that flexo, a novel hypomorphic mutant of mouse IFT88, forms multiple digits in all four limbs despite the lack of ectopic Hh signaling and the downregulation of Hh activity in its normal domain. Hypomorphic mutants for another mouse IFT gene, Ift52, exhibit similar phenotypes to flexo mice in multiple organs. We also show, by double mutant analysis, that IFT proteins are required for tissues to respond to Hh ligand. Finally, we show that IFT proteins regulate Gli activity in part through the proteolytic processing of the Gli3 protein. Thus, IFT protein function is required for both the activator and repressor activities of the Gli proteins in the Hh pathway.

The generation and genotyping of Ift88^null and Ift88^hypo are as described previously(Huangfu et al., 2003). cDNA sequence of mouse Ngd5 gene (Ift52 homolog; Genbank Accession Number L38481) was used to Blast against Bay Genomics database(http://baygenomics.ucsf.edu/). Two gene-trap ES cell clones, XL826 and RRJ295, were identified to carry mutations in Ngd5 genes. The Ift52^hypo mice were generated by blastocyst injection of these ES cells and kept on C3H background. Mice carrying these mutations were genotyped by PCR using lacZ-specific primers (Liu et al., 1999) or MIT markers (D2mit49 and D2mit226) flanking the IFT52 gene. Ptch1-lacZ and Shh mutant mice are genotyped as described (Goodrich et al.,1997; Chiang et al.,1996).

Whole-mount RNA in situ hybridization on embryos was performed as described(Liu et al., 1998). Immunohistochemical study on frozen cryosectioned tissue was performed as described (Timmer et al.,2001).

Protein lysate was prepared from E10.5 wild type, Gli3^-/-,Ift88^null Ift88^hypo or Ift52^hypo mouse whole embryo, dissected spinal cord or limb buds. Equal amounts of protein were loaded onto 7% SDS-PAGE gels and western blotting was performed as described(Wang et al., 2000). Antiβ-tubulin antibody (Sigma, #T4026) was used as loading control. The result of western blot was quantitated using NIH Image 1.60.

Flexo (Fxo) is a recessive mouse mutation that causes multiple defects, including defects in dorsoventral patterning of the CNS (see Fig. 4) and polydactyly (6-9 digits) in all four limbs (Fig. 1A-D). Sequence analysis of DNA from Fxo embryos indicates that a splice mutation in mouse Ift88 leads to a 29 amino acid deletion in the IFT88 protein(Huangfu et al., 2003). However, a normal Ift88 transcript is detected at very low level and shortened cilia are detected in ventral node cells (data not shown). The similar but milder phenotype observed in Fxo mutant embryos relative to Ift88-null embryos indicates Fxo is a hypomorphic allele of Ift88 (also Fig. 7for more evidence). In order to avoid unnecessary confusion, we refer to the Fxo allele as Ift88^hypo and to the null allele as Ift88^null.

Ift88^hypo mutants survive until E12.5, allowing us to examine anteroposterior patterning in the Ift88^hypo limb bud. Normally, Fgf4 is expressed in the posterior aspect of the AER at E10.5 (Fig. 1E) and begins to be downregulated at E11.5 (Niswander and Martin, 1992). We found that Fgf4 expression in Ift88^hypo embryos is expanded to the entire length of AER at E10.5 (Fig. 1F) and its expression is maintained at E11.5 in the anterior AER (data not shown). Homeobox genes Hoxd11 and Hoxd13 are both normally expressed in the posterior mesenchyme of the limb buds at E10.5 and E11.5(Fig. 1G,I; data not shown),and ectopic expression of Hoxd genes in the anterior parts of the limb has been shown to cause polydactyly and homeotic transformation(Knezevic et al., 1997; Morgan et al., 1992). We found that both genes are expressed in an expanded region at both E10.5 and E11.5(Fig. 1H,J; data not shown). Therefore, the formation of extra digits in the limb coincides with an anterior-to-posterior transformation.

Hh signaling is required for maintaining posterior genes such as Fgf4,Hoxd11 and Hoxd13, and plays essential roles in digit formation(Chiang et al., 2001; Kraus et al., 2001). Ectopic Hh expression and/or Hh activity revealed by Hh target gene expression such as Ptch1 and Gli1, has been found to be associated with polydactyly as well as ectopic Fgf4 and Hoxdexpression in mouse and chick (Hui and Joyner, 1993; Yang et al.,1998). We therefore examined whether the polydactyly phenotype in Ift88^hypo limbs is associated with ectopic expression of Shh and Ptch1. In mutant limbs at both E10.5 and E11.5,there is no ectopic Shh expression and Shh is expressed normally in the posterior mesenchyme (Fig. 2A,B; data not shown). There is also no ectopic Ptch1expression in Ift88^hypo limbs(Fig. 2C,D; data not shown). Therefore, the formation of extra digits does not appear to correlate with ectopic Hh expression or activity. Surprisingly, the endogenous Ptch1 expression is greatly reduced, at both E10.5 and E11.5(Fig. 2C,D and data not shown). Gli1 is another target gene of Hh signaling and its expression appears to depend strictly on Hh signaling(Bai et al., 2004). Gli1 expression in Ift88^hypo limb is also downregulated at E10.5 (Fig. 2E,F). Thus, although Shh is normally expressed, there appears to be a loss of Hh signal transduction in the Ift88^hypo mutant limb bud.

It has been suggested that a mutual repression between Gli3 and Hand2 in the early limb buds sets up a limb prepattern that determines the anteroposterior polarity of the limb(te Welscher et al., 2002a). This prepattern exists before Shh expression initiates in the limb and hence is not affected by a mutation in the Shh gene(Chiang et al., 2001). We therefore examined the expression of Hand2 and Gli3 at E10.25 and E10.5 to determine whether the limb prepattern is altered in Ift88^hypo limbs. Normally, Hand2 is expressed in the posterior mesenchyme of the early and later limb buds(Fig. 2G,I). In Ift88^hypo limbs, Hand2 expression is both upregulated and anteriorly expanded (Fig. 2H,J). This suggests that the limb prepatterning defect results from a failure of Gli3 to repress Hand2 expression; however, Gli3 RNA is present in Ift88^hypo limbs(Fig. 2K-N). Later, Gli3 expression expands posteriorly(Fig. 2N), which is probably due, in part, to the loss of Shh signaling. Taken together with the data above, the analysis of Ift88^hypo limbs indicates loss of both Gli activator function (downregulation of Ptch1 and Gli1 expression) and Gli repressor function (Hand2misexpression and polydactyly).

Our results that both Hh signaling and limb patterning along the anteroposterior axis are altered in Ift88^hypo mutants are in contrast to a previous report in which there was no change detected in the expression patterns of Shh, Ptch1 or Hoxd genes in limbs of another hypomorphic allele of mouse Ift88, Orpk(Zhang et al., 2003). In contrast to Ift88^hypo, the Orpk mutant mice can survive to postnatal stage and the polydactyly phenotype is much milder than Ift88^hypo (Moyer et al., 1994). Therefore, it is possible that an alteration in gene expression would be too mild to detect in the Orpk mutant embryos.

Ift88^hypo is so far the only mouse IFT mutant to survive long enough to reveal the role for IFT88 in the regulation of both activator and repressor functions of Gli proteins. In order to determine whether the phenotype seen in Ift88^hypo reflects a general requirement for IFT proteins in Gli-positive and -negative regulation, we sought additional IFT mutant alleles that allow a characterization of both CNS and limb development.

Mouse Ngd5 shares sequence homology with Chlamydomonas Ift52 (Wick et al., 1995)(data not shown). In addition, Ngd5 protein physically interacts with other known mammalian IFT proteins, indicating that it is the mammalian ortholog of IFT52 (Baker et al., 2003). Using a Blast search, we identified two ES cell lines (Bay Genomics) in which the Ngd5 gene is disrupted by insertion of the lacZsequence. In the first line (XL826), the C-terminal 99 amino acids of Ngd5 protein are replaced with lacZ sequence, whereas in the second allele(RRJ295), the last 47 amino acids are replaced with lacZ(Fig. 3A). We derived mutant embryos from both mutant ES cell lines and they both exhibited similar defects in embryogenesis. Therefore, we refer to both lines as Ift52^hypo.

At E10.5, homozygous Ift52^hypo mutants have a tighter mesencephalic flexure compared with their wild-type littermates(Fig. 3B,C) and/or a neural tube closure defect (Fig. 3G). In wild-type embryos, the heart tube always loops to the left(Fig. 3D), whereas about half of the Ift52^hypo homozygous mutant embryos display right-looping heart tubes, indicating defects in left-right axis determination(Fig. 3E). Craniofacial development is also abnormal in Ift52^hypo mutant embryos. At E10.5, normal mouse embryos have separated olfactory pits and maxillary arches (red and blue arrowheads, respectively, in Fig. 3F). By contrast, in the E10.5 Ift52^hypo mutant embryos both the olfactory pits and maxillary arches are fused at the ventral midline (red and blue arrowhead,respectively in Fig. 3G). At E12.5, polydactyly (6-7 digits) is seen in all four limbs(Fig. 3H,I). As Ptch1expression is downregulated in Ift88^hypo limb buds,indicating defective Hh signaling, we examined whether similar downregulation also occurs in Ift52^hypo limb buds. Ptch1, which is normally expressed in the posterodistal region of the limb buds at E10.5(Fig. 3J), is not expressed in the Ift52^hypo limb buds(Fig. 3K), indicating IFT52 is also required for Hh signaling.

The tight mesencephalic flexure, left-right and ventral midline defects as well as polydactyly seen in Ift52^hypo embryos are very similar to what is seen in Ift88^hypo embryos. Loss of Ptch1 expression in the Ift52^hypo limb buds further indicates that, similar to IFT88, IFT52 is also required for normal Hh signaling. We therefore examined ventral patterning of the spinal cord in both Ift52^hypo and Ift88^hypo mutant embryos. Our published studies of Ift88 null mutants(Ift88^null) show that the floorplate, V3 interneurons and motoneurons are all absent (Huangfu et al., 2003). In Ift88^hypo and Ift52^hypo mutant spinal cord, Shh-expression in the floor plate is absent (Fig. 4A,E,I). Nkx2.2-expressing cells that give rise to V3 interneurons and that are normally dorsal and lateral to floorplate(Fig. 4B) are still present in the mutants, but they occupy the ventral most region of the spinal cord and the number is greatly reduced (Fig. 4F,J). Motoneurons that express Isl1/2 are normally located dorsally to V3 interneurons (Fig. 4C); in the two hypomorphic IFT mutants, these cells expand into more ventral regions (Fig. 4G,K). Interestingly, scattered motoneurons are found in the ventral midline of the spinal cord, presumably among V3 interneurons. Pax6 is normally strongly expressed in motoneuron precursors and weakly expressed in more dorsal regions of the spinal cord(Fig. 4D). In the two mutants, Pax6 is ectopically expressed in scattered cells in more ventral regions of the spinal cord (Fig. 4H,L). Therefore, both Ift88^hypo and Ift52^hypo mutant embryos exhibit similar defects in ventral patterning of the spinal cord, and this phenotype is consistent with compromised Hh signaling.

In order to further address the relationship between IFT proteins and Hh signaling, we characterized the double mutant between Ift88 and Shh. Ift88^null mutants have defects in ventral spinal cord patterning that is similar to Shh mutants. However, interesting differences exist between the two mutants. In Shh mutants, the spinal cord is severely dorsalized. Pax7, which labels the dorsal progenitor cells, is expanded throughout the entire spinal cord(Fig. 5A,B). Strong Pax6 expression, which labels neural progenitor cells in the ventral-intermediate region of the spinal cord, is shifted to the ventral most part of the spinal cord (Fig. 5E,F). Lhx3 is normally expressed in both motoneurons and V2 interneurons, and En1 is expressed in differentiated V1 interneurons that are dorsally adjacent to V2 interneurons(Fig. 5I,M). In Shhmutants, very few En1- and Lhx3-expressing cells are present in the ventral midline of the spinal cord(Fig. 5J,N). By contrast, in the Ift88^null spinal cord, Pax7 expression remains dorsally restricted (Fig. 5C). Strong Pax6 expression is detected in a broader domain expanded both ventrally and dorsally(Fig. 5G). En1- and Lhx3-expressing cells are both present but expanded to the ventral midline (Fig. 5K,O). One interpretation for the apparent milder ventral defects in Ift88^null is that Hh can partially signal through Gli in the absence of IFT88, which would predict that loss of Hh ligand in addition to IFT88 leads to a severe dorsalizing defect similar to that seen in Shh mutants. We hence generated double Ift88^null; Shh^-/- mutant embryos to determine whether this is indeed the case.

In the spinal cord of the double homozygous mutants between Ift88^null and Shh (Ift88^null; Shh^-/-), Pax7 expression is dorsally restricted(Fig. 5D) and Pax6 is strongly expressed in most of the spinal cord, except for the most dorsal region, where it is expressed at a lower level(Fig. 5H). En1- and Lhx3-positive cells are present and expanded into the ventral midline(Fig. 5L,P). These patterns of expression of spinal cord neural progenitor and differentiated cell markers resemble that in Ift88^null mutants, indicating that IFT88 is epistatic to Shh and it is an essential regulator of Hh signaling or Gli function. It has been suggested that severe ventral spinal cord patterning defects in Shh mutants may result from unregulated Gli3 repressor function (Litingtung and Chiang,2000). Therefore, a more plausible interpretation for the Ift88^null spinal cord phenotype is that there is a loss of both Gli activator and repressor function. Indeed, when all Gli activities are ablated in mouse embryos, the dorsoventral patterning of the spinal cord is similar to that in Ift88^null embryos(Bai et al., 2004). Thus, we suggest that in Shh mutants, Gli3 repressor function remains active leading to a severe patterning defect, whereas in Ift88^null,Ift88^hypo and Ift52^hypo mutants, there is a loss of both Gli activator and repressor functions leading to less severe patterning defects.

The double mutants between Shh and Ift88^nulldie around E10.5. Therefore, in order to address the relationship between IFT proteins and Hh signaling during limb patterning, we generated double mutants between Shh and Ift88^hypo. At E12.5, Shhembryos are smaller than their littermates and exhibit severe midline defects with the loss of most craniofacial structures replaced with a proboscis-like structure and a cyclopic eye (Fig. 6A) (Chiang et al.,1996). The double mutants between Shh and Ift88^hypo are of normal size, and some of the craniofacial structures are restored and the two eyes are separated by midline tissue(Fig. 6B; data not shown). The general morphology of the double mutants is similar to Ift88^hypo single mutants(Fig. 6C), consistent with our observation in the spinal cord of Shh/Ift88^null double mutants.

It has been shown that anteroposterior patterning and formation of digits in the mammalian limbs depends on the interaction between Shh and Gli transcription factors, mainly Gli3(Niswander, 2003), and that the major function of Shh in the limb is likely to antagonize the repressive function of Gli3 (Litingtung et al.,2002; te Welscher et al.,2002b). In E12.5 Shh mutant limbs, the posterior tissue degenerates, leaving a narrow pointy bud with a single digit condensation(Fig. 6D)(Chiang et al., 2001; Kraus et al., 2001). By contrast, the Shh/Ift88^hypo double mutants display polydactyly (Fig. 6E), similar to Ift88^hypo mutants(Fig. 6F). Moreover, the polydactyly is similar to that observed following loss of Gli3. In the absence of Gli3, multiple digits are formed, with the anterior expansion of otherwise posteriorly located genes such as Fgf4, Hoxd11 and Hoxd13 (Buscher et al.,1997; Hui and Joyner,1993). Ift88^hypo mutant limb buds also display an anterior expansion of Fgf4, Hoxd11 and Hoxd13, along with the formation of extra digits, suggesting a loss of Gli3 repressor function. The fact that the polydactyly of Ift88^hypomutants is not altered by loss of Shh function also suggests that there is a loss of Gli3 repressor function in Ift88^hypo mutants. However, the loss of the Shh targets Ptch1 and Gli1 in both the limb and other places in Ift88^hypo embryos, but not Gli3 mutant embryos, also indicate a loss of Gli activator functions. The simplest explanation for such paradoxical phenotype is that IFT proteins are required, independently of Hh input, for both the activator and repressor function of Gli transcription factors.

To address directly the question of how IFT proteins may regulate the Gli transcriptional activities at molecular level, we examined whether mutations in IFT genes affect Gli3 protein processing. We performed western blot analysis using an antibody against the N terminus of Gli3 that recognizes both the Gli3 activator and repressor forms(Wang et al., 2000). In E10.5 wild-type whole embryos, most full-length Gli3 protein (Gli3-190) is proteolytically processed into the repressor form Gli3-83, and the amount of Gli3-190 is minimal (Fig. 7A,lane 1). By contrast, in Ift88^null mutants, the amount of processed Gli3 is significantly reduced as the full-length protein is dramatically increased (Fig. 7A, lane 2). As a result, the ratio of Gli3-190 versus Gli3-83 is markedly increased in the Ift88^null mutant embryos when compared with that in wild type (1.13 to 0.39)(Fig. 7B). Interestingly,although Gli3 processing was also disrupted in Ift88^hypomutant embryos (Fig. 7, compare lane 3 with lane 4), the quantitative analysis revealed that Ift88^hypo mutation has less effect on Gli3 processing than Ift88^null mutation does, as measured by the ratio of Gli3-190 to Gli3-83 (Fig. 7B). Similar results were obtained in Ift52^hypo mutant embryos at the same stage (data not shown), indicating that the severity of the mutant phenotype correlates well with the defects in Gli3 processing. We also examined the extent of Gli3 processing in spinal cord and limb buds and found a similar accumulation of Gli3-190 and a reduction of Gli3-83 (data not shown), indicating that IFT proteins are essential for regulating Gli3 processing in diverse tissues. Our results in three different mutants strongly suggest that IFT regulates Gli3 at a step prior to its proteolysis. It is of note that, despite the accumulation of Gli3 full-length proteins, the IFT mutant embryos lack the molecular and embryological hallmarks of Gli activator function, indicating the full-length Gli proteins present in the mutants lack the ability to fully activate downstream genes.

Our analysis indicates that newly synthesized, unmodified Gli proteins do not possess full transcriptional activities. IFT proteins, directly or indirectly, modify the Gli proteins (perhaps by targeting them to the appropriate location in the cell for modification), allowing them to acquire full transcriptional activity. Such modification by IFTs also primes the Gli proteins for further processing through proteolytic cleavage to generate Gli repressors. This proteolytic processing is negatively regulated by Hh. The modified Gli proteins, either the activator or the repressor, are then transported to the nucleus to regulate gene transcription. A novel gene, iguana, has been identified in zebrafish to regulate the nuclear import of Gli proteins (Sekimizu et al.,2004; Wolff et al.,2004). Our results have not addressed the possibility of an additional role for IFT proteins at this important step. Thus, in the absence of IFT function, Gli proteins lack the modifications that are crucial for their activator and repressor activities, as well as their competence to respond to Hh signaling.

In Drosophila, Costal 2 (Cos2) and Fused(Fu) appear to play similar bipartite roles in Hh signaling such that both gain and loss of Hh signaling phenotypes are observed in such mutants(Lefers et al., 2001; Wang and Holmgren, 2000). Cos2, a Kinesin II-related protein without a functional motor, seems to play multiple roles in Hh signaling, including recruiting Ci to the cell surface through Smo and cytoplasmic retention of Ci(Jia et al., 2003; Lum et al., 2003; Ogden et al., 2003; Ruel et al., 2003; Wang and Holmgren, 2000). Fu is a Kinase whose only function appears to be antagonizing the function of Sufu, a negative regulator of Ci. Interestingly, when Fu is mutated, Ci cleavage is blocked and full-length Ci accumulates in the cells(Lefers et al., 2001).

Hh signaling in vertebrates appears to be more complex than that in insects. First, three Gli family members play different roles in mediating Hh signals, whereas in Drosophila there is only a single Ci gene that encodes both an activator and a repressor. Second, the existence and function of Cos2 and Fu in the vertebrate has yet to be confirmed. Third, IFT proteins that are crucial for mouse Hh signaling do not appear to mediate Hh signaling in Drosophila (Han et al.,2003; Sarpal et al.,2003). Therefore, vertebrates may use different sets of proteins to regulate Gli3 processing and subcellular localization of Gli proteins. Finally, as the loss of IFT functions leads to simultaneous loss of cilia and abnormal Hh signaling, it is difficult to distinguish a more direct function of IFTs as a component of the Hh pathway in the regulation of Gli protein modification, or a more indirect function, through signals associated with cilia.

Despite the divergence on the molecules involved, the vertebrate and insects may share similar strategy in regulating Gli/Ci activities. Kif3a, a subunit of mouse kinesin 2, plays important roles in Hh signaling(Huangfu et al., 2003). Although Kif3a does not seem to be the mouse ortholog of Cos2 according to sequence homology, it may be functionally equivalent. IFT proteins may also be a novel addition to the Gli-Kinesin (Ci-Cos2) complex that serves to bring Gli proteins to Smo upon activation of the pathway similar to what has been shown in Drosophila (Jia et al.,2003; Lum et al.,2003; Ogden et al.,2003; Ruel et al.,2003; Wang and Holmgren,2000). It will be interesting in the future to elucidate whether a functional relationship exists between these proteins, and to determine the subcellular site(s) of IFT action and their interacting protein partners.

In summary, we have shown that IFT proteins are required for the proteolytic processing of Gli3 protein. Although there is an increase in the ratio of full-length activator to repressor Gli3 protein forms, this is not reflected genetically as the IFT mutants lack the hallmarks of in vivo Gli activator functions. Our analysis of two novel IFT mutants, as well as double mutants for Ift88 and Shh, demonstrate that Gli activator as well as Gli repressor functions are disrupted and that IFT is an essential component of the Hh signaling pathway. Thus, IFT proteins are required for both the activator and repressor functions of Gli transcription factors.

We thank Dr P. Beachy for Shh mutant mice and Dr M. Scott for the Ptc-lacZ mice. For in situ probes, we thank Dr D. Duboule for Hoxd11 and Hoxd13, Dr K. Ligon for Hand2, and Dr A. Joyner for Gli1 and Gli3. We are grateful to the MSKCC Mouse Transgenic Core Facility for the blastocyst injections. We thank Drs J. Hooper, L. Sussel and I. Zohn for critically reading the manuscript. Monoclonal antibodies against Shh, Isl1/2, Nkx2.2 (developed by Jessell), Pax6 and Pax7 (developed by Kawakami) were obtained from Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa. The ES cell lines used to generate Ift52 mutant mice are from Bay Genomics. This work was supported by a fellowship from NICHD to A.L. (F32-HD045090), an NIH grant to B.W. (R01 CA111673) and an NIH grant to L.A.N. (RO1 HD32427). L.A.N. is a HHMI investigator.