|Home | About | Journals | Submit | Contact Us | Français|
Cilia formation requires intraflagellar transport (IFT) proteins. Recent studies indicate that mammalian Hedgehog (Hh) signaling requires cilia. It is unclear, however, if the requirement for cilia and IFT proteins in Hh signaling represents a general rule for all vertebrates. Here we examine zebrafish ift57, ift88, and ift172 mutants and morphants for defects in Hh signaling. Although ift57 and ift88 mutants and morphants contained residual maternal protein, the cilia were disrupted. In contrast to previous genetic studies in mouse, mutations in zebrafish IFT genes did not affect the expression of Hh target genes in the neural tube and forebrain and had no quantitative effect on Hh target gene expression. Zebrafish IFT mutants also exhibited no dramatic changes in the craniofacial skeleton, somite formation, or motor neuron patterning. Thus, our data indicate the requirement for cilia in the Hh signal transduction pathway may not represent a universal mechanism in vertebrates.
Cilia are microtubule-based organelles that extend from the surface of almost all vertebrate cells. Originally regarded as vestigial remnants of unicellular eukaryotes, work in the last decade has revealed that cilia perform a wide array of functions in multicellular organisms (reviewed in (Satir and Christensen, 2007). Motile cilia beat in a rhythmic fashion to generate fluid flow across surfaces, whereas immotile primary cilia typically function as specialized sensory structures. Primary cilia can respond to fluid flow via mechanosensitive ion channels or to specific ligands via chemoreceptors concentrated in the ciliary membrane. In addition, a number of neurons contain specialized ciliary structures, such as vertebrate photoreceptor outer segments, olfactory cilia, and the hair cell kinocilia of the inner ear.
The formation and maintenance of cilia requires the conserved process of intraflagellar transport (IFT) (reviewed in (Scholey, 2003). Originally discovered in Chlamydomonas (Kozminski et al., 1993), IFT refers to the bidirectional movement of multisubunit IFT particles along ciliary microtubules and has been described in a number of species, including flies, zebrafish, C. elegans, and mouse (Pazour et al., 2000; Haycraft et al., 2001; Avidor-Reiss et al., 2004; Tsujikawa and Malicki, 2004). The IFT particle is composed of at least 17 distinct IFT proteins that assemble into two smaller subunits, Complex A and Complex B (Scholey, 2003). The kinesin-II motor moves the IFT particle to the ciliary tip in an anterograde fashion, whereas IFT-dynein transports the particle back to the basal body (Rosenbaum and Witman, 2002). In addition to ciliogenesis, it is widely held that IFT is also responsible for ciliary localization of numerous membrane-bound receptors, such as rhodopsin and the TRPV channels OSM-9 and OCR-2, (Marszalek et al., 2000; Pazour et al., 2002; Qin et al., 2005). Thus, mutations in IFT proteins, kinesin-II, or the IFT-dynein cause ciliary defects and result in multiple phenotypes including left-right patterning defects, retinal degeneration, kidney cysts, hydrocephaly, and many other developmental abnormalities.
Sonic hedgehog (Shh) is a member of the Hedgehog (Hh) family of secreted ligands and acts as a morphogen to regulate the patterning and growth of many tissues during vertebrate embryogenesis (Ingham and McMahon, 2001). Shh binds to the membrane protein Patched (ptc), which relieves the inhibition of the membrane protein Smoothened (smo). Upon Shh-dependent activation, smo transduces the signal to the Gli family of zinc-fingered transcription factors. Three gli genes have been identified in vertebrates, gli1, gli2, and gli3. In zebrafish and mammals, both Gli1 and Gli2 function as transcriptional activators (Lee et al., 1997; Karlstrom et al., 2003). In the absence of Hh signaling, the full length Gli3 protein is proteolytically cleaved to form a repressor (Gli3R). This processing is inhibited in the presence of Hh signaling and results in Gli3 functioning as a transcriptional activator (Wang et al., 2000). Components of the Shh pathway are also direct target genes, as an active pathway results in upregulation of ptc and gli1 gene expression.
Several recent studies demonstrated that mutations affecting ciliogenesis in mice also produce phenotypes consistent with abnormal Shh signaling downstream of ptc and upstream of the gli transcription factors (Huangfu et al., 2003; Haycraft et al., 2005; Huangfu and Anderson, 2005; Liu et al., 2005; May et al., 2005; Houde et al., 2006; Caspary et al., 2007). In the neural tube, Shh secreted from the notochord and floorplate generates a concentration gradient with the highest levels of Shh found in the ventral neural tube. This ventral to dorsal gradient of Shh in the neural tube results in varying transcriptional response by the Gli proteins and is necessary for neuronal specification within the ventral neural tube. Null or strong hypomorphic mutations in the mouse ift88 ortholog polaris, the ift172 ortholog wimple, the ift57 ortholog Hippi, and the Ndg5 gene (the murine IFT52 ortholog) all resulted in a loss of motor neuron specification in the ventral neural tube due to loss hedgehog signaling (Huangfu et al., 2003; Liu et al., 2005; Houde et al., 2006). In the mammalian limb bud, Shh signaling activates expression of the Hh target genes Ptc-1 and Gli1. In Ift88 mutant mice, ectodermal and mesenchymal cells of the limb bud lacked cilia and Ptc-1 and Gli1 expression was absent (Haycraft et al., 2005). Similar results were found in Ift52, Ift57, and Ift172 mutant mice (Huangfu and Anderson, 2005; Liu et al., 2005; Houde et al., 2006). Biochemical analysis revealed that processing of Gli3A to the Gli3R repressor form was significantly reduced in Ift88 mutant mice, indicating a requirement for cilia in Gli3 processing. Consistent with a defect in Gli function in the limb bud and similar to defects seen in Gli3 mutants (Hui and Joyner, 1993), all IFT mutant mice examined to date have extra digits in the forelimbs and hindlimbs (Haycraft et al., 2005; Liu et al., 2005; Houde et al., 2006). Furthermore, mutations in the IFT-dynein retrograde motor, Dnchc2, also exhibit defects in spinal cord patterning, limb patterning, the loss of cilia in limb mesenchyme, as well as reduced Gli3 processing (Huangfu and Anderson, 2005; May et al., 2005). Finally, several components of the Hh pathway, including Smo and the Gli proteins, exhibited localization to cilia in limb bud cells and primary cilia of the mouse node (Corbit et al., 2005; Haycraft et al., 2005; May et al., 2005). Combined, these results clearly indicate mammalian Hh signaling requires the presence of a ciliary structure.
Comparisons among different species suggest the requirement for cilia in mediating the Hh response is not universal. Drosophila harboring mutations in the orthologs of ift88 or ift172 survive and do not exhibit phenotypes associated with defective Hh signaling (Han et al., 2003; Avidor-Reiss et al., 2004). Consistent with a role of IFT in the formation and maintenance of sensory neurons, however, these mutants exhibited structural defects a variety of sensory cilia as well as loss of mechanosensation of bristle neurons and chordotonal organs. Mutations also exist in the zebrafish IFT genes ift57, ift88, and ift172. These mutations disrupt cilia in sensory neurons and motile cilia in the pronephric duct (Sun et al., 2004; Tsujikawa and Malicki, 2004; Gross et al., 2005) and morpholino knock-down of ift57 or ift88 causes left-right asymmetry defects resulting from loss of nodal cilia (Kramer-Zucker et al., 2005). To determine if cilia are required for Hh signaling in zebrafish, we analyzed three zebrafish IFT mutants for phenotypes similar to those seen in mutants of the Hh pathway. In contrast to mammals, we did not find evidence that cilia play a role in Hh signaling. We find that neuronal fates in the ventral spinal cord and the expression of several Hh target genes were unaffected in zebrafish IFT mutants. Both somite and neurocranial structures were consistent with normal Hh signaling. The addition of low levels of cyclopamine to sensitize the mutants to perturbations in the Hh pathway failed to produce a phenotype. Although Hh signaling was normal, both IFT mutant and morphant embryos lacked cilia in the otic vesicle and neural tube. The data indicate that zebrafish do not require these IFT genes for the Hh pathway and suggest that the role of cilia in mediating Hh signaling may not be evolutionarily conserved among all vertebrates.
For mutant analysis, we used the ift57hi3417/curly, ift88tz288b/oval, and ift172hi2211/moe mutant alleles, which we will refer to herein as the ift57, ift88, and ift172 mutations, respectively. The ift57 and ift172 mutations resulted from retroviral insertions (Amsterdam et al., 1999), whereas the ift88 mutation resulted from ENU mutagenesis (Malicki et al., 1996) (Fig. 1). Database searches have not revealed other homologs for these genes, which strongly suggest the genes exist as single copies within the genome. These mutants were previously described as having defects in ciliated sensory neurons and pronephric cilia (Sun et al., 2004; Tsujikawa and Malicki, 2004; Gross et al., 2005). When first identified in large-scale mutagenesis screens, however, none were reported to exhibit midline defects or other phenotypes associated with abnormal Hh-signaling, such as abnormal floor plate or somite formation, defects in motor neuron branching, or neurocranial defects (Brand et al., 1996; Sun et al., 2004). Western blot analysis revealed a dramatic reduction in IFT88 protein at 48 hpf and a milder, but noticeable reduction in IFT57 protein in the respective mutants (Fig. 1D).
To complement our analysis of IFT mutants, we injected antisense morpholinos against each of the IFT genes into wild type embryos at the 1-cell stage. We used both translation-blocking and splice-blocking morpholinos that were previously shown to fully phenocopy the IFT mutants (Sun et al., 2004; Tsujikawa and Malicki, 2004; Kramer-Zucker et al., 2005) and found no phenotypic difference between translation-blocking or splice-blocking morpholinos for each of the three targeted genes. Following each series of injections, we confirmed the effectiveness of the morpholinos by randomly reserving approximately 20–30 embryos and examining them for a ventral body curvature and kidney cysts at 48 hpf, as well as by performing RT-PCR for splice-blocking morpholinos (Fig. 9E). For embryos injected with translation-blocking morpholinos against ift57 or ift88, Western blot analysis was performed on protein extracts of 24 hpf embryos (Fig. 9D’). In both instances, the translation-blocking morpholinos significantly reduced the amount of IFT protein present, although some residual protein still remained. Injection of higher concentrations of morpholinos led to nonspecific phenotypes and widespread cell death.
Hh signaling in mammals requires intact cilia and the effects of IFT mutations on the Hh pathway are likely caused by the inability to form ciliary structures. We examined IFT morphants and IFT mutants for the presence of cilia in the neural tube by staining with a monoclonal antibody against acetylated tubulin at 24 hpf. As expected, wild type animals at 24 hpf exhibited numerous elongated cilia extending from cells in the ventral neural tube (Fig. 2A). Cilia were severely disrupted in the IFT mutants and morphants and the limited staining likely reflected staining of basal bodies with acetylated tubulin (Fig. 2B-H). To confirm this hypothesis, sections were stained with acetylated tubulin and gamma tubulin, which is a marker for basal bodies. Consistent with the previous results, cilia were seen in the wild type sections but acetylated tubulin and gamma tubulin colocalized in the IFT mutants and no cilia were observed (Fig. 2I-L). Cilia within the neural tube are likely motile and it is non-motile primary cilia that are required for Hh signaling. Primary cilia are believed to only be present on cells during growth arrest (Christensen et al., 2008), yet zebrafish undergo rapid cell divisions during the first 24 hours of development (Kane et al., 1992; Kimmel et al., 1994). At 18–20 hpf, however, the otic vesicle contains two populations of cilia: tether cilia located at the anterior and posterior ends of the ear and a transient population of “short cilia” distributed throughout the lumen. The characteristics of these two cilia populations remain somewhat controversial. Early studies suggested the short cilia are motile and the tether cilia are immotile (Riley et al., 1997). However, a very recent report (Colantonio et al., 2009) reached an opposite conclusion and stated that the short cilia are immotile (perhaps primary cilia) and tether cilia are motile. We examined IFT morphants at 18 hpf for cilia within the otic vesicle and found that tether cilia were shorter than wild type and that the short cilia were missing or reduced from the IFT morphants (Fig. 2M-P). Taken together, these results demonstrate that mutations in zebrafish IFT genes severely affected cilia structure.
As cilia are required for mammalian Hh signaling, we asked if Hh signaling requires IFT components in zebrafish. We analyzed the expression of several Hh pathway components and Hh target genes at 24 hours post fertilization (hpf) in IFT mutants. As zebrafish IFT mutants cannot be reliably distinguished from wild type embryos by morphological criteria at 24 hpf, we analyzed large numbers (n ≥ 30) of progeny from heterozygous incrosses and looked for differences in approximately 25% of the embryos. We then photographed and subsequently genotyped 10–14 embryos to identify wild type and mutant fish. Although mutations in components of the IFT pathway strongly reduced or eliminated ptc expression in mouse models, no difference in the expression of ptc-1 was observed among IFT mutants (Fig. 3, Panels A1-C4). We did not observe any difference in the staining intensity or pattern between several identified homozygous IFT mutants and non-mutant siblings. To examine the spatial expression of ptc-1 within the posterior neural tube, we isolated DNA from the heads of individual fish following in situ hybridization with ptc-1 and identified mutant and wild type embryos by genotyping. We then analyzed 10 μm-thick cryosections of tail tissue of wild type and mutant embryos (Fig. 3P-S). In both wild type and mutant embryos the expression of ptc-1 appeared similar and extended to the most dorsal aspect of the neural tube, consistent with robust Hh signaling from the floorplate. The zebrafish genome contains two Ptc related genes, ptc-1 and ptc-2 (Concordet et al., 1996). Both zebrafish ptc genes are downstream targets of the Hh pathway (Lewis et al., 1999). We also examined the expression of ptc-2 among progeny from heterozygous incrosses and again did not observe any differences in expression (Fig. 3, Panels D1-F4). Hh signaling also positively regulates the expression of the Gli1 transcription factor. Furthermore, Gli1 is the major activator of Hh target genes in zebrafish (Karlstrom et al., 2003). Loss of individual IFT components or the IFT-dynein motor resulted in loss of Gli1 expression in mouse (Haycraft et al., 2005; May et al., 2005). In contrast, gli1 expression remained unaffected in all three zebrafish IFT mutants (Fig. 3, Panels G1-I2). We next investigated how loss of IFT function affected the expression of gli3. In both mouse and zebrafish, the Hh pathway negatively regulates gli3 expression and embryos lacking Hh signaling (e.g.. smoothened mutants) exhibit a ventral expansion of the gli3 expression domain (Wang et al., 2000; Wijgerde et al., 2002; Tyurina et al., 2005). Although the expression of Gli3 in mouse IFT mutants has not been reported, Gli3 does localize to cilia and loss of cilia disrupts mammalian Hh signaling. One hypothesis is that gli3 expression should be expanded in these mutants as Hh signaling is disrupted. In both mouse and zebrafish, however, the loss of gli1 or gli2 alone or in combination does not affect gli3 expression, suggesting that repression of gli3 occurs independently of gli1 and gli2 (Matise et al., 1998; Tyurina et al., 2005). We did not, however, observe any expansion of the gli3 expression domain in zebrafish IFT mutants (Fig. 3, Panels J1-L2). Finally, we examined the expression of nkx2.2, a marker of the ventral neuroectoderm. Mutations in the zebrafish gli1 or gli2 genes strongly reduce or abolish expression of nkx2.2 in regions of the midbrain and hindbrain (Karlstrom et al., 2003) and mutations in mouse IFT genes reduce the number of nkx2.2 expressing cells in the ventral neural tube (Liu et al., 2005). If the zebrafish IFT mutants disrupt Hh signaling, we expected changes in the expression of nkx2.2, particularly in the midbrain and hindbrain. The highly defined nkx2.2 domain along the ventral neuroectoderm permits the identification of subtle changes in expression. We did not, however, observe any differences in nkx2.2 expression in the anterior CNS between wild type and genotyped IFT mutants, suggesting that Hh signaling was intact (Fig. 3, Panels M1-O2).
Hh signaling negatively regulates the expression of the paired-box (Pax) gene pax6 and a reduction in Hh signaling leads to expansion of the pax6 domain in the eye and ventral neural tube (Macdonald et al., 1995; Karlstrom et al., 2003). We did not observe an increase in pax6 expression in either the eye or neural tube of whole-mounted embryos (Fig. 4A-C). To more carefully analyze pax6 expression, we examined plastic sections of wild type and genotyped IFT mutant embryos following in situ hybridization. Whereas reduction of Hh signaling expands the pax6 domain, we found that pax6 expression was similar, if not slightly reduced, in the lens and neuroepithelium in all IFT mutants as compared to wild type (Fig. 4D-G). pax6 also controls many aspects of retinal development and mutations in pax6 or overexpression of pax6 dramatically disrupt eye development (Schedl et al., 1996; Kozmik, 2005). It should be noted that zebrafish IFT mutants only present with photoreceptor degeneration and were not reported to show any noticeable defects in early eye morphology, dorsal/ventral patterning of the retina, or lens development, all of which are processes controlled by pax6 (Tsujikawa and Malicki, 2004; Gross et al., 2005).
In Hh pathway mutants such as smu (smoothened) and yot (gli2), the specification of the anterior pituitary fails to occur during neurulation and an ectopic lens forms (Kondoh et al., 2000; Varga et al., 2001). At 20 somite stage (~18–20 hpf), the marker pitx3 is lost from the anterior pituitary and from a domain within the diencephalon of smu mutants (Zilinski et al., 2005). We therefore examined the expression of pitx3, a marker of the anterior pituitary in 18 hpf IFT mutants and morphants. We never observed ectopic lens formation in any embryos lacking IFT function (data not shown) and the expression of pitx3 was similar between wild type embryos and those with reduced IFT function (Fig. 4H-K).
To exclude the possibility that the onset of Hh-dependent phenotypes occurs later in zebrafish IFT mutants, IFT mutants were studied at more advanced developmental timepoints when IFT protein levels are reduced (Fig 1D). We examined the expression of shh and ptc-1 at 48 hpf when mutant embryos can be readily identified from non-mutant siblings by a downward curvature of the tail and body axis. Both shh and ptc-1 were expressed in the diencephalon, notochord, floor plate, and in the pectoral fins, structures which are the equivalent to the mammalian limb bud (Fig. 5). Mutations in the mouse IFT genes reduce expression of shh in the floor plate, but do not affect shh expression in the limb bud (Huangfu et al., 2003; Liu et al., 2005; Houde et al., 2006). There is, however, a severe to complete loss of ptc-1 expression in the limb bud of mouse IFT mutants at E10.5 (Haycraft et al., 2005; Liu et al., 2005). In zebrafish IFT mutants, we did not observe changes in the expression of either shh or ptc-1 in any tissues, including the pectoral fins (Fig. 5). This provides further evidence that zebrafish IFT mutants exhibit normal Hh signaling in the limb bud and CNS, even with reduced IFT protein levels.
As in situ hybridization may not accurately reveal subtle changes in gene expression, we performed quantitative real-time PCR (qPCR) for Hh target genes at both 24 hpf and 48 hpf. We obtained RNA from genotyped IFT mutants and wild type siblings at 24 hpf and generated cDNA to use as template for qPCR. Each experiment was performed in triplicate. The relative gene expression of Hh target genes was normalized against beta-actin expression. The expression values from mutant embryos were then compared against values obtained from wild type embryos, which was set at 100%. Consistent with the previous in situ hybridization results at 24 hpf, qPCR revealed that the expression of ptc1, ptc2, gli1, pax6, or nkx2.2 in IFT mutants was highly similar to wild type values (Fig. 6A). Compared to wild type levels, the level of ptc1 was not reduced in ift57 mutants, ift88 mutants, or ift172 mutants. Similar results were seen for ptc2, nkx2.), pax6, or gli1. Furthermore, we did not observe a decrease in Hh target gene expression at 48 hpf, when IFT protein levels were significantly reduced (Fig. 6B). We only examined expression in ift57 and ift88 mutants, as these were the only mutants that we could analyze protein levels by western blotting. Nevertheless, the level of ptc1 was similar to wild type in both ift57 mutants and ift88 mutants and the level of gli1 was also similar in ift57 mutants and ift88 mutants. These results indicate that Hh signaling remained unaffected in the IFT mutants, even with a significant reduction in the level of IFT protein and disruption of cilia.
Loss of Hh signaling results in cyclopia, severe craniofacial defects, and the loss of head and jaw cartilage, as seen in smoothened (smo) mutants (Barresi et al., 2000; Chen et al., 2001). We therefore examined the craniofacial skeleton of IFT mutants by Alcian blue staining (Fig. 7). We also compared the IFT mutant embryos to detour (gli1) mutants and embryos treated with 100 μM cyclopamine, which exhibit the different extremes of craniofacial defects due to alterations in Hh signaling. The alkaloid molecule cyclopamine blocks Hh signaling by directly binding to the Smoothened protein (Chen et al., 2002). In zebrafish, cyclopamine concentrations of 100 μM have been used to completely block Hh signaling and phenocopy smoothened mutants (Karlstrom et al., 2003; Vanderlaan et al., 2005). The detour mutants, however, exhibit some of the mildest craniofacial defects among midlines signaling mutants (Kimmel et al., 2001). Cyclopamine (100 μM) mimicked the phenotype of smo mutants and craniofacial skeletal elements were lost (Fig. 7P). Although we observed subtle differences between wild type and IFT mutants in the length of pharyngeal arches, these phenotypes did not resemble those seen in zebrafish treated with 100 μM cyclopamine. The detour mutant phenotype was less severe than that of cyclopamine treated embryos. As previously noted (Brand et al., 1996; Kimmel et al., 2001), detour mutants showed a reduction in the mediolateral aspect of ethmoid plate and a near fusion of the trabeculae (Fig. 7N, black arrow). In contrast, the IFT mutants possessed all major elements of the pharyngeal skeleton, with no evidence of cyclopia, no reduction in the ethmoid plate, and no fusion of trabeculae (Fig. 7D-L). Taken together, our phenotypic data indicates that the loss of IFT function in zebrafish does not affect Hh-dependent craniofacial development.
As mutations in zebrafish IFT genes do not alter the expression of Hh target genes, we attempted to sensitize the embryos such that subtle changes in Hh signaling might provoke a more robust phenotypic response. We found that cyclopamine concentrations of 10 μM noticeably reduced, and at higher concentrations eliminated, expression of ptc-1, as well as disrupted somite formation, phenotypes which indicate a reduction of Hh signaling (data not shown). We therefore performed all experiments at a cyclopamine concentration (5 μM) that did not qualitatively affect ptc-1 expression in order to observe any synergistic effects of cyclopamine on the IFT mutants. Progeny from incrosses of IFT heterozygotes were exposed to 5 μM cyclopamine at 50% epiboly (~5.5 hpf) and ptc-1 expression was examined at 24 hpf. Even in the presence of a known antagonist to Hh signaling, a consistent level of ptc-1 expression was observed in all embryos (Fig. 8A), indicating that loss of IFT does not affect Hh signaling even in a sensitized background. As a positive control, we utilized the leprechaun (lep) mutant, which disrupts the ptc-2 gene (Koudijs et al. 2005). As ptc is a negative regulator of Smoothened, lep mutants exhibit hyperactive Hh signaling, which can be determined by the increased expression of ptc-1. In embryos from a lep heterozygous incross, 23% (7 of 30) embryos exhibited noticeably stronger ptc-1 staining in both the forebrain and in the trunk (Fig. 8B). Following treatment with 3 μM cyclopamine, all embryos (30 of 30) exhibited normal expression of ptc-1 (Fig. 8B). Loss of lep also causes mild lens degeneration, which manifests as a reduction in pupil size (data not shown) and retraction of the lens into the eye (Fig. 8C). In animals from an incross of lep heterozygous animals, 25% (5 of 20) exhibited this phenotype. Addition of 3 μM cyclopamine also rescued this phenotype, all animals (30 of 30) showed normal eye phenotypes. While low doses of cyclopamine can perturb the Hh pathway in sensitized backgrounds, loss of IFT function did not act synergistically with low concentrations of cyclopamine.
Consistent with our analysis of the IFT mutant alleles, expression of ptc-1 was not altered in the IFT morphant embryos at 24 hpf (Fig. 9A-D). Disruption of the Hh pathway leads to somites that are “flattened” with a more obtuse angle that is greater than 125° (Koudijs et al., 2005). Embryos injected with either translation-blocking or splice-blocking morpholinos possessed somites with the characteristic chevron-shape and we did not observe any defects in floor-plate formation (Fig. 9F-J), both of which are developmental processes requiring a functional Hh pathway (Barresi et al., 2000; Chen et al., 2001; Koudijs et al., 2005). On average, somite angles in wild type animals were 97° (st. dev. 8.8°). Morphants of ift57 (102°), ift88 (92°) or ift172 (98°) revealed no statistical difference (p> 0.1) in somite angles from wild type. To verify these effects were consistent with the IFT mutants, we photographed individual embryos from incrosses of heterozygous IFT mutants and subsequently genotyped each sample (Fig. 9K-O). Among genotyped embryos, we did not observe any statistical difference in somite angles from ift57 (102°), ift88 (98°), or ift172 (98°) homozygous mutants. As Hh signaling from the floor plate and Gli function within the neural tube are essential for motor neuron induction in the ventral spinal cord of zebrafish, we stained IFT mutants and morphants at 24 hpf for islet-1, a marker for primary and secondary motor neurons. Mouse IFT mutants exhibit phenotypes ranging from a severe reduction in islet-1 expression (Houde et al., 2006) to a ventral expansion of islet1/2 positive cells within the neural tube at the expense of V3 interneurons and motor neurons (Liu et al., 2005). Furthermore, zebrafish gli1 and gli2DR mutants show a significant (≥50%) loss of islet-1 expressing cells in the ventral neural tube (Vanderlaan et al., 2005). In zebrafish IFT morphants or genotyped homozygous mutants, however, we did not observe any statistically significant reduction in the pattern or number of islet-1 positive cells at 24 hpf (Fig. 9P-Y). Finally, we examined IFT mutants for the presence of slow muscle cells using the S58 antibody, which specifically labels slow muscle fibers in zebrafish (Devoto et al., 1996). smoothened mutants lack almost all slow muscle fibers detected by the S58 antibody (Barresi et al., 2000). In contrast, we found robust S58 labeling in all three IFT mutants (Fig. 8Z-C’), which further demonstrates that loss of IFT function does not affect zebrafish Hh signaling.
In this study, we examined zebrafish with mutations in three different IFT Complex B genes and demonstrated that these mutants did not show phenotypes consistent with defects in the Hh pathway. Recent evidence has shown importance of cilia, and therefore the IFT process, in facilitating mammalian Hh signaling at a step between the Smo receptor and the Gli transcription factors. Genetic studies in mice have revealed that mutations in several IFT complex B proteins (IFT52, IFT57, IFT88, IFT172), the anterograde motor protein Kif3A, and the retrograde IFT motor (Dnchc2) cause limb and neural tube patterning defects similar to phenotypes seen in Gli mutants (Huangfu et al., 2003; Haycraft et al., 2005; Huangfu and Anderson, 2005; Liu et al., 2005; May et al., 2005; Houde et al., 2006). In zebrafish, the expression of many Hh target genes was unaffected at both 24 hpf and 48 hpf in IFT mutants and craniofacial patterning was grossly normal. Furthermore, the zebrafish mutants did not appear to be sensitized to perturbations in Hh signaling, as addition of low doses of cyclopamine did not reveal any differences in ptc-1 expression. Loss of ift57, ift88, or ift172 function by mutation or morpholino knockdown did not affect somite formation or motor neuron differentiation, phenotypes which were observed in IFT mutant mice. Most significantly, loss of IFT function significantly disrupted cilia in the otic vesicle at 18 hpf and the ventral neural tube at 24 hpf without affecting Hh signaling, which also confirmed the role of IFT in cilia formation and maintenance.
Our results are at odds with a recent report showing zebrafish targeted with a splice-blocking morpholino against another IFT gene, ift80, exhibited reduction in the levels of Hh signaling as assayed with ptc1 transcription (Beales et al., 2007). During a separate investigation of the role of IFT80 protein in photoreceptor development, we injected lower concentrations of the same ift80 splice-blocking morpholino and could achieve complete knockdown of the wild type gene without a concomitant loss of ptc-1 expression (S. L. and B. D. P., unpublished observations). Injecting higher concentrations of ift80 morpholino led to necrosis in the brain and liver and other effects (e.g. very small eyes, severe heart edema, and large kidney cysts), which we considered to be nonspecific.
Could compensation from the maternal protein explain our results? Morpholino knockdown with translation-blocking morpholinos targeted both zygotic and maternal transcripts and resulted in much lower protein levels in ift57 and ift88 morphants at 24 hpf. As discussed below, even partial reduction of IFT function in mice causes measureable defects in Hedgehog signaling. Previous studies found that wild type zebrafish embryos injected with translation-blocking morpholinos against ift88 showed dramatically shortened cilia by 19 hpf in Kupffer’s Vesicle, an organ similar in function to the mammalian node (Bisgrove et al., 2005), which resulted in left-right asymmetry defects. The zygotic null mutants used in this study retain cilia in certain tissues (Sun et al., 2004; Tsujikawa and Malicki, 2004) and this may reflect translation from maternal IFT transcripts or from residual maternal protein in the egg. The ift57 and ift172 genes are expressed from maternal transcripts and cilia were reported in the pronephric duct of zygotic ift57 and ift172 mutants at 50 hpf, suggestive of translation from maternal transcripts (Sun et al., 2004). By 48 hpf, the amount of IFT88 protein is dramatically reduced in the mutants and the levels of IFT57 protein are lower than that observed in wild type embryos, without a concomitant reduction in Hh target gene expression. It should also be emphasized that the principal function of IFT is cilia formation. It is generally assumed that mammalian Hh signaling requires the presence of a primary cilium structure and not IFT per se, as demonstrated by the Hh-deficient phenotypes observed in mutants lacking ciliary proteins such as dynein (May et al., 2005) or Arl13b (Caspary et al., 2007). In this study, both mutant and morphant animals exhibited severely affected cilia in the otic vesicle and neural tube.
It is important to consider if shorter cilia could compromise the Hh signaling pathway. The Tg737orpk mouse line is a hypomorphic mutation in ift88 caused by a DNA integration into an intron near the 3′end of the gene, which partially disrupts expression (Moyer et al., 1994). In contrast to the embryonic lethality associated with targeted deletion of the ift88 gene, the Tg737orpk mice are viable but exhibit numerous phenotypes associated with cilia defects (reviewed in (Lehman et al., 2008). The Tg737orpk mice also exhibit phenotypes consistent with altered Hh signaling, including craniofacial patterning and polydactyly (Zhang et al., 2003), as well as defects in pancreas function and glucose homeostasis (Zhang et al., 2005), which is suggestive of Hh-related defects in the production of β-cells. Further evidence for shorter cilia affecting Hh signaling comes from the flexo (fxo) mutant, which is a stronger hypomorphic allele of Ift88 (Liu et al., 2005, Development). The fxo mutants die at E12.5 but are less severe than complete null alleles of Ift88. Liu et al. (2005) state that fxo mutants retain short cilia in ventral node cells, again suggestive of a partial IFT function. However, the fxo mutants exhibit polydactyly, lack ptc and gli1 expression and exhibit a reduction in shh expression in the floor plate of the neural tube. Thus, the effects on Hh signaling range in decreasing severity between ift88 null allele, the fxo mutant allele, and the Tg737orpk allele. Taken together, data from the literature clearly indicate that shorter cilia do compromise the Hh signaling pathway in mammals. In contrast, our data demonstrate that mutations or morpholino targeting of zebrafish IFT genes reduce protein levels and disrupt cilia but do not compromise Hh signaling by all assays tested.
A second explanation, first suggested by Huangfu and Anderson (2006), is that the IFT pathway may have substituted for the kinesin-like motor Costal2 (Cos2) following the divergence of fish and tetrapods. It should be noted that Cos2 is unrelated to kinesin-II, the motor used in IFT. In flies, Cos2 interacts directly with the C-terminal tail of Smo and helps regulate the activation of the Ci transcription factor, the homolog to vertebrate Gli proteins (Huangfu and Anderson, 2006), and references therein). Cos2 sequesters Ci on microtubules and prevents localization to the nucleus. Similarly, Kif7, a vertebrate ortholog to Cos2, negatively regulates the Hh pathway in zebrafish and directly binds to Gli1 (Tay et al., 2005). Whether Kif7 interacts with the vertebrate Smo protein, however, remains unclear. The C-terminal tail of Drosophila Smo, which interacts with Cos2, shares almost no homology to the zebrafish or mouse Smo protein (Huangfu and Anderson, 2006). The function of Kif7 in mammals has not been tested. While Cos2 negatively regulates the Hh pathway in flies and fish, the function of IFT is to positively regulate the Hh pathway in mice. Thus, zebrafish Hh signaling may only require negative regulation by Kif7, similar to the Drosophila Cos2 paradigm, whereas mammalian Hh signaling may have evolved to require parallel pathways utilizing Kif7 and IFT or the IFT pathway exclusively.
Work in Xenopus has demonstrated that loss of the inturned and fuzzy genes, which result in convergent-extension defects, also disrupt Hh signaling by blocking ciliogenesis (Park et al., 2006). Recent evidence now links the planar cell polarity (PCP) pathway and ciliogenesis (Ross et al., 2005; Park et al., 2006), including a genetic link between Ift88 and PCP in mice (Jones et al., 2008). Intriguingly, Hh may be required for proper PCP signaling in the Drosophila epidermis (Colosimo and Tolwinski, 2006), although cilia are not required for Hh signaling in flies.
Perhaps activation of the Hh pathway in zebrafish requires only cell surface localization of key signaling components. In Drosophila, Hh responsive cells lack cilia but do require Smo localization to the cell membrane (Denef et al., 2000; Zhu et al., 2003; Nakano et al., 2004). The Smo protein requires a hydrophobic and basic residue motif at the C-terminus for proper ciliary localization in MDCK cells (Corbit et al., 2005). While Smo RNA with mutations in this motif failed to rescue zebrafish smu–/– mutants, changes to this motif could also interfere with other aspects of Smo function that are independent of ciliary localization. This does not exclude the possibility that the pathways required for Smo translocation to the membrane or for internalization following activation may utilize some components of the cilia proteome in zebrafish. Recent work by Tobin et al. (2008) has shown that morpholino knock-down of the Bardet-Biedl Syndrome (BBS) proteins BBS4, BBS6, and BBS8 reduced Hh signaling in zebrafish. The BBS proteins localize to the basal body and/or cilia in mammalian cells and C. elegans sensory neurons and the BBS proteins appear required for membrane trafficking to the cilia (Nachury et al., 2007). Loss of BBS protein function, however, does not prevent cilia formation in mice or zebrafish (Mykytyn et al., 2004; Nishimura et al., 2004; Fath et al., 2005; Yen et al., 2006). The BBS proteins also regulate the intracellular Wnt response and the PCP pathway (Gerdes et al., 2007). As BBS6 and BBS12 have homology to chaperone proteins (Kim et al., 2005; Stoetzel et al., 2007) and BBS4 physically interacts with the proteasome (Gerdes et al., 2007), the role of BBS proteins may be in signal integration at periocentriolar regions associated with cilia and near the cell membrane (Gerdes et al., 2007). Perhaps the reduction of Hh signaling in zebrafish BBS morphants also reflects perturbation of a more primitive mechanism that allows the cell to interpret signals at the plasma membrane that must occur regardless of the presence or absence of IFT and cilia. While such a mechanism remains highly speculative, the current understanding remains unclear of precisely how the BBS and IFT complexes cooperate to regulate cilia formation and various development pathways in different species, much less how and when these interactions arose during evolution. Indeed, it is interesting to note that zebrafish lacking the conserved cilia protein ofd1 also show normal Hh signaling but did display shortened cilia, laterality defects, and hydrocephalus, which are all phenotypes associated with defective cilia (Ferrante et al., 2009). Our results demonstrate that mutations in zebrafish ift57, ift88, and fit172 genes disrupt cilia but do not affect Hh signaling. As these results appear different from those in Xenopus or mice, future work is needed to determine when vertebrates first required cilia for Hh signaling.
Zebrafish were maintained as previously described and developmental stages were determined according to Kimmel et al. (Kimmel et al., 1995; Perkins et al., 2005). Animals were maintained in accordance with approved IACUC protocols. The ovaltz288b allele was a gift from Jarema Malicki (Doerre and Malicki, 2002) and the ift57hi3417/curly and ift172hi2211/moe mutations were previously described (Sun et al., 2004; Gross et al., 2005). Live embryos were anesthetized in tricaine (0.4 mg/mL) and photographed with a Canon digital camera mounted on an AxioImager microscope. Images were obtained with a PlanApo 20x objective using Nomarski optics and processed using Adobe Photoshop.
To identify homozygous mutant embryos at 24 hpf, genotyping was performed on whole embryos, heads, or tails (depending on the experimental needs) as described below. Embryos processed for ptc-1 and nkx2.2 in situ hybridization were digitally photographed with identical camera settings and the entire embryo subsequently genotyped to match photographs with mutants and non-mutant siblings. Similarly, following whole-mount antibody staining with 39.4D5, which recognizes islet-1, individual embryos were photographed and subsequently genotyped. Following whole-mount immunostaining with the S58 antibody, the heads were removed from individual embryos using a razor blade. DNA was extracted from individual heads and used for genotyping reactions. The tails from confirmed homozygous mutant and homozygous wild type embryos were subsequently imaged for S58 immunoreactivity. Embryos processed for cilia staining via acetylated tubulin immunohistochemistry were fixed overnight in 4% paraformaldehyde. Following fixation, the tails were removed from individual embryos at the most posterior edge of the yolk. Tail DNA was extracted and used for genotyping. The remaining trunks from genotyped mutant and wild type embryos were used for immunohistochemistry. A similar procedure was done for embryos following in situ hybridization with a pax6 probe.
ift88 homozygous mutants and non-mutant siblings were identified at 24 hpf by genotyping the Bcl1 RFLP (MTJM10RFLP) that was previously identified (Tsujikawa and Malicki, 2004). This RFLP eliminates a Bcl1 restriction site. It is located within the ift88 locus and tightly linked to the ovaltz288b allele. The 202 bp RFLP was amplified by PCR using primers 5′ ATGGTGCAGGATTGCCTATT 3′ and 5′ CTTTACATTGGGAGTCGGGT 3′. To genotype embryos from heterozygous incrosses, Bcl1 was added to samples immediately following PCR amplification, digested for 2 hrs at 50 °C, and analyzed on 2.5% agarose gels. PCR from the mutant allele produces a 202 bp fragment, whereas the wild type allele is digested into two smaller fragments, the higher of which can be easily resolved on agarose gels (Figure 1C).
ift57 and ift172 mutants were genotyped by PCR, which used a three-primer combination in each reaction to identify the presence of the viral insert. In ift57 mutants, the ~6 kb virus sequence inserted approximately halfway into the first exon, 50 nucleotides from the ATG start codon (Figure 1A). We used the forward primer 5′ GCTGCTGCAGAATAGCCGTG 3′ and the reverse primer 5′ GGGGACCAGAAACTAACTTTACTC 3′ to generate a 448 bp product from the wild type allele. MSL4 is a viral-specific reverse primer that generates a 210 bp product with the forward primer (Figure 1B). The MSL4 sequence is GCTAGCTTGCCAAACCTACAGGT. In homozygous mutant embryos, only the 210 bp product is generated. In heterozygous animals, the 210 bp product is generated from the mutant allele and the 448 bp product is generated from the wild type allele (Figure 1C). In ift172 mutants, the virus inserted at the splice donor of the second exon (Figure 1A). We used the forward primer 5′
GATGGAGCTGCTAAAGTCACCTGTAT 3′ and the reverse primer 5′
GGCCCAGCCATAAGTGTAAC 3′ to generate a 461 bp product from the wild type allele.
NLTR3 is a different viral-specific reserve primer that generates a 265 bp product with the forward primer (Figure 1C). The NLTR3 sequence is CTGTTCCATCTGTTCCTGAC.
Morpholino knockdown of was carried out using morpholinos directed against the splice site (SP) or translation initiation site (AUG). The following sequences were used: ift57-AUG,
5′-CCTCCATCAACACGAACATCTGATA- 3′; ift57-SP,
5′-AGTGTTATCGCCTCACCAGGGTTCG-3′; ift88-AUG, 5′-
5′-CAACTCCACTCACCCCATAAGCTGT-3′; ift172-AUG, 5′-
GACTCAGGGCAGTTATAAGAACGTA-3′. Control morpholino, 5′-
Localization of mRNA by in situ hybridization was done using digoxigenin labeled riboprobes as described (Jowett and Lettice, 1994). Embryos from heterozygous incrosses or morpholino injections were analyzed at 24 hpf or 48 hpf for differences in the expression of specific markers among progeny. The following researchers kindly provided cDNA constructs to generate riboprobes: shh (Stenkamp et al., 2000), Deborah Stenkamp (Univ. of Idaho, Moscow, ID), ptc-1, ptc-2, gli1, gli2, gli3, and nkx2.2, (Karlstrom et al., 2003), Rolf Karlstrom (Univ. of Massachusetts, Amherst, MA). Images were obtained on a Zeiss StereoLumar fluorescent stereomicroscope using an AxioCam digital camera.
Cyclopamine (Sigma; (Taipale et al., 2000) was made as a 10 mM stock solution in 95% ethanol. Embryos were collected and staged at 50% epiboly prior to plating in embryo medium at 5.5 hpf. Cyclopamine was diluted to the desired concentration in embryo medium containing 0.5% DMSO and embryos were raised to 24 hpf. Control animals were plated in embryo medium containing 0.5% DMSO without cyclopamine. The working concentration used for the described experiments was 5 μM. For experiments with lep1 mutants, the working concentration was 3 μM.
Western blot analysis was performed as previously described (Krock and Perkins, 2008), with the following modifications. Embryos were deyolked and prepared according to previously published protocols (Link et al., 2006). Pre-case polyacrylamide gels (10%, BioRad) were loaded with 15 embryo equivalents per lane. Following transfer to PVDF membranes, immunoblotting was performed using a rabbit polyclonal IFT88 antibody (1:3000) or rabbit polyclonal IFT57 antibody (1:3000). Alpha tubulin was used as a loading control and detected by immunoblotting with monoclonal anti-α-tubulin (1:2000; Sigma T6199). We used HRP-conjugated anti-mouse or anti-rabbit secondary antibodies for detection as previously described (Krock and Perkins, 2008).
Whole-mount immunostaining was performed as described previously (Chitnis and Kuwada, 1990) with the modifications that 1% DMSO was included in all solutions and 0.2% Triton X-100 was used. The 39.4D5 antibody, which recognizes islet (Korzh et al., 1993), was used at 1:200 and a monoclonal antibody against acetylated alpha-tubulin (Sigma, St Louis, MO) was used at 1:100 dilution to identify cilia. S58 is a monoclonal antibody that recognizes slow-muscle fibers in zebrafish (Devoto et al., 1996) but does not label aldehyde-fixed tissues. For S58 staining, embryos were fixed in Carnoys fixative (60% ethanol, 30% chloroform, 10% glacial acetic acid) for one hour on ice. Embryos were rehydrated through an ethanol series (10 min × 95%, 85%, 70%, 50%, and 30% EtOH) at room temperature. At this point, immunostaining was performed similarly to that described above. S58 was used at 1:10 dilution. The appropriate Alexa secondary antibodies (Invitrogen) were diluted 1:50 and used for fluorescent detection. Both S58 and 39.4D5 were obtained from the Developmental Studies Hybridoma Bank (DSHB, University of Iowa)
Immunohistochemistry was performed as previously described (Perkins et al., 2005). Anti-acetylated tubulin (Sigma, St Louis, MO) was used to label cilia and was diluted 1:500. Anti-mouse Alexa-488 secondary antibody was diluted 1:500 and rhodamine phalloidin (1:500) was included in the secondary antibody solution to label actin. Slides were counterstained with DAPI (Invitrogen) to label DNA. Embryos for whole-mount immunohistochemistry were mounted on depression slides in PBST. Whole-mount images were obtained on a Zeiss AxioImager fluorescent microscope using a PlanNeoFluar 10x objective and an AxioCam digital camera. Immunohistochemistry sections were viewed with a Zeiss ApoTome and a 63x PlanApo objective and digital images obtained with an AxioCam. Images were prepared using Adobe Photoshop software. Alcian Blue staining was done as previously described (Kimmel et al., 1998).
For real-time RT-PCR experiments from samples at 24 hpf, embryos were anesthetized in tricaine and the tails surgically removed for genotyping while the head/trunk region was immediately placed in Trizol reagent and stored at –80 °C. Homozygous mutants and wild type sibling were identified by genotyping (24 hpf). For experiments performed from samples at 48 hpf, embryos were genotyped by visual inspection and the entire embryo was placed in Trizol for RNA extraction. For each IFT gene, RNA from 5–7 mutants or wild type siblings were pooled and cDNA was synthesized using the Superscript kit (Invitrogen). cDNA yield was similar between wild type and mutant samples at each timepoint. The concentration of cDNAs was then normalized to 400 ng/μL. Real-time PCR reactions were prepared using SYBR-Green reagents and performed in triplicate. Gene expression was standardized against an internal beta-actin reference sample and gene expression in mutant samples was then compared to that of wild type sibling samples. At 24 hpf, homozygous wild type siblings of each genotype were used comparison against mutants. At 48 hpf, phenotypically wild type sibling were used for comparison. Transcript levels were quantified and analyzed according to published protocols (Rajeevan et al., 2001).
We are grateful to Deborah Stenkamp and Rolf Karlstrom for sharing reagents and probes. We also thank Rolf Karlstrom for sharing detour mutants and Jeff Gross for sharing leprechaun mutants. We are indebted to Tom Schilling, Tailin Zhang, and Courtney Alexander for assistance with the neurocranium dissections. We also thank Jeff Gross, Arne Lekven, Bruce Riley, and Suma Datta for comments on early versions of the manuscript and for the valuable efforts of reviewers to make this manuscript better. This work was supported by the NEI RO1 EY017037 award (B. P.)
This work is supported by NIH Grant: EY017037 to B.P.