The intraflagellar transport protein IFT57 is required for cilia maintenance and regulates IFT-particle–kinesin-II dissociation in vertebrate photoreceptors

Vertebrate photoreceptors are highly specialized neurons that possess a modified sensory cilium known as the outer segment. The outer segment develops as an extension of a nonmotile primary cilium (De Robertis, 1960). As the outer segment lacks the machinery for protein synthesis, all protein destined for the outer segment must pass through the connecting cilium. Large amounts of protein synthesized in the inner segment must be efficiently transported to the outer segment to replenish material lost from the distal tips each day. Estimates from mammalian systems have calculated ∼2000 rhodopsin molecules per minute must be transported to the outer segment to compensate for lost material (Besharse, 1990). Hence, both the development and survival of the photoreceptor require this continual transport of protein to the outer segment (Marszalek et al., 2000).

Studies of rhodopsin trafficking in several species have linked defects in protein transport to photoreceptor degeneration and the disease retinitis pigmentosa. The C-terminal tail of rhodopsin contains a sorting sequence that is necessary and sufficient for transport to the outer segment (Perkins et al., 2002; Tam et al., 2000). Mutations in this region result in protein accumulation in the inner segment and at the base of the connecting cilium in mice, rats and frogs (Green et al., 2000; Li et al., 1996; Sung et al., 1994; Tam et al., 2000), leading to photoreceptor degeneration. Indeed, mutations in the C terminus of human rhodopsin, such as P347L and S344Ter, can cause retinitis pigmentosa (Berson et al., 1991). Rhodopsin mislocalization also occurs in animals with mutations in the molecular motors kinesin II (Marszalek et al., 2000) and the dynein light chain Tctex-1 (Tai et al., 1999), both of which show severe retinal degeneration. It is imperative, therefore, that cargo targeted for the outer segment reach its destination or retinal degeneration will occur. Thus, both mutations within the opsin gene and mutations in the transport machinery can cause retinal degenerative diseases.

Protein transport along a ciliary axoneme, such as the connecting cilium, occurs via the process known as intraflagellar transport (IFT) (Rosenbaum and Witman, 2002). Both the assembly and maintenance of cilia require IFT and defects in ciliogenesis have been linked to retinal degeneration, polycystic kidney disease, Bardet-Biedl syndrome, Jeune asphyxiating thoracic dystrophy, respiratory disease and defective left-right axis determination (Beales et al., 2007; Pazour and Rosenbaum, 2002; Snell et al., 2004). IFT refers to movement of the IFT particle, a multisubunit protein complex that consists of at least 17 IFT proteins that form two subcomplexes: complex A and complex B (Cole et al., 1998). The IFT particle associates with heterotrimeric kinesin II, which comprises two motor subunits and an accessory subunit, known as KIF3A, KIF3B and KAP, respectively (reviewed by Cole, 1999). Kinesin II cooperates with a homodimeric kinesin known as OSM-3 to mediate transport of the IFT particle and its associated cargo toward the ciliary tip (Cole et al., 1998; Orozco et al., 1999; Ou et al., 2005; Snow et al., 2004). The process of IFT is highly conserved, as mutations in IFT proteins perturb ciliary assembly and/or maintenance in organisms as diverse as Chlamydomonas, C. elegans, Drosophila, mouse, humans and zebrafish (Beales et al., 2007; Cole et al., 1998; Han et al., 2003; Murcia et al., 2000; Pazour et al., 2000; Pedersen et al., 2005; Sun et al., 2004; Tsujikawa and Malicki, 2004).

Recent biochemical studies, predominantly in Chlamydomonas, have started to reveal the structural composition of the IFT particle and specific interactions between individual IFT proteins, particularly within complex B. Eleven proteins constitute the Chlamydomonas complex B, a subset of these forms a core consisting of an IFT72/74-IFT80 tetramer along with IFT88, IFT81, IFT52 and IFT46 (Lucker et al., 2005). The outer surface of complex B is composed of IFT20, IFT57, IFT80 and IFT172. Data from yeast two-hybrid experiments indicate direct interactions between IFT72/74 and IFT81, and between IFT57 and IFT20. Similar approaches have indicated interactions between IFT20 and the KIF3B subunit of kinesin II (Baker et al., 2003; Lucker et al., 2005). Although the IFT72/74-IFT80 interaction probably forms the structural core of complex B, the functional nature of the interactions described for the outer surface IFT proteins remains unclear.

Previous studies investigating mutations in IFT genes have revealed few phenotypic differences in ciliated structures of any tissue. In Chlamydomonas, mutations in genes coding for complex B proteins, such as IFT52, IFT88 and IFT172, result in a complete absence of flagella (Cole, 2003). IFT88 mutations have been shown to abolish cilia in the sensory neurons of C. elegans and Drosophila (Han et al., 2003; Haycraft et al., 2001). In zebrafish, mutants of IFT88 and IFT172 lack outer segments entirely, and IFT88 mutants lack all sensory cilia at 4 days post fertilization (dpf) (Gross et al., 2005; Tsujikawa and Malicki, 2004). In mice, all null alleles of IFT88 and IFT172 cause embryonic lethality before E12, thereby preventing analysis of photoreceptor structure, though nodal cilia are completely absent in these animals (Huangfu et al., 2003; Murcia et al., 2000). In Tg737^orpk mutants, which have a hypomorphic mutation in murine IFT88, photoreceptors display aberrant outer segment disk stacking, accumulation of vesicles and progressive photoreceptor degeneration (Pazour et al., 2002; Pazour et al., 2000). However, recent evidence suggests that loss of individual IFT proteins may not completely abolish ciliogenesis. Although not completely normal, cilia do remain in Chlamydomonas cells that lack IFT27, which plays a role in cell cycle regulation (Qin et al., 2007), or IFT46, which facilitates transport of outer dynein arms (Hou et al., 2007). Phenotypic differences have not yet been described in other tissues or species.

Although the photoreceptor phenotypes associated with the partial or complete loss of function of IFT88 have been well characterized in both mouse and zebrafish, no such analysis has been made for most of the remaining 16 or so IFT peptides. Loss-of-function studies with the zebrafish IFT140 and IFT81 did not reveal a retinal phenotype, although the IFT81 mutation did cause cystic kidneys (Gross et al., 2005; Sun et al., 2004; Tsujikawa and Malicki, 2004). Morpholino knockdown of the zebrafish IFT52 and IFT57 genes resulted in a loss of photoreceptors (Tsujikawa and Malicki, 2004); however, the ultrastructure, development and morphology of photoreceptors in these animals were not analyzed. Although photoreceptors clearly require the IFT process for proper outer segment biogenesis, the composition of the IFT particle functioning in the photoreceptor may be different from the one in Chlamydomonas. Many cargo molecules destined for the outer segments, such as rhodopsin, are unique to photoreceptors. Vertebrate photoreceptors also have a simpler axonemal structure (9+0 microtubule arrangement) than the one found in the Chlamydomonas flagellum or vertebrate motile cilia (9+2 arrangement).

Herein, we analyze zebrafish with an insertional mutation in the ift57 gene, which have a photoreceptor phenotype that is distinct from IFT88 mutant zebrafish. Our data show that the process of IFT can occur, albeit inefficiently, in the absence of IFT57. Our data also attribute specific functions to IFT57 and IFT20 within the IFT complex, and provide novel insights into how kinesin II dissociates from the IFT particle. This work has implications for both the molecular mechanism of IFT and the molecular requirements for photoreceptor outer segment formation.

To determine the effects different IFT mutations on photoreceptor development, we examined the phenotypes of zebrafish IFT57 and IFT88 mutants. In a screen for photoreceptor defects, we previously identified a mutation in the zebrafish IFT57 homolog (Gross et al., 2005). The hi3417 allele is a retroviral insertional mutation (Amsterdam and Hopkins, 1999) in the first exon of the IFT57 gene. This mutant has been reported to form kidney cysts (Sun et al., 2004), but the retinal phenotype of IFT57 mutants has yet to be fully characterized. Zebrafish oval mutants carry an ENU-induced point mutation in the IFT88 gene that introduces a premature stop codon, thereby eliminating function (Tsujikawa and Malicki, 2004).

At 4 days post-fertilization, both IFT57 and IFT88 mutants exhibited a ventral body curvature, had slightly smaller eyes and developed kidney cysts (supplementary material Fig. S1). To confirm that the retroviral insertion in IFT57 causes the observed phenotype, we injected splice site-directed morpholino oligonucleotides into wild-type embryos. Injection of gene-specific morpholinos phenocopied the morphological and kidney phenotypes of both IFT57 and IFT88 mutants (supplementary material Fig. S1). These results show that the general phenotype of both mutants is highly similar, and suggest that the IFT57 mutation represents a functional null allele.

To compare the retinal anatomy of IFT57 and IFT88 mutants, we analyzed histological sections of 4 dpf animals by light microscopy. Retinal lamination and normal cellular differentiation was unaffected in IFT57 and IFT88 mutants at 4 dpf (Fig. 1A-C). Both IFT57 and IFT88 mutants exhibited holes within the photoreceptor layer, which is indicative of cell death, whereas other cell types within the retina were unaffected. Cell death specifically within the outer nuclear layer indicated that photoreceptors are the only cell type within the retina whose survival was affected by the loss of IFT proteins. Consistent with previous findings (Doerre and Malicki, 2002; Tsujikawa and Malicki, 2004), higher magnification images (Fig. 1D-F) found no photoreceptor outer segments in IFT88 mutants. By contrast, IFT57 mutant photoreceptors retained short outer segments in both the periphery and central regions of the retina. Morpholino-injected animals (morphants) were phenotypically identical to the mutants at 4 dpf (Fig. 1E,F, compare with Fig. 1H,I). To test whether the IFT57 mutant phenotype reflected a hypomorphic mutation, we performed western blots on lysates of 4 dpf IFT57 mutant embryos with a polyclonal antibody against the C terminus of zebrafish IFT57 (see Materials and Methods). We did not detect any IFT57 protein in mutant lysates (Fig. 1G), and concluded that the hi3417 allele caused a null mutation in the IFT57 gene. These results demonstrate phenotypic differences resulting from mutations in two different IFT complex B proteins.

As IFT57 mutants produced outer segments, we hypothesized that components of the phototransduction cascade would be transported to the outer segments. Immunohistochemical analysis with an antibody against rhodopsin, 1D1, and an antibody against blue cone opsin revealed both rhodopsin and blue cone opsin were present within the outer segments of IFT57 mutant rods and blue cones, respectively (Fig. 2A-F). Opsin mislocalized to the inner segment in IFT57 mutant photoreceptors, although transport to the outer segments did occur (Fig. 2B,E). Both rhodopsin and blue opsin were completely mislocalized throughout the plasma membrane of IFT88 mutant photoreceptors (Fig. 2C,F). These data indicate that transport to the outer segment via IFT was disrupted but not abolished in the absence of IFT57 but did not occur in the absence of IFT88. To address the issue of whether IFT57 mutants generated connecting cilia, we stained retinal sections with anti-acetylated tubulin (Fig. 2J-L). Consistent with the observation that outer segment formation and opsin transport occurs in IFT57 mutants, we found cilia projecting apically from the inner segment of wild type and IFT57 mutants. Ciliary projections were not observed in IFT88 mutants. However, when stained with ZPR1, a marker for red-green double cone morphology, we found that IFT57 and IFT88 mutant photoreceptor morphology was abnormal (Fig. 2G-I).

As mislocalization of opsin can cause photoreceptor degeneration, and defects in opsin transport are associated with disorganized outer segments (Pazour et al., 2002), we used transmission electron microscopy (TEM) to examine the photoreceptor morphology of 4 dpf IFT57 and IFT88 mutants (Fig. 3A-C). Wild-type photoreceptors had elongated outer segments, while IFT88 mutants exhibited no photoreceptor outer segments. By contrast, both rod and cone outer segments were seen in IFT57 mutants. Consistent with results from immunohistochemistry, basal bodies and connecting cilia were observed in IFT57 mutant photoreceptors. We observed vesicle membranes accumulating near the connecting cilium in IFT57 mutants. These vesicles were similar to those seen in photoreceptors expressing a dominant-negative Rab8, which disrupted transport of vesicular cargo through the connecting cilium (Moritz et al., 2001). These membranes were probably post-Golgi vesicles that reflect inefficient protein transport through the outer segment, which is consistent with results of opsin mislocalization in the inner segment. Although the outer segments were shorter and ciliary transport was affected, the disk membranes remained well ordered and tightly stacked in the IFT57 mutants. Upon quantification, IFT57 mutants were found to have photoreceptor outer segments that were reduced in length by 75% compared with wild-type photoreceptors (Fig. 4A).

To quantify how the loss of IFT57 and IFT88 affected rhodopsin transport, we performed immunogold analysis on IFT57 and IFT88 mutants using the anti-rhodopsin antibody 1D1 (Fig. 3F-H). Although the rhodopsin immunogold label localized primarily to the rod outer segments in IFT57 mutants (Fig. 3G), gold particle density was reduced by 59% when compared with age-matched wild-type controls (Fig. 4B). In IFT88 mutants, gold-labeled rhodopsin localized predominantly to the apical part of the inner segment but was also distributed in the plasma membrane (Fig. 3H), indicating that vesicular trafficking from the Golgi to the apical surface is unaffected by loss of IFT88. These data demonstrate that outer segment formation and opsin transport requires IFT88, whereas IFT can function at reduced efficiency without IFT57.

Biochemical studies have shown that IFT57 directly interacts with IFT20 and that IFT20 directly interacts with the KIF3B subunit of kinesin II (Baker et al., 2003). Baker and colleagues (Baker et al., 2003) proposed a model wherein IFT57 recruits IFT20 to the IFT particle and IFT20 serves as the link between kinesin II and the IFT particle. To investigate the nature of the IFT57-IFT20 interaction in vivo, we analyzed the localization of IFT20 in IFT57 mutant photoreceptors. In immunohistochemical analysis, the polyclonal antibodies we generated against IFT20 failed to label with sufficient specificity for conclusive results. We therefore injected one-cell embryos with a plasmid containing an IFT20-GFP fusion protein under the control of a Xenopus laevis rhodopsin promoter to express IFT20-GFP transiently in a subset of zebrafish rod photoreceptors. In transgenic rods of 4 dpf wild-type embryos, IFT20-GFP was observed at the base of the connecting cilium, as demonstrated by colocalization with anti-acetylated tubulin (Fig. 5A). IFT20-GFP also localized to the base of the cilium in IFT57 mutant rods, indicating that ciliary localization of IFT20 does not require IFT57 (Fig. 5B). Consistent with recent reports that IFT20 also localized to the Golgi apparatus (Follit et al., 2006), we observed that IFT20-GFP fluorescence colocalized with antibodies against syntaxin 6, a marker for the trans-Golgi network, in both wild type and IFT57 mutants (Fig. 5C,D). As mutation of IFT88 completely abolishes IFT and our IFT57 mutant exhibits some functional IFT, we predicted that loss of IFT57 would not affect the localization of IFT88 to the connecting cilium. When we immunolabeled IFT57 mutant retinal sections with an anti-IFT88 antibody, we observed IFT88 staining that colocalized with acetylated tubulin. The pattern of localization of IFT88 also indicated that IFT88 was present within the connecting cilium itself, consistent with our assertion that IFT57 mutants retain some functional IFT. Taken together, these results indicate that IFT57 is not required for the normal localization of IFT20 or IFT88 to the base of the connecting cilium.

We next analyzed the composition of the IFT particle in IFT57 mutants in order to understand how the process of IFT could occur in the absence of the IFT57 protein. We first performed western blotting experiments on wild-type and IFT57 mutant larval lysates (Fig. 6A). Protein blots probed with antibodies against KIF3A, IFT88, IFT52 and IFT20 showed that IFT protein levels were present at levels that would probably be sufficient for biological activity. As shown previously, we did not detect measurable levels of IFT57 protein in mutant lysates. These results suggested that loss of IFT57 did not significantly alter the expression of other IFT components.

We next performed co-immunoprecipitation experiments on 4 dpf wild-type and IFT57 mutant lysates. Briefly, 4 dpf zebrafish larvae were homogenized and IFT proteins co-immunoprecipitated using a polyclonal antibody against zebrafish IFT88 (Fig. 6B). We found that IFT52, IFT20 and the KIF3A subunit of kinesin II could be co-precipitated from wild-type zebrafish lysates. When these experiments were performed on IFT57 mutant lysates, we found that KIF3A and IFT52 co-precipitated, but IFT20 did not. We were surprised to find that KIF3A co-precipitated from IFT57 mutant lysates, indicating that kinesin II can interact with the IFT particle independent of IFT20.

When we performed the co-immunoprecipitation experiments on IFT57 mutant lysates, we noticed that significantly more KIF3A was present in our precipitates from IFT57 mutant lysates than from wild-type lysates. Given that the process of IFT is defective in IFT57 mutants, but still occurs, we hypothesized that kinesin fails to dissociate from the IFT particle in the absence of IFT57. The interaction between kinesin II and the IFT particle is salt sensitive and ATP dependent, such that addition of 1 mM ATP to lysates blocks precipitation of kinesin II with the IFT particle (Baker et al., 2003). Consistent with these results, we found that kinesin II did not co-precipitate from wild-type lysates in the presence of 1 mM ATP (Fig. 6C). When IFT57 mutant lysates were incubated with 1 mM ATP, however, the association between kinesin II and the IFT particle is maintained (Fig. 6C). However, if protein lysates were incubated with 10 mM ATP, the association between kinesin II and the IFT particle is lost in both wild type and IFT57 mutants (Fig. 6D). These data show that the ATP-dependent dissociation of kinesin II from the IFT particle is impaired, and suggest a role for IFT57 and IFT20 in mediating motor dissociation from the IFT particle.

In this study, we investigated and compared the roles of IFT57 and IFT88 in vertebrate photoreceptor development. Unlike the mouse Tg737^orpk allele of IFT88, which forms outer segments, the zebrafish IFT57 mutant phenotype is not the result of a hypomorphic allele. Although a genome duplication occurred during the evolution of the lineage leading to zebrafish, only about 30% of these genes were retained (Postlethwait et al., 2000). IFT57 paralogs have not been identified in the zebrafish genome databases that might compensate for its function and generate a hypomorphic phenotype (Sun et al., 2004) (this study). Our western blotting analysis of IFT57 mutants also excludes maternal contribution of protein or mRNA as an explanation for the ciliary phenotype of IFT57 mutants at 4 dpf. Even if an undetectable amount of IFT57 protein exists in IFT57 mutants, the stoichiometric composition of the IFT particle would be perturbed, probably resulting in a null phenotype. Taken together, these results demonstrate that the loss of IFT57 and IFT88 can produce distinct phenotypes that reflect functional roles for these proteins within the IFT complex, as discussed below.

Many studies of loss-of-function phenotypes in vertebrate IFT genes focus attention on phenotypic similarities and rarely emphasize differences, which implies that all IFT mutant phenotypes should be identical. Mutations in the mouse IFT52, IFT57, IFT88 and IFT172 orthologs all lead to identical defects in cilia formation, sonic hedgehog signaling and motoneuron differentiation (Houde et al., 2006; Huangfu et al., 2003; Liu et al., 2005). A mouse knockout of IFT57 was reported to have loss of nodal cilia; however, the analysis did not exclude the possibility of cilia formation and subsequent degeneration (Houde et al., 2006). Morpholinos that disrupt zebrafish IFT88 and IFT57 gene expression produce similar defects in left-right asymmetry, kidney function and cilia motility, although subtle differences in cilia length and phenotypic outcomes were mentioned but not elaborated (Bisgrove et al., 2005; Kramer-Zucker et al., 2005). In zebrafish, a retroviral insertion mutation in the complex B gene IFT81 did not affect photoreceptor development (Gross et al., 2005) but did cause kidney cysts (Sun et al., 2004), whereas morpholino knockdown of a complex A protein, IFT140, did not affect photoreceptor survival (Tsujikawa and Malicki, 2004). Our results provide the first in vivo comparative study of photoreceptor structure and function in two null IFT mutants, and demonstrate that mutations in IFT57 and IFT88 affect different aspects of photoreceptor development.

Our data also indicate that ciliary transport and outer segment morphogenesis requires IFT88, whereas IFT57 is required for effective transport and outer segment maintenance. Transport is also eliminated in IFT88 mutants of Chlamydomonas, mice and worms (Haycraft et al., 2001; Pazour et al., 2000; Qin et al., 2001). By contrast, the presence of connecting cilia and opsin labeling of outer segment structures indicate ciliary transport does occur in IFT57 mutants. It is important to note, however, that transport remains compromised in IFT57 mutants, as indicated by opsin mislocalization to the inner segment, vesicle accumulation at the base of the cilium and an overall reduction in outer segment length.

A previous analysis of a mutation in CHE-13, the C. elegans ortholog of IFT57, noted that the CHE-13 phenotype was identical to that of a previously characterized IFT88 mutation (Haycraft et al., 2003). Our analysis, however, indicates that mutation of these two genes results in different outcomes. This may be due to inherent differences in the mechanism of IFT between these two systems. In addition, the authors did not exclude the possibility of formation of these structures and subsequent degeneration, which is what we observed in our study.

Although transport is compromised, disk stacking in IFT57 mutant outer segments was unaffected. Opsin is believed to play an important role in the formation and stabilization of disk membranes within the outer segment (Nathans, 1992). Studies of rhodopsin heterozygous knockout mice found that reducing opsin by ∼50% causes disorganization of outer segments and disk stacking, with some abnormally oriented membranes and a complete loss of outer segments in homozygous rhodopsin knockout mice (Humphries et al., 1997; Lem et al., 1999). Rhodopsin levels were reduced by almost 60% in IFT57 mutant photoreceptors, yet no such disorganization of the outer segment was observed. One explanation is that opsin mislocalization kills the photoreceptor before outer segment disorganization occurs. Alternatively, zebrafish outer segments may require less opsin for disk membrane organization and outer segment integrity than do mammalian systems.

Our phenotypic and biochemical analyses of IFT57 support a model for protein interactions within the IFT particle that facilitate ciliary transport and anterograde motor dissociation. We show that IFT20 does not bind the IFT particle in the absence of IFT57, indicating that IFT57 mediates the interaction between IFT20 and the particle. Additionally, kinesin II co-precipitates with the IFT particle in IFT57 mutant lysates. These results indicate that the previously described IFT20-kinesin II interaction (Baker et al., 2003) is not the primary link between kinesin II with the IFT particle. We suggest that kinesin II binds to the IFT particle through an interaction with an unidentified IFT protein. Finally, we have shown that the ATP-dependent dissociation of kinesin II from the IFT particle is inhibited by loss of IFT57. Consistent with this, we also observe a greater quantity of KIF3A in IFT57 mutant precipitates than in wild type precipitates. It is important to note, however, that kinesin II has the ability to dissociate from the IFT particle in IFT57 mutants, as addition of super-physiological levels of ATP (10 mM) can cause dissociation of kinesin II.

We propose that IFT57 mediates the interaction between IFT20 and the IFT particle, whereas an undefined IFT protein or adaptor protein serves as the primary bridge between kinesin II and the IFT particle (Fig. 7). Our data indicates this bridge must be independent of both IFT57 and IFT20. Additionally, we propose that IFT57 and/or IFT20 mediates dissociation of heterotrimeric kinesin II from the IFT particle, potentially by enhancing the ATPase activity of KIF3B. As IFT20 is the only IFT protein known to interact with kinesin II, loss of IFT20 from the IFT particle is the likely explanation for the observed kinesin II dissociation defect. However, we cannot rule out a direct role for IFT57 in this process, though yeast 2-hybrid analysis showed no interaction between IFT57 and any of the kinesin II subunits (Baker et al., 2003). Others have shown that the ATP-mediated dissociation of kinesin II from the IFT particle is independent from the ATP-dependent dissociation of kinesin II from microtubules, illustrating that our data reflect a novel mechanism for motor dissociation and are not a result of impaired microtubule association. According to our model, neither the assembly of the IFT particle nor the recruitment of kinesin II requires IFT57.

Knockdown of IFT20 has been shown to abolish cilia in cultured mammalian cells, indicating that IFT20 is essential for IFT. Our data indicate that IFT20 is not essential for IFT but does play a vital role in effective transport. The more severe phenotype described in cultured cells may reflect additional roles of IFT20 in post-Golgi to cilium trafficking of proteins. Consistent with this, Follit et al. (Follit et al., 2006) observe a reduction in the amount of polycystin 2 that is trafficked to the cilium when IFT20 is partially knocked down in cultured mammalian cells. As we show that loss of IFT57 abolishes the interaction between IFT20 and the IFT particle, we are able to separate the Golgi and ciliary functions of IFT20 and demonstrate its role within the IFT particle and that IFT20 is necessary only for efficient IFT.

In conclusion, we have provided novel details about IFT function in vertebrate photoreceptors that probably have broader implications for IFT function in a wide variety of tissues. First, we have demonstrated that IFT57 is not absolutely required for ciliogenesis and transport in photoreceptors but is required for ciliary elongation (e.g. outer segment growth) and the long-term maintenance of photoreceptor survival. Combined with previous results demonstrating that loss of IFT81 or IFT140 does not affect photoreceptor development, it is likely that most, but not all, IFT proteins are required for ciliogenesis in all vertebrate tissues, and the composition of the IFT particle varies from tissue to tissue. Second, we have provided evidence that presents novel insights into how IFT20 associates with the IFT particle, and the mechanism by which kinesin II dissociates from the IFT particle. These data illustrate how proper dissociation of kinesin II is necessary for efficient IFT and show the functional nature of two previously described interactions within the IFT particle. Additional biochemical studies are necessary to elucidate which IFT subunit(s) mediate the interaction of the particle with kinesin II, the mechanism by which the IFT particle assembles and the mechanisms that mediate kinesin II dissociation from the IFT particle. Finally, the structure and function of cilia varies with different tissues and the composition of the IFT particle may reflect this diversity. Identification of IFT genes required for photoreceptor development may help predict whether mutations in specific IFT genes will lead to retinal disease.

The oval allele is an ENU-induced point mutation that caused a T-A transition, resulting in a premature stop codon in exon 11 and a null mutation in the ift88 gene. The IFT88 mutant line was a gift from Jarema Malicki (Doerre and Malicki, 2002) and the ift57 mutation has been previously described (Gross et al., 2005). Zebrafish were maintained according to standard procedures (Westerfield, 1995).

Morpholino knockdown of IFT57 was carried out using morpholinos directed against the splice site, sequence 5′-GTTATCGCCTCACCAGGGTTCGAAG-3′. A morpholino designed against the 5′UTR of IFT88, sequence 5′-TTATTAAACAGAAATACTCCCA-3′ was used to knock down IFT88 gene expression. Both morpholinos used have been described previously (Tsujikawa and Malicki, 2004) and were synthesized by Gene Tools.

An IFT20-GFP fusion construct driven by the zebrafish opsin promoter was obtained from Dr Joe Besharse (Medical College of Wisconsin). Transgenic photoreceptors were generated as described (Perkins et al., 2002) and analyzed by immunohistochemistry as described below.

Embryos were processed for light microscopy and TEM histology as previously described (Schmitt and Dowling, 1999). Histological sections were stained with a solution of 1% Azure Blue, 1% Methylene Blue and 1% sodium borate (Electron Microscopy Sciences). For immunogold labeling, staged embryos were fixed in 4% paraformaldehyde, 0.5% glutaraldehyde and 1% tannic acid in 0.1 M sodium cacodylate buffer (pH 7.4) overnight at 4°C. Embryos were washed in 0.15 M sodium cacodylate buffer at 4°C and dehydrated in a graded ethanol series with 1% p-phenylenediamine. Specimens were embedded in LR-White resin (Electron Microscopy Sciences) and polymerized overnight at 55°C. For immunolocalization, incubations were carried out in a PELCO Biowave (Ted Pella) at low wattage, at 30°C. Grids were incubated in blocking solution (PBST + 4% cold water fish gelatin), then in primary antibody (1D1 1:50) diluted in blocking solution. Grids were washed with PBS then incubated in secondary antibody, donkey anti-mouse 18 nm colloidal gold (Jackson Immuno Research 1:30) diluted in TBST (TBST+2% normal donkey serum +4% coldwater fish gelatin). Grids were washed with TBS and then fixed with 1% glutaraldehyde and post stained with 2% uranyl acetate. Photographs were obtained on a JEOL 1200EX transmission electron microscope and images processed using Adobe Photoshop.

Rabbit polyclonal antibodies were generated against synthetic peptides corresponding to the C terminus of IFT88, IFT57, IFT52 and IFT20. The antigens were LEFADGELGDDLLPE, CMHATHLLEPNAQAY, CKKLNEEHDVDTAEARFSMY and CEAEQSEFIDQFILQK, respectively. The IFT88, IFT57 and IFT20 antibodies were generated and affinity purified by Bethyl Labs, while the IFT52 antibody was generated and affinity purified by Open Biosystems.

Immunohistochemistry was performed as previously described (Perkins et al., 2005). Images were obtained on an Olympus FV1000 confocal microscope or a Zeiss ImagerZ1 fluorescence microscope fitted with an ApoTome. Images were prepared using Adobe Photoshop software. The following list of primary antibodies were used, followed by the dilution used: 1D1 (1:200) (Fadool, 1999), cone opsin antibodies (1:200) (Vihtelic et al., 1999), ZPR1 (Zebrafish International Resource Center 1:200), anti-acetylated tubulin (Sigma 1:500), anti-syntaxin 6 (BD Bioscience 1:200) and anti-IFT88 (1:5000). The appropriate fluorescently conjugated antibodies (Jackson Immunological) were used at 1:500 dilution. Slides were counterstained with DAPI (Invitrogen) to label DNA.

Wild type and IFT57 mutants (4 dpf) were collected and homogenized using 200-250 larvae per reaction in IP lysis buffer (PBS + 1% Triton + 5 mM EDTA) supplemented with a cocktail of protease inhibitors (Complete-mini EDTA Free, Roche). Samples were lightly sonicated and lysates subsequently clarified by centrifugation at 16,100 g at 4°C for 15 minutes. Lysates were then precleared with ExactaCruz F preclearing matrix (Santa Cruz Biotechnology) according to manufacturer's instructions. Polyclonal rabbit anti-IFT88 antibody was bound to ExactaCruz F IP matrix per instructions. An equivalent amount of normal rabbit IGG (Santa Cruz Biologicals) was bound to ExactaCruz F IP matrix as a negative control. Lysates were then pooled and normalized for total protein concentration before incubation overnight at 4°C with antibody-IP matrix on a rocking platform. The IP matrix was pelleted by low speed centrifugation and washed twice with cold PBS and then five times with cold IP lysis buffer. Samples were eluted from IP matrix by boiling 5 minutes in 45 μl of 2× Laemmli buffer and then 20 μl loaded on an SDS-PAGE gel for analysis. For ATP experiments, prior to addition of antibody-IP matrix to the lysates, ATP (Roche) was added to achieve either 1 or 10 mM final concentration. The lysates were incubated with ATP at 4°C for 1 hour prior to the addition of antibody-IP matrix to the lysates.

Heads of 4 dpf zebrafish were homogenized in IP lysis buffer at a volume of 3 μl per head, then lightly sonicated. SDS sample buffer (4×) was added to the lysate and the samples were then boiled for 5 minutes. prior to centrifugation at 16,100 g for 10 minutes and subsequent loading on a 10% PAGE gel. Following electrophoresis, proteins were blotted onto PVDF membranes (BioRad) and subjected to immunodetection using standard protocols. The following dilutions of primary antibodies were used: K2.4 (mouse anti-KIF3A, Covance 1:10,000), rabbit anti-IFT88 (1:5000), rabbit anti-IFT52 (1:5000), rabbit anti-IFT20 (1:5000), rabbit anti IFT57 (1:5000) and mouse anti-acetylated tubulin (Sigma, 1:10,000).

The authors thank Jarema Malicki for providing the oval^tz288b (IFT88) mutant line, Joseph Besharse for providing the xops:IFT20-GFP plasmid, and Tom Vihtelic for providing the cone opsin antibodies. The authors are indebted to Ann Ellis in the Texas A&M Microscopy Imaging Center for assistance with transmission electron microscopy. This work was supported by the NIH grant RO1EY017037 to B.D.P.