Analysis of xbx genes in C. elegans

Sensory behavior in higher animals depends on the correct recognition and processing of signals from the environment or from within the organism, and then on the appropriate reaction to those signals. The first step in this chain of events, signal recognition, is mediated by specialized cells of the nervous system, the sensory neurons. They create functional compartments for the localization and exposure of the signal reception and transduction machineries. There is considerable diversity in these compartments. Prominent examples are ciliated endings at the tip of dendrites of sensory neurons. Specialized mammalian sensory cilia include those in the photoreceptor cells in the eye, in the hair cells in the ear, and in olfactory neurons in the nose. The universal architecture of cilia and their close relatives, flagella,consists of a microtubular axonemal core enclosed by a membrane, exposed on the cell surface. A convergence of results from several systems has led to detailed structural descriptions of cilia and flagella(Perkins et al., 1986; Dutcher, 1995; Rosenbaum and Witman,2002).

In the nematode C. elegans 60 of the 302 neurons of the hermaphrodite are ciliated sensory neurons (CSN)(Ward et al., 1975; White et al., 1986), forming many structurally distinct types of sensory cilia. Whereas many ciliary mutants are available in C. elegans, there is only one known gene mutation that completely eliminates all classes of sensory cilia and all functional components of cilium structure. This gene is daf-19, and it encodes the sole C. elegans member of the RFX-type transcription factors (Swoboda et al.,2000), found widely in the eukaryotic kingdom. All members of the RFX transcription factor family are characterized by the presence of a conserved DNA binding domain (DBD). The RFX-DBD binds to special motifs(X-boxes) in promoters of its target genes. Genes containing the X-box promoter motif are called xbx genes.

Apart from C. elegans DAF-19, the RFX family currently contains nine characterized members: five in mice and humans (RFX1-5)(Emery et al., 1996a), two in Drosophila melanogaster (dRFX1-2)(Dubruille et al., 2002; Otsuki et al., 2004), and one member each from Schizosaccharomyces pombe(Wu and McLeod, 1995) and Saccharomyces cerevisiae (Huang et al., 1998). The data obtained so far suggest diverse biological roles of RFX proteins. In yeasts they regulate some aspects of the cell cycle(Wu and McLeod, 1995; Huang et al., 1998). In humans RFX factors are involved in the transcriptional regulation of major histocompatibility complex class II genes (RFX5)(Reith and Mach, 2001) and in the modulation of Ras signaling in epithelial cells (RFX3)(Maijgren et al., 2004).

The finding of X-boxes in promoters of ciliary genes in C. eleganshas revealed an important role of the RFX family in the regulation of ciliogenesis (Swoboda et al.,2000). Since then, the conservation of RFX-binding elements has been reported in several distantly related species. For example, some ciliary genes in D. melanogaster contain X-box-like sequences in their promoters (Avidor-Reiss et al.,2004). However, an experimental demonstration of an RFX-dependence for Drosophila ciliary gene candidates exists so far only for the nompB gene (Han et al.,2003). Recently, data about the possible RFX regulation of ciliogenesis in mammals were also obtained. Rfx3-deficient mice exhibit frequent left-right asymmetry defects, which are caused by ciliary abnormalities in mutant embryos (Bonnafe et al., 2004). Mouse RFX3 regulates the expression of D2lic,the mouse ortholog of the C. elegans ciliary gene xbx-1, but does not affect the expression of Tg737, the mouse ortholog of the C. elegans ciliary gene osm-5. These observations suggest that RFX regulation of ciliogenesis in higher organisms is more complicated,and different subtypes of RFX proteins may be restricted to particular components of ciliary structure.

In our current work we first concentrated on the isolation of genes important for cilium structure and function using a genome-wide X-box promoter motif search in the nematode species C. elegans and C. briggsae. In this computational approach we focused our efforts only on 5′ flanking regions of genes, since X-box motifs have previously been shown to be functional in those regions(Swoboda et al., 2000). Subsequently, we performed expression analyses of the group of positive C. elegans matches in wild-type and daf-19 mutant backgrounds,together with X-box mutagenesis experiments. Results of these analyses established the X-box consensus for C. elegans, the approximate number of xbx genes in the C. elegans genome and assigned already known and newly found ciliary genes to specific structural and functional groups. Because the organization of C. elegans sensory cilia is very similar to sensory cilia in mammals, the results obtained with the C. elegans model will have general significance.

Growth and culture of C. elegans strains were carried out following standard procedures (Brenner,1974). The following strains were used for this study: wild type N2 Bristol; JT8651 daf-19(m86)/mnC1; lin-15(n765ts); JT6924 daf-19(m86); daf-12(sa204); JT204 daf-12(sa204); RB773 nud-1(ok552); RB819 xbx-4(ok635); RB857 xbx-6(ok852); NL2099 rrf-3(pk1426); CB3323 che-13(e1805); CB1033 che-2(e1033). Extrachromosomal arrays were used for all GFP expression analyses. All strains used and strain construction details are available on request.

The X-box motif search was performed primarily with a Perl-based algorithm that searches through a given genome sequence for all possible matches. The algorithm first finds all sequences that match a defined consensus. After that step, the main module of the program implements a cross-match file (P. Green,personal communication), which compares a 3 kb window downstream of each match to a file containing the DNA sequences for all predicted genes, and a file containing assembled ESTs. Cross-match parameters – `minmatch' and`minscore' were set to 40. All other parameters were kept at default values. Minimal and maximal distances from positive matches to predicted genes were set to a range of 0-1000 nucleotides. To obtain a copy of the algorithm,contact kbubb@u.washington.edu. Cross-match must be obtained separately (see www.phrap.comfor access/download information).

Subsequent to genome analyses using Cross-match, we made use of another program, DNA Motif Searcher, which takes a set of user-definable X-box sequences to search for additional motif instances in the genome. The set of motifs is interpreted into a position-specific score matrix (PSSM). Using this PSSM, the program can then identify the closest matching occurrences of the motif based on a score cutoff, or it can identify the top number of occurrences of the motif. For download or for more detailed explanations about Motif Searcher, please contact jht@u.washington.edu.

Genome sequence information, EST files, gene predictions and identities for X-box searches were obtained from the following sources: C. eleganscomplete genome sequence, WS122 release(ftp://ftp.sanger.ac.uk/pub/wormbase/WS122); C. briggsae draft genome sequence, cb25.agp8 version(ftp://ftp.wormbase.org/pub/wormbase/briggsae).

GFP expression constructs were designed by inserting about 2 kb of promoter regions and the first several codons of a gene of interest into the GFP expression vector pPD95.77 (gift from A. Fire). PCR fragments of promoter regions were obtained from wild-type N2 genomic DNA and cloned into appropriate sites of pPD95.77. For some genes, the wild-type X-boxes within promoters were mutated by overlap extension mutagenesis, replacing X-box sequences with nonspecific nucleotides containing indicative restriction enzyme sites. To check for correct translational reading frames and promoter regions, junctions between vector and amplified inserts were verified by sequencing for all constructs. For the XBX-2::GFP translational fusion, the entire coding sequence of the gene with about 1 kb of promoter were fused to the pPD95.77 vector.

The following worm strains were used for injections and expression analyses: JT8651, JT6924 and JT204. The strain JT8651 daf-19(m86)/mnC1;lin-15(n765ts) served as the wild-type background, since daf-19(m86) is fully recessive. daf-19 mutants are strongly Daf-c (dauer larva formation – constitutive) across the normal temperature range. Therefore, segregating dauers were recovered at 15°C to obtain a daf-19 homozygous background. Alternatively, the strain JT6924 daf-19(m86); daf-12(sa204) was used as a daf-19mutant background. Worms of this genotype exhibit a Daf-d (dauer larva formation – defective) phenotype and do not require the recovery of dauers. In this case, JT204 daf-12(sa204) worms were used as a wild-type background with regard to daf-19.

Adult hermaphrodites were transformed using standard protocols(Mello et al., 1991). Constructs were injected typically at 10-100 ng/μl along with coinjection markers such as pRF4 (contains the dominant marker rol-6(su1006)) or pBLH98 (contains the wild-type lin-15 gene to rescue lin-15(n765ts)).

GFP expression patterns were analyzed in stable transgenic lines at 1000× magnification by conventional fluorescence microscopy (Zeiss Axioplan 2). Expression patterns were examined in at least two independent transgenic lines at most developmental stages of the worm. Neuronal cell anatomies and identities followed published descriptions(Ward et al., 1975; White et al., 1986).

For the analysis of XBX-2::GFP movement properties, worms were mounted on agarose pads and anesthetized with 10 mM levamisol. Adult worms were analyzed with a Leica confocal imaging spectrophotometer TCS SP unit mounted on a Leica DMIRBE inverted microscope, and the obtained images were processed using Leica Confocal Software 2.5. Images were taken with a 63× objective and a 488 nm GFP filter. At least 40 stacked images were converted into an AVI file with a rate of two frames per second.

RNA-mediated interference (RNAi) was performed according to standard methods (Timmons et al.,2001). PCR fragments for genes of interest were generated from N2 genomic DNA and cloned into the double T7 promoter-containing vector L4440(gift from A. Fire). All constructs were transformed into HT115(DE3) bacterial cells and plated onto NGM plates with antibiotics and IPTG. L4-stage hermaphrodites were transferred to plates with induced bacteria and F₃ progenies were analyzed for possible phenotypes.

Fluorescent dye-filling assays were performed essentially as described previously (Starich et al.,1995) using the fluorescent dye DiI. Worm strains N2 and CB3323 were used as positive and negative controls, respectively. Stained adult hermaphrodites were analyzed at 1000× magnification by conventional fluorescence microscopy (Zeiss Axioplan 2).

All deletion alleles analyzed in this study were generated by the C. elegans Gene Knockout Consortium(http://celeganskoconsortium.omrf.org/)using publicly available methodology(http://www.mutantfactory.ouhsc.edu/protocols.asp).

The original mutated strains RB819 xbx-4(ok635) IV and RB957 xbx-6(ok852) V were outcrossed three times with N2 and the following worm strains: JT7146 egl-4(n478) unc-33(e204) IV and DR108 dpy-11(e224) unc-42(e270) V, respectively. Outcrossed worms resulted in homozygous mutant strains containing the xbx-4(ok635) IV and xbx-6(ok852) V deletion alleles. These strains were then used as the basis for further analysis.

The discovery of X-boxes in promoter regions of certain ciliary genes (e.g. Swoboda et al., 2000; Haycraft et al., 2001; Haycraft et al., 2003; Fan et al., 2004) prompted us to analyze the whole C. elegans genome for the presence of these motifs. In order to perform this analysis of all C. elegans promoters we implemented an in-house searching algorithm – the X-box searcher,which searches for all possible matches to a defined motif sequence(Fig. 1A). For the initial search we used a `relaxed' consensus (RYYNYY WW RRNRAC), that fits published mammalian X-box sequences (Emery et al.,1996b) and the first emerging C. elegans X-boxes from previously known ciliary genes (Swoboda et al., 2000). Using the X-box searcher with the relaxed consensus generates the highest number of output matches (1927) that were equally spread within 1000 bp of promoter regions (Fig. 1B,C). Through ongoing work with xbx gene candidates (see below) we obtained a `refined' X-box consensus (GTHNYY AT RRNAAC), which corresponds to most of the experimentally proven xbx genes. We further re-analyzed the C. elegans genome using the refined consensus and found a significant reduction in the number of matches (128). Unlike the relaxed consensus, most of the refined matches show clustering around the region of 100 bp upstream of the ATG. This clustering was not observed when the X-box search sequence was scrambled(Fig. 1B,D). The refined xbx gene list includes the majority of experimentally proven xbx genes, but not all of them. At the same time, the list of genes obtained with the relaxed X-box consensus contains too much `search noise'. According to our expression analysis (see below) we estimate that up to 90% of the relaxed matches can be found in promoter regions of non-ciliary genes or genes that are not specifically expressed in ciliated sensory neurons (CSN)(see Table S2 in supplementary material). Therefore, we additionally tried several different X-box sequences and found an `average' consensus (RTHNYY WT RRNRAC). This consensus permitting ambiguities at three positions as compared to the refined consensus (A or G at position 1, A or T at position 7 and A or G at position 12) makes it possible to find all known X-boxes and reduces the search noise by about 2.5 times (758 matches)(Fig. 1B; see Table S1 in supplementary material). This consensus can be used for further analyses in C. elegans, as well as for X-box search efforts in other organisms.

The functional repertoire of the average candidate xbx gene list contains different molecular groups, including possible components of the ciliary structure and transport machinery (6%), transcription factors (5%),receptors (11%) and ion channels (1%). About 37% of the genes have no identified function and the rest of the list (40%) is composed of genes of various molecular identities (Fig. 2; see Table S1 in supplementary material).

Our primary goal was to find new xbx and ciliary genes with high efficiency. Our X-box search might not be exhaustive, because the search algorithm is based on an originally small set of experimentally proven X-box sequences. To determine whether additional xbx gene candidates could be found in the C. elegans genome, we tried a different search approach, which is position-specific score matrix (PSSM) based (J.H.T.,unpublished). Using the output data from the average list as a training set for the PSSM searcher we achieved almost the same list of xbx genes(data not shown), strengthening our overall strategy. However, we additionally obtained several prominent candidates that had a slightly different X-box motif than the average consensus and therefore were not found with the X-box searcher: for example dlc-1 (GTTATT AT AACTAC, which encodes a dynein light chain), C01B12.4 (GTTTCC AT AGCTAC, which encodes a predicted seven transmembrane receptor of the rhodopsin family).

Expression patterns of orthologous genes are often conserved. Because many orthologous transcription factors are also functionally conserved, one possible model to account for homologous gene expression patterns is conservation of specific binding sites within regulatory elements of orthologous genes (Ruvinsky et al., 2003). The nematodes C. briggsaeand C. elegans are closely related species with very similar overall genome organizations (Stein et al.,2003). To find possible conservations of X-box regulatory elements between those two organisms we applied the C. elegans X-box search strategy to the C. briggsae genome. The number of obtained matches was slightly less than with C. elegans, probably because of the draft quality of the C. briggsae genome. Nevertheless, the profile of X-box distribution within promoter regions was similar to that of C. elegans (Fig. 1D).

In summary, using two different X-box promoter motif search approaches,X-box consensus sequences with varying degrees of refinement, together with cross-species comparisons and gene expression analysis (see below), we were able to identify a large number of bona fide xbx genes, a significant part of which we expect to also be ciliary- or CSN-specific genes.

Our computational search has revealed a large, heterogeneous group of X-box matches (see Table S1 in supplementary material). To determine the specificity of our computational search with regard to ciliary and CSN structure and function we initiated expression analyses for some of the genes. For this purpose we isolated a group of candidates with different nucleotide compositions of the X-box motifs, different positions within promoter regions and different proposed molecular functions of matching genes.

Our previous model associated the expression of xbx genes with CSN(Swoboda et al., 2000). We predicted that DAF-19 function, a particular X-box composition and its position within the promoter should be required for the expression of this group of genes. Therefore, expression patterns of selected genes were analyzed in both wild-type and daf-19 mutant backgrounds. For some genes we also analyzed the actual X-box sequence by replacing it with nonspecific nucleotides within the respective expression construct.

Based on our obtained expression data(Table 1), we subdivide xbx genes into groups: (1) genes that are strongly regulated by DAF-19 and required for the general development of cilia; (2) genes that are partially regulated by DAF-19 and are probably required for more specific ciliary and/or CSN functions.

Genes of the first group are characterized by expression in many, most or all CSN. All of them are DAF-19 dependent and for some genes tested (through mutagenesis) we also demonstrated a dependence on the actual X-box sequence(Table 1).

Most genes from this group encode known participants of intraflagellar transport (IFT): osm-1, osm-5, osm-6, che-2, che-13, xbx-1. The mechanism of IFT was originally described in the biflagellate alga Chlamydomonas reinhardtii(Kozminski et al., 1993). It is characterized by the movement of IFT particles along ciliary/flagellar axonemal microtubules by means of kinesin and dynein motor molecules. C. elegans xbx genes encoding IFT proteins are well described. IFT components such as OSM-1, OSM-5, OSM-6, and CHE-13 are associated with the heterotrimeric motor protein kinesin-II. They are essential for anterograde transport (Signor et al.,1999; Haycraft et al.,2001; Haycraft et al.,2003). The gene xbx-1 encodes a dynein light intermediate chain (DLIC) that is important for retrograde transport within cilia(Schafer et al., 2003).

Herein we report a new IFT gene, xbx-2 (D1009.5). The xbx-2::gfp promoter fusion was strongly expressed in most of the ciliated sensory organs of the worm – amphids, phasmids, inner and outer labial quadrants (Fig. 3A). The XBX-2 protein contains a Tctex-1 domain that belongs to the family of dynein light chain (DLC) proteins. These molecules are essential for dynein assembly and participate in specific motor-cargo interactions(DiBella et al., 2001; Tai et al., 2001). To analyze the possible role of XBX-2 in the IFT process, we generated transgenic worms expressing XBX-2::GFP protein. Using time-lapse confocal microscopy, we observed movement of XBX-2::GFP particles along the ciliary axoneme in both retrograde (Fig. 4) and anterograde directions (see also Movie 1 in supplementary material).

A major part of the first group was also composed of C. elegansorthologs of human Bardet–Biedl syndrome genes: bbs-1, bbs-2, bbs-5,bbs-7 and bbs-8. Recent data suggest that bbs genes are probably involved in the development of the basal body during cilium formation(Ansley et al., 2003; Li et al., 2004) and are required for assembly and proper function of some IFT components(Blacque et al., 2004; Fan et al., 2004). We expressed bbs gene members both in wild-type and in daf-19mutant backgrounds (in parallel with X-box mutagenesis experiments) and found that this group of genes strongly requires both DAF-19 function and proper X-box composition (Table 1). These results confirm that genes encoding general cilium structure molecules are strongly regulated by DAF-19.

The second group of xbx genes includes many novel genes, which are expressed in various subsets of CSN. For example, xbx-3::gfp was strongly expressed in all amphid and phasmid neurons, but not in other ciliated sensilla (Fig. 3B). In daf-19 mutant worms, expression was restricted to one amphid neuron and abolished in phasmids, with occasional ectopic expression in other tissues. Expression of the novel gene xbx-4 was also observed in some amphid and phasmid neurons (Fig. 3C), but in daf-19 mutants it was completely abolished in both organs. The xbx-5::gfp promoter fusion was characterized by faint, punctate expression in phasmids and some amphid neurons(Fig. 3D). In a daf-19mutant background expression was absent in amphids but still visible in phasmids. The predicted XBX-5 protein contains seven transmembrane domains and can be considered as a possible receptor. The xbx-6::gfp construct was abundantly expressed in many cell types: pharyngeal muscles, numerous neurons in the head and tail regions, the ventral nerve cord and body wall muscles. Because of high overall expression levels, we were not able to identify individual CSN in the head region, but we observed expression in phasmids (Fig. 3E), which was strictly DAF-19 dependent. The expression in other cells was unchanged in daf-19 mutants. The xbx-6 gene encodes an N-methyl-D-aspartate receptor-associated protein. The novel gene xbx-7 was expressed in phasmids, some amphid neurons and in interlabial neurons (Fig. 3F),while in daf-19 mutants expression was significantly reduced. The gene che-11 encodes a large protein, which is orthologous to Chlamydomonas IFT140 (Qin et al.,2001). The expression level of che-11::gfp was significantly reduced in phasmids and amphids in daf-19 mutants(Table 2), while in the labial neurons it was almost unchanged. The nuclear hormone receptor gene nhr-44 was expressed in head neurons, including the ciliated sensory neuron ASK, and also in other cell types, and its overall expression properties fit those of other xbx genes(Table 1).

Within the second group of xbx genes we especially note the genes odr-4, nud-1 and tub-1, the molecular identities of which were previously described.

The gene odr-4 has been shown to be important for the localization of some seven transmembrane domain odorant receptors to cilia(Dwyer et al., 1998). The expression of an ODR-4::GFP translational fusion was significantly reduced both in daf-19 mutants and after mutation of the X-box sequence (data not shown). This indicates that not only transport mechanisms within cilia(IFT), but also to cilia (ODR-4) are under DAF-19 control.

The C. elegans ortholog of the NudC gene of the fungus Aspergillus nidulans, nud-1, was identified as a candidate xbx gene during our search. NUD-1 is an important component in microtubule-dependent nuclear positioning, which is required for proper growth, development and cellular function in both lower and higher eukaryotes. Sustained expression of nud-1::gfp in CSN was previously described(Dawe et al., 2001). We introduced this construct into a daf-19 mutant background and found its expression drastically reduced (Table 2).

The gene tub-1 is the worm ortholog of murine tubbywhich, when mutated, leads to neuronal deficits and late-onset obesity(Carroll et al., 2004). tub-1 mutant worms exhibit functional defects in CSN and show a mild elevation of lipid accumulation (H.Y.M., unpublished data). The translational GFP fusion for tub-1 was expressed only in the cytoplasm in all CSN(Fig. 1G). The expression level of this construct was significantly reduced in a daf-19 mutant background (Table 2). We analyzed the 5′-UTR region of the human TUB gene and found two X-box-like sequences that perfectly match the C. elegans consensus:GTTGCC AT GGAAAC (–296) and GTTGCT AT AGTAAC (–339). Intriguingly,microtubule-associated protein 1A (MAP1A) can modify hearing defects in tubby mice (Ikeda et al.,2002), and its expression is regulated by RFX molecules(Nakayama et al., 2003). These observations suggest that the regulation of tubby pathways by RFX transcription factors can be conserved in evolution. Based on the data from the second group of xbx genes, we conclude that DAF-19 only partially regulates the expression of certain genes. These genes may only be required for specialized functions in CSN or during ciliogenesis.

The remaining genes analyzed were mostly expressed in many other different cell types and only in very few cases was expression observed predominantly in CSN (Table 1; see Table S2 in supplementary material). We checked some genes from this group (zag-1,aqp-2, gpa-9, F55D12.1, F17A2.3) in a daf-19 mutant background and found that expression patterns were unchanged in the absence of DAF-19 function. Most of these X-box matches differ from the refined consensus or are located further upstream of the ATG (Table 1).

An additional confirmation and filtering mechanism used was a cross-species comparison, where candidates were classified as likely or unlikely xbx genes depending on the conservation of the X-box motif sequence between C. elegans and C. briggsae(Table 1). We found that most of the genes from the first group have putative C. briggsae orthologs with nearly identical X-box sequences and positions in their promoters. In the second group, the frequency of X-box occurrences in C. briggsaeorthologs was reduced. C. elegans xbx gene candidates that were expressed predominantly in cell types other than CSN typically have no X-box-like sequences in promoters of their C. briggsae orthologs(Table 1; see Table S2 in supplementary material). Thus, the actual X-box sequence and position conservation between C. elegans and C. briggsae (and maybe other organisms) can be used as an additional measure to deduce possible molecular roles of xbx genes in cilia or CSN.

The expression data from 22 GFP fusions that show dependence on DAF-19 function and actual X-box sequence (Table 1) allowed us to derive an in vivo refined consensus for C. elegans xbx genes (Fig. 5A). This consensus can be characterized as a 14 nt imperfect palindromic sequence (GTHNYY AT RRNAAC) consisting of conserved nucleotides at 5′ (GT) and 3′ (AC) ends of the two half-sites separated by a conserved AT spacer. Most of the X-boxes matching this consensus are located in immediate proximity to the gene start (in the range of 50-250 bp upstream of the ATG). Lack of DAF-19 function or changes in the actual X-box sequence lead to drastic reduction or variation of gene expression patterns, suggesting a crucial role of the given motif sequence in the regulation of target genes(see also Discussion).

Numerous mutations have been generated in C. elegans that affect cilia and their formation (Perkins et al.,1986; Starich et al.,1995). They cause various mutant phenotypes, including Dyf(fluorescent dye filling defective), Osm (avoidance of high osmotic strength defective), Mat (male mating defective), Che and Odr (chemo- and odorant sensation defective).

xbx genes from group 1 show Dyf and to various extents also different sensory phenotypes (Table 1) (Collet et al.,1998; Signor et al.,1999; Fujiwara et al.,1999; Haycraft et al.,2001; Haycraft et al.,2003; Schafer et al.,2003; Blacque et al.,2004; Li et al.,2004). Mutations that reduce dye filling of amphid and phasmid neurons are indicative of general defects in cilium structure and are often accompanied by various sensory mutant phenotypes(Starich et al., 1995).

The availability of a large group of xbx gene candidates obtained in our computational search prompted us to try a screening approach for new members of general cilium structure and function using genetic interference mediated by double stranded RNA (RNAi). We analyzed already known dyfgenes (che-13, osm-5) together with novel candidates (bbs-2,bbs-7, xbx-1, xbx-2) with regard to the Dyf phenotype. It was known that RNAi is less efficient in neuronal types of cells(Simmer et al., 2002). To increase possible effects of interference, in parallel to wild-type worms we also tested rrf-3 mutants, which are sensitive to RNAi in diverse tissues, especially in neurons (Simmer et al., 2002). We found that RNAi of the ciliary genes tested did not result in strong Dyf phenotypes in wild type or rrf-3 mutants (data not shown). Therefore, RNAi cannot be implemented as a quick and easy screening technique with regard to the Dyf phenotype in C. elegans.

Unlike those in group 1, the roles of many genes in group 2 are largely unknown. To begin the functional investigation of the second group we analyzed two genes, xbx-4 and xbx-6, mutants of which were available. The xbx-4(ok635) deletion extends over 951 bp starting in the promoter region and ending in the first intron, completely eliminating the beginning of the gene. The xbx-6(ok852) deletion extends over 1720 bp starting in the promoter region and covering five of the six exons of the gene. Since the expression of these genes is associated with CSN, we first focused our efforts on phenotypes related to defects in cilium structure or sensory abnormalities. The following results were obtained: both analyzed deletion alleles demonstrated wild-type responses with regard to fluorescent dye filling, high osmotic strength avoidance and in odorant sensation assays using three different odors (data not shown).

These xbx-4 and xbx-6 results suggest that in some instances members of the xbx gene family may have specialized molecular functions and therefore mutants have more specialized sets of sensory phenotypes, although we cannot formally exclude genetic redundancy with other (xbx) genes expressed in CSN. For example, two other members of group 2, the genes odr-4 and tub-1, when mutated,also do not produce general structural defects of cilia, but more specialized functional ciliary abnormalities, such as selective defects in odorant sensation (Dwyer et al., 1998)(H.Y.M., unpublished).

In conclusion, our data support the sorting of xbx genes into different groups, where members of group 1 are typically required for more general aspects of cilia formation, while genes from group 2 are typically required for more specialized functions within cilia and/or CSN.

To find groups of genes that belong to general and to specialized ciliary gene classes we used a promoter motif, the X-box, which participates in the regulation of certain ciliary genes in the nematode Caenorhabditis elegans. To make our search as efficient as possible we repeated,expanded and refined it with new sets of parameters (in a bootstrap-like fashion). As a result, we extracted from the C. elegans genome a list of 758 xbx gene candidates (Table S1 in supplementary material). We predict around 150 candidates from our list to be ciliary/ciliated sensory neuron (CSN)-specific genes, since they meet the following combination of criteria, found through experimental work:

• Out of more than 30 xbx genes tested for GFP expression patterns,about 60% of them, having an X-box motif fitting the refined consensus, were expressed in CSN, and most of them only in this group of neurons. When tested by mutational analyses expression in CSN was nearly always DAF-19 and X-box dependent.

• Most of these experimentally confirmed xbx genes have the X-box promoter motif fairly close upstream of the ATG (up to –250), a pattern found previously for the first few cilium-specific xbx genes analyzed(Swoboda et al., 2000). By using various randomly scrambled X-box sequences for searches we found that the concentration of hits closely upstream of the ATG disappeared and became uniform throughout the search space (Fig. 1D). We also analyzed available expression data for genes, where the X-box match was located further upstream of the ATG and found that these genes are typically not expressed in a ciliary- or CSN-specific manner (Table S2 in supplementary material). By using the X-box as an anchor it is possible that gene predictions for some of our candidates could be re-evaluated. Therefore, some X-box matches located further upstream can still be considered as functional promoter motifs (Table S1 in supplementary material).

• Finally, cross-species comparisons with the closely related nematode C. briggsae show strong conservation of the X-box sequence and position within promoters of many of the C. elegans experimentally confirmed xbx genes.

We note that our search is not completely unbiased, because the search algorithm is based on an originally small set of experimentally proven xbx genes. Therefore, as further xbx genes will be shown to be ciliary- or CSN-specific genes, their X-boxes will be included into the search parameters (bootstrap mechanism).

Other genome search approaches, in part utilizing X-box matches as a parameter, were used in different organisms to find general components of cilia formation (Li et al.,2004; Avidor-Reiss et al.,2004), yielding a set of conserved ciliary genes and gene candidates. We compared information from flagellar and basal body genes(Li et al., 2004) with our list of X-boxes and found an overlap of 15 X-box matches. At the same time, we found eight additional xbx gene candidates (Table S1 in supplementary material). Despite overlaps, many of the X-box matches were different between the respective searches, possibly because of different experimental parameters. Another important comparison was obtained from a recent study of olfactory neuron-specific genes (Colosimo et al., 2004), where we found 56 genes in common (Table S1 in supplementary material). All approaches together with further filtering mechanisms will give a complete, exhaustive list of genes important for structure and function of cilia and CSN.

cis-Regulatory elements are information processing devices hardwired into the genomic DNA sequence, the function of which is to regulate gene expression (Davidson,2001). Frequently, they are organized into modules that include many sites for DNA binding proteins(Howard and Davidson, 2004). In C. elegans, only some cell-type-defining transcription factors(CEH-10/TTX-3, MEC-3) target single binding sites in promoters of regulated genes (Zhang et al., 2002; Wenick and Hobert, 2004). Typically, also DAF-19 targets contain only a single X-box motif in their promoters. Nevertheless, different xbx genes show different expression properties, suggesting the presence of specific DAF-19 co-regulators. We propose that the information about particular gene expression profiles could already be stored at the level of the X-box sequence. For example, all genes for cilium structure and IFT from group 1 have perfect matches to the refined X-box motif consensus. Each gene of this group is expressed in most or all CSN and is strongly dependent on DAF-19 function. Whereas a perfect match to the `refined consensus' does not automatically predict expression in all CSN, variations in the X-box motif sequence, especially in the more variable half-site (GTHNYY), predict different gene expression properties (xbx-3, che-11, odr-4, tub-1). These genes are either expressed only in a subset of CSN (xbx-3,odr-4) or are only partially dependent on DAF-19 function (che-11,tub-1).

Structural experiments have shown that each half-site of a symmetric X-box interacts with both DBDs of the RFX homodimer(Gajiwala et al., 2000). Nevertheless, binding of RFX is not ultimately dependent on dimerization and monomers can bind to a single `high-affinity' half-site (RGYAAC)(Siegrist et al., 1993; Emery et al., 1996b).

We hypothesize that DAF-19 is a crucial transcription factor of genes required for general cilium formation. In this case, DAF-19 recognizes symmetric X-boxes as a homodimer and strongly activates their expression in most or all CSN (Fig. 5B). If a target gene contains a more asymmetric X-box sequence, it can be recognized by heterodimers of DAF-19 together with other, as yet unidentified factors,leading then to specific expression patterns in subsets of CSN(Fig. 5B). Hypothetically,X-box motif distances from the ATG could also contribute to target gene expression variability.

The development of sensory neurons in C. elegans is a complicated process that includes many stages and interactions of different transcription factors (Melkman and Sengupta,2004). The gene daf-19 acts at late stages of sensory neuron development when the respective cell fates have already been determined and specification and subsequent differentiation occurs. Several different types of genes are required to produce functional cilia in the worm: genes for their molecular structure, genes implicated in ciliary transport machineries and genes involved in signal reception and transduction(Jansen et al., 1999; Troemel, 1999; Rosenbaum and Witman, 2002; Melkman and Sengupta,2004).

Our previous model associated DAF-19 regulation with only a certain group of genes functioning in cilium morphogenesis and architecture(Swoboda et al., 2000). The data obtained in our current research suggest that the repertoire of DAF-19-dependent genes is much broader(Fig. 2). For the first time we have found that DAF-19 can also regulate genes of ciliary function. For example, expression of the gene odr-4 requires both a correct X-box sequence and the presence of DAF-19. The ODR-4 protein has been shown to be an important factor for localizing a subset of seven transmembrane domain odorant receptors to cilia (Dwyer et al.,1998). Moreover, we have shown that DAF-19 can directly regulate expression levels of some putative receptor proteins. For example, the genes xbx-5 and xbx-6 encode a seven transmembrane domain protein and an N-methyl-D-aspartate receptor-associated protein,respectively.

In addition to the group of genes with signal reception properties, we found X-boxes in promoters of proposed transcription factors(Fig. 2). Most of these factors contain a C2H2-type zinc-finger domain. The important role of this type of transcription factors for the development of cell-specific properties in CSN was already described for the gene che-1(Uchida et al., 2003). Thus,we predict the presence of cilium-specific developmental cascades directed by DAF-19-dependent transcription factors. These cascades may be required for specialized ciliary functions as well as being necessary for the possible parallel regulation of CSN specification and their final functional differentiation. For example, it has been demonstrated that DAF-19 can affect the expression of indirect targets in the HOB-specific pathway through some unknown factor(s) (the male-specific ciliated HOB neuron is necessary for sensation of the hermaphrodite vulva during mating)(Yu et al., 2003). Our own data also suggest that DAF-19 could be required not only for general cilia formation, but in some instances for the development of cell-specific properties as well. For example, we observed that the gene nhr-44 was expressed in a DAF-19-dependent manner in the ciliated sensory neuron ASK. This gene belongs to the nuclear hormone receptor family, which includes many ligand-regulated transcriptional modulators involved in many developmental processes (Miyabayashi et al.,1999).

Another example is the gene nud-1. We propose that certain microtubule-associated molecules (like NUD-1) acting during early developmental stages could later be recruited by DAF-19 for the purposes of cilia development. The important role of nud-1 in nuclear migration during embryogenesis in C. elegans was previously described(Dawe et al., 2001). It has also been shown that the mammalian ortholog of NUD-1, NudC, associates with the dynein motor complex during neuronal migration(Aumais et al., 2001). Therefore, we suggest a possible role for NUD-1 as a component of IFT during ciliogenesis. In addition to nud-1, we extracted two further X-box-containing genes from our list of candidates, spd-5(Hamill et al., 2002) and dlc-1, which are also involved in nuclear migration during early embryonic development and might later be recruited by DAF-19 for the development of CSN.

Considering the above we propose a model where DAF-19 regulates the development of a `ciliary module' during the differentiation of sensory neurons in C. elegans (Fig. 6). According to this model, DAF-19 is a key factor for the general development of cilia. At the same time, together with other factors,it can drive the expression of genes required for specialized functions in cilia.

In mammals cilia are near ubiquitous organelles that project from the surfaces of many different cell types to carry out motility and sensory functions (Afzelius, 2004). Cilia have been implicated directly in many developmental processes such as generation of left-right asymmetry, heart development, maintenance of the renal epithelium, respiratory function, and physiological balance of the cerebrospinal fluid. Defects in cilia function and structure lead to a wide range of developmental problems and diseases(Pazour and Rosenbaum, 2002; Afzelius, 2004), among them:Bardet–Biedl syndrome, polycystic kidney disease (PKD), X-linked retinitis pigmentosa 2, nephronophthisis, maturity-onset obesity, etc. Many C. elegans genes that are orthologous to genes of human cilia-dependent diseases are X-box-containing genes. Therefore, we believe that the understanding of ciliogenesis in C. elegans will have a significant impact on the understanding and treatment of cilia-based pathologies in humans.

The C. elegans gene names xbx-2 and dylt-2describe the same gene, D1009.5.

We thank the following people for technical assistance, helpful discussions, unpublished information, C. elegans strains, cosmid clones, cDNA clones, RNAi clones and GFP vectors: J. Ahringer, M. Barr, G. Caldwell, A. Coulson, A. Fire, C. Haycraft, Y. Kohara, G. Merrihew, E. Onischenko, S. Tat, H. Tian, B. Yoder and especially members of our and the T. Bürglin laboratories. We thank G. Gahmon and A. Laurençon for critical reading of the manuscript. We thank the C. elegans Genome Sequencing Consortium for providing genome sequence information and the Caenorhabditis Genetics Center, which is funded by the National Institutes of Health, for providing some of the C. elegans strains used in this study. We thank the C. elegans Knockout Consortium for providing gene deletion mutants. Work by H.Y.M. was supported by a postdoctoral fellowship from the Human Frontier Science Promotion Organization (Strasbourg, France). M.R.L. acknowledges scholar awards from the Canadian Institutes of Health Research (CIHR) and Michael Smith Foundation for Health Research (MSFHR), and funding provided by the Heart and Stroke Foundation of B. C. and Yukon and the Natural Sciences and Engineering Research Council of Canada (NSERC). This work was supported in part by Public Health Service Grants from NIH to J.H.T. and G.R., and by a grant to P.S. from the Swedish Foundation for Strategic Research (SSF) in Stockholm, Sweden.