SDF-9, a protein tyrosine phosphatase-like molecule, regulates the L3/dauer developmental decision through hormonal signaling in C. elegans

Environmental cues regulate various aspects of animal development. For example, sex determination in the atherinid fish is partly regulated by temperature (Conover and Kynard,1981), and anti-juvenile hormones produced by some plants cause precocious metamorphosis in certain insects(Bowers et al., 1976). Such phenomena have attracted biologists' attention, because they show how animals adapt to their environment and how their phenotypes are determined by the interaction of their genome and environment. However, the mechanisms of such regulation mostly remain to be studied.

The dauer larva of the nematode Caenorhabditis elegans serves as a good model system for these issues. Under favorable conditions, C. elegans develops successively through four larval stages (L1-L4) to the reproductive adult in 3 days. However, under unfavorable conditions, it forms an alternative third-stage larva called dauer larva and arrest development. The dauer larva can survive for periods of several weeks or months without aging, while the adult animal can live only for about 2 weeks. When favorable conditions are encountered, the dauer larva begins to feed and resumes development to the adult stage (Riddle and Albert, 1997). It is known that the decision to enter the dauer stage is regulated by three environmental cues: food, temperature and a dauer-inducing pheromone that reflects population density(Golden and Riddle, 1982; Golden and Riddle, 1984a; Golden and Riddle, 1984b). At least two of them, food and dauer pheromone, are thought to be sensed by chemosensory neurons with ciliated endings directly exposed to the external environment (Albert et al.,1981; Perkins et al.,1986; Bargmann and Horvitz,1991; Vowels and Thomas,1992; Thomas et al.,1993; Schackwitz et al.,1996). The sensory neurons that participate in dauer formation have been identified by laser microsurgery. When sensory neurons ADF, ASG, ASI and ASJ are killed at the L1 stage, wild-type animals become dauer larvae and arrest development regardless of environmental conditions(Bargmann and Horvitz,1991).

To identify genes involved in dauer signaling, many mutants that show abnormality in the dauer decision have been isolated(Riddle et al., 1981; Albert et al., 1981; Swanson and Riddle, 1981; Malone and Thomas, 1994; Inoue and Thomas, 2000). Such mutants, named daf (dauer formation abnormal), consist of two groups:dauer formation-constitutive (Daf-c) mutants, which enter the dauer stage even under conditions appropriate for reproductive growth, and dauer formation-defective (Daf-d) mutants, which develop as non-dauers even under harsh conditions. By epistasis analyses, these daf genes have been ordered into three pathways that act in parallel to regulate dauer formation(Vowels and Thomas, 1992; Thomas et al., 1993; Gottlieb and Ruvkun, 1994). Molecular analyses revealed that these pathways correspond to cell-signaling pathways conserved among many animals, namely, cGMP(Coburn et al., 1998; Birnby et al., 2000),TGFβ (Georgi et al.,1990; Estevez et al.,1993; Schackwitz et al.,1996; Ren et al.,1996; Patterson et al.,1997; Inoue and Thomas,2000) and insulin/IGF-I(Morris et al., 1996; Kimura et al., 1997; Ogg et al., 1997; Lin et al., 1997; Ogg and Ruvkun, 1998)pathways. The signals through the three pathways are integrated by a nuclear hormone receptor DAF-12, suggesting that the developmental decision is regulated by a lipophilic hormone(s)(Antebi et al., 2000; Snow and Larsen, 2000). Recently it was reported that one of the Daf-c genes, daf-9, encodes a cytochrome P450 and acts upstream of daf-12. The results indicate that DAF-9 is probably involved in the biosynthetic pathway for a ligand(s) of DAF-12 (Gerisch et al., 2001; Jia et al., 2002). daf-9 mutants as well as special daf-12 mutants that show a Daf-c phenotype form dauer-like larvae, which differ from normal dauer larvae in that their pharynx pumps and is not radially constricted. Furthermore, the Daf-c phenotypes resulting from weak alleles of these genes are greatly enhanced by cholesterol deprivation, consistent with the involvement of steroids in the function of daf-12 and daf-9. C. eleganscannot synthesize sterols, and steroid hormones in C. elegans are considered to be made from cholesterol provided by culture media(Chitwood, 1999).

Conventional Daf-c genes contain mutations that cause highly penetrant Daf-c phenotypes. In addition, there are many other genes that, when mutated,result in only low-penetrance Daf-c, synthetic Daf-c (syn-Daf) and/or Hid(high temperature-induced dauer formation) phenotypes. Syn-Daf mutants and Hid mutants form dauer larvae only in a certain mutant background and at 27°C(a temperature too high for the reproduction of C. elegans),respectively (Bargmann et al.,1990; Avery, 1993; Katsura et al., 1994; Iwasaki et al., 1997; Prasad et al., 1998; Take-Uchi et al., 1998; Ailion et al., 1999; Sze et al., 2000; Sym et al., 2000; Daniels et al., 2000; Ailion and Thomas, 2000; Lanjuin and Sengupta, 2002). The presence of such genes suggests that the daf pathways may be bifurcated further or modified by other pathways. Analysis of these genes will help to elucidate the complicated regulatory network of dauer decision and provide a new insight into the environmental regulation of development.

Here we present genetic and molecular analyses of a new syn-Daf gene, sdf-9. It encodes a protein tyrosine phosphatase-like molecule, which seems to increase the activity of DAF-9 or help the execution of the DAF-9 function. The cells expressing sdf-9 are the same as those expressing daf-9 in the head and have been identified as XXXL/R cells in this study. These cells seem to play a key role in the metabolism of the steroid hormone(s) for DAF-12 that regulates dauer formation.

Worms were grown by standard methods(Brenner, 1974; Sulston and Hodgkin, 1988). Normal NGM agar plates contained 5 μg/ml cholesterol, while cholesterol-poor NGM agar plates (NGM minus cholesterol) were the same but without the cholesterol. Escherichia coli strain OP50 was used as a food for worms. The following C. elegans strains were used in this work: wild-type N2, CB4856, daf-2(e1370), daf-3(e1376), daf-5(e1386),daf-7(e1372), daf-9(e1406), daf-12(m20), daf-16(m27), dpy-5(e61), lon-2(e678),rol-9(sc148), sdf-9(ut157), sdf-9(ut163), sdf-9(ut169), sdf-9(ut174),sdf-9(ut187), unc-31(e169) and unc-51(e369). The sdf-9mutants were backcrossed at least three times.

Four to six adult hermaphrodites were allowed to lay eggs at 20°C for 4-7 hours on 6-cm NGM plates on which E. coli had been grown. Parent animals were then removed and the plates were incubated at the assay temperature. Dauer and non-dauer animals were counted after 3-3.5 days at 20°C, 2.5-3 days at 25°C, and 2-2.5 days at 27°C.

The sdf-9 mutations were previously mapped to the right arm of chromosome V (N. Suzuki, T. Ishihara and I. Katsura, unpublished). More precise mapping was performed as follows. From the genotype sdf-9(ut187)/unc-51(e369)rol-9(sc148), 12/12 Rol non-Unc recombinants were non-syn-Daf, and thus sdf-9(ut187) mapped near or on the right side of rol-9. Using SNPs (Wicks et al., 2001), sdf-9(ut187)was placed to the right of cosmid ZC15 (data not shown). Cosmids and PCR fragments from this region of the genome were introduced into unc-31(e169); sdf-9(ut187) animals, and the resulting transgenic lines were tested for the rescue of the Daf-c phenotype. A 14.8 kb PCR fragment containing the predicted ORFs Y44A6D.4 and Y44A6D.5(C. elegans Sequencing Consortium, 1998) rescued the Daf-c phenotype. This fragment was digested with SpeI and NaeI, and the resulting 8.6 kb fragment containing Y44A6D.4 was subcloned into the SpeI-EcoRV site of pBluescript KS(+) (Stratagene Co.) to generate the plasmid pKO1. pKO1 was digested with EcoRV and HindIII, and the 5.7 kb fragment was cloned into the HincII-HindIII site of pBluescript KS(+) to generate the plasmid pKO2, which also rescued the Daf-c phenotype (data not shown).

Genomic DNA was isolated from the sdf-9 mutant animals, and the sdf-9 region was PCR-amplified. Mutation sites were determined by sequencing the mixed products of two independent PCR reactions.

The exon-intron structure of sdf-9 gene was deduced by sequencing RT-PCR products amplified with gene-specific primers. The 5′ end of the gene was determined by sequencing PCR products obtained with SL1 leader sequence and gene-specific primers, while the 3′ end was determined by sequencing 3′ RACE products with gene-specific and artificial adapter primers.

Germline transformation was carried out as described previously(Mello et al., 1991), at a concentration of 1-60 ng/μl test DNA with 5-15 ng/μl coinjection marker(myo-3::GFP, gcy-10::GFP, or pRF4(rol-6)) and 40-95 ng/μl carrier DNA (pBluescript KS(+)).

In this work, we used the modified Fire vectors pKOGzero and pKORzero as GFP and RFP (dsRed) vectors, respectively, for studying gene expression. pKOGzero was made by ligating fragments of three Fire vectors: a 2.9 kb XhoI-ApaI fragment of pPD95.67, a 98 base NcoI-XhoI fragment of pPD104.53 and a 1.6 kb XhoI-ApaI fragment of pPD104.91. pKORzero was made by substituting RFP cDNA for the GFP cDNA of pKOGzero. An 8.3-kb C. elegans genomic DNA containing the sdf-9 gene was amplified with primers 5′-ATACGGAGCGCAAGGCTGTG-3′ and 5′-GGCTTTGGGATCCACCACGGGCGGC-3′ and digested with HincII and BamHI. The resulting 6.2 kb fragment was subcloned into the HincII-BamHI site of pKOGzero to make pKOG6. pKOG6 was digested with KpnI to remove the nuclear localization signal, and self-ligated to generate pKOG7 (SDF-9::GFP; a 3.5 kb promoter and the entire coding region fused to GFP cDNA). pKOG6 was digested with HindIII and PvuII and the 3.5 kb fragment was cloned into the HindIII-HincII site of pKOGzero and pKORzero to generate pKOG8 (sdf-9p::GFP1; a 3.5 kb promoter region fused to GFP cDNA) and pKOR1 (sdf-9p::RFP; a 3.5 kb promoter region fused to RFP cDNA). pKOG8 was digested with KpnI to remove the nuclear localization signal and self-ligated to generate pKOG9 (sdf-9p::GFP2). The coding region (ApaI fragment) of the sdf-9 gene from pKO1 was cloned into pBluescript KS(+) to generate pKO20 (sdf-9 promoterless gene). The daf-9 promoter region was amplified by PCR as described previously (Jia et al., 2002),and inserted into the PstI site of pKO20 to generate pKO20-d9(daf-9p::sdf-9(promoterless)). For DAF-9::GFP, a 7.7 kb PCR product from the daf-9 genomic DNA containing a 5.4 kb 5′upstream sequence and a 2.3 kb coding sequence was cloned into pKOGzero.

The following cell-specific GFP markers were used: egl-4.a::GFP for IL1VL/R (Fujiwara et al.,2002), gcy-8::GFP for AFDL/R(Yu et al., 1997), gpa-4::GFP for ASIL/R (Jansen et al., 1999), T23G5.5::GFP for CEPDL/R(Jayanthi et al., 1998) (T. Ishihara, unpublished) and ttx-3::GFP for AIYL/R(Hobert et al., 1997). egl-4.a::GFP, gcy-8::GFP, a mixture of gpa-4::GFP and ttx-3::GFP, and a mixture of gpa-4::GFP and T23G5.5::GFP, respectively, were injected into wild-type animals together with sdf-9p::RFP (pKOR1) to make animals that carry an extrachromosomal array containing one or two GFP markers and sdf-9p::RFP. The transgenic animals were grown on normal NGM plates to generate spontaneous mosaic animals, which were observed under a microscope at the L3 to adult stage. Mosaic animals were identified by the presence of sdf-9p::RFP in only one of the two sdf-9-expressing cells. Such animals were found at a frequency of several percent among animals showing RFP fluorescence, and examined for the presence or absence of each GFP marker, using both Nomarski and fluorescence optics.

Laser microsurgery was carried out essentially as described previously(Bargmann and Avery, 1995). pKOG8 (sdf-9p::GFP1) was injected into unc-31(e169) animals together with gcy-10::GFP and pBluescript KS(+), and the established array was introduced into N2, daf-3(e1376), daf-16(m27), and daf-12(m20) animals. The cells expressing sdf-9p::GFP(XXXL/R cells, according to our assignment) were ablated in L1 larvae, which were then cultivated on NGM plates at 25°C. After 2 days, the numbers dauer larvae, dauer-like larvae and non-dauers were scored. The success of cell ablation was judged by the loss of sdf-9p::GFP fluorescence, and animals showing the fluorescence were ignored. Control animals, that were treated in the same way as the operated animals except that no cells were killed, did not form dauer or dauer-like larvae.

pKO20 (sdf-9 promoterless gene) and pKO20-d9(daf-9p::sdf-9(promoterless)), were injected into unc-31(e169); sdf-9(ut187) animals together with gcy-10::GFP and pBluescript KS(+), and the transgenic lines were tested for dauer formation.

A 9.3-kb PCR product from the daf-9 genomic region containing 5.4-kb 5′ upstream sequence, 2.3-kb coding sequence, and 1.6-kb 3′sequence was injected with gcy-10::GFP and pBluescript KS(+) into unc-31(e169), unc-31(e169);sdf-9(ut187), daf-7(e1372) and daf-2(e1370) animals, respectively. Multiple independent transgenic lines were tested for dauer formation. For the experiments with the sdf-9(ut163) mutant, the arrays were established in the unc-31(e169) background and introduced into sdf-9(ut163)animals by crossing. sdf-9(ut187); Ex[daf-9(+), gcy-10::GFP]animals were made from unc-31(e169); sdf-9(ut187); Ex[daf-9(+),gcy-10::GFP].

Although the screen for single mutants with a strong Daf-c phenotype at 25°C is near saturation (Malone and Thomas, 1994), many syn-Daf mutants remain to be isolated(Avery, 1993; Katsura et al., 1994; Iwasaki et al., 1997; Take-Uchi et al., 1998; Ailion et al., 1999; Sym et al., 2000; Daniels et al., 2000; Ailion and Thomas, 2000; Lanjuin and Sengupta, 2002). To identify new genes that regulate dauer formation, we isolated 44 mutants showing Daf-c phenotypes in the unc-31(e169) mutant background and named them sdf (synthetic dauer formation abnormal; N.S., T.I. and I.K., unpublished). Five of them mapped in the same gene, which we named sdf-9. At 25°C, the sdf-9 mutants exhibited a strong Daf-c phenotype in the unc-31(e169) background and a weak Daf-c phenotype in the wild-type background, while at 20°C, they showed a weak Daf-c or a Daf⁺ phenotype even in the unc-31(e169)background (Tables 1, 2). The dauer larvae of the unc-31; sdf-9 double mutants were abnormal in movement due to the unc-31 mutation, but normal in morphology. However, the dauer larvae of the sdf-9 single mutants had a pharynx that pumps and that does not show the radial constriction of the normal dauer pharynx(Fig. 1). In these aspects they resembled the dauer-like larvae of daf-9 or daf-12 Daf-c mutants (Albert and Riddle,1988; Antebi et al.,1998; Antebi et al.,2000; Gerisch et al.,2001; Jia et al.,2002), and hence we call them dauer-like larvae. The sdf-9 dauer-like larvae recovered spontaneously and resumed development to fertile adults after 1-2 days. (78% (n=37) of sdf-9(ut163) dauer-like larvae recovered at 25°C, 94%(n=16) of sdf-9(ut163) and 100% (n=23) of sdf-9(ut187) dauer-like larvae, recovered at 20°C.) The adults of sdf-9 had a normal gonads and vulvas, unlike daf-9 adults. Although the sdf-9 single mutants formed dauer-like larvae at 25°C under non-starving conditions, they formed normal dauer larvae under starving conditions.

Since some syn-Daf mutants show a strong Daf-c phenotype at 27°C (Hid)(Ailion and Thomas, 2000), we examined whether sdf-9 mutants also show a Hid phenotype. Of the five sdf-9 alleles, ut163 produced a strong Hid phenotype, ut157, ut169 and ut187 weaker Hid than unc-31(e169)but stronger than the wild type (Table 2). Such allele-specific Hid phenotypes have been reported for tax-2 mutants (Ailion and Thomas,2000).

To determine the position of the sdf-9 gene in the dauer pathway,we carried out genetic epistasis experiments. Under reproductive growth conditions, the sdf-9(ut163) single mutant formed dauer-like larvae at 25°C. The daf-16(m27) and daf-12(m20) mutations suppressed this phenotype, while daf-3(e1376) enhanced it(Table 3). The presence of the daf-5(e1386) mutation had only a very small effect on the Daf-c phenotype of sdf-9: N2, daf-5, ut163 and daf-5;ut163 mutants produced 0% (n=236), 0%(n=416), 80% (n=256), and 52% (n=256) dauer and dauer-like larvae, respectively, at 25.5°C on NGM minus cholesterol. We also tested suppression of the Daf-c phenotype of unc-31(e169); sdf-9at 20°C, using three sdf-9 alleles, ut163, ut174 and ut187. The results were essentially the same as the suppression of sdf-9(ut163), i.e., daf-16 and daf-12 mutations suppressed the unc-31; sdf-9 Daf-c phenotypes, while daf-3enhanced it (data not shown). The results indicate that sdf-9 acts upstream of daf-16 and daf-12 but downstream of or in parallel with daf-3 and daf-5. Unidentified interaction may exist between daf-3 and sdf-9, because the daf-3(e1376) mutation slightly enhanced the Daf-c phenotype of sdf-9(ut163), as in the case of daf-3(mgDf90) and daf-12(rh273daf-c) (Gerisch et al., 2001).

We found that the sdf-9 gene corresponds to the predicted ORF Y44A6D.4 by three-factor cross, SNP mapping, and rescue experiments(Fig. 2 and Materials and Methods). The cDNA analysis showed that it has the SL1 trans-splice leader sequence, that it encodes a protein of 345 amino acids, and that the prediction of exons was correct except for the 3′ region(Fig. 3).

Blast search revealed that the predicted SDF-9 protein is homologous to protein tyrosine phosphatases (PTPs; Fig. 4). PTPs form a family of diverse proteins characterized by the presence of at least one conserved PTP domain (240-250 amino acids)(Tonks and Neel, 2001). SDF-9 is a non-transmembrane PTP that has only one PTP domain and no other known domains. The PTP domain normally contains the active site motif HCxxxxxR(where x represents any amino acid), but the corresponding sequence of SDF-9 was QSARGSSR. Since the cysteine residue in the active site is essential for the activity of some PTPs (Streuli et al.,1989; Streuli et al.,1990; Guan and Dixon,1990; Guan et al., 1991; Zhou et al., 1994), SDF-9 may not have the activity of protein tyrosine phosphatase.

The coding regions and exon-intron boundaries of the five sdf-9alleles were sequenced to identify mutation sites. The ut163 mutation was a C to T transition resulting in the change of Pro 175 to Leu. The ut157 and ut187 mutations were a G to A transition that changed the conserved Arg 264 to Lys. The ut169 and ut174mutations were a G to A transition, and the TGG codon for Trp 169 was replaced by a stop codon, TGA.

This nonsense mutation would result in a truncated protein lacking the PTP active site and hence these alleles are likely to be null.

We investigated the expression pattern of sdf-9 using GFP and RFP fusion genes. pKOG7 (SDF-9::GFP) rescued the Daf-c phenotype of unc-31(e169); sdf-9(ut187) (data not shown), indicating that the GFP fusion protein is functional and that the place and the time of this expression are sufficient for normal dauer regulation. The expression of SDF-9::GFP was observed from late embryos to adults, and primarily in two cells anterior to the nerve ring. The SDF-9::GFP fusion protein was localized in the cytoplasm. The fluorescence intensity was essentially constant from the threefold stage of embryos to L4 larvae, but slightly weaker at the adult stage. To analyze the cell shape, a promoter-fusion GFP construct, sdf-9p::GFP2, was introduced into wild-type animals. Like SDF-9::GFP, sdf-9p::GFP2 was expressed strongly in two ventral cells anterior to the nerve ring (Fig. 5A), but additional faint signals were occasionally observed in several neurons posterior to the nerve ring. The cells expressing sdf-9p::GFP2 and SDF-9::GFP had two (or sometimes three) short processes, which had a complex shape and looked different from those of sensory neurons. In some transgenic animals, one of the processes occasionally extended to the anterior tip of the head (Fig. 5B). In dauer larvae, one of the processes was long and extended posteriorly, but did not seem to enter the nerve ring (Fig. 5C). The positions of the two cells were variable, with the right-hand cell occasionally found anterior to the metacorpus (19%, n=508).

These results were similar to the expression pattern of daf-9 in the head region (Jia et al.,2002). We therefore made transgenic animals carrying both sdf-9p::RFP and DAF-9::GFP, and compared the expression of the two fusion genes. As shown in Fig. 6A, the cells expressing sdf-9p::RFP were the same as those expressing DAF-9::GFP in the head.

These cells were identified tentatively as IL1VL/R or URAVL/R in other laboratories, mainly on the basis of their positions(Gerisch et al., 2001; Jia et al., 2002). However,since it is difficult to identify cells anterior to the nerve ring only by their positions (Bargmann and Avery,1995), we ascertained the identification of these cells by a newly designed genetic mosaic experiment using known cell-specific markers, as follows.

We first chose the candidates of the sdf-9-expressing cells(Fig. 7) by their positions,using figure 2 of White et al.(White et al., 1986). Then,ciliated sensory neurons (hatched in Fig. 7) were excluded from the candidates, because we found that the cells expressing sdf-9p::RFP did not express che-2p::GFP(data not shown), which is known to be expressed in most ciliated sensory neurons including those shown in Fig. 7 (Fujiwara et al.,1999).

We then carried out mosaic analysis using sdf-9p::RFP and some cell-specific GFP markers (Table 4). When a mixture of two DNA clones is injected into germline cells of C. elegans, they join together and form an extrachromosomal array containing both DNAs (Mello et al.,1991), which is inherited relatively stably but lost at a certain probability by segregation through cell division(Miller et al., 1996). The rationale of this mosaic analysis is that the co-loss or co-presence of GFP fluorescence and RFP fluorescence in mosaic animals should correlate with the degree of lineal relatedness between the GFP- and RFP-expressing cells. Fig. 6B shows the cell lineage of the GFP marker cells and the candidates of sdf-9-expressing cells(white cell bodies in Fig. 7),which was used as a reference in this experiment.

The results (Table 4)indicate, for instance, that the left-hand cell that expresses sdf-9::RFP (abbreviated as sdf-9L below) is nearer to ASIL(88% co-loss or co-presence) and ASIR (80%) in the cell lineage than the right-hand cell that expresses sdf-9::RFP (sdf-9R) (12% and 20%, respectively). However, sdf-9L is nearer to AFDR (55%) and CEPDL(89%) in the cell lineage than sdf-9R (45% and 11%, respectively),while sdf-9R is nearer to AFDL (64%) and CEPDR (79%) than sdf-9L (36% and 21%, respectively). These and other results are summarized as the cell lineages shown in Fig. 6C, which shows that sdf-9R is derived from the ABa blastomere, while sdf-9L is from the ABp blastomere. The cells that satisfy this condition are limited to XXXL/R among the candidate cells shown in Fig. 6B.

There is additional evidence supporting the proposal that the sdf-9::RFP-expressing cells are XXXL/R. If we assume that sdf-9L/R are XXXL/R, as many as 94% of the mosaic animals could be explained by a single loss of the extrachromosomal array, and only 6% of the mosaic animals require two losses and none require three losses, for their formation (see supplemental data). In contrast, if sdf-9L/R were any of the other cells shown in Fig. 6B, much higher percentages of mosaic animals would have suffered multiple losses (see Table S3). Since multiple losses of the extrachromosomal array in one animal do not occur so frequently, the results support the above conclusion.

To determine whether the sdf-9-expressing cells, which we identified as XXXL/R cells, are involved in dauer formation, we killed them in L1 larvae by laser microbeam surgery. At 25°C, about 30% of the operated wild-type animals became dauer-like larvae resembling those of sdf-9and having a non-constricted, pumping pharynx(Table 3). Furthermore, about 60% of the operated unc-31(e169) animals formed Unc but otherwise normal dauer larvae that resemble unc-31; sdf-9 dauer larvae(Table 3). We also killed these cells in some Daf-d mutants: daf-3(e1376), daf-16(m27) and daf-12(m20). When the XXXL/R cells were killed, daf-3animals formed dauer-like larvae, whereas daf-16 and daf-12animals formed neither dauer nor dauer-like larvae(Table 3). These results suggest that XXXL/R cells are important in preventing formation of dauer-like larvae and that the position of their function in the dauer pathway is the same as that of sdf-9.

Because sdf-9 is expressed in the daf-9-expressing cells in the head, we expressed the sdf-9 gene under the control of the daf-9 promoter and tested whether this is sufficient for the wild-type phenotype of sdf-9 gene. The Daf-c phenotype of the unc-31(e169); sdf-9(ut187) double mutant was suppressed by the sdf-9 gene under the control of a daf-9 promoter (see Table S4),but not by the promoterless sdf-9 gene (data not shown). The result indicates that the expression driven by the daf-9 promoter is sufficient for the function of the sdf-9 gene in normal dauer regulation. By using GFP and RFP fusion genes, we already showed that the cells in which sdf-9 is strongly expressed are a subset of cells in which daf-9 is strongly expressed. The result in this section confirms this conclusion through the function of sdf-9.

We also expressed sdf-9 under the control of other promoters. npc-1 and npc-2 are homologs of the Niemann-Pick type C disease gene, which plays a role in intracellular sterol transport(Hoekstra and van IJzendoorn,2000; Ioannou,2001). In C. elegans, it is known that the double mutant npc-2; npc-1 shows a Daf-c phenotype(Sym et al., 2000). We found that npc-1p::GFP was expressed in the sdf-9-expressing cells. Furthermore, the expression of sdf-9 under the control of the npc-1 promoter rescued the Daf-c phenotype of unc-31(e169);sdf-9(ut187) (data not shown). These results suggest that the sdf-9-expressing cells, which we identified as XXXL/R, may be important in the metabolism of steroids or steroid hormone signaling.

It was reported that cholesterol deprivation enhances the weak Daf-c phenotype of daf-9(rh50) and daf-12(rh284) mutants(Gerisch et al., 2001). Since sdf-9 resembles daf-9 in dauer-like larvae and expression pattern, we investigated the dauer formation of sdf-9 mutants on NGM plates without a supplement of cholesterol (NGM minus cholesterol). The sdf-9 Daf-c phenotype was enhanced by cholesterol deprivation(Table 2). We confirmed that a functional SDF-9::GFP (pKOG7) rescued this Daf-c phenotype of the sdf-9(ut163) mutant (see Table S5). We also found that increased concentrations of cholesterol tend to suppress the Daf-c phenotype of sdf-9 mutants(Table 2). These results show that cholesterol or its metabolite regulates dauer formation in cooperation with the regulatory pathway including the sdf-9 gene.

We examined the expression of DAF-9::GFP in the wild type and the following mutant backgrounds: unc-31(e169), sdf-9(ut174), sdf-9(ut187),unc-31(e169); sdf-9(ut174), and unc-31(e169); sdf-9(ut187). The expression did not change in these mutant backgrounds (data not shown), but the transgene weakly suppressed the Daf-c phenotype of unc-31; sdf-9mutants, although it weakly enhanced the dauer formation of the unc-31 mutant (data not shown). We therefore examined whether the wild-type daf-9 gene suppresses the Daf-c phenotypes of sdf-9 single mutants on NGM minus cholesterol and unc-31;sdf-9 double mutants on NGM. As shown in Table 5, the extrachromosomal array of the wild-type daf-9 gene strongly suppressed the Daf-c phenotypes of sdf-9(ut163), sdf-9(ut187) and unc-31(e169);sdf-9(ut187), but not those of daf-7(e1372) and daf-2(e1370), in which daf-9p::GFP was also expressed (data not shown). These results suggest that SDF-9 enhances the function but not the expression of DAF-9.

The sdf-9 gene was identified first by the syn-Daf phenotype of its mutants. Then we cloned the gene by mapping and mutant rescue experiments. The predicted amino acid sequence revealed that the protein product of sdf-9 has homology to protein tyrosine phosphatases (PTPs) but that the active site cysteine of PTP is missing in SDF-9. This cysteine residue forms a bond with a phosphate group in a reaction intermediate(Guan and Dixon, 1991), and many mutant PTPs in which this cysteine is replaced by serine have no detectable phosphatase activity (Streuli et al., 1989; Streuli et al.,1990; Guan and Dixon,1990; Guan and Dixon,1991; Zhou et al.,1994). These mutant PTPs, however, have the activity of binding to phosphorylated substrates (Jia et al.,1995). Many organisms including C. elegans, Drosophila,yeast and humans have such PTP-like molecules as wild-type proteins. These proteins may function as novel regulatory factors in signal transduction systems involving protein tyrosine phosphorylation, by binding to phosphorylated tyrosine residues in proteins(Wishart et al., 1995; Wishart and Dixon, 1998). Alternatively, these PTP-like molecules may have the activity of protein tyrosine phosphatase, but their correct substrates have not been found. In any case, the results of this study suggest that dauer regulation involves a regulatory system containing a tyrosine-phosphorylated protein.

The results of this study showed that the function of sdf-9 is closely related to that of daf-9. First, like daf-9 mutants, sdf-9 mutants form dauer-like larvae under reproductive growth conditions, and normal dauer larvae under dauer-inducing conditions. Moreover,like daf-9, the formation of dauer-like larvae by sdf-9mutants is enhanced by cholesterol deprivation. Second, sdf-9 is expressed in the two head cells in which daf-9 is expressed. Third,like the Daf-c phenotype of daf-9, the Daf-c and syn-Daf phenotypes of sdf-9 are suppressed by daf-12 but not by daf-3.

In some aspects, however, sdf-9 mutants show weaker phenotypes than daf-9 mutants, or the sdf-9 gene has only a part of the properties of daf-9 gene. First, even a putative null mutant of sdf-9 shows only a weak, temperature-sensitive Daf-c phenotype, while many daf-9 mutants show strong, nonconditional Daf-c phenotypes. Second, the formation of dauer-like larvae is suppressed by daf-16 in the case of sdf-9 mutants but not in daf-9 mutants. Third, sdf-9 is expressed essentially only in two cells in the head, while daf-9 is expressed in hypodermis, vulval blast cells and spermatheca at some stages of development, besides the two head cells. Fourth, sdf-9 mutants seem to show normal longevity, at least at 25°C(K.O., T.I. and I.K., unpublished), while daf-9 mutants show increased longevity under certain conditions(Gerisch et al., 2001; Jia et al., 2002). These results may be explained, if the sdf-9 function is partially redundant, if sdf-9 controls only part of the multiple functions of daf-9, or if sdf-9 just enhances the daf-9function.

The functional relationship between sdf-9 and daf-9 can be further speculated on. This study revealed that the Daf-c and syn-Daf phenotypes of sdf-9 are suppressed by the overproduction of the wild-type daf-9 gene, and that DAF-9::GFP is expressed even in the sdf-9 mutant background. The results show that SDF-9 enhances the function but not the synthesis of DAF-9. SDF-9 may increase the activity of the DAF-9 P450 molecule or may help the execution of the DAF-9 function, for instance, through the metabolism or transport of the substrate or product of DAF-9.

Although we have discussed that sdf-9 is closely related to daf-9 in its function, the position of the sdf-9 gene in the dauer regulatory pathway is yet another issue to be discussed. The Daf-c phenotypes of sdf-9 as well as daf-9 were suppressed by daf-12 but not by daf-3 or daf-5. These results indicate that the sdf-9 gene, like daf-9, acts upstream of daf-12, but downstream of or in parallel with the TGFβ pathway genes daf-3 and daf-5. Since the sdf-9 Daf-c phenotype is not enhanced by daf-5, its enhancement by daf-3is probably related to the Daf-c phenotype of daf-3 (but not daf-5) at 27°C (Ailion and Thomas, 2000).

Unlike daf-9, the Daf-c phenotypes of sdf-9 mutations were suppressed by daf-16. This result can be interpreted in the light of the laser microsurgery experiments. The dauer larva formation induced by the ablation of XXXL/R showed the same suppression pattern as that of sdf-9 mutants. Furthermore, like sdf-9 mutations, the operated larvae produced mainly dauer-like larvae in the wild-type background,and normal dauer larvae in the unc-31 background. Those results suggest that the function of XXXL/R is located upstream of daf-16 in the dauer regulatory pathway and that the function of daf-9 in other cells is located downstream of or in parallel with daf-16 in the pathway.

The XXXL/R-specific suppression by daf-16 suggests that the DAF-2 insulin receptor signal transduction may play a tissue-specific role in these cells, which involves SDF-9 function and whose defect leads to dauer-like larvae formation. Experiments on daf-2 mosaic animals(Apfeld and Kenyon, 1998)revealed that animals that are daf-2(–) in AB, ABa, ABar, ABp,or ABplaa cell lineages form dauer-like larvae (`class II dauers'). Since XXXL and XXXR cells form in the ABplaa and ABarpa lineages, respectively, the result is in agreement with the idea that loss of daf-2 gene activity in XXXL/R results in the formation of dauer-like larvae. It remains to be examined whether daf-2 is expressed in XXXL/R and whether SDF-9 interacts with the DAF-2 tyrosine kinase, especially its tyrosine-phosphorylated form.

We identified the sdf-9- and daf-9-expressing cells in the head as XXXL/R by mosaic experiments. These cells are known as embryonic hypodermal cells whose function at later stages is unknown(Sulston et al., 1983). White et al. (White et al., 1986)included these cells in the list of neurons and their associated cells in the head (figure. 2 of White et al.). This study revealed a function of XXXL/R, namely, regulation of dauer formation. The most direct evidence for this function is that killing of these cells induces formation of dauer-like larvae in the wild-type background and that of dauer larvae in the unc-31 background.

Gerisch et al. (Gerisch et al.,2001) and Jia et al. (Jia et al., 2002) also carried out laser microsurgery experiments on these cells. Although their results are partly different from ours, we think there is no major contradiction. Gerisch et al.(Gerisch et al., 2001)reported that ablation of daf-9-expressing head cells in 41 daf-9(dh6); Is[daf-9::GFP] animals resulted in dauer-like larva formation in only four (10%) of them. The difference may be due to overexpression of daf-9 by Is[daf-9::GFP] as well as other differences in experimental conditions. Jia et al.(Jia et al., 2002) found that none of the six N2; Ex[daf-9p::GFP] animals formed dauer-like larvae after killing of the daf-9-expressing head cells,but they argued that the timing of the surgery might be too late. They also killed these cells in daf-9(m540); Ex[daf-9p::daf-9 cDNA::GFP] animals, and all the four animals in which the surgery was successful formed dauer-like larvae.

XXXL/R cells seem to regulate dauer formation through the functions of daf-9 and sdf-9, which are expressed in these cells and whose mutants form dauer-like larvae. The following results show that these functions are probably related to steroid hormonal signaling and/or steroid metabolism. First, daf-9 encodes a cytochrome P450 of the CYP2 family, whose members are involved in steroid metabolism(Gerisch et al., 2001; Jia et al., 2002). Second, the daf-12 gene, which acts downstream of daf-9, encodes a steroid hormone receptor (Antebi et al.,2000). Third, the Daf-c phenotypes of daf-9, sdf-9 and daf-12(daf-c) mutations are enhanced by the deprivation of cholesterol (Gerisch et al.,2001; Jia et al.,2002; this study). Fourth, the expression of SDF-9 under the control of the npc-1 promoter rescues the Daf-c phenotype of unc-31(e169); sdf-9(ut187), where npc-1 is a homolog of the Niemann-Pick type C disease gene, which plays a role in intracellular sterol transport (Hoekstra and van IJzendoorn,2000; Ioannou,2001).

In conclusion, we showed that SDF-9, a protein tyrosine phosphatase-like molecule, is involved in the regulation of dauer formation. Its function is closely related to that of DAF-9, a P450 molecule: it probably enhances the activity of DAF-9 or helps the execution of the DAF-9 function. Furthermore,SDF-9 is expressed in two head cells in which DAF-9 is expressed. We identified these cells as XXXL/R cells, which are known as embryonic hypodermal cells but whose function at later stages remains to be studied. Since this study on SDF-9 and former studies on DAF-9 suggested that the functions of these proteins are related to steroid metabolism or steroid hormonal signaling, XXXL/R cells seem to play a key role in the metabolism or function of a steroid hormone(s) that acts in dauer regulation.

We thank Hiroshi Kagoshima for gcy-8::GFP, Oliver Hobert for ttx-3::GFP, Andy Fire for GFP vectors, Shohei Mitani for RFP cDNA,the Caenorhabditis Genetics Center for strains, Manabi Fujiwara for communicating unpublished results, Shuji Honda for help in the measurements of longevity, and Mariko Sugiura for help in preparing the manuscript.