The C. elegans gene lin-36 acts cell autonomously in the lin-35 Rb pathway

Receptor tyrosine kinase (RTK) and Ras-mediated signal transduction pathways have been found to function normally in the control of the development of numerous organisms (Pawson and Bernstein, 1990; Perrimon, 1994) and have been implicated in oncogenesis (Egan and Weinberg, 1993). The core signal transduction pathway is well conserved and well understood (Egan and Weinberg, 1993). However, little is known about how this pathway is regulated. To approach this problem, we are analyzing the regulation of the RTK- and Ras-mediated signal transduction pathway of vulval induction in Caenorhabditis elegans (for reviews, see Horvitz and Sternberg, 1991; Eisenmann and Kim, 1994; Sundaram and Han, 1996).

In vulval induction a signal from a single cell in the developing gonad, the anchor cell, activates an RTK- and Ras-mediated signal transduction pathway to induce three nearby hypodermal blast cells, P5.p, P6.p and P7.p, to adopt vulval fates. These three cells are members of the vulval equivalence group, a set of six cells collectively referred to as P(3-8).p, all of which have the potential to adopt either of two vulval fates (called 1° and 2°) or a non-vulval fate (called 3°). Mutations that decrease activity in this signaling pathway block vulval induction and result in a vulvaless (Vul) phenotype, whereas mutations that increase activity in this pathway can cause the P3.p, P4.p and P8.p cells to adopt vulval fates and result in a multivulva (Muv) phenotype.

A number of genes have been identified that appear to regulate the activity of the signal transduction pathway of vulval induction. Mutations in the genes lin-2, lin-7 and lin-10 produce a Vul phenotype similar to that produced by reduction-of-function mutations in components of the let-60 ras pathway. lin-2, lin-7 and lin-10 appear to act to localize the LET-23 receptor to the basolateral membrane of the P(3-8).p cells, thereby upregulating the activity of the receptor tyrosine kinase pathway (Ferguson et al., 1987; Hoskins et al., 1996; Simske et al., 1996).

Antagonists of the vulval inductive pathway have also been identified. A group of genes known as the synthetic Multivulva (synMuv) genes inhibit the adoption of vulval fates by the P(3-8).p cells. A synthetic interaction between two different synMuv mutations is required to produce a synMuv phenotype (Horvitz and Sulston, 1980; Ferguson and Horvitz, 1985, 1989). Specifically, the synMuv mutations define two classes of genes, A and B. Animals must carry both a class A and a class B mutation to have a Muv phenotype; animals that carry mutations in either class but not in both classes have a wild-type vulval phenotype. Ferguson and Horvitz (1989) proposed that the synMuv genes encode the components of two functionally redundant pathways that negatively regulate vulval development: if both pathways are blocked, P(3-8).p cells that would normally be prevented from adopting vulval fates instead adopt vulval fates. Four class A genes (lin-8, lin-15A, lin-38 and lin-56) and ten class B genes (lin-9, lin-15B, lin-35, lin-36, lin-37, lin-51, lin-52, lin-53, lin-54, lin-55) have been identified (Horvitz and Sulston, 1980; Ferguson and Horvitz, 1989; J. H. Thomas and H. R. Horvitz, unpublished observations). lin-15 is a complex locus with genetically separable class A and class B activities encoded by two nonoverlapping transcripts (Ferguson and Horvitz, 1989; Clark et al., 1994; Huang et al., 1994). Five synMuv genes have been cloned. Three, lin-15A, lin-15B and lin-9, encode novel proteins (Clark et al., 1994; Huang et al., 1994; Beitel, 1994). lin-35, a class B gene, encodes a protein similar to the retinoblastoma-susceptibility gene product Rb, and lin-53, another class B gene, encodes a protein similar to the RbAp48 protein, which interacts with Rb (Lu and Horvitz, 1998). Genetic mosaic analyses suggest that lin-15A, lin-15B and lin-37 act cell non-autonomously and most likely in the hypodermal syncytium, suggesting that the synMuv genes encode the components of two redundant systems by which the hypodermis inhibits vulval induction (Herman and Hedgecock, 1990; Hedgecock and Herman, 1995).

In this paper, we describe studies of the class B synMuv gene lin-36. We report that lin-36 encodes a novel protein, is likely expressed in the nuclei of the P(3-8).p cells, and acts cell autonomously.

C. elegans strains were maintained and genetically manipulated as described by Brenner (1974) and were grown at 20°C unless otherwise stated. N2 was the wild-type strain. Mutations used in this study are described by Hodgkin et al. (1988) or as indicated in the text.

Triple and double mutants for gene interaction studies were constructed essentially as described by Ferguson et al. (1987). For all genes tested, either null alleles or the strongest known mutations were used.

L4 males of genotype lin-8(n111) were mutagenized with 4,5´,8-trimethylpsoralen (TMP) and ultraviolet radiation (Yandell et al., 1994) and mated with hermaphrodites of genotype lin-8(n111); sma-3(e491) lin-36(n766) unc-32(e189). A total of 23,666 haploid genomes were screened, and one Muv non-Sma non-Unc F₁ animal was isolated. The new mutation was made homozygous and backcrossed three times.

In a second non-complementation screen, L4 males of genotype lin-8(n111) were mutagenized with ethyl methanesulfonate (EMS) (Brenner, 1974) and mated with hermaphrodites of genotype lin-8(n111); lin-36(n766) unc-32(e189); lon-2(e678) xol-1(y70). A total of 21,003 haploid genomes were screened, and five Muv non-Unc non-Lon F₁ animals were isolated. The new mutations were made homozygous and backcrossed three or four times.

P(3-8).p cell lineages were observed using Nomarski optics as described by Sulston and Horvitz (1977).

General DNA manipulations and analyses were performed essentially according to the protocols of Sambrook et al. (1989). The RNA filter used for northern blots was a gift from L. Bloom.

Subclones were constructed from cosmid E02E3. MluI and NcoI were used to delete cosmid DNA to construct pJHT1 and pJHT2, respectively. The following subclones contained the DNA between the listed restriction sites: pJHT9, MluI and KpnI; pJHT12, SpeI and KpnI; pJHT13, SalI and KpnI; pJHT15, SalI and SacI; pJHT16, SalI and XbaI; pJHT18, ClaI and XbaI.

Germline transformation by microinjection was done according to Mello and Fire (1995). DNA for injection was purified using Qiagen columns (Qiagen Inc., Chatsworth, CA). Cosmids and plasmids for rescue experiments were injected at concentrations of 20-80 µg/ml. The lin-36::GFP reporter was injected at 50 µg/ml. pRF4, a plasmid containing the rol-6(su1006) dominant allele, was used as a coinjection marker in all experiments at a concentration of 80 µg/ml (Mello and Fire, 1995).

We screened a mixed-stage cDNA library (Barstead and Waterston, 1989) with the ³²P-labeled 5.0 kb insert in pJHT16 and isolated one cDNA. After limited ExoIII digestion, M13 phage was used to prepare single-stranded DNA (Sambrook et al., 1989), and DNA sequences were determined using the method of dideoxy chain termination and the Sequenase enzyme (US Biochemical, Cleveland, OH). One region in the 3´ untranslated region required the use of Taq polymerase at 95°C for sequence determination (Promega, Madison, WI).

Genomic DNA was purified from N2 and strains carrying all known lin-36 mutations. lin-36(n3090) DNA was digested with SalI and XbaI, NcoI, NcoI and XbaI, EcoRV and HinfI, and analyzed by Southern blot analysis (Sambrook et al., 1989) using the 3²P-labeled insert of pJHT16 as a probe. The lin-36 coding region and the regions of introns near the splice sites were amplified using the polymerase chain reaction (PCR), and the sequences of these PCR products were determined using an automated ABI 373A cycle sequencer (Applied Biosystems, Foster City, CA).

Genetic mosaic analysis of lin-36 was conducted using the method of free duplication loss described by Herman (1984). Two strategies were adopted to find mosaic animals using a strain of genotype lin-8(n111); dpy-17(e164) ncl-1(e1865) lin-36(n766) unc-36(e251); sDp3[dpy-17(+) ncl-1(+) lin-36(+) unc-36(+)]. First, mosaic L3 and L4 larvae were isolated on the basis of gross morphology and then observed using Nomarski optics to define the cell division at which duplication loss occurred by observing the enlarged nucleoli seen in cells mutant for ncl-1 (Hedgecock and Herman, 1995). Specifically, Unc non-Dpy animals result from duplication loss in AB or ABp; Dpy non-Unc animals result from loss in P₁, P₂ or C (Yuan and Horvitz, 1990). In these animals, a P₂ loss can be distinguished from a C loss by the segregation of all Dpy Ncl Unc progeny. Semi-Unc non-Dpy animals are the result of a loss in either ABpl or ABpr. The second strategy was to examine 1,515 seemingly wild-type L3 and L4 larvae for the Ncl phenotype using Nomarski optics. This approach enabled us to identify ABa losses and losses in the ABp lineages later than ABpl/r. The Ncl phenotype was scored primarily in neurons; it cannot be scored reliably in the Pn.p cells themselves. From the P₁ lineage, the pharyngeal neurons I4, I6, M1, M4 and M5, the tail neurons PVR and DVC, and occasionally hyp11 were scored. From ABp, the HSNs and most of the ABp-derived neurons of the head and tail were scored. The subset of ABp-derived neurons with a Ncl phenotype and their lineage relationship (Sulston and Horvitz, 1977; Sulston et al., 1983; White et al., 1986) allowed us to determine the cell division at which duplication loss occurred. From ABa, the BDUs, the ALMs, the pharyngeal neurons I5, MI, the M2s and the M3s, and most of the ABa-derived neurons of the head were scored.

A lin-36::GFP reporter was constructed by ligating a 4.4 kb SalI SacI fragment from pJHT16 to a linker fragment and SalI XmaI double-digested pPD95.79, a vector containing a green fluorescent protein (GFP) coding region with a C. elegans unc-54 3´ end and a fluorescence enhancing S65C mutation (Chalfie et al., 1994; Heim et al., 1994, 1995; A. Fire, personal communication). The linker fragment was constructed by using the primers pG5F (5´ AGAGCAATGGAAGACGC) and pG3W (5´ TCCCCCCGGGTTGTTGAGAATCCGATG) to amplify DNA from pJHT16 and digesting with XmaI and SacI. pG3W anneals to the last codons of lin-36, and adds an XmaI site followed by four additional nucleotides.

Five rescued lines were isolated after germline transformation of lin-36(n766); lin-15(n767) animals. To induce chromosomal integration, transformants were exposed to 4,100 rads of γ-irradiation from a ₆₀Co source (Mello and Fire, 1995). Two lines that retained lin-36 rescue ability and GFP staining were isolated.

Laser microsurgery was conducted essentially as described previously (Sulston and White, 1980; Avery and Horvitz, 1987; Thomas et al., 1990). 60 mM sodium azide was used as an anesthetic. Mock-ablated animals were used as an operation control.

Seven lin-36 mutations had been identified previously in screens for class B synMuv mutations (Ferguson and Horvitz, 1989; J. H. Thomas and H. R. Horvitz, unpublished observations). To isolate additional lin-36 mutations, we conducted two non-complementation screens using a lin-8(n111) class A background. In one screen, 4,5´,8-trimethylpsoralen was used as a mutagen. One lin-36 mutation, n3090, was isolated and shown by Southern hybridization to be associated with an EcoRV restriction fragment length polymorphism (RFLP). This RFLP was produced by a point mutation rather than a deletion (see below). In the other screen, EMS was used as a mutagen, and five lin-36 mutations (n3093, n3094, n3095, n3096, n3097) were isolated. This frequency of EMS-induced lin-36 alleles, 2.4×10⁻⁴ per haploid genome, is close to that expected for the average EMS-induced loss-of-function mutation in C. elegans, 5×10⁻⁴ (Brenner, 1974; Meneely and Herman, 1979; Greenwald and Horvitz, 1980).

The phenotypes caused by the strongest lin-36 mutant alleles (n747, n766, n2243, n3093, n3095, n3096) were essentially identical (Table 1), suggesting that they may be complete loss-of-function alleles. The frequency of isolation in our non-complementation screen is consistent with this hypothesis. However, animals heterozygous for a strong lin-36 allele and a deficiency of the lin-36 region displayed a phenotype weaker than that of the corresponding homozygote. For example, the penetrance of the Muv phenotype of lin-8; lin-36(n766)/nDf20 animals was 87% (n=259), whereas that of lin-8; lin-36(n766) homozygotes carrying the same heterozygous marker, unc-32(e189), was 95% (n=214). Similar results were observed for lin-36(n3096). In addition, animals heterozygous for a weak lin-36 allele and a strong lin-36 allele displayed a stronger phenotype than animals heterozygous for a weak allele and a deficiency. Specifically, the penetrance of the Muv phenotype of lin-8; lin-36(n2235)/nDf20 animals was 74% (n=194), whereas that of lin-8; lin-36(n2235)/lin-36(n766) animals was 87% (n=214) and that of lin-8; lin-36(n2235)/lin-36(n747) animals was 91% (n=188); all animals scored were heterozygous for unc-32(e189). These observations suggest that strong alleles of lin-36 reduce lin-36 function more than does a deficiency. Using the polymerase chain reaction (PCR), we showed that all of the lin-36 coding sequence was deleted from the nDf20 chromosome (data not shown). It is possible that a linked, haplo-insufficient suppressor of the synMuv phenotype is deleted by this deficiency. Alternatively, the strong lin-36 alleles may interfere with the activities of other class B synMuv genes in addition to reducing lin-36 activity. Although our data are consistent with the hypothesis that we have isolated complete loss-of-function alleles, it remains possible that the strongest lin-36 alleles are not nulls but rather partially reduce lin-36 activity and also act to antagonize the activity of other members of the lin-35 Rb pathway.

We analyzed the vulval defects of strains carrying lin-36 mutations at the cell lineage level by directly observing the lineages of the P(3-8).p descendants. Animals carrying only a strong lin-36 mutation, n766, displayed wild-type P(3-8).p cell lineages, as did animals carrying only the strong class A mutation, lin-15(n767) (Table 2). The cells of the vulval equivalence group that are not induced in the wild type, P(3,4,8).p, were induced in animals of genotype lin-36(n766); lin-15(n767). In the 10 animals examined, 29/30 of these cells were completely induced and the remaining 1/30 (a P8.p cell that divided to give an OOS lineage; see Sternberg and Horvitz, 1986, 1989, for nomenclature) showed partial induction.

Previous data placed lin-36 between lin-13 and unc-36 on linkage group III (Ferguson and Horvitz, 1989). We mapped lin-36 under nDf20 and between egl-5 and unc-36 (Fig. 1A), both of which had been cloned and positioned on the physical map (Wilson et al., 1994). The genetic data suggested that lin-36 was located about midway between egl-5 and unc-36, and we injected cosmid pools from this region into animals bearing a class A synMuv mutation and a lin-36 mutation in germline transformation experiments. Cosmid E02E3 rescued the Muv phenotype of both lin-8(n111); lin-36(n747) and lin-36(n766); lin-15(n767) animals and was used to construct subclones. These experiments defined a 5.0 kb minimal rescuing fragment, the XbaI SalI insert of pJHT16 (Fig. 1B).

This 5.0 kb fragment was used to probe a northern blot containing poly(A)+ RNA from mixed-stage animals and recognized a single band of approximately 3.5 kb. The same fragment was used as a probe to screen a cDNA library and isolate a single clone. The 3,400 nucleotide cDNA included a poly(A) tail and nine nucleotides of the SL1 trans-spliced leader (Krause and Hirsh, 1987; Huang and Hirsh, 1989), indicating that it was full-length. The sequence of the cDNA was compared to the genomic sequence from the region generated by the C. elegans genome sequencing project (Wilson et al., 1994) to define the exon and intron structure of the gene (Fig. 1C).

The coding sequence of lin-36 was determined from the cDNA clone (Fig. 2). Conceptual translation yielded a novel gene product of 962 amino acids with a predicted molecular mass of 108 kDa. The predicted protein product is highly hydrophilic and has no hydrophobic stretches of significant length. There are few obvious protein motifs. There is a potential nuclear localization sequence (AKKRK) at amino acids 528-532 (for review, see Goldfarb, 1989) and a poly(glutamine) stretch at the C terminus. The region between residues 101 and 376 is cysteine- and histidine-rich; it contains 15 histidines and 14 cysteines. One region containing two cysteines (residues 268 and 271), two histidines (residues 284 and 288) and a leucine (residue 281) fits zinc-finger spacing requirements; however, other amino acids are not conserved (Rosenberg et al., 1986; Tautz et al., 1987), and we suspect that this region does not form a zinc finger.

We used PCR to amplify lin-36 DNA from mutants to determine the sequence changes in these strains. Six mutations are missense mutations, five are nonsense mutations, one is a splice-site mutation, and one contains a missense mutation as well as a silent mutation (Table 3; Fig. 1C). Three of the ochre mutations, lin-36(n766), lin-36(n3095), lin-36(n3096) and the amber mutation lin-36(n3093) eliminate 167, 394, 521 and 384 amino acids of the 962 amino acid coding sequence, respectively. All produced a strong phenotype (Table 1). The fourth ochre mutation, lin-36(n2235), eliminates only 65 amino acids of the coding sequence and produced a weaker phenotype.

To determine the site of action of lin-36, we conducted a genetic mosaic analysis. A strain of genotype lin-8(n111); dpy-17(e164) ncl-1(e1865) lin-36(n766) unc-36(e251); sDp3(f) was used to generate genetic mosaics as described in Materials and methods. Mosaics were identified and characterized as L3 and L4 larvae and subsequently scored for the Muv phenotype upon reaching adulthood. Strains mutant for lin-36 exhibit maternal rescue of the Muv phenotype (Ferguson and Horvitz, 1989; J. H. Thomas and H. R. Horvitz, unpublished observations). Presumably, maternally deposited wild-type lin-36 RNA or protein persists through several cell divisions and is present in sufficient quantities in the relevant cells to affect the penetrance of the Muv phenotype. This maternal effect affects the mosaic analysis, so we determined that the penetrance of the Muv phenotype of Dpy Ncl Unc animals homozygous for lin-36 descended from duplication-bearing mothers was 69% (n=240). This number sets an upper limit on the fraction of Muv animals we expected from a mitotic loss of the duplication in the lineage from which the cellular focus of lin-36 is derived.

Duplication loss in AB, ABp, ABpl or ABpr often produced Muv animals (Fig. 3). Loss in ABa or P₁, P₂ or C failed to result in Muv animals. These data are consistent with an anatomical focus in both the ABpl and ABpr lineages. The P(3-8).p cells derive from both of these lineages. The somatic gonad is derived from EMS, a daughter of P₁. By contrast, the major hypodermal syncytium, hyp7, derives from ABa, ABp and C; C, a descendant of P₁, makes the largest hyp7 contribution during embryogenesis. ABa makes a larger contribution to hyp7 than either ABpl or ABpr, yet losses in ABpl or ABpr often resulted in a Muv phenotype, whereas losses in ABa or P₁ did not (Fig. 3). Since vulval development is thought to involve the P(3-8).p cells, hyp7 and the gonadal anchor cell, these observations indicate that the P(3-8).p cells, which are derived from ABpl and ABpr, are the best candidates for the focus of lin-36 activity. Thus, lin-36 probably acts cell autonomously. As discussed below, synMuv genes with a postulated focus in hyp7 show very different clonal behavior. We observed three double losses (Fig. 3 legend). One occurred in AB and MS/EMS and resulted in a Muv phenotype, presumably because of the AB loss. Another occurred in ABpl and C and resulted in a Muv phenotype, presumably because of the ABpl loss. A third occurred in ABprapp and ABa and resulted in a Muv phenotype, presumably because of the ABprapp loss. Thus, we can account for the phenotypes of mosaic animals with double losses on the basis of one of the two losses.

The simplest interpretation of these results is that lin-36 acts cell autonomously in the P(3-8).p cells, although it is conceivable that lin-36 is instead required in close relatives of the P(3-8).p cells, possibly hyp7 cells derived from V3 or V5 in both the ABpl and ABpr lineages. It is also conceivable that lin-36 acts in both the P(3-8).p cells and critical role in the P(3-8).p cells. In such a model, loss of lin-36 activity in both the P(3-8).p cells and some hyp7 nuclei results in a Muv phenotype, although loss in only hyp7 does not. Although the penetrance of the Muv phenotype in animals that have lost lin-36 activity in AB or ABp (64%, n=44) is essentially the same as that in animals that have lost lin-36 activity meiotically (69%, n=240), the penetrance of the Muv phenotype in animals that have lost lin-36 activity in ABpl or ABpr is substantially lower (2/10, excluding double losses). These data are consistent with a lin-36 contribution from hyp7 in addition to that from P(3-8).p cells. However, these data are also consistent with the cell autonomous model. Specifically, the control animals that lost lin-36 activity meiotically displayed not only incomplete penetrance but also variable expressivity of the Muv phenotype, in that not all P(3-8).p cells adopted vulval fates. Since ABpl and ABpr mosaic animals have only one or two P(3-8).p cells that can adopt ectopic cell fates, the cell autonomous model predicts a reduced frequency of Muv animals from these mosaic classes compared to the AB and ABp mosaic classes.

We determined the expression pattern of a lin-36::GFP reporter construct containing 1 kb of sequence 5´ to the ATG start site, the entire lin-36 open reading frame, all lin-36 introns and the GFP gene fused in-frame to the last codon of lin-36. This lin-36::GFP reporter transgene rescued the Muv phenotype of a lin-36; lin-15(n767) strain, indicating that it was expressed in cells needed for lin-36 function. lin-36::GFP was expressed in the nuclei of the P(3-8).p cells and their descendants during vulval cell determination, division and invagination (Fig. 4). Similar observations have been made with an independently constructed lin-36::GFP reporter (A. Hajnal and S. Kim, personal communication). The expression pattern of lin-36 in the P(3-8).p cells and their descendants is consistent with our conclusion from the genetic mosaic analysis that lin-36 acts cell autonomously. The P(3-8).p cell expression of lin-36::GFP was not altered by mutations in other class B genes, specifically lin-9, lin-15, lin-35, lin-51, lin-52, lin-53 and lin-55 (Ferguson and Horvitz, 1989; J. H. Thomas and H. R. Horvitz, unpublished observations) (data not shown).

Besides the P(3-8).p cells, many other cells expressed GFP in strains bearing our reporter construct. Most notably, neurons of the head, tail and ventral cord expressed lin-36::GFP throughout development, and sporadic fluorescence was infrequently observed in the germline (data not shown). Very weak staining was observed in hypodermal and intestinal nuclei.

lin-36::GFP expression in all cases was localized to nuclei. The GFP reporter gene used in the construction of this reporter construct did not contain a nuclear localization sequence, but as mentioned above, the lin-36 coding region contains a potential nuclear localization sequence. These data suggest that lin-36 may act in the nuclei of the P(3-8).p cells to mediate its effect on vulval development.

The Vul phenotype of mutations in the let-23 receptor tyrosine kinase gene is epistatic to the Muv phenotype of the lin-8; lin-9 and lin-15AB synMuv mutations (Ferguson et al., 1987; Huang et al., 1994). To determine if lin-36 acts similarly in the genetic pathway controlling vulval induction, we constructed triple mutants using let-23 mutations and a lin-36 mutation (Table 4). The Vul phenotype of let-23 mutations is epistatic to the lin-36; lin-15A synMuv phenotype. Thus, let-23 activity is required for the adoption of vulval fates by P(3-8).p cells in synMuv mutants. The let-23 receptor tyrosine kinase acts genetically downstream of, or in parallel to, lin-36 as well as the other synMuv genes.

lin-2, lin-7 and lin-10 positively regulate the let-23-mediated signal transduction cascade, apparently by localizing the LET-23 receptor to the basolateral membrane of the P(3-8).p cells (Simske et al., 1995; Hoskins et al., 1996). A mixture of Muv and Vul phenotypes, i.e. animals both with ectopic ventral protrusions and bloated with eggs (or bags of worms), has been reported for animals triply mutant for lin-8; lin-9 and either lin-2, lin-7 or lin-10; however, double mutants between the synMuv mutation lin-15(n309) and a lin-2, lin-7 and lin-10 mutation have been reported to be Muv (Ferguson et al., 1987). Since the lin-9 mutation used in these experiments was not a null allele (Ferguson and Horvitz, 1989), we repeated these experiments using the strong lin-36(n766) allele. Some animals were Muv, some were Vul, some had a wild-type phenotype, and some expressed both a Vul and a Muv phenotype (Table 4). Thus, in these triple mutants the Muv and Vul phenotypes were coexpressed. Since this result is similar to that observed by Ferguson et al. (1987) with lin-8; lin-9 and lin-2, lin-7 or lin-10 triple mutants, we repeated the lin-15AB experiment, but used another allele, lin-15(e1763), which is a slightly larger deletion of lin-15 than is lin-15(n309). Both alleles eliminate both the lin-15A and lin-15B gene products and their activities (Clark et al., 1994). The lin-15AB double mutants with lin-2, lin-7 and lin-10 mutations showed coexpression of the Muv and Vul phenotypes (Table 4). These results are consistent with our observations of the lin-36; lin-15A mutants; however, unlike the lin-36 triple mutants, most animals of these genotypes were either Muv or Muv Vul; few were wild type or Vul. Since almost all animals expressed the Muv phenotype and some expressed the Vul phenotype in addition to the Muv phenotype, it is likely that in the previous studies by Ferguson et al. (1987) the Vul phenotype was not noticed. We conclude that in mutants carrying class A and class B synMuv mutations and a lin-2, lin-7 or lin-10 mutation there is a coexpression of the Muv and Vul phenotypes.

The anchor cell is necessary for the adoption of vulval cell fates by the P(3-8).p cell in wild-type animals, as demonstrated by experiments in which the anchor cell was ablated by laser microsurgery (Kimble, 1981). lin-8; lin-9 and lin-15AB synMuv mutants in which the anchor cell was ablated using laser microsurgery are Muv, suggesting that in the absence of negative regulation by the synMuv genes, P(3-8).p cells do not require the anchor cell signal to adopt vulval cell fates (Ferguson et al., 1987). We tested whether anchor cell-deficient synMuv mutants carrying lin-36 mutations were similarly Muv. We eliminated the entire somatic gonad, which includes the anchor cell progenitor, using laser microsurgery to ablate Z1 and Z4, the precursors of the somatic gonad (Kimble and Hirsh, 1979). These operations were performed at the early L1 larval stage. The ablation of Z1 and Z4 prevents germline proliferation and produces a sterile animal with no gonad (Kimble and White, 1981). This phenotype served as an internal control to confirm the ablation of the somatic gonad. We ablated Z1 and Z4 in lin-36; lin-15A animals and found that 11/11 of these animals produced ectopic vulval tissue.

We tested whether the anchor cell signal is responsible for the adoption of vulval fates by P(3-8).p cells in which receptor localization is disrupted and the negative regulatory pathways are eliminated by triple mutants carrying synMuv mutations and mutations in lin-2, lin-7 or lin-10. We ablated Z1 and Z4 in lin-7; lin-36; lin-15A animals and observed that 14/14 operated animals had ectopic vulval tissue. Thus, in animals in which the gonadal signal, the receptor localization system and the synMuv-mediated inhibitory pathways have been disrupted, the P(3-8).p cells can adopt vulval fates. These results, together with the triple mutant analyses, are consistent with the hypothesis that a ligand-independent and receptor localization-independent activity of the let-23 signal transduction pathway can lead to the adoption of vulval cell fates. Presumably, it is this ligand-independent and receptor localization-independent activity that the synMuv genes normally inhibit.

Our results suggest that lin-36 encodes a novel protein that is required within the P(3-8).p cells to negatively regulate vulval development. Our lin-36 genetic mosaic data are in marked contrast to the lin-15 and lin-37 genetic mosaic data that led to the conclusion that these synMuv loci act cell non-autonomously (Herman and Hedgecock, 1990; Hedgecock and Herman, 1995). Specifically, Herman and Hedgecock (1990) found that loss of lin-15 function in P₁, AB or ABp, ABpl or ABpr can result in a Muv phenotype. hyp7, which derives from these lineages, was postulated to be the anatomical focus of both lin-15A and lin-15B activity, since both activities must be lost to produce a Muv phenotype. Hedgecock and Herman (1995) found that loss of lin-37 function in P₁, ABp, ABpl or ABpr did not cause a Muv phenotype and that loss in AB infrequently resulted in a Muv phenotype. Loss in AB results in many hyp7 nuclei lacking lin-37 function. Hedgecock and Herman (1995) postulated that lin-37 is also required in hyp7, but that a relatively low dosage is sufficient for wild-type function. Our results, which show a requirement for lin-36 function in only those lineages that generate the P(3-8).p cells (e.g. ABp but not P₁), suggest that lin-36 acts in the P(3-8).p cells and not in hyp7. This hypothesis is consistent with our lin-36 expression data.

Taken together, our findings and those of Herman and Hedgecock (1990) and Hedgecock and Herman (1995) suggest that some class B synMuv genes function in the P(3-8).p cells and others function in hyp7. Thus, the class B synMuv genes appear to encode components of an intercellular signaling system between hyp7 and P(3-8).p. Since lin-36 is expressed in the nucleus, it is likely that the inhibitory signal from the class B pathway is interpreted within the nucleus of the P(3-8).p cell.

The class B synMuv genes lin-35 and lin-53 encode proteins similar to the products of the mammalian retinoblastoma (Rb) tumor suppressor gene and the Rb-binding protein RbAp48, respectively (Lu and Horvitz, 1998). Both Rb and RbAp48 are believed to be involved in transcriptional regulation (Qian et al., 1993; Taya, 1997). lin-35 Rb appears to be expressed in the nuclei of the P(3-8).p cells (Lu and Horvitz, 1998). Since lin-36 is expressed in the nuclei of the P(3-8).p cells and seems to act in these cells, it is possible that lin-36 acts as a regulator or effector of lin-35 Rb activity.

How do lin-36 and the other class B synMuv genes antagonize the let-23 signal transduction cascade? Our observations suggest that the class B synMuv genes may inhibit the adoption of vulval cell fates by acting in the nucleus, possibly at the level of transcription. If so, the class B genes would provide an input into the cell-fate decision independently of the let-23 signal transduction pathway. The nuclear role of Rb-like proteins is consistent with this hypothesis.

SynMuv mutants are Vul in a strain in which let-23 receptor gene activity is reduced but are Muv in gonad-ablated animals. These observations indicate that whereas activation of the inductive signal transduction cascade by the gonadal ligand is not necessary for the adoption of vulval fates in the absence of negative regulation by lin-36 and the other synMuv genes, the presence of the intracellular signal transduction cascade nonetheless is necessary. These results are consistent with a model in which the two classes of synMuv genes antagonize either the gonadal ligand-independent (basal) activity of the let-23-mediated inductive pathway or a weaker activity stimulated by a second, nongonadal signal of the let-23-mediated pathway. In either case, a gonadal ligand-independent activity is necessary for the adoption of vulval fates by the P(3-8).p cells in a synMuv mutant. The inhibition of this activity by the synMuv genes can be overcome by the activation of the inductive pathway by the binding of the lin-3 ligand to receptor in the subset of P(3-8).p cells nearest the anchor cell.

lin-2, lin-7 and lin-10 positively regulate the inductive signal transduction cascade by localizing the LET-23 receptor to the P(3-8).p cell junctions (Hoskins et al., 1996; Simske et al., 1996). Our gene interaction studies showed that the synMuv phenotype is coexpressed with the Vul phenotype of lin-2, lin-7 and lin-10 in mutants with lin-36 and lin-15A or with lin-15AB. This coexpression suggests that the synMuv genes act to antagonize the let-23 pathway independently of the lin-2, lin-7, lin-10 regulatory system, since neither set of genes requires the presence of the other to exert its effect on vulval development. In the absence of the activity of the synMuv genes, the proper subcellular localization of LET-23 receptor is sufficient to allow all P(3-8).p cells to adopt vulval fates. However, in the absence of both synMuv gene-mediated negative regulation and proper subcellular localization of LET-23 receptor, the P(3-8).p cells variably adopt either vulval or nonvulval fates, despite the presence of the inductive signal.

Our laser ablation experiments showed that the P(3-8).p cell fates seen in animals lacking the stimulatory anchor cell signal, positive regulation by lin-7 and negative regulation by lin-36 and lin-15A, are similar to the fates seen in animals lacking just the positive regulatory pathway mediated by lin-2, lin-7, lin-10 and the negative regulatory pathway mediated by the synMuv genes. This observation suggests that the P(3-8).p cells that adopt vulval cell fates in animals lacking both synMuv activity and receptor localization do not adopt these fates in response to the anchor cell signal. We hypothesize that LET-23 receptor clustering can amplify the signal independently of ligand binding. Not all P(3-8).p cells adopt vulval cell fates, however. Unstimulated and unlocalized let-23 activity is sufficient only for some cells to adopt vulval fates in the absence of synMuv gene-mediated inhibition.

The data described above suggest the following model. The six Pn.p cells are predisposed to adopt vulval fates by an unstimulated basal activity of the let-23 receptor tyrosine kinase (Fig. 5A). In the absence of this activity, the P(3-8).p cells cannot adopt vulval fates. However, this predisposition to adopt vulval fates is inhibited by the action of the two functionally redundant signaling pathways encoded by the synMuv genes, which antagonize the basal activity of the let-23 signal transduction pathway. The inductive signal from the anchor cell overcomes this inhibition and causes nearby P(3-8).p cells to adopt vulval cell fates, either by overriding or by inhibiting the synMuv genes (Fig. 5B).

We thank Greg Beitel, Kerry Kornfeld and Shai Shaham for helpful discussions during the course of this work. We thank Xiaowei Lu, Craig Ceol and Rachel Hoang for comments concerning the manuscript, Beth James for technical assistance and Alex Hajnal and Stuart Kim for sharing unpublished data. We thank Stuart Kim and the Caenorhabditis Genetics Center (supported by the National Institutes of Health National Center for Research Resources) for providing some of the strains used in this work. J. H. T. was a Predoctoral Fellow of the Howard Hughes Medical Institute. H. R. H. is an Investigator of the Howard Hughes Medical Institute.