Cell fate patterning during C. elegans vulval development

The nematode Caenorhcibdilis elegcuis has been used to study the mechanisms that control cell fate determination reviewed by Greenwald and Rubin, 1992; Lambie and Kimble. 1991a). The development of C elegcuis follows a rgely invariant cell lineage (Kimble and Hirsh. 1979; Sulston and Horvitz. 1977; Sulston. 1976; Sulston et al., 1980; Sulston et al., 1983). Although some of the cell cages of C. elegcuis appear to be specified by cell-tonomous mechanisms, other lineages are specified by cell-cell interactions. Some of these interactions were first ted from instances of naturally occurring variation in die cage in which the development of a cell correlated with its position, thus suggesting that the cell was responding to. nvironmental cues (Sulston and Horvitz. 1977). Other examples of cell-cell interactions have been defined by periments in which groups of cells arc ablated by irradi-lon with a laser microbcam to sec if the development of c remaining cells is affected by the change in their envi-rnnent (Kimble, 1981; Priess and Thomson. 1987; Schnabel. 1991; Sulston et al., 1983; Sulston and White, 1980). C. elegans is well suited for genetic analysis Brenner. 1974) and many mutations that affect its cell cage have been identified (Horvitz. 1988). The genetic alysis of these mutations and the characterization of the. corresponding genes with the techniques of molecular lology has been a driving force in elucidating the mecha-isms that control cell fate in C. elegans.

The development of the hermaphrodite vulva has been extensively studied by cell ablation experiments and by genetic analysis. The results of these studies support a three signal model of vulval development summarized below (and reviewed by Horvitz and Sternberg. 1991) (Fig. I). The vulva normally develops from three of six ectodermal blast cells called Vulval Precursor Cells (VPCs) that are located within the ventral epidermis. Although in wiki type devel opinent each of the VPCs always assumes a particular fate, a variety of experiments indicate that each of the VPCs is equivalent in its ability to assume either of three fates and thus the VPCs constitute an equivalence group (Sternberg and Horvitz. 1986; Sulston and White. 1980; Thomas ct al.. 1990). Each fate consists of a distinct cell lineage that produces a particular set of cell types (Fig. 2). The 1 and 2° fates both produce vulval tissue although they contribute to different regions of the vulva. The 3° fate produces non-vulval epidermis (Sulston and Horvitz, 1977; Sulston. 1976). The fate chosen by a VPC depends upon three intercellular signals. The anchor cell of the somatic gonad produces a signal that induces the three more proximal VPCs to assume the 1 anil 2 fates (Kimble. 1981). An inhibitory signal inhibits the VPCs from assuming vulval fates in the absence of the inductive signal (Herman and Hedgecock. 1990). A lateral signal among the induced VPCs regulates the pattern of 1 and 2° fates (Sternberg. 1988a). In wild-type development the combined action of the three signals causes the six VPCs to assume a 3° 3° 2° 1° 2° 3°pattern of fates. In the follow ing sections, each of the three signals is discussed separately. This is then followed by a discussion of how the pattern of VPC fates is controlled by the three signals.

Cell ablation experiments indicate that the vulva develops as the result of an inductive interaction between the somatic gonad and the vulval precursors (Sulston and White, 1980). Kimble showed that the anchor cell of the somatic gonad is necessary and sufficient to induce vulva development (Kimble, 1981). If the anchor cell or its precursors are ablated, then the three VPCs that would normally assume the 1° and 2° fates instead assume the 3° fate (Table 1C). This induction occurs 1-3 hours before the VPCs divide (Kimble, 1981). If the anchor cell is ablated after the VPCs have divided, the cell lineage of their daughters is unaffected (Kimble, 1981; Sternberg and Horvitz, 1986). A large number of mutations that disrupt vulva induction have been characterized (reviewed by Horvitz and Sternberg, 1991). The cloning of several of these genes indicates that this induction is mediated by an Epidermal Growth Factor (EGF)-like signalling pathway that includes an EGF-like growth factor encoded by lin-3 (Hill and Sternberg, 1992) and an EGF-Receptor like molecule encoded by let-23 (Aroian et al., 1990).

Lin-3 is proposed to be a nematode member of the Epidermal Growth Factor (EGF) family of growth factors (Hill and Sternberg, 1992). This family includes ligands of the EGF-Receptor (EGF-R) such as EGF (Gregory, 1975; Savage et al., 1972), Transforming Growth Factor-alpha (TGF-a) (Derynck et al., 1984), Heparin Binding-EGF (HB-EGF) (Higashiyama et al., 1991), amphiregulin (Shoyab et al., 1989), and also ligands of ?ieu/HER2, a homologue of the EGF-R, (Holmes et al., 1992, Wen et al., 1992) and the predicted product of the spitz locus of Drosophila (Rutledge et al., 1992). The EGF growth factors are usually made as membrane spanning proteins that contain at least one extracellular EGF domain (reviewed by Carpenter and Wahl, 1990). EGF domains are sequence motifs of approximately 50 amino acids that consist of six cysteine residues with semi-conserved spacing (reviewed by Davis, 1990). In many cases it is known that the EGF domain can be processed away from the rest of the protein to produce a secreted factor that can activate its receptor (reviewed by Carpenter and Wahl, 1990). In the case of TGF-a, the EGF domain can also activate the receptor without being released from the membrane (reviewed by Massague, 1990).

The gene lin-3 was identified by reduction-of-function mutations that cause a recessive, vulvaless phenotype in which up to all three of the VPCs that usually assume vulval fates instead assume epidermal fates (Ferguson and Horvitz, 1985; Horvitz and Sulston, 1980; Sulston and Horvitz, 1981). Thus, wild-type lin-3 activity is required for vulval induction. In contrast, animals bearing lin-3 transgenes (transgenes made up of wild-type genomic DNA cloned from the lin-3 locus) can have a dominant multivulva phenotype in which up to all six of the VPCs assume vulval fates (Hill and Sternberg, 1992). The multivulva phenotype of the transgenes is proposed to be a gain-of-fu action phenotype because the transgenes are created by a method that concatenates hundreds of copies of the injected DNA into an extrachromosomal array (Mello et al., 1991). Such an array would likely express levels of Lin-3 protein that are greater than the levels expressed by the two chromosomal copies of the lin-3 locus. Together, the reduction-of-function and the gain-of-function phenotypes of lin-3 suggest that the dose of lin-3 activity controls the number of VPCs that assume vulval fates.

The analysis of lin-3 transgenes suggests that lin-3 can act in the anchor cell to cause the VPCs to assume vulval fates (Hill and Sternberg, 1992). A transgene in which the lacZ gene of E. coli is inserted in frame within a lin-3 genomic DNA clone directs expression Of p-galactosidase activity specifically in the anchor cell at the time of vulval induction (Fig. 3). This transgene, which should produce a fusion protein that contains the extracellular and transmembrane domains of the Lin-3 protein and a cytoplasmic domain consisting of the p-galactosidase protein, retains the ability to induce vulval fates. This result suggests that expression of lin-3 in the anchor cell is sufficient to induce vulval fates. This hypothesis is further substantiated by the results of ablation experiments. Ablation of the four gonadal precursor cells at hatching, which prevents the development of the anchor cell, greatly reduces (he ability of lin-3 transgenes to induce vulval development.

It has been proposed that the inductive signal is a secreted factor because induction of vulval fates can occur without apparent direct contact between the anchor cell and the induced VPCs (Sternberg and Horvitz., 1986; Sulston and White. 1980; Thomas et al., 1990). As mentioned previously. the EGF growth factors can produce secreted factors that consist of an EGF domain. We have made a transgene to lest whether a secreted form of the EGF domain of Lin-3. without the rest of the Lin-3 protein, would be sufficient to induce vulval fates (Hill and Sternberg, unpublished data) (Fig. 4). This transgene uses a heat shock promoter that should express the Lin-3 EGF domain in a tissue general manner (Stringham et al., 1992). The transgene is able to stimulate 1° and 2° vulval fates even when the gonadal precursors have been ablated prior to the development of the anchor cell. This result suggests that Lin-3 can act as a secreted factor in a manner similar to the EGF growth factors. Since overexpression of the EGF domain of Lin-3 is sufficient to induce vulval fates in the absence of the gonad, it is possible that Lin-3 is the only vulval inducing signal made specifically by the anchor cell. It is not yet known if the Lin-3 protein is normally processed in vivo.

A pathway of genes has been characterized that is believed to mediate the response of the VPCs to the inductive and inhibitory signals (Fig. 5). This pathway includes the following genes, which are believed to act in the following order on the basis of genetic epistasis experiments: let-23, which encodes a homologue of the EGF-Receptor (EGF-R) (Aroian et al., 1990); sem-5, which encodes an adapter protein that contains SI 12 and SI 13 domains (Clark et al., 1992); let-60, which encodes a ras protein (Han and Sternberg. 1990); and lin-45, which encodes a raf serinethreonine kinase (Han et al., 1993). Wild-type activity of each of these genes is required for vulval induction. A reduction of function mutation in let-23 is epistatic to the ability of lin-3 transgenes to stimulate vulval fates indicating that let-23 acts downstream of lin-3. (I lili and Sternberg. 1992). This pathway is similar to pathways studied in other experimental systems. For example, a ras protein, a raf kinase, and a Scm-5-like adapter protein also act downstream of the receptors encoded by the torso (N. Perrimon. this Volume) and sevenless (E. Hafen. this Volume) genes of Drosophila, and also downstream of growth factor receptors in mammals (McCormick. 1993).

A number of other genes have been implicated in the process of vulval induction. The genes lin-2. Un-7. and Un-10 are required for wild-type levels of vulval induction (Ferguson and Horvitz. 1985). These genes may act al a step near let-23 in the pathway of vulval induction (Ferguson et al. 1987). The gene Un-10 encodes a novel protein (Kim and Horvitz. 1990). The gene lin-l acts downstream of lin-45( 1 Ian et al., 1993). Wild-type lin-l activity acts lo repress vulval tales. (Ferguson and Horvitz. 1985; Ferguson et al., 1987).

lin-3 and let-23 are both required for vulval induction, larval viability (Aroian and Sternberg, 1991; Clark et al., .988; Ferguson and Horvitz, 1985; Herman. 1978. and R. Hill and P. Sternberg, unpublished results), hermaphrodite tertility (Aroian and Sternberg, 1991 ; Ferguson and Horvitz, 1985) and proper specification of cell fate in the male B lineage (11. Chamberlin and P. Sternberg, unpublished results). This suggests that the same growth factor and cceptor are used together in multiple decisions during the ievelopmcnt of C. elegans. let-23 is also required for proper cell fate specification in the P11-P12 equivalence group Aroian and Sternberg. 1991). but it has not been demon-crated whether lin-3 is also required for PI I-PI 2 fate spcc-ication. sem-5 (Clark et al., 1992), let-60 (Beilel et al., 990; Clark et al., 1988: Han et al., 1990), and lin-45 (Han et al., 1993) are also required for larval viability and a simple model is that they act downstream of let-23 in all tissues that ise the lin-3 let-23 pathway, lin-2. lin-7. and lin-10, arc nique in that they arc only required for vulval induction Ferguson and Horvitz, 1985). In contrast, sem-5 may be a generalized signal transduction component that acts downcream of multiple receptors, since sem-5 activity is required for the proper positioning of the sex myoblasts (Clark ct al.. 992), a process that does not require lin-3 or let-23.

Genetic experiments suggest that a pathway of genes that includes lin-l5 (Ferguson and Horvitz, 1985; Ferguson and Horvitz, 1989) produces an intercellular signal that inhibits vulval fates (Herman and Hedgecock, 1990). Loss-of-function mutations in the lin-l5 locus cause a multivulva phenotype in which all six VPCs can assume vulval fates (Table IE1). lin-15 mutant hermaphrodites display a multivulva phenotype even if the anchor cell has been ablated (Tabic IE2). Thus, lin-l5 does not act only in the anchor cell (Ferguson ct al., 1987). Genetic mosaic experiments indicate that lin-15 does not act cell autonomously in the VPCs. and therefore that Lin-15 is part of an intercellular signalling pathway (Herman and Hcdgecock. 1990). The overall pattern of mosaic phenotypes is complex and has been interpreted to indicate that lin-15 can act in the epidermal syncytium hyp7. a tissue generated by several different cell lineages (Herman and Hcdgecock. 1990). hyp7 is the main body epidermis and surrounds all of the VPCs. Thus a simple model is that the inhibitory signal would affect all of the VPCs equivalently and would thus provide no spatial specificity.

The inhibitory signal appears lo act in parallel to the inductive signal lo regulate the activity of the response pathway of the VPCs. Reduction-of-function mutations in the let-23 locus that confer a vulvaless phenotype are epistatic lo the multivulva phenotype of lin-15 mutations (Aroian and Sternberg. 1991; Ferguson et al., 1987; L. Huang and P. Sternberg, unpublished data). Therefore, Un-75, like the inductive signal, is an upstream regulator of let-23. The inductive and inhibitory signals can regulate vulval fates independently of each other suggesting that they act in parallel. First, in a mutant animal that lacks lin-15 activity, the cell P6.p will always assume the 1° fate if the anchor cell is present, but can assume the 2° or 1° fate if the anchor cell is absent (Table 1 El, 2). This indicates that the inductive signal can influence vulval fate choice independently of Un-15 (Sternberg, 1988a). Second, the multivulva phenotype of putative null mutations of lin-15 are epistatic to the vulvaless phenotype of strong reduction-of-function Un-3 genotypes. (Ferguson et al., 1987; R. Hill and P. Sternberg unpublished results). This suggests that the inhibitory signal can affect vulval fates independently of the inductive signal encoded by lin-3. The inhibitory signal does not appear to be a competitive inhibitor of the inductive signal since multicopy Lin-15 transgenes that presumably overexpress the Lin-15 protein do not interfere with vulval induction (L. Huang, and P. Sternberg, unpublished results). In wild-type development, a VPC that receives both the inductive signal and the inhibitory signal will assume a vulval fate. Thus, the inhibitory signal prevents the VPCs from assuming vulval fates only in the absence of the inductive signal. One possible interpretation of these observations is that the inhibitory signal negatively regulates the basal activity of let-23. The molecular mechanism by which this inhibitory pathway acts is not understood.

In addition to the inductive and inhibitory signals that originate from tissues other than the VPCs, a lateral signal among the VPCs also regulates VPC fate choice. Evidence for this signal came from observations on the pattern of VPC fates found in lin-15 mutant animals (Sternberg, 1988a). As described above, all six VPCs assume 1° and 2° vulval fates in a lin-15 mutant animal due to a defect in an inhibitory signal from the surrounding epidermal tissue. In Lin-15 mutant animals, it is commonly observed that two adjacent VPCs both assume the 2^° fate, but it is rarely observed that adjacent VPCs both assume the 1° fate (Table 1E1). This suggests the action of a signal that prevents adjacent VPCs from both assuming the 1° fate. The action of this signal can also be observed in experiments in which groups of VPCs are isolated in a lin-15 mutant animal by laser ablation of the other VPCs. A single isolated VPC in a Lin-15 mutant animal assumes the 1⁰ fate (Table 1E3). However, when two adjacent VPCs are left after surgery, one of them will assume the 1° fate and the other will assume the 2° fate (Table 1E4). The lateral signal acts only at a short range and may require direct cell contact since two isolated VPCs in a lin-15 mutant are only inhibited from both assuming the 1° fate if they are close to each other (Table 1E 5).

This lateral signal is believed to act through the gene Un-12 (Sternberg and Horvitz, 1989). Un-12 encodes a cell surface protein homologous to the protein encoded by the Notch locus of Drosophila (Greenwald, 1985; Yochem et al., 1988). Both Un-12 and Notch function in lateral interactions that control cell fate choice between cells of equivalent developmental potential (Greenwald et al., 1983; Greenwald and Rubin, 1992; see P. Simpson, this volume). lin-12 acts during vulval development to promote the 2° fate. In animals homozygous for lin-12 loss-of-function mutations, no VPCs assume the 2° fate, and in animals homozygous for lin-12 gain-of-function mutations, all six VPCs assume the 2° fate independently of the inductive signal (Table IF; Greenwald et al., 1983; Sternberg and Horvitz, 1989). lin-12 also promotes the 2° fate in the AC (1° fate)-Ventral Uterine (VU) precursor (2° fate) equivalence group (Greenwald et al., 1983). Genetic mosaic analysis of Un-12 activity in the AC-VU decision indicates that Un-12 acts in a cell autonomous manner to specify the 2° fate (Seydoux and Greenwald, 1989). It has thus been predicted that the Un-12 product acts as the receptor for the lateral signal during vulval development.

Lateral signalling acts through a separate pathway from the inductive pathway. It is not clear precisely where the two pathways converge to control VPC fate although it is likely downstream of I in-2, I in-7, and Un-10 (Sternberg and Horvitz, 1989). Currently other components of the lateral pathway, including the ligand for lin-12 and genes that act downstream of Un-12, are not well characterized. One explanation for this is that these other components of the lateral pathway may be required for an essential decision prior to the time of vulval development. The C. elegans gene glp-1 is structurally very similar to Un-12 and both genes may have originated from a gene duplication of a common ancestor (Yochem and Greenwald, 1989). glp-1 is required zygotically for maintenance of the mitotic germline (Austin and Kimble, 1987) and is required maternally for the proper fate specification of early blastomeres (Priess et al., 1987). The lin-12 glp-1 double mutant has a synthetic zygotic lethal phenotype called Lag (Lambie and Kimble, 1991b). This indicates that glp-1 and Un-12 are functionally redundant for certain essential developmental processes. Other genes that can be mutated to give a Lag phenotype could be shared components of the lin-12 and glp-1 pathways and thus might be required for lateral interactions during vulval development (Lambie and Kimble, 1991b).

Vulval development allows the examination of how different intercellular signals interact to control pattern formation. During vulval development, three intercellular signals direct the six VPCs to adopt the pattern of fates 3° 3° 2° 1° 2° 3°. Mutations that disrupt any of the three signals affects the fates assumed by the VPCs. The wild-type pattern is precise in the number of VPCs induced to assume the 1° and 2° vulval fates, the position of the VPCs induced to assume vulval fates, and in the pattern of 1° and 2^° fates assumed by the induced cells. This section summarizes our knowledge of what signals control the extent, location, and pattern of vulval induction.

As mentioned previously, the genetic dose of lin-3 is proposed to control the number of VPCs induced to assume vulval fates. This hypothesis is based on the observations that reduction-of-function mutations in lin-3 reduce the number of the VPCs that assume vulval fates and that multicopy lin-3 transgenes, which presumably overexpress the Lin-3 protein, increase the number of VPCs that assume vulval fates (Ferguson and Horvitz. 1985: Hill and Sternberg. 1992). In lin-12 mutant animals in which lateral signalling is disrupted, the number of VPCs that assume vulval fates is near the wild-type level (Table I Fl ; Sternberg and Horvitz. 1989). Thus the lateral signal does not appear to regulate the number of VPCs that assume vulval fates. The inhibitory signal acts in a spatially general manner to inhibit vulval fates, but the action of the inhibitory signal is overridden by the inductive signal. Thus the inhibitory signal ensures that vulval fates are assumed only in response to the inductive signal, but does not itself determine the number of VPCs that assume vulval fates.

The anchor cell controls the location of vulval development. I his is illustrated by the pattern of fates assumed by the VPCs in dig-1 mutant animals. In dig-1 mutants the position of the gonad and (he anchor cell can be shifted anteriorly and/or dorsally of their wild-type positions (Thomas et al., 1990). In animals in which the anchor cell is shifted anteriorly of its wild-type position and remains ventral, the pattern of vulval fates is shifted commensurably. For example, when the anchor cell is positioned over P5.p. P5.p assumes the 1° fate and the overall pattern ol vulval fates is 3° 2° 1 °2° 3° 3° (Table IG).

As just discussed, in wild-type development, the induction 4 1° and 2° vulval fates is limited to P5.p. P6.p and P7.p because the anchor cell is located above P6.p and because it only makes enough inductive signal to induce the three most proximal VPCs. What signals direct the three induced VPCs to assume a 2° 1° 2° pattern of fates? The experimental analysis of animals that have only one VPC indicates that the inductive signal plays an important role in patterning. The fate of a single, isolated VPC correlates with its iistance from the anchor cell: a single VPC that is relatively. lose to the anchor cell will assume the I fate, a single VPC hat is further away will assume (he 2° fate, and a single VPC that is distant will assume the 3° fate (Table Hl) Sternberg and Horvitz. 1986; P. Sternberg, unpublished csults. and M. Herman, and II. R. Horvitz. personal communication). Since these isolated VPCs have no neighborrig VPCs. they should not be subject to (he effects of (he .itérai signal. Thus, the tael (hat the fate of an isolated VPC. prelates with its distance from the anchor cell suggests that ere is a gradient of inductive signal centered on the anchor .1. and that the VPCs assume different fates in response to afferent levels of the inductive signal.

The inhibitory signal mediated through the lin-15 pathway does not have an important role in vulval fate patterning. Although all six VPCs assume vulval fates in lin-15 mutant animals, the patterning of fates is still normal. In a lin-15 mutant animal, the inductive signal still promotes P6.p to assume the 1° fate and (he lateral signal still functions to prevent P5.p and P7.p from assuming the 1° fate. Thus if the anchor cell is present in a lin-15 mutant tmal. then P5.p. P6.p and P7.p. will assume the same 2° 1° 2° pattern of fates that (hey assume in wild-type animals (Table IEI). However, if the anchor cell is ablated, then P5.p. P6.p, and P7.p can each assume cither (he 1 ° or 2° fate (Table 1E2) (Sternberg. 1988a).

The lateral signal appears to play an important role in the patterning of vulval fates assumed by the induced VPCs. In lin-12 loss-of-function mutant animals, the pattern of fates is usually 3° 3° 1° 1° 1° 3° (Table 1F1). Thus, lateral signalling may be necessary to specify the 2° fate. Moreover, the lateral signal is believed to measure differences between two VPCs in the amount of inductive signal received and to act in response to this difference lo direct the VPCs to assume different laics. This model is supported by indirect evidence that certain levels of the inductive signal can specify a VPC to have different fates depending upon the states of the VPCs neighbors. For example, a VPC in the position of P5.p always assumes the 2° fate in wild-type development. But an isolated VPC in this position has been observed to assume the I late in animals in which the other VPCs have been ablated (P. Sternberg, unpublished observations). This isolated VPC may assume (he I fate for cither of two reasons; it might be receiving more inductive signal than it docs in wild type because it lacks neighbors that absorb the inductive signal, or it may receive the same amount of inductive signal as it does in wild type, but lacks lateral signalling from its neighbors that specify it to be 2°. Another example is the pattern of vulval fates in dig-1 mutant animals in which the anchor cell is displaced to the dorsal side of the animal (Thomas et al., 1990). In these animals the VPC closest to the anchor cell is at a distance equal lo or greater than the distance of P5.p is lo the anchor cell in a wild-type animal. Thus, no VPC in these dig-1 mutant animals should receive more inductive signal then does P5.p in a wild-type animal. Yet. although P5.p always assumes the 2 fate in wild-type development. VPCs al an equivalent distance from the anchor cell in dig-1 initiants often assume the 1° fate. Thus it is possible that there arc intermediate doses of the inductive signal that will specify a VPC to have the I fate in the context of neighboring VPCs that arc receiving less inductive signal, but that will specify a VPC lo have the 2° fate in the context of neighboring VPCs that are receiving more inductive signal.

The action of the lateral signal in vulva development is also modeled on the role of lin-12 in the AC-VU equivalence group (Greenwald et al., 1983). In this equivalence group, either the cell Zl.ppp or Z4.aaa can assume the fate of the anchor cell with approximately equal likelihood. Lateral signalling among these two cells mediates this decision without apparent influence from outside cells (Kimble. 1981; Seydoux and Greenwald. 1989). Although either cell can assume cither fate, the lateral signal ensures that only one cell always assumes the fate of (he anchor cell and that the other cell always assumes the fate of VU(Kimble and Hirsh. 1979). Based upon this observation, it has been proposed that the lateral signal acts in a feedback loop that establishes mutually exclusive stales in adjacent cells. In this model both cells start with the ability to produce both the lateral signal and the receptor for the lateral signal. Cells that receive the lateral signal down-regulate the activity of the lateral signal and up-regulale the activity of the receptor. Conversely, cells that produce the lateral signal up-regulate the activity of the lateral signal and down-regulate the activity of the receptor. This feedback mechanism should result in only one cell producing active lateral signal and the other cell producing only active receptor (Seydoux and Greenwald, 1989; Sternberg, 1988b). In vulval development, the outcome of lateral signalling among the VPCs is invariant and is believed to be biased by the inductive signal produced by the anchor cell. It is proposed that the lateral signalling initially acts in proportion to the amount of inductive signal received by each VPC. The feedback mechanisms of the lateral signal then act to accentuate the differences in the amount of lateral signalling between the VPCs. This ensures that adjacent cells assume different fates in response to different levels of the inductive signal (Sternberg and Horvitz, 1989).

Another way of comparing the relative contribution of the inductive signal and the lateral signal in controlling pattern formation is by examining the genetic interactions of the two signals. In animals with lin-12 gain-of function mutations, all six VPCs usually assume the 2° fate. There is also usually no anchor cell in these animals and thus no inductive signal (Table 1F2). However, in those lin-12 gain-of-function mutant animals that have an anchor cell, P6.p always assumes the 1° fate (Table 1F3) (Sternberg and Horvitz, 1989). Thus, for a single cell, the action of the inductive signal in specifying the 1° fate overrides the action of the lateral signal in specifying the 2° fate. A function of the lateral signal in patterning though, is to prevent two adjacent VPCs from both assuming the 1° fate. Thus an important experiment is to determine if the lateral signal can still promote adjacent cells to assume different fates when they both receive high levels of the inductive signal. In lin-3 transgenic animals, which overexpress the inductive signal in the anchor cell, adjacent VPCs often both have the 1 ° fate (Table II) (Hill and Sternberg, unpublished data). In this situation, two adjacent VPCs both assume the 1° fate, even though each of them should be laterally signalling their neighbors to be 2°. Thus, either a high level of the inductive signal specifies the 1° fate and overrides the action of the lateral signal, or, alternatively, Lin-3 protein might bind to, and inhibit the action of either the ligand or the receptor of the lateral signal. In either case, the dose of the inductive signal must be limited to achieve wild-type patterning in which P5.p and P7.p assume the 2° fate.

The pattern of 1° and 2° fates assumed by the VPCs is determined both by the action of an inductive signal encoded by lin-3, and a lateral signal that acts though lin-12. The inductive signal appears to provide the specific information that sets up the pattern of fates assumed by the VPCs. First, the inductive signal originates only from the anchor cell and is the only signal that clearly provides spatially specific information. Second, the fact that the fate of an isolated VPC correlates with its distance from the anchor cell, suggests that different doses of the inductive signal promote the 1°, 2°, and 3° fates, and that a spatial gradient of the inductive signal promotes the VPCs to assume a graded pattern of fates. Third, a high level of the inductive signal specifies the 1° fate and overrides the action of the lateral signal in specifying the 2° fate. This indicates that the production of the inductive signal must be limited to get VPCs with the 2° fate. The properties of the lateral signal and lin-12 are intriguing. On the one hand the lateral signal seems to reinforce a pattern that has already been determined by the inductive signal; i. e. the lateral signal ensures that P5.p and P7.p assume the 2° fate in response to intermediate amounts of the inductive signal, and that P6.p assumes the 1° fate in response to high amounts of the inductive signal. On the other hand, wild-type activity of the gene lin-12, which is believed to encode the receptor for the lateral signal, is required to specify the 2° fate. Thus a VPC that receives a dose of the inductive signal that promotes the 2° fate still apparently requires input from the lateral signal to assume the 2° fate. Thus, the mechanism by which an isolated VPC assumes the 2° fate is unclear, since it lacks neighbors with which it can laterally signal. Such a VPC may engage in autocrine signalling through Un-12 to assume the 2° fate. The inductive signal and the lateral signal appear to cooperate to promote the graded 2° 1° 2° pattern of fates assumed by the induced VPCs. The use of two signals to promote this pattern may ensure that the process of pattern formation has a reproducible outcome. Thus for example, the lateral signal could ensure that the VPCs assume a 2° 1 ° 2° pattern of fates over a broad dose range of the inductive signal, as long as the inductive signal is distributed in a gradient.

Our current model of VPC fate determination is summarized in Fig. 1. An inhibitory signal that arises in a spatially general manner inhibits the VPCs from assuming vulval fates in the absence of the inductive signal. Localized production of the inductive signal encoded by lin-3 controls the number and location of VPCs that assume the vulval fates. Lateral signalling among the VPCs acts in response to a spatial gradient of inductive signal to specify P5.p and P7.p to assume the 2° fate.

A question for the future is how the different signals interact to control the fate of the VPCs. The molecular cloning of several key genes in these intercellular pathways means that the interactions of these signals can be analyzed both genetically and by the techniques of molecular biology. The molecular mechanism by which the inhibitory signal represses the response pathway of the VPCs has not been established. Vulva development provides an example where the outcome of lateral interactions among cells of equivalent developmental potential appears to be controlled by the action of an outside signal. Elucidating the mechanisms by which the lateral signal and inductive signal interact to control the choice between 1⁰ and 2° will be a goal of future research. The use of genetics to identify additional components of the lateral signalling pathway will be important in understanding the mechanism of action of the lateral signal. Genetic screens for suppressers of existing mutations that disrupt vulval development should be useful in identifying new genes involved in the pathways of vulval fate determination.