GENETIC CONTROL OF CELL INTERACTIONS IN NEMATODE DEVELOPMENT

Annu. Rev. Genet. 1991. 25. 411-36 Copyright 199l by Annual Reviews Inc. All rights reserved GENETIC CONTROL OF CELL INTERACTIONS IN NEMATODE DEVELOPMENT E. J. Lambie and Judith Kimble Laboratory of Molecular Biology and Department of Biochemistry, University of Wisconsin, Madison, Wisconsin 53706 KEY WORDS: Caenorhabditis elegans, embryogenesis, induction, lateral signaling, signal transduction CONTENTS INTRODUCTION... 411 SUMMARY OF DEVELOPMENT OF C. ELEGANS... 412 CELL INTERACTIONS... 413 Identification of cell interactions... 413 Distal tip cell control of germline proliferation... 414 Cell interactions that control vulval development... 415 Interactions in the lateral hypodermis... 417 Other cell interactions... 418 GENETICONTROL OF CELL INTERACTIONS... 419 glp-1 and lin-12... 419 Genes regulating the VH/SH decision... 426 pal-1... 429 Other genetic pathways... 430 CONCLUSIONS, SPECULATIONS AND PROSPECTS... 431 INTRODUCTION Cell interactions arc vital to the regulation of cell division, differentiation, and pattern formation during the development of most multicellular creatures. The nematode Caenorhabditis elegans is no exception to this rule. Like other 0066-4197/91/1215-0411502.00 411

412 LAMBIE & KIMBLE metazoans, this small worm relies heavily on cell interactions during development. With the powerful genetics of C. elegans and its relatively simple anatomy, researchers have been able to delineate the genetic mechanisms regulating cell interactions with an unparalleled degree of precision. In addition, as the analysis of cell fate regulation in C. elegans has advanced to the molecular level, it has merged to a remarkable degree with the rapidly expanding field of signal transduction biochemistry, which has been intensively studied in vertebrate cells. The conjunction of these two areas of research is certain to enhance our understanding of the mechanisms, regulation, and evolution of cell interactions during development. In this review, we emphasize those interactions in C. elegans that have been best characterized genetically. These include examples of both induction and lateral signaling. Induction occurs between separate tissues, with one tissue regulating the fate of another, whereas lateral signaling occurs within a single tissue and results in the adoption of distinct fates by cells with equivalent developmental potential. In addition to our discussion of the genetics of these better characterized interactions, we briefly mention other cell interactions and other genetic controls that promise to provide special insight into how cell interactions influence cell fate. SUMMARY OF DEVELOPMENT OF C. ELEGANS Figure 1 diagrams the essential features of the development and anatomy of C. elegans. The development of C. elegans, which unfolds in about 3 1/2 days, is known at the level of individual cells. For a detailed account of the development and anatomy of C. elegans see refs. (1, 16, 32, 44, 49, 74, 76, 77, 79); for extensive reviews see refs. (47, 75, 87, 88). During early embryogenesis, the zygote generates five somatic founder cells (AB, MS, E, C, and D) and one germline founder cell P4 (Figure 1A). Most founder cells contribute to several tissues; moreover, most major tissues (e.g. hypodermis, nerve, muscle) arise from several founder cells. In contrast, the gut and germ line are each generated by clonal divisions of a single founder cell, E and P4, respectively (Figure 1A). In the first half embryogenesis, the somatic founder cells generate 559 descendants by a fixed cell. lineage, whereas P4 divides only once, producing two germline precursor cells. The second half of embryogenesis is characterized by extensive differentiation and morphogenesis, resulting in a tiny wormabout 150/zm long. After hatching, the animal passes through four larval stages (L1, L2, L3, L4) to reach adulthood. During this period of postembryonic development, the somatic and germline tissues grow and mature sexually. Somatic development occurs by a nearly invariant lineage. Typically, each cell in C. elegans has a fate that is coupled to its position and ancestry. By cell fate, we mean either differentiation as a specific cell type (e.g. a neuron) or execution of

NEMATODE CELL INTERACTIONS 413 A ZYGOTE [ 1 hypodermis anterior pharynxmuscle neurons I MS muscle posterior pharynx somatic gonad B I E intestine I I hypodermis neurons muscle D muscle P4 germ line distal tip _ ce s [syncytiol hypodermis mrn I inure " anterior ovoteshs /.^,..^ bogy,, ~neur ons pharynx intestine somatic vulva differentiating gonoo gametes Figure 1 A) An outline of the embryonicell lineage of C. elegans. Beneath the name of each blast cell is a list of the tissues it generates. B) Major anatomical features of the adult C. elegans hermaphrodite. particular pattern of divisions, generating a fixed number of cells with specific fates. In contrast to somatic development, the pattern of cell divisions in the germ line is variable. CELL INTERACTIONS In this section, we first describe methods by which cell interactions have been discovered, then discuss selected interactions that exemplify the types of interactions known to influence cell fate during development, and, finally, mention other interactions that are generally less well characterized. Identification of Cell Interactions Most cell interactions have been identified by observing the consequences of altering a cell s environment during development. Three different experimental approaches have been developed. First, individual cells can be onus

414 LAMBIE & KIMBLE killed. In larvae, laser microsurgery is extremely effective in eliminating cells with little or no damage to neighboring cells (80). However, in embryos laser ablation has been less effective (79). Yet early embryonic blastomeres have been successfully removed by extrusion through a punctured eggshell (60, 65) and the laser has been used to arrest cells in the embryo rather than killing them (66). After a cell is killed/removed, the fates of its neighbors are observed. If abnormalities are seen, it is likely that the removed cell influences the development of its neighbor(s). Second, the relative positions blastomeres in the early embryo can be changed by micromanipulation (60, 89). Here again, if the predicted development is changed, the fates of the blastomeres are likely to be regulated by cell interactions. Third, some interactions have been discovered by the analysis of genetic mosaics (reviewed in ref. 30). In these experiments, an animal is generated with one group of wild-type cells and a second group of mutant cells. If the fate of a cell is affected by the genotype of its neighbor, these cells probably interact. These experiments have clearly demonstrated that cell interactions are critical to development in C. elegans from early embryogenesis through adulthood. Distal Tip Cell Control of Germline Proliferation The control of germline proliferation by the distal tip celt is an example of induction. The adult germ line is a syncytial tissue consisting of about 2500 P4 descendants in hermaphrodites and about 1000 in males (32, 49). In hermaphrodites, the germline tissue is split into two ovotestes, each with a single somatic cell, called the distal tip cell, located at its apex (Figure 1B; 2. See also 44). In males, the germ line occupies a single testis with two distal tip cells at the end. Induction of the germline from two cells at hatching to the one or two thousand descendants in adults depends on the presence of the distal tip cell (Figure 2; 48). In either sex, if the distal tip cells are killed during larval development or adulthood, all germline nuclei that are in mitosis enter meiosis. In hermaphrodites, if a single distal tip cell is killed, germline mitoses are arrested only in the ovotestis to which that distal tip cell belongs. Therefore, the presence of the distal tip cell is essential for the larval mitotic divisions that generate the germline tissue and for the adult divisions that maintain a stem cell population in the mature gonad. A simple model to explain this phenomenon is that the distal tip cell signals germline nuclei to remain in mitosis. Germline proliferation can also be stimulated by a second somatic gonadal cell, the anchor cell (68). Normally the anchor cell is separated from the germ line by other somatic gonadal cells. However, if these other cells are removed, the anchor cell induces germline mitoses in the proximal region of the ovotestis. This induction differs from the distal tip cell induction in that the continued presence of the anchor cell is not required to maintain proximal mitoses (68).

NEMATODE CELL INTERACTIONS 415 A INTACT B DTC REMOVED Figure 2 A) The distal tip cell promotes mitosis (MI) in neighboring germline nuclei. B) animals in which the distal tip cell has been removed by laser microsurgery, all germline nuclei enter meiosis (ME). Cell Interactions that Control Vulval Development The hermaphrodite vulva is an opening in the body wall through which insemination occurs and embryos are extruded (Figure 1B). The vulva is specialization of the body wall epithelium, or hypodermis, that is controlled by specific nerves and muscles. Here we limit our discussion to those interactions required for development of the vulval hypodermis. These include both inductive and lateral signaling interactions. Recent reviews provide more comprehensive summaries of vulval development and its genetic control (34, 71). VULVAL INDUCTION The bulk of the hypodermis is syncytial, consisting of numerous nuclei in a common cytoplasm (87). However, in young larvae, six cells in the ventral hypodermis remain separate. These six cells are the vulval precursor cells (or VPCs). The vulva is produced by three of the vulval precursor cells, which generate a total of 22 vulval hypodermal (VH) descendants (77). One of the vulval precursor cells follows the VH1 fate, producing eight vulval hypodermal cells, and the two flanking vulval precursor cells each follow the VH2 fate, generating seven more vulval hypodermal cells on each side (Figure 3). The remaining three vulval precursor cells divide once and their descendants join the syncytial hypoderm (SH) (Figure Induction of the vulval hypodermal fates (VH1 and VH2) depends on the presence of a single cell, the anchor cell, in the somatic gonad (43). If the anchor cell is killed, all vulval precursor cells follow the SH fate and join the syncytial hypodermis (Figure 3B). Furthermore, if all somatic gonadal cells except the anchor cell are killed, vulval induction occurs normally (43). Therefore, the anchor cell regulates the VH/SH decision. This inductive interaction occurs just prior to the first division of the vulval precursor cells in the early L3 stage (43).

416 LAMBIE & KIMBLE A INTACT T T T,,,T B AC REMOVED T T T T T T Figure 3 A) Lateral signaling occurs between the members of the anchor cell (AC)/ventral uterine precursor cell (VU) equivalence group and the members of the vulval equivalence groups. The AC induces adjacent vulval equivalence group members to adopt vulval hypodermal (VH 1, VH2) fates, whereas the syncytial hypoderm (SH) is shown as inhibiting vulval hypodermal fates. B) When both members of the AC/VU equivalence group are removed by laser microsurgery, none of the members of the vulval equivalence group are induced to adopt vulval hypodermal fates. Two lines of evidence support the idea that the anchor cell induces vulval development at a distance. First, the anchor cell is normally positioned directly over the VH1 cell, but at some distance from the two VH2 cells. Since the anchor cell can induce VH2 at a distance, even when the cell normally adopting the VH 1 fate is not present (72, 81), it seems likely that the anchor cell induces both VHI and VH2 fates directly. Second, the anchor cell can induce vulva formation when positioned on the opposite side of the animal. In dig-1 (for displaced gonad) mutants, vulval induction occurs even though the somatic gonad is located on the dorsal side of the animal (81). The most likely mechanism for anchor cell control is production of a diffusible signal. Yet it remains possible that the anchor cell extends a cellular process to induce VH fates at a distance or that it can act via the syncytial hypoderm. Identification of the anchor cell signal and electron microscopic analysis of the cellular contacts between the anchor cell and vulval precursor cells should distinguish among these possible mechanisms for action at a distance. Based on a genetic mosaic analysis of lin-15, a gene essential to the VH/SH decision (see below), Herman & Hedgecock have proposed that the syncytial hypoderm surrounding the vulval precursor cells promotes the SH fate and/or inhibits the VH fates in the vulval precursor cells (31). This interaction was previously unknown, presumably because ablation of the syncytial hypoderm

NEMATODE CELL INTERACTIONS 417 that surrounds the vulval precursor cells would cause lethality. Therefore, the vulval precursor cells appear to make the VH/SH decision by integrating the effects of two antagonistic inductive interactions, one with the anchor cell and the other with the surrounding hypoderm. LATERAL SIGNALING AND VULVAL DEVELOPMENT The correct specification of cell fates in the vulval hypodermis is dependent on lateral signaling within two sets of equivalent cells. Such arrays of cells with equivalent developmental potential are known as equivalence groups (46). The first group that we consider is the two-celled AC/VU equivalence group, located in the somatic gonad (Figure 3). Either member of this group can differentiate an anchor cell (AC) or a ventral uterine precursor cell (VU)(43, 44). described above, the anchor cell induces vulval development in the hypodermis. The VU cell, on the other hand, generates ten uterine cells (44). If either cell of the AC/VU pair is killed, the remaining cell becomes an anchor cell (43, 67). Therefore, the AC fate is known as primary. The fate is followed only when an anchor cell is present and therefore is known as secondary. If both cells of the AC/VU pair are left intact, one always becomes an anchor cell and the other always adopts the VU fate. One model to explain these results is that the anchor cell influences its neighbor to assume the VU fate, either by inhibiting the AC or promoting the VU fate. A second model is that the two cells of the AC/VU pair compete for some site or factor that promotes anchor cell differentiation. The first model is favored by the observation that both members of the AC/VU pair can assume the AC fate if the close association of the two ceils is disrupted (28). The six vulval precursor cells constitute the second equivalence group required for vulval development (Figure 3). As described above ~, the VH/SH decision among the vulval precursor cells depends on an inductive interaction with the anchor cell. However, each of the six vulval precursor cells follows one of three fates: VH1, VH2, or SH. These fates, like the AC and VU fates, can be placed into a hierarchy by laser ablation experiments: VH1 is primary, VH2 secondary, and SH tertiary (72, 80). A simple model to explain these results is that the cell determined to follow the VH1 fate signals its neighbors to adopt a different fate and, similarly, that cells determined to follow the VH2 fate signal their neighbors. However, competition for sites or factors that promote specific fates remains a formal possibility. Genetic evidence supports the idea of an interaction between VH1 and VH2, but not between VH2 and SH (73). Therefore, only one lateral signal, between VHI and its neighbors, is diagrammed in Figure 3. Interactions in the Lateral Hypodermis In the L1 larva, there are seven lateral hypodermal cells distributed along the antero-posterior axis on each side of the animal (77). In males, the two most

418 LAMBIE & KIMBLE posterior cells (V6 and T) give rise to sensory structures known as rays, while their anterior counterparts (V1-V4) generate seam cells (hypodermal cells that produce cuticular ridges called alae). The cell that lies between these two groups (V5) gives rise to both a seam cell and a ray. If T is killed, the lineages of V1-V6 are unaffected; however, if either V5 or V6 is killed, its anterior neighbor will shift to occupy the vacant position and assume the fate of the ablated cell (80). The decision to produce either rays or seam cells appears be controlled by a combination of the position of a cell along the anteroposterior axis and signaling between neighboring cells (38, 85; see section below on pal-l). Other Interactions Other interactions in addition to those discussed above dramatically affect development in C. elegans. A detailed discussion of these interactions is beyond the scope of this review. However, we mention them briefly here and provide references for interested readers. EMBRYONIC INDUCTION OF ASYMMETRY Recent experiments involving micromanipulation and laser irradiation of blastomcres have demonstrated that cell interactions determine left-right asymmetry (89, 90) and possibly govern the establishment of the dorso-ventral axis (66). EMBRYONIC INDUCTION OF THE ANTERIOR PHARYNX The pharynx is a neuromuscular organ that pumps food into the intestine (Figure 1B). The anterior pharynxderives primarily from the AB blastomere and the posterior pharynx.from EMS (Figure 1A; 79). If EMS is physically removed, AB longer produces pharyngeal tissue (60). Therefore, EMS or one of its descendants is likely to induce pharyngeal development in AB or one of its descendants. Experiments involving the removal of EMS or its descendants at various times have demonstrated that this interaction must occur between the 4- and 28-cell stages of embryogenesis (60). OTHER EQUIVALENCE GROUPS Besides the AC/VU and vulval equivalence groups described above, several other equivalence groups have been revealed by laser ablation experiments (for review, see ref. 75). In each case, hierarchy of fates has been established. Of particular note are equivalence groups that are affected by the genes discussed below. These include the G2/W.pair, which generates one ectoblast and one neuroblast, the P1 l/p12 pair, which produces two different ectoblasts, two equivalence groups in the descendants of the B precursor cell that produce parts of the male copulatory apparatus (Bc~/Bfl and B,//BtS), and a three-celled equivalence group (P9.p, P10.p, and P11.p) located in the ventral hypodermis of the male, which also is responsible for generating specific cells of the male copulatory apparatus.

NEMATODE CELL INTERACTIONS 419 MIGRATIONS Although most cells in C. elegans stay where they are born, some cells migrate long distances to adopt new positions within the animal. For example, the somatic gonadal precursor cells migrate from their place of birth in the head to a ventral position midway along the animal s anteroposterior axis (79). Similarly, the sex myoblasts migrate from the posterior region of the ani~nal to the gonadal region, where they generate vulval and uterine muscles (77). Evidence is beginning to accumulate in support long-range signaling as the mechanism that may direct these migrations (81). In addition, axonal outgrowth and branching are dependent upon cell-cell and cell-extracellular matrix interactions (28, 53, 81). Mutants have been identified in which these interactions are likely to be perturbed. (For a recent review, see (29).) CELL DEATH During male development, the programmed death of two different cells depends on specific interactions with an engulfing cell (76). This type of cell death differs from most other cell deaths in C. elegans. (For review, see (17).) OTHER TYPES OF CELL INTERACTIONS Some cells react to the deaths of their neighbors by altering their own morphology, generating additional descendants, or reversing their lineage polarity (75). A proliferative response has been observed specifically in the lateral hypodermis (80) and changes lineage polarity have been seen in isolated cells of various tissues (11, 43, 80). An understanding of these "other" interactions requires a detailed description of the cellular morphology involved; since these changes remain largely unexplored genetically, we do not consider them further. GENETIC CONTROL OF CELL INTERACTIONS Among the genes known to regulate cell interactions in C. elegans, we focus on those that mediate induction and lateral signaling. First, we discuss glp-1 and lin-12, homologous genes that mediate germline induction and lateral signaling, respectively. Second, we discuss genes that mediate the VH/SH decision during vulval development. Third, we describe the pal-1 gene and its role in silencing an interaction. Finally, we mention two intensively analyzed genetic pathways for cell fate regulation that have recently become candidates for genetic models of signal transduction. glp- 1 and lin- 12 Two genes, glp-1 and lin-12, control several distinct cell interactions in C. elegans. The glp-1 gene mediates germline induction by the distal tip cell and induction of the anterior pharynx during embryogenesis (5, 59). In contrast, the lin-12 gene primarily regulates cell fate decisions among members of

420 LAMBIE & KIMBLE equivalence groups (e.g. the AC/VU and VH1/VH2 decisions) (22). These two genes are discussed together here, because it has become evident that the glp-1 and lin-12 proteins are remarkably similar in both structure and function (5, 51, 55a, 68, 91). MUTANT PHENOTYPE OF glp-1 Null mutations of glp-1, glp-l(o), severely curtail the mitotic proliferation of the germ line (4, 60). In glp-l(o) mutants, the two germline precursor cells present at hatching undergo only one or two mitotic divisions and then enter meiosis. Yet, the distal tip cell is present and meiosis and gametogenesis appear to be unaffected. Therefore, the glp-1 product is required in the germ line for the decision between mitosis and meiosis. One class of glp-l(ts) mutants has a germ line virtually identical to that of a null mutant when raised at restrictive temperature. However, when glp-l(ts) mutants are raised at permissive temperature and then shifted to restrictive temperature as adults, a second phenotype, embryonic lethality, is seen (4, 60). The cellular defects observed in the dead embryos indicate that glp-1 is also required for the embryonic induction of the anterior pharynx and the development of the embryonic hypodermis (60). The expression of glp-1 is required zygotically for germline development and maternally for embryogenesis (4, 60). The maternal requircment for glp-1 is absolute: a wild-type allele introduced by mating does not rescue any aspect of the glp-1 embryonic lethality. Five glp-1 mutations have unusual phenotypes (4, 55a, 60; V. Kodoyianni, E. M. Maine, J. Kimble, manuscript in preparation). Four have more severe effects on the embryo than the germ line. These mutations may identify regions of glp-1 specific for embryonic functions. Alternatively, the embryo may be more sensitive than the germ line to the level of glp-1 activity, in which case these alleles could simply be weak loss-of-function mutations. The other unusual mutation, glp-1(q35), causes both a recessive Glp and a dominant Muv phenotype; this allele is discussed further below. MUTANT PHENOTYPE OF lin-12 Two classes of lin-12 mutations exist: dominant gain-of-function mutations, lin-12(d), and recessive loss-offunction mutations, lin-12(r) (25). The lin-12(d) alleles were identified by their Multivulva (Muv) phenotype and the lin-12(r) have been obtained primarily as revertants of lin-12(d). Both genetic and molecular evidence support the notion that lin-12(r) alleles cause a decrease in fin-12 activity, while lin-12(d) alleles cause an increase (22, 25). The lin-12(r) and lin-12(d) mutations transform cell fates in several equivalence groups and have reciprocal effects (25). Generally, equivalent cells follow a primary fate in lin-12(r) mutants and a secondary fate in lin-12(d) mutants. For example, in the AC/VU equivalence group (see section above on

NEMATODE CELL INTERACTIONS 421 vulval development), both cells follow the primary AC fate in lin-12(r), but the secondary VU fate in lin-12(d) mutants. In animals lacking lin-12, the anchor cell appears to be no longer able to signal its neighbor to assume the VU fate. In males, lin-12 mutations have comparable effects on the fates adopted by the cells of the P9.p, P10.p, P11.p equivalence group (25). simple interpretation of these results is that the lin-12 product mediates lateral signaling (25, 70, 73). Other effects of lin-12 mutations indicate that this gene plays a fairly broad role in controlling cell fates during development. First, in the G2/W equivalence group, the effects of lin-12 are opposite that described above (25). Both cells follow the secondary W fate in lin-12(r) mutants and the primary G2 fate in lin-12(d) mutants. This may be a case where lin-12 is required for reception of a signal that specifies the primary fate. Second, some pairs of cells that exhibit reciprocal fate transformations in lin-12(r) and lin-12(d) mutants do not change their fates in response to ablation (e.g. sex muscle and coelomocyte fates) (25, 80). This suggests that lin-12 may not always be required for cell interactions per se. An alternative mechanism, consistent with the molecular identity of lin-12 as a membrane protein (see below), invokes autocrine stimulation by these cells. Finally, in lin-12(r) mutants, germ cells in the proximal region of the gonad undergo ectopic mitotic proliferation (68). Therefore, the effects of lin-12 activity are not limited to somatic cells. However, Seydoux and colleagues have shown that this abnormal proximal germline proliferation occurs because lin-12 activity is lacking in the somatic gonad and suggest that lin-12(+) expressed on somatic gonadal cells normally blocks an interaction between the anchor cell and the germ line (68). MOSAIC ANALYSIS AND TEMPERATURE SHIll" EXPERIMENTS WITH glp-1 AND fin-12 Genetic mosaic analyses of glp-i and lin-12 have demonstrated that each is required in the receiving cell of its relevant cell interaction. Thus, glp-l(+) activity is required in the germ line but not the distal tip cell for germline induction (4) and lin-12(+) activity is required in the ventral uterine precursor cell but not the anchor cell for lateral signaling within the AC/VU equivalence group (67). In addition, temperature-shift experiments with glpl(ts) and lin-12(ts) mutants have shown that each activity is required at the time when the relevant cell interaction occurs. Thus, glp-1 activity is required throughout postembryonic development Ibr continued germline mitoses and only from the 4- to 28-cell stage for embryonic induction (4, 60). Similarly, lin-12 appears to act during the L2 stage for lateral signaling in the AC/VU equivalence group (25), just when the AC/VU interaction occurs (43). MOLECULAR ANALYSIS OF glp-i AND lin-12 The glp-1 and lin-12 genes appear to encode proteins that are similar to each other both in sequence and

422 LAMBIE & KIMBLE overall organization (5, 22, 91, 92). The glp-1 and lin-12 proteins are 48% identical and 66% similar in amino acid sequence (V. Kodoyianni, personal communication) and have conserved exon/intron boundaries (91). Furthermore, the lin-12 and glp-1 genes arc located very close to each other (approximately 20kb apart) on chromosome II1 (4). Therefore, these two genes likely to have evolved by gene duplication. The glp-1 and lin-12 proteins share the same structure (5, 91, 92). Both possess a putative signal sequence at the N-terminus and an internal hydrophobic region followed by positively charged amino acids that might function as a transmembrane domain. Therefore, these are likely to be transmembrane proteins. In addition, both are composed of the same types of structural motifs arranged in the same order. The predicted extracellular domain is composed almost entirely of two different types of cysteine-rich motifs. One motif is EGF-like and is repeated 13 times in lin-12 and 10 times in glp-1; the other, the LNG motif (see below), is present in three copies both lin-12 and glp-1. Intracellularly, there are six copies of a motif observed originally (8) in two yeast genes, cdclo (required in Schizosaccharomyces pombe for initiation of the cell cycle) and SWI6 (required in Saccharomyces cerevisiae for HO gene transcription during G0. Recently, this repeat has been found in several other regulatory proteins, including the transcription factor NF-KB (40) and vertebrate erythrocyte ankyrin (54), a protein thought to attach cytoskeletal elements to the cell membrane. The structure of the glp-1 and lin-12 proteins is strikingly similar to that of the protein encoded by Notch (5, 39, 86, 91, 92), Drosophila gene th at al so is res ponsible for developmental cell interactions (52, 58). Moreover, vertebrate homologues with all the features of lin-12, Notch, and glp-1 have recently been discovered (13; G. Lemke, personal communication; J. Sklar, personal communication). We call this group of genes the LNG gene family, for its founding members: lin-12, Notch, and glp-1. The molecular characterization of glp-1 and lin-12 mutations provides critical information about the functional and regulatory domains of these proteins. For glp-1, severe loss-of-function alleles are either nonsense mutations in the first half of the coding region or missense mutations in the EGF-like, LNG, or SWI6 repeats (V. Kodoyianni, E. Maine, J. Kimble, manuscript in preparation). For lin-12, one loss-of-function mutant (n302n653) carries a missense mutation in one of the SWI6 repeats (23). The glp-1(q35) allele, which phenotypically mimics lin-12(d), is a nonsense mutant predicted to generate a protein lacking 122 amino acids from the C-terminus (55a). This terminal region is not highly conserved between lin-12 and glp-1 and may be required for somatic regulation of glp-1 activity. The other partial loss-of-function glp-1 alleles all map to the EGF-like repeats (V. Kodoyianni, E. Maine, J. Kimble, manuscript in preparation). The lin-12 gain-of-function alleles are all missense mutations in the extracellular region

NEMATODE CELL INTERACTIONS 423 adjacent to the transmembrane domain (23). Because the analogous amino acids are highly conserved in glp-1, the two proteins are likely to be regulated by a similar mechanism. Based on experiments that indicate an interaction between lin-12 alleles, Greenwald & Seydoux (23) propose that the gain-offunction lin-12 mutations might promote dimerization. However, these mutations could equally well interfere with dimerization or disrupt any downregulation of protein activity. glp-i AND lin-12 OVERLAP IN FUNCTION In addition to being similar in structure and function, the lin-12 and glp-1 proteins appear to be biochemically interchangeable. Three lines of evidence support this idea. First is the phenotype of the lin-12 glp-1 double mutant (51). Whereas both lin-12 and glp-1 single mutants usually reach adulthood, the lin-12 glp-1 double mutant invariably arrests in the first larval stage. Therefore, in glp-1 and lin-12 single mutants, the presence of the wild-type product of the other gene is essential for survival. Therefore, certain developmental processes must be able to rely on either glp-1 or lin-12 function. Second is the similarity in the effects of the lin-12(d) and glp-l(q35) mutations. Both cause a Muv phenotype that is anchor cell-independent (5, 25, 55a). Third is the effect of lin-12(r) on the germ line. In lin-12(r) mutants, mitotic germline nuclei are present at the proximal end of the gonad in addition to the normal site at the distal end (68). Seydoux and colleagues speculate that the anchor cell signal mimics the distal tip cell signal and that the positive regulatory signal normally received by lin-12 product may also activate the glp-1 protein. The major differences between glp-i and lin-12 may be regulatory in nature. The glp-1 gene is primarily responsible for germline-dependent interactions (embryonic induction of the pharynx depends on maternal glp-1 product, which is presumed to be produced in the germ line). Although glp-1 plays a role in the embryonic soma in the absence of lin-12, the lin-12 gene appears to be more critical to these decisions (51). The lin-12 gene, in contrast to glp-1, is primarily responsible for somatic interactions and specifically for lateral signaling. One simple mechanism to achieve such a disparity in the roles of functionally similar proteins is to confine the expression of each to a separate tissue. By Northern blot analysis, it has been shown that glp-1 RNA accumulates primarily in the germline (5). In addition, analysis of the glp- 1(q35) mutation indicates that the somatic activity of glp-1 may be inhibited posttranslationally (55a). Therefore, two regulatory mechanisms have already been discovered that could explain the restriction of glp-1 activity to the germ line. GENES ACTING WITH glp-1 AND/OR lin-12 In screens for mutants with the glp-1 and lin-12 mutant phenotypes, only glp-1 and lin-12 were discovered.

Collagen suppressors of glp-1 Mutations in eight genes required for the normal morphology of the animal act as recessive suppressors of glp-1 (55). These genes had been previously identified: dpy-l, dpy-2, dpy-3, dpy-7, dpy-8, dpy-9, dpy-lo, and sqt-1 (9, 15). Three of these genes have been sequenced and found to encode collagens (50; J. Kramer, personal com- Annual Reviews 424 LAMBIE & KIMBLE However, two other approaches have successfully identified genes that act with glp-1 and/or lin-12. The lag genes In screens for mutants with a larval-lethal phenotype similar to that of the lin-12 glp-i double mutant, two genes, lag-1 and lag-2 (for /in-12 and glp-1), were identified (51). For both lag-1 and lag-2, the strongest alleles are probable null mutations and result in a phenotype that is virtually indistinguishable from that of the lin-12 glp-1 double mutant. Also, for each lag gene, weak mutations exhibit postembryonic phenotypes diagnostic of lin-12 and glp-1 single mutants. Therefore, each lag gene is involved not only in the embryonic functions of lin-12 and glp-1 but also in the postembryonic functions of lin-12 and glp-1. The roles played by lag-1 and lag-2 are not understood. Double mutant experiments have not been informative. Since the lin-12 glp-1 double mutant phenotype is the same as the lag-1 and lag-2 probable null phenotypes, epistasis analysis cannot establish their functional relationship. Furthermore, the double mutants that might have been useful, the lin-12(d); lag double mutants, are Lag (E. Lambie, S. Christensen, J. Kimble, unpublished observations). This result does not distinguish among the various relationships that the lag genes might have with lin-12/glp-1. One or both might be an obligate positive regulator of lin-12/glp-1, act together with lin-12/glp- 1, or function downstream of lin-12/glp-l. A combined mosaic and molecular analysis of the lag genes will likely be required to understand their functional relationships. The sog and sel genes Isolation of suppressors of glp-l(ts) and lin-12(d) mutations has revealed eight sog (for suppressors of glp-1) genes (E. Maine, J. Priess, personal communication) and six sel genes (for suppressors/ enhancers of/in-12) (J. Thomas, I. Greenwald, personal communication). The sog and sel mutants have no visibly abnormal phenotype. Among the suppressors of lin-12(d), sel-3 mutations are now known to be dominant gain-of-function mutations of lag-2 (F. Tax, J. Thomas, R. H. Horvitz, personal communication). The lag-2/sel-3 locus was first identified in a screen for lethals and was originally called let-461 (37). Suppression lin-12(d) by lag-2(gf) indicates an interaction between lin-12 and lag-2, but the nature of that interaction is unclear.

NEMATODE CELL INTERACTIONS 425 munication); the molecular identities of the others are not known. This finding suggests an interaction between the extracellular matrix and glp-1. If the collagen mutations alter the structure of the gonadal basement membrane, this may lead to an increase in the availability of the glp-1 protein to receive the distal tip cell signal. This model requires that a partially functional glp-1 gene be present and explains the lack of suppression of null mutations (55). MODELS FOR THE ACTION OF glp-i/lin-12 Figure 4 presents a speculative model for regulation by glp-1, based on the glp-i mutant phenotype (4, 59), the requirement for glp-1 in the receiving cell (4), and the probable status glp-1 as a transmembrane protein (5, 91; S. Crittenden, J. Kimble, unpublished). The glp-1 protein is proposed to be a receptor for a signal produced by the distal tip cell. Although the hypothetical signal is drawn as diffusible, a membrane-bound signal is equally possible. Furthermore, the glp-1 receptor is proposed to be activated by the signal and to generate an Figure 4 A model for the control of germlinc proliferation. The distal tip cell is shown as producing a diffusible signal that activates glp-1, which is expressed on the surface of the syncytial germ line. The activation ofglp-i triggers a transduction pathway that results in mitosis. glp-1 is not activated in regions of the germ line that are more distant from the distal tip cell, resulting in the entry into meiosis.

426 LAMBIE & KIMBLE intracellular second messenger, which in turn activates mitosis, e. g. by stimulating a p34/cdc2-1ike (56) activity. One simple modification of this general model incorporates the idea that the extracellular matrix may negatively regulate the glp-1 receptor (55). In this case, the primary function of the signal might be to interfere with the negative regulatory interaction between glp-i and the extracellular matrix. Clearly the modelof Figure 4 is speculative and other mechanisms are conceivable. For example, the glp-1 protein might act by stabilizing a receptor rather than by functioning as one itself. Alternatively, there may be no signal; the glp-1 protein might be in dynamic equilibrium between a free activated state and an inhibited state in which it is bound by a negative regulator (e. g. the extracellular matrix). Given that lin-12 is so similar to glp-1 in structure and function, one might think that the model presented for induction in Figure 4 might also apply to lateral signaling. To some degree this is true, but lateral signaling differs from induction in an important way. In induction, the two interacting cells are distinct at the outset, whereas in lateral signaling, they are the same. Therefore, some additional mechanism must be postulated to generate two different cells. One such mechanism has been proposed by Seydoux & Greenwald (67), who suggest that each equivalent cell initially produces both the lin-12 receptor and a limited amount of signal, but not enough signal to activate lin-12. At some point, one of the cells begins to produce enough signal to activate lin-12 in the apposing cell, either because of random fluctuation or because of some bias imposed by position or lineage. The activated lin-12 stimulates a negative regulatory circuit that down-regulates its own signal production, resulting in the establishment of two distinct cell types. Genes Regulating the VH/SH Decision The genes that regulate the VH/SH decision can be placed in two categories based on the phenotypes of loss-of-function or reduced function alleles. First, mutations in six genes, lin-2, lin-3, lin-7 (35), lin-lo, let-23 (18), and let-60 (7, 26), have a Vulvaless (Vul) phenotype. In Vul mutants, most vulval precursor cells follow the SH fate (78). Second, mutations in three genes, lin-1 (35), lin-13, and lin-15 (18), have a Multivulva (Muv) phenotype. Muv mutants, most vulval precursor cells adopt a VH fate (either VH1 or VH2) (78). In strong Muv mutants, all vulval precursor cells execute fates, even in the absence of the anchor cell (20). In addition, recessive mutations in certain pairs of genes (e. g. lin-8 and lin-9) result in a Muv phenotype (see below); these are termed the SynMuv loci, because of the synthetic nature of the Muv phenotype (19, 35). MOSAIC ANALYSIS OF vul AND Muv GENES Based on laser ablation experiments (43, 70, 80), the Vul and Muv genes might be predicted to act

NEMATODE CELL INTERACTIONS 427 either in the anchor cell as part of the signaling mechanism or in the vulval precursor cells as part of the receiving mechanism. Because the phenotypes of the Muv mutants are essentially anchor cell-independent (20), the prediction was that the Muv genes might act in the vulval precursor cells. To date this prediction has been tested for only one gene, lin-15, with surprising results. In a landmark study, Herman & Hedgecock found that expression of the wildtype lin-15 gene in the vulval precursor cells is neither necessary nor sufficient to rescue the lin-15 Muv phenotype (31). This result, in combination with an analysis of which cells do need to express lin-15 to ensure normal vulval development, led them to speculate that lin-15 is normally expressed within the syncytial hypoderm (hyp7), where it would act to repress the fates in the adjoining vulval precursor cells. By this model, the anchor cell would function by relieving repression from the surrounding hypodermis. The anchor cell could accomplish this by sending an overriding signal to the vulval precursor cells or by down-regulating lin-15 (and/or other Muv genes) within hyp7. An alternative model is that hyp7 might relay the inductive signal from the AC to the vulval precursor cells, in which case lin-15 would function by preventing expression/transmission of the inducing signal. Additional studies are needed to distinguish among these models. SUPPRESSORS OF lin-15 In an effort to identify additional loci required for the VH/SH decision, suppressors of the fin-15 Muv phenotype have been isolated (7, 26). Among the lin-15 suppressors, alleles of let-60, a gene previously identified by recessive lethal mutations (12, 64), were obtained. Some lin-15 suppressor alleles are recessive Vul or recessive lethal (7) and others are both dominant Vul and recessive lethal mutations (7, 26). separate class of dominant let-60 mutations [previously thought to define a separate gene, lin-34 (18)] cause a Muv phenotype (7). Together, these suggest that the activity state of let-60 acts as a switch, with a high level being necessary and sufficient for the induction of VH fates and a low level resulting in the SH fate. Weak alleles of let-23, another gene first identified by recessive lethal mutations (69), were also obtained as suppressors of fin-15 (3). These weak let-23 alleles have a Vul phenotype. Significantly, the dominant Muv alleles of let-60 are epistatic to the Vul let-23 mutations (26). The analysis of the coding sequences of let-60 and let-23 (see below) has provided a credible biochemical basis for this relationship. MOLECULAR CHARACTERIZATION OF Vul AND Muv GENES Recent molecular data provide insight into the functions of two Vul genes. First, let-23 encodes a member of the EGF receptor subfamily of receptor tyrosine kinases (2). Although the tissue within which let-23 functions is unknown, an

428 LAMBIE & KIMBLE exciting possibility is that let-23 is the receptor for the anchor-cell signal. Second, the let-60 gene encodes a protein homologous to mammalian ras (27). All five dominant Muv mutations of let-60 are associated with a change in amino acid 13, which is comparable to one of the alterations of mammalian ras resulting in oncogenic activation (7). Because let-60 dominant Muv mutations are epistatic to let-23 Vul mutations (26), let-60 is likely to function downstream from let-23 in a pathway of signal transduction that specifies VH fates. Only one other Muv/Vul gene has been cloned and sequenced: lin-lo encodes an apparently unique protein (41). Therefore, the analysis of these genes will provide more than a genetic model for known proteins. It will reveal new activities involved in either the function or regulation of wellknown proteins such as ras. A MODEL FOR VULVAL INDUCTION Classically, the functional relationships among regulatory genes have been deduced from double mutant experiments. However, in the case of the Muv and Vul genes, the interpretation of such experiments is considerably hampered by the inability to use null mutations (45). Certain Muv and Vul genes have a lethal null phenotype, which means that only partial loss-of-function or gain-of-function mutations can be used for epistasis. Other Muv and Vul genes have not been sufficiently characterized to know the nature of their mutations. Therefore, much of the double mutant data remains difficult to interpret, although certain double mutants (see below) have helped our understanding of how these genes may interact. Figure 5 presents a speculative model for the genetic control of the VH/SH decision, based largely on proposals made by several investigators, notably Hedgecock, Herman, Horvitz, and Sternberg. The key results essential to the model are as follows: (a) all Muv mutants are anchor-cell independent (20) and thus these genes are unlikely to function as part of the mechanism by which the anchor cell signals the vulval precursor cells. (b) fin-15 appears to act in hyp7 (31). Since the lin-15 locus has components required for both functionally redundant Muv pathways identified by the SynMuv genes (19), we suggest in Figure 5 that all Muv genes (Muv and SynMuv) direct interaction between hyp7 and the vulval precursor cells that normally promotes SH or inhibits the VH fates. Some Muv genes, e.g. lin-15, may act in hyp7 while others act in the vulval precursor cells. Indeed, some Muv genes (including lin-15) may act in both hyp7 and the vulval precursor cells. (c) let-23 is similar to the EGF receptor, which stimulates proliferation in mammalian cells (2). This gene is therefore a good candidate for the receptor of the anchor cell signal and is placed in the vulval precursor cells as part of the mechanism for stimulating the VH fate. (d) Dominant gain-of-function mutations of let-60 are epistatic to partial loss-of-function mutations of let-23 (26). Therefore, let-60 appears to act downstream of let-23 in the pathway

NEMATODE CELL INTERACTIONS 429 SYNCYTIAL HYPODERH ANCHOR CELL let-6o Syncytlal Hypoderm Vulval Hypoderm VULVAL PRECURSOR CELL Figure 5 A model for the control of vuval precursor cell fates. Vulval hypodermal fates are induced by the anchor cell and inhibited by the syncytial hypoderm. The Vul genes are required for induction, whereas the Muv genes are required for inhibiton, let-23 is proposed to act as the receptor l~r an inductive signal. When activated, let-23 activates let-60, which promotes vulval hypodermal fates. stimulating the VH fate. (e) Partial loss-of-function mutations of the Muv gene, lin-15, are epistatic to putative null mutations in three Vul genes, lin-2, lin-7, and lin-lo (20). Although a simple interpretation of this result is that lin-15 acts downstream of these three Vul genes, another interpretation, which takes into account the fact that lin-15 may act in hyp7, is that the Muv pathway is antagonistic to the Vul pathway. By this model, when the Muv pathway is defective, certain components of the Vul pathway are unnecessary. Further investigations of the VH/SH decision will test the primary tenets of the model shown in Figure 5. Both the involvement of hyp7 in repressing the VH fate and the action of let-23 in the vulval precursor cells to stimulate the VH fate remain speculative. Mosaic analyses and localization of the proteins involved will provide important tests for each of these hypotheses. The idea that let-23 encodes the receptor for the anchor-cell signal stands as one of the central features of this model and the identification of this signal will be essential for understanding how vulval fates are induced. pal-1 The fates of the V1-V6 cells of the male lateral hypodermis (see above) are subject to several genetic controls. In part, the decision between an anterior (seam cell) fate and a posterior (sensory ray) fate is determined by the position

430 LAMBIE & KIMBLE of a cell along the antero-posterior axis. The perception of this position appears to be antagonistically regulated by two genes, lin-22 and mab-5, the latter of which encodes a homeodomain containing protein (14, 38). The best evidence that the decision between sensory ray and seam cell fates is also regulated by direct cell interactions comes from studies ofpal-i mutants (for posterior alae) (85). In toss-of-function pal-1 mutants, the V6 cell produces seam cells instead of rays. However, if the posterior neighbor of V6 (T) ablated, V6 produces its normal complement of rays. To explain these results, Waring & Kenyon propose that T sends a signal to V6, causing it to produce alae instead of rays and that this signal is normally rendered ineffective, or "silenced", by pal-1 (85). Mosaic analysis indicates that the silencing of this signal requires wild-type pal-1 activity within V6, but not T (84). Since pal-1 encodes a homeodomain containing protein (84), Waring & Kenyon suggest that pal-1 acts within V6 (or its descendants), either to prevent the expression of genes required for the transduction of a signal from T or to promote the expression of genes (e. g. mab-5) that are required for the production of rays. Similar signaling and silencing events may regulate the fates of other cells within the lateral hypodermis. Other Genetic Pathways We have focused above on the genetic analysis of cell interactions for which most is known both at the cellular and genetic levels. Here we mention two other pathways that may also provide insight into how signal transduction pathways regulate cell fate. SEX DETERMINATION The genetic pathway controlling the sexual phenotype of the soma in C. elegans is one of the best understood regulatory pathways in this organism (33, 83). Essentially, a cascade of genes interprets the X/A ratio to set the activity state of a master regulator, tra-1. The tra-1 gene functions cell-autonomously (36) and appears to encode a transcription factor (D. Zarkower, J. Hodgkin, personal communication). If tra-1 is active, the soma is female; if it is inactive, the soma is male. Genetic and molecular analyses of the sex-determining genes indicate that cell interactions may be involved in the regulatory cascade. A gene mediating an early step in the pathway, her-l, acts nonautonomously (C. Hunter, W. B. Wood, personal communication) and encodes a small protein with an apparent signal sequence (M. Perry, W. B. Wood, personal communication). The tra-2 gene, which is regulated by her-l, appears to encode a membrane protein (P. Kuwabara, J. Kimble, unpublished). Together, these data raise the intriguing possibility that tra-1 activity depends on a signal transduction event mediated by her-1 and tra-2.

(a) Cell interactions play a significant role in regulating development in C. elegans. Although the first cell interactions identified in C. elegans were considered to be exceptions to rule by ancestry (43, 48, 80), the continued discovery of interactions, including embryonic induction (60) and control of left-right asymmetry (89), have demonstrated that this organism much like other metazoans in its reliance on cell interactions during development. (b) The molecular controls of cell interactions in C. elegans are homologous to those in other metazoans, including higher vertebrates. In the past year, two genes from C. elegans that regulate vulval interactions, let-23 and let-60, have been shown to be similar to the EGF receptor and ras of higher organisms (2, 27). Furthermore, vertebrate members of the LNG family were discovered during that same period (13; G. Lemke, personal communication; J. Sklar, personal communication). This similarity implies that at least some molccular mechanisms regulating cell interactions during metazoan development may have evolved in a progenitor to both nematodes and vertebrates. The molecular similarity between nematode regulators and the vertebrate signal transduction machinery promises to be very informative. In ver- Annual Reviews NEMATODE CELL INTERACTIONS 431 DAUER DEVELOPMENT If growth conditions are favorable, larval development consists of four stages (L I-L4). However, if overcrowded or starved, animal will enter a facultative stage, called a dauer larva, at the second molt (reviewed in ref. 62). Environmental stimuli, received by sensory neurons the head called amphids, are transmitted to divert development into the dauer pathway (reviewed in refs. 6, 62). The decision to develop as a dauer larva regulated by a set of 20 daf genes that have been ordered into a pathway (6, 61, 63). Two genes that when mutant lead to constitutive dauer formation have recently been found to encode receptor protein kinases (21; D. Riddle, personal communication). The sites of action of these receptors are not known. Two reasonable possibilities are that they might be involved in transduction of environmental cues leading to dauer formation or the neural integration of this signal. CONCLUSIONS, SPECULATIONS, AND PROSPECTS What progress have we made over the past decade towards understanding the genetic control of cell interactions during development in C. elegans? In this section we present what we consider to be the most important conclusions that have emerged and discuss their implications, both for the evolution of these controls and for the direction of future investigations.

432 LAMBIE & KIMBLE tebrates, the developmental role of EGF and EGF receptor is not understood. Known vertebrate mutations in EGF receptor cause a gain of gene function and therefore may be misleading with regard to the normal role played by this gene during development (10). An understanding how a protein similar to EGF receptor controls interactions in worms will provide ideas about how the EGF receptor may work in vertebrates. In addition, the functional relationship between tyrosine kinase receptors and ras has proven to be virtually intractable to date. The similarities of let-23 and let-60 to these proteins provide a genetic inroad into the problem. Furthermore, the discovery of previously undescribed proteins, such as that encoded by lin-lo (41), will certainly help to identify additional components involved in this type of transduction event. (c) Induction and lateral signaling rely, at least in part, on the same molecular mechanism. The glp-1 and lin-12 genes are similar in structure (5, 91) and appear to be functionally interchangeable in some contexts (51, 55a, 67). Yet glp-1 primarily regulates inductive events (4, 59) and lin-12 primarily regulates lateral signaling (25, 70, 75). Furthermore, lag-1 and lag-2 mediate both inductive and lateral signaling types of interactions. Therefore this classification is likely to be artificial when viewed from the perspective of molecular mechanism. (d) Cell interactions can be negatively regulated or "silenced". Interactions that are regulated in this manner were first detected in mutant backgrounds, e. g. pal-l(if) (85) or lin-12(o) (68). These findings suggest that the wild-type function of pal-1 and lin-12 is, at least in part, to prevent or silence a cell interaction. The extent to which the silencing of cell interactions is used for developmental regulation is not yet known. (e) Many genes identified by their regulation of one cell interaction control other interactions as well. Such pleiotropy has important consequences for the genetic analysis of these genes. For example, the effect of glp-1 on pharyngeal induction was not appreciated until an unusual glp-1 allele was isolated in which some animals escaped the germline and hypodermal defects (59). Similarly, the role of let-23 and let-60 in cell interactions was unknown until weak alleles were isolated that had specific vulval defects (7, 18, 26). These findings underscore the necessity of analyzing weak alleles in addition to null alleles. Both kinds of mutation are informative and useful. (f) Redundancy is not uncommon among genes controlling cell interactions. Good examples are the requirement for either lin-12 or glp-1 during embryogenesis (51) and the SynMuv genes in vulval development (19). Redundancy can make the identification of genes by mutation extremely difficult. Their discovery depends on the identification of rare double mutants (35) or dominant alleles (24, 57). Furthermore, partial redundancy can delay identification of other genes functioning in the same process.

NEMATODE CELL INTERACTIONS 433 Thus, for lin-12 and glp-1, finding the lin-12 glp-1 double mutant phenotype allowed identification of the two lag genes, which are required for both glp-i and lin-12 action (51). (g) Regulators that are similar at the molecular level can have both overlapping and unique functions during development. For example, glp-1 and lin-12 are homologs that are redundant during embryogenesis (51) but have distinct roles postembryonically (4, 25, 59). The molecular basis the separate developmental roles played by such similar genes may rely on gene-specific regulation (e. g. differences in expression) or partially divergent functions. (h) The extracellular matrix is likely to influence genes that regulate cellular interactions. The idea that the extracellular matrix plays a critical role in controlling development is not new (see ref. 82); however, little genetic evidence has accrued to support it. The finding that mutant collagens alter the activity of glp-1 provides supporting in vivo evidence and may provide a genetic model for dissecting the role of the extracellular matrix in regulatory cell interactions (55). CONCLUDING REMARKS The powerful molecular genetics of C. elegans coupled with the simplicity of its anatomy and development has provided a remarkable experimental accessibility for the analysis of regulatory cell interactions during development. Over the past decade it has become clear that interactions dictating development in C. elegans are similar both in scope and type to those analyzed in other metazoans, including both invertebrates and vertebrates. In addition, the idea is emerging that the molecular controls of nematode cell interactions are similar to those present in vertebrates. We can therefore look forward during the next decade to an increasingly productive interplay between the molecular genetic analysis of cell interactions during development and the biochemical analysis of signal transduction in vertebrate cells. Literature Cited 1. Albertson, D. G., Thomson, J. N. 1976. The pharynx of Caenorhabditis elegans. Philos. Trans. R. Soc. London Ser. B Biol. SoL 275:299-325 2. Aroian, R. V., Koga, M., Mendel, J. E., Ohshima, Y., Stcrnberg, P, W. 1990. The let-23 gene necessary for Caenorhabditis elegans vulval induction encodes a tyrosine kinase of the EGF receptor subfamily. Nature 348:693-99 3. Aroian, R. V., Sternberg, P. W. 1991. Multiple functions of let-23, a C, elegans receptor tyrosine kinase gene required for vulval induction. Genetics. 128:251-267 Austin, J., Kimble, J. 1987. glp-1 is required in the germ line for regulation of the decision between mitosis and meiosis in C. elegans. Cell 51:589-99 Austin, J., Kimble, J. 1989. Transcript analysis of glp-1 and lin-12, homologous genes required for cell interactions during development of C. elegans. Cell 58:565-71

434 LAMBIE & KIMBLE 6. Bargmann, C. I., Horvitz, H. R. 1991. Control of larval development by chemosensory neurons in C. elegans. Science 251:1243-1246 7. Beitel, G. J., Clark, S. G., Horvitz, H. R. 1990. Caenorhabditis elegans ras gene let-60 acts as a switch in the pathway of vulval induction. Nature 348:503-9 8. Breeden, L., Nasmyth, K. 1987. Cell cycle control of the yeast HO gene: cisand trans-acting regulators. Cell 48:389-97 9. Brenner, S. 1974. The genetics of Caenorhabditis elegans. Genetics 77:71-94 10. Carpenter, G., Wahl, M. I. 1990. Handbook of Experimental Pharmacology. Volume 95/1. Peptide Growth factors and their receptors I. The epidermal growth factor family, p. 69-171. M. B. Sporn and A. B. Roberts eds. Springer- Verlag 11. Chisholm, A. D., Hodgkin, J. 1989. The mab-9 gene controls the fate of B, the major male-specific blast cell in the tail region of Caenorhabditis elegans. Genes Dev. 3:1413-23 12. Clark, D. V., Rogalski, R. M., Donati, L. M., Baillie, D. L. 1988. The unc- 22(lV) region of Caenorhabditis elegans: genetic analysis of lethal mutations. Genetics 119:345-53 13. Coffman, C., Harris, W., Kintner, C. 1990. Xotch the Xenopus homolog of Drosophila Notch. Science 249:1438-41 14. Costa, M., Weir, M., Coulson, A., Sulston, J., Kenyon, C. 1988. Posterior pattern formation in C. elegans involves position-specific expression of a gene containing a homeobox. Cell 55:747-56 15. Cox, G. N., Laufer, J. S., Kusch, M., Edgar, R. S. 1980. Genetic and phenotypic characterization of roller mutants of Caenorhabditis elegans. Genetics 95:317-39 16. Deppe, U., Schierenberg, E., Cole, T., Krieg, C., Schmitt, D., et al. 1978. Cell lineages of the embryo of the nematode Caenorhabditis elegans. Proc. Natl. Acad. Sci. USA 75:376-80 17. Ellis, R. E., Yuan, J., Horvitz, H. R. 1991. Mechanisms and functions of cell death. Annu. Rev. Cell Biology 7:000- O0 18. Ferguson, E., Horvitz, H. R. 1985. Identification and characterization of 22 genes that affect the vulval cell lineages of Caenorhabditis elegans. Genetics 110:17-72 19. Ferguson, E., Horvitz, H. R. 1989. The multivulva phenotype of certain C. elegans mutants results from defects in two functionally-redundant pathways. Genetics 123:109-21 20. Ferguson, E. L., Sternberg, P. W., Horvitz, H. R. 1987. A genetic pathway for the specification of the vulval cell lineages of Caenorhabitis elegans. Nature 326:259-67 21. Georgi, L. L., Albert, P. S., Riddle, D. L. 1990. daf-1, a C. elegans gene controlling dauer larva development, encodes a novel receptor protein kinase. Cell 61:635-45 22. Greenwald, I. 1985. lin-12, a nematode homeotic gene, is homologous to a set of mammalian proteins that includes epidermal growth factor. Cell 43:583-90 23. Greenwald, I., Seydoux, G. 1990. Analysis of gain-of-function mutations of the lin-12 gene of Caenorhabditis elegans. Nature 346:197-99 24. Greenwald, I. S., Horvitz, H. R. 1980. unc-93(e1500): A behavioral mutant of Caenorhabditis elegans that defines a gene with a wild-type null phenotype. Genetics 96:147-64 25. Greenwald, I. S., Sternberg, P. W., Horvitz, H. R. 1983. The lin-12 locus specifies cell fates in C. elegans. Cell 34:435-44 26. Han, M., Aroian, R. V., Sternberg, P. W. 1990. The let-60 locus controls the switch between vulval and non-vulval cell types in C. elegans. Genetics 126: 899-913 27. Han, M., Sternberg, P. W. 1990. let-60, a gene that specifies cell fates during C. elegans vulval induction, encodes a ras protein. Cell 63:921-31 28. Hedgecock, E. M., Culotti, J. G., Hall, D. H. 1990. The unc-5,unc-6, and unc- 40 genes guide circumferential migration of pioneer axons and mesodermal cells on the epidermis in C. elegans. Neuron 2:61-85 29. Hedgecock, E. M., Culotti, J. G., Hall, D. H., Stern, B. D. 1987. Genetics of cell and axon migration in Caenorhabditis elegans. Development 100:365-82 30. Herman, R. K. 1989. Mosaic analysis in the nematode Caenorhabditis elegans. J. Neurogenet. 5:1-24 31. Herman, R. K., Hedgecock, E. M. 1990. Limitation of the size of the vulval primordium of Caenorhabditis elegans by lin-15 expression in surrounding hypodermis. Nature 348:169-71 32. Hirsh, D., Oppenheim, D., Klass, M. 1976. Development of the reproductive system of Caenorhabditis elegans. Dev. Biol. 49:200-19 33. Hodgkin, J. 1990. Sex determination compared in Drosophila and Caenorhabditis Nature 344:721-28 34. Horvitz, H. R., Sternberg, P. W. 1991.

58. Poulson, D. F. 1937. Chromosomal deficiencies and embryonic development of Drosophila melanogaster. Proc. Natl. Acad. Sci. USA 23:133-37 59. Priess, J. R., Schnabel, H,, Schnabel, R. 1987. The glp-1 locus and cellular interactions in early C. elegans embryos. Cell 51:601-11 60. Priess, J. R., Thomson, J. N. 1987. Cellular interactions in early C. elegans embryos. Cell 48:241-50 61. Riddle, D. L. 1977. A genetic pathway for dauer larva formation in Caenorhab- Annual Reviews NEMATODE CELL INTERACTIONS 435 Multiple intercellular signalling systems control the development of the C. elegans vulva. Nature. 351, 535~.1 35. Horvitz, H. R., Sulston, J. E. 1980. Isolation and genetic characterization of cell-lineage mutants of the nematode Caenorhabditis elegans. Genetics 96: 435-54 36. Hunter, C. P., Wood, W. B. 1990. The tra-1 gene determines sexual phenotype cell-autonomously in C. elegans. Cell 63:1193-204 37. Johnsen, R. C. 1990. Genetic analysis of the left half of linkage group V in Caenorhabditis elegans. PhD thesis. Simon Fraser Univ., Buruaby, BC 38. Kenyon, C. 1986. A gene involved in the development of posterior body region of C. elegans. Cell 46:477-87 39. Kidd, S., Kelley, M. R., Young, M. W. 1986. Sequence of the Notch locus of Drosophila: relationship of the encoded protein to mammalian clotting and growth factors. Mol. Cell Biol. 6:3094-108 40. Kieran, M., Blank, V., Legeat, F., Vandekerckhove, J., Loftspeich, F., et al. 1990. The DNA binding subunit of NF- KB is identical to Factor KBF1 and homologous to the tel oncogene product. Cell 62:1007-18 41. Kim, S. K., Horvitz, H. R. 1990. The Caenorhabditis elegans gene lin-lo is broadly expressed while required specifically for the determination of vulval cell fates. Genes Dev. 4:357-71 42. Deleted in proof 43. Kimble, J. 1981. Lineage alterations after ablation of cells in the somatic gonad of Caenorhabditis elegans. Dev. Biol. 87:286-300 44. Kimble, J., Hirsh, D. 1979. Postembryonic cell lineages of the hermaphrodite and male gonads in Caenorhabditis elegans. Dev. Biol. 70: 396-417 45. Kimble, J., Schedl, T. 1987. Developmental genetics of Caenorhabditis elegans. In Developmental Genetics, A Primer in Developmental Biology, ed. G. M. Malacinski. 171-192. London: Macmillan 46. Kimble, J., Sulston, J,, White, J. 1979. Regulative development in the postembryonic lineages of Caenorhabditis elegans. In Cell Lineage, Stem Cells and Cell Determination, ed. N. LeDouarin, pp. 59-68. New York: Elsevier 47. Kimble, J., Ward, S. 1988. Germ-line development and fertilization. In The Nematode Caenorhabditis elegans, ed. W. B. Wood, pp. 191-213. New York: Cold Spring Harbor Lab. 48. Kimble, J., White, J. G. 1981. On the control of germ cell development in Caenorhabditis elegans. Dev. Biol. 81:208-19 49. Klass, M., Wolf, N., Hirsh, D. 1976. Development of the male reproductive system and sexual transformation in the nematode Caenorhabditis elegans. Dev. Biol. 69:329-35 50. Kramer, J. M., Johnson, J. J., Edgar, R. S., Basch, C,, Roberts S. 1988. The sqt-1 gene of C. elegans encodes a collagen critical for organismal morphogenesis. Cell 55:555-65 51. Lambie, E. J., Kimble, J. 1991. Two homologous regulatory genes, lin-12 and glp-1, have overlapping functions. Development 112:231-240 52. Lehmann, R., Jimnez, F., Dietrich, U., Campos-Ortcga. J. A. 1983. On the phenotype and development of mutants of early neurogenes~s in Drosophila melanogaster. Wilhelm Roux s Arch. Dev. Biol. 192:6~74 53. Li, C., Chalfie, M. 1990. Organogenesis in C. elegans: positioning of neurons and muscles in the egg-laying system. Neuron 4:681-95 54. Lux, S. E., John, K. M., Bennett, V. 1990. Analysis of edna for human erythrocyte ankyrin indicates a repeated structure with homology to tissuedifferentiation and cell-cycle control proteins. Nature 344:36-42 55. Maine, E., Kimble, J. 1989. Identification of genes that interact with glp-l, a g.ene required for inductive cell interaction in C. elegans. Development 106: 133-43 55a. Mango, S. E., Maine, E. M., Kimble J. 1991. A carboxy-terminal truncation activates the glp-1 protein to specify vulval fates in Caenorhabditis elegans. Nature 352:811-15 56. Nurse, P. 1990. Universal control mechanism regulating onset of M-phase. Nature 344:503-8 57. Park, E. C., Horvitz, R. 1986. C. elegans unc-105 mutations affect muscle and are suppressed by other mutations that affect muscle. Genetics 113:853-67

436 LAMBIE & KIMBLE ditis elegans. Stadler Genet. Symp. 9:101-20 62. Riddle, D. L. 1988. The dauer larva. See Ref. 47, pp. 393-412 63. Riddle, D. L., Swanson, M. M., Albert, P. S. 1981. Interacting genes in nematode dauer larva formation. Nature 290:668-71 64. Rogalski, T. M., Moerman, G. G., Baillie, D. L. 1982. Essential genes and deficiencies in the unc-22 IV region of Caenorhabditis elegans. Genetics 102: 725 36 65. Schierenberg, E. 1987. Reversal of cellular polarity and early cell-cell interaction in the embryo of Caenorhabditis elegans. Dev. Biol. 122:452-63 66. Schnabel, R. 1991. Cellular interactions involved in the determination of the early C. elegans embryo. Mech. Dev. 34:85-100 67. Seydoux, G., Greenwald, I. 1989. Cell autonomy of lin-12 function in a cell fate decision in C. elegans. Cell 57:1237-45 68. Seydoux, G., Schedl, T., Greenwald, I. 1990. Cell-cell interactions prevent a potential inductive interaction between soma and gennline in C. elegans. Cell 61:939-51 69. Sigurdson, D. C., Spanier, G. J., Herman, R. K. 1984. Caenorhabditis elegans deficiency mapping. Genetics 108: 331-45 70. Sternberg, P. W. 1988. Control of cell fates within equivalence groups in C. elegans. Trends Neurosci. 11:259-64 71. Sternberg, P. W. 1990. Genetic control of cell type and pattern formation in Caenorhabditis elegans. Adv. Genet. 27:63-116 72. Sternberg, P. W., Horvitz, H. R. 1986. Pattern formation during vulval asymmetric cell divisions. Cell 44:761-72 73. Sternberg, P. W., Horvitz, H. R. 1989. The combined action of two intercellular signalling pathways specifies three cell fates during vulval induction in C. elegans. Cell 58:679-93 74. Sulston, J. E. 1976. Post-embryonic development in the ventral cord of Caenorhabditis elegans. Philos. Trans. R. Soc. London Ser. B 275:287-98 75. Sulston, J. E. 1988. Cell lineage. See Ref. 47, pp. 123-55 76. Sulston, J. E., Albertson, D. G., Thomson, J. N. 1980. The Caenorhabditis elegans male: postembryonic development of nongonadal structures. Dev. Biol. 78:542-76 77. Sulston, J. E., Horvitz, H. R. 1977. Postembryonic cell lineages of the nematode Caenorhabditis elegans. Dev. Biol. 56:110-56 78. Sulston, J. E., Horvitz, H. R. 1981. Abnormal cell lineages in mutants of the nematode Caenorhabditis elegans. Dev. Biol. 82:41-55 79. Sulston, J. E., Schierneberg, E., White, J. G., Thomson, J. N. 1983. The embryonic cell lineage of the nematode Caenorhabditis elegans. Dev. Biol. 100: 64-119 80. Sulston, J. E., White, J. G. 1980. Regulation and cell autonomy during postembryonic development of Caenorhabditis elegans. Dev. Biol. 78:577-97 81. Thomas, J. H., Stern, M. J., Horvitz, H. R. 1990. Cell interactions coordinate the development of the C. elegans egglaying system. Cell 62:1041-1052 82. Trelstad, R. L. 1984. The Role of Extracellular Matrix in Development. New York: Liss 83. Villeneuve, A. M., Meyer, B. J. 1990. Regulatory hierarchy controlling sex determination and dosage compensation in Caenorhabditis elegans. Adv. Genet 27: 117-88 84. Waring, D. A., Kenyon, C. 1991. Regulation of cellular responsiveness to inductive signals in the developing C. elegans nervous system. Nature 350: 712-15 85. Waring, D. A., Kenyon, C. 1990. Selective silencing of cell communication influences anteroposterior pattern formation in C. elegans. Cell 60:123-31 86. Wharton, K. A., Johansen, K. M., Xu, T., Artavanis-Tsakonas, S. 1985. Nucleotide sequence from the neurogenic locus Notch implies a gene product which shares homology with proteins containing EGF-like repeats. Cell 43: 567-81 87. White, J. G. 1988. The anatomy. See Ref. 47, pp. 81-122 88. Wood, W. B. 1988. Embryology. See Ref. 47, pp. 215-41 89. Wood, W. B. 1991. Evidence from reversal of handedness in C. elegans embryos for early cell interactions determining cell fates. Nature 349;536-38 90. Wood, W. B., Kershaw, D. 1991. Handed asymmetry, handedness reversal, and mechanisms of cell fate determination in nematode embryos. Ciba Found. Symp. 162: Biological Asymmetry and Handedness In press 91. Yochem, J., Greenwald, I. 1989. glp-i and lin-12, genes implicated in distinct cell-cell interactions in C. elegans, encode similar transmembrane proteins. Cell 58:553-63 92. Yochem, J., Weston, K., Greenwald, I. 1988. C. elegans lin-12 encodes a transmembrane protein similar to Drosophila Notch and yeast cell cycle gene products. Nature 335:547-50