Two maternal genes, apx-1 and pie-1, are required to distinguish the fates of equivalent blastomeres in the early Caenorhabditis elegans embryo

In a developing organism, cells initially equivalent in devel- opmental potential become influenced to adopt different fates by interacting with neighboring cells (for a review, see Greenwald and Rubin, 1992). Studies of the nematode Caenorhabditis elegans led to the first demonstrations that certain pairs or groups of cells, called equivalence groups, are born with the potential to express either of two alternative fates (Kimble and Hirsh, 1979; Kimble, 1981; Sulston and White, 1980). Cell interactions are then required to determine which cell adopts which fate. In some cases, variability is observed in the decision, but in other cases precursors always make the same choice, even though they initially are capable of express- ing more than one fate (e.g. Sulston et al., 1983; Bowerman et al., 1992a; Chamberlin and Sternberg, 1993). Studies in the grasshopper (Taghert et al., 1984; Kuwada and Goodman, 1985; Doe and Goodman, 1985), leech (Weisblat and Blair, 1984), and more recently zebrafish (Eisen et al., 1990; Eisen, 1992) have demonstrated that equivalence groups are widely conserved as a developmental strategy.

The earliest example of an equivalence group in C. elegans is found at the 4-cell stage of embryogenesis. Two sister blas- tomeres called ABa and ABp are born equivalent but ulti- mately express very different patterns of cell fate (Fig. 1; Sulston et al., 1983, Priess et al., 1987). For example, the ante- riorly positioned blastomere ABa generates cells that form the pharynx, an organ used for feeding. In contrast, the dorsal-most blastomere ABp produces no pharyngeal cells, but does con- tribute many specialized cell types not made by ABa. These ABp-specific fates include a group of cells that form the excretory system, several cell types that form parts of the anus and rectum, and a variety of specialized neurons (Sulston et al., 1983). In spite of their ultimately divergent patterns of development, ABa and ABp initially are capable of replacing each other if their positions are switched at the beginning of the 4-cell stage (Priess and Thomson, 1987). Thus ABa and ABp constitute an equivalence group in the early C. elegans embryo, and the specification of their different identities is essential for establishment of the nematode body plan.

The initial equivalence of ABa and ABp suggests that cell interactions with neighboring blastomeres might be reponsible for distinguishing their eventual fates. A likely source for an inductive signal that breaks the ABa/ABp equivalence is the P₂ blastomere. Because P₂ touches ABp but not ABa, a non- diffusible signal from P₂ would influence only ABp (see Fig. 1). Indeed, a previous analysis has shown that if P₂ is destroyed, ABp is unable to produce at least one cell type it normally makes, the intestinal-rectal valve cells (Bowerman et al., 1992a). However, the ultimate fate of ABp in the absence of P₂ signalling, beyond the absence of valve cells, has not been addressed. In this paper, we show that mutations in a newly identified maternal gene, apx-1, and in the maternal gene pie-1, appear to prevent the P₂/ABp interaction. In apx-1 and in pie-1 mutant embryos, ABp fails to produce its normal cell types and instead develops more like ABa. We also show that after removal of P₂, operated wild-type embryos produce extra pharyngeal cells and no intestinal-rectal valve cells, consistent with both ABa and ABp adopting ABa-like fates in the absence of a P₂ induction. We propose that the P₂/ABp interaction breaks the equivalence of ABa and ABp, and we discuss models in which apx-1(+) encodes a component of the P₂/ABp signalling pathway.

N2 was used as the standard wild-type strain and cultured as previ- ously described (Brenner, 1974).The following alleles were used: dpy-18 (e364), pie-1(zu127); pie-1(zu154), glp-1(zu24), unc-25 (e156) LG III; lag-2(s1486), dpy-11(e224), unc-46(e177) LG V. The balancers DnT1 (translocation balancer for LG IV and V), eT1 (translocation balancer for LG III and V), qC1 (inversion balancer for LG III) were used, as well as the deficiencies sDf70 and lag-2(q387). apx-1 embryos were obtained from homozygous apx-1 mothers produced by apx-1/DnT1 heterozygotes. The glp-1 allele zu24 was isolated in a previous screen for maternal-effect mutants (B. Bowerman and J. Priess, unpublished data). zu24 resembles the weak glp-1 allele e2142 (Priess et al., 1987; Kodoyianni et al., 1992) in that many zu24 embryos hatch (89/112) but lack an anterior pharynx. Double mutant glp-1;apx-1 embryos were obtained from homozygous hermaphro- dites produced by unc-32 glp-1/eT1 III; dpy-11 apx-1(or3)/eT1 parents. pie-1 embryos were obtained from the strains dpy-18 pie- 1(zu127)/qC1 and pie-1(zu154) unc-25/qC1.

Three mutant alleles (or3, or15 and or22) of apx-1 (anterior pharynx excess) were isolated in a genetic screen (see Kemphues et al., 1988) of roughly 28,000 haploid genomes, in which we searched for mutants with altered ABp development. For the first allele, we searched for embryos with unusual patterns of epidermal cells. Hypodermal cells in wild-type embryos at the ‘bean stage’ of embryogenesis form stereotyped rows of ventral, lateral and dorsal hypodermal cells (Sulston et al., 1983). Inspection of apx-1 mutant embryos at the bean stage indicates that they often have rows of what resemble dorsal hypodermal cells but no recognizable patterns of lateral or ventral hypodermal cells (not shown). After isolating the first allele, we sub- sequently looked for embryos with extra pharyngeal cells. The frequency at which we found the alleles (1×10⁻⁴) is similar to that at which reduction of function mutations in other maternal genes have been found. For example, in this same screen, we isolated one allele of glp-1 and three alleles of skn-1. The three apx-1 alleles map to chromosome V, between lag-2 and unc-46 [2/102 Unc nonLag recom- binants from apx-1(or3)/lag-2 unc46 did not pick up or3]. The alleles or15 and or22 were shown to be linked to chromosome V and failed to complement apx-1(or3). The three alleles result in indistinguish- able embryonic phenotypes (data not shown) and are completely penetrant for embryonic-lethality [e.g. 0/2340 embryos from 14 homozygous apx-1(or3) mothers hatched]. The apx-1 allele zu215 (C. Mello) fails to complement or3, and Mello et al., 1994, have reported a similar phenotypic description of apx-1 mutant alleles. apx-1/Df animals were made by mating apx-1/+ males into lag-2(q387)/DnT1 (dominant Unc) hermaphrodites. Instead of 1/2 of the nonUnc outcross hermaphrodite adults being apx-1/Df and producing dead embryos, only 1/5 of the nonUnc adults produced apx-1 mutant embryos (10/50). Other nonUnc animals died as larvae, most with apparently abnormal rectums and were presumably hemizygous for apx-1. Mutations in the apx-1 homologue lag-2 (Tax et al., 1994) cause rectal defects (Lambie and Kimble, 1991b), but the defects in hemizygous apx-1 animals appear different, often having anal openings connected to the posterior end of the intestine.

The monoclonal antibodies 5.6 and 9.2.1 (Miller et al., 1983), and an anti-intermediate filament antibody (Pruss et al., 1981), were used to detect the production of non-pharyngeal muscle cells, pharyngeal muscle cells, and pharyngeal marginal cells, respectively. The mon- oclonal antibody J126, from S. Strome, shows a staining pattern identical to that of the antibody 2CB7 described in Bowerman et al., 1992a (B. Bowerman and S. Mango, unpublished data) and was used to detect intestinal-rectal valve cells and pharyngeal gland cells. The monoclonal antibodies 3NB12, NE2/1B4 and NE8/46C.3 were isolated in a screen for cell-type-specific monoclonal antibodies (see Okamoto and Thomson, 1985). 3NB12 (Priess and Thomson, 1987) was used to detect pharyngeal muscle cells in Fig. 4, and in Fig. 6. NE2/1B4 (Okamoto and Thomson, unpublished data) was used to detect epidermal seam cells produced by ABa and ABp in apx-1(or3) embryos. NE8/46C.3 (Okamoto and Thomson, unpublished data) was used to stain and count the body wall muscle cells produced by P3 and C in apx-1(or3) embryos. Because the muscle cells often are in clusters, it is difficult to count cell numbers precisely. The monoclonal antibody OIC1D4 (from S. Strome) was used to detect P-granules. The promoter trap strain UL8 (Hope, 1991; Young and Hope, 1993) was used to score for the production of rectal-epithelial cells (Mango et al., unpublished data) by staining with X-GAL (Fire, 1992).

Cell lineage analysis was done as described previously (Sulston et al., 1983; Mango et al., 1991), using Nomarski optics with a video- recording system (Hird and White, 1993). Blastomere names reflect their lineage. For example, ABa and ABp are the anterior and posterior daughters, respectively, of the 2-cell stage blastomere AB. Laser ablations, embryo fixation and antibody staining procedures, microscopy, and photography were performed as described previously (Bowerman et al., 1992b).

To analyze the production of ABp-specific cell types in wild-type, apx-1, and pie-1 embryos (Fig. 2 and Tables 1 and 2), all blastomeres except ABp were killed in late 4-cell stage embryos, shortly before the next set of embryonic cleavages. This was done to allow any inter- action between P₂ and ABp to occur (one can prevent the P₂/ABp interaction only by killing P₂ immediately after it is born (see Bowerman et al., 1992a). The development of the blastomeres P₃, C, and EMS in apx-1(or3) embryos were analyzed by killing all other blastomeres besides P₂ or EMS at the 4-cell stage (and then one of the daughters of P₂ to analyze P₃ and C).

To test the ability of ABa and ABp to produce pharyngeal cells in response to a signal from MS in wild-type, apx-1, and pie-1 embryos (Fig. 3, Tables 1 and 2), all blastomeres other than either ABa and EMS, or ABp and EMS, were killed at the 4-cell stage, the E daughter of EMS was killed at the 8-cell stage, and the MS daughter of EMS was killed at the 12-cell stage, by which time one can no longer prevent the MS-dependent induction of ABa pharyngeal cells in wild- type embryos (Hutter and Schnabel, 1994; Mango et al., unpublished data). Laser ablation experiments to assay the production of induced pharyngeal cells by individual descendants of ABa and ABp are described in the legend to Table 1.

P₂ was removed from 4-cell stage wild-type embryos through a hole in the eggshell, according to S. Gendreau et al., manuscript in preparation. Embryos were allowed to develop 20 hours at 15°C, and then stained either with J126 (for intestinal-rectal valves), or first with 3NB12 (for pharyngeal muscle), and second with J126. In most embryos, different J126+ cell types could be distinguished by their different cellular morphology, making it possible to score for the presence of intestinal-rectal valve cells and intestine. We scored pharyngeal cell production only in embryos that made intestinal cells.

The ability of each ABp grand-daughter to produce epidermal seam cells (see Sulston et al., 1983) in apx-1(or3) embryos was analyzed by killing P₂ and EMS at the 4-cell stage, and all AB descendants but one ABp grandaughter at the 12-cell stage. After developing to the equivalent of hatching, the operated embryos were fixed and stained with the seam cell-specific antibody NE2/1B4 (see above). Neither ABpla or ABpra produced differentiated descendants that stained pos- itively (0/17 and 0/13 operated embryos, respectively). When ABprp was allowed to develop, only 4/27 operated embryos stained posi- tively. In contrast, 13/18 operated or3 embryos stained positively when ABplp was allowed to develop.

To identify genes required for the P₂/ABp cell interaction, we screened for maternal-effect, embryonic-lethal mutations that result in abnormal ABp development (see Materials and Methods). We isolated three recessive alleles that map to the left arm of chromosome V and fail to complement each other. These alleles define the gene apx-1 (anterior pharynx excess; also see Mello et al., 1994). Mutations in apx-1 are strictly maternal in that sperm from wild-type males do not rescue the embryonic lethality [0/431 apx-1(or3)/+ embryos hatched after mating +/+ males into homozygous apx-1(or3) mothers purged of their own maternal supply of sperm] and do not cause any apparent change in the phenotype of the mutant embryos (data not shown). While the frequency with which we isolated these alleles and their recessive nature argue that the mutations reduce gene function, they may not be null. Essentially all homozygous apx-1(or3) animals [from apx-1(or3)/+ parents] survive to adulthood (212/213) and produce dead embryos, but over 50% of hemizygous animals [apx-1(or3)/Deficiency] die during larval development (see Materials and Methods), possibly due to abnormalities in the structure of the rectum (data not shown). We see a low frequency (<1%) of similar rectal defects in animals homozygous for any of the three alleles we have isolated (data not shown). Hemizygous mothers that survive to adulthood produce dead embryos with a phenotype indistinguishable from that of embryos produced by homozygous apx-1(or3) mothers (as judged by morpho- logical criteria using Nomarski optics to score 50 mutant embryos for the presence of excess pharyngeal cells, as well as for the presence of neurons, epidermal cells, body wall muscles, intestinal cells, germ cells, and cell deaths). We conclude that the embryonic phenotype represents a loss of apx-1 function, but apx-1 may have additional function(s) late in embryogenesis or during larval development.

In a wild type embryo, the ABp blastomere produces two intestinal-rectal valve cells, three rectal-epithelial cells and four muscles (Sulston et al., 1983). To determine the fate of ABp in apx-1 embryos, we examined whether ABp from apx- 1 mutants produces these specific cell types. All blastomeres but ABp were destroyed with a laser microbeam late in the 4- cell stage. ABp was then allowed to develop to the equivalent of hatching and analyzed for the production of ABp-specific cell types by staining for cell-type-specific markers (see Materials and Methods). Whereas almost all ABp blastomeres from wild-type embryos produce intestinal-rectal valve cells, rectal epithelial cells and muscle cells, ABp from apx-1 mutants fails to produce each of these ABp-specific cell types (Fig. 2; Table 1). Thus, apx-1(+) activity is required for the proper development of the ABp blastomere.

To determine if blastomeres other than ABp develop normally in apx-1(or3) embryos, we examined the cell types produced by P₂, EMS and ABa. We used a laser ablation strategy similar to the one described above to isolate the devel- opment of each blastomere. To examine the fate of P₂ in apx- 1 embryos, we analyzed the development of its two daughters, P₃ and C. As in wild-type embryos (Sulston et al., 1983), P₃ produces about 20 body wall muscle cells: 18-21 muscle cells could be counted in each of 8/8 operated embryos stained with an antibody specific for body wall muscle. C, which normally produces epidermal cells and 32 body wall muscle cells (Sulston et al., 1983), also appears to develop normally in apx-1 embryos. In 6/6 operated apx-1 embryos, C produced epidermal cells (scored by their characterstic morphology using Nomarski optics) and body wall muscle cells (26-34 cells stained positively in each operated embryo). Similar numbers of body wall muscle cells were produced by P₃ and C in operated wild-type embryos (data not shown). We also analyzed the development of the P₂-derived germ-line precur- sors by staining a mixed-stage population of apx-1(or3) embryos with antibodies that recognize germ-line-specific P- granules (Strome and Wood, 1983). During wild-type embryo- genesis, P-granules are segregrated to P₂, and subsequently to two descendants of P₂ that give rise to the germ line post- embryonically (Strome and Wood, 1983). In apx-1(or3) embryos, P-granules segregation was normal (data not shown), and two germ-line precursors were present in all late stage embryos (n=163; scored by staining with anti- bodies that recognize P-granules). Thus the development of P₂ appears unaffected by mutations in apx-1. These results suggest that the altered development of ABp in apx-1 mutant embryos is not due to generalized defects in P₂ identity leading to a loss of P₂ signalling (see below).

We next examined the fate of the EMS blastomere, which normally produces pharyngeal cells, body wall muscles and intestine (Sulston et al., 1983). Similarly, in 24/24 operated apx-1(or3) embryos, EMS produced pharyngeal muscle and body wall muscle cells, as scored by antibody staining; and 26/26 operated embryos generated intestinal cells, as scored by their production of birefringent gut granules. Finally, we examined the development of ABa. We found that ABa from apx-1 embryos produces a normal pattern of pharyngeal cells (see below and Table 1) and special- ized epidermal cells called seam cells (5/5 operated embryos stained posi- tively for seam cells). These results suggest that the 4-cell stage blas- tomeres P₂, EMS, and ABa develop normally in apx-1 embryos.

Because ABa and ABp are born equivalent in a wild-type embryo, we Nomarski optics, it appeared likely that apx-1 mutant embryos produce extra pharyngeal cells, an observation confirmed by staining intact apx-1 embryos with an antibody that recognizes pharyngeal muscles (Fig. 4). Indeed, if all blastomeres other than ABp are destroyed with a laser microbeam, ABp, like ABa, can produce pharyngeal cells in apx-1 mutant embryos (Fig. 3).

In a wild-type embryo, ABa and one other 4-cell stage blas- tomere, EMS, produce pharyngeal cells (Sulston et al., 1983). Pharyngeal cells produced by ABa differ from those made by EMS with respect to the specific cell types made and the devel- opmental pathways used to generate them. For example, only EMS generates pharyngeal gland cells, while ABa and EMS both produce pharyngeal muscles and support cells called pha- wanted to test the possibility that in the absence of apx-1 function ABp might adopt a fate more like that of ABa. One important difference in the fates of these two blastomeres is that ABa, but not ABp, normally produces pharyngeal cells (Sulston et al., 1983). From visual inspection using ryngeal marginal cells (Sulston et al., 1983). In apx-1(or3) embryos, ABp produces muscle and marginal cells, but no gland cells, as does ABa from either wild-type or apx-1(or3) animals (Fig. 3). Moreover, approximately the same number of pharyngeal muscle cells are produced by ABp from apx-1(or3) as are made by ABa from either wild-type or apx-1(or3) embryos (12-16 cells, see Fig. 3). Thus, ABp produces the number and types of pharyngeal cells characterstic of ABa, not EMS.

Another characteristic feature of pharyngeal cell production by ABa is that it depends on cell signalling. Specifically, a signal from the MS daughter of EMS at about the 12-cell stage induces two grand-daughters of ABa to produce pharyngeal cells (see Fig. 7; Priess and Thomson, 1987; Hutter and Schnabel, 1994; Mango et al., submitted). Because of the require- ment for cell signalling, ABa does not produce pharyngeal cells if MS is destroyed at the 8-cell stage. However, laser ablation of MS at the 12-15 cell stage cannot block ABa from producing pharyngeal cells (Hutter and Schnabel, 1994; Mango et al., unpub- lished data). Similarly, the production of pharyngeal cells by ABp in apx- 1(or3) embryos depends on MS: neither ABa nor ABp produces pha- ryngeal muscles if MS is killed at the 8-cell stage (0/11 operated embyos stained positively with antibodies that recognize pharyngeal muscle), but they both produce pharyngeal cells if MS is killed at the 12-cell stage (Fig. 3 and Table 1). We conclude that ABp in apx-1 embryos depends on an interac- tion with MS to produce pharyngeal cells.

In a wild-type embryo, the interac- tion between MS and ABa descendants depends on glp-1, a putative receptor for cell signals (Austin and Kimble, 1987; Priess et al., 1987; Austin and Kimble, 1989; Yochem and Greenwald, 1989; Evans et al., 1994; Crittenden et al., in press). Like ABa in glp-1 mutant embryos, neither ABa nor ABp produce pharyngeal muscle cells in embryos doubly mutant for both apx-1 and glp-1 (Fig. 4). In summary, the production of pharyngeal cells by ABp in apx-1 embryos depends on the MS blastomere and on glp-1 activity, as does the production of pharyngeal cells by ABa in both wild-type and apx-1(or3) embryos. These data suggest that ABp in apx-1 mutants develops more like its sister ABa.

The observation that ABa and ABp in apx-1 mutants produce similar numbers and types of pharyngeal cells suggests that ABp descendants may adopt fates similar or even identical to those of ABa descendants in a wild-type embryo. To test this hypothesis, we observed the patterns of cell division and pro- grammed cell death produced by certain ABp descendants (see drawing in Fig. 7 showing the positions of ABp descendants). We found that two posterior ABp grand-daughters, ABplp and ABprp from apx-1(or3) embryos each generate a cell lineage pattern that is characteristic of the posterior wild-type ABa grand-daughter ABalp (Fig. 5). These results suggest that ABplp and ABprp in apx-1 embryos adopt ABalp-like identi- ties.

We also examined the patterns of differentiated pharyngeal cells produced by ABplp and ABprp from apx-1(or3) embryos. In a wild-type embryo, two ABa grand-daughters produce pha- ryngeal cells, ABalp and ABara. Both daughters of ABara generate pharyngeal cells, but only the anterior daughter of ABalp produces pharyngeal cells (Sulston et al., 1983). In apx- 1(or3) embryos, only the anterior daughters of ABplp and ABprp produce pharyngeal muscles, similar to ABalp in wild- type (Fig. 5 and Table 1). These patterns of cell lineage and terminal cell fate argue that the two ABp grand-daughters ABplp and ABprp in apx-1 embryos recapitulate the develop- ment of the ABa grand-daughter ABalp. We also note that because the cell lineages followed correspond to a portion of the ABalp lineage that does not generate pharyngeal cells, the similarity between ABplp and ABprp in apx-1 mutants and ABalp in wild-type embryos extends beyond the production of pharyngeal cells.

The failure of ABp to produce its characteristic cell types in apx-1 embryos suggests that the P₂/ABp interaction, shown previously to be required for the proper development of ABp (Bowerman et al., 1992a), may be defective in apx-1 mutants. Based on our analysis of the apx-1 phenotype, we suggest that in the absence of P₂/ABp signalling, the equivalence between ABa and ABp is not broken and therefore ABp develops more like ABa. As an independent test of this model, we examined the development of ABp in pie-1 mutant embryos. In pie-1 mutants, P₂ appears to adopt a fate identical to its sister EMS (Mello et al., 1992), thereby ‘genetically ablating’ the P₂ blas- tomere. Consistent with a loss of P₂ identity resulting in a loss of P₂/ABp signalling, ABp in pie-1 embryos appears to develop similarly to ABa in wild-type embryos: neither intesti- nal-rectal valve cells nor muscle cells are produced by ABp, which instead makes pharyngeal cells in response to inductive signals from MS (Table 2). These results support a model in which signalling from P₂ breaks the equivalence of ABa and ABp.

One prediction of our P₂/ABp interaction model is that removal of P₂ from wild-type embryos should result in the wild-type ABp blastomere producing pharyngeal cells. To test this pre- diction, P₂ was physically removed from wild-type embryos 4-5 minutes after the P₁ blastomere divided into P₂ and EMS. These operated embryos fail to produce intestinal-rectal valve cells, indicating that the interaction between P₂ and ABp was prevented (3/3, data not shown). Significantly, these operated embryos often produce extra pharyngeal muscle cells (7/10; Fig. 6). The P₂/ABp interaction appears to occur early in the cell cycle, as operated embryos rarely produce extra pharyn- geal cells if P₂ is removed more than 5 minutes after the division of P₁ (1/6; Fig. 6). We suggest that after removal of P₂ in wild-type embryos, ABp develops like its sister ABa.

We have shown that two maternal genes, apx-1 and pie-1, are required for proper development of the ABp blastomere. In apx-1 and pie-1 mutants, ABp fails to produce its characteris- tic cell types and instead develops more like its sister ABa, producing pharyngeal cells in response to an inductive signal from MS. We can phenocopy the apx-1 and pie-1 production of extra pharyngeal cells by removing the P₂ blastomere from wild-type embryos. Therefore, we propose that the alteration in ABp fate observed in apx-1 and pie-1 embryos reflects a block in P₂/ABp signalling at the 4-cell stage. We suggest that normally in embryogenesis, the P₂/ABp interaction breaks the initial equivalence of ABa and ABp, leading to the production of ABp-specific cell types and rendering ABp incapable of producing pharyngeal cells in response to the subsequent signal from MS (see Fig. 7). While the effect of pie-1 on the P₂/ABp interaction is probably indirect, due to the changed identity of P₂, we suggest that apx-1 participates directly in the P₂/ABp signalling pathway.

The apx-1 locus could play one of two possible roles during ABp specification. First, apx-1 could encode a component of the P₂/ABp signalling pathway. Alternatively, apx-1 could code for a localized factor, or determinant, that contributes to the proper specification of ABp (for a review, see Wood and Edgar, 1994). Given that ABa and ABp are born developmen- tally equivalent, mutations in such a ‘determinant of AB fate’ would be expected to affect both ABa and ABp. Because Aba development appears wild-type in apx-1 mutants, we favor the first hypothesis, that apx-1 is involved in P₂/ABp signalling. Moreover, the apparently normal development of P₂ in apx-1 embryos suggests that apx-1 functions directly in the P₂/ABp signalling pathway, and not in other processes that might indi- rectly affect signalling by altering the identity of P₂. These arguments are strengthened by recent sequence information that shows apx-1 encodes a predicted transmembrane protein and is similar in sequence to Delta and Serrate in Drosophila and lag-2 in C. elegans (Mello et al., 1994). These three genes encode candidate ligands for Notch-like receptors (Artavanis- Tsakonas et al., 1991; Tax et al., 1994; Henderson et al., unpublished data). The glp-1 gene is a Notch homologue (Austin and Kimble, 1989; Yochem and Greenwald, 1989), and is required for the P₂/ABp interaction (Mello et al., 1994; Hutter and Schnabel, 1994; Moskowitz et al., unpublished data). Therefore, by analogy, apx-1 may be a P₂-produced ligand for the glp-1 receptor.

Two types of intercellular interactions have been shown to regulate C. elegans development: lateral signalling and inductive interactions (for a review, see Lambie and Kimble, 1991a). In lateral signalling, cells of equivalent developmental potential interact with each other to determine their fates. By contrast, inductive interactions involve non-equivalent cells, one signalling and the other receiving. The P₂/ABp interaction discussed here is an example of an inductive interaction, in which two equivalent cells, ABa and ABp, become different due to an interaction with a third non-equivalent cell, P₂. Because interactions that rely on lateral signalling can result in variable fate choices by individual cells (e.g. the AC/VU decision; see Kimble, 1981), it may be that the P₂/ABp inductive interaction prevents variability in the fates adopted by ABa and ABp.

In addition to the P₂ and MS inductions discussed here, other interactions may also be necessary to specify the different fates of ABa and ABp. Consistent with this idea, cell lineages of the two anterior ABp grand-daughters (ABpla and ABpra) in apx- 1(or3) embryos did not appear to recapitulate ABa cell fates (S. Mango and B. Bowerman, unpublished observations). Specifically, the ABp descendant ABplapapaa in apx-1(or3) embryos did not appear to undergo programmed cell death (n=4), in contrast to the lineally equivalent descendant of ABa in wild-type embryos, ABalapapaa (see Sulston et al., 1983). Nor does ABpla produce pharyngeal cells or epidermal seam cells, as do other ABa descendants. Furthermore, although another ABp descendant, ABprp, generates a cell lineage pattern characteristic of ABarp, in the absence of MS induction (Mello et al., 1994), we have found that ABprp fails to produce at least one differentiated marker for ABarp fate (epidermal seam cells; see Materials and Methods). Thus, while apx-1 function, and by extension the P₂ interaction, appears to play a critical role in breaking the initial equivalence of ABa and ABp, additional interactions may influence the ultimate fates of ABa and ABp descendants.

The family of proteins with homology to glp-1 and Notch are required for multiple cell interactions to specify many different cell fates (Austin and Kimble, 1987; Priess et al., 1987; Artavanis-Tsakonas et al., 1991). How do these proteins par- ticipate in so many different cell fate decisions? For example, how can ABa and ABp, which presumably are born with identical intracellular components, both use glp-1 to specify completely different developmental pathways? One possibility is that P₂ and MS produce distinct ligands that interact with glp-1 to trigger different downstream effects. Alternatively, the P₂ and MS inductions may act through distinct receptors that also require a common glp-1 signalling pathway to specify different fates. The Delta-like sequence of apx-1, and the observation that apx-1 is required for P₂ but not MS signalling, are consistent with either of these possibilities. Alternatively, P₂ and MS could send qualitatively identical signals, with ABa and ABp descendants responding differently depending on the time at which the signal is received. For example, a highly labile product that influences the response to the P₂ signal could be present in both ABa and ABp at the 4-cell stage when P₂ signalling occurs, but decay before the MS induction at the 12- to 15-cell stage. In this case, the specificity of apx-1 for the P₂/ABp interaction may reflect a requirement for precise temporal and spatial control of the expression of a signalling molecule, and not a requirement for qualitatively different signals.

During wild-type embryogenesis, the P₂ blastomere is required for two inductive events: the P₂/ABp interaction discussed here, and an interaction between P₂ and EMS that is required for EMS to produce intestinal cells (Goldstein, 1992). Both interactions occur early in the 4-cell stage (Goldstein, 1992; Bowerman et al., 1992a; this work), and it is conceivable that P₂ might produce one signal that is interpreted differently by ABp and EMS. This scenario appears unlikely, however, as these two interactions are separable. In both apx-1 and pie-1 mutants, ABp development is disrupted, but EMS still generates its normal complement of intestinal cells. Moreover, after P2 removal, the operated wild-type embryos often produce intestinal cells, but not the ABp-derived intestinal- rectal valve cells. Therefore, we suggest that signalling between P₂ and ABp utilizes a different mechanism than does signalling between P₂ and EMS.

The orientation of the early embryonic cleavages and the rigidity of the chitinous eggshell in C. elegans result in ABa and ABp occupying anterior and dorsal positions, while the sisters P₂ and EMS occupy posterior and ventral positions, respectively (see Hyman and White, 1987, and Fig. 7). Thus, at the 4-cell stage, P₂ touches ABp, but not ABa. Because of this geometry, an inductive signal from P₂ that depends on cell contact would influence only ABp, providing a simple mechanism for breaking the ABa/ABp equivalence. The extent of contact between different blastomeres may be important at the 12-cell stage as well, when ABa descendants that are induced to produce pharyngeal cells have broad contacts with MS, whereas ABa descendants that do not produce pharyngeal cells have little or no contact. Similarly in apx-1 embryos, only the two ABp grand-daughters that make extensive contact with MS are induced to produce pharyngeal cells. Another correla- tion between cell contact and the induction of AB descendants is observed in pie-1 mutant embryos. In pie-1 embryos, a daughter of P₂ (called C) develops like a wild-type MS blas- tomere (Mello et al., 1992). Consistent with C adopting an MS identity early in development, it can sometimes induce descen- dants of ABp, which touch C, to produce pharyngeal muscle cells. However, C cannot induce descendants of ABa, which do not contact C, to produce pharyngeal cells (see Table 2). These data argue that cell contact is necessary for the inductive interactions of P₂ and MS. Thus, in the early C. elegans embryo, as in the early Drosophila embryo, cell position appears to be critical for specifying cell fate. In Drosophila, the embryonic axes are specified in a syncytium, and identity appears to be determined largely by position within gradients of Bicoid, Nanos and nuclear Dorsal proteins (Driever, 1993; Chasan and Anderson, 1993; St. Johnston, 1993). In C. elegans, cell position is critical for determining cell contacts that may delimit which cells participate in different inductive interactions. Together, the early positioning of blastomeres and the inductive interactions that follow are crucial for properly establishing the nematode body plan.

Two exceptions to the correlation between cell contact and signalling are observed with MS and ABp descendants. First, the ABp grand-daughters ABplp and ABprp have extensive contacts with MS, but as yet MS has not been found to influence their wild-type development (e.g. Bowerman et al., 1992a). Thus apx-1(+) activity may render ABplp and ABprp completely insensitive to the MS signal. Second, ABpra depends on MS signalling to develop properly (Hutter and Schnabel, 1994), but makes relatively little contact with MS. These less extensive contacts may be sufficient for contact- mediated cell signalling. Alternatively, it is possible that ABpra is influenced by a diffusible MS signal, or that the effects of MS on ABpra are indirect.

The first two authors contributed equally to the experiments described. We are grateful to S. Gendreau for removing P₂ from wild- type embryos. We thank J. Ahringer, J. Priess, D. Miller and S. Strome for antibodies; the I. Hope, J. Priess and J. Thomas laboratories, and the Caehorhabditis Genetics Center at the University of Minnesota, for providing strains; J. Rothman for use of his videomicroscopy equipment; C. Mello, B. Draper, J. Priess, H. Hutter, R. Schnabel, I. Moskowitz, S. Gendreau and J. Rothman for communicating unpub- lished results; S. Crittenden, J. Eisen, T. Evans, B. Pickett, H. Roehl and Chuck Kimmel’s Bi 410/510 students at the University of Oregon for comments on the manuscript; and J. Kimble for her generosity. This work was supported by a training grant from the NIH (C. J. T.); a fellowship from the NIH (S. E. M.); and grants from the Medical Research Fund of Oregon and the American Cancer Society (B. B.).