Establishment of POP-1 asymmetry in early C. elegansembryos

The pattern of cell cleavage and differentiation in C. elegans is largely invariant (Sulston et al.,1983; Schnabel et al.,1997). The spindles of most cells are oriented such that mitosis generates an anterior daughter and a posterior daughter, and most of the early anterior/posterior (a/p) cell divisions result in sister cells with different fates. Although these fates are determined through diverse pathways, several studies have suggested that cells throughout the embryo recognize a common a/p coordinate system in choosing their fates(Mello et al., 1992; Hutter and Schnabel, 1994; Mango et al., 1994; Moskowitz et al., 1994; Way et al., 1994). Recent studies have provided considerable evidence for this hypothesis, and have begun to identify molecular components of the a/p system(Lin et al., 1995; Lin et al., 1998; Kaletta et al., 1997; Meneghini et al., 1999; Rocheleau et al., 1999; Rocheleau et al., 1997; Thorpe et al., 1997; Maduro et al., 2002). The transcription factor POP-1 is present in the nuclei of all early embryonic,and several postembryonic, cells. After an embryonic cell divides, its anterior daughter invariably shows a higher level of nuclear POP-1 than its posterior daughter. Thus each a/p pair of sister cells exhibits POP-1 asymmetry with a reproducible high/low polarity. Conditions that result in high POP-1 in the posterior sister cause an anterior transformation in fate,whereas the absence of POP-1 in the anterior sister causes a posterior transformation in fate. POP-1 functions in a variety of cell-type decisions,including mesodermal/endodermal choices and epidermal/neuronal choices. These observations suggest that POP-1 asymmetry provides the a/p coordinate system that collaborates with more widely expressed transcription factors to diversify sister cells.

POP-1 is related to TCF/Pangolin, a transcriptional effector of the canonical Wnt signaling pathway, and POP-1 has been shown to function in canonical Wnt signaling during larval development(Lin et al., 1995; Herman, 2001). However, POP-1 asymmetry in the early embryo is regulated by a non-canonical Wnt pathway,with parallel input from a mitogen-activated protein kinase (MAPK) pathway(for a review, see Korswagen,2002). Components of these pathways include MOM-2/Wnt,MOM-5/Frizzled, WRM-1/beta-catenin, MOM-4/MAPKKK/TAK1 and LIT-1/Nemo. Sister cells show POP-1 asymmetry because they differ in their nucleo/cytoplasmic distributions of POP-1 (Maduro et al.,2002). Studies with cultured vertebrate cells suggest that WRM-1/beta-catenin can activate LIT-1/Nemo, resulting in phosphorylated POP-1 that accumulates in the cytoplasm(Rocheleau et al., 1999).

How is the a/p polarity system established? The most detailed experimental studies to date have focused on the development of the EMS cell(Goldstein, 1992; Goldstein, 1993). EMS divides into an anterior, mesodermal precursor and a posterior, endodermal precursor. This a/p polarity is induced during the 4-cell stage of embryogenesis by a neighboring cell called P₂. For example, removing P₂causes both EMS daughters to have anterior fates (high POP-1) and repositioning P₂ on the opposite surface of EMS reverses the polarity of the division.

Relatively little is known about the cellular events that establish a/p polarity in other embryonic cells. At the 2-cell stage, the posterior cell is called P₁ (the parent of P₂ and EMS) and the anterior cell is called AB (see Fig. 1). Within the AB lineage, POP-1 asymmetry is first detectable after the third division of AB, when there are four a/p sister pairs of AB descendants(Lin et al., 1998); for convenience we refer to this stage as the AB⁸ stage. POP-1 function is essential for a/p differences in cell fate within each of the four sister pairs of AB⁸ cells (Lin et al.,1998). Previous studies on how a/p differences are generated in the AB lineage have reached contradictory conclusions. In one set of experiments, AB was separated from P₁ and allowed to develop to the AB¹⁶ stage (Wittmann et al.,1997). As many as eight of the AB¹⁶ cells expressed a transgenic marker that normally is expressed in the eight posterior AB¹⁶ cells, suggesting that AB has an inherent a/p polarity. In a different study, videomicroscopy was used to follow AB development after killing P₁ or P₁ descendants(Hutter and Schnabel, 1995). Several AB⁸ cells showed posterior to anterior transformations in fate after killing P₁, but not P₁ descendants,suggesting that P₁ induces an a/p polarity in AB that is maintained in a latent form until the AB⁸ stage.

To further analyze the cellular basis for a/p polarity in the AB lineage,we analyzed POP-1 levels directly by immunostaining isolated and cultured embryonic cells. Our results indicate that POP-1 asymmetry at the AB⁸ stage results from interactions with specific P₁descendants, rather than with P₁. These interactions are mediated in part by MOM-2/Wnt signaling. Surprisingly, by the AB¹⁶ stage embryonic cells have acquired an ability to generate POP-1 asymmetry that appears to be independent of MOM-2/Wnt signaling or prior interactions with other cells, but that requires MOM-5/Frizzled.

Nematodes were cultured as in Brenner(Brenner, 1974). The wild-type Bristol strain N2 was used. The following mutations were used. LGI, mom-5(or57), dpy-5(e61), mom-4(or39) and unc-13(e1091); LGV, mom-2(or42) and dpy-11(e1180); LGX, mom-1(or10) and unc-6(n102). The following integrated transgene containing green fluorescent protein (GFP) reporter was used: zuIs3(end-1::GFP) (Nance and Priess,2002).

Individual embryonic cells were isolated from devitellinized embryos by gently drawing embryos in and out of a drawn-out capillary needle as described previously (Edgar, 1995). Cells were cultured in medium consisting of 5% L-15 (Gibco), 10% fetal calf serum(Gibco), and 4.7% sucrose, adjusted to 320-330 mOsm. To devitellinize embryos,eggs were placed in hypochlorite solution [6% NaOCl, 2.5 N KOH] on an inverted microscope slide for 3.5 minutes, then rinsed three times with 0.25 M HEPES,pH 7.0 before transfer to culture medium.

Isolated cells were micropipetted directly into a drop of fixative [2%paraformaldehyde, 60 mM PIPES, 25 mM HEPES (pH 6.8), 10 mM EGTA, 2 mM MgCl₂] on a poly-L-lysine (Sigma)-coated glass slide. After five minutes, excess fixative was removed and slides were placed in -20°C acetone for 5 minutes. Cells were rinsed twice in Tris-Tween [100 mM Tris-HCl(pH 7.5), 200 mM NaCl, 0.1% Tween], then incubated with 10% normal goat serum(Gibco) in Tris-Tween at room temperature for 30 minutes. Cells were then incubated overnight at room temperature with primary antisera. Fixation and immunostaining of intact embryos was as described previously(Lin et al., 1998). The following dilutions of antibodies/antisera were used: anti POP-1 [1:2500 mouse mABRL2 (P4G4) (Lin et al.,1998)]; midbody staining [1:50 rabbit anti-PKL-2 (kindly provided by M. Land)]; P-granules [1:17000 rabbit anti-PGL-1(Kawasaki et al., 1998)]. Secondary antibodies were conjugated to either Cy-3 (Jackson ImmunoResearch Laboratories) or FITC (Tago). Cells were stained with DAPI(4,6-diamidino-2-phenylindole) at 60 ng/ml for five minutes. POP-1 was scored only when the PKL-2 staining pattern provided an unambiguous indication of sister pairs, thus some results are scored by the number of sister pairs rather than by the number of experiments.

Standard techniques were used to synthesize double-stranded RNA (dsRNA)from T7 promoter-tagged, PCR-amplified genomic DNA for mom-2, mom-5,goa-1, and gpa-16. PCR primers were chosen to span exons and generate fragments between 0.5 and 2 kb in size. L4 or young adult hermaphrodites were soaked overnight with dsRNA(Tabara et al., 1998).

The first division of the fertilized egg results in an anterior cell called AB and a posterior cell called P₁(Fig. 1). Each of the descendants of these cells has a distinct name in C. elegansnomenclature, however it is convenient here to refer to the AB descendants collectively. For example, there are two AB daughters in a 4-cell embryo, so we refer to these daughters as the AB² cells and to the developmental stage as the AB² stage. The descendants of the P₁ cell are indicated by their standard names such as MS or E. Previous reports have analyzed POP-1 asymmetry in all cells up to the 28-cell stage, and POP-1 asymmetry has been observed in E descendants and a few MS descendants in later embryos (Lin et al.,1998; Maduro et al.,2002). To identify sister cells in older embryos and in cell culture experiments, we used a POP-1 antibody in conjunction with an antiserum against the kinase PKL-2. This antiserum stains the midbody, or cell-division remnant, between sister cells (small arrows in Fig. 2B) (M. Land and C. S. Rubin, unpublished). Using fixed and immunostained embryos at various stages,we identified each of the first 32 descendants of the MS cell, all of the AB⁸, AB¹⁶ and AB³² sister pairs, and a large subset of the AB⁶⁴ and AB¹²⁸ sister pairs(Fig. 2A,B and Materials and Methods). For each sister pair, POP-1 was high in the anterior sister and low in the posterior sister, suggesting that POP-1 asymmetry is reiterated with the same polarity after most, if not all, a/p divisions within the embryo.

In normal embryogenesis, POP-1 asymmetry is first evident in AB descendants at the AB⁸ stage (Lin et al.,1998). To examine POP-1 in cultured cells, we removed the eggshell from 2-cell stage embryos and allowed the devitellinized embryos to develop to the AB⁸ stage (see Fig. 1 for a comparison of cell positions in normal and devitellinized embryos). POP-1 asymmetry was observed between most sister pairs of AB⁸ cells in devitellinized but otherwise intact embryos (78%, 32 sister pairs scored) and in embryos in which AB and P₁ were separated then immediately recombined (78%, 32 sister pairs). POP-1 asymmetry was not detectable between any sister pairs of AB⁸ cells when AB was kept separate from P₁ (0%, 56 sister pairs scored; Fig. 3A,B) or when AB was allowed to contact P₁ only through the first cell cycle (0%, 44 sister pairs). These results indicate that signaling from P₁ is not sufficient for POP-1 asymmetry between sister pairs of AB⁸ cells,and instead suggest that P₁ descendants influence POP-1 asymmetry.

In C. elegans and other animals, Wnt proteins are secreted and can signal non-adjacent target cells (Whangbo et al., 2000; Zecca et al.,1996; Neumann and Cohen,1997). Because P₂, a daughter of P₁,expresses MOM-2/Wnt and signals EMS, in principle P₂ might signal the AB descendants. However, a previous study showed that P₂ is not essential for POP-1 asymmetry in the AB⁸ cells(Lin et al., 1998). We have confirmed this result using our modified conditions of cell isolation and culture; most sister pairs of AB⁸ cells show POP-1 asymmetry after P₂ is removed from a 4-cell embryo (85%, 48 sister pairs). Moreover, P₂ is not sufficient for POP-1 asymmetry; although P₂ normally contacts one of the AB² cells, called ABp,the daughters of ABp do not exhibit POP-1 asymmetry.

Why does P₂ signaling fail to alter POP-1 levels in the ABp daughters? Because P₂ has been shown to interact with ABp, but not EMS, through a separate, Notch-related signaling pathway, we wondered whether this pathway precluded POP-1 asymmetry. However, embryos that were depleted of the receptor GLP-1/Notch by glp-1(RNAi) did not show POP-1 asymmetry at the ABp division, similar to wild-type embryos (0/5). A second possibility is that the transverse divisions of the AB² cells, perpendicular to P₂, prevent POP-1 asymmetry. A previous study noted that EMS may fail to respond to induction if it divides perpendicular to, rather than in line with, P₂, although this analysis is complicated by the fact that signaling from P₂ normally orients the EMS division(Goldstein, 1995). To examine whether division orientation influenced POP-1 asymmetry, we used RNAi to inhibit the functions of the G alpha proteins encoded by goa-1 and gpa-16 (see Materials and Methods); such embryos appear to have random spindle orientations (Gotta and Ahringer, 2001; Zwaal et al.,1996). We observed that a/p, but not transverse, divisions of EMS resulted in POP-1 asymmetry (6/6 and 0/5, respectively; Fig. 4A). Similarly, a/p, but not transverse, divisions of the AB² cells resulted in POP-1 asymmetry (4/4 and 0/4, respectively; Fig. 4A). When the functions of goa-1 and gpa-16 were inhibited in mom-2(or42) mutant embryos, POP-1 asymmetry was not detectable after either an a/p, or transverse, division of the AB²daughters (0/12 and 0/6, respectively). As a second method for altering the AB² division axis, a laser microbeam was used to fuse the AB² daughters together immediately after their birth. As a fused AB² cell enters mitosis, it often develops two largely separate spindles that orient a/p. After the tetrapolar division, the posterior daughters usually showed lower levels of nuclear POP-1 than the anterior daughters (16/18 sister pairs; Fig. 4C).

We used devitellinized embryos to further examine interactions between P₂ and the AB² cells. In devitellinized embryos, as in normal embryos, the division axes of the AB² cells are oriented transversely with respect to the position of EMS (asterisks in Fig. 1). We found that the AB² divisions remained approximately transverse after P₂was removed (32/32 embryos), or after a graft P₂ was placed on various surfaces of one or both AB² cells (22/22 embryos). We conclude that the presence or absence of P₂ does not significantly alter the transverse division axis of an AB² cell. When a graft P₂ was positioned perpendicular to the predicted AB²division axis (short red arrow in Fig. 1), POP-1 asymmetry was not observed between either pair of AB² daughters (0/8 embryos). In the next set of experiments the graft P₂ was placed in line with the predicted division axis of one of the AB² cells (long red arrow in Fig. 1); we refer to this experimental paradigm as the AB² polarization assay. In each of these experiments, the AB² cell in contact with the graft P₂ divided into daughters with POP-1 asymmetry (6/6; Fig. 4E,F). The daughter distal to P₂ had high POP-1 and the proximal daughter had low POP-1,similar to the high/low POP-1 polarity that P₂ normally induces in the distal/proximal daughters of EMS. The AB² cell that was not in contact with the graft P₂ divided into daughters lacking POP-1 asymmetry (6/6). In a reciprocal experiment, a pair of AB² cells was isolated and grafted onto the P₂ cell of an intact,devitellinized host embryo (position indicated by blue arrow in Fig. 1). In this configuration,the P₂ cell of the host embryo is in contact simultaneously with the host EMS and one of the graft AB² cells, analogous to the pattern of cell contacts that P₂ makes in normal 4-cell embryos. In each case in which the graft AB² cell divided in line with P₂, the daughters of that AB² cell showed high/low POP-1 polarity with respect to P₂ (5/5). Taken together, these results suggest that ABp in normal embryogenesis has the potential to respond to signaling from P₂, and that interactions between P₂ and EMS do not preclude P₂ from simultaneously signaling ABp. Instead,the transverse division of ABp appears to prevent it from responding to P₂, such that POP-1 remains high in both ABp daughters.

The AB² polarization assay described above (long red arrow in Fig. 1) was used to examine the signaling properties of several cells in the early embryo. We found that graft AB², AB⁴, AB⁸, P₁ and EMS cells could not induce POP-1 asymmetry in this assay(Table 1). In contrast, the P₁ descendants E, C and P₃ appeared equivalent to P₂ in their ability to induce POP-1 asymmetry with high/low polarity (Table 1). The P₁ descendant MS usually induced POP-1 asymmetry, however the asymmetry was markedly less than that observed with the other P₁descendants (data not shown).

Previous experiments by others have demonstrated that a signal from P₂ aligns the EMS division axis and induces the proximal EMS daughter (E) to lengthen its cell cycle and to undergo endodermal differentiation (Goldstein,1992; Goldstein,1993). To further compare the signaling properties of MS, E and C with P₂, each of these cells was grafted onto an isolated EMS cell. Similar to P₂, both E and C were able to align the EMS division axis and to induce the proximal EMS daughter to lengthen its cell cycle and to undergo endodermal differentiation (6/7, 8/8 and 8/8 experiments,respectively; Fig. 5E,F). MS was unable to align the EMS spindle and did not induce endoderm-specific differentiation (0/9 experiments). We conclude that E, C and P₂,and to a lesser extent MS, can induce POP-1 polarity, and that the signaling properties of E and C appear identical to those reported previously for P₂.

MS, E and C are born during the AB⁴ stage and make extensive cell contacts with AB descendants (see below). Thus interactions at the AB⁴ stage might generate POP-1 asymmetry at the AB⁸stage. We found that the daughters of an isolated AB⁴ cell did not show POP-1 asymmetry (Table 1). However, when an MS, E or C cell was grafted onto an isolated AB⁴cell, the AB⁴ daughters showed high/low POP-1 polarity in almost all cases (Fig. 5G,H; Table 1). Surprisingly, in each experiment the spindle of the AB⁴ cell appeared to align with the signaling cell, in contrast to the spindles of AB² cells. Grafting an AB⁴ cell onto a second AB⁴ cell never resulted in POP-1 asymmetry (Table 1), and these cells showed only infrequent, and probably random, alignment of their spindles (7/38 had aligned spindles). Thus, MS, E and C have the ability to align the AB⁴ spindles and to induce high/low POP-1 polarity in the AB⁴ daughters. The ability of AB descendants to respond to signaling persists until at least the AB¹⁶ stage; isolated AB¹⁶ cells that were combined with C divided into daughters with high/low POP-1 polarity relative to C(Table 1).

Videomicroscopy was used to examine the cell contacts and spindle orientations of the four AB⁴ cells in live embryos (n=12);these cells are named ABal, ABar, ABpl and ABpr. ABal contacts only MS, and its spindle normally orients toward MS. ABpr and ABpl contact MS, E and C simultaneously, however their spindles orient toward E and often move slightly toward E before cell division (Fig. 5A,C). Finally, ABar contacts C and MS, and its spindle is oriented approximately between C and MS. The observation that three of the AB⁴ cells contact multiple signaling cells simultaneously suggests that the AB⁴ cells distinguish, or integrate, the various signals. We used devitellinized 8-cell embryos to test whether an AB⁴ cell could distinguish between signals from MS and E. In devitellinized embryos,the spindles of the AB⁴ cells usually orient toward MS (see Fig. 1), and the AB⁴daughters that show POP-1 asymmetry invariably have high/low POP-1 polarity with respect to the position of MS. We therefore grafted an E cell onto one of the AB⁴ cells of a host embryo at a site either opposite, or orthogonal, to the site of contact with MS (yellow arrow in Fig. 1). In each of six experiments, the spindle of the AB⁴ cell aligned toward the graft E, rather than the host MS, and divided with high/low POP-1 polarity relative to the graft E (Fig. 5I,J). We conclude that signaling from E predominates over signaling from MS. Similarly,we infer that signaling from C may predominate over MS signaling in normal development because POP-1 polarity in ABar is high/low relative to C, rather than MS.

Consistent with previous results by others, we found that mom-2mutants at the AB⁸ stage had a partial loss of POP-1 asymmetry between sister pairs of AB⁸ cells(Table 2) (see also Lin et al., 1998; Meneghini et al., 1999). To examine the role of MOM-2/Wnt in signaling, we isolated individual cells from mom-2(or42) mutant embryos to use as graft cells in the AB² polarization assay with a wild-type host embryo. Graft P₂ or C cells from a mom-2(or42) mutant were unable to induce POP-1 asymmetry in this assay, whereas the MS cell sometimes induced weak POP-1 asymmetry (Table 1;the E cell adopts an MS-like fate in mom-2 mutants, and was not tested separately). We next grafted either a C or MS cell from a mom-2(or42) mutant onto an isolated, wild-type AB⁴ cell. Similar to the above results, C did not align the AB⁴ spindle (0/6)and did not induce POP-1 asymmetry between the AB⁴ daughters(Table 1). In contrast, MS aligned the AB⁴ spindle (15/15) and induced high/low POP-1 polarity with either a wild-type or reduced level of POP-1 asymmetry(Fig. 5G and Table 1). We conclude that MOM-2/Wnt is essential for signaling from C, but not MS. As described above,in wild-type development one AB⁴ cell (ABar) contacts MS and C simultaneously, and the ABar daughters show high/low POP-1 polarity with respect to C. In mom-2(or42) mutant embryos, POP-1 asymmetry often was not detectable between the daughters of ABar (7/13 embryos). However, when present, the POP-1 polarity was reversed and oriented high/low with respect to MS (6/6). Thus, MS, rather than C, may orient the polarity of the ABar division in mom-2 mutant embryos.

Previous studies have described defects in the orientation of the ABar spindle in mom-2 mutant embryos(Rocheleau et al., 1997; Thorpe et al., 1997). We found that all of the AB⁴ cells have variable defects in spindle orientation (Fig. 5B,D). For example, the ABpl spindle often aligned toward MS or midway between MS and E(Fig. 5D), whereas a wild-type ABpl spindle aligns toward E (Fig. 5C). When an MS, E or C cell from a mom-2(or42) embryo was grafted onto an isolated AB⁴ cell from a second mom-2(or42) embryo, in nearly all cases the AB⁴ spindle aligned toward MS or E (7/7 and 6/7, respectively) but only rarely aligned toward C (2/13). Thus, the AB⁴ spindle defects in mom-2mutant embryos appear to result from the absence of a signal from C, and from an inability to discriminate between competing signals from MS and E.

The above results provide evidence that P₁ descendants play an important role in generating POP-1 asymmetry between the AB⁸ sister pairs, a stage when POP-1 plays a critical role in specifying a/p differences in cell fate. Surprisingly, we found that if an isolated AB cell was allowed to develop to the AB¹⁶ or AB³² stages, most of the resulting sister pairs showed POP-1 asymmetry (80%, 45 sister pairs and 97%,107 sister pairs, respectively; Fig. 3C,D). Although POP-1 asymmetry was never observed between the daughters of an isolated EMS cell (0%, 8 sister pairs), we often observed POP-1 asymmetry between the granddaughters and great-granddaughters (56%, 18 sister pairs and 95%, 40 sister pairs, respectively). Therefore AB and EMS descendants after the AB⁸ stage have an ability to generate POP-1 asymmetry that appears to be independent of prior exposure to the signaling cells P₂, MS, E, P₃ or C.

We wanted to determine whether POP-1 asymmetry in sister pairs of AB¹⁶ cells required signaling between AB descendants. For these experiments we separated AB from P₁, then separated each of the successive descendants of AB immediately after each cell division. The sequentially isolated cell was allowed to divide one additional time before staining the resulting sister pair for POP-1. As expected, POP-1 asymmetry was not observed between sequentially isolated sister pairs of AB⁴ or AB⁸ cells (0/11 and 0/10, respectively). In contrast, POP-1 asymmetry was usually present between sequentially isolated sister pairs of AB¹⁶ and AB³² cells (18/35 and 29/29, respectively; Fig. 3E,F).

Intact mom-2(or42) mutant embryos show a variable reduction or loss of POP-1 asymmetry between the EMS daughters and between sister pairs of AB⁸ cells (Table 2). However, mom-2(or42) mutants and mom-2(RNAi) embryos analyzed after the AB⁸ stage showed strong POP-1 asymmetry between most sister pairs of AB or EMS descendants(Fig. 2C,D; Table 2 and data not shown). POP-1 polarity was abnormal in these mutants, for example, many transverse sister pairs showed POP-1 asymmetry, and some a/p sister pairs had low/high polarity. We sequentially isolated AB⁸ cells from mom-2(or42) embryos and from mom-1(or10); mom-2(RNAi) double mutant embryos, then allowed these cells to divide once or twice in culture before immunostaining for POP-1. The mom-1 gene encodes the only C. elegans protein related to Prc, and should thus be required for secretion of MOM-2 and other Wnt family members(Kadowaki et al., 1996; Rocheleau et al., 1997). We found that the daughters and granddaughters of the sequentially isolated AB⁸ cells usually showed robust POP-1 asymmetry(Table 3). Thus, MOM-2/Wnt is not essential for POP-1 asymmetry after the AB⁸ stage, but has a role in POP-1 polarity.

We found that neither the daughters nor granddaughters of a sequentially isolated AB⁸ cell from a mom-5(or57) mutant embryo showed POP-1 asymmetry (Table 3). Similarly, an AB cell that was isolated from a mom-5(or57) mutant and allowed to divide to the AB¹⁶ or AB³² stages did not show POP-1 asymmetry between any sister pairs (0/8 and 0/4 experiments,respectively). These results suggest that MOM-5/Frizzled has a role in POP-1 asymmetry that is independent of MOM-2/Wnt signaling. Because sister pairs of AB¹⁶ or AB³² cells usually show POP-1 asymmetry in intact embryos that have been depleted of MOM-5/Frizzled by RNAi or by mutation, we examined intact embryos that were depleted of both MOM-2/Wnt and MOM-5/Frizzled. These latter embryos lacked POP-1 asymmetry in both AB and EMS descendants at the AB¹⁶ and AB³² stages(Fig. 2E,F; Table 2).

A wild-type, sequentially isolated AB¹⁶ cell can divide into a pair of sister cells with POP-1 asymmetry. However, when two adjacent AB¹⁶ cells were allowed to divide, their spindles almost invariably aligned to generate a line of two sister pairs with low/high-high/low POP-1 polarity (Fig. 3G). This was the case irrespective of whether two sequentially isolated AB¹⁶cells were combined and allowed to divide once (19/21 experiments), or a sequentially isolated AB⁸ cell was allowed to divide twice (30/32). Identical results were obtained for the sequentially isolated daughters of the EMS cell (5/5). These results suggest that each parental cell aligned the spindle of the other, with a reciprocal induction of low/high POP-1 polarity. Thus, these cells induce low/high POP-1 in the distal/proximal daughters of the responding cell, in contrast to the high/low POP-1 induced by MOM-2/Wnt signaling.

We observed two additional examples in which low/high POP-1 polarity appeared to be induced in the distal/proximal daughters of a responding cell. First, experiments combining an AB⁴ cell with an MS, E or C cell,or combining an AB¹⁶ cell with a C cell, resulted in a high/low-high/low pattern of POP-1 polarity(Fig. 5G,H). This pattern is consistent with the hypothesis that the AB⁴ cell induced low/high POP-1 polarity in the MS daughters, whereas MS simultaneously induced high/low POP-1 polarity in the AB⁴ daughters. The second example involved POP-1 polarity in C and P₃, the daughters of the P₂ cell(see legend to Fig. 6 for details). In wild-type or mom-2(or42) mutant embryos, C is born proximal to ABp and has high POP-1, whereas P₃ is distal and has low POP-1 (Fig. 6A, and see Fig. 1). In a devitellinized embryo, in which P₂ does not contact ABp, we found that POP-1 polarity was reversed in the P₂ daughters(Fig. 6C). We observed the same polarity reversal when isolated EMS and P₂ cells were combined and allowed to divide (Fig. 6E). However, the normal pattern of POP-1 polarity was restored after ABp and ABa cells were grafted onto the P₂ cell of a devitellinized wild-type host embryo (position of blue arrow in Fig. 1). This was the case irrespective of whether the graft ABp and ABa cells originated from wild-type embryos (10/10), or from mom-2(or42) mutants (5/5; Fig. 6G). Thus, AB² cells can induce low/high POP-1 polarity in the P₂ daughters.

We examined whether MOM-2/Wnt, MOM-1/Prc, MOM-5/Frizzled or MOM-4/MAPKKK are required for the polarized division of adjacent AB¹⁶ cells (the granddaughters of a sequentially isolated AB⁸ cell; Table 3). The mom-2and mom-2;mom-1 mutant cells usually had aligned spindles and low/high-high/low POP-1 polarity. In contrast, the cells from mom-5or mom-4 mutants did not show POP-1 asymmetry. The mom-4mutant cells usually aligned their spindles, however the mom-5 mutant cells did not (Table 3). We conclude that MOM-5/Frizzled and MOM-4/MAPKKK, but not MOM-2/Wnt, play a role in low/high POP-1 polarity.

The a/p sister cells in early C. elegans embryos have an invariant high/low POP-1 polarity that is propagated in successive cell divisions. The establishment of this simple pattern appears to involve a complex series of cell interactions. As the 4-cell embryo divides, the EMS spindle is aligned with P₂, and P₂ signaling results in low levels of nuclear POP-1 in the daughter of EMS proximal to P₂. The AB² cells, called ABa and ABp, divide perpendicular to P₂ and do not respond to P₂ signaling. We have shown that an AB² cell can respond to signaling if the AB²spindle and P₂ are aligned, either by altering the position of P₂ or by altering the orientation of the AB² spindle. In normal development the perpendicular, left/right division of ABp produces daughters with very similar fates, and the parallel differentiation of these left/right daughters is a major source of bilateral symmetry in C. elegans (Sulston et al.,1983). If the ABp spindle were to align with P₂, the a/p division would result in daughters with different levels of POP-1 and presumably different fates.

At the next round of cell division, each AB⁴ cell divides into daughters with POP-1 asymmetry, as do the subsequent AB⁸ and AB¹⁶ cells. Previous studies reached contradictory conclusions about whether asymmetry in the AB lineage was determined intrinsically or through cell interactions (see Introduction). Our present study provides a resolution of this paradox. We observed POP-1 asymmetry in sister pairs of AB¹⁶ cells, but not AB⁸ cells, which were derived from an isolated AB. These results are consistent with those from a previous study that killing P₁ caused posterior to anterior fate transformations at the AB⁸ stage; POP-1 remains high in all of the AB⁸cells after P₁ is removed, so all cells should adopt anterior fates(Hutter and Schnabel, 1995). Our results also support the study of Wittmann et al., which suggested AB descendants had an intrinsic asymmetry independent of P₁; their study used a transgenic marker that is expressed at the AB¹⁶ stage(Wittmann et al., 1997). Thus,cell interactions are essential for POP-1 asymmetry at, but not after, the AB⁸ stage. Reported examples of AB⁸ cells that correctly adopted posterior fates after killing or removing P₁ may represent cells in transition between the two modes of generating POP-1 asymmetry(Gendreau et al., 1994; Hutter and Schnabel,1995).

Although we propose that cell interactions determine POP-1 asymmetry in the early AB lineage, our results argue against a previous model that the primary interaction is between AB and P₁ (latent polarity model), rather than between AB descendants and P₁ descendants(Hutter and Schnabel, 1995). We have shown that exposing AB to P₁ is not sufficient to generate POP-1 asymmetry at the AB⁸ stage, and that P₁ does not provide signaling activity in our assays. In the previous study, laser irradiation of the P₁ descendants did not prevent a/p polarity at the AB⁸ stage, suggesting that the P₁ descendants were not essential for polarization (Hutter and Schnabel, 1995). However, laser-irradiation does not effectively prevent P₂ from signaling EMS, nor does it prevent P₂from signaling ABp through a separate, Notch-related signaling pathway(Mello et al., 1994). All of the known genes that are involved in POP-1 asymmetry in the early embryo are expressed maternally, as are the components of the Notch pathway. Thus, it may be difficult to eliminate translation of maternally provided mRNAs by laser-irradiating the early embryonic cells.

The AB⁴ cells contact one or more of the P₁descendants MS, E, C and P₃. We have shown that each of these P₁ descendants can induce high/low POP-1 polarity in AB descendants, similar to the ability of P₂ to induce high/low POP-1 polarity in the daughters of EMS. Indeed, at least two of the P₁descendants, E and C, appear to be fully equivalent to P₂ in their ability to polarize EMS. C and P₂ require MOM-2/Wnt for signaling,however the MS cell appears to provide an additional, or alternative, high/low signal. We presume this unidentified signal is the source of POP-1 asymmetry in the AB⁴ daughters in mom-2 mutants. The C. elegans genome can encode five Wnt proteins, raising the possibility that other Wnts contribute to high/low signaling during the first few embryonic cell divisions (Ruvkin and Hobert,1998).

The AB⁴ spindles, in contrast to the AB² spindles,align with the signaling cells P₂, MS, E or C in cell culture experiments. This difference might result from the synthesis of new regulators in the AB⁴ cells, or from the degradation of inhibitors present in the AB² cells. In normal development each AB⁴ division results in an anterior/posterior pair of daughters. However, all of these divisions are oblique, and some are nearly transverse, with respect to the a/p axis of the egg (see Fig. 5A,C). The alignment of the AB⁴ spindles with posterior-localized signaling cells such as E provides a partial explanation for the a/p pattern of high/low POP-1 polarity at the AB⁸ stage. In addition, the ability of MS to align the ABal spindle may be crucial for the normal pattern of Notch-mediated interactions. MS can interact with the ABal daughters through a Notch-like signaling pathway, and the alignment of the ABal spindle by MS ensures that only one daughter undergoes this interaction(Mello et al., 1994; Hutter and Schnabel, 1994).

Detailed cell-killing experiments on early C. elegans embryos have revealed unexpected and surprisingly complex networks of positive and negative interactions that are involved in the specification of muscle fates and that are presumably integrated for proper development(Schnabel, 1994; Schnabel, 1995). Although the molecular basis of muscle specification is not yet understood, our present study shows that the embryo contains multiple cells with the potential to influence POP-1 polarity. In our cell culture experiments, signaling from E predominated over signaling from MS when an AB⁴ cell simultaneously contacted both E and MS. Similarly, we propose that signaling from C predominates over signaling from MS in polarizing the division of ABar. The polarity of POP-1 at the ABar division is reversed in mom-2 mutants,presumably because of the absence of a signal from C and the persistence of a non-MOM-2/Wnt signal from MS. An analogous example was found in a previous analysis of the C. elegans Wnt family member, EGL-20, which functions in the asymmetric division of a larval epidermal cell called V5. In the absence of EGL-20/Wnt, the normal asymmetry of the V5 division is reversed by signaling from a neighboring cell (Whangbo et al., 2000).

By the AB¹⁶ stage, the descendants of both AB and EMS have an ability to generate POP-1 asymmetry that appears to be independent of prior contact with signaling cells that express MOM-2/Wnt. This conclusion is based on experiments in which cells were sequentially isolated in culture, but is supported by the observation that intact mom-2;mom-1 mutant embryos show robust POP-1 asymmetry between sister pairs at the AB¹⁶ and later stages. Although sequentially isolated AB¹⁶ cells can divide with POP-1 asymmetry, contact with other AB descendants can determine POP-1 polarity. These interactions induce low/high POP-1 polarity, in contrast to the high/low polarity induced by MOM-2/Wnt signals. Because adjacent AB¹⁶ cells divide with mirror image POP-1 polarity, we assume that each cell can both signal and respond to signaling. Thus, a single, isolated AB¹⁶ cell may generate POP-1 asymmetry through self-signaling.

MOM-5/Frizzled is essential for POP-1 asymmetry in isolated cells that have not been exposed to MOM-2/Wnt signaling. Therefore MOM-5/Frizzled may be a component of the signaling pathway that generates low/high POP-1 polarity independent of MOM-2/Wnt. Drosophila Frizzled is an essential component of the planar cell polarity pathway, however the role of Wnt proteins has not been determined (Lawrence et al., 2002). It will be of interest to determine whether other genes involved in Drosophila planar cell polarity have functions in low/high signaling in C. elegans. MOM-4/MAPKKK and proteins such as LIT-1/Nemo and WRM-1/Beta-catenin are essential for POP-1 asymmetry in AB descendants, and thus appear to be core components of the asymmetry-generating machinery (this study) (Kaletta et al.,1997; Lin et al.,1998; Meneghini et al.,1999; Rocheleau et al.,1997; Rocheleau et al.,1999).

The observation that POP-1 asymmetry is present in intact mom-5mutant embryos, but not in mom-2;mom-5 double mutant embryos,suggests that MOM-2/Wnt signaling can induce POP-1 asymmetry independent of MOM-5/Frizzled. These results support the view from previous genetic studies that MOM-2/Wnt and MOM-5/Frizzled have overlapping, but distinct roles in the early embryo (Rocheleau et al.,1997; Schlesinger et al.,1999). A survey of gene expression patterns in C. elegansembryos has detected mRNAs corresponding to at least two additional Frizzled-related proteins that are candidate receptors for MOM-2/Wnt (Y. Kohara, personal communication; http://nematode.lab.nig.ac.jp/). However, our present study indicates that these Frizzleds cannot be functionally redundant with MOM-5 for the POP-1 asymmetry shown by isolated,cultured cells.

When two adjacent AB⁸ cells divide in culture they generate a line of two sister pairs with low/high-high/low POP-1 polarity. In an intact,normal embryo, similarly oriented divisions of adjacent AB⁸ cells would be expected to produce sister pairs with high/low-high/low POP-1 polarity. Thus, the behavior of the isolated cells does not reproduce the normal pattern of POP-1 polarity. Among the cell culture experiments described here, the only condition that resulted in high/low-high/low POP-1 polarity involved combining a low/high signaling cell (AB⁴) with a high/low signaling cell such as MS, E or C. Thus, it is possible that the normal pattern of POP-1 involves two distinct signaling pathways. We have shown that isolated AB descendants remain responsive to MOM-2/Wnt signaling until at least the AB¹⁶ stage, however we do not yet know whether Wnt signals persist in normal embryos at the AB¹⁶ and later stages.

We are indebted to Marianne Land and Charles S. Rubin for providing the PLK-2 antiserum prior to publication. We thank Rafal Ciosk and Jeremy Nance for a critical reading of the manuscript, Russell Hill for stimulating interest in this project, and current members of the Priess laboratory for advice and assistance. Some of the nematode strains used in this study were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources (NCRRR). F. Park is supported by the Poncin Scholarship Fund, and J. Priess is supported by the Howard Hughes Medical Institute.