Cell polarity and gastrulation in C. elegans

Numerous studies on C. elegans embryos have revealed the basic strategies that establish the anterior-posterior and dorsal-ventral axes. Axis specification in C. elegans embryos begins with fertilization of the egg, where the point of sperm entry defines the posterior pole (Goldstein and Hird, 1996). Fertilization induces the association of a group of proteins, collectively called PAR proteins, to either the anterior or posterior cortex. For example, PAR-3 and PAR-6 associate with the anterior pole while PAR-2 and PAR-1 associate with the posterior pole (Boyd et al., 1996; Etemad-Moghadam et al., 1995; Guo and Kemphues, 1995; Hung and Kemphues, 1999). The localization of the PAR proteins is interdependent; mutations in par-3 cause the anterior mislocalization of PAR-2, and mutations in par-2 cause the posterior mislocalization of PAR-3 (Boyd et al., 1996; Etemad-Moghadam et al., 1995).

The dorsal-ventral axis is determined as the embryo divides from two cells to four (reviewed by Schnabel and Priess, 1997). The division of the anterior cell generates two initially equivalent daughters that express a receptor related to the Notch protein. The division of the posterior cell generates non-equivalent daughters, one of which expresses a ligand for the receptor. In an apparently random manner, only one of the receptor-expressing cells contacts the ligand-expressing cell; this interaction leads to the specification of dorsal cell types such as hypodermis (skin).

In addition to anterior-posterior and dorsal-ventral axes, the early embryo must establish an outer-inner polarity. Cells on the ventral surface of the embryo become committed to endodermal and mesodermal fates through a combination of cell signaling events and asymmetrically localized factors (Bowerman et al., 1993; Bowerman et al., 1992; Goldstein, 1992; Lin et al., 1995; Rocheleau et al., 1997). These endodermal and mesodermal precursors must then move from the outer surface of the embryo into the interior for the formation of a functional intestine and muscular system.

Animal embryos use several strategies to position endodermal and mesodermal precursors into their interior, a process called gastrulation. In many embryos, such as those of sea urchins and amphibians, the early embryonic cleavages generate a cluster of cells (the blastula), and a central cavity (the blastocoel) develops within this cluster. Certain cells detach from their neighbors on the cell surface and enter the blastocoel (ingression), or large groups of cells fold into the blastocoel (invagination and involution) (Gilbert and Raunio, 1997). In some animals, gastrulation occurs without formation of a blastocoel. In ctenophores, for example, ectodermal cells spread over and internalize endodermal cells (epiboly) (Komai, 1968; Martindale and Henry, 1999).

Gastrulation in C. elegans begins at the 26-cell stage, when two endodermal precursors ingress from the surface of the embryo into the interior. Shortly thereafter, mesodermal precursors and germline precursors follow the endodermal precursors into the interior of the embryo (Sulston et al., 1983). Very little is known about the cellular or molecular basis for these events. Several mutations have been identified that prevent or delay ingression of the endodermal precursors, most of which cause the endodermal precursors to divide prematurely (Denich et al., 1984; Knight and Wood, 1998). Inhibiting embryonic transcription also causes the endodermal precursors to divide precociously and prevents ingression (Powell-Coffman et al., 1996). Ingression of the endodermal precursors is likely to require embryonic transcription of one or more genes in a chromosomal region called the endoderm determining region, or EDR; a chromosomal deficiency that deletes the EDR delays or prevents ingression (Zhu et al., 1997). There are at least two genes in the EDR that are involved in specification of the endodermal cell fate, suggesting a possible link between cell fate and ingression (Zhu et al., 1997).

In this report we investigate the cellular basis for gastrulation in C. elegans. PAR proteins that show anterior-posterior asymmetry in 1-cell and 2-cell embryos subsequently develop outer-inner, or apical-basal, polarity as the blastula forms. We demonstrate that the apical-basal localization of the PAR proteins is dependent on cell-cell contacts by generating double embryos with abnormal patterns of cell contact. We show that ingression is associated with changes in the shape of the ingressing cells and a redistribution of non-muscle myosin. Finally, we present evidence that cell fate, rather than cell position, is the predominant factor regulating ingression, although steric interactions can play an important role.

Nematodes were cultured and manipulated as described (Brenner, 1974). Unless otherwise indicated, experiments were performed on the wild-type N2 (var. Bristol). The following mutants were utilized: chromosome II LG II: unc-4(e120) (Brenner, 1974), mex-1(zu120) (Mello et al., 1992); unc-32(e189) (Brenner, 1974), lit-1(t1512ts) (Kaletta et al., 1997), unc-45(e286ts) (Brenner, 1974), par-2(lw32) (Cheng et al., 1995), par-3(it71) (Cheng et al., 1995), lon-1(e185) (Brenner, 1974); LGV: zuDf2 (Zhu et al., 1997). The following integrated transgenes containing green fluorescent protein (GFP) reporters were used: zuIs3 (end-1::GFP) (J. Nance, unpublished), ruIs32 (pie-1::GFP::HIS-11) (Pratis et al., 2001), pxIs6 (pha-4::GFP::HIS-11) (Portereiko and Mango, 2001), itIs153 (pie-1::PAR-2::GFP) (Wallenfang and Seydoux, 2000). Mex-1 embryos were mex-1 unc-4. Par-2 embryos were par-2 unc-45. Par-3 embryos were par-3 lon-1; zuIs3.

Standard techniques were used to synthesize double-stranded RNA (dsRNA) from T7 promoter-tagged, PCR-amplified cDNA or genomic DNA. For analysis of ama-1(RNAi) embryos (2-4 μg/μl dsRNA made from cDNA region 2876-3450), young adult hermaphrodites carrying both a maternal (pie-1::GFP::HIS-11) and zygotic (end-1::GFP) GFP reporter were injected in the gonad or soaked overnight with dsRNA (Fire et al., 1998; Tabara et al., 1998). After allowing worms to recover for 30 hours, embryos were analyzed for end-1::GFP expression. In a representative experiment, the end-1::GFP reporter was detectable in only 5/224 ama-1(RNAi) embryos. For analysis of hmr-1(RNAi) embryos (2.5 μg/μl dsRNA made from genomic DNA region 13344-13989), approximately 20 hours after injection of dsRNA into wild-type adults, embryos were collected and fixed for immunofluorescence. Embryos were immunostained to confirm HMR-1 depletion.

To examine blastocoel formation, 1-cell or 2-cell embryos were pipetted onto a poly-L-lysine (Sigma)-coated coverslip containing clay spacers; the coverslip was inverted onto a slide for observation. In all other experiments 2-cell or 4-cell embryos were mounted for videomicroscopy on 4% agarose (see Sulston et al., 1983). Isolated AB blastomeres were mounted in embryonic culture medium (Shelton and Bowerman, 1996) on slides and covered with coverslips coated with 1.5% agarose; techniques used to isolate and culture embryos or individual blastomeres were essentially as described (Shelton and Bowerman, 1996). Embryo combination experiments were performed at 15°C. To inhibit transcription, embryonic culture medium was supplemented with 50 μg/ml α-amanitin (Sigma). Laser ablations were performed as described elsewhere (Mello et al., 1992).

4D images were acquired as described (Thomas et al., 1996) using 4D Grabber v1.32 software (C. Thomas, Integrated Microscopy Resource, University of Wisconsin-Madison, USA). Movies were analyzed using 4D Viewer v4.11 (C. Thomas, Integrated Microscopy Resource, University of Wisconsin-Madison, USA) and Nematode Navigator software (kindly provided by J. Pitt, http://www.fhcrc.org/labs/priess/nn.html). The volume of the blastocoel and embryo (± s.d.) were calculated using NIH Image1.62 (Wayne Rasband, National Institutes of Health, USA).

Ingressions were scored when a cell sank permanently from the surface of the embryo. Comparisons of ingression times were based on the analysis of 4D videorecordings of three embryos; all cells that ingressed during gastrulation were followed in each embryo. Ingression times were normalized to cell division timings presented by Sulston et al. (Sulston et al., 1983) using the interval between the MSa and MSaaa divisions as a measure of developmental time. Descriptions of wild-type gastrulation were based on observations of these and additional 4D videorecordings. Ectopic ingressions in mex-1 mutants were scored in the ABpr lineage. lit-1 temperature shifts were from 15°C to 25°C at the 24-cell stage. Because of the variable expressivity of the lit-1 mutant phenotype, analysis of MS ingression was restricted to the 7/11 embryos showing simultaneous ingression of all wishbone cells. In the E ablation experiments, analysis was restricted to the anterior MS central cells that do not contact the E corpse.

Mixed-stage embryos were fixed, embedded, sectioned and stained for transmission electron microscopy (TEM) as described (Priess and Hirsh, 1986). Several thick sections were cut from each block to determine the precise stage and orientation of specific embryos. Sections were analyzed from >50 embryos embedded in three different blocks. Embryos for scanning electron microscopy were fixed as for TEM, omitting the tannic acid treatment. Embryos were transferred by mouth pipet to a poly-L-lysine coated coverslip, then dehydrated and dried as described (Braet et al., 1997).

Embryos were fixed and processed for overnight immunostaining at 4°C as described (Leung et al., 1999). Cultured blastomeres were fixed in –20°C methanol for 15 minutes, then immunostained at 37°C for 1 hour. A strain containing the itIs153 reporter was used for analysis of PAR-2 in cultured blastomeres. Unless indicated otherwise, immunofluorescence studies were based on the observation of 15-100 embryos at the indicated stage. The following dilutions of antibodies/antisera were used: 1:200 mouse anti-GFP (‘α’-GFP; Clontech); 1:10 rabbit α-HMR-1 (Costa et al., 1998); 1:50 rabbit α-NMY-2 (Guo and Kemphues, 1996); 1:10 rabbit α-PAR-2 (Boyd et al., 1996); 1:10 chicken α-PAR-3 (Tabuse et al., 1998); 1:15 rabbit α-PAR-3 (Etemad-Moghadam et al., 1995); 1:10 rabbit α-PAR-6 (Hung and Kemphues, 1999); 1:1000 rabbit α-PGL-1 (Kawasaki et al., 1998). Secondary antibodies were conjugated with Alexa Fluor 488 (Molecular Probes) or Cy-3 (Jackson ImmunoResearch Laboratories, Inc.). In some experiments, DNA was stained with 60 ng/ml 4,6-diamidino-2-phenylindole (DAPI).

Confocal images were acquired on a Leica TCS SP spectral confocal microscope. All other images were acquired using a standard epifluorescence microscope equipped with a digital camera; some epifluorescence images were numerically deconvolved using DeltaVision algorithms (Applied Precision, Inc.).

For the analysis of NMY-2 localization in MS descendants, embryos expressing pha-4::GFP::HIS-11 were stained with antibodies or antisera against PGL-1, GFP and NMY-2; GFP and PGL-1 staining were used as reference points to identify specific MS descendants after fixation. NMY-2 was scored as ‘enriched’ when the apical concentration in an ingressing cell appeared greater than in other surface cells, exclusive of dividing cells.

Gastrulation in C. elegans involves the movement of a subset of cells from the ventral surface of the embryo into the blastocoel. These movements, described here as cell ingressions, occur over a period of about 3 hours and begin at the 26-cell stage of embryogenesis. Some, or all, of the descendants of each of the early embryonic blastomeres ingress (see Fig. 4A,B). The ingressing cells form tissues such as the endoderm (descendants of the E blastomere), germline (P4 descendants) and mesoderm (D descendants and a subset of AB, MS, and C descendants; see Fig. 4A).

Prior to gastrulation, the embryo consists of a hull of cells one cell in thickness. These cells surround a small, central cavity called the blastocoel (arrowheads, Fig. 1A). For convenience, we distinguish three types of membrane surfaces for each cell. The apical membrane faces the perimeter of the embryo, the basal membrane faces the blastocoel, and the lateral membranes face adjacent cells within the hull (Fig. 1A). In three-dimensional timelapse (‘4D’) videorecordings of living embryos, the blastocoel forms as the basal surfaces of diametrically opposed cells gradually detach from each other. Small openings between basal surfaces are first visible at the 4-cell stage (arrowhead, Fig. 1B) and numerous openings of variable size are present by the 26-cell stage (arrowheads, Fig. 1C). The blastocoel achieves its maximum volume by the 26-cell stage when it measures 450±100 μm³ (n=5). This volume is only 2% of the volume of the embryo (25,800±980 μm³), and less than half the volume of a single cell (about 1000 μm³). Because of the small size of the blastocoel, the basal surfaces of diametrically opposed cells transiently come into contact after they separate, temporarily obscuring regions of the blastocoel. In contrast to the often large spaces separating basal surfaces (Fig. 1E), lateral surfaces either are juxtaposed or are separated by only small spaces (Fig. 1D).

How might the various surfaces of embryonic cells become specified, such that the blastocoel forms at the basal surface? In the 4-cell embryo, the basal surface corresponds to the region where the two daughters of the AB blastomere (ABa and ABp) contact the two daughters of the P1 blastomere (EMS and P2) (Fig. 1B). We asked whether blastocoel formation required interactions between specific AB and P1 descendants by separating AB from P1 at the 2-cell stage and allowing each blastomere to develop in isolation (n=8). The AB descendants generated by the first two cycles of cell division formed a tightly adherent spherical cluster of cells. By the third cycle of cell division, the clusters developed a central cavity of variable size that resembled the normal blastocoel (Fig. 1F). Because in these experiments P1 descendants form a linear array of cells rather than a spherical cluster (see Fig. 1H) (Schierenberg, 1987), we did not examine the behaviour of these cells further. We asked whether AB descendants from separate embryos could generate a blastocoel by combining two 2-cell embryos, or two early 4-cell embryos, head to head with the anteriormost surfaces of the AB cells in contact (n=16). At the point of contact, the surfaces of the combined blastomeres flattened and adhered tightly. All of these ‘double’ embryos developed a central cavity of variable size surrounded by both sources of AB descendants (labeled 1 and 2 in Fig. 1H). We draw several conclusions from these experiments. First, the surfaces of cells at the 2-cell and 4-cell stages of embryogenesis are adhesive; thus this adhesion must be overcome for cell surfaces to separate during blastocoel formation. Second, interactions between AB descendants are sufficient to form a blastocoel. Finally, we conclude that cell contact can reorient the polarity of AB descendants; in double embryos a blastocoel can form at a surface that would normally have been an apical surface.

Analysis of the PAR proteins has focused primarily on their anterior-posterior localization at the 1-cell stage and their roles in subsequent anterior-posterior polarity (see Introduction). After cell division begins, however, this anterior-posterior asymmetry is reiterated only in the lineage of cells that form the germline (Boyd et al., 1996; Etemad-Moghadam et al., 1995; Guo and Kemphues, 1995; Hung and Kemphues, 1999) (the germline cells P2 and P3 are indicated by cyan asterisks in Fig. 2). In contrast, some somatic (non-germline) cells show an apical-basal polarity in PAR localization (Boyd et al., 1996; Etemad-Moghadam et al., 1995; Guo and Kemphues, 1995; Hung and Kemphues, 1999). By the 4-cell stage, the posterior PAR protein PAR-2 is localized to basolateral surfaces (arrows, Fig. 2A,B), and is not detected on the apical surfaces of somatic cells. By the middle of the 4-cell stage, the anterior PAR proteins PAR-3 and PAR-6 are present over the entire cortex of each somatic blastomere (arrows, Fig. 2C). However by the end of the 4-cell stage and at later stages PAR-3 and PAR-6 are concentrated in a broad ‘cap’ centered on the apical surface (arrow, Fig. 2D). The apical-basal polarity of the PAR proteins persists through early gastrulation, although the level of PAR-2 diminishes (Fig. 2E). After gastrulation begins, cells in the interior of the embryo accumulate PAR-3 and PAR-6 on their blastocoel-facing surfaces (arrowheads, Fig. 2F,F′); these and later changes in PAR distribution were not analyzed further (see also Leung et al., 1999).

We used 4D videomicroscopy to examine formation of the blastocoel in par-2 and par-3 mutant embryos. par-2 mutant embryos (n=6) developed a central cavity resembling the normal blastocoel (data not shown). par-3 mutant embryos (n=6) also developed a central cavity, but in addition showed frequent separations between the lateral membranes of cells (arrowheads, Fig. 1G); these lateral separations persisted through several cell divisions. The lateral separations in par-3 mutant embryos were comparable to the spaces observed between the basal surfaces of cells in wild-type embryos, but were much larger than the spaces normally present between lateral surfaces. These results suggest that PAR-3 is required, directly or indirectly, for the normal adhesiveness of lateral membranes.

How do the PAR proteins switch from anterior-posterior polarity to apical-basal polarity? Germline blastomeres, which maintain anterior-posterior PAR polarity during the early embryonic stages, are transcriptionally quiescent, while somatic blastomeres that switch to apical-basal asymmetry are transcriptionally active (Seydoux et al., 1996). However, we found that PAR-3 and PAR-2 switch from anterior-posterior to apical-basal polarity after embryonic transcription is prevented by double-stranded RNA inhibition (RNAi) of the ama-1 gene [Fig. 2G and data not shown; the ama-1 gene encodes the large subunit of RNA polymerase II (Rogalski and Riddle, 1988)]. In addition, a blastocoel appeared to form normally in ama-1(RNAi) embryos (n=6) and in wild-type embryos exposed to α-amanitin (n=4). Thus apical-basal PAR polarity, as well as blastocoel formation, appears to be determined by maternally provided gene products.

We asked whether cell contacts determine apical-basal PAR polarity by constructing and analyzing double embryos with ectopic cell contacts, as described above. 2-cell embryos were joined head to head, then allowed to develop between 15 and 45 minutes (1-3 additional cell cycles) before fixation. We found that PAR-2 was excluded from the apical regions of somatic blastomeres in the double embryos, but was localized to all basal and lateral membranes including ectopic sites of cell contact (n=6; arrow, Fig. 3A). In contrast, PAR-3 and PAR-6 were localized predominately to the contact-free, apical surfaces of somatic blastomeres (n=6 and 8, respectively; Fig. 3B-D). As in wild-type embryos, there appeared to be a cap of PAR-3 and PAR-6 toward the center of the apical cortex.

We asked if PAR localization was dependent on HMR-1, a C. elegans homologue of vertebrate cadherin; HMR-1 is localized exclusively to sites of cell contact in a wild-type 4-cell embryo (Costa et al., 1998), a pattern very similar to that of PAR-2. However hmr-1(RNAi) embryos, with markedly depleted HMR-1 levels, had the wild-type pattern of both PAR-2 and PAR-3 localization (data not shown). We next asked if the apical-basal polarity in PAR localization was dependent on par gene function. As reported previously by others (Boyd et al., 1996; Etemad-Moghadam et al., 1995), we observed that PAR-2 was not restricted to the posterior pole in 1-cell stage par-3 mutant embryos, and that PAR-3 was not restricted to the anterior pole in par-2 mutant embryos. During the 2-cell to 16-cell stages, PAR-2 was not restricted to basal and lateral surfaces in par-3 mutant embryos, and was instead associated with the entire cortex of each blastomere (Fig. 2H) (see also Boyd et al., 1996). PAR-3 showed the wild-type pattern of apical localization in par-2 mutant embryos (n=40), but failed to accumulate at the cortex in par-6 mutants (n=15). We conclude that PAR-3 is required to exclude PAR-2 from the apical surface. PAR-3 localization to the apical surface appears to be independent of both HMR-1 and PAR-2, but requires PAR-6.

Ingression occurs on the ventral surface of the embryo. Many of the cells that ingress are born on the ventral surface (E, MS and P4 descendants; Fig. 4B). Other cells, such as the C descendants (green cells in Fig. 4B), are born in dorsal or lateral positions and move to the ventral surface prior to ingression. We analyzed 4D videorecordings of gastrulating embryos to determine when and where cells ingress. Cells do not ingress from a single location, but rather enter the blastocoel from several positions over the ventral surface of the embryo; these patterns of ingression are summarized in Fig. 4C, and some of the major groups of ingressing cells are described here briefly.

The daughters of the E blastomere are the first to ingress, beginning at the 26-cell stage (90 minutes after the first embryonic division). Ingression of the E daughters (yellow asterisks in Fig. 4D) takes about 16 minutes, and creates a transient opening on the surface of the embryo. This opening is sealed as neighboring cells, primarily MS descendants and P4 (red and cyan arrowheads, respectively, Fig. 4D), spread across the surface of the E daughters. In scanning electron micrographs of the ventral surface, the MS descendants show wedge-shaped processes that are oriented toward the ingressing E daughters (Fig. 5A,B). The leading edges of these processes taper into thin sheets over the surfaces of the E daughters (Fig. 5C). Although the P4 blastomere moves across the surfaces of the E daughters, it does not extend processes resembling those from the MS descendants (Fig. 5).

By 150 minutes most of the ventral surface of the embryo is occupied by the 16 descendants of the MS blastomere. 12 of the MS descendants ingress at about 150 minutes (red ‘wishbone’ shape in Fig. 4C). The 4 remaining MS descendants stay on the surface of the embryo for an additional cell cycle before ingressing at about 190 minutes (red oval shape in Fig. 4C, and red asterisks in Fig. 4E). The ingression of the ‘wishbone’ group of 12 MS descendants, as well as the ingression of the P4 descendants, each creates a transient surface gap that is quickly enclosed by neighboring cells. In contrast, the surface gaps created by ingression of the D descendants (arrow, Fig. 4E) and by ingression of the remaining MS descendants (arrows, Fig. 4F) persist for 20-30 minutes. In all cases examined, sister cells ingressed either simultaneously or within a few minutes of each other (see Discussion). After the cycles of ingression are completed, the ventral surface of the embryo is occupied primarily by neuronal precursors. These precursors do not appear to enter the blastocoel by ingression, but are internalized by epiboly of skin cells [Fig. 4G; reviewed by Simske and Hardin (Simske and Hardin, 2001); see also Sulston et al. (Sulston et al., 1983)].

Cells in the left half of the ‘wishbone’ group of MS descendants ingress at the same time and have the same fate as cells in the right half of the wishbone [see Fig. 4C; see also Sulston et al. (Sulston et al., 1983)]. The observation that cells with identical fates, but with different positions, can ingress at the same time suggests that cell fate could determine the time of ingression. In their analysis of the C. elegans cell lineage, Sulston et al. (Sulston et al., 1983) documented several examples of pairs of cells that were located in different parts of the embryo but that had identical, or nearly identical, patterns of division and differentiation. We selected nine such examples of AB descendants (cell pairs numbered a and b in Table 1), and determined their ingression times from videorecordings of live embryos. Cells with the same fate usually had remarkably similar ingression times (8/9 cases). For example, the ingression times of cells 4a and 4b differed by only 1 minute (Table 1). In contrast, cells that were born in equivalent cell cycles, but that had different fates could differ significantly in ingression times. For example, the ingression times of cells 4a and 9b differed by 77 minutes (Table 1).

If cell fate, rather than position, is the sole determinant of ingression, mutations that affect cell fate should alter the pattern of ingression. Mutations in the mex-1 gene cause AB descendants to express a transcription factor that normally specifies the MS fate (Bowerman et al., 1993; Mello et al., 1992). We found that several AB descendants that do not ingress in wild-type embryos did ingress in mex-1 mutant embryos analyzed over the period corresponding to normal MS ingression (n=2). However, many of the descendants of the transformed AB cells did not ingress, in contrast to wild-type MS descendants, and the pattern of ingression varied between each of the mex-1 embryos.

To understand how ingression times are determined in the wild-type MS lineage, we focused on the difference in ingression times between the MS ‘wishbone’ descendants and their closely related relatives in the center of the wishbone (hereafter called the central cells; see Fig. 4C). As described above, the MS central cells (asterisks, Fig. 6A) remain on the surface for one cell cycle after the wishbone cells ingress. The central cells that remain on the surface overlie E descendants in the center of the embryo, while the wishbone cells ingress along the perimeter of the E descendants. We therefore asked whether preventing E ingression would allow the central cells to ingress with the wishbone cells. Although killing the E blastomere with a laser microbeam prevented ingression of the E daughters, the central cells did not ingress along with the wishbone cells (12/12 embryos, Fig. 6B). We obtained identical results in experiments where the fate of the E blastomere was altered by a chromosomal deficiency of the endoderm determining region (Zhu et al., 1997); although the E daughters and many of the subsequent E descendants remained on the surface of the mutant embryos, the wishbone and central cells showed the wild-type difference in ingression times (6/6 embryos).

By lineage, the central cells are posterior relatives of the anterior wishbone cells. In several cases studied, anterior-posterior differences between closely related cells in C. elegans are determined by anterior-posterior asymmetry in the expression of the transcription factor POP-1 (Lin et al., 1998; Lin et al., 1995), and POP-1 asymmetry can be controlled by the kinase LIT-1 (Kaletta et al., 1997; Meneghini et al., 1999; Rocheleau et al., 1999). We therefore shifted embryos homozygous for the temperature-sensitive mutation lit-1(t1512ts) to the restrictive temperature before the birth of the central cells; this temperature shift should have caused the central cells to have the same pattern of POP-1 expression as the wishbone cells. However, the central cells did not ingress at the same time as the wishbone cells in any of these experiments (n=7; Fig. 6C). We next asked whether the presence of the E descendants in the blastocoel might block ingression of the central cells by killing the E blastomere in temperature-shifted lit-1 mutant embryos. In 4 of 6 embryos tested, the central cells ingressed at the same time as the wishbone cells (arrow, Fig. 6D). Because the lit-1(t1512ts) allele is incompletely penetrant, we consider it likely that POP-1 expression was not altered in the two embryos with normal ingression times (see Kaletta et al., 1997). We conclude that the ingression pattern of the MS central descendants is controlled by redundant mechanisms involving cell fate specification and interactions with E descendants.

In living embryos, the apical surfaces of the E daughters initially appear rounded (arrowhead, Fig. 7A), similar to the apical surfaces of neighboring cells. Shortly before ingression, their apical surfaces flatten (Fig. 7B), and remain flattened during ingression; small, transient protrusions of the apical membranes appear during flattening (data not shown). We observed a similar apical flattening during ingression of the MS descendants (Fig. 7E), suggesting that apical flattening is a general characteristic of ingressing cells.

Apical flattening during ingression could be autonomous or a response to tension generated by neighboring cells. In normal development, the MS descendants extend processes over the E daughters during ingression (see Fig. 5). However, we found that the E daughters underwent apical flattening and ingressed even after the MS blastomere was killed with a laser microbeam [10/10 embryos; see also Junkersdorf and Schierenberg (Junkersdorf and Schierenberg, 1992)], suggesting apical flattening is a cell-autonomous process.

Cell shape changes often are driven by actin and non-muscle myosin. We therefore examined the distribution of NMY-2, a C. elegans non-muscle myosin (Guo and Kemphues, 1996). NMY-2 is present at low levels at the cell cortex during all stages of the cell cycle (Fig. 7C,D), and is strongly enriched in the contractile ring or midbody of mitotic cells (arrows, Fig. 7C,D). NMY-2 began to show enrichment on the apical surfaces of the E daughters just prior to ingression (24-cell stage, 6/36 embryos). This enrichment was evident in most E daughters at the onset of ingression (26-cell stage, 36/43 embryos; Fig. 7D), and in all E daughters near the end of ingression (28-cell stage, 20/20 embryos. The level of NMY-2 at the apical surface decreased once the E daughters entered the blastocoel (data not shown). We observed a similar apical enrichment of NMY-2 on ingressing MS descendants (arrowhead, Fig. 7F).

In comparison to other animal systems (see Introduction), gastrulation in C. elegans involves both cell ingression and epiboly. Embryogenesis occurs within a fixed space limited by the eggshell, and the volume of the blastocoel is very small, less than the volume of a single cell at the onset of gastrulation. Thus the first cell ingressions must be coupled to simultaneous redistributions in the masses of other embryonic cells. During ingression of the E daughters, this redistribution appears to be accomplished primarily by the MS descendants that immediately spread across the site of ingression. At later embryonic stages the ingressing cells are relatively small and there is less need for the surrounding cells to redistribute their mass; at these stages surface gaps persist over the sites of ingression for long periods of time.

Gastrulation in embryos like Drosophila involves sheets of cells that invaginate to form furrows and pockets. While adherens junctions link the surface cells of Drosophila embryos at the time of gastrulation (Oda et al., 1998; Tepass and Hartenstein, 1994), adherens junctions are not visible in C. elegans embryos until most of the cell movements associated with gastrulation are complete (Costa et al., 1998) (our unpublished observations). The absence of adherens junctions in early C. elegans embryos may thus facilitate ingression by small groups of cells. In sea urchins, primary mesenchyme cells appear to lose adherens junction connections with neighboring surface cells during ingression (Miller and McClay, 1997a; Miller and McClay, 1997b). In our survey of ingression times in C. elegans embryos, we noted that sister cells ingressed at approximately the same time; solitary cells did not ingress. A likely possibility is that the midbody linkage between sister cells prevents either sister from ingressing alone. Both electron microscopic and immunocytochemical studies have shown that the embryonic cells in C. elegans remain coupled by midbodies to their sisters through most of the cell cycle (Krieg et al., 1978) (J. N. and J. R. P., unpublished observations).

Ingressing cells in C. elegans embryos show an apical flattening and an apical accumulation of non-muscle myosin. Apical flattening could promote ingression by redistributing cytoplasm toward the basal region of an ingressing cell, and create space at the surface of the embryo for neighboring cells to redistribute their cytoplasm laterally. The invagination of epithelial sheets in Drosophila gastrulation is also associated with an apical flattening and accumulation of non-muscle myosin (Leptin and Grunewald, 1990; Sweeton et al., 1991; Young et al., 1991). Thus the cellular basis for invagination by epithelial sheets and for ingression by small groups of cells may be conserved, with the principal difference being whether the cells are linked by junctional complexes.

In principle, gastrulation could position a subset of unspecified cells into the interior of the embryo, where subsequent events would specify the endodermal and mesodermal fates appropriate for internal organs. Instead, C. elegans and many other animals appear to specify the fates of endodermal and mesodermal progenitors when these progenitors are on the embryo surface. Thus there must be direct or indirect mechanisms that couple gastrulation movements with cell fate.

Our present study provides evidence that cell fate is tightly linked to cell ingression in C. elegans. For example, mex-1 mutants with ectopic MS-like cells show ectopic ingressions during the period MS descendants normally ingress, and a mutation in lit-1 that changes the fate of MS descendants can change the ingression time of those descendants. Similarly, in Drosophila the cell fate regulators snail and twist appear to control invagination (reviewed by Leptin, 1995).

The molecular connections between cell fate and ingression or invagination are not known. Embryonic transcription is required for gastrulation in C. elegans (Powell-Coffman et al., 1996), though we have shown here that transcription is not required for the apical-basal localization of the PAR proteins or for blastocoel formation. Instead, embryonic transcription is required for the apical concentration of non-muscle myosin in ingressing cells and for the apical flattening of these cells. A simple model is that the cellular machinery required for ingression or invagination is regulated by one master control gene, and that this gene is in turn regulated by any one of several embryonically expressed transcription factors that specify cell fate. By analogy, the Drosophila cell cycle regulator string has a large, complex promoter controlled by numerous transcription factors that also control region-specific differentiation, creating regions of mitosis that match regions of common cell fate (Edgar et al., 1994; Johnston and Edgar, 1998; Lehman et al., 1999).

The control of MS ingression in C. elegans appears to be an example of fine-tuning in a developmental system. After ingression of the E daughters, the largest available spaces in the blastocoel are anterior to, and on both sides of, the E daughters. The MS wishbone cells overlie these spaces, and ingress into them. In contrast, the MS central cells overlie the E daughters, and do not ingress at the same time as the wishbone cells. Our analysis of lit-1 mutant embryos demonstrates that the presence of the E daughters in the blastocoel is sufficient to prevent ingression of the central cells during ingression of the wishbone cells; we consider it likely that this is a steric interaction, but cannot exclude more specific cell signaling. However, the MS central cells are subject to an additional, lit-1-dependent, control that prevents ingression even when space in the blastocoel is made available by killing the E blastomere. We propose that in normal development this mechanism prevents the central cells from attempting to ingress when there is not sufficient space to do so.

The events of gastrulation indicate that early embryonic cells have apical-basal polarity; for example, NMY-2 accumulates at the apical surface of ingressing cells, and the blastocoel forms at the basal surfaces. Apical-basal asymmetry is evident in embryonic cells of C. elegans as early as the 4-cell stage: PAR-3 and PAR-6 are localized to the apical surfaces, and PAR-2 and HMR-1(cadherin) are localized to the basal and lateral surfaces of cells. Drosophila and vertebrate homologues of PAR-3 and PAR-6, together with the atypical protein kinase PKC-3, have been shown to function in establishing apical-basal polarity in epithelial cells and neuroblasts (reviewed by Ohno, 2001). We have shown here that large cell separations are observed between both basal and lateral surfaces of cells in par-3 mutants. This result suggests that par-3(+) may have a role in distinguishing the basal from lateral surfaces during blastocoel formation in wild-type embryos, where large separations only occur between basal surfaces.

We have shown that cell contacts restrict PAR-3 and PAR-6 to the contact-free, apical surfaces. This mechanism differs, at least in part, from the mechanism that localizes these same proteins to the anterior surface of the 1-cell embryo. While PAR-2 has a role in determining PAR-3 localization at the 1-cell stage, PAR-2 is not required for the apical localization of PAR-3 at the 4-cell stage, nor is it required for blastocoel formation. HMR-1(cadherin) also is localized to cell contacts, but does not appear to have a role in PAR localization or blastocoel formation (our present results) (Costa et al., 1998). Interestingly, genetic or immunological inhibition of E-cadherin function in early mouse embryos does not prevent individual cells from becoming polarized, but rather causes a randomization in the axis of polarity (Johnson et al., 1986; Larue et al., 1994; Riethmacher et al., 1995). HMR-1 appears to be the only ‘classical’ cadherin with a β-catenin binding site, similar to mouse E-cadherin (Hill et al., 2001), although the C. elegans genome sequence predicts several cadherin-related proteins whose functions and localization have not been determined (Hill et al., 2001).

Localization of PAR-3, or associated proteins, to the apical surface could in principle differentiate the basal surface from the lateral surface. For example, the localization of ion channels to the apical surface could create a gradient that affects the opposite, basal, surface differently than the lateral surfaces. Vectorial ion transport is essential for formation of the blastocoel in mouse embryos, and channel proteins appear to be localized with apical-basal polarity in trophectodermal cells lining the blastocoel [reviewed by Watson and Barcroft (Watson and Barcroft, 2001)]. It will be interesting in future studies to determine how apical-basal polarity of the PAR proteins directs subsequent asymmetries. In the 1-cell embryo, the anterior-posterior polarity of the PAR proteins establishes a parallel gradient of MEX-5 in the cytoplasm. MEX-5 in turn functions to prevent the anterior expression of posterior proteins (Schubert et al., 2000). However, MEX-5 is uniformly distributed in somatic blastomeres that have an apical-basal polarity in PAR protein distribution, suggesting that MEX-5 does not mediate apical-basal polarity.

We thank Judith Austin, Ken Kemphues, Susan Mango and Geraldine Seydoux for providing transgenic strains; Ichiro Kawasaki and Susan Strome for PGL-1 antibodies; and Kathryn Baker and Ken Kemphues for the extremely generous gifts of PAR and NMY-2 antibodies. We thank Judy Groombridge and Franque Remington for assistance with electron microscopy, Jason Pitt for providing advice and Nematode Navigator software, and members of the Priess laboratory for useful discussions and suggestions. 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 (NCRR). J.N. and J.R.P. are supported by the Howard Hughes Medical Institute.