C. elegans PAR-3 and PAR-6 are required for apicobasal asymmetries associated with cell adhesion and gastrulation

Fertilization of the C. elegans egg initiates a cascade of events that defines the anterior-posterior axis of the embryo (reviewed by Pellettieri and Seydoux,2002). Key components of this cascade are a group of proteins,collectively called PAR proteins, that associate asymmetrically with the cell cortex. PAR-3, PAR-6 and PKC-3 associate with the anterior cortex, whereas PAR-1 and PAR-2 associate with the posterior cortex(Boyd et al., 1996; Etemad-Moghadam et al., 1995; Guo and Kemphues, 1995; Hung and Kemphues, 1999; Tabuse et al., 1998). PAR functions are crucial for the subsequent anterior-posterior asymmetries of the early cells (reviewed by Kemphues and Strome, 1997). The anterior and posterior cells that arise from the first division of the egg differ in size, cleavage rate, spindle orientation, distribution of germline granules (P granules) and expression of several proteins, such as PIE-1, that are encoded by maternally provided mRNAs. par mutant embryos lack each of these asymmetries.

Although the PAR proteins have been studied primarily in newly fertilized,1-cell embryos, they are expressed continuously during the early cell cycles and subsequently in epithelial cells during organogenesis. After the 1-cell stage, the anterior-posterior asymmetry of the PAR proteins is reiterated only in the lineage of cells that eventually produce the germline (germline precursors; see Fig. 1A). PAR-3, PAR-6 and PKC-3 associate with the anterior surface of each germline precursor prior to division, and PAR-1 and PAR-2 associate with the posterior surface (Boyd et al., 1996; Etemad-Moghadam et al., 1995; Guo and Kemphues, 1995; Hung and Kemphues, 1999; Tabuse et al., 1998). By contrast, the embryonic cells that produce only somatic cell types (somatic precursors) undergo a dramatic reorganization of PAR proteins(Boyd et al., 1996; Etemad-Moghadam et al., 1995; Guo and Kemphues, 1995; Hung and Kemphues, 1999; Nance and Priess, 2002; Tabuse et al., 1998). At the early 4-cell stage, the formerly anterior PAR proteins, PAR-3, PAR-6 and PKC-3, localize transiently to the entire cell cortex. By the late 4-cell stage, however, PAR-3, PAR-6 and PKC-3 redistribute to the contact-free,apical, surfaces of cells. The formerly posterior PAR proteins, PAR-1 and PAR-2, localize in a reciprocal manner to sites of cell contact, the basolateral surfaces. The PAR proteins gradually disappear from embryos after the 26-cell stage. PAR-3, PAR-6 and PKC-3 are expressed again at the∼400-cell stage in the epithelial cells of developing organs, where they are localized asymmetrically toward the apical surface(Leung et al., 1999; McMahon et al., 2001). The epithelial functions of the PAR proteins have not been studied in C. elegans. However, Drosophila homologs of PAR-3, PAR-6 and PKC-3 are expressed in epithelial cells where they appear to distinguish apical from nonapical membrane domains (reviewed by Knust and Bossinger, 2003). Mutations in the Drosophila par homologs result in gross defects in the polarity of epithelial cells and can cause epithelial cell sheets to become multilayered (Müller and Wieschaus, 1996; Petronczki and Knoblich, 2001; Wodarz et al., 2000).

What function might the PAR proteins perform in the interval between the 1-cell stage and organogenesis in C. elegans embryos? One possibility is that the anterior-posterior to apicobasal transition in PAR protein distribution has a role in gastrulation. During gastrulation, cells lack asymmetries typical of epithelial cells, such as localized adherens junctions and basement membranes, but have apicobasal asymmetries(Costa et al., 1998; Nance and Priess, 2002) (J.N.,E.M.M. and J.R.P., unpublished). Gastrulation involves the movement, or ingression, of surface cells into the interior of the embryo(Nance and Priess, 2002; Sulston et al., 1983). Ingressing cells show a flattening and contraction of their apical surfaces and nonmuscle myosin becomes enriched at these surfaces(Lee and Goldstein, 2003; Nance and Priess, 2002). The ingressing cells move into a central cavity called the blastocoel. Blastocoel formation appears to involve specialization of the basal surfaces of cells,where PAR-2 and PAR-1 are located (Nance and Priess, 2002). By the end of the 4-cell stage small separations appear between cells along their innermost, basal surfaces; these separations increase in size over the next few cell cycles to form the blastocoel. Similar separations do not develop between the lateral surfaces of cells, which indicates that the lateral surfaces remain tightly adherent. Although the outer, apical surfaces of the early cells do not normally contact other cells, experiments on cultured embryos demonstrate that these surfaces also remain adhesive (Nance and Priess,2002).

Previous studies have shown that the pattern of cell contacts in the early embryo determines the basal position of the blastocoel; 4-cell embryos that are combined head-to-head along their former apical surfaces generate an ectopic blastocoel between these surfaces(Nance and Priess, 2002). Interestingly, the PAR proteins have been shown to redistribute in these same experiments, with PAR-3 and PAR-6 moving to the new contact-free surfaces. Abnormal separations can develop between the lateral surfaces of cells in par-3 and par-6 mutants, but not par-2 mutants, and cell ingressions are either absent or abnormal(Nance and Priess, 2002)(J.N., E.M.M. and J.R.P., unpublished). Thus, PAR-3 and PAR-6 might function in apicobasal asymmetry of early embryonic cells.

Determining whether PAR-3 and PAR-6 function in apicobasal asymmetry in the early embryo is complicated by their roles in anterior-posterior asymmetry at the 1-cell stage. For example, par-3 mutant embryos have highly abnormal patterns of cell cleavage and altered cell fates that might disrupt apicobasal asymmetries indirectly(Kemphues et al., 1988). PAR-3 and PAR-6 are encoded by maternally supplied mRNAs(Kemphues et al., 1988; Watts et al., 1996), and the transition from anterior-posterior to apicobasal PAR asymmetry occurs without embryonic gene transcription (Nance and Priess, 2002). Thus, preventing PAR-3 and PAR-6 function after the 1-cell stage requires a method that selectively removes maternally supplied gene products.

In normal embryonic development, the maternal protein PIE-1 is distributed asymmetrically to the germline precursors, in part through the degradation of PIE-1 in somatic precursors (see Fig. 1A) (Mello et al.,1996; Reese et al.,2000). Analysis of PIE-1 has identified a peptide sequence, the ZF1 domain, which is necessary and sufficient for the degradation of PIE-1 in somatic precursors (Reese et al.,2000). Because the anterior-posterior asymmetry of the PAR proteins is established before the asymmetric degradation of PIE-1(Tenenhaus et al., 1998), we reasoned that a PAR protein coupled to ZF1 might function in anterior-posterior asymmetry before being degraded. We show here that ZF1 coupled to PAR-3 and PAR-6 proteins rescue all the anterior-posterior defects associated with mutations in par-3 and par-6, respectively. However, the coupled proteins disappear prior to gastrulation and the resulting embryos have defects in lateral adhesion and cell ingression. Thus,PAR-3 and PAR-6 have a role in apicobasal asymmetry in the early embryo that is independent of their earlier role in anterior-posterior asymmetry.

Nematodes were cultured and manipulated genetically as described by Brenner(Brenner, 1974). The following mutant alleles and chromosomal rearrangements were utilized: chromosome I(LGI): unc-101(m1), par-6(zu222)(Watts et al., 1996), hIn; LGIII: lon-1(e185), par-3(it71)(Cheng et al., 1995),unc-32(e189), qC1, unc-119(ed3); LGIV: spe-26(hc138ts),him-8(e1489). Transgene insertions zuIs20(par-3::zf1::gfp), zuIs52, zuIs54, zuIs57, zuIs58 (all par-6::zf1::gfp), zuIs45 (nmy-2::gfp),zuIs3 (end-1::gfp) and extrachromosomal array zuEx69(par-6::gfp) were created in this study and are described below. Additional references for mutations listed above can be found in(Hodgkin, 1997).

par-3(ZF1) strains were lon-1 par-3; par-3::zf1::gfp(zuIs20);him-8. par-6(ZF1) strains were unc-101 par-6;par-6::zf1::gfp(zuIs54); other alleles of par-6::zf1::gfp were used where indicated. Because par-3(ZF1) and par-6(ZF1)strains were marked with lon-1 and unc-101 mutations,respectively, lon-1 and unc-101 worms were used as controls for these strains. The him-8 mutation present in the par-3(ZF1) strain increases the frequency of males and does not alter early embryogenesis (Hodgkin et al.,1979) (J.N., E.M.M. and J.R.P., unpublished). A strain of genotype unc-32 par-3; par-3::zf1::gfp(zuIs20) was used to assess the viability of par-3(ZF1) embryos.

Standard techniques were used to manipulate and amplify DNA. Genomic sequences containing par-3, par-6 and nmy-2 were identified using the Wormbase web site(http://www.wormbase.org,release WS54, 2001). Cosmid DNA containing each gene was digested(par-3: F54E7, 16526 bp SalI fragment; par-6:T26E3, 9080 bp XbaI-SmaI fragment; nmy-2: F20G4,13965 bp XbaI fragment) and subcloned into the pBluescript KS+ vector(Stratagene). Subclones of each gene are predicted to include the entire coding region, 3′-untranslated region and 3-5 kb of sequence 5′ of the gene. A PstI site was introduced either just before or in the stop codon of each gene by site-directed mutagenesis (Quickchange kit,Stratagene) or recombinant PCR. To construct par-6::gfp and nmy-2::gfp, the coding region of gfp was amplified by PCR from plasmid pPD95.75 (1995 Fire Lab vector kit, www.ciwemb.edu)and cloned into the introduced PstI site. To construct par-3::zf1::gfp and par-6::zf1::gfp, sequences encoding the ZF1 domain (Reese et al.,2000) were first amplified by PCR from pie-1 cDNA p661(Mello et al., 1996) and introduced into the KpnI site 5′ of gfp in plasmid pPD95.75; the resulting zf1::gfp coding region was amplified by PCR and cloned into the introduced PstI site. Prior to transformation of each construct, the unc-119(+) coding region was inserted into the NotI site in the vector (Maduro and Pilgrim, 1995).

To construct end-1::gfp, the end-1 promoter (2126 bp 5′ to 75 bp 3′ of the end-1 start codon) was amplified by PCR and fused 5′ of gfp coding sequences in plasmid pPD95.75(Cassata et al., 1998).

Strains expressing par-3::zf1::gfp, par-6::gfp, par-6::zf1::gfpand nmy-2::gfp were obtained by microparticle bombardment of unc-119 worms with plasmids described above(Praitis et al., 2001). A strain expressing the end-1::gfp reporter was obtained by injecting spe-26 worms with end-1::gfp and a spe-26(+)cotransformation marker; the resulting end-1::gfp extrachromosomal array was integrated by γ-irradiation(Mello and Fire, 1995).

Anti-PAR-3 monoclonal antibody P4A1 was produced in collaboration with Ken Kemphues. Purified recombinant PAR-3(Etemad-Moghadam et al., 1995)was injected into mice at the FHCRC Hybridoma Production Facility as described(Wayner and Carter, 1987). Hybridoma supernatants were assayed by immunostaining early embryos fixed in bulk with paraformaldehyde and methanol(Costa et al., 1997). Antibody P4A1 stained early embryos in the same pattern as previously described PAR-3 polyclonal sera (Etemad-Moghadam et al.,1995) and did not stain either par-3(it71) or par-3(RNAi) early embryos (data not shown).

For most immunostaining experiments, embryos were fixed on slides using the freeze-crack methanol procedure and incubated with primary antibodies and fluorochrome-conjugated secondary antibodies(Leung et al., 1999); embryos were fixed for PIE-1 immunostaining as described(Mello et al., 1996). The following primary antibodies/antisera and dilutions were used: chicken anti-GFP, 1:200 (Chemicon); mouse anti-HMP-1, 1:10(Costa et al., 1998); rabbit anti-HMR-1, 1:10 (Costa et al.,1998); rabbit anti-LAD-1, 1:300(Chen et al., 2001); rabbit anti-PAR-1, 1:30 (Guo and Kemphues,1995); rabbit anti-PAR-2, 1:3(Boyd et al., 1996); mouse anti-PAR-3, 1:20 (this study); rabbit anti-PAR-6, 1:20(Hung and Kemphues, 1999);mouse anti-PIE-1, 1:10 (Mello et al.,1996); rat anti-PKC-3, 1:10(Tabuse et al., 1998); rabbit anti-PGL-1, 1:1000 (Kawasaki et al.,1998). Images of immunostained embryos were captured on a Deltavision microscope (Applied Precision) and deconvolved. Where not indicated, immunostaining observations were based on the analysis of >15 embryos at the appropriate stage.

Embryos were combined, cultured and immunostained for PAR-3 as described(Nance and Priess, 2002). Wild-type and par-3 mutant embryos were combined at the 2-cell stage such that the anterior cell of the wild-type embryo contacted a par-3mutant cell. Chimeric embryos were cultured for 2-3 division cycles before fixation. Wild-type cells were recognized by their distinctive pattern of cell division and the presence of cortical PAR-3.

lon-1 par-3; par-3::zf1::gfp; him-8 and control lon-1embryos were fixed and processed for electron microscopy as described(Priess and Hirsh, 1986). For each genotype, sections of 40-50 fixed embryos at the 12-15 cell stage were analyzed.

Embryos were mounted and imaged for 3D-timelapse microscopy as described(Nance and Priess, 2002). Fluorescence images of embryos expressing GFP were acquired on a Leica TCS scanning confocal microscope.

Reported cell division times are in minutes from the beginning of the 2-cell stage. Times were normalized to those reported by Sulston et al.(Sulston et al., 1983). Initiation of ingression of mesodermal cells was scored when these cells first began to sink below the surface of the embryo.

Double-stranded RNA (dsRNA) corresponding to bases 5425-5889 of the nmy-1 cDNA was synthesized as described(Nance and Priess, 2002). Young adult hermaphrodites were injected with nmy-1 dsRNA (3.5 μgμl^–1) and 24 hours later eggs at the 1-4 cell stage were collected, mounted and video-recorded for 150 minutes as described above. Ingression of endodermal cells was analyzed as described in the legend to Table 1. Endodermal cells failed to ingress within the recording period in 1 out of 12 nmy-1(RNAi) embryos; reported ingression times are for the remaining 11 embryos. Although it was not possible to monitor NMY-1 protein levels to quantitatively determine the efficacy of RNAi, all injected control(lon-1) embryos displayed defects in the elongation stage of embryogenesis (n=20).

To examine the role of par-3 and par-6 in the apicobasal polarity of early embryonic cells, we first constructed par-6::zf1::gfp and par-3::zf1::gfp transgenes, and a control par-6::gfp transgene. These transgenes were integrated into chromosomes using techniques that allow stable expression in the germ line(see Materials and methods). The PAR-3^ZF1-GFP and PAR-6^ZF1-GFP proteins encoded by the transgenes were localized to the anterior cortex of 1-cell embryos in a pattern indistinguishable from that of the normal PAR-3 and PAR-6 proteins, and control PAR-6^GFPprotein (Fig. 1Ba,b). At each cell cycle after the 1-cell stage, the embryo consists of one or more somatic precursor cells (Fig. 1A, cells outlined in bold) and one germline precursor(Fig. 1A, dark red cells). At each stage, the PAR-6^ZF1-GFP and PAR-3^ZF1-GFP proteins showed an anterior-posterior asymmetry in the germline precursor that was indistinguishable from normal PAR-3 and PAR-6 proteins(Fig. 1Bc,d, small arrows). Therefore, PAR proteins that are linked to the ZF1 domain reproduce the normal anterior-posterior asymmetries of the PAR-3 and PAR-6 proteins in the 1-cell embryo and in germline precursors at subsequent stages.

In contrast to the germline precursors, the levels of PAR-3^ZF1-GFP and PAR-6^ZF1-GFP in somatic precursors diminished to undetectable levels. During the 4-cell stage, when PAR-3, PAR-6 and PAR-6^GFP redistributed to the apical cortices of all three somatic precursors (Fig. 1Bc,large arrow, and data not shown), PAR-3^ZF1-GFP was absent in the two oldest somatic precursors (the ABa and ABp cells; Fig. 1Bd, large arrow). In EMS,the youngest somatic precursor, PAR-3^ZF1-GFP began to redistribute from the cell periphery to the apical cortex(Fig. 1Bd, arrowhead), but disappeared during the following cell cycle. PAR-6^ZF1-GFPdisappeared similarly in the somatic precursors, but usually required an additional cell cycle to do so (Fig. 1Bf). After the 4-cell or 8-cell stages, PAR-3^ZF1-GFPand PAR-6^ZF1-GFP could not be detected in the older somatic precursors by either GFP fluorescence or immunostaining for GFP. For example,at the 26-cell stage, when the endodermal precursors normally begin the first cell ingressions, PAR-3^ZF1-GFP and PAR-6^ZF1-GFP were not detected in the endodermal precursors or the neighboring cells that flank the anterior and lateral sides of the endodermal precursors(Fig. 1Bh and data not shown).

Although normal embryos express PAR-3 and PAR-6 until the 26-cell stage(Fig. 1Bg and data not shown),the levels of these proteins gradually decline during subsequent cell cycles. A second phase of expression coincides with the beginning of organogenesis(∼400-cell stage), when PAR-3 and PAR-6 appear in nascent epithelial cells(Fig. 1Bi and data not shown). We found that PAR-3^ZF1-GFP and PAR-6^ZF1-GFP were expressed in nascent epithelia in a pattern similar to that of endogenous PAR-3 and PAR-6 (Fig. 1Bj and data not shown). This result indicates that the machinery that degrades proteins with the ZF1 domain does not operate at the 400-cell and later stages. In summary, we conclude that the ZF1 domain effectively removes the PAR-3^ZF1-GFP and PAR-6^ZF1-GFP proteins from somatic precursors between the 4- and 8-cell stages until the beginning of organogenesis.

To test whether the PAR-3^ZF1-GFP and PAR-6^ZF1-GFPproteins could provide PAR functions essential for anterior-posterior asymmetry, we crossed the corresponding transgenes into par-3 or par-6 mutant strains, which lack detectable maternal PAR-3 or PAR-6,respectively (Etemad-Moghadam et al.,1995; Hung and Kemphues,1999). For simplicity, we refer to a par-3 mutant with an integrated par-3::zf1::gfp transgene as par-3(ZF1), and to par-3(ZF1) and par-6(ZF1) embryos collectively as par(ZF1) embryos. Although the par-3(ZF1) and par-6(ZF1) embryos had abnormalities in the appearance of early embryonic cells and in cell movements during gastrulation (see below), most of the embryos developed to hatching and grew to fertile adults [eggs hatched in wild type, 1271/1281 (99%); par-3(ZF1), 1061/1085 (98%); par-6(ZF1), 2443/2548 (96%)]. The par-6(ZF1) strain with the lowest level of PAR-6^ZF1-GFP expression (zuIs52, see Table 1) produced some embryos that grew to agametic adults (30/396). This phenotype has been described in strains with hypomorphic alleles of par-3 and par-6, and is likely to reflect suboptimal levels of PAR-6^ZF1-GFP in germline cells (Kemphues et al., 1988; Watts et al., 1996).

PAR-6^ZF1-GFP was expressed asymmetrically at the 1-cell stage and in the germline precursors in par-6(ZF1) embryos and wild-type embryos, and similar results were observed for PAR-3^ZF1-GFP in par-3(ZF1) embryos (Fig. 1Ba,b and data not shown). Thus, the ZF1-tagged proteins provide sufficient par-6(+) or par-3(+) function to promote their own asymmetric localization. We asked whether the anterior-posterior asymmetries that characterize the first division of a wild-type embryo occurred in the par(ZF1) embryos. The first division is unequal in wild-type embryos (Fig. 2A),resulting in a small posterior daughter (the germline precursor), but in par-3 and par-6 mutant embryos the first division is equal(Fig. 2E, Table 1). In all par(ZF1) embryos examined the first cell division was unequal, as in wild-type embryos (Fig. 2I, Table 1). Similarly, the subsequent three divisions of the germline precursors were unequal in the par(ZF1) embryos, as in wild-type embryos (wild type, n=14; par-3(ZF1), n=15; par-6(ZF1), n=14). In wild-type embryos,the first division results in the asymmetric localization of PIE-1 and cytoplasmic granules, called P granules, to the posterior daughter(Fig. 2B,C). These asymmetries are absent in par-3 and par-6 mutant embryos(Fig. 2F,G), but were present in all par-3(ZF1) and par-6(ZF1) embryos(Fig. 2J,K) (P granules: wild type, n=54; par-3(ZF1), n=26; par-6(ZF1), n=62. PIE-1: wild type, n=58; par-3(ZF1), n=35; par-6(ZF1),n=44).

Each of the early cells has a distinctive cell-cycle period in wild-type embryos, whereas all the early divisions are approximately synchronous in par mutant embryos. We determined the cell-cycle periods for each cell through the 50-cell stage in par-3(ZF1) and par-6(ZF1)embryos and found these to be indistinguishable from wild type(Table 1 and data not shown). In normal embryogenesis, the descendants of the E cell, and no other cells,express an endoderm-determining factor called END-1; this asymmetry results from the combination of asymmetrically localized transcriptional regulators and asymmetric cell-cell interactions (reviewed by Maduro and Rothman, 2002). Wild-type and par-3(ZF1) embryos both showed correct expression of an end-1::gfp transgene only in the E daughters(Fig. 2D,L), whereas par-3 embryos either lacked end-1::gfp expression or had faint expression in multiple cells (n=30)(Fig. 2H). Taken together,these observations indicate that the expression of a par::zf1::gfptransgene in a par mutant embryo fully rescues defects in anterior-posterior asymmetry.

PAR^ZF1-GFP proteins were degraded rapidly in the somatic precursor cells of the par(ZF1) strains in a pattern similar to the degradation of PAR^ZF1-GFP proteins in otherwise wild-type strains(Fig. 1Bc-h, Fig. 3C). Because degradation of the PAR^ZF1-GFP proteins occurred progressively with the age of the somatic precursor, we observed several examples where a cell with apical PAR^ZF1-GFP was adjacent to a cell that lacked PAR^ZF1-GFP. This pattern indicates that apical restriction of PAR-3 does not require PAR-3-mediated interactions with neighboring cells. To test this hypothesis further, we combined wild-type embryos with par-3mutant embryos at the 2-cell stage and allowed the chimeric embryos to divide in culture. In each of seven chimeric wild-type/par-3 mutant embryos,PAR-3 was excluded from all surfaces where a wild-type cell contacted a par-3 mutant cell. Instead, PAR-3 was concentrated on the contact-free surfaces of the wild-type cells(Fig. 1Bk,l and Discussion). These results are similar to those of a previous study where wild-type embryos were combined with wild-type embryos(Nance and Priess, 2002) (data not shown).

Genetic studies have shown that the localization of PAR proteins in the 1-cell embryo is interdependent; removal of any one protein can alter the pattern of, or prevent, the cortical localization of the other PAR proteins(Boyd et al., 1996; Etemad-Moghadam et al., 1995; Hung and Kemphues, 1999; Tabuse et al., 1998; Watts et al., 1996). We therefore asked whether the apicobasal localization of the PAR proteins after the 1-cell stage involved a similar interdependence. PAR-1 and PAR-2 are normally associated with the basolateral surfaces of somatic precursors(Fig. 3B,B′ and data not shown). After the degradation of either PAR-3^ZF1-GFP or PAR-6^ZF1-GFP in the somatic precursors of par(ZF1)embryos, PAR-1 and PAR-2 localized to both the apical and basolateral surfaces(Fig. 3D,D′ and data not shown). Thus PAR-3 and PAR-6 functions appear to be required to restrict PAR-1 and PAR-2 to the basolateral surfaces.

We next asked if the apical localizations of PAR-3, PAR-6 and PKC-3 were interdependent. The youngest somatic cells of par-3(ZF1) embryos contained apical PAR-3^ZF1-GFP(Fig. 4C, arrow) and PKC-3 colocalized with PAR-3^ZF1-GFP(Fig. 4D, arrow). The older somatic cells lacked PAR-3^ZF1-GFP and contained cytoplasmic rather than apical PKC-3 (Fig. 4D). Similarly, PKC-3 also failed to localize to the apical cortex in the somatic cells lacking PAR-6^ZF1-GFP in par-6(ZF1) embryos(Fig. 4E,F). By contrast,endogenous PAR-3 showed a robust association with the apical cortex of cells lacking PAR-6^ZF1-GFP in par-6(ZF1) embryos(Fig. 4G,H). The apical localization of PAR-3 in the par-6(ZF1) strain does not result simply from the perdurance of cortical PAR-3 after the degradation of PAR-6^ZF1-GFP: in both wild-type embryos and in par-6(ZF1)embryos, PAR-3 disappeared transiently from the cortex during cell division and reappeared during the next cell cycle. In summary, PAR-3, but not PKC-3,associates specifically with the apical cortex in cells that lack PAR-6. However, neither PAR-6 nor PKC-3 can associate with the apical cortex of cells that lack PAR-3.

In early wild-type embryos the lateral, but not basal, surfaces of interphase cells are tightly adherent. We reported previously that par-3 mutant embryos have abnormally large spaces between the lateral surfaces of cells (Nance and Priess,2002). We therefore examined cell-cell contacts in video recordings of par-3(ZF1) and par-6(ZF1) embryos. Embryos from both strains developed separations between the basal surfaces of cells to form a central blastocoel. However, additional, large separations were present between the lateral surfaces of cells in the par(ZF1) embryos(Fig. 5C, Table 1) but not wild-type embryos (Fig. 5A, Table 1). Focal planes through the par(ZF1) embryos often showed abnormal spaces 1-2 μm wide between the otherwise contiguous lateral surfaces of cells(Fig. 5D, Table 1). Electron microscopy confirmed that the images visible in the light microscope were intercellular spaces rather than intracellular structures(Fig. 5E). We consider it unlikely that lateral adhesion was disrupted by transgene overexpression of the PAR proteins because the same transgenes did not disrupt adhesion when crossed into otherwise wild-type embryos(Table 1).

Cell-adhesion defects in par(ZF1) embryos could result from the mislocalization of adhesive proteins that normally localize to the basolateral surfaces where cells contact other cells. HMR-1, the sole C. eleganshomolog of E-cadherin, is found in wild-type early embryos exclusively at basolateral surfaces, where it colocalizes with HMP-1/α-catenin and HMP-2/β-catenin (Costa et al.,1998). However both HMR-1 and HMP-1 were localized normally to the basolateral surfaces of all par-3(ZF1) embryos examined(Fig. 3E and data not shown). LAD-1, the C. elegans homolog of L1CAM, is believed to have a role in cell adhesion and has been reported to localize to the basolateral surfaces of some embryonic cells (Chen et al.,2001). However, LAD-1 was not restricted to the basolateral surfaces of early embryonic cells in wild-type embryos, and this distribution did not change in par-3(ZF1) embryos (data not shown).

The first cells to ingress during gastrulation are the two endodermal precursors. The endodermal precursors move from the ventral surface into the interior, beginning ∼90 minutes after the first cleavage of the embryo(Sulston et al., 1983). During ingression, neighboring cells spread across the apical surfaces of the endodermal precursors (Lee and Goldstein,2003; Nance and Priess,2002). We found that the ingression of the endodermal precursors in par(ZF1) embryos was markedly slower than wild-type endodermal precursors. For example, at a time when wild-type endodermal precursors had moved 2±1.2 μm (n=5) away from the ventral surface, the endodermal precursors in par-3(ZF1) embryos had either not moved or moved only 0.4±0.5 μm (n=6, P<0.05). The endodermal precursors in wild-type embryos completed ingression in 23 minutes,which corresponds to a single cell cycle(Table 1, Fig. 6Aa-c, Movie 1A). By contrast, the endodermal precursors in par-3(ZF1) and par-6(ZF1) strains required 51 and 43 minutes to complete ingression,respectively, and invariably divided before ingression was complete(Table 1, Fig. 6Ad-f, Movie 1B). Expression of PAR-3^ZF1-GFP or PAR-6^ZF1-GFP in otherwise wild-type embryos did not slow ingression, indicating that the ingression defect was not caused by overexpression of the PAR proteins from the transgenes (Table 1).

Because NMY-2/nonmuscle myosin has been implicated in ingression (see Introduction), we used an nmy-2::gfp transgene to examine cells in living embryos. In both wild-type and par-3(ZF1) embryos,NMY-2^GFP was present at low, uniform levels at the cortices of interphase cells, and concentrated at either the cleavage furrow or midbodies of dividing cells (Fig. 6Bb,arrowhead). In wild-type embryos, the ingressing endodermal precursors showed a progressive enrichment of apical NMY-2^GFP as their apical surfaces flattened (Fig. 6Ba-c,Movie 2A)(4/4 embryos). By contrast, the endodermal precursors in most par-3(ZF1) embryos had no enrichment of apical NMY-2^GFPduring the period ingression normally occurs and little, if any, flattening of their apical surfaces (Fig. 6Bd-f, Movie 2B)(6/7 embryos not enriched, 1/7 enriched slightly). Thus, the slow ingression of the endodermal precursors in par-3(ZF1) embryos was associated with a failure to contract their apical surfaces and correlated with reduced levels of apical NMY-2.

Two genes in C. elegans, nmy-2 and nmy-1, encode nonmuscle myosin heavy chains. Although nmy-2 is required for anterior-posterior polarity and cell division, roles for nmy-1 in the early embryo have not been described (Guo and Kemphues, 1996). Because endodermal cells in par-3(ZF1) embryos eventually ingress, we wondered if NMY-1 might compensate for the reduced levels of apical NMY-2. However, the pattern of endodermal cell ingression was not altered by depleting NMY-1 by RNA-mediated interference and endodermal cells were internalized at 136±6.7 minutes after the 2-cell stage (n=11, compare to par-3(ZF1) and wild type in Table 1).

In normal development, ingression of the endodermal precursors is followed by ingression of a group of mesodermal cells that are descendants of an early embryonic cell called MS. Ingression of these MS descendants begins ∼1 hour after the E daughters begin their ingression(Nance and Priess, 2002; Sulston et al., 1983). We compared the ingression of a pair of MS descendants (MSaaaa and MSaaap) in wild-type embryos and par-3(ZF1) embryos. In wild-type embryos, both cells ingressed ∼11 minutes after their birth (11±2 minutes, n=5). In each of six par-3(ZF1) embryos examined, the same cells had variable defects in ingression. In one embryo, neither cell ingressed during an observation period of two cell cycles. In embryos where the cells eventually ingressed, the first ingression movements were evident 7 minutes later than in wild-type embryos (18±3 minutes after their birth, n=5, P<0.05). Thus ingressions in both endodermal and mesodermal lineages occur more slowly in par-3(ZF1) embryos than in wild-type embryos.

In C. elegans, many maternally expressed gene products perdure over several cell cycles and some can persist throughout embryogenesis into larval stages (Swanson and Riddle,1981). Delineating when such products function often requires the availability of temperature-sensitive mutant alleles of the protein. We have taken an alternative approach to selectively removing PAR proteins by exploiting a protein degradation system that normally removes the PIE-1 protein from somatic cells. We have shown that fusing the ZF1 domain of PIE-1 to the PAR proteins does not impair PAR functions at the 1-cell stage of embryogenesis when they play a crucial role in establishing anterior-posterior asymmetry. However, the PAR^ZF1 proteins disappear in the somatic precursor cells after the 4-cell stage and remain absent until late development when the par genes are transcribed by the embryo. This technique should prove useful for examining additional maternal proteins that are implicated in early asymmetry, and for analyzing somatic functions of proteins that are expressed in both somatic and germline precursors.

During the 4-cell stage of embryogenesis, the PAR proteins undergo a dramatic redistribution along the apicobasal axis. Our results indicate that recruitment of PAR-3 to the apical cortex is a key step in this redistribution, analogous to previous observations on the role of PAR-3 at the 1-cell stage (Watts et al.,1996; Tabuse et al.,1998). We showed that PAR-3 localization to the apical cortex occurs independently of PAR-6 (this study) and PAR-2(Nance and Priess, 2002). Moreover, PAR-3 localization is crucial for recruiting PAR-6 and PKC-3 to the apical cortex, and restricting PAR-2 to basolateral surfaces. Localization of PAR-3 to the apical cortex is not sufficient for the colocalization of PAR-6 and PKC-3: PAR-6 does not colocalize with apical PAR-3 in pkc-3(RNAi)embryos (J.N., E.M.M. and J.R.P., unpublished), and PKC-3 does not colocalize with apical PAR-3 in par-6(ZF1) embryos. Thus both PAR-6 and PKC-3 must be present for either protein to associate with apical PAR-3. Biochemical studies of PAR-3, PAR-6 and PKC-3 homologs in mammalian cells have shown that these proteins can bind to one another directly(Joberty et al., 2000; Lin et al., 2000), indicating that interactions between all three proteins might be necessary to stabilize a complex with apical PAR-3.

How is PAR-3 recruited to the apical cortex? Our previous experiments with recombined embryonic cells demonstrated that PAR-3 is excluded from surfaces that are in contact with neighboring cells(Nance and Priess, 2002). Thus, in a normal embryo PAR-3 would be restricted to the contact-free, apical surface. We have shown here that the exclusion of PAR-3 from contact surfaces is not dependent on the presence of PAR-3 in the neighboring cells. First, par-3(ZF1) embryos can contain young somatic cells with apically restricted PAR-3^ZF1-GFP that are adjacent to older somatic cells that lack PAR-3^ZF1-GFP. Second, wild-type cells that are recombined with par-3 mutant cells correctly localize PAR-3 to their contact-free surfaces. In normal development, PAR-3 must distinguish the apical surface from contact surfaces at each cell cycle because PAR-3 dissociates from and reassociates with the apical cortex before and after each cell division, respectively. In principle, non-PAR proteins could define a location on the apical surface that is maintained during cell division. However, because there is natural variability in cell contacts during cell division, and because our cell recombination experiments show that the apicobasal axis is not fixed (Nance and Priess, 2002), we favor the hypothesis that the apical surface is redefined after each of the early cell divisions.

The observation that PAR-3 is excluded from contact surfaces indicates that proteins that are involved in cell adhesion either directly or indirectly influence PAR-3 localization. In a reciprocal manner, the apical PAR-3 complex appears to either directly or indirectly modulate cell adhesion. We have shown that par(ZF1) embryos can develop prominent gaps between the lateral surfaces of cells, whereas these surfaces are tightly adherent in normal embryos. Similarly, embryo culture experiments have shown that apical surfaces have the potential to adhere on contact with other cells(Nance and Priess, 2002). Thus, the apical PAR-3 complex might concentrate or modify adhesive factors at the apical and adjacent lateral surfaces. Alternatively, the apical PAR complex might direct the vectorial transport to the basal surface of proteins that inhibit cell adhesion; such proteins would mislocalize to lateral surfaces in par(ZF1) embryos. By analogy, an apical PAR complex in Drosophila neuroblasts directs the localization of the Miranda protein to the opposite end of the cell (reviewed by Doe and Bowerman, 2001).

Because cell separations occur between many different types of cells in par(ZF1) embryos, they appear to be caused by a general defect in cell-cell adhesion. This defect is unlikely to result from a failure to transcribe the genes required for cell adhesion because similar separations are not observed when transcription is inhibited during the early cleavage stages (Nance and Priess,2002). Thus, we favor the hypothesis that PAR proteins regulate either the localization or activity of maternally provided proteins that have roles in cell adhesion. Maternally expressed HMR-1/E-cadherin associates with HMP-1/α-catenin and HMP-2/β-catenin at the basolateral surfaces of embryonic cells (Costa et al.,1998). However, we observed that both HMR-1 and HMP-1 localize properly in par-3(ZF1) embryos. Depletion of HMR-1 or HMP-1 does not lead to noticeable defects in the adhesiveness of the early embryonic cells(Costa et al., 1998),indicating that additional adhesive proteins remain to be identified.

Defects in cell adhesion could contribute to the abnormally slow cell ingressions observed in par(ZF1) embryos. When an ingressing cell separates from its neighbors at the surface of a normal embryo, a transient gap is created that is closed by the rapid spreading of neighboring cells. This spreading presumably is mediated by lateral adhesion between the newly exposed surfaces of the neighboring cells. Thus, adhesion between the neighboring cells might exert a squeezing force on the ingressing cell that contributes to the normal speed of ingression.

Defects in apical contraction are a second likely cause of the slowed cell ingressions of par(ZF1) embryos. In normal embryos, NMY-2/nonmuscle myosin concentrates at the apical cortex of ingressing cells, reaching a level comparable to that found in the cleavage furrows of dividing cells(Nance and Priess, 2002 and this study). The apical contraction of an ingressing cell could concentrate a fixed, but initially dispersed, population of cortical NMY-2, and lead to an apparent increase in the level of NMY-2. However, membrane-associated proteins such as LAD-1 do not show a similar behavior during cell ingression, raising the possibility that additional NMY-2 is recruited to the apical surfaces of ingressing cells (Chen et al.,2001) (J.N., E.M.M. and J.R.P., unpublished). Irrespective of the mechanism by which NMY-2 is concentrated in normal ingression, this concentration is either markedly reduced or does not occur in par(ZF1) embryos. Because myosin activity is essential for apical contraction and ingression (Lee and Goldstein, 2003), failure to either concentrate or activate NMY-2 is likely to lead to defects in ingression.

Why doesn't cell ingression fail completely in the par(ZF1)embryos? It is possible that a small but significant amount of PAR^ZF1-GFP persists in ingressing cells. Although we cannot eliminate this possibility, the residual level would have to be below the level of detection of immunocytochemistry using antibodies against either GFP or the various PAR proteins. Moreover, all of the lines of transgenic animals we generated had identical defects in ingression despite considerable variation in their initial level of PAR^ZF1-GFP at the 1-cell stage. A second possibility is that the low level of NMY-2 remaining at the apical surface is sufficient for a weak contraction and slow ingression. It is unlikely that NMY-1 functions redundantly with NMY-2 in this process because depleting NMY-1 in par-3(ZF1) embryos does not further impair endodermal cell ingression. Last, it is possible that the PAR-dependent concentration of NMY-2 at the apical surface functions primarily to increase the efficiency of an otherwise PAR-independent pathway for ingression. Cells that lack PAR^ZF1-GFP proteins retain at least one apicobasal asymmetry, the basolateral localization of HMR-1/E-cadherin, and basal localization of structures such as lamellipodia or filopodia could contribute to ingression. Studies in other systems have shown that the ARP2/3 complex of proteins functions in the nucleation and branching of microfilaments and that it localizes to the leading edges of crawling cells (reviewed by Higgs and Pollard, 2001). Interestingly, depletion of the C. elegans ARP2/3 complex prevents ingression of the endodermal precursors(Severson et al., 2002). Large filopodial-like projections are apparent on the C. elegans endodermal precursors after ingression (J.N., E.M.M. and J.R.P., unpublished), however it is not known whether smaller filopodia and lamellipodia are present during ingression. The possibility that multiple mechanisms contribute to cell ingression in C. elegans is reminiscent of studies on invagination in Drosophila. During Drosophila gastrulation, invaginating sheets of cells flatten and contract their apical surfaces, and apical contraction is associated with an apical accumulation of nonmuscle myosin(reviewed by Leptin, 1999). These shape changes are regulated in part by the Folded gastrulation (Fog)signaling pathway. However, mutations that disrupt the Fog pathway slow, but do not prevent, invagination (Costa et al.,1994; Parks and Wieschaus,1991).

The fact that cell ingressions can be slowed in par(ZF1) embryos in C. elegans without causing embryonic lethality is surprising given the essentially invariant positions of embryonic cells during normal tissue morphogenesis (Sulston et al.,1983). However, examples of natural variability in cell positions have been documented in wild-type embryos, where the mispositioned cells can migrate to their normal location (Schnabel et al., 1997). In addition, mutations that block cell death result in embryos with mispositioned cells. These embryos develop into viable animals that appear superficially normal, although they have numerous defects in cellular anatomy (White et al.,1991). We do not yet know whether the mispositioned cells in par(ZF1) embryos undergo compensatory migrations or whether the resulting animals have anatomical defects that are not apparent by light microscopy.

C. elegans embryos have at least three distinct periods in which the PAR-3 complex must distinguish different cell surfaces. At the 1-cell stage PAR-3 associates with the anterior surface, and at the 4-cell stage PAR-3 associates with the apical surface. In late embryogenesis PAR-3 is localized asymmetrically in epithelial cells, and the apicobasal axis of the internal epithelia is inverted with respect to that of earlier embryonic cells(Leung et al., 1999; McMahon et al., 2001) (J.N.,E.M.M. and J.R.P., unpublished). These localization patterns appear to be specified de novo during each period. Disruption of PAR asymmetry at the 1-cell stage by mutations in par-2 does not prevent apical localization of PAR-3 after the 4-cell stage(Nance and Priess, 2002). Similarly, we showed that the absence of the PAR-3 complex between the 4-400-cell stages in par-3(ZF1) embryos does not prevent the subsequent apical localization of PAR-3 during organogenesis.

The molecular cues used to localize the PAR-3 complex remain to be identified and, at some level, these are likely to vary. For example, sperm position and cell contacts specify polarity at the 1- and 4-cell stages,respectively. Although the mechanism of PAR localization has not been studied extensively in the epithelial cells of C. elegans, genetic studies in Drosophila have identified homologs of proteins in the C. elegans PAR-3 complex that regulate apicobasal polarity in epithelial cells. E-cadherin-mediated cell adhesion is required for apical PAR-3 complex localization in Drosophila epithelial cells(Bilder et al., 2003), whereas HMR-1/E-cadherin is not essential for PAR-3 complex asymmetry at either the 1-cell or 4-cell stage in C. elegans(Costa et al., 1997; Nance and Priess, 2002). Apical localization of the PAR-3 complex in Drosophila epithelia is antagonized by a basolateral complex of proteins that includes Discs large and Scribble (Bilder et al., 2003; Tanentzapf and Tepass, 2003). The C. elegans homologs of the latter proteins, DLG-1/Discs large and LET-413/Scribble, are expressed in epithelial cells, and depletion of these proteins causes epithelial defects (reviewed by Knust and Bossinger, 2003). However, these proteins do not appear to function in apicobasal polarity of early embryonic cells because they are either not expressed in the early embryo (DLG-1) or are not required for apical localization of PAR-3(let-413) (J.N., E.M.M. and J.R.P., unpublished). Thus, identifying the molecular basis of cell-contact-dependent PAR localization remains an important goal for future studies on apicobasal PAR asymmetry.

We thank Lihsia Chen, Andy Fire, Ben Leung and Susan Strome for providing antisera and plasmids, and Ken Kemphues for generously supplying PAR antibodies. We thank Judy Groombridge and Bobbie Schneider for assistance with electron microscopy, Liz Wayner for antibody production, and Rafal Ciosk and Susan Parkhurst for providing helpful comments on the manuscript. 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. J.N. is supported by Postdoctoral Fellowship Grant # PF-02-007-01-DCC from the American Cancer Society, E.M.M. is supported by the NIH (NIGMS 5P50 GM66050-02) and J.R.P. is supported by the Howard Hughes Medical Institute.