The maternal gene spn-4 encodes a predicted RRM protein required for mitotic spindle orientation and cell fate patterning in early C. elegans embryos

The asymmetric partitioning of developmental potential to the daughters of polarized cells is a central theme in development. During embryogenesis in the nematode Caenorhabditis elegans, the first mitotic division is asymmetric (Sulston et al., 1983). The zygote, a cell called P₀, divides into a larger anterior daughter, AB, and a smaller posterior daughter, P₁ (Fig. 1). Subsequently, P₁ and three of its immediate descendants undergo a sequence of four additional asymmetric divisions that establish much of the nematode body plan (Bowerman and Shelton, 1999).

Each asymmetric division in the early embryo is preceded by a rotation of the nucleus and its associated pair of centrosome-nucleated microtubule asters. These rotations align mitotic spindle axes with the anteroposterior (AP) axis of cell polarity (Hyman and White, 1987; Schierenberg, 1987; Hyman, 1989), and spindle rotation appears to promote the differential segregation of cytoplasmic determinants that become polarized in distribution along the AP axis before division. For example, cytoplasmic ribonucleoprotein particles called P granules, required for germline fate, are segregated to a single germline precursor during each of the four asymmetric divisions that produce the germline progenitor P₄ (Strome and Wood, 1983; Hird et al., 1996; Kawasaki et al., 1998). Before division in P₀ and P₁, P granules move towards the posterior pole and accumulate at the cortex. The P₀ and P₁ centrosome pairs initially are aligned roughly perpendicular to the AP axis, but during prophase they rotate about 90° (Hyman and White, 1987). Thus, the posterior daughters, P₁ and P₂, inherit nearly all the P granules.

Genetic and molecular studies have shown that, in addition to P granules, many cytoplasmic developmental regulators in the early C. elegans embryo become polarized in distribution along the AP axis prior to asymmetric divisions and, thus, are differentially segregated to anterior and posterior daughter cells (Guo and Kemphues, 1995; Etemed-Mogadham and Kemphues, 1995; Boyd et al., 1996; Draper et al., 1996; Guedes and Priess, 1996; Mello et al., 1996; Tabuse et al., 1998; Hung and Kemphues, 1999; Schubert et al., 2000). These polarized regulators include a group of cortically localized kinases and PDZ domain proteins encoded by the par genes and the pkc-3 gene. The widely conserved PAR and PKC-3 proteins are required for the asymmetric division of P₀, and homologues in other organisms also regulate cell polarity (Doe and Bowerman, 2001). In C. elegans, the P₀ zygote normally produces two daughters of unequal size that subsequently divide asynchronously with orthogonally oriented mitotic spindles (Fig. 1). In par/pkc-3 mutant embryos, both P₀ daughters are equal in size and divide synchronously. At the two-cell stage in par-2 mutant embryos, both mitotic spindles remain transversely oriented, while in par-3 mutants, both spindles rotate to become longitudinally oriented. In par-2; par-3 double mutants, as in par-3 single mutants, both spindles rotate, indicating that PAR-2 and PAR-3 are not required for spindle rotation itself but restrict it to P₁ (Cheng et al., 1995; Hung and Kemphues, 1999).

Asymmetric cell divisions in the early embryo are alike in that mitotic spindles rotate to lie along the AP axis, but each rotation differs in detail. In P₀, cortically localized components of the dynein microtubule motor complex are required for spindle rotation (Skop and White, 1998; Gönczy et al., 1999). By interacting with astral microtubules throughout the cortex, dynein-mediated forces may align the P₀ spindle axis with the longer AP axis of the oblong zygote. In P₁, spindle rotation requires a more localized capture of astral microtubules by a discrete ‘remnant’ complex, left after the completion of P₀ cytokinesis and marked by a small, persistent ring of microfilaments (Hyman and White, 1987; Hyman, 1989; Waddle et al., 1994). Upon capture of astral microtubules emanating from one pole, the nucleus and its associated centrosomal asters rotate as a complex, in response to force(s) that pull one centrosome towards the remnant site. While apparently intrinsic mechanisms act in P₀ and P₁, spindle rotation in a daughter of P₁ called EMS requires an inductive Wnt signal (Goldstein, 1995; Schlesinger et al., 1999). Thus, at least in part, distinct mechanisms act during each rotation. Consistent with this view, rotation is abnormal at the two-cell stage in par and pkc-3 mutants (see above), but P₀ rotation appears roughly normal (Rose and Kemphues, 1998).

To identify the mechanisms that position mitotic spindles, we used genetic screens to find embryonic-lethal mutants with defects in mitotic spindle orientation. We isolated three mutant alleles of a C. elegans gene we name spn-4, for spindle-orientation defective-4. The predicted SPN-4 protein contains a single RNA recognition motif (RRM) and is a member of a small subfamily of conserved worm, fly and mammalian RRM proteins. In spn-4 mutant embryos, the first mitotic division appears normal, but at the two-cell stage, the posterior mitotic spindle fails to rotate and by metaphase re-specifies the cellular axis of polarity. Maternal expression of spn-4 also is required for at least three other developmental processes. We therefore suggest that the predicted SPN-4 RRM protein post-transcriptionally influences the expression of multiple factors, each of which acts more specifically to regulate developmental pathways in early C. elegans embryos.

We cultured C. elegans as described (Brenner, 1974). N2 Bristol was used as the wild-type strain. The following alleles and balancer chromosomes were used. Linkage group (LG) III lon-1(e185), par-3(it71) and qC1 (LGIII inversion balancer: dpy-19ts, glp-1). LG IV him-8(e1489). LG V dpy-11(e224), unc-42(e270), nT1 (LG IV/V translocation balancer. let-?), DnT1(LG IV/V translocation balancer, Dominant unc-?(n754) let-?), spn-4 (or25, or80 and or191ts). LG X lin-2(e1309). The following strains were used: KK571 par-3 lon-1/qC1 III, DR108 dpy-11 unc-42 V, EU4 lin-2 X, EU923 spn-4(or191ts) V, EU588 him-8(e1489) IV; spn-4(or191ts) V, EU769 +/nT1 IV; spn-4(or25) unc42/nT1 V, EU370 +/DnT1 IV; spn-4(or25)/DnT1, EU772 +/DnT1 IV; spn-4(or80)/DnT1 and EU918 par-3 lon-1/qC1 III; spn-4(or191ts)V. The temperature-sensitive allele of spn-4(or191ts) was maintained by growing homozygous animals at 15°C. To observe the mutant phenotype, L4 larvae were raised overnight at 25°C before collecting early embryos from young adult hermaphrodites.

The following strains carrying integrated cell-type-specific green fluorescent protein (GFP) reporter transgenes were used to facilitate identification of specific cell types or specific gene expression.

PD4251: myo-3::gfp ccIs4251[myo-3::GFP]I, GFP expressed in all body wall muscles and vulval muscles (Liu and Fire, 2000) (provided by A. Fire, Carnegie Institute of Washington).

JH227: pie-1::gfp axEx73[pJH3.92;pRF4], PIE-1::GFP expressed in germ line precursors (Reese et al., 2000) (provided by G. Seydoux, Johns Hopkins University School of Medicine).

OK0029: ceh-22::gfp cuIs1[ceh-22::GFP;pRF4], GFP expressed in pharyngeal muscle cells (Okkema et al., 1997) (provided by P. Okkema, University of Illinois).

DP132: unc-119::gfp [pDP#MMUGF12;pRF4], GFP expressed in most or all neuronal cells (D. Pilgrim personal communication; provided by D. Pilgrim, University of Alberta).

JR186: med-1::gfp wIs93[pMM280;pRF4], GFP expressed in E and MS and their descendants (Maduro et al., 2001) (provided by M. Maduro and J. Rothman, University of California, Santa Barbara).

JR553: end-1::gfp wIs28[end-1::GFP::lacZ;pRF4] (Maduro et al., 2001) (provided by M. Maduro and J. Rothman, UCSB).

JM62: elt-2::GFP wIs84[elt-2::GFP, pFR4] (Maduro et al., 2001) (provided by J. McGhee, U. Calgary).

HC14: vab-7::gfp qtIs1[vab-7::GFP; pRF4], GFP expressed in the posterior great-grand-daughters of C (Cxxp cells; C. Hunter, personal communication, provided by C. Hunter, Harvard University).

The or25 and or80 alleles were isolated in genetic screens for non-conditional maternal-effect embryonic-lethal mutants (Kemphues et al., 1988), while the or191ts allele was isolated in genetic screens for temperature-sensitive embryonic-lethal mutants (Encalada et al., 2000). The or191ts allele is 100% embryonic lethal when homozygous mothers are raised at 25°C (n=579). At 15°C, 84% of the embryos hatch (555/661). All three mutant alleles were mapped to center of linkage group V using meiotic segregation and recombination to position the alleles relative to previously identified loci (data not shown). The three alleles failed to complement each other in crosses that generated all possible hetero-allelic outcross progeny (data not shown). Linkage group and three-factor mapping were used to position the or191ts allele at approximately +0.1149 on chromosome V. From a strain with the genotype spn-4(or191ts)/dpy-11 unc-42 V, 27/29 Unc-nonDpy recombinants picked up spn-4(or191ts) while 2/29 did not. Of 48 Dpy-nonUnc recombinants, two picked up spn-4(or191ts) and 46 did not. The essential requirements for spn-4 are strictly maternal, as wild-type males crossed with homozygous spn-4(or80) hermaphrodites do not rescue the embryonic-lethality (broods with more than 290 embryos were scored; none of 1047 embryos from three broods hatched). To test for zygotic requirements during embryogenesis, broods of embryos were collected from heterozygous spn-4(or80)/+ hermaphrodites at room temperature. Nearly all the embryos hatched (1262/1269). 99 progeny from or80/+ hermaphrodites were followed from hatching until adulthood. 21/99 grew to adulthood but produced broods of dead embryos, the remaining 78 produced viable progeny. No other phenotypes were observed. We conclude that maternal expression is sufficient for all essential spn-4 requirements. Homozygous spn-4(or191ts) L1 larvae raised at 25°C grow to adulthood and produce broods of dead embryos, suggesting there are no zygotic requirements for spn-4 during post-embryonic development.

To clone spn-4, we used a combination of germline transformation with wild-type genomic DNA (Mello et al., 1991) and RNA-mediated interference (RNAi) (Fire et al., 1998). Germline transformation with a single cosmid, T11B11, rescued the broods of embryos produced by homozygous spn-4(or191ts) hermaphrodites at 25°C. The concentration of injected cosmid DNA was 5 μg/ml. A dominant rol-6 gene was used as marker for transformation by co-injecting the plasmid pRF4 at 100 μg/ml (Mello et al., 1991). Genomic yeast DNA was used as carrier at 100 μg/ml. Eight stable transmitting lines were obtained using cosmid T11B11; in four lines the embryonic lethality of or191ts was complemented. We then used RNAi to reduce the function of each predicted gene in T11B11. Microinjecting into the ovaries of wild-type hermaphrodites dsRNA made from a 640 base pair fragment of DNA amplified by PCR from exon 5 of the predicted gene ZC404.8 resulted in embryonic lethality with a spn-4-like mutant phenotype. To confirm the gene identity we sequenced ZC404.8 (GenBank Accession Number, CE07598) in the three independently isolated spn-4 mutant strains. Five fragments of 500-600 bp, overlapping roughly 100 bp, were amplified from genomic DNA using PCR (Williams, 1995). Bands were excised from agarose gels and purified with GeneClean II (Bio 101, La Jolla, CA) and cloned into pGEM-T vector (Promega, Madison, WI). Sequencing was carried out at the University of Oregon DNA sequencing facility, using an ABI 377 Prism automated fluorescent sequencer. Clones from two independent PCR reactions for each allele were sequenced and compared with sequences from lin-2(e1309) animals, the parental strain used for mutagenesis. Codon 75 of or191ts allele is mutated from GGA to GAA, resulting in a G to E substitution. Codon 91 of the or25 allele is mutated from GGG to GAG also resulting in a G to E substitution. The sequence of or80 revealed a 617 bp deletion in the 3′ untranslated region (3′ UTR).

To identify related proteins, we compared sequences in databases using the gapped Blast program (http://www.ncbi.nim.nih.gov). We found two C. elegans, one fly, one mouse and two human sequences related to SPN-4. The related proteins exhibit substantial sequence identity within their RRMs, in addition to having conserved residues throughout their length. For phylogenetic analysis, we included the full sequences of SPN-4 and related proteins, and the Saccharomyces cerevisiae RRM protein NSR1 was used as the outgroup. Protein sequence alignments and phylogenetic tree construction were done using the MacVector software.

To reduce gene function using RNAi (Fire et al., 1998), dsRNA was made in vitro from cDNA clones and microinjected into wild-type ovaries at a concentration of at least 1 μg/μl. The cDNA clones used and the corresponding genes were the following: pie-1, yk613c12; skn-1, yk18e11; pal-1, yk213g9; par-3, yk344b2 (all provided by Yuji Kohara) and mom-2, pJC359 (a full-length cDNA cloned into the pGEM-T vector). spn-4 dsRNA was made from the clone pZG05, a PGEM-T vector containing a 640 bp insert that was amplified from ZC404.8 exon 5, using the T11B11 cosmid as template. RNAi depleted embryos were harvested 16-24 hours post injection.

Laser ablations were performed as described (Avery and Horvitz, 1989), using a Zeiss Axioskop equipped with Nomarski DIC optics, a VSL-337 laser (Laser Science) and a MicroPoint fiberoptic laser adaptor (Photonics Instruments). Embryos were collected from a watch glass filled with M9 medium using a mouth pipette, after cutting open young adult hermaphrodites with a scalpel blade. Embryos were transferred to 3% agarose cushions and overlaid with a 22×22 mm glass cover-slip for viewing with Nomarski optics. To calculate the volume of sister blastomeres, the diameter of each blastomere was measured at the end of cytokinesis of the parent cell, spherical shape was assumed for volume calculation (Tabara et al., 1999). Mitotic spindle orientation angles were measured from digital images of intact embryos made using Nomarski DIC optics. Spindle orientations in P₀ and P₁ are shown as the angle made by a line connecting the centrosomes relative to the long (AP) axis of the embryo. Measurements were made just after nuclear envelope breakdown. A similar distribution of spindle orientation defects was observed in embryos mounted within a hanging drop of M9 medium suspended from a glass coverslip mounted over a depression slide. In spn-4(or80) mutants, the AB spindle was transversely oriented in 6/8 embryos and was at about 45° in 2/8 embryos; the P₁ spindle was transversely oriented in 2/7 embryos, about 45° in 4/7 and resembled wild type in 1/7 embryos.

Immunostaining was carried out using the following mouse monoclonal antibodies and dilutions: 3NB12 (1:100) to detect a myosin specific for pharyngeal muscle cells (Miller et al., 1983), NE8 4C6.3 (1:100) to detect body wall muscle cells (antibody collection MRC LMB, Cambridge) (Kaletta et al., 1997), J126 (1:100) to detect intestinal cells (Mango et al., 1994), Dm1α to detect α-tubulin (Sigma), CS1 to detect MEX-5 (Schubert et al., 2000) and F2A to detect SKN-1 protein (Bowerman et al., 1993). The following rabbit polyclonal antibodies and dilutions were also used: anti-PGL-1 (1:100) to detect P granules (Kawasaki et al., 1998) (gift from S. Strome) and anti-PAR-2 (Boyd et al., 1996) (gift from K. Kemphues). DNA was stained with DAPI or TOTO (Molecular Probes). Immunostaining with the antibodies J126, 3NB12, NE8 4C6.3 and rabbit anti-PGL-1 was done using the cold methanol/acetone fixation protocol (Bowerman et al., 1993). SKN-1 (Bowerman et al., 1993), PAR-2 (Shelton et al., 1999), MEX-5 (Lin et al., 1995) and α-tubulin (Severson et al., 2000) immunostaining was as described. Birefringent gut granules characteristic of intestinal cells were detected using polarizing optics (Bowerman et al., 1992). Hypodermal cells were identified based on their morphology as observed using Nomarski DIC optics: they have characteristically flattened cell shapes with relatively large smooth nuclei containing a single prominent nucleolus. Digital Nomarski images were acquired using a DAGE MTI VE1000 camera under the control of Scion image software. Digital fluorescence microscopy image acquisition was performed using either a BioRad MRC 1024 confocal microscope, or a Zeiss Axioskop equipped for epifluorescence with a Micromax EBF512 cooled CCD camera (Roper Scientific).

To identify genes required for mitotic spindle orientation, we screened mutagenized populations of worms for embryonic-lethal mutants with cell fate patterning and/or cell division defects (Materials and Methods). In one conditional mutant and two non-conditional mutants, the posterior mitotic spindle at the two-cell stage frequently fails to rotate prior to cell division (Fig. 1). These recessive, maternal-effect mutant alleles define a single genetic locus we call spn-4 (Materials and Methods). The essential requirements for spn-4 are strictly maternal, and we have not detected any zygotic requirements (Materials and Methods). In all three mutant strains, spindle positioning is normal at the one-cell stage, and although the posterior two-cell stage spindle fails to rotate, embryonic cell divisions remain asymmetric (Fig. 1).

After identifying spn-4, we used genetic and molecular methods to determine that this locus corresponds to the predicted gene ZC404.8 (Materials and Methods). We sequenced ZC404.8 in spn-4 mutants and found mis-sense mutations in the alleles or25 and or191ts (Fig. 2A), and a 617 base pair 3′UTR deletion in or80 (Materials and Methods). Reducing ZC404.8 function using RNA interference (RNAi; see Materials and Methods) results in a mutant phenotype nearly identical to that caused by our strongest mutant allele. We therefore have used RNAi, in addition to mutant alleles, to define requirements for spn-4.

The predicted ZC404.8 open reading frame encodes a 351 amino acid protein with a single RNA recognition motif (Fig. 2A), the most common RNA binding motif known (Haruhiko and Dreyfuss, 1997; Varani and Nagai, 1998). BLAST searches show that SPN-4 is a member of a small subfamily of RRM proteins (Fig. 2; see Discussion). The mis-sense mutations in spn-4(or25) and spn-4(or191ts) affect highly conserved residues in the RRM (Fig. 2A).

We conclude that the SPN-4 RRM protein may act indirectly to regulate spindle rotation by post-transcriptionally modulating the expression of protein(s) that participate more directly.

In two-cell stage spn-4 mutants, both mitotic spindles remain transversely oriented, as in par-2 mutant embryos (see Introduction). In contrast to par-2 mutants, though, the first division is unequal in spn-4 embryos, and the larger anterior daughter divides before the smaller posterior daughter, as in wild type (Fig. 1). Thus, spn-4 appears to be required more specifically than is par-2 for rotation in P₁. To further compare the roles of the par-2 and spn-4, we examined spindle orientation in par-3; spn-4 double mutants. In contrast to par-3 mutants and par-2; par-3 double mutants (see Introduction), both mitotic spindles fail to rotate in most par-3; spn-4 double mutant embryos (Fig. 3). In other respects, the par-3; spn-4 double mutants resemble par-3 mutants: the first division is equal and the two daughters divide synchronously (Fig. 3).

We conclude that PAR-2 and PAR-3 have earlier and more general roles in the regulation of anterior-posterior asymmetry, and that SPN-4 acts downstream of, or in parallel to the PAR proteins to more specifically regulate spindle rotation in P₁.

Although the posterior mitotic spindle at the two-cell stage in a spn-4 mutant fails to rotate, the cell in which it resides nevertheless divides asymmetrically. In wild-type embryos the volume of the anterior daughter, EMS, is two to three times the volume of the posterior daughter, P₂ (n=11; Materials and Methods), and EMS divides before P₂ (Fig. 1E,F). In spn-4 mutant embryos the same relative volumes are maintained (or25 n=6, or80 n=5, or191ts n=6), and the ‘P₁’ descendants divide asynchronously with roughly normal spindle orientation (Fig. 1). Thus spn-4 is required for rotation specifically in P₁ and not in other asymmetrically dividing embryonic cells.

We also examined other asymmetries that normally appear during the division of P₁, including the segregation of cytoplasmic P granules to germline precursors. In wild-type embryos, P granules move posteriorly prior to the asymmetric division of germline precursors (see Introduction). In spn-4 mutant embryos, P-granule segregation in P₀ appears normal. Early in mitosis at the two-cell stage in spn-4 mutants, during prophase in ‘P₁’, P granules again appear normal, being enriched at the posterior cortex (Fig. 4E,F; n=12). However, by pro-metaphase, P granules are localized laterally, near one pole of the transversely oriented posterior mitotic spindle (7/8 embryos). In all two-cell stage spn-4 embryos fixed during metaphase (n=3; Fig. 4H) or during anaphase or telophase (n=6), P granules were always centered around one spindle pole. In four-cell stage embryos, P granules were present only in the smaller daughter of ‘P₁’ (Fig. 4I,J; n=18). In embryos allowed to develop in culture medium after removal of their eggshells at the two-cell stage, P granules were present in only a single cell at the four-cell stage in wild-type (n=4) and in spn-4 mutant embryos (n=3). Therefore constraints imposed by the eggshell on spindle orientation do not appear to influence asymmetric segregation of P granules in spn-4 embryos.

We next examined the distribution of PAR-2 in spn-4 mutant embryos. As in wild type, we found that PAR-2 is restricted to the posterior cortex in P₀ (J. E. G., data not shown) and initially in ‘P₁’ as well (n=6; Fig. 4A,B). However, PAR-2 redistributes when the P₁-like cell enters mitosis, forming by metaphase or anaphase a laterally positioned cortical crescent centered around one spindle pole (n=7; Fig. 4D).

As a final assay for cell polarity in spn-4 mutant embryos, we examined the distribution of MEX-5, a CCCH finger protein that becomes enriched in the anterior cytoplasm of dividing germline precursors; MEX-5 and a nearly identical protein MEX-6 are required to restrict PIE-1 and other proteins to posterior cytoplasm (Schubert et al., 2000). As in wild type, MEX-5 was present at higher levels in ‘AB’ than in ‘P₁’ in all fixed two-cell stage spn-4 mutant embryos (n=10), and we also detected MEX-5 at higher levels in ‘EMS’ than in ‘P₂’ in four-cell stage embryos (n=12).

We conclude that the P₁-like cell still divides asymmetrically and correctly segregates many cell fate determinants in spn-4 mutants, at least in part because the mitotic spindle can reorient the axis of cell polarity.

After determining that the division of the P₁-like cell in spn-4 mutant embryos remains asymmetric, we further assessed its development by examining the differentiated cell types it produces. In a wild-type embryo, P₁ divides to produce a posterior mesectoderm and germline precursor called P₂, and a ventral mesendoderm precursor called EMS (Sulston et al., 1983) (see Fig. 1). To assess EMS fate in spn-4 mutants, we first examined the production of intestinal cells. In a wild-type embryo, the posterior daughter of EMS, called E, makes all 20 of the intestinal cells produced in C. elegans. Using polarizing optics to score terminally differentiated embryos for the presence or absence of birefringent gut granules, we determined that only 43% of spn-4(RNAi) mutant embryos (41/95) and 59% of spn-4(or80) embryos (83/140) make intestinal cells. About 80% of spn-4(or25) embryos (66/89), and nearly all spn-4(or191ts) embryos (81/84) make gut. Similarly, using an antibody that recognizes a surface antigen on intestinal cells, we found that 39% of spn-4(or80) embryos (47/121) produce positively staining cells (Fig. 5A; see Materials and Methods). Even though the ‘P₁’ spindle remains transversely oriented in spn-4 mutants, one daughter is always larger, and the larger daughter always divides first, at least in part along the AP axis, to produce a posterior E-like daughter (Fig. 1). We used a laser micro-beam to kill this E-like cell in spn-4(RNAi) and spn-4(or80) mutant embryos and found that no gut granules were made (n=9 and n=13, respectively). Thus endoderm potential is properly segregated to one ‘P₁’ granddaughter in those spn-4 embryos that make intestinal cells, in spite of the spindle rotation defect at the two-cell stage.

Finally, in an attempt to further reduce gene function, we injected spn-4 dsRNA into homozygous spn-4(or80) hermaphrodites, but this did not change the fraction of embryos that made gut granules (38/86 or80RNAi embryos, or 42%) relative to spn-4(RNAi) embryos (43%; see above). We conclude that the large 3′UTR deletion in spn-4(or80) strongly reduces gene function, and RNA interference may completely eliminate it.

We next examined the fate of the anterior daughter of EMS, called MS, which normally produces the posterior part of the pharynx, a neuromuscular feeding organ in the head of a nematode. Two 12-cell stage anterior embryonic cells called ABara and ABalp produce the rest of the pharynx in a wild-type embryo (Sulston et al., 1983), in response to signal(s) from MS (Mango et al., 1994). We used a monoclonal antibody that recognizes a pharyngeal myosin to score the production of pharyngeal muscle cells in fixed, terminally differentiated spn-4 mutant embryos (see Materials and Methods). We found that 83% of spn-4(or80) mutant embryos (77/93) fail to produce any pharyngeal muscle cells (Fig. 5A), while the remainder make only two to five cells, instead of the 39 made by a wild-type embryo (Sulston et al., 1983). Similarly, using a transgenic strain that expresses green fluorescent protein (GFP) in pharyngeal muscle cells (Materials and Methods), we found that 73% of spn-4(RNAi) mutant embryos (91/124) fail to make any pharyngeal muscle, and the remainder (33/124) make only a few GFP-positive cells.

In summary, most spn-4 mutant embryos fail to produce both MS-derived and AB-derived pharyngeal muscle cells, while a smaller fraction of spn-4 mutants fail to make E-derived intestinal cells.

The loss mesendoderm in spn-4 embryos resembles the mutant phenotype observed in embryos lacking SKN-1, a maternally expressed transcription factor required for EMS to produce pharyngeal and intestinal cells (Bowerman et al., 1992) and for MS to induce pharyngeal cell production by AB descendants (Shelton and Bowerman, 1996). Other maternally expressed proteins limit SKN-1 function to EMS, and SKN-1 is required for the ectopic production of pharyngeal cells and intestinal cells in all mutants that fail to limit mesendoderm fate to EMS (Bowerman, 1998). To further compare their requirements, we asked if spn-4, like skn-1, is required for the ectopic specification of pharyngeal and intestinal cells in par-3, pie-1 and mom-2 mutants (Fig. 5B). In par-3 mutant embryos, several AP fate asymmetries are lost (Kemphues et al., 1988; Etemad-Moghadam and Kemphues, 1995), and all four-cell stage blastomeres can produce SKN-1-dependent pharyngeal mesoderm (Bowerman et al., 1997). We found that while par-3(it71) embryos nearly always produce large numbers of pharyngeal muscle (61/66 embryos), most par-3(it71); spn-4(or191ts) embryos fail to produce any pharyngeal muscle (75/87; Fig. 5B; see Materials and Methods). In pie-1 mutant embryos, both EMS and its sister P₂ adopt EMS-like fates, producing twice the normal number of pharyngeal and intestinal cells. PIE-1 is a putative transcriptional repressor that is localized to germline precursors, including P₂, where it prevents SKN-1 from specifying an EMS-like fate (Mello et al., 1992; Mello et al., 1996) (see Fig. 6). We can detect large numbers of intestinal and pharyngeal cells in nearly all differentiated pie-1(RNAi) mutants (35/35 and 33/35, respectively), while most pie-1(RNAi); spn-4(RNAi) double mutant embryos fail to produce any intestinal or pharyngeal cells (59/87 and 75/87, respectively; Fig. 5B). In mom-2 mutant embryos, both EMS daughters adopt MS-like fates and produce extra pharynx at the expense of intestine (Thorpe et al., 1997; Rocheleau et al., 1997). However, no pharyngeal muscle cells are made by most mom-2(RNAi); spn-4(RNAi) double mutant embryos (139/157 or 89%, Fig. 5B). We conclude that like the transcription factor SKN-1, the SPN-4 RRM protein is required to specify mesendoderm fate in C. elegans.

In skn-1 mutants, the two daughters of ‘EMS’ develop like C, a daughter of P₂ that produces mesectoderm (Bowerman et al., 1992). We therefore asked if ‘EMS’ in spn-4(RNAi) embryos also produces C-like daughters. In a wild-type embryo, C produces 13 epidermal cells, 32 body wall muscle cells and two neurons (Sulston et al., 1983). To examine EMS fate in spn-4(RNAi) embryos, we used a laser micro-beam to kill all four-cell stage blastomeres except ‘EMS’ in embryos from one of two transgenic strains. One strain expresses GFP specifically in body wall muscle cells and the other expresses GFP in neurons (Materials and Methods). Most isolated spn-4 mutant ‘EMS’ blastomeres (13/18) produced at least 10-20 body wall muscle cells, although 3/18 failed to produce any GFP-positive cells. In 9/9 operated spn-4 embryos, ‘EMS’ produced 1-3 GFP-positive neuronal cells. In all cases, the isolated precursors appeared to produce numerous differentiated descendants with morphologies characteristic of epidermal cells (data not shown; Materials and Methods). These cell fate transformations are consistent with EMS, or its daughters, adopting C-like fate(s) in spn-4 mutant embryos. To further test this hypothesis, we examined the activity of a maternally expressed homeodomain protein called PAL-1 that is required to specify C fate (Hunter and Kenyon, 1996). Normally, PAL-1 activates the expression of another homeodomain protein called VAB-7, which is first detectable in half of the great grand-daughters of C (Ahringer, 1997). We used a laser micro-beam to kill embryonic cells in spn-4(RNAi) and wild-type embryos, using a strain that expresses a vab-7::GFP transgene (Materials and Methods). As expected, we detected no VAB-7::GFP expression after killing P₂ in wild-type embryos (9/9). However, after killing all cells but ‘EMS’ in spn-4(RNAi) embryos, roughly half of the descendants of this cell express VAB-7::GFP (9/9). No VAB-7::GFP-positive cells were detected in 28/29 pal-1(RNAi);spn-4(RNAi) double mutant embryos. We conclude that in spn-4 mutant embryos, as in skn-1 mutants, the daughters of EMS adopt C-like fates as a result of ectopic PAL-1 activity in EMS and its descendants.

In wild-type four-cell stage embryos, both SKN-1 and PAL-1 are present at high levels in the nucleus of EMS, but only SKN-1 appears active (Bowerman et al., 1993; Hunter and Kenyon, 1996). Because PAL-1 is active in EMS in spn-4 mutant embryos, we asked if the normal role of SPN-4 might be to negatively regulate PAL-1 in EMS and thereby permit the specification of mesendoderm by SKN-1. We reasoned that if SPN-4 blocks PAL-1 function in EMS, then depleting PAL-1 function in spn-4 mutant embryos should restore the production of SKN-1-dependent pharyngeal cells. We therefore stained fixed spn-4(or80); pal-1(RNAi) double mutants with an antibody that recognizes pharyngeal muscle (see Materials and Methods). We found that, as in spn-4(or80) single mutants, most spn-4(or80); pal-1(RNAi) double mutants fail to produce any pharyngeal muscle (81/91 embryos), while the remainder (10/91) make only one to five cells.

We conclude that the failure to specify pharyngeal cell fates in spn-4 mutant embryos does not result from the ectopic activity of PAL-1 in EMS, and that SPN-4 does not simply prevent PAL-1 from functioning in EMS.

To account for the lack of mesendoderm in spn-4 mutant embryos, we considered three additional explanations: (1) mis-localization of PIE-1, a negative regulator of SKN-1, might prevent mesendoderm specification; (2) SKN-1 might not accumulate to sufficiently high levels to specify EMS fate; (3) spn-4 might be required to specify EMS fate independent of SKN-1 and PIE-1 function. In wild-type four-cell stage embryos, SKN-1 accumulates to high levels in both P₂ and EMS, but not in the anterior blastomeres ABa and ABp (Bowerman et al., 1993) (see Fig. 6). At the same time, PIE-1 accumulates to high levels only in P₂, where it blocks SKN-1 function (Mello et al., 1996; Seydoux et al., 1996) (see Fig. 6). Because the mitotic spindle in ‘P₁’ fails to rotate in spn-4 mutant embryos, we initially hypothesized that PIE-1 might be mis-localized to both ‘P₁’ daughters and thus prevent SKN-1 from specifying mesendoderm fate in either cell. However, as we observed with P granules and with PAR-2 (see above), the P₁-like cell in a spn-4 mutant embryo still divides asymmetrically, and nuclear PIE-1::GFP is detectable only in its smaller daughter (n=12; Fig. 6D). Because PIE-1 localization appears normal in spn-4 mutant embryos, and because pie-1; spn-4 double mutants fail to produce mesendoderm (Fig. 5B), we conclude that mis-localization of PIE-1 function cannot account for the lack of mesendoderm in spn-4 mutants.

We next asked if SPN-4 is required for the normal expression of SKN-1, using a monoclonal antibody to stain fixed spn-4 mutant embryos (Fig. 6F,H; see Materials and Methods). In 5/10 two-cell stage spn-4 embryos, SKN-1 appeared normal, with detectable levels confined to ‘P₁’. However, SKN-1 is undetectable in some two-cell stage spn-4 mutant embryos (5/10), while we always detect SKN-1 in P₁ at the two-cell stage in wild-type (n=10). To our surprise, high levels of SKN-1 are present throughout four-cell stage spn-4 mutants (8/9 embryos), instead of being confined to the two daughters of ‘P₁’ (1/9), as in wild type (9/9 embryos). In summary, the accumulation of SKN-1 in ‘P₁’ may be delayed in spn-4 embryos, but spn-4 also is required to prevent high levels of SKN-1 from accumulating in AB daughters at the four-cell stage.

Because SKN-1 levels in EMS appear normal in spn-4 mutant embryos, we asked if SKN-1 can activate the zygotic transcription of a downstream target gene called med-1, which encodes a GATA transcription factor required zygotically for mesendoderm fate (Maduro et al., 2001). MED-1 expression requires both SKN-1 and SKN-1-binding sites in the med-1 promoter, and ectopic expression of SKN-1 is sufficient to drive ectopic expression of a MED-1::GFP translational fusion (Maduro et al., 2001). To determine if SPN-4 and SKN-1 act in the same pathway, or in different pathways required for mesendoderm fate, we examined the expression of MED-1::GFP in spn-4(RNAi) embryos. Intriguingly, we found that MED-1::GFP expression is normal in spn-4(RNAi) embryos, appearing at the four-cell stage in EMS and remaining on in EMS descendants (15/15 embryos; Fig. 6J). We also examined the expression of END-1 and ELT-2, two additional GATA transcription factors required for endoderm fate that are expressed in E descendants later in embryogenesis and act downstream of MED-1 (Zhu et al., 1997; Zhu et al., 1998; Fukushige et al., 1998). Using two transgenic strains that express GFP driven by the end-1 and elt-2 promoters (Material and Methods), we detected no end-1 driven GFP expression in all 19 spn-4(RNAi) embryos analyzed. Similarly, we detected elt-2 driven GFP expression in only 19/101 spn-4(RNAi) embryos (data not shown). Although spn-4(RNAi) mutant embryos produce intestinal cells much more frequently than we detect end-1 or elt-2 driven GFP expression, all GFP-positive embryos did produce intestinal cells (data not shown). These results suggest that SPN-4 acts in parallel to SKN-1, in pathways that may converge downstream of MED-1 but upstream of END-1 and ELT-2.

As a final test of how spn-4 and skn-1 might interact to promote mesendoderm fate, we examined the production of intestinal cells in spn-4(RNAi); skn-1(RNAi) double mutants. Reducing or eliminating either spn-4 or skn-1 function results in partially penetrant losses of endoderm: 57% of spn-4(RNAi) embryos, 80% of skn-1(zu67) and 70% (26/38) of skn-1(RNAi) mutant embryos fail to make gut (Bowerman et al., 1992) (data in this paper). All spn-4(RNAi); skn-1(RNAi) double mutants fail to produce birefringent gut granules (n=66). The increased penetrance of the endoderm defect in the double mutant is consistent with SPN-4 and SKN-1 acting in distinct, parallel pathways to promote the specification of endoderm fate.

To further characterize the P₁-like cell in spn-4 mutant embryos, we next scored the cell types produced by its P₂-like daughter. In a wild-type embryo, P₂ divides asymmetrically to produce another germline precursor called P₃ and a larger somatic daughter called C. P₃ divides into the body wall muscle precursor D and the germline progenitor P₄, while C produces skin and body wall muscle and two neurons (Sulston et al., 1983). In spn-4(RNAi) embryos, isolated P₃ and C blastomeres produced body wall muscle cells (5/5 and 9/9 operated embryos, respectively), and 5/5 isolated ‘P₂’ blastomeres produced a normal pattern of VAB-7::GFP expression, as in wild-type embryos.

Although the descendants of P₂ produce roughly normal patterns of cell fate, spn-4 mutant embryos appear to make extra germline (Fig. 5A). In nearly all terminally differentiated spn-4(or80) mutant embryos (61/62), we detected at least six to eight P-granule-positive cells, in contrast to the two produced by a wild-type embryo (Sulston et al., 1983). To determine the origin of the excess germline, we examined P granules segregation in more detail. In wild-type embryos, P granules are sequentially segregated to the germline progenitor P₄, and then to both P₄ daughters, Z₂ and Z₃ (Fig. 7), which then remain quiescent until larval development (Sulston et al, 1983). In spn-4 embryos, P granules are still segregated to P₃ -like (8/8) and P₄-like (13/13) cells. The P₄-like cell then divides prematurely to produce two P granule-positive cells (7/7; Fig. 7). These results suggest that the extra germline we detect in terminally differentiated spn-4 mutants (Fig. 5A) reflects a loss of germline quiescence during embryogenesis, rather than an ectopic specification of germline fate. We also examined expression of the germline protein PIE-1 in more detail. When the P₂-like daughter in a spn-4 mutant divides, PIE-1::GFP is segregated to a single daughter, as in wild type (5/5 embryos). However, when this P₃-like cell divides, PIE-1::GFP is not detectable in either daughter in most embryos (11/13). In 2/13 embryos, PIE-1::GFP was detected only transiently in one of the two ‘P₃’ daughters. Normally, PIE-1 is detectable in P₄ until it divides and is detected transiently in both P₄ daughters (Mello et al., 1996; Seydoux and Strome, 1999).

In summary, development of the P₂-like cell in a spn-4 mutant embryo is largely normal, except for a premature decline in PIE-1 levels and an apparent loss of germline quiescence.

In addition to lacking intestine and pharynx and making extra germline, large numbers of body wall muscle cells are present throughout most terminally differentiated spn-4 mutant embryos (Fig. 5A). As 80/81 body wall muscle cells are made by P₁ in a wild-type embryo (Sulston et al., 1983), the distribution of body wall muscle in spn-4 embryos suggests that both two-cell stage blastomeres might be making body wall muscle. To determine the origin of the muscle cells in spn-4 mutants, we used a laser micro-beam to isolate ‘P₁’ and ‘AB’ in spn-4(RNAi) embryos from a transgenic line that expresses GFP in body wall muscle (Materials and Methods). In most operated embryos, both posterior (11/11) and anterior (8/11) two-cell stage blastomeres produced at least 10-20 GFP-positive cells. Thus anterior blastomeres in spn-4 mutant embryos produce ectopic body wall muscle, indicating that spn-4 is required to restrict muscle specification to descendants of P₁.

Ectopic specification of body wall muscle in AB descendants also occurs in mutants that fail to restrict either SKN-1 or PAL-1 to P₁ descendants (Bowerman, 1998). In mex-1 mutant embryos, SKN-1 is present at high levels in all four-cell stage blastomeres, and ‘AB’ descendants produce ectopic, SKN-1-dependent pharyngeal and body wall muscle (Mello et al., 1992; Bowerman et al., 1993). In mex-3 mutant embryos, PAL-1 is present at high levels in all four-cell stage blastomeres, and ‘AB’ descendants produce ectopic, PAL-1-dependent skin and body wall muscle (Draper et al., 1996; Hunter and Kenyon, 1996). Because spn-4 mutant embryos fail to produce pharyngeal and intestinal cells, it seemed unlikely that mis-localization of SKN-1 (Fig. 6) could account for the ectopic production of body wall muscle by AB descendants. As expected, when we used a laser micro-beam to isolate ‘AB’ in skn-1(RNAi); spn-4(RNAi) double mutants, 9/10 operated embryos still produced numerous GFP-positive body wall muscle cells. By contrast, after staining fixed terminally differentiated embryos with a cell type-specific antibody (Materials and Methods), we found that 35% (28/80) of spn-4(or80); pal-1(RNAi) double mutant embryos failed to produce any body wall muscle cells. Thus body wall muscle production in spn-4 mutant embryos appears largely dependent on PAL-1. Consistent with PAL-1 functioning ectopically in AB descendants, we also detected expression of a PAL-1-dependent VAB-7::GFP fusion protein in descendants of ‘AB’ in 9/9 operated spn-4(RNAi) embryos in which ‘P₁’ was killed with a laser micro-beam.

To summarize, in addition to resembling skn-1 mutant embryos in lacking mesendoderm, spn-4 mutants also resemble mex-3 mutants in failing to restrict PAL-1 activity, and hence mesectoderm fate, to P₁ descendants.

Our analysis of the spn-4 locus suggests that the encoded RRM protein influences multiple developmental pathways early in C. elegans embryogenesis. The posterior two-cell stage mitotic spindle fails to align with the AP axis, mesendoderm is usually absent, mesectoderm is specified ectopically throughout the embryo, developmental regulators are mis-localized, and the germline proliferates abnormally. These defects may reflect requirements for spn-4 in multiple, independent processes. For example, the spindle rotation defect does not appear responsible for the loss of mesendoderm, given the normal segregation of PIE-1 and other germline factors, the general requirement for SPN-4 in other mutants that produce mesendoderm ectopically and the more normal development of mesendoderm in other spindle orientation-defective mutants (Zwaal et al., 1996; Rose and Kemphues, 1998). To influence these different developmental processes, we suspect that SPN-4 regulates multiple more specifically acting regulatory factors.

SPN-4 belongs to a small subfamily of RRM proteins that has three members in C. elegans, one in Drosophila and two in humans. One of the three family members in C. elegans is FOX-1, which acts post-transcriptionally to prevent expression of XOL-1, a key regulator of dose compensation and sex determination in C. elegans (Nicoll et al., 1997). Thus it seems likely that SPN-4 also acts post-transcriptionally, presumably by binding and regulating the expression of target RNAs. Although no analysis of the single Drosophila locus has been reported, one of the two human family members, A2BP1, binds Ataxin 2. Glutamine-expansion mutations in Ataxin 2 are associated with spinocerebellar ataxia (Shibata et al., 2000).

Although spn-4 appears to regulate multiple developmental processes, to our knowledge spn-4 mutants exhibit the most specific defect in P₁ spindle rotation yet described. We do observe some anterior spindle orientation defects at the two-cell stage, particularly in spn-4(or191ts) embryos, curiously the weakest allele with respect to the endoderm defect. However, the posterior rotation defect is substantially more severe and more highly penetrant, particularly in spn-4(or25) and spn-4(or80) embryos. Other maternal genes involved in spindle rotation at the two-cell stage appear to act less directly or less specifically. The PAR and PKC-3 proteins spatially restrict but are not required for mitotic spindle rotation in two-cell stage embryos (Rose and Kemphues, 1998) (see Introduction). Mutational inactivation of three additional genes (let-99, ooc-3 and ooc-5), and RNAi depletion of heterotrimeric G protein subunits and a protein called RIC-8, each result in spindle orientation defects at the two-cell stage (Zwaal et al., 1996; Rose and Kemphues, 1997; Basham and Rose, 1999; Gotta and Ahringer, 2000; Miller and Rand, 2000; Pichler et al., 2000). In let-99 mutant embryos, the ‘AB’ spindle orientation defects are more severe and more penetrant than those in spn-4 mutants, and P₀ spindle rotation is partially defective (Rose and Kemphues, 1997). Furthermore, PAR-2 distribution is normal in let-99 mutant embryos, while P granules are mis-localized to both ‘P₁’ daughters. The let-99 gene encodes a novel protein. Embryos lacking heterotrimeric G protein subunits show defects in centrosome migration and in P₀ spindle positioning, in addition to spindle orientation defects in both P₁ and in AB descendants (Zwaal et al., 1996; Gotta and Ahringer, 2000; Miller and Rand, 2000). Both ooc-3 and ooc-5 single mutants exhibit a highly penetrant failure in ‘P₁’ spindle rotation, but P granules are mis-localized to both daughters and PAR-2 is depleted or absent (Basham and Rose, 1999; Pichler et al., 2000). Moreover, ooc-3 and ooc-5 mutants also have defects in oogenesis and in P₀ spindle rotation, and in ooc-3; ooc-5 double mutants, both two-cell stage mitotic spindles are frequently mis-oriented. The ooc-3 gene encodes a novel protein that localizes to the endoplasmic reticulum, suggesting a possible link between protein trafficking and spindle rotation (Pichler et al., 2000). In summary, spn-4 mutants exhibit a relatively specific defect in P₁ spindle rotation, and thus far only in spn-4 mutants does the mitotic spindle appear capable of re-orienting the axis of cell polarity.

Rotation of the nucleus and centrosomes in P₁ appears to require the capture of astral microtubules by factors present at the remnant site left upon the completion of cytokinesis when P₀ divides (Hyman, 1989; Waddle et al., 1994). Capture of either pole is thought to be sufficient for proper rotation (Hyman, 1989), and the minus-end-directed microtubule motor protein dynein, and associated dynactin components, localize to this site (Gönczy et al., 1999; Skop and White, 1998). In a wild-type embryo, rotation in P₁ is preceded by anterior movement of the centrosomal/nuclear complex towards the remnant site. In spn-4 mutants, anterior movement of the nucleus appears normal (data not shown), but spindle rotation fails or is only partially complete. Thus, spn-4 may be required not for the initial capture of astral microtubules but later and more specifically for rotation itself. Similarly, in two-cell stage ooc-3 and ooc-5 mutant embryos, the posterior nucleus moves anteriorly but fails to rotate (Basham and Rose, 1999). These findings indicate that the capture and rotation machinery may be complex and involve different kinds of force application and movement.

Even when spindle rotation fails completely, the division of the posterior blastomere in two-cell stage spn-4 embryos remains highly asymmetric. The smaller daughter develops much like a wild-type P₂, except for abnormal proliferation of the germline. As in a wild-type embryo, the EMS-like cell in spn-4 embryos divides along the a-p axis, before its P₂-like sister, although it usually fails to produce the pharyngeal and intestinal cells made by EMS in a wild-type embryo. When spn-4 embryos do make endoderm, the posterior daughter of the EMS-like cell produces it, as in wild type. The surprising lack of polarity defects in spn-4 mutant embryos suggests that spindle rotation is not essential for the proper segregation of many cell fate determinants during the asymmetric division of P₁.

To explain the persistence of asymmetry during the division of ‘P₁’ in spn-4 embryos, we suggest that the poles of the mitotic spindle become polarized and, in the absence of rotation, re-specify the cellular axis of polarity. We propose this model based on our observations that both P granules and PAR-2 redistribute by late in the two-cell stage to form a more transverse axis of polarity, and on our results showing that the asymmetric segregation of PIE-1 and MEX-5 appear normal in spn-4 mutants. Previous studies have shown that two distinct mechanisms contribute to P granule localization. P granules move towards the cortex at one pole of asymmetrically dividing cells, and P granules left behind are unstable (Hird et al., 1996). In addition to directional movement and degradation contributing to polarity prior to division, we suggest that the mitotic spindle in P₁ maintains polarity during cell division. Such a role for the P₁ spindle may be particularly important when, owing to constraints imposed by the eggshell, the P₁ and AB spindles adopt variably skewed axes as they elongate during anaphase (see Fig. 1). Intriguingly, P granules in wild-type and spn-4 mutant embryos frequently appear to form linear arrays that project out from one P₁ spindle pole (Fig. 4, Fig. 7). Based on inhibitor studies, P-granule localization to the posterior cortex in P₀ does not appear to require microtubules (Strome and Wood, 1983; Hill and Strome, 1988), although proper P-granule localization in P₂ does (Hird et al., 1996). Based on our results, we suggest that astral microtubules might be important for maintaining an axis of polarity in P₁.

The apparent ability of the P₁ mitotic spindle in spn-4 mutants to re-specify an axis of cell polarity is surprising because previous experiments have suggested that random capture of either spindle pole by the cytokinetic remnant at the anterior cortex appears sufficient to align the P₁ spindle along the AP axis (Hyman, 1989). For example, the remnant site is capable of capturing the opposite spindle pole after the original connection to one pole is severed with a laser micro-beam, although the developmental consequences of such manipulations were not reported (Hyman, 1989). If spindle poles become polarized such that cytoplasmic factors are directed to one pole, either polarization occurs after the normal time of astral microtubule capture and spindle rotation, or capture is not random. A more detailed genetic and molecular analysis of this process should improve our understanding of these issues.

Precedent for mitotic spindles having a role in cell polarity has come from studies of asymmetrically dividing Drosophila neuroblasts (Doe and Bowerman, 2001). Mutations in genes that regulate neuroblast polarity, such as Inscuteable and Pins, cause cortical proteins to be mis-localized early in neuroblast cell division. However, by late in mitosis, these cortical proteins become localized to one pole and the neuroblasts divide asymmetrically, a phenomenon referred to as ‘telophase rescue’ (Peng et al., 2000). Thus in both Drosophila neuroblasts and in the early C. elegans embryo, mitotic spindles may be capable of maintaining and even re-directing axes of cell polarity. In neither case, however, is it known whether the spindle poles are in fact responsible for re-specifying the axis of polarity. For example, in Drosophila neuroblasts it is possible that the cortical proteins become polarized and then orient the mitotic spindle (Peng et al., 2000). To address this issue more conclusively in C. elegans, we are currently analyzing microtubule requirements using GFP translational fusions to visualize the microtubule cytoskeleton and P granules in live embryos.

SPN-4 appears to mediate the establishment of a subset of the asymmetries specified more generally by the earlier acting, cortically localized PAR and PKC-3 proteins. Consistent with this model, spn-4 mutant embryos retain most of the cell size, cell cycle and cell fate asymmetries observed in wild-type embryos, while par mutant embryos suffer widespread, nearly complete losses of these early asymmetries. Moreover, spn-4 is required for mitotic spindle rotation at the two-cell stage in par-3 mutant embryos, and PAR-2 is localized normally in one-cell and pre-prometaphase two-cell stage spn-4 mutant embryos. These results indicate that SPN-4 acts either downstream of, or in parallel to PAR-2 and PAR-3 to regulate P₁ mitotic spindle rotation, but not all asymmetries specified by the PAR proteins.

Our view that SPN-4 mediates a subset of the asymmetries specified by the PAR proteins is similar to the role proposed for MEX-5 and MEX-6, two nearly identical and partially redundant proteins that each contain two CCCH finger motifs (Schubert et al., 2000). MEX-5 is present at high levels in the anterior cytoplasm of asymmetrically dividing germline precursors. PAR-1 at the posterior cortex restricts MEX-5 to the anterior cytoplasm in P₀, while MEX-5 restricts a group of three additional CCCH regulatory proteins to the posterior cytoplasm. The mechanism by which MEX-5 and MEX-6 act is unknown, although mammalian CCCH fingers have been implicated in RNA binding (Bai and Tolias, 1996; Barabino et al., 1997; Lai et al., 2000). Perhaps SPN-4 regulates different target RNAs through a post-transcriptional mechanism distinct from that used by MEX-5 and MEX-6. Consistent with SPN-4 and MEX-5/6 acting in distinct pathways, MEX-5 localization is unaffected by mutations in SPN-4.

Beyond its role in spindle positioning, spn-4 is required for proper cell fate patterning in the early embryo. One such pathway affected by mutational inactivation of spn-4 is the specification of mesectoderm by the homeodomain protein PAL-1 (Hunter and Kenyon, 1996). PAL-1 is present at high levels in P₁ descendants, but not in AB descendants. The KH domain protein MEX-3 is present at high levels in anterior blastomeres and appears to restrict translation of PAL-1 to posterior blastomeres, perhaps by binding 3′UTR sequences in pal-1 mRNA through two KH motifs (Draper et al., 1996; Hunter and Kenyon, 1996). In mex-3 mutant embryos, PAL-1 is present at high levels in the descendants of both P₁ and AB, and the AB descendants ectopically produce PAL-1-dependent body wall muscle. Intriguingly, SPN-4 may influence PAL-1 through MEX-3, as SPN-4 has been identified as a MEX-3-interacting protein (N. Huang and C. Hunter, personal communication). Thus SPN-4 may have important protein partners in addition to presumed RNA targets.

Although spn-4 and mex-3 both restrict PAL-1 function to P₁ descendants, spn-4 and mex-3 mutants exhibit different germline phenotypes. In mex-3 mutant embryos, P₃ divides equally and both daughters, instead of just one, adopt a germline fate, producing extra germline progenitors (Draper et al., 1996). In spn-4 mutants, a single germline progenitor appears to be produced but its two daughters apparently proliferate instead of remaining quiescent during the rest of embryogenesis. Perhaps the premature decline in PIE-1 levels we detect in the spn-4 germline lineage is responsible for the loss of germline quiescence.

SPN-4 also influences at least one other early developmental pathway, the specification of mesendoderm at the four-cell stage in EMS. In this process, SPN-4 appears to act independently of the transcription factor SKN-1, shown previously to be required for mesendoderm fate specification. Like skn-1, spn-4 is required for the specification of mesoderm and endoderm in mutants that produce these cell types ectopically. Moreover, SKN-1 is expressed at roughly normal levels in spn-4 mutant embryos, and med-1, a target gene regulated by SKN-1, is activated in EMS. Curiously, even though SKN-1 is present at high levels in all four-cell stage blastomeres, MED-1 expression in spn-4 mutant embryos is limited to EMS descendants. Presumably SPN-4 regulates the expression of other factors that either contribute directly to the specification of mesendoderm, or somehow provide a permissive environment for the specification of mesendoderm fate by the transcription factors SKN-1 and MED-1.

The maternal contributions to the specification of mesendoderm in C. elegans appear complex. In addition to SKN-1 and SPN-4, a CCCH finger protein called POS-1 also is required to specify the fate of the mesendoderm precursor EMS (Tabara et al., 1999). Intriguingly, SPN-4 is the same as the POS-1 binding protein, PIP-1 (K. Ogura and Y. Kohara, personal communication). Thus, SPN-4 and POS-1 may act in one pathway that contributes to mesendoderm fate specification, or they might integrate independent inputs. In addition to SPN-4, SKN-1 and POS-1, the specification of endoderm also requires a Wnt signal (Rocheleau et al., 1997; Thorpe et al., 1997) and the input of a MAP kinase-related pathway (Meneghini et al., 1999; Rocheleau et al., 1999). How these different maternal inputs interact remains largely unknown. The identification of SPN-4 targets, and the targets of other maternal regulators, should help clarify our understanding of mesendoderm development in C. elegans.

We thank Andrew Fire, Craig Hunter, Morris Maduro, Jim McGhee, David Pilgrim, Joel Rothman and Geraldine Seydoux for providing strains; Ken Kemphues, Jim Priess and Susan Strome for providing antibodies; Yuji Kohara for providing cDNAs; Stephan Schneider for advice on sequence analysis; Yuji Kohara and Craig Hunter for sharing unpublished results; Bruce Draper, Chris Doe, Danielle Hamill and Rebecca Lyczak for helpful comments on the manuscript. J.-E. G. was supported by the Fundação para a Ciência e a Tecnologia (BD18141/98); C. A. S. and K. A. S. were supported by NIH postdoctoral fellowships, S. E. Encalada by an NIH training grant, J. C. C. and B. B. by NIH research grant R01GM49869.