The C. elegans MELK ortholog PIG-1 regulates cell size asymmetry and daughter cell fate in asymmetric neuroblast divisions

Much of metazoan cellular diversity is generated by asymmetric cell divisions in which a polarized mother cell divides to produce daughters that adopt different fates (Horvitz and Herskowitz, 1992). The C. elegans zygote requires the serine/threonine kinase PAR-1 to establish polarity along the presumptive anteroposterior (AP) axis. Loss of PAR-1 activity disrupts asymmetric positioning of the mitotic spindle and the asymmetric distribution of cell fate determinants to daughter cells (Guo and Kemphues, 1995; Kemphues et al., 1988). PAR-1 orthologs in other organisms also regulate cell polarity. In Drosophila, PAR-1 polarizes the microtubule cytoskeleton in the germ line and establishes the embryonic AP axis by regulating the asymmetric localization of cell fate determinants in the oocyte(Cox et al., 2001; Shulman et al., 2000; Tomancak et al., 2000). PAR-1 also regulates epithelial polarity, both in the Drosophila follicular epithelium (Doerflinger et al.,2003) and in cultured MDCK cells(Cohen et al., 2004).

PAR-1 is a member of a superfamily of serine/threonine kinases that regulate cell polarity. In addition to PAR-1, this group includes Kin1p from Schizosaccharomyces pombe and the conserved SAD-1 family. Kin1p regulates bipolar growth and the correct positioning of the nucleus and cell cleavage plane (Drewes and Nurse,2003; La Carbona et al.,2004). Nematode SAD-1 may regulate neural polarity(Crump et al., 2001), and the mouse homologs of SAD-1 are required for neurons to develop processes with normal dendritic and axonal properties(Kishi et al., 2005). Expression of vertebrate PAR-1 homologs in S. pombe can partially rescue kin1 phenotypes, suggesting an ancestral role for this kinase family in regulating cell polarity (Drewes and Nurse, 2003).

Another member of the PAR-1/Kin1/SAD-1 superfamily of serine/threonine kinases is Maternal Embryonic Leucine zipper Kinase (MELK), also known as pEg3 kinase and MPK38 (Blot et al.,2002; Gil et al.,1997; Heyer et al.,1997). MELK proteins share sequence homology and a similar domain architecture with PAR-1 family members. Both contain an N-terminal kinase domain and a C-terminal kinase-associated region. MELK was originally identified in a screen for maternal genes expressed during preimplantation of the mouse embryo (Heyer et al.,1997). MELK is also expressed in other vertebrate tissues,especially in proliferating cell populations, including the blast cells of the early embryo (Heyer et al.,1999), embryonic stem cells(Nakano et al., 2005), adult germ cells (ovaries and spermatogonia)(Heyer et al., 1999),hematopoietic stem cells (Easterday et al., 2003; Saito et al.,2005) and neural stem cells(Easterday et al., 2003; Nakano et al., 2005). MELK has been implicated in the regulation of spliceosome assembly(Vulsteke et al., 2004), gene expression (Saito et al.,2005) and cell proliferation(Davezac et al., 2002; Gray et al., 2005; Nakano et al., 2005). Its role in regulating cell division is controversial, with conflicting evidence suggesting it has both stimulatory and inhibitory functions.

In a screen for C. elegans mutants with defective asymmetric neuroblast divisions, we identified alleles of the gene pig-1(par-1-like gene), which encodes the worm ortholog of MELK. We show that PIG-1 controls cell size asymmetry and neuroblast daughter cell fate in certain neuroblast lineages. PIG-1 acts cell autonomously in affected lineages, and we propose that PIG-1 regulates asymmetric neuroblast divisions by controlling neuroblast polarity. These data are the first to implicate MELK family proteins in regulating cell polarity.

Nematodes were cultured as previously described(Brenner, 1974). N2 Bristol was the wild-type strain used in this study, and experiments were performed at 20°C. The following mutations and transgenic arrays were used.

LG I: ynIs45 [Pflp-15::gfp](Kim and Li, 2004), zdIs5 [mec-4::gfp] (Clark and Chiu, 2003), kyIs39 [Psra-6::gfp](Troemel et al., 1995); LG II: ynIs25 [Pflp-12::gfp](Kim and Li, 2004; Li et al., 1999a), rrf-3(pk1426) (Simmer et al.,2002), juIs76 [Punc-25:gfp](Jin et al., 1999), gmIs20 [hlh-14::gfp](Frank et al., 2003);

LG III: gmIs12 [Psrb-6::gfp](Hawkins et al., 2005; Troemel et al., 1995);

LG IV: pig-1 (gm280, gm300, gm301, gm344) (this study), ced-3(n717) (Ellis and Horvitz,1986), ham-1(n1811, n1438, gm214, gm279)(Desai et al., 1988; Frank et al., 2005; Guenther and Garriga,1996);

LG V: egl-1(n1084 n3082)(Conradt and Horvitz, 1998), ayIs9 [Pegl-17::gfp](Branda and Stern, 2000), muIs102 [Pgcy-32::gfp](Yu et al., 1997) (Y. Lie and C. Kenyon, personal communication). gmIs22 [nlp-1::gfp](Frank et al., 2003; Li et al., 1999b);

Unmapped: mgIs21 [Plin-11::gfp](Hobert et al., 1998); and Extrachromosomal arrays: gmEx320 and gmEx321[Ppig-1::gfp; rol-6(d)], gmEx326 [Ppig-1::pig-1::gfp;rol-6(d)], gmEx394, -95, and -96[Pmab-5::pig-1::gfp; Pdpy-30::NLS::dsRed2; rol-6(d)] (this study).

The pig-1 alleles gm280, gm300 and gm301 were isolated in forward genetic screens for mutations that altered the number of PHB neurons. Mutagenesis was carried out as described by Brenner(Brenner, 1974). Mutagenized gmIs12 P0 worms were allowed to lay eggs. After several days, we transferred 50 F1 animals onto fresh plates. From each plate we transferred 20 F2 hermaphrodites to individual plates and screened their progeny. We screened the progeny of 2400 F2 hermaphrodites.

pig-1(gm344) was isolated from a library of worms mutagenized with UV-trimethylpsoralen. The construction of this mutant library has been described (Withee et al.,2004). The gm344 breakpoints are defined by the following two sequences that flank the 524 bp deletion:5′-AGCTCGTGTCAGGACGCGAA-3′ and 5′-TGGCCACCTTTTGATTTGTC-3′.

The HSN and NSM neurons were detected by staining adult hermaphrodites with rabbit anti-serotonin antibodies as previously described(Garriga et al., 1993). All other neurons were detected with the following integrated transcriptional reporters that express GFP under control of the indicated C. eleganspromoter. PHB neurons: gmIs12 [Psrb-6::gfp](Hawkins et al., 2005; Troemel et al., 1995), which is expressed both in PHB and the adjacent PHA neurons, and gmIs22[nlp-1::gfp], which is PHB-specific(Frank et al., 2003; Li et al., 1999b). PHA, I2 and CAN neurons: ynIs45 [flp-15::gfp](Kim and Li, 2004). AQR, PQR and URX neurons: muIs102 [Pgcy-32::gfp](Yu et al., 1997) (Y. Lie and C. Kenyon, personal communication). ALM, AVM, PVM and PLM neurons: zdIs5 [mec-4::gfp] (Clark and Chiu, 2003). SDQ, SMBD and SMBV neurons: ynIs25[flp-12::gfp] (Kim and Li,2004; Li et al.,1999a). M4 neuron: ayIs9 [Pegl-17::gfp](Branda and Stern, 2000). PVQ neuron: kyIs39 [Psra-6::gfp](Troemel et al., 1995). VC neurons: mgIs21 [Plin-11::gfp](Hobert et al., 1998). VD and DD neurons: juIs76 [unc-25::gfp](Jin et al., 1999).

Live animals were observed as previously described(Sulston and Horvitz, 1977). Embryos were fixed and stained as described by Guenther and Garriga(Guenther and Garriga, 1996). Dye filling of the phasmid neurons with DiI was performed as previously described (Hedgecock et al.,1987). Worms were examined using a Zeiss Axioskop 2 microscope. Images were collected using an ORCA-ER CCD camera (Hammamatsu) and Openlab imaging software (Improvision). Images were prepared for publication using Adobe PhotoShop (Adobe Systems).

L1 animals expressing zdIs5[mec-4::gfp] were mounted and followed by Nomarski optics as previously described(Sulston and Horvitz, 1977). We began following the left or right Q lineage in animals where the Q cell had divided once. After observing an animal, it was transferred to a plate with food. When the animal reached the L4 stage, we used epifluorescence to score the number of A/PVM neurons.

Single nucleotide polymorphism mapping was used to position pig-1on the left arm of LG IV between snp_Y38C1A.2 and snp_K11H12.1(Wicks et al., 2001). We then used RNAi to systematically inactivate each of the ORFs in this region by feeding bacterial strains expressing dsRNA complementary to these ORFs to gmIs12; ced-3(n717) hermaphrodites(Kamath et al., 2001; Timmons and Fire, 1998). RNAi feeding strains used to inactivate genes in this region were obtained from the library designed by the Ahringer Laboratory(Fraser et al., 2000).

We characterized the transcript produced by the pig-1 open reading frame by sequencing eight cDNA clones (yk554F7, yk338H11, yk400B12, yk150E3,yk13C10, yk88F11, yk516B10, yk167G5) kindly provided by Y. Kohara (National Institute of Genetics, Mishima, Japan). To characterize the 5′ terminus of the pig-1 message, we amplified a pig-1 cDNA from an oligo(dT)-primed embryonic cDNA library using the SL1 splice-leader-specific forward primer pSL1 (5′-GGTTTAATTACCCAAGTTTG-3′) and a pig-1-specific reverse primer W03G16N(5′-gagatccacaacacgtatcc-3′). Sequencing this PCR product revealed that SL1 is trans-spliced to the pig-1 message one nucleotide upstream of the pig-1 start codon.

Sequence alignments were performed using ClustalX(Thompson et al., 1997) and dendograms were generated with NJplot(Perriere and Gouy, 1996). The GenBank Accession numbers for sequences used in phylogenetic analyses are: Homo sapiens MELK, NP_055606; MARK2 (PAR-1), NP_059672; SAD1A,AAS86442; SAD1B, AAS86443; Mus musculus MELK, NP_034920; Xenopus laevis pEG3 (MELK), CAA78913; Danio rerio MELK, BAC75706; C. elegans PAR-1, NP_001024019; PIG-1, NP_001023420; SAD-1, NP_510253; Drosophila melanogaster PAR-1, AAF69801; CG6114 (SAD-1), NP_648814; S. pombe kin1, P22987.

To generate the pig-1 transcriptional fusion(Ppig-1::gfp), we amplified the intergenic region between pig-1 and the nearest upstream gene, and cloned the product into the GFP vector pPD95.77 (A. Fire, S. Xu, J. Ahnn and G. Seydoux, personal communication). gmEx320 and gmEx321 were generated by injecting pPD95.77::Ppig-1 (50 ng/μl) and the co-injection marker pRF4 [rol-6(d); 50 ng/μl](Mello et al., 1991) into N2 hermaphrodites.

To generate the pig-1 translational fusion(Ppig-1::pig-1::gfp), we amplified a full-length pig-1 cDNA from plasmid yk400B12 and cloned the product between the pig-1promoter region and GFP coding sequences. gmEx326 was generated by injecting this plasmid into N2 hermaphrodites at 50 ng/μl with 50 ng/μl pRF4.

The Pmab-5::pig-1::gfp construct was generated by ligating a full-length pig-1 cDNA into plasmid pPD95.77::Pmab-5. pPD95.77::Pmab-5 contains the entire mab-5 promoter driving expression of GFP (Fleming et al.,2005). gmEx394, gmEx395 and gmEx396 were generated by injecting this plasmid into zdIs5; pig-1(gm344)hermaphrodites at 10 ng/μl with 50 ng/μl of pRF4 and 50 ng/μl of plasmid Pdpy-30::NLS::DsRed2.

The Pdpy-30::NLS::DsRed2 plasmid was constructed by first amplifying the dpy-30 promoter from plasmid pTY1003 (a gift from Barbara Meyer, University of California, Berkeley, CA, USA) using a reverse primer that contained a start codon followed by the SV40 T antigen nuclear localization signal (NLS) (Kalderon et al., 1984). This product was cloned into vector pPD95.75::DsRed2 (Bulow et al.,2004).

The daughters of the HSN/PHB neuroblast were identified using gmIs20, a nuclear-localized translational fusion between hlh-14 and GFP (Frank et al.,2003). The size of nuclei in these cells is proportional to cell size: these cells have very little cytoplasm and are roughly spherical(Frank et al., 2005). Nuclear diameter was calculated by averaging the length of the long axis of a nucleus with the length of the perpendicular axis of the same nucleus.

The daughters of the Q.p neuroblast were identified using ayIs9(Pegl-17::gfp) (Branda and Stern,2000). We measured cell area in a single plane of focus. These cells are extremely flat, and thus measurements of cell area are good approximations of cell size. We calculated cell area by circumscribing the cell and measuring its interior area with Openlab software (Improvision). We averaged two measurements per cell.

To identify molecules that regulate asymmetric neuroblast divisions in C. elegans, we screened for mutants with defective asymmetric division of the HSN/PHB neuroblast. In wild-type embryos, the bilaterally symmetric HSN/PHB neuroblast divides to produce a smaller anterior daughter that undergoes programmed cell death and a larger posterior daughter, the HSN/PHB precursor (Fig. 1D). Each precursor subsequently divides to produce an HSN, a serotonergic motoneuron that stimulates egg laying by the hermaphrodite, and a PHB, a chemosensory neuron. The primary goal of our screen was to identify mutations that transform one neuroblast daughter into the other, and thereby alter the number of HSN and PHB neurons produced.

We screened for mutants with abnormal numbers of PHBs using the reporter gmIs12, which expresses GFP in two neurons in the tail: PHA and PHB(Hawkins et al., 2005; Troemel et al., 1995). Three maternal-effect recessive alleles of pig-1 were isolated in these screens: gm280, gm300 and gm301. A deletion allele, gm344, was subsequently isolated using a reverse-genetic approach. These mutants are relatively healthy but have a low penetrance of extra gmIs12- positive cells in the tail that we determined were extra PHB neurons by using PHA- or PHB-specific GFP reporters(Fig. 1G,H; Fig. 2A,B). If the extra PHB phenotype of pig-1 mutants resulted from the production of an extra HSN/PHB precursor, pig-1 mutants should also produce a low frequency of extra HSNs. Staining pig-1 mutants with an anti-serotonin antibody verified this prediction (Fig. 1E,F; Fig. 2A).

Could pig-1 act prior to the birth of the HSN/PHB neuroblast to regulate the production of HSNs and PHBs? The transcription factor hlh-14, for example, specifies the fate of the progenitor that generates the HSN/PHB neuroblast and its sister the PVQ neuroblast(Fig. 1B). In hlh-14mutants, the HSN, PHB and PVQ neurons are lost(Frank et al., 2003). We reasoned that pig-1 mutant lineages would produce extra PVQs if the progenitor that produces the HSN/PHB and PVQ neuroblasts were duplicated, or fewer PVQs if the PVQ neuroblast were transformed into its sister, the HSN/PHB neuroblast. Using the PVQ-specific GFP reporter kyIs39[Psra-6::gfp] (Troemel et al.,1995), we scored the number of PVQs in pig-1 mutants and found no deviation from wild type [pig-1(gm280), n=138; pig-1(gm301), n=118]. These observations suggest that the extra HSNs and PHBs in pig-1 mutants arise because of defects in the HSN/PHB neuroblast or its descendants and not because of earlier lineage defects. The data are consistent with a model where pig-1 mutations transform the anterior daughter of the HSN/PHB neuroblast, which is normally destined to die, into a second HSN/PHB precursor.

To address whether pig-1 mutations perturb other lineages, we scored the number of cells produced using cell-specific GFP reporters and identified six lineages that produce extra neurons in pig-1 mutants:the HSN/PHB, I2, M4 and PLM/ALN precursors divide during embryogenesis, and the Q.a and Q.p precursors divide during the first larval stage(Fig. 1, Fig. 2A, Fig. 3). Many other lineages appear to be unaffected (Fig. 1C). The only feature shared by the affected lineages is the presence of an apoptotic cell. Our observations are consistent with a model where mutations in pig-1 transform these cells into their sisters.

The core programmed cell death machinery in C. elegans consists of four conserved proteins: EGL-1, CED-9, CED-4 and CED-3 (for reviews, see Conradt and Xue, 2005; Kaufmann and Hengartner,2001). In wild-type animals, the BCL2-like protein CED-9 sequesters the Apaf1-like protein CED-4 to mitochondria, preventing it from interacting with the caspase CED-3. Somatic cells that undergo programmed cell death express the BH3-domain protein EGL-1. EGL-1 binds to CED-9, resulting in the release of CED-4 from mitochondria and CED-3 activation.

Two observations suggested that pig-1 could be a component of the programmed cell death pathway. First, the low frequency of extra HSNs and PHBs observed in pig-1 mutants closely approximates that observed in ced-3 and ced-4 mutants(Guenther and Garriga, 1996). Second, all of the lineages affected by pig-1 mutations produce apoptotic cells. We hypothesized that if pig-1 acted in a linear pathway with ced-3, animals with mutations in both pig-1 and ced-3 would have the same phenotype as the single mutants. Alternatively, if pig-1 and ced-3 acted in parallel, then the double mutant would have a stronger phenotype than the individual single mutants. We found that pig-1 ced-3 double mutants had a dramatically stronger phenotype than either single mutant, resulting in nearly 100% of lineages producing extra HSNs and PHBs(Fig. 2A,B). Synergistic interactions between mutations in pig-1 and ced-3 were also observed in three other lineages (Q.p, I2 and PLM/ALN) that we analyzed(Fig. 3A,C,D). pig-1mutations also interacted synergistically with a loss-of-function mutation in egl-1 in the Q.p and PLM lineages(Fig. 3A,D). Together, these data indicate that pig-1 acts at least partly in parallel to the cell death pathway. Our favored interpretation of these results is that mutations in pig-1 almost always transform the apoptotic daughter of the neuroblast into its sister, but that this transformation is usually masked by apoptosis. When this mask is removed by a mutation in a cell death gene like ced-3, the underlying transformation is revealed.

We postulated that pig-1 mutants have extra neurons because the cells normally fated to die in affected lineages occasionally survive and are transformed into their sister cells. To test this hypothesis, we followed the divisions of the left Q.p neuroblast, which divides to produce a posterior cell that dies (Q.pp) and the precursor to the PVM and SDQ neurons(Fig. 1B). In four out of seven Q.p divisions that we followed in pig-1(gm344) larvae, both neuroblast daughters survived and divided to produce two extra neurons. Because we followed the divisions in zdIs5 [mec-4::gfp];pig-1 mutants, we determined whether the lineages that we followed produced extra PVM neurons. There was a perfect correlation between the ability of Q.pp to survive and divide and the presence of extra PVMs. In the four divisions where Q.pp survived and divided, the lineage produced an extra PVM. In the three divisions where Q.pp died, the lineage produced a single PVM. We also noted that the sizes of the pig-1 Q.p daughter nuclei were similar, which is different from wild type where the Q.pp nucleus is much smaller than the Q.pa nucleus.

To verify that the ectopic divisions of the surviving cells in pig-1 mutants could not be explained merely by defective apoptosis,we followed the divisions of six Q.p neuroblasts in zdIs5; ced-3mutants. In all of these lineages, Q.pp survived but did not divide, and none produced an extra AVM or PVM. As in wild type, the ced-3 Q.p division was highly asymmetric, producing a large Q.pa nucleus and a small Q.pp nucleus. These observations are consistent with previous analyses of the Ced-3 phenotype in other lineages (Ellis and Horvitz, 1986; Guenther and Garriga, 1996). Together, they indicate that the apoptotic daughters of neuroblasts have different fates than their sisters, independent of their ability to survive. These data support the model that the extra neurons observed in pig-1 mutants were generated by the inappropriate survival and transformation of cells that normally die into their sisters.

The similar sizes of the Q.p daughter nuclei in pig-1 mutants suggested that pig-1 regulates the asymmetric division of neuroblasts into daughters of unequal size. To confirm a role for pig-1 in regulating cell size, we quantified daughter cell size for both the HSN/PHB and Q.p neuroblast divisions. In the HSN/PHB neuroblast division, we used the hlh-14::gfp transgene gmIs20 to measure the size of the neuroblast daughter nuclei (Frank et al.,2003). The HSN/PHB neuroblast normally divides to produce a smaller anterior daughter, which is destined to die, and a larger posterior HSN/PHB precursor. In wild-type embryos, the diameter of the nucleus of the anterior daughter was approximately three-quarters that of the nucleus of its sister, whereas, in pig-1 mutants, the daughter nuclei were more equal in size (Fig. 4). The size differences between pig-1 and wild-type HSN/PHB neuroblast daughter nuclei cannot be explained merely by inappropriate survival of the anterior daughter, as a mutation in ced-3 has no phenotype in this assay (Fig. 4E). We also quantified the size of the Q.p neuroblast daughters using the cytoplasmic reporter ayIs9 (Pegl-17::gfp)(Branda and Stern, 2000). Although the anterior daughter of Q.p (a neural precursor) is normally larger than its sister (which undergoes apoptosis), the daughters are more symmetric in size in pig-1 mutants (Fig. 4F). Similar effects on daughter cell size were also observed in the Q.a division (data not shown). Together, these data indicate that pig-1 controls daughter cell size in asymmetric neuroblast divisions.

We mapped pig-1 to a 500-kbp interval on LGIV containing approximately 100 genes. Three observations indicate that the open reading frame (ORF) W03G1.6 from this region is pig-1. First, we inactivated each ORF in this region using RNA-mediated interference (RNAi)(Kamath et al., 2001; Timmons and Fire, 1998). Only RNAi-mediated inactivation of W03G1.6 in gmIs12; ced-3(n717)animals induced a PHB duplication phenotype(Fig. 2A). Second, we sequenced the gene in three pig-1 mutant strains and identified a single missense mutation for each allele (see below). Third, we generated a deletion allele of W03G1.6 and found that the mutant displayed Pig-1 phenotypes.

Sequence analysis of pig-1 cDNAs indicates the gene transcribes a single SL1 trans-spliced mRNA that encodes a conserved serine/threonine kinase that is homologous to vertebrate MELK (Fig. 5). Like MELK family members, PIG-1 has an N-terminal kinase domain, a middle region of reduced homology, and a kinase-associated domain at its C terminus (Fig. 5B). Protein sequence comparisons indicate that PIG-1 and its vertebrate homologs form a separate subfamily of serine/threonine kinases distinct from the Kin1p,PAR-1 and SAD-1 families, indicating that PIG-1 is the C. elegansortholog of MELK (Fig. 5C). We found that gm300 and gm301 alleles are missense mutations in the kinase domain that change conserved glycines to charged amino acids(Fig. 5B). The weaker allele gm280 is a missense mutation in the kinase-associated domain that changes a conserved glycine to arginine(Fig. 5B). This analysis indicates that both the kinase and kinase-associated domains are important for pig-1 activity.

The pig-1 deletion allele gm344 is a 524 bp deletion that removes the pig-1 promoter, all of the first exon, and part of the second exon of the pig-1 coding sequence. gm344 appears to be a molecular null based on RT-PCR and western blot assays (data not shown). Like the other pig-1 mutants, gm344 mutants are viable and the strength of Pig-1 mutant phenotypes is identical to what we observed in pig-1(gm301) mutants (Fig. 3A,D). These comparisons suggest that the alleles gm300,gm301 and gm344 eliminate PIG-1 function, and that the allele gm280 reduces it.

Pig-1 phenotypes are maternally rescued, consistent with the presence of pig-1 transcripts in the hermaphrodite germ line (NEXT database, Y. Kohara, personal communication). To characterize zygotic pig-1expression, we examined transgenic worms expressing a transcriptional reporter containing 850 bp upstream of the pig-1 start codon fused to GFP coding sequences. This construct contained all of the noncoding sequence between pig-1 and the next predicted upstream gene, and expressed GFP ubiquitously in early embryos. The expression became progressively more restricted in older embryos and young larvae, and was not observed in adults(data not shown). In larvae, we observed expression in dividing cells: ventral nerve cord neuroblasts, vulval precursors, dividing hypodermal seam cells, and the Q neuroblasts and their descendants(Fig. 6A-F; data not shown). To investigate the subcellular distribution of the PIG-1 protein, we examined transgenic worms expressing a full-length pig-1::gfp translational fusion. This construct contains the same promoter used in the transcriptional reporter fused to a full-length pig-1 cDNA tagged with GFP. This transgene rescued Pig-1 phenotypes in the PLM/ALN and Q.p lineages (data not shown). In all cells, the fusion protein was localized to the cytoplasm and excluded from nuclei (Fig. 6G,H).

To determine whether pig-1 acts cell autonomously in neuroblasts to regulate their divisions, we focused on the Q.p neuroblast, which produces the A/PVM and SDQ neurons (Fig. 1B). The pig-1 promoter drives expression in the Q neuroblast and its descendents (Fig. 6A-F), consistent with a cell autonomous role for pig-1.To test this hypothesis, we expressed a pig-1 cDNA from the mab-5 promoter and asked whether its expression would rescue the Q.p defects of pig-1 mutants. mab-5 encodes a homeobox transcription factor expressed in a bilaterally symmetric group of posterior cells with one exception: it is expressed in the left Q lineage (which produces PVM), but not the right Q lineage (which produces AVM)(Cowing and Kenyon, 1992; Salser and Kenyon, 1992). We reasoned that if pig-1 acts in the Q lineage, pig-1expression from the mab-5 promoter should rescue the extra PVM but not the extra AVM phenotype of pig-1 mutants. Consistent with the hypothesis that PIG-1 acts in Q.p to promote its asymmetric division, the Pmab-5::pig-1 transgene rescued the Pig-1 PVM but not the AVM phenotype (Fig. 7).

The C. elegans HAM-1 protein regulates the asymmetric divisions of many embryonic neuroblasts, including the HSN/PHB neuroblast(Frank et al., 2005; Guenther and Garriga, 1996). Like mutations in pig-1, mutations in ham-1 alter both HSN/PHB neuroblast daughter cell size asymmetry and cell fate. In ham-1(gm279) null mutants, the relative sizes of the neuroblast daughters are reversed, and the anterior daughter is larger than its sister. The anterior daughter is transformed into its sister, resulting in the production of extra HSNs and PHBs. ham-1 HSNs and PHBs also often fail to differentiate properly, and occasionally mutant lineages fail to produce HSNs or PHBs that express cell-type specific markers; this latter phenotype may be a consequence of the failure of the posterior daughter of the HSN/PHB neuroblast to divide or differentiate correctly because of its small size. ham-1 encodes a novel protein that is asymmetrically distributed to the posterior cortex of the dividing neuroblast and is inherited specifically by the precursor. HAM-1 is hypothesized to control the asymmetric segregation of developmental potential into the neuroblast daughters, in part by regulating the volume of neuroblast cytoplasm distributed to each daughter cell (Frank et al., 2005).

The similarity of pig-1 and ham-1 phenotypes led us to determine whether the two genes act together to regulate neuroblast divisions. To address this possibility, we examined the output of the HSN/PHB neuroblast lineage in pig-1 ham-1 double mutants. While pig-1 mutants have a low penetrance of extra PHBs, ham-1 mutants have a higher penetrance of extra PHBs, as well as missing PHBs(Fig. 2C). We were surprised to discover that although ham-1 mutants have a stronger phenotype, pig-1 mutations were epistatic to ham-1 mutations: pig-1 ham-1 double mutants had a low penetrance of extra PHBs and were not missing PHBs. This relationship was observed with multiple pig-1 and ham-1 alleles (Fig. 2C; data not shown). We obtained similar results when we scored the number of HSNs (Fig. 2C)and the number of PLMs (Fig. 3D), indicating that the genetic interaction between the two genes is not PHB or lineage specific. Mutations in pig-1 are also epistatic to ham-1 in a ced-3 background: whereas ham-1(n1438)ced-3 double mutants have extra PHBs about 60% of the time, pig-1 ced-3 and pig-1 ham-1(n1438) ced-3 mutants almost always have extra cells (Fig. 2D).

pig-1 mutations also suppressed Ham-1 differentiation phenotypes. The PHB phasmid neurons in ham-1 mutants frequently fail to fill with the dye DiI because of defective sensilla development(Guenther and Garriga, 1996). pig-1 mutants have no dye-filling defects, and mutations in pig-1 suppress the dye-filling defects of ham-1 mutants(data not shown). pig-1 mutations also suppress the HSN migration defects observed in ham-1 mutants (data not shown).

Mutations in both pig-1 and ham-1 perturb the asymmetric division of the HSN/PHB neuroblast into daughters of unequal size. While strong ham-1 mutations reverse the relative sizes of the daughters(Frank et al., 2005),mutations in pig-1 cause the neuroblast to divide more symmetrically. When we compared the sizes of the HSN/PHB neuroblast daughters in pig-1 ham-1 double mutants, we found that they were similar to pig-1mutants (Fig. 4). In summary, pig-1 ham-1 double mutants seem to phenocopy pig-1 single mutants in assays that examine the cell lineage, cell size, and the differentiation of neuroblast descendants in the HSN/PHB neuroblast lineage.

Many genes that regulate asymmetric cell divisions control cell polarity. For example, the PAR proteins are required in the C. elegans zygote for the asymmetric positioning of the mitotic spindle and the asymmetric segregation of cell fate determinants to daughter cells(Kemphues et al., 1988; Rose and Kemphues, 1998). pig-1 mutants share phenotypic features with these cell polarity mutants, and PIG-1 is related molecularly to PAR-1. However, although pig-1 and par-1 phenotypes have similar features, PAR-1 localizes asymmetrically in the mother cell, whereas PIG-1 and its vertebrate orthologs do not (Chartrain et al.,2005; Heyer et al.,1999). The symmetric distribution of PIG-1 suggests that it may play a permissive role in the regulation of neuroblast polarity. In this sense, PIG-1 may resemble the symmetrically distributed 14-3-3 protein PAR-5,which regulates both spindle positioning and cell fate in the C. elegans zygote (Morton et al.,2002).

We propose that PIG-1 controls the asymmetric positioning of the mitotic spindle and asymmetric segregation of neural fate determinants to daughter cells. In this model, loss of pig-1 activity disrupts neuroblast polarity, resulting in symmetric positioning of the cleavage plane and mislocalization of cell fate determinants. The neuroblast therefore divides symmetrically, and cell fate determinants are symmetrically distributed to daughter cells. The segregation of these determinants into the normally apoptotic daughter of the neuroblast transforms it into its sister.

HAM-1 is also thought to regulate asymmetric neuroblast divisions by regulating neuroblast polarity, and mutations in ham-1 also disrupt cell size asymmetry and cell fate in the HSN/PHB neuroblast division. HAM-1 has previously been proposed to regulate neuroblast daughter cell fate in part by controlling cell size (Frank et al.,2005). One important question that arises from our analysis is whether HAM-1 and PIG-1 could function solely to regulate the position of the neuroblast cleavage plane, and the ensuing cell fate transformation could result from altered daughter cell size. This model predicts that determinants are symmetrically distributed in the neuroblast and that cell fate is determined by the concentration of determinants in a neuroblast daughter. Alternatively, PIG-1 and HAM-1 could regulate the position of the neuroblast cleavage plane and the asymmetric segregation of cell fate determinants to daughter cells independently. In this model, determinants are asymmetrically distributed in the neuroblast, and their segregation into daughter cells is mostly independent of cleavage plane position.

Two observations lead us to favor a modified version of the second model,in which the size of daughter cells does not directly specify their fate, but might influence their survival (Fig. 8). First, cell size alone cannot control the apoptotic fate of the daughter cells. Although the posterior daughter of the HSN/PHB neuroblast is smaller in ham-1 mutants, it is still the larger anterior daughter that dies (Frank et al.,2005). Second, if cleavage plane position were to regulate quantitative differences in the segregation of cell fate determinants to the anterior daughter of the HSN/PHB neuroblast, the frequency at which this cell is transformed into its sister in pig-1 and ham-1 mutants should be proportional to the deviation of the cleavage plane from its normal position. This does not seem to be the case, as analyses of pig-1 ced-3 and ham-1 ced-3 double mutants suggest that pig-1and ham-1 mutations always transform the anterior daughter of the HSN/PHB neuroblast into its sister, independent of daughter cell size. However, the penetrance of extra cells in pig-1, ham-1 and pig-1 ham-1 mutants does appear to be proportional to the deviation of the cleavage plane from its normal position, so cell size could affect the frequency at which the anterior daughter of the neuroblast escapes programmed cell death. Thus, increasing the size of the anterior daughter of the HSN/PHB neuroblast may increase the frequency at which this cell escapes programmed cell death.

The epistatic interactions between the pig-1 and ham-1mutations are intriguing. If daughter cell size influences the penetrance of extra and missing neurons, then pig-1's epistasis to ham-1at the level of daughter size is sufficient to explain why pig-1mutations suppress the extra neuron and missing neuron phenotypes of ham-1 mutants (Fig. 8). pig-1 might be epistatic to ham-1 because HAM-1 inhibits PIG-1-dependent positioning of the mitotic spindle. HAM-1 may restrict PIG-1 activity to the anterior of the HSN/PHB neuroblast, and the mispositioning of the cleavage plane in ham-1 mutants could be a consequence of ectopic PIG-1 activity in the posterior of the neuroblast. Alternatively, PIG-1 might be a general factor required for spindle positioning. PIG-1, for example, could regulate, or be regulated by, both HAM-1 and an activity that has been postulated to antagonize HAM-1(Frank et al., 2005).

The vertebrate orthologs of PIG-1 have been implicated in regulating cell proliferation, although the nature of this regulation has been the subject of debate. MELK kinase activity exhibits cell cycle dependence, with maximal activity during mitosis (Blot et al.,2002; Davezac et al.,2002). MELK has been shown to bind and phosphorylate CDC25B, a phosphatase that promotes G2/M progression by removing inhibitory phosphate groups from the cyclin-dependent kinase CDC2(Davezac et al., 2002). MELK phosphorylation was proposed to inhibit CDC25B, as several groups have shown that MELK overexpression results in a block at the G2/M boundary(Davezac et al., 2002; Gray et al., 2005; Vulsteke et al., 2004). The C. elegans genome encodes three CDC25 homologs. To determine whether Pig-1 phenotypes require CDC-25 activity, we used RNAi(Fraser et al., 2000) to inactivate these genes in pig-1 and pig-1 ced-3 animals. We examined the output of the HSN/PHB and PLM/ALN lineages, and found that RNAi of cdc-25.2 suppressed the extra PHB and PLM phenotypes of these mutants (A. Singhvi and G.G., unpublished). While these observations are consistent with a role for PIG-1 as a negative regulator of CDC-25.2, it is also possible that the cdc-25.2 RNAi effects reflect a nonspecific effect on cell division.

MELK has also been proposed to positively regulate the cell cycle. Overexpression of MELK in mouse neurosphere cultures results in increased proliferation, whereas MELK knockdown by siRNA has the opposite effect(Nakano et al., 2005). MELK has also been shown to positively regulate the growth rate of cultured human tumor cell lines, both in vitro and after injection into a murine host(Gray et al., 2005). Nakano et al. (Nakano et al., 2005)proposed that MELK positively regulates the cell cycle through its association with the zinc-finger protein ZPR9 and the proto-oncogene B-myb. MELK binds and phosphorylates ZPR9 (Seong et al.,2002), which in turn binds B-myb and increases its transcriptional activity (Seong et al., 2003). Although an ortholog of ZPR9 exists in C. elegans (ORF C16A3.4), RNAi of this gene does not generate a Pig-1-like phenotype (S.C. and G.G.,unpublished).

Altered cell cycle progression in the neuroblast could lead to the phenotypes observed in pig-1 mutants, as cell cycle regulators are known to control asymmetric divisions. In Drosophila, Cdc2/B type cyclins regulate asymmetric divisions of neuroblasts in the central nervous system. Loss of zygotic cdc2 activity results in both defective spindle positioning and sister-sister cell fate transformations(Tio et al., 2001). In C. elegans, the anaphase-promoting complex (APC), the APC activator cdc20 and the APC effector separin are required during the first cell division for proper positioning of the cleavage plane and the asymmetric distribution of cell fate determinants(Rappleye et al., 2002). PIG-1 could regulate the timing of cell cycle events in the neuroblast, which in turn, could indirectly regulate neuroblast polarity. In conclusion, our analysis of PIG-1 suggests that MELK could regulate polarity and daughter cell fate in asymmetrically dividing cell lineages such as stem cells.

We thank A. Singhvi and N. Hawkins for comments on the manuscript, A. Singhvi for sharing unpublished observations, and members of the Garriga,Meyer and Dernburg labs for helpful discussions. We also thank Y. Kohara for cDNAs and C. Li, Y. Lie, C. Kenyon, O. Hobert and B. Meyer for bacterial and C. elegans strains. Some of the C. elegans strains used in this study were obtained from the C. elegans Genetics Center, which is funded by the NIH National Center for Research Resources. This work was supported by an Ernest Brown Babcock Scholarship to S.C., a National Science Foundation predoctoral fellowship to C.A.F., and by National Institutes of Health grant NS42213 to G.G.