TLP-1 is an asymmetric cell fate determinant that responds to Wnt signals and controls male tail tip morphogenesis in C. elegans

Two fundamental questions in developmental biology are understanding how morphogenesis is controlled and how changes in morphogenesis during evolution lead to the elaboration of different forms. For morphogenesis to occur properly during development, many cellular processes must be correctly controlled; among these are the establishment of different cell fates through asymmetric cell divisions, cell polarity and cell fusion. Changes in any of these processes can lead to changes in morphology that over time could lead to the genesis of different forms.

The polarities of specific cells in the C. elegans tail are controlled by Wnt signaling. The conserved Wnt signaling pathway is one of the major signaling pathways controlling animal development (reviewed by Cadigan and Nusse, 1997; Wodarz and Nusse, 1998). Through genetic and molecular studies, the identification of the components that function in a canonical signaling pathway, as well as the probable order of their action, has been determined. In short, Wnt signals are short-range secreted signals that act through Frizzled (Fz) receptors at the cell surface. Through the action of Dishevelled inside the cell, β-catenin is stabilized, causing it to accumulate in the cytoplasm and the nucleus, where it interacts with Tcf factors to activate target genes.

Mutations in the Wnt gene lin-44 cause the polarities of certain cells in the C. elegans tail – the B, TL and TR cells – to be reversed (Herman and Horvitz, 1994; Herman et al., 1995). LIN-44 is expressed by the epidermal cells at the tip of the developing tail, which are posterior to the cells whose polarities are affected by lin-44 mutations. LIN-44 function is required within the tail tip cells, suggesting that LIN-44 is secreted by these cells and affects the polarity of more anterior asymmetric cell divisions (Herman et al., 1995). Mutations in the Frizzled-related gene lin-17 cause a loss of polarity in the same cells in which lin-44 mutations cause a reversal of polarity (Sternberg and Horvitz, 1988; Sawa et al., 1996), suggesting that LIN-17 serves as the receptor for LIN-44. The Wnt pathway that controls the polarity of the TL and TR cells shares some components with the canonical Wnt pathway, such as the requirement for the Tcf homolog, POP-1, but may not require other components such as β-catenin (Herman, 2001), suggesting that it is a novel Wnt pathway. The defects in cell polarity observed in lin-44 and lin-17 mutants lead to defects in morphogenesis as evidenced by the formation of abnormal male tail structures in both mutants (Sternberg and Horvitz, 1988; Herman and Horvitz, 1994).

Another cellular function that often affects morphogenesis is cell fusion. In vertebrates, the formation of multinucleated cells, or syncytia, is involved in the formation of skeletal muscles, bones and, in mammals, the placenta. In C. elegans, one third of all the somatic nuclei in the adult animal are part of one of a few syncytia that are formed in reproducible spatial and temporal patterns (Shemer and Podbilewicz, 2000). For example, a detailed analysis of male tail development has revealed that a specific temporal order of cell fusions and cell shape changes lead to the formation of the male tail. C. elegans male tail tip morphogenesis begins after the bursal ray cell divisions are complete (Nguyen et al., 1999). The tail tip cells then lose adhesion with the cuticle and recede anteriorly. This tail tip retraction is accompanied by changes in cell shape, position and cell contacts. The retraction of the tail tip causes the adult male tail to be blunt ended, or ‘peloderan’, where the bursal fan extends posteriorly and envelopes the tail tip. By contrast, these changes in cell shape, position and cell contacts do not occur in hermaphrodites, and the pointed larval shape is retained in the adult. In other nematode species, such as Oscheius myriophila, the male tail tip cells also fail to retract during morphogenesis of the tail, causing the adult male tail to be pointed, or ‘leptoderan’, and protrude beyond the posterior edge of the adult fan (Nguyen et al., 1999).

To further understand the roles cell polarity and cell fusions play in morphogenesis, we have isolated mutations identifying a new gene, tlp-1 (T cell lineage defective and leptoderan tail). We show that tlp-1 is required for the specification of asymmetric cell fates in the descendants of the T cell, as well as for cell fusions and cell shape changes of the tail tip hypodermal cells that lead to drastic changes in male tail morphology. tlp-1 encodes a nuclear-localized C2H2 zinc-finger protein and is asymmetrically expressed during the divisions of the T cell lineage. Expression studies and cell lineage analysis in double mutants suggest that tlp-1 is expressed in response to lin-44 and lin-17 genes in the control of T cell polarity. Our results suggest that tlp-1 encodes a transcription factor required for morphogenesis of the C. elegans tail by regulating asymmetric cell fates in response to Wnt signaling.

Nematodes were cultured by standard techniques (Sulston and Hodgkin, 1988). The following mutations were used: LGI, lin-44(n1792), lin-17(n671), unc-29(e1072); LGIV, bli-6(sc16), unc-24(e138), dpy-20(e1282), unc-26(e205), tlp-1(bx85, mh17, ny14), egl-23(n601), tra-3(e1107), dpy-4 (e1166); and LGV, him-5(e1490). Rearrangements used are LGIV, V, nT1(IV, V), nDf27, sDf21, sDf23. Male defects were studied using strains containing the him-5(e1490) mutation, which increases the frequency of spontaneous male progeny (Hodgkin et al., 1979).

tlp-1 alleles were isolated in separate screens for phasmid dye-filling defective (Pdy) mutants (mh17), male abnormal (Mab) mutants (bx85) and males with leptoderan-like (Lep) tails (ny14) using ethyl methanesulfonate (EMS), diepoxybutane (DEB) and trimethypsoralen followed by u.v. irradiation as a mutagens, respectively. All three alleles failed to complement each other for the Pdy and Lep defects, indicating that all were alleles of tlp-1. tlp-1 was mapped to LGIV (not shown, Fig. 4). Three-factor crosses further defined the map position of tlp-1 to the region between unc-26 and egl-23 (see Fig. 4A). Complementation tests revealed that tlp-1 is not uncovered by sDf21 and sDf23, but may be partially uncovered by nDf27 (Table 1).

Living animals were observed using Nomarski optics; cell nomenclature and cell lineage analysis were as previously described (Sulston and Horvitz, 1977). Fates of the T cell descendants were determined by nuclear morphologies according to Herman and Horvitz (Herman and Horvitz, 1994). Phasmid dye-filling as an indicator of normal T cell polarity was scored as previously described (Herman and Horvitz, 1994). jam-1::gfp, a marker of adherens junctions, was used to visualize cell boundaries (Mohler et al., 1998) and to monitor cell fusions (accompanied by adherens junction degradation).

Cosmids, plasmids and gel-purified PCR fragments derived from cosmid T23G4 or F52D4 were injected (10-60 ng/μl) using either pRF4 [a plasmid containing the semidominant rol-6(su1006) allele], pTG96 (a sur-5::gfp fusion construct kindly provided by Dr Min Han, University of Colorado) or pEL166 (a str-1::gfp fusion construct kindly provided by Dr Erik Lundquist, University of Kansas) (Troemel et al., 1997) (40-100 ng/μl) as markers into the mitotic germline of tlp-1 hermaphrodites (Mello and Fire, 1995). Yeast genomic DNA isolated from strains containing the desired YAC DNAs were isolated (Qiagen^R) and used for injection (40-100 ng/μl). A 10,110 bp PCR product amplified from F52D4 was digested with ScaI and cloned into pBluescript to obtain pXZS5, which contained a 7519 bp fragment that rescued the T cell defect of tlp-1 mutants. We scored each line for either phasmid dye-filling or Lep defects and sometimes both.

The sequences of mutant alleles mh17 and bx85 were determined from PCR-amplified genomic DNA fragments using primers that amplified candidate genes in the tlp-1 region. DNA lesions in T23G4.1 were identified from at least two independent PCR-amplified mh17 and bx85 genomic DNA fragments. ny14 is a deletion allele for which both approximate end points were identified by PCR (see Fig. 4A). Briefly, PCR products from ny14 were obtained from regions covered by YAC segment Y45F10D and cosmid H08M01 on the right, but not from any of the tested regions covered by cosmid C47A4. Additionally, segments including regions of ORFs T23G4.1, T23G4.2 and T23G4.3 could not be amplified from ny14.

The nucleotide sequence of genomic DNA through the tlp-1 locus was determined by the C. elegans Genome Consortium and primers used were based on this sequence. The 1065 bp tlp-1 coding region was amplified by PCR from cDNAs synthesized using Promega^TM Reverse Transcription System. The trans-spliced leader 1 (SL1) primer was used to obtain a 292 bp 5′ fragment. An oligo(dT) primer was used to obtain 273 bp and 527 bp fragments that included the full 3′-UTR. All PCR fragments were cloned into pTOPOXL (Invitrogen^TM) and the sequence of both strands of the cDNA clones were determined (accession AY046083).

A tlp-1::gfp fusion that placed five additional residues, the GFP-coding sequence and the unc-54 3′ UTR at the end of the TLP-1 coding sequence was constructed by PCR. Briefly, a 7973 bp PCR fragment amplified from F52D4 was cloned into pTOPOXL to obtain pXZS4, which was digested with KpnI and EcoRV and ligated into pPD95.75 (from A. Fire, Carnegie Institute, Baltimore, MD) digested with KpnI and SmaI to obtain pZXE7. The tlp-1::gfp construct begins 5767 bp upstream of exon 1 and places the gfp tag at the end of the tlp-1-coding region. The cosmid C45D10, containing unc-29(+), was used as a co-injectable marker (Lackner et al., 1994). mhEx50 is a transgenic array generated in the unc-29 background containing the tlp-1::gfp fusion and C45D10 that rescues the phasmid dye-filling defect and T cell defects. Similar results were obtained with two other independently generated arrays. JAM-1::GFP expression was observed using a strain of genotype unc-30(e191) tlp-1(bx85) jcIs1[unc-29(+); jam-1::gfp rol-6(su1006)]; him5(e1490).

To identify sequences related to TLP-1, available data banks were searched using BLASTP and PSI-BLAST using fragments of TLP-1 and Drosophila NocA that included the SPLALLA or C2H2 zinc-finger motifs. Additionally, Entrez and WormPD were queried with keywords ‘Sp1’ and ‘C2H2’ to identify additional candidate proteins. Preliminary neighbor-joining analyses were used to identify sequences most similar to TLP-1. We limited our in-depth phylogenetic analysis to 45 of these predicted protein sequences (listed in Fig. 4D), all of which have one or more predicted C2H2 zinc-finger motifs and 15 of which have the highly conserved SPLALLA motif. Although we assumed that the database sequences were correct, the human Sp3 sequence is clearly truncated at the N terminus, and C. elegans F45H11.1 appears to be truncated at the C terminus in the middle of the second zinc finger.

Multiple alignments of predicted protein sequences were performed using ClustalX 1.81 (Thompson et al., 1997) and MacClade 4.03 (Maddison and Maddison, 2001). Initially, the SPLALLA motif (alignment positions 154-173), Btd box (Supp et al., 1996; positions 979-985) and putative C2H2 zinc-finger motifs (positions 1020-1154) were aligned by hand in MacClade, and the sequences flanking these motifs were aligned as separate regions in ClustalX (accession ALIGN_000176 – can be retrieved from the EMBLALIGN database at http://srs.ebi).

PAUP* (Swofford, 2001) was used to perform weighted-parsimony, in which amino acid changes were weighted by the minimum number of nonsynonymous nucleotide changes required, and neighbor-joining distance analyses to determine the possible relationships among the 45 predicted C2H2 proteins, with particular emphasis on the proteins most closely related to TLP-1. Phylogenetic analyses were restricted to the 123 least ambiguously aligned characters (positions) of 1409 total; alignment between divergent sequences was mostly arbitrary for the remaining characters. Of these 123 characters, 113 were informative for the cladistic analyses. Phylogenetic analyses on alternative alignments or additional characters were consistent with those using only the 123 characters. Trees collected from exhaustive and heuristic searches were compared statistically and strict consensus trees were constructed from trees that were not rejected in these tests. Uncollapsed branches in these consensus trees were deemed to be unambiguously supported. Additional tests of branch stability included bootstrap and jackknife tests and decay indices. Details of these analyses are available from D. F., as are the NEXUS data files.

In wild-type hermaphrodites, the bilaterally symmetric TL and TR cells (collectively referred to as T cells) divide asymmetrically during the L1 stage: T.a generates four hypodermal cells and one neuron whereas T.p generates five neural cells (Fig. 1). In lin-44 mutants, the polarity of the division of the T cell lineage is reversed: T.a generates five neural cells and T.p generates four hypodermal cells and one neuron (Herman and Horvitz, 1994). In lin-17 mutants, the polarity of division of the T cell lineage is lost: both T.a and T.p generate hypodermal cells (Sternberg and Horvitz, 1988). The defects in T cell polarity observed in both lin-17 and lin-44 mutants cause the two neurons of the phasmid, a sensory structure in the tail, to fail to fill with fluorescent dyes (Herman et al., 1995; Sawa et al., 1996).

Screens for additional mutations that resulted in phasmid dye-filling (Pdy) defects resulted in the identification of the mh17 allele of tlp-1. In addition, separate screens for mutations that caused male tail defects yielded two additional tlp-1 alleles, bx85 and ny14, which also display phasmid dye-filling defects (Table 1). Both the Pdy and Lep defects are recessive. The penetrances of the Pdy defect of mh17 and bx85 mutations are similar but ny14 is more penetrant, although this may not be due to the removal of tlp-1 function alone.

Analysis of the T cell lineages of each tlp-1 mutant showed that cell fates of the posterior T cell daughter, T.p, and the posterior daughter of the T.ap cell, T.app, were variably defective (Fig. 1). We observed four basic patterns of defective cell lineages from which we conclude that tlp-1 mutations cause a loss of asymmetry in the divisions of the T.p and T.ap cells, most often resulting in the loss of cell divisions and neural cell fates (Fig. 1B).

Tail tip morphogenesis in C. elegans males involves changes in cell adhesion of the tail tip cells to the cuticle followed by an anterior movement of the tip cells which causes the adult male tail to be blunt ended, or ‘peloderan’.

To further understand male tail morphogenesis and tail tip cell retraction, we have isolated mutations that result in a leptoderan (Lep) male tail. The leptoderan tails of bx85 males, as well as those of ny14 and mh17, are prominent and variable in length (Fig. 2). The Lep defect is completely penetrant in all three tlp-1 mutant alleles (Table 1). Interestingly, the tail tips in these mutants seem to differentiate fan cuticle, as excess cuticle surrounds the pointy tail tip, unlike lep-1 mutants (Nguyen et al., 1999). Regions of the fan are often narrow and misshapen. In addition, rays 8 and 9 (60% of sides, n=58) or 7-9 (17% of sides) were often missing, consistent with the T cell lineage defects we observed. In a few animals we observed fusions of rays 3 and 4 or 4 and 5 and an ectopic T cell-derived ray. Gross morphologies of the tail tips of tlp-1 hermaphrodites are unaffected, although there are often defects at the cellular level (see below).

As cell fusion is a prominent event in male tail tip morphogenesis, we wanted to know if the Lep defect of tlp-1 mutants might be due to failure in cell fusions. The distribution of adherens junctions was observed in live worms using JAM-1::GFP (Fig. 3), a fusion of the green fluorescent protein (gfp) (Chalfie et al., 1994) to the junction associated molecule, jam-1, that localizes to cell boundaries (Mohler et al., 1998). In wild-type males, the tail tip cells fuse before the Rn.p cells fuse to form the ‘set’ (‘tail seam’) syncytium. These fusions correspond to a punctate ‘broken’ pattern of JAM-1::GFP, owing to the degradation of the junction (Fig. 3A). In tlp-1(bx85) mutants a punctate JAM-1::GFP pattern usually appears long after the ‘set’ has formed (Fig. 3B). That is, compared with wild type, fusion of the male tail tip hypodermal cells is heterochronically delayed in tlp-1 mutants.

In wild-type hermaphrodites and larval males, lateral adherens junctions surrounding the posterior epidermal cells are contiguous from anterior to posterior: adherens junctions border the lateral seam (‘se’), surround the phasmids at the posterior end of the seam, continue posteriorly between hyp8 and hyp9 (on the ventral side) and hyp11 (on the dorsal side), finally extending along the pointy hyp10 cell (Fig. 3C,E). In tlp-1 mutants, the adherens border between hyp8 and hyp11 is often missing, such that a ‘blank space’ appears between the ‘se’ and the tail tip cells (Fig. 3D). Because of abnormal arrangements of nuclei in these tails, we could not discern whether hyp11 or hyp8 were present, but hyp9 and hyp10 nuclei were always present. The lack of adherens junctions could indicate that specific cells (e.g. hyp8 and hyp11) have ectopically fused with the hyp7 body syncytium. If so, such fusions must have occurred very early in development, as these defects can already be seen in L1 and L2 animals.

In other cases, extra adherens junctions are formed in the tail tip (Fig. 3F), indicating either the presence of an extra epidermal cell or extra adherens junctions built by one of the tail tip cells. These presumed extra cells might be generated by the abnormal T cell lineages observed in tlp-1 mutants. These cells are not phasmid socket cells transformed from their normal fate, as characteristic phasmid structures can be seen in some of these animals (Fig. 3F). All of these variations suggest that tlp-1 mutations cause changes in the presence or fates of particular cells in the tails of both sexes, thereby affecting tail tip cell fusions.

The fusion and retraction of the male tail tip cells might be triggered by cell signals emanating from an anterior source or by a developmental program intrinsic to the tail tip cells or both. It is unlikely that a fusion signal emanates from the T cell lineage as other mutations that affect the T cell lineage, such as lin-17, do not cause Lep tails, although lin-44 males have a low penetrance Lep tail defect. Furthermore, we killed the T cells with a laser microbeam (Sulston and White, 1980) in five wild-type males, but did not observe a Lep defect. It is also possible that the abnormal T cell fates physically interfere with such a signal emanating from other cells. This was tested by killing both TL and TR cells in five tlp-1 males; all had Lep tails, suggesting that the abnormal T cell lineages in tlp-1 males do not block an anterior signal for fusion and retraction.

Microinjection of either cosmid F52D4 or T23G4 rescued both the T cell and Lep defects of tlp-1(mh17) and tlp-1(bx85) mutants. Additionally, a 7519 bp fragment derived from F52D4 rescued the phasmid dye-filling defects of tlp-1(mh17) and tlp-1(bx85) mutants, but not the phasmid dye-filling defect of tlp-1(ny14) mutants (Fig. 4A). Interestingly, a rescuing tlp-1::gfp transgene did not rescue the phasmid dye-filling defect of ny14 mutants either, but did restore the presence of phasmid socket cells, suggesting the presence of another gene in the region deleted by ny14 that may be involved in phasmid function but not in asymmetric cell fate determination (Table 1). Furthermore, neither the 7519 bp fragment nor the tlp-1::gfp construct rescued the Lep defect of bx85 males but a 10,110 bp fragment partially rescued the Lep defect of mh17 and bx85 males (Fig. 4A). The tlp-1::gfp construct contained all the sequences 5′ to the coding region present on the 10,110 bp genomic fragment and extends 3421 bp further 5′ than T23G4, suggesting that other sequences, which must lie 3′ to the coding region, are necessary for the expression of tlp-1 required for proper retraction and fusion of the tail tip cells. Alternatively, the GFP moiety may interfere with tail tip-specific activity of TLP-1.

We found DNA lesions in each tlp-1 mutation that affected the coding region of the predicted gene T23G4.1 (Fig. 4B), confirming that we have identified the tlp-1 locus. ny14 is a deletion that extends beyond cosmid C47A4 on the right and predicted gene T23G4.4 on the left (not shown), thus removing all of T23G4.1, the tlp-1-coding region. The mh17 mutation is a point mutation that changed glutamine residue 77 to an ochre stop codon. The bx85 mutation is a 5 bp deletion that causes a frameshift at residue 89 and a termination codon 162 residues later.

Overlapping cDNA clones specific to tlp-1 were identified by RT-PCR from cDNA prepared from mixed-stage wild-type animals and sequenced. The positions of the four exons in the gene structure are the same as those the C. elegans genome project predicted for T23G4.1, indicating that this predicted gene corresponds to tlp-1. The composite cDNA contains the SL1 transpliced leader sequence (Krause and Hirsh, 1987), a 1062 nucleotide open reading frame, a 25 nucleotide 5′ untranslated region, a 459 nucleotide 3′ untranslated region and a poly(A) tail. Developmental northern analysis using the RT-PCR product containing the complete coding region as a probe detected a single 1.8 kb transcript present at all stages (not shown), consistent with the size of the 1664 bp composite cDNA. Together, these data suggest that the composite cDNA represents the full-length tlp-1 message.

Translation of the tlp-1 composite cDNA yielded a predicted protein of 354 amino acid residues containing two serine/threonine-rich regions, a glutamine-rich region and a zinc finger domain of the C2H2 type (Fig. 4C). Another highly conserved, previously undescribed motif of unknown function (SPLALLA) occurs near the N terminus (Fig. 4C). The C2H2-type zinc-finger domains are characterized by sequences that approximate the form (Tyr/Phe)-X-Cys-X_2-4-Cys-X₃-Phe-X₃-Leu-X₂-His-X_3-5-His, where X represents more variable amino acids (Berg and Shi, 1996). TLP-1 bears all the most conserved sequences of the C2H2 motif. However, instead of two to four amino acids between the two Cys residues, the variable part contains eight amino acids. The variable regions are speculated to mediate different interactions with other macromolecules (Berg and Shi, 1996). All C2H2 zinc-finger proteins studied so far are DNA-binding proteins (Iuchi, 2001), suggesting that TLP-1 functions as a transcription factor. Both mh17 and bx85 mutations truncate the protein before the zinc finger, suggesting its function is important.

Phylogenetic analyses show that tlp-1 is a member of a monophyletic Sp1-like family of C2H2-type zinc-finger transcription factor genes (thick lineages in Fig. 4D), and is most closely related to the D. melanogaster nocA (no ocelli) gene (Cheah et al., 1994), the noz1 (nocA-related zinc finger 1) gene from zebrafish (Andreazzoli et al., 2001), and two uncharacterized genes predicted from the D. melanogaster and human genome projects. Because the relationships of these five sequences reasonably reflect the relationships of the organisms that harbor them, this NocA/TLP-1 subfamily is an orthology group.

Predicted proteins of this subfamily uniquely share the SPLALLA motif with the proteins of the monophyletic Sp subfamily (Sp1-Sp5) of vertebrates, constituting the primary evidence of common descent by gene duplication from an ancestral sequence with a SPLALLA motif (lineage 12, Fig. 4D). However, five other Sp-like sequences (Buttonhead, D-Sp1, Y40B1A.4, T22C8.5, and F45H11.1) do not share this SPLALLA motif. Because of poor resolution of relationships among these latter sequences and between these sequences and the vertebrate Sp and NocA/TLP-1 subfamilies, several alternative evolutionary scenarios are possible besides that depicted in Fig. 4D. Specifically, one or all of the sequences without the SPLALLA motif could be most closely related to the vertebrate Sp subfamily. For example, D-Sp1 could be the Drosophila ortholog of the vertebrate Sp subfamily. Although possible, these alternative hypotheses are less parsimonious than that proposed in Fig. 4D.

Proteins in the NocA/TLP-1 subfamily have only one convincing C2H2 zinc finger, whereas the other Sp family members have three (as noted previously for NocA) (Cheah et al., 1994). Based on the supported phylogenetic hypotheses, the NocA/TLP-1 subfamily members lost the first and third zinc fingers. Having three zinc fingers is likely to be ancestral, because the most closely related C2H2 family, the KLF family, also has three zinc fingers (Knight and Shimeld, 2001).

All of the proteins in the Sp family have retained a sequence adjacent to the zinc-finger region called the ‘Btd (Buttonhead) box’ (Wimmer et al., 1996), although the consensus is reduced from that originally described: R-X(0-4)-C-X-C(/D/N)-P-N/Y-C. However, TLP-1 has an A instead of N or Y as the penultimate residue, and the other NocA/TLP-1 subfamily members do not have the R. Thus, there has been considerable evolution in the structure of the NocA/TLP-1 proteins after they diverged from other members of the Sp family, as indicated by the longer branch lengths of their lineages in the phylogeny (Fig. 4D). Consistent with this increased accumulation of differences, phenetic (distance-based) phylogenetic analyses artefactually place the NocA/TLP-1 subfamily outside of either the Sp or KLF families. In this case, the 5 non-SPLALLA Sp sequences and the vertebrate Sp subfamily are grouped on the basis of plesiomorphic (primitive) similarity alone. We do not know what adaptive significance there may be, if any, to this increased divergence.

A transgene in which gfp was fused in frame with tlp-1 and whose expression was designed to be driven by the tlp-1 promoter rescued the T cell defect but not the Lep defects of tlp-1 mutants (Table 1). tlp-1::gfp expression begins to be barely detectable at the beginning of gastrulation at about 100 minutes of embryonic development (Fig. 5A). Shortly after the gastrulation begins, the level of expression increases and is detectable in the nuclei of most embryonic cells (Fig. 5B). This pattern appears to persist through gastrulation. However, at about 260 minutes, expression in the anterior of the embryo fades and expression in the posterior of the embryo persists and gets stronger (Fig. 5C). At the 1.5-fold stage, at about 400 minutes, we primarily observe expression in posterior nuclei (Fig. 5D) and this pattern continues throughout the rest of development. We have not characterized this expression pattern in detail, as we do not observe any embryonic defects in tlp-1 mutants. However, we have observed expression in hyp9 and hyp10 in threefold embryos but not in the T cells (not shown).

During larval development, we observed tlp-1::gfp expression in the posterior intestinal cells, several neuronal nuclei in the head (not shown), the tail tip cells in hermaphrodites and males and in descendants of the T cell lineage. The tail tip cell expression we observed was weak, but consistent with a role for tlp-1 in the differentiation of these cells that, when reduced or absent, could lead to the Lep tail defect we observed in tlp-1 males.

We have examined tlp-1::gfp expression in the T cell descendants in some detail. We observed weak GFP expression in the T cell in 6% (n=124) of animals (Fig. 5E). After the division of the T cell, we observed stronger GFP expression in the posterior T cell daughter, T.p. but not in T.a (Fig. 5F). More specifically, of the animals that showed GFP expression in the T cell lineage (57/150), 88% of the animals examined showed GFP expression in T.p alone and 12% showed expression in T.a and T.p. This expression pattern correlates well with the cell lineage defects we observed in the T.p sublineage. We also observed GFP expression in the posterior T.ap daughter cell, T.app but not in T.ap or its anterior daughter cell T.apa (Fig. 5G). Specifically, all of the animals that showed expression in the T.ap lineage (11/56) showed GFP expression in T.app alone. We did not observe expression in T.apa in any animals. Although we did not observe expression in T.ap, the expression pattern correlates with the cell lineage defects we observed in the T.app division. In summary, these results suggest that TLP-1 is asymmetrically segregated or asymmetrically expressed in the T cell descendants.

Mutations in lin-44 cause the polarity of the first division of the T cell to be reversed (Herman and Horvitz, 1994). LIN-44 encodes a Wnt protein that is expressed from the tail tip hypodermis and signals polarity information to the more anterior T cells. We examined the expression of tlp-1::gfp in a lin-44 mutant to determine whether TLP-1 localization might respond to LIN-44 polarity signals. We observed weak GFP expression in the T cell in 8% (n=108) of lin-44 animals (not shown). Interestingly, we found that tlp-1::gfp expression is often reversed in lin-44 mutants, reflecting the reversal in T cell polarity (Fig. 6A). Specifically, of the animals in which we observed GFP expression in the T cell lineage (20/86), 60% showed expression in T.a only, 20% showed expression in T.p only, and 20% showed expression in T.a and T.p. These results correlate with the effect of lin-44 mutations on the T cell division: roughly 70% of the T cell divisions are reversed, 20% are symmetric and 10% are normal (Herman and Horvitz, 1994). This suggests that the asymmetry of tlp-1 expression is controlled by the LIN-44 signaling pathway.

To further explore how lin-44 might regulate tlp-1, we analyzed T cell lineages in a lin-44; tlp-1double mutant (Fig. 6B). In each lineage it appeared that the polarity of the T cell division was reversed, as it is in lin-44 single mutants. However, the subsequent cell division patterns were reminiscent of tlp-1 single mutants, in that we observed the loss of both neural cell fates and cell divisions in four lineages and only the loss of cell divisions in another four lineages. Thus it appears that in lin-44; tlp-1double mutants we observed the effects of both mutations, indicating that tlp-1 mutants retain some T cell polarity information that can be affected by mutations in lin-44. In addition, lin-44 and tlp-1 appear to interact in the formation of the male tail. lin-44 males display a weak Lep tail defect that is significantly enhanced by one mutant copy of tlp-1 (Table 1). This suggests that tlp-1 might also function with, and perhaps downstream of, lin-44 in the development of the male tail.

We also examined the localization of tlp-1::gfp expression in lin-17 mutants. We observed weak tlp-1::gfp expression in 4.0% of T cells examined (n=202) in lin-17 animals. Of the animals that showed GFP expression in the T cell lineage after the division of the T cell (11/140), 1.0% showed GFP expression in T.a alone, 5.0% showed expression in T.p alone and 2% showed expression in T.a and T.p (not shown). We also analyzed the T cell lineages in a lin-17; tlp-1 double mutant. In each lineage, there was a loss of cell polarity, resulting in the generation of only hypodermal cell fates. In fact, both the pattern of cell divisions and cell fates generated was indistinguishable from that observed in lin-17 single mutants, suggesting that lin-17 is epistatic to tlp-1 and that the neural cell fates we sometimes observed in tlp-1 single mutants were dependant upon lin-17 function (Fig. 6C); consistent with the observation that lin-17 mutations cause a loss of tlp-1::gfp expression.

tlp-1 is required for proper morphogenesis of the C. elegans tail. Mutations in tlp-1 cause defects in tail morphogenesis by affecting two fundamental processes: specification of asymmetric cell fates and cell fusion. The T cell lineage defects observed in tlp-1 mutants suggest that tlp-1 is involved in the generation or execution of T cell fate asymmetry. Mutations in tlp-1 also cause the normally rounded, peloderan male tail to be pointed, or leptoderan, apparently owing to a heterochronic delay in the highly ordered set of cell fusions and retractions of the tail tip hypodermis.

Molecular analysis revealed that tlp-1 encodes a C2H2-type zinc-finger protein that is a member of the Sp family of transcription factors. Although the zinc-finger motifs of TLP-1 and its orthologs (Fig. 4D) are quite divergent from other Sp family members, they are probably most closely related to the Sp subfamily of vertebrate transcriptional activators. NocA and Noz1, as well as members of the other Sp subfamilies (e.g. mouse Sp4 and D. melanogaster Buttonhead and D-Sp1), are also transcription factors with roles in developing brain, CNS and peripheral sensory organs (Cheah et al., 1994; Supp et al., 1996; Wimmer et al., 1996; Schöck et al., 1999; Andreazzoli et al., 2001). TLP-1 is expressed in nuclei and has a functional role in the postembryonic development of the phasmid sensory organs in the C. elegans tail. Thus, the most parsimonious prediction for TLP-1 function would be as a transcription factor involved in patterning neuronal organs.

By the same logic, one might also predict an embryonic role for tlp-1 in the developing brain. Although tlp-1 has a dynamic and highly regulated embryonic expression pattern, we do not observe any embryonic defects in any of our tlp-1 mutants, including ny14, which completely deletes the locus. That there are no major phenotypic consequences in homozygous null embryos suggests that there may be another gene acting redundantly with tlp-1 in its embryonic role. As a precedent for such redundancy, buttonhead and Sp1 are redundant for postembryonic sensory organ development in Drosophila, although buttonhead alone is required for embryonic head segment formation (Schöck et al., 1999). There are three other as yet uncharacterized genes in C. elegans that belong to the Sp family (Y40B1A.4, T22C8.5, and F45H11.1) that could be good candidates for having overlapping function with tlp-1. However, if TLP-1 does not function in embryonic development, this would be a novel trait that arose subsequent to the divergence of TLP-1 from its orthologs; the embryonic expression pattern might then be a nonfunctional vestige of the ancestral embryonic function.

Loss of tlp-1 function leads to the loss of specific T cell divisions and cell fates in the divisions of the posterior T cell daughter, T.p, and the T.ap cell. We observed the loss of neural cell fates and cell divisions in both affected cells (Fig. 1). What role does tlp-1 play in controlling these divisions? One possibility is that tlp-1 is involved only in the specification of neural cell fates. However, the loss of divisions in the T.p and T.ap sublineages, including the T.apa cell division which does not generate neural cell fates, suggests that tlp-1 is not just involved in the specification of neural cell fates. For example, tlp-1 may a play a role in the asymmetric expression of cell fate factors in response to cell polarity cues.

In fact, TLP-1 itself is asymmetrically expressed in the T cell lineage in response to cell polarity cues. We observed tlp-1::gfp expression in the T cell and the T.p cell. Later, tlp-1::gfp expression reappears in T.app, a descendant of the T.a and T.ap cells which never expressed it, suggesting that tlp-1 is asymmetrically expressed rather than asymmetrically segregated in the T cell lineage (Fig. 7). However, it is also possible that tlp-1 is expressed in T.ap, and perhaps T.apa, at levels below detection, which may explain why we observed defects in the division of T.apa. Alternatively, tlp-1 may not be expressed in T.ap and T.apa fate may be determined by signals from T.app, which expresses tlp-1. Finally, tlp-1 may be involved in a process required for proper T.apa division that gets set up in the T cell and is subsequently segregated to T.apa.

The precise mechanism by which tlp-1 is asymmetrically expressed is not clear. However, the asymmetric expression of tlp-1 is controlled by LIN-44/Wnt signals (Figs 6A, 7). This suggests that tlp-1 may be involved in the specification of cell fates in response to LIN-44 signals that control cell polarity. Taken together with the effects of lin-44 mutations on TLP-1 localization, it appears that a LIN-44 signal functions to polarize the T cell lineage but that TLP-1 is needed for correct cell fate specification within the polarized lineage. Thus, tlp-1 appears to be a downstream target of a Wnt signal that functions to polarize cells.

In addition, we rarely observed tlp-1::gfp expression in lin-17 animals, consistent with the idea that both T cell daughters take on a T.a-like cell fate, in which one would not expect to observe tlp-1 expression (Fig. 7). Another transcription factor whose expression is asymmetric in the T cell lineage is POP-1/Tcf (Herman, 2001). POP-1 is also asymmetrically distributed at many divisions during embryonic and postembryonic development, with anterior daughter cells having higher levels than do posterior daughter cells (Lin et al., 1998; Herman, 2001). Interestingly, the expression of POP-1 is the opposite of TLP-1; POP-1 is expressed at high levels in T.a and at low levels in T.p. Furthermore, like TLP-1, POP-1 expression also responds to LIN-44/Wnt signals; in lin-44 mutants, POP-1 expression is reversed, being expressed at high levels in T.p and at low levels in T.a. The mechanism by which asymmetric POP-1 expression is also not clear. In addition, for the T cell division, it does not appear that the different levels of POP-1 per se, specify cell polarity, as reducing pop-1 function caused both T cell daughters to take on a hypodermal, or anterior, cell fate. This suggests that the high level of POP-1 observed in T.a may be nonfunctional and that a low level of a perhaps modified form of POP-1 is required for the posterior cell fate (Herman, 2001). Possibly this form of POP-1 is involved in regulating the expression of TLP-1, however this has not yet been tested. Alternatively, TLP-1 may be involved in regulating the distribution and modification of the active form of POP-1.

Male tail morphogenesis involves a specifically timed set of cell fusions and cell retractions (Nguyen et al., 1999). We observed that these specific cell fusions and retractions are heterochronically delayed in tlp-1 males, leading to the failure of tail tip retractions and resulting in the leptoderan tail (Lep) defect. How are these cell fusions and retractions initiated and controlled and what role might TLP-1 play in the process? We considered two possibilities: either the fusion and retraction program is intrinsic to the tail tip cells; or an extrinsic cell signal(s), from an unknown source(s) triggers the beginning of the fusions and retractions. TLP-1 is expressed in the tail tip cells hyp9 and hyp10 during embryogenesis. If the cell fusion and retraction program is intrinsic to the tail tip cells, tlp-1 might function to control the fate of the tail tip cells or be required to initiate the cell fusion and retraction program. Interestingly, the tail tip cells are also posterior daughters of asymmetric divisions in the embryo, suggesting the possibility that tail tip cell fate may be part of a asymmetric cell fate governed by TLP-1. Our results did not rule out the existence of an extrinsic cell fusion signal. However, they did suggest that such a signal does not emanate from the T cell descendants and that the abnormal tlp-1 T cell fates do not interfere with such a signal. As TLP-1 probably functions as a transcription factor, it could control the fate or the production of the signal within the signaling cell. Alternatively, it could control the fate of the tail tip cells or their competence to receive a signal. There is yet no evidence for such a signal, although lin-44/Wnt mutant males display a low level of Lep tails, suggesting that a role for Wnt signaling in the control of male tail tip morphogenesis. Furthermore, the frequency of Lep tails observed in lin-44 males is significantly enhanced by tlp-1 mutations, suggesting that tlp-1 functions with lin-44 in male tail morphogenesis as it does in the asymmetric divisions of the T cell.

Other C. elegans transcription factors have been shown to play a role in cell fusions. Three C. elegans Hox genes have been implicated in the control of cell fusions during both embryonic and postembryonic development (Clark et al., 1993; Wang et al., 1993; Brunschwig et al., 1999). During postembryonic development, lin-39 and mab-5 function to control the fusions of the ventral hypodermal cells, the Pn.p cells, to the hyp7 syncytium. In particular, the Hox gene lin-39, which encodes the C. elegans homolog of the Drosophila Deformed and Sex combs reduced genes (Clark et al., 1993; Wang et al., 1993), functions in hermaphrodites to control the fusion of the P(3-8).p cells to allow the formation of the vulval precursor cells. In this case, lin-39 expression is required to prevent cell fusions (Clark et al., 1993; Wang et al., 1993). Another C. elegans Hox gene, mab-5, which encodes the C. elegans homolog of the Drosophila Antennepedia gene, functions to block P(9-11).p cell fusions in males (Kenyon, 1986). The expression of both lin-39 and mab-5 are controlled by egl-27 (Ch’ng and Kenyon, 1999). egl-27 encodes a product that contains an N-terminal region that has similarity to mammalian MTA (metastasis-associated) proteins. (Herman et al., 1999; Solari et al., 1999). MTA proteins have been biochemically linked to nuclear protein complexes that are involved in chromatin remodeling and histone deacetylation (Xue et al., 1998). Interestingly, mutations in egl-27 also cause defects in the polarity of the T cell (Herman et al., 1999), suggesting a link between the control of cell fusion and cell polarity at the level of chromatin remodeling.

The Lep tails observed in tlp-1 males are morphologically distinct from those observed in leptoderan species of rhabditid nematodes related to C. elegans. Thus, tlp-1 is not solely responsible for the observed variation, but is involved in processes responsible for such variation. Interestingly, mutations in lep-1 also lead to the failure of male tail tip retraction and cause C. elegans males to have a tail morphology similar to that of leptoderan nematodes. The lep-1 male tails resemble the tails of leptoderan nematodes more closely than do tlp-1 male tails. Initial analysis of lep-1 males revealed that the fusions of the tail tip cells are delayed. The tail hypodermal cells also fail to fuse with each other in closely related nematode species with naturally occurring leptoderan tails (Nguyen et al., 1999). Together these results suggest that cell fusions are integral to nematode male tail morphogenesis (Shemer and Podbilewicz, 2000), and that morphological diversity can thus result from evolutionary changes in cell fusions and cell shape. However, more work is needed to determine what the morphogenetic mechanism is and how the program is initiated. Finally, the Lep defects of both lep-1 and tlp-1 mutants suggest that changes in cell fusions can lead to changes in morphogenesis that may contribute to natural variation in form.

Besides Noc, the other D. melanogaster ortholog of TLP-1 is ElbowB (accession no. AAF44874) (M. Wernet and C. Desplan, pers. comm.).

The authors thank members of both the Fitch and Herman laboratories for useful discussions during the course of this work. Some strains used were obtained from the Caenorhabditis Genetic Center, which is supported by NIH NCRR. This work was supported by NSF grants IBN-9506844 and DEB-9981632 to D. F. and by NIH grant GM56339 to M. H.