ABSTRACT
In screens for mutants defective in vulval morphogenesis, multiple mutants were isolated in which the uterus and the vulva fail to make a proper connection. We describe five alleles that define the gene cog-2, forconnection of gonad defective. To form a functional connection between the vulva and the uterus, the anchor cell must fuse with the multinucleate uterine seam cell, derived from uterine cells that adopt a π lineage. In cog-2mutants, the anchor cell does not fuse to the uterine seam cell and, instead, remains at the apex of the vulva, blocking the connection between the vulval and uterine lumens, resulting in an egg-laying defective phenotype. According to lineage analysis and expression assays for two π-cell-specific markers, induction of the πfate occurs normally in cog-2mutants. We have cloned cog-2and shown that it encodes a Sox family transcription factor that is expressed in the π lineage. Thus, it appears that COG-2 is a transcription factor that regulates a late-stage aspect of uterine seam cell differentiation that specifically affects anchor cell-uterine seam cell fusion.
INTRODUCTION
The elucidation of organogenesis, in which groups of cells with various identities coordinate to form structurally intact and functional organs, is clearly important to the understanding of the development of multicellular organisms. Since the structure of the C. elegans vulva is well-defined at the anatomical level and the process of cell-fate determination in the vulva is well- characterized, vulva development provides an ideal system to study this problem.
C. elegans vulval development is an elaborate process requiring the generation of vulval precursor cells (VPCs), the determination of specific vulval cell fates, and finally the execution of vulval fates which involves an intricate morphogenesis process (Horvitz and Sternberg, 1991). VPC generation requires the action of the homeobox gene lin-39 in the first larval stage, allowing a group of six cells, P3.p-P8.p, to maintain competency to adopt vulval fates (Clark et al., 1993; Salser et al., 1993). Vulval cell fate determination occurs during the third larval stage when three of the VPCs, P5.p-P7.p, are induced by a RTK/Ras/MAPK-mediated signal from the anchor cell (AC) and then divide to form the 22 vulval cells. P6.p adopts a 1° vulval fate and produces eight progeny. P5.p and P7.p adopt a 2° vulval fate and produce seven progeny each. The 1° versus 2° fate decision is determined by each VPC in response to the RTK/Ras/MAPK-mediated pathway activated by the inductive signal as well as lin-12 mediated, lateral signals between VPCs (reviewed by Sternberg and Han, 1998).
During the execution phase of vulval development, in the third and fourth larval stages, the 22 descendants of the VPCs cooperate in a complex morphogenesis process to form the mature vulva. The most distal vulval cells migrate towards the center of the body and begin to invaginate. Ultimately, a stacked series of ring-like, multinucleate cells forms the tube-like structure connecting the uterus to the outside of the body (Fig. 1; J. White, personal communication). Vulval cells also form attachments to the uterus, the sex muscles and the hypodermis during this period (J. White, personal communication).
The AC plays multiple, essential roles during vulval development (Fig. 1). It invades the basement membrane surrounding the gonad and serves as the source of the inductive signal that promotes vulval fates in the VPCs (Hill and Sternberg, 1992; Fig. 1). In addition, it plays an integral role in establishing a functional connection between the vulva and the uterus. The AC signals six of the twelve granddaughters of the ventral uterine cells, causing them to adopt a fate, π? different from the default fate, ρ? of their six sisters and cousins (Fig. 1; Newman et al., 1995). The six π precursors (three per side) divide to form twelve π cells. Eight of the π cells (four per side) fuse to form the large multinucleate uterine seam (utse) cell that lies over the top of the vulva orifice and underlies the developing uterus (Newman et al., 1996). The remaining four π cells (two per side) adopt the uv1 fate and also make connections with vulval cells (Newman et al., 1996). The AC must also penetrate the dorsal vulval toroid cell (vulF) to form a channel through the top of the vulva (Fig. 1). This places the AC in a position that blocks the entrance to the vulva from the uterus. The AC then plays a final essential part by fusing with the utse cell (Fig. 1; Newman et al., 1996). Its nucleus and the eight nuclei of the utse cell move to the distal corners of the utse cell, leaving a thin layer of cytoplasm overlaying the vulval orifice, separating the vulval and uterine lumens. This thin laminar process can presumably be broken when the first egg is laid to establish a direct channel between the two organs (Newman et al., 1996).
The complexity of the morphogenesis process as well as previous genetic studies suggest that vulval morphogenesis requires the action of many gene products (Seydoux et al., 1993). We have initiated a genetic dissection of organogenesis by focusing on one of the last steps in vulval morphogenesis, the formation of a functional connection between the vulva and the uterus. We have cloned and characterized a gene, cog-2, that plays a key role in the AC-utse cell fusion. In the absence of cog-2 gene activity, the AC performs its initial functions normally so that all the components for establishing the vulval- uterine connection (the vulval cells and the π cells) are coordinated properly in time and space. However, fusion between the AC and utse cell is defective, preventing the formation of a functional vulva.
MATERIALS AND METHODS
Strains and genetic methods
Methods for manipulating the wild-type Bristol N2 and N2-derived strains were as described by Brenner (1974). Strains were maintained and experiments were conducted at 20°C, unless otherwise noted. Mutations used in this study, described by Riddle et al. (1997), are: unc-119(ed3) III, dpy-20(e1282, e1362) IV, dpy-8(e130) X, lon-2(e678) X, mec-2(e75) X, mom-1(zu188) X, and unc-6(e78) X. The lin-11::gfp line contains mgIs21[pRF4; plin-11-ABCDE-GFP] (Hobert et al., 1998). The cdh-3::gfp construct is from strain NL1008: dpy-20(e1362) IV; pkEX246 [dpy-20(+); cdh-3::gfp] (Pettitt et al., 1996).
Isolation of cog-2 mutants
cog-2 alleles were isolated in screens for temperature-sensitive (TS) and non-TS mutants. The cog-2 alleles are non-TS. N2 L4 hermaphrodites were treated with 50 mM ethylmethane sulfonate (Brenner, 1974) for 4 hours, cloned and maintained at 15°C. A synchronized population of F2 animals was prepared by picking 30 adult F1 hermaphrodites to each of multiple plates and allowing them to lay eggs for only 3 hours. The F2 eggs were allowed to hatch at 15°C and develop to the L2 stage. F2 progeny were then switched to 25°C for approximately 15 hours, during the stages of vulval induction and morphogenesis, and then returned to 15°C. F2 progeny were screened after 2-3 days for egg-laying defective (Egl) hermaphrodites in which progeny have hatched inside the Egl mother. This population of mutants was subjected to a secondary screen in which the morphology of the vulva was directly observed using Nomarski optics to identify mutants with vulval morphogenesis defects. Five recessive alleles that define cog-2 were identified, among others. ku205, ku209 and ku241 all fail to complement ku207. ku205 and ku194 are rescued by the same cosmid. ku194, ku207 and ku241 were outcrossed at least 5 times and further characterized. ku205 and ku209 hermaphrodites fail to mate and have not been extensively outcrossed. Since ku205 and ku209 were not extensively outcrossed, only the phenotypic characterizations of ku194, ku207 and ku241 are reported. However, characterization of ku205 and ku209 indicated that these alleles are similar in severity to ku194 (data not shown). cog-2 mutant males have no apparent defects and mate with high efficiency (data not shown).
Genetic mapping and rescue of cog-2
ku194 was mapped to the X chromosome between the genetic markers mec-2 and unc-6. 27/47 Lon non/Unc and 10/19 Unc non-Lon recombinants from lon-2 unc-6/cog-2 heterozygotes segregated cog-2. 34/37 Lon non-Cog and 1/6 Cog non-Lon recombinants from lon-2 cog-2/mec-2 heterozygotes segregated mec-2. Transgenic strains were generated by germline transformation (Mello et al., 1991). Cosmids were obtained from A. Coulson (Sanger Center). For rescue experiments, 5-10 μg/ml of cosmid or plasmid was coinjected with 100 μg/ml of the transformation marker sur-5::gfp, pTG96 (Gu et al., 1998), into ku194/dpy-8 mom-1 unc-6 heterozygotes. Clones from each line that stably expressed GFP and that no longer segregated Dpy Unc animals were selected to obtain transgenic homozygous ku194 lines. Positive rescuing activity was scored as restoration of egg-laying ability to transgenic ku194 hermaphrodites. Other transgenic lines were obtained using unc-119(+) (Maduro and Pilgrim, 1995) as a transformation marker to inject unc-119(ed3) hermaphrodites. Extrachromosomal arrays were integrated as described by Mello and Fire (1995).
cog-2 cDNA cloning and sequencing
An 1816 base pair BamHI-HindIII genomic fragment encoding all of the predicted T22B7.1 gene and part of the predicted W01C8.2 gene in the rescuing fragment was used to probe the Stratagene (La Jolla, CA) Nematode UniZAP®XR λ mixed-stage cDNA library. From one million plaques, five positive clones were isolated, excised and sequenced using an ABI automated sequencer. Two cDNAs appeared full-length. Three were incomplete cDNAs likely derived from the same transcript as the full-length clones. The full-length cDNA contained 5 base pairs derived from the SL1 spliced leader sequence, a 3 base pair 5′ UTR, a single open reading frame coding for 471 amino acids and a 672 base pair 3′ UTR. Molecular lesions were identified by sequencing PCR amplified genomic DNA from lysates of mutant worms.
Plasmid construction
T22B7 was digested with KpnI and religated to produce pKG6. The two largest PstI-KpnI fragments from pKG6 were separately ligated to pBluescript S/K+ to produce pKG6-3′ and pKG6-5′. The 5.8 kilobase pair EcoRI fragment from pKG6-3′ was cloned into pBluescript to make pWH4. pWH4 was ligated to the 8.5 kilobase pair KpnI-SalI fragment from pKG6-3′ to make pWH5. pRC1 was created by digesting pWH4 with AflII and religating and then ligating to the KpnI-SalI fragment as above. pRC2 was created by digesting pWH4 with SalI and religating, digesting with HindIII and EcoRI, creating blunt ends with Klenow and then religating, and finally, cloning in the KpnI-SalI fragment, as above.
To make the cog-2::gfp translational fusion, pWH14, a fragment that contains sequences including the HindIII site in the seventh exon through the last amino acid was amplified using primers that introduce a NotI and a BamHI site, 5′GCGATGGTATTATTCTAAAGCT- TTCAC3′ and 5′CGGGATCCGCGGCCGCTTCAGCTGTTTGTA- GGAGATGATGTGA3′, and cloned into the HindIII and BamHI sites of pBluescript. This plasmid was ligated to a 2.3 kilobase pair SalI- HindIII fragment from pWH5 containing the remaining cog-2 coding sequence and part of the promoter. The remainder of the promoter was ligated upstream in the form of a 8.5 kilobase pair KpnI-SalI fragment from pWH5. The resulting construct was digested with NotI and ligated to a fragment with EagI ends containing the cog-2 3′ end amplified from cosmid W01C8 and having ApaI and PstI sites introduced 5′ to the UTR. The primers used in this amplification were 5′CGCGGCCGGGGGCCCAATCTGCAGGTAACTTTAACCAAGC GTGTCATTCC3′ and 5′GCGG-CCGCGCGGCCGCACATTTCCT- ACCATAATCCG3′. Finally, to produce pWH14, an ApaI-PstI fragment from pPD102.33 containing a GFP cassette (A. Fire, S. Xu, J. Ahnn and G. Seydoux, personal communication) was cloned into the vector cut with ApaI and PstI.
To produce pWH17, the cog-2::gfp transcriptional fusion, pWH14 was digested with NotI and HindIII to release a 7.4 kilobase pair fragment containing the vector, the promoter and the first 19 codons of cog-2. This fragment was ligated to a 1.9 kilobase pair HindIII- EagI fragment from pPD95.69 (A. Fire, S. Xu, J. Ahnn and G. Seydoux, personal communication) to produce a cog-2::gfp transcriptional fusion containing the unc-54 3′ end. The cog-2::gfp translational fusion was injected into unc-119(ed3) worms along with an unc-119(+) marker to allow selection of transformed animals and the fluorescence pattern was determined in a non Cog background.
RESULTS
cog-2 mutants have a defective vulval-uterine connection
To identify vulval morphogenesis defective mutants, the F2 generation of mutagenized animals was screened for Egl mutants. All Egl mutants were subjected to a secondary screen in which the vulva structure was examined directly using Nomarski optics to identify mutants with normal vulval lineages but specific structural defects indicative of abnormal morphogenesis. Eight mutants that had defects in establishing a functional vulval-uterine connection were identified in screens covering 13,000 haploid genomes. Five of these alleles define a previously unidentified gene, cog-2 (connection of gonad defective). The remaining three alleles define two additional cog genes that will be described elsewhere.
cog-2 mutant hermaphrodites display a highly penetrant Egl phenotype, an incompletely penetrant protruding vulva phenotype, and show a diminished brood size (Table 1). Any Egl defect may result indirectly in a small brood size because fertilized eggs hatch inside the hermaphrodite and kill it before the end of its reproductive age. However, a more direct effect of cog-2 on brood size is suggested by the low frequency of ku194 and ku207 hermaphrodites that have no progeny and by the small brood size of the ku207 homozygous animals that do lay eggs (approximately 134, compared to the wild-type brood size of 300) (Table 1) (Wood, 1988). Even though all alleles are recessive, we were not able to determine the genetic nature of the cog-2 alleles because cog-2 is not uncovered by genetic deficiencies in the region (data not shown). However, ku194 is a likely molecular null (see below), consistent with our analysis of its highly penetrant Egl phenotype.
Additionally, a gonad migration defect is associated with all cog-2 alleles between two and four times more often than in a wild-type background (Table 2). 13% of gonad arms follow abnormal migration patterns in ku194 hermaphrodites (Table 2). In 2 of 11 abnormal migrations that were scored, the gonad reflexed prematurely. In the remaining 9 abnormal cases, the arms migrated distally for the correct time and distance and then reflexed and migrated proximally again as expected. However, they often followed an abnormally ventralized or dorsalized path. One of these 9 abnormal ku194 gonad arms failed to move dorsally at all, three moved dorsally too soon (Fig. 2D), two migrated dorsally and then returned abnormally to the ventral side to continue the migration, and three briefly migrated either dorsally or ventrally at the improper time and then returned to their normal migration path.
To investigate the Egl phenotype, we examined by Nomarski optics the AC, VPCs, and uterine cells in cog-2 mutants at various developmental stages. The vulval structure appears normal at the L3 molt in cog-2 mutants. All 22 nuclei are positioned properly and the correctly shaped invagination is formed. However, as organogenesis proceeds throughout L4, the vulva structure becomes obviously abnormal (Fig. 2). In wild-type, mid-L4 hermaphrodites, a thin planar process derived from utse cell cytoplasm can be seen in a lateral view lying across the top of the vulval lumen, separating the vulval and uterine lumens (Fig. 2A). The AC that formally occupied the top of the vulva has fused with the utse cell, and its nucleus has migrated distally to clear the orifice (Newman et al., 1996). In cog-2(ku194) mutants, the thin planar process at the top of the L4 vulva is not visible, and the AC nucleus abnormally appears in its place (Fig. 2B). cog-2 mutants with normal vulval morphogenesis are rarely observed (Table 2). The AC-block phenotype is highly penetrant (81-95%) for all cog-2 alleles (Table 2), and the remainder of the mutants (5-19%) have more grossly affected vulval structures (Fig. 2D, and Table 2). A majority of the animals with severe defects in vulval morphology also have defects in gonad migration, although this correlation is not absolute (Fig. 2D). Frequently, it also appears that the AC fails to form a channel through the vulF cell in cog-2 mutants because the anterior and posterior regions of the vulva come into close contact at the apex rather than appearing widely separated as in the wild-type vulva (compare Fig. 2A and B).
The AC and the utse cell do not fuse in cog-2 mutants
Because the AC nucleus fails to move away from the vulval orifice to the corner of the utse cell in cog-2 mutants, we investigated whether the AC-utse cell fusion occurs. A cdh- 3::gfp reporter was used to monitor fusion in wild-type and mutant worms (Pettitt et al., 1996). cdh-3::gfp carried on an extrachromosomal array is expressed in wild-type hermaphrodites during L3 in the cytoplasm of the AC, resulting in a distinct spot of fluorescence at the apex of the vulva (Fig. 3A,B). As the AC fuses to the utse cell during L3 lethargus, fluorescence spreads from the AC cytoplasm throughout the larger utse cell, resulting in a diffuse fluorescent pattern above the vulva (Fig. 3C,D). In cog-2(ku207) mutants, the fluorescence pattern of the CDH-3::GFP fusion protein is observed in the area of the AC cytoplasm in L3 hermaphrodites as is seen in wild-type hermaphrodites. However, in L4 hermaphrodites, after fusion should have occurred, the distinct, AC pattern of expression is still visible (Fig. 3E,F), and AC fluorescence at the apex of the vulva continues into the adult stages of development (Fig. 3G,H). The diffuse staining pattern indicative of AC-utse cell fusion is rarely seen. 30 of 30 wild-type worms displayed the fluorescence pattern indicative of fusion. In contrast, 26 of 33 L4 and young adult cog-2(ku207) hermaphrodites displayed a distinct fluorescence pattern indicative of an unfused AC, and the remaining 7 animals displayed an ambiguous expression pattern. These results suggest that the AC fails to fuse with the utse cell in cog-2 mutants.
π cell fate appears normal in cog-2 mutants
In cog-2(ku207) mutants, the AC nucleus occasionally adopts an abnormal morphology, appearing large and bloated (Fig. 2C and Table 2). A similar bloated AC phenotype was described previously, under conditions in which the π cells had been ablated (Newman et al., 1996). Because both π cell ablation and a cog-2 mutation can lead to a similar bloated AC phenotype and because the AC could fail to fuse with the utse cell due to a defect in the π cell lineage from which the utse cell is derived, we investigated the status of the π cell lineage in cog-2 mutants. π cell precursors that are not induced by a signal from the AC adopt the default ρ lineage. A π lineage is easily distinguishable from a ρ lineage based on the characteristic number and orientation of divisions (Newman et al., 1995). π cell precursors divide once with a dorsal-ventral division plane to produce a larger, dorsal cell and a smaller, ventral cell (Fig. 4). In contrast, two divisions in a mostly anterior-posterior plane are characteristic of a ρ lineage (with the exception of the outer ρ precursor on each side of the animal, each of which undergoes an initial lateral division to restore symmetry to the uterus) (Kimble and Hirsh, 1979; Newman et al., 1995).
In cog-2 mutants, the π cell precursors adopt a π fate, rather than the default ρ fate, based on their division patterns (Fig. 4). All three π precursors per side of the gonad were lineaged from 21 sides of ku194 hermaphrodites for a total of 63 cells lineaged. 56 of these cells underwent an asymmetric division in a dorsal-ventral plane (Fig. 4). Furthermore, two worms in which normal lineages were observed for all three π cells on both sides of the gonad were recovered from the slide after lineaging, and these animals manifested a completely Egl phenotype suggesting that the Cog phenotype of ku194 mutants is the result of something other than a lack of π cell induction. A single π cell was observed to divide in a manner reminiscent of a ρ lineage, and the remaining 6/63 cells were from a single worm in which both the π and ρ precursors failed to divide by the beginning of L4 (Fig. 4). This failure to divide phenotype does not appear to be the cause of the highly penetrant Egl phenotype associated with ku194 animals. It is possible that the less penetrant grossly abnormal morphogenesis phenotype could be associated with a failure of these lineages to progress. Vulval lineages were normal in all cog-2 mutants examined (data not shown).
Further support for the observation that normal π cell induction takes place in cog-2 mutants is that the mutant π cells properly establish expression of two π-specific markers, lin-11::gfp and cog-2::gfp. LIN-11 is a LIM homeodomain transcription factor that is expressed in the π cells as well as in cells of the vulval secondary lineage (Freyd et al., 1990; Newman and Sternberg, 1996). An integrated array containing a lin-11::gfp fusion reporter with GFP fused in frame to the fifth exon of lin-11 (Hobert et al., 1998) is expressed by wild-type hermaphrodites in the π cells at the beginning of L4 (Fig. 5A,B). ku194 animals carrying the lin-11::gfp integrated array express LIN-11::GFP in a similar pattern, indicating that the π cell fate is intact with regard to expression of this lineage marker (Fig. 5C,D). Identical results were observed with a lin- 11::lacZ reporter (data not shown). Since cog-2 is also expressed in the π cells (see below), a cog-2::gfp transcriptional fusion was used to mark the π cells in wild-type (Fig. 5E,F) and cog-2(ku194) (Fig. 5G,H) hermaphrodites, and once again all six mutant π cells express this reporter (Fig. 5). Expression of these two π-cell-specific markers in cog-2 mutants indicates that the π cell fate is not severely disrupted in cog-2 mutants, but does not exclude a later utse cell-specific defect. In fact, a late defect is consistent with observations that while the expression of the cog-2::gfp transcriptional reporter remains visible during late L4 in the twelve π cells in a wild-type background, expression becomes fainter in these cells and additional faintly expressing cells are sometimes seen during late L4 and young adult stages in a cog-2 mutant background (data not shown).
cog-2 encodes a Sox domain protein
ku194 was mapped to the X chromosome between mec-2 and unc-6. cog-2 was cloned using microinjection transformation to rescue the Egl phenotype of ku194. Cosmid T22B7 and subclones containing the 3′ end of T22B7 can efficiently restore the ability of transgenic ku194 hermaphrodites to lay eggs (Fig. 6A). Genefinder (Wilson et al., 1994) predicted two transcripts in this region, T22B7.1 and W01C8.2. However, small deletions in either predicted gene eliminated rescuing activity (Fig. 6A). By screening a cDNA library using a probe that included portions of both predicted transcripts, we identified a single cDNA class. The complete cog-2 cDNA contains the last three exons of T22B7.1 and all five exons of WO1C8.2 (with some slight differences from the predicted intron-exon boundaries in both cases), as well as two additional exons from 1.7 and 3 kilobase pairs upstream that were not predicted by Genefinder. cog-2 appears to be transpliced to SL1, contains 10 exons and encodes a predicted protein of 471 amino acids (Figs 6B, 7A). The 672 base pair 3′UTR identified in the full-length cDNA is dispensable for functional activity since the rescuing cosmid, T22B7, and the T22B7-derived rescuing plasmids end just 31 base pairs downstream of the predicted stop codon, thereby removing most of the 3′ UTR.
The predicted product of the cog-2 gene is homologous to a large family of putative, transcription factors called SOX (human) or Sox (murine) domain proteins. All Sox proteins have a single high mobility group (HMG) box with significant similarity to the HMG box of the mammalian sex determining factor SRY (Laudet et al., 1993). COG-2 is most similar to the group D Sox domain proteins Sox5 and Sox6 (Fig. 7B; Connor et al., 1995; Denny et al., 1992b; Prior and Walter, 1996). COG-2 is greater than 70% identical to Sox5 and Sox6 in the region of the DNA-binding, HMG box, and the COG-2 HMG box is 96% identical to the Sox family HMG consensus (Nambu and Nambu, 1996; Fig. 7B). However, like many other Sox proteins, COG-2 shows no similarity outside of the HMG box to any other Sox protein or any other protein in the data bases.
The location of the lesion associated with each of the five cog-2 alleles was determined. ku209 and ku194 are amber mutations that result in the truncation of over half of the protein including the Sox domain, consistent with the hypothesis that ku194 is a strong, loss-of-function allele and a likely molecular null (Figs 6, 7A). ku207 is a missense mutation resulting in an alanine to threonine substitution within the Sox domain (Figs 6, 7A). This alanine is highly conserved among HMG boxes in general, not just within the SRY related Sox family (Fig. 7B; Laudet et al., 1993). ku205 and ku241 are splice site mutations occurring prior to the region coding for the Sox domain (Figs 6, 7A). Failure to splice at either of these sites would again lead to a severely truncated protein. Unfortunately, splicing in these mutants could not be assayed because no cog-2 transcript could be detected using northern blot analysis (data not shown).
cog-2 is expressed in the π cells
We investigated the expression pattern of cog-2 since it seems to be acting specifically on fusion between the AC and the utse cell. We examined the expression pattern of cog-2 using a reporter, pWH14, that contains 6.4 kilobase pairs of sequence upstream of the cog-2 coding region, the entire coding region fused to GFP, and the 3′UTR of cog-2 (Fig. 8A). The cog-2::gfp translational fusion encodes a functional protein since it can partially rescue the Egl phenotype of ku194 mutants as an extrachromosomal array (data not shown).
Four stable lines containing an extrachromosomal array of the translational fusion were examined, and one line was integrated and reexamined. All lines shared consistent expression patterns. COG-2::GFP is a nuclear protein consistent with COG-2 homology to transcription factors. cog-2::gfp is expressed in a dynamic pattern in cells of the π lineage (Fig. 8B). No expression is visible in the gonad until the π cell precursors are born. At this point, expression is evident in all π cell precursors as well as in the single ρ precursor on each side that has a π cell precursor as a sister (Fig. 8B,C). After the π cell precursors undergo their single and final division, cog-2::gfp expression is still bright in the π cells, but expression is usually no longer evident in the ρ lineage (Fig. 8B,D). Most animals show no expression of COG-2::GFP expression past the ρ precursor stage in the ρ lineage, although in a few animals, faint GFP expression was observed in the daughters, but never the granddaughters, of the expressing ρ precursor (Fig. 8B). cog-2::gfp expression remains strong in the π lineage throughout the L4 stage, becoming slightly brighter in the π cells that adopt the uv1 fate (Fig. 8B). In addition, the AC nucleus begins to express cog-2::gfp after fusion with the utse cell (not shown). Expression in the AC is not detectable before the time of fusion. cog-2::gfp expression is not detectable in the vulva at any stage, in any other part of the gonad including the distal tip cells, or in the hypodermis (data not shown). cog-2::gfp is expressed in most, if not all, neurons in the head and the tail of the animal (data not shown). However, four unidentified cells in the head express cog-2::gfp at a higher level than others (data not shown). Expression is also faintly visible in the body wall muscles and the intestinal cells (data not shown).
Because cog-2 is expressed in the π cells, a cog-2::gfp transcriptional fusion was produced to use as a marker for the π cell fate. The transcriptional fusion construct deletes all but the first 19 codons of the cog-2 coding region and the cog-2 3′UTR and replaces them with a GFP cassette containing a NLS and the unc-54 3′ UTR. The cog-2::gfp transcriptional fusion is expressed in a similar pattern to the translational fusion in the π and ρ lineages. The expression pattern of the transcriptional fusion differs from the translational fusion in that the body wall muscles express the fusion protein at a higher level than any other cell and fewer neurons expressing cog-2::gfp can be detected with the transcriptional fusion, although the same brightly expressing cells that stand out with the translational fusion are visible.
DISCUSSION
Mutations in cog-2 were identified in a screen for vulval morphogenesis defective mutants. cog-2(ku194) is likely a null mutation that primarily affects the egg-laying system of C. elegans and, to a lesser extent gonad migration. We have shown that the Egl phenotype is a result of a failure of the vulval and uterine lumens to make a functional connection. The AC-utse cell fusion event, a prerequisite for formation of a functional connection, is defective in cog-2 mutants. Instead, the AC remains at the dorsal opening of the vulva, blocking the passage of eggs from the uterus. We propose that a late-stage defect in utse cell determination, specifically affecting the ability of the utse cell and the AC to fuse, is the likely explanation for the cog-2 phenotype. This hypothesis is consistent with our observations that a cog-2 reporter is expressed in the π cells, which are utse cell precursors, but is not expressed in the AC. Furthermore, in cog-2 mutants the AC and the π cells appear to adopt normal fates and yet fail to fuse. The AC adopts a proper fate as assayed by its ability to perform its normal roles up until the time of fusion with the utse cell, including induction of the vulva cells and π cells, with the possible exception of penetration of vulF at a high frequency. The π cell fate is induced as assayed by lineage pattern and the wild-type expression pattern of two π cell-specific markers. Therefore, the defect in cog-2 mutants could be a specific late- stage cellular defect within the utse cell such that it is unable to fuse with the AC. Alternatively, there may be a lack of utse- AC signaling such that fusion is never initiated.
Cell-to-cell fusion is an important component of multiple developmental events such as fertilization (Wilson and Snell, 1998), formation of the placental lining (Hernandez-Verdun, 1974), myogenesis (Wakelam, 1985), and formation of ostoeclasts (Roodman, 1991), and is a mechanism employed frequently during C. elegans development. More than 20% of the nuclei in an adult hermaphrodite are found within syncytial cells that can have upwards of 100 nuclei (Podbilewicz and White, 1994; Sulston and Horvitz, 1977; Sulston et al., 1983). Although the purpose of many of these fusion events in C. elegans remains mysterious, the AC-utse cell fusion appears to be required for disposing of the AC, which appears to be simply in the way after it has carried out its previous functions. Since cog-2 mutants have a rather specific defect affecting primarily uterine and vulval morphogenesis, it is most likely that the role of COG-2 in cell fusion may be to allow identification of a specific cell fusion partner rather than as a general factor required for multiple cell fusion events. However, the identification of an important regulator of a cell fusion event opens the door for further investigation into the mechanism of cell fusion. Perhaps identification of targets of the COG-2 transcription factor will allow elucidatation of cell fusion processes.
COG-2 is a putative transcription factor
cog-2 encodes a putative transcription factor containing a Sox domain. Sox family members contain an HMG box that is greater than 50% identical to the HMG box of SRY. Sox genes, which are present in the genomes of many vertebrates, have diverse tissue-specific expression patterns during early development and have been implicated in numerous developmental processes (Denny et al., 1992a; Laudet et al., 1993; Pevny and Lovell- Badge, 1997; Prior and Walter, 1996). Examples include the murine genes Sox2, Sox4 and Sox9. Sox2 is expressed embryonically in the nervous system as well as other tissues and acts in association with Oct-3 to regulate fibroblast growth factor 4 in embryonic carcinoma cell lines (Ambrosetti et al., 1997; Collignon et al., 1996; Yuan et al., 1995). Sox4 is expressed in the embryonic heart and neural tissue and in the thymus after birth (Wetering et al., 1993). Sox9, which plays roles in both sex determination and skeletal formation, is expressed in the developing gonad and in chondrocytes (Morais-da-Silva et al., 1996; Wright et al., 1995). Additionally, a Sox gene from the Drosophila genome, fish-hook/Dichaete, is a segmentation gene that specifically regulates the even-skipped gene (Ma et al., 1998; Nambu and Nambu, 1996; Russell et al., 1996).
The Sox domain of COG-2 is most similar to that of the Sox5 and Sox6 proteins. Sox5 is a site-specific DNA-binding protein that binds to the minor groove of the double helix and severely bends DNA as do many other Sox proteins (Connor et al., 1994). Sox5 is likely an architectural type transcription factor. Architectural factors distort the DNA structure and may act to bring other factors into close proximity after binding distant sequences, to recruit components that bind specific structural elements or to prevent the binding of other factors. Based on the sequence similarity to Sox5 and the expression pattern of COG-2, we speculate that COG-2 is an architectural type transcription factor that targets a gene or genes required for terminal utse cell fate, perhaps acting to allow recognition of the utse cell fusion partner, the AC, or otherwise promote fusion.
cog-2 mutants display a partially penetrant protruding vulva phenotype. We speculate that this phenotype may be the result of defective connections between uterine and vulval cells resulting in abnormal eversion of the vulva at the last stage of morphogenesis. While the basis for any effect of cog-2 on brood size is unknown, cog-2 may play a role in early embryogenesis since it is expressed in the embryo prior to elongation (data not shown). Expression of cog-2 reporters in body wall muscles suggests a possible explanation for the gonad migration defects since the hermaphrodite gonad migrates along the body wall muscles (Kimble and Hirsh, 1979).
Expression of COG-2 in the nervous system of C. elegans is intriguing in light of the large number of Sox genes expressed during mammalian CNS development (Pevny and Lovell-Badge, 1997). cog-2 mutants have no obvious neural or behavioral defects. However, the C. elegans genome has multiple Sox genes, and COG-2 may act redundantly with any of these factors as has been postulated for Sox genes in mouse and human.
ACKNOWLEDGEMENTS
We thank K. Grant, R. Chuang and D. Fay for assistance with genetic and molecular analysis, J. Pettitt for the cdh-3::gfp reporter, O. Hobert for the lin-11::gfp reporter, A. Fire for GFP vectors, and members of the Han and Wood labs for discussions and critical reading of the manuscript. WH-R is supported by postdoctoral fellowship no. PF-98-113-01-DDC from the American Cancer Society. M. H. is a HHMI investigator. This work was supported by NIH grant no. GM47869.