RESULTS

Comparison of C. elegans lin-3 and C. briggsae lin-3: Outside of the EGF motif, C. elegans LIN-3 has no sequence similarity to any other known protein. To infer functional importance of other domains, we cloned the C. briggsae lin-3 by low-stringency hybridization using a C. elegans lin-3 genomic probe. The genomic structure of C. briggsae lin-3, therefore, is inferred from the comparison to lin-3 structure in C. elegans.

We first tested whether C. elegans lin-3 and C. briggsae lin-3 have interchangeable functions. Most C. elegans animals bearing multiple copies of lin-3 as a transgene exhibit a Muv phenotype (n > 50; also see Hill and Sternberg 1992; Changet al. 1999). Likewise, we found that C. elegans animals become Muv when bearing multiple copies of C. briggsae lin-3 as a transgene (n > 50). When C. briggsae animals bear a transgene with multiple copies of C. briggsae lin-3, the animals have vulval defects. Similarly, the expression of multiple copies of C. elegans lin-3 from a transgene induces similar defects in C. briggsae (Figure 1). However, instead of the pseudovulvae seen in C. elegans, many C. briggsae animals have one big, protruding vulva. Closer examination revealed that this is because some C. briggsae transgenic animals have adjacent 1° vulval cells (four of the five animals we observed had more than three VPCs induced; among the four animals, three had anatomy consistent with the presence of adjacent 1° vulval precursor cells), which can cause the formation of protruding vulvae (P. Sternberg and R. Palmer, unpublished observations). The phenotypic difference caused by the expression of multiple copies of lin-3 in the two species might reflect subtle differences in the signaling mechanism of lin-3 or in the responses of VPCs to lin-3. However, it is clear that lin-3 in both species plays a role in vulval formation. Other evidence supporting the functional conservation of lin-3 in the two species is that a C. elegans lin-3::lacZ construct was expressed in the AC of C. briggsae when introduced as a transgene (data not shown).

We then compared the number of exons, intron locations, nucleotide sequences, and the inferred protein sequences of C. briggsae and C. elegans lin-3. The number of exons and intron locations of lin-3 are almost identical in the two species (Figure 2). The most notable difference is that the EGF motif spans three exons in C. elegans lin-3, but only two in C. briggsae lin-3. The exons have identity at the nucleotide level ranging from 65 to 91%. For introns, 39% of the C. elegans sequence are identical to that of C. briggsae, and 53% of the C. briggsae intron sequences are identical to that of C. elegans. The difference arises because the introns in the two species are of different total lengths. The identity between individual introns ranges from 38 to 63%. This is an overestimate, since we always use the shorter sequence as the reference when two corresponding introns have different sizes in the two species. The overall amino acid sequence identity in the coding region is 76% (Figure 3). Previous sequence comparisons between C. elegans genes and their homologs in C. briggsae have demonstrated that exon and intron structures and protein sequences are highly conserved in the two species, but intronic and flanking sequences are not. The nucleotide identity for the coding region is generally >60%, and an even higher degree of identity is present for protein sequences (for examples, see Kennedyet al. 1993; Maduro and Pilgrim 1996). Results from our comparison of lin-3 in the two species are consistent with the previous studies.

LIN-3 protein can be divided into several domains: the signal sequence (which will be described later), the region between the signal sequence and the EGF motif, EGF motif, the region between the EGF motif and the transmembrane domain, the transmembrane domain, and the cytoplasmic domain. However, we will not attempt to determine which domain is functionally important using sequence identity as the sole criterion, since the overall degree of identity between LIN-3 from the two species is very high. The EGF motif is the most conserved region with 96% amino acid identity. This is consistent with the functional importance of this motif. The region between the signal sequence and the EGF motif is less conserved, with 65% amino acid identity. The 16 amino acids preceding the EGF motif, however, are completely conserved and contain several lysines and arginines. A region rich in lysines and arginines N-terminal to the EGF motif is required for the heparin binding of heparin-binding EGF (Thompsonet al. 1994), and it is possible that this domain of LIN-3 also interacts with extracellular matrix proteins. Most precursors of EGF family members are cleaved at both the N and C terminus of the EGF motif, and the lysines and arginines N-terminal to the EGF motif are required for the cleavage of amphiregulin precursor (Thorne and Plowman 1994). This region of LIN-3 might also have functions in the secretion of a soluble LIN-3. The region between the EGF motif and the transmembrane domain should also be involved in the cleavage of the LIN-3 precursor, and 89% of the residues in this region are conserved. The transmembrane domain in the two species has 88% of the amino acids conserved. Although none of the lin-3 alleles has a mutation in this domain (see below), it should nonetheless be important for membrane localization of the LIN-3 precursor. The cytoplasmic domain has an overall amino acid identity of 79%. The N-terminal portion of the cytoplasmic domain is less conserved compared to the more C-terminal region. The cytoplasmic domains of various EGF growth factors are quite dissimilar and usually very short (Massagué and Pandielle 1993). However, the cytoplasmic domain of LIN-3 is unusually long, and our findings from the study of lin-3 phenotypes and molecular lesions suggests that the cytoplasmic domain is functionally important (see below).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

lin-3 and its C. briggsae homolog are functionally interchangeable in vulval induction. Adult animals are shown in all panels. The anterior of the animal is to the left and the ventral side is down. Wild-type vulvae in (A) C. elegans and (B) adult C. briggsae, respectively. (C) Overexpression of C. elegans lin-3 causes a Muv phenotype in C. elegans. (D) Overexpression of C. briggsae lin-3 causes the formation of a big, protruding vulva in C. briggsae. (E) Overexpression of C. briggsae lin-3 causes a Muv phenotype in C. elegans. (F) Overexpression of C. elegans lin-3 causes the formation of a big, protruding vulva in C. briggsae. Scale bars represent 0.04 mm in A–C and ~0.04 mm in D–F. The arrows indicate the positions of vulvae or pseudovulvae.

There are two alternative splice sites in C. elegans lin-3, and both are conserved in the lin-3 homolog of C. briggsae (Figures 2 and 3). One alternative splice site lies between the EGF motif and the transmembrane domain (Hill and Sternberg 1992). The two putative protein products in C. elegans may differ in the range of their signaling (J. Liu and P. W. Sternberg, unpublished results). The alternatively spliced region consists of 15 amino acids and is 87% identical in the two species. This is consistent with the possibility that both proteins are functionally important. LIN-3 may also have two alternative N termini: one encoded by exons A, B, and C, and the other encoded by exon C2. Either exon C or C2 could provide a signal sequence for the membrane localization of the LIN-3 precursor (Hill and Sternberg 1992; Figure 2). Although all known cDNAs of C. elegans lin-3 contain exon C but not C2, exon C2 appears to be functional since transgenes containing exon C2, but not C, can induce vulval differentiation (Hill and Sternberg 1992; J. Liu and R. Hill, unpublished data). It is possible that in a genomic sequence lacking exon C, C2 can substitute for its function. When we performed RT-PCR on RNA extracted from C. elegans of mixed developmental stages, we were indeed able to amplify sequences encoded by exon C2, although with very low efficiency (data not shown). A total of 64% of the amino acids of the signal sequence encoded by exon C and 63% of those encoded by exon C2 are conserved between the two species (Figure 3).

Exons A, B, and part of C constitute the 5′ untranslated region (UTR) in C. elegans cDNA. The 5′ UTR is 71% identical (out of 152 nucleotides) with a region in the C. briggsae genomic DNA that is 5′ to the open reading frame. The degree of identity is higher than that between the introns. This region may have some regulatory functions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Genomic structures of C. elegans and C. briggsae lin-3. Exons are represented as boxes. Three exons code for the EGF motif in C. elegans, E, F, and G, but only two in C. briggsae. We named these two exons EF and G to facilitate comparison. The black region is the EGF motif. The shaded region 3′ to the EGF motif is the transmembrane domain. The shaded regions 5′ to the EGF motif are signal sequences. Exon C is present in C. elegans lin-3 cDNA, but not the original genomic clone, which nonetheless provided vulval-inducing function (Hill and Sternberg 1992). Exon C2 may provide an alternative signal sequence.

lin-3 alleles can cause either tissue-specific or tissue-general defects: Seven lin-3 mutant alleles have been obtained previously in various genetic screens: e1417 and n378 in a screen for recessive mutations with vulval defects (Horvitz and Sulston 1980; Ferguson and Horvitz 1985), n1058 and n1059 in a screen for mutations that fail to complement lin-3(e1417) (Ferguson and Horvitz 1985), sy91 as a transposon-induced allele in a screen for mutations that fail to complement e1417 (Hill and Sternberg 1992), and s751 and s1263 in a screen for recessive lethal mutations (Clarket al. 1988). We carried out an additional noncomplementation screen against e1417, and we obtained sy51, sy52, and sy53. All three are recessive alleles (data not shown) and cause completely penetrant lethality when homozygous. All 10 alleles of lin-3 are recessive, loss-, or reduction-of-function alleles. sy91 has a transposon insertion in one of the introns (Hill and Sternberg 1992), and it is not clear what effect such an insertion has on LIN-3 protein. We therefore focused our study on the 9 EMS-induced alleles.

Mutations in lin-3 affect vulval induction as well as hermaphrodite fertility, male spicule development, and viability of both hermaphrodites and males. Ferguson et al. (1985) demonstrated that the lethality and Vul phenotype caused by lin-3 are loss-of-function phenotypes. The loss-of-function phenotype of lin-3 in male spicule development is an anterior-to-posterior fate transformation in the B lineage (Chamberlin and Sternberg 1993). Some lin-3 alleles cause hermaphrodite sterility. A partial description of this phenotype can be found elsewhere (Clandininet al. 1998). lin-3 is also involved in the specification of P12 neuroectoblast in the ventral posterior epidermis. Since recessive lin-3 alleles do not alter P12 fate specification unless combined with other mutations, we will not further discuss this phenotype (Jiang and Sternberg 1998).

Six alleles cause completely penetrant larval lethality: n1059, s751, s1263, sy51, sy52, and sy53 (Table 1). n1059 is a genetic null allele and causes animals to arrest at the first larval stage (L1; Ferguson and Horvitz 1985). s1263 and s751 animals arrest at L2 or L3 (Clarket al. 1988). sy51 and sy52 cause L2 to L3 lethality, and sy53 causes L1 lethality, as judged by the size of dead worms. Animals homozygous for these alleles die without undergoing the other developmental events mediated by lin-3 activity. We therefore studied the functions of these alleles in fertility and vulval and male spicule development by examining animals heterozygous for one lethal allele and a viable lin-3 allele. These animals survive but show defects in vulval development, male spicule formation, and hermaphrodite fertility, indicating that the lethal alleles also disrupt these other developmental processes (Table 1). However, there is inhomogeneity between different lethal alleles. For example, animals heterozygous for n1059 and n378 display stronger defects in fertility and vulval and spicule development, compared to animals heterozygous for n378 and any other lethal allele. This suggests that the other lethal alleles are not null. sy53 retains some activity in hermaphrodite fertility when in trans to n1058, a feature unique among the lethal alleles.

Three other alleles, e1417, n378, and n1058, cause little or no lethality. Since they disrupt only some of the developmental processes mediated by lin-3, we refer to them as having tissue-specific defects. e1417 causes defects only in vulval development, both when homozygous and when in trans to the null allele n1059 (Sulston and Horvitz 1981; Ferguson and Horvitz 1985; Table 1). Thus, e1417 appears to retain wild-type lin-3 activity in all other developmental processes. n378 is defective only in the development of vulva and male spicules, causing a severe Vul phenotype and a weak spicule defect (Ferguson and Horvitz 1985; Table 1). When in trans to n1059, the defects become more severe, but there is still no defect in viability and fertility. n1058 animals have a weak Vul phenotype and weak defects on viability, but hermaphrodites are completely sterile and males have strong spicule defects. In trans to n1059, the Vul defect and the lethality are enhanced (Ferguson and Horvitz 1985; Table 1). n1058 also exhibits intragenic complementation on vulval induction with n378 (Table 1) and e1417 (Ferguson and Horvitz 1985).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

(A). Comparison of protein sequences of C. elegans (CE3) and C. briggsae (CB3) lin-3. Letters in black boxes represent conserved amino acids. The triangles indicate intron positions. The EGF motif is underlined. The alternatively spliced region is doubly underlined. Overlined sequences are the putative cleavage sites of the transmembrane LIN-3. The transmembrane region is indicated with double overline. (B) The comparison of the alternative exon C2 in the two species. A total of 63% of the amino acids are conserved.

We examined the severity of the phenotypes caused by each lin-3 allele for four separate developmental processes. For each process, we constructed an allelic series by ranking the alleles according to the severity of their effects on that process (Figure 4). We find that the allelic series is different for different processes. For example, n1058 confers a much more severe defect in fertility than n378, but has a less severe effect in vulval induction. This suggests that not only do some alleles of lin-3 cause tissue-specific defects, but also that their functions in different developmental processes are independently mutable.

To understand how various domains of LIN-3 may affect its function, we sequenced nine EMS-induced lin-3 alleles to determine their molecular lesions. All except e1417 have mutations within the coding region. e1417 has mutations in neither the exons, nor the exon-intron junctions, nor the 5′ UTR, and the mutation may thus lie in a noncoding regulatory region.

Mutations in an extracellular region other than the EGF motif: n378 changes Glu21 to Lys, which is the first amino acid after the signal sequence in the extracellular domain (Figure 5), regardless of which signal sequence is used. In many other EGFs, the region surrounding the signal sequence is removed to form the mature protein product (reviewed by Massagué and Pandielle 1993). Therefore, it is unlikely that the mutation in n378 affects the properties of the final product of LIN-3. The effect of n378 is thus more likely to be on the LIN-3 precursor. Since a signal sequence is critical in the translocation of proteins from the cytoplasm to the plasma membrane or extracellular space, the mutation in n378 may result in a lower concentration of LIN-3 on the cell surface by affecting the structure and stability of the LIN-3 precursor.

EGF motif is essential for the function of lin-3: Four lin-3 lethal alleles have mutations in the EGF motif (Figure 5). n1059 changes Trp188 to a stop codon, which would result in a protein missing the last 17 residues of the EGF motif, including the sixth cysteine. This constitutes the molecular evidence that n1059 is a null allele. s1263, another lethal allele, changes the first cysteine in the EGF motif, Cys154, to a tyrosine. sy53 changes the last nucleotide in exon D from G to A. This nucleotide alteration could have two effects on the lin-3 protein product. First, Asp156, between the first and second cysteine, would be changed to Asn. Second, it alters a partially conserved G in front of a splice donor into an A and could make RNA splicing less efficient (Blumenthal and Steward 1997). A failure to splice out the downstream intron would result in a message encoding a protein product with an incomplete EGF motif having only the first cysteine. However, we do not know which effect is the principal cause of the phenotypes associated with this mutation. sy51 changes a conserved splice donor from gt to at. The mutation could result in a protein with a truncated EGF motif having only the first cysteine. That all four alleles cause tissue-general defects demonstrates that the EGF motif is crucial for lin-3 function in all developmental processes that require lin-3.

Why, if the EGF motif is crucial, is only one of the four alleles a null allele? There are two possible explanations. First, the position of the stop codon in n1059 is 3′ to the altered splice donor in sy51. Therefore, sy51 should produce a protein shorter than that in n1059. However, sy51 causes a less severe defect than n1059. It is possible that some splicing still occurs in sy51, resulting in a small amount of full-length protein. In fact, it has been shown that altering the conserved nucleotides of splice donors diminishes, but does not abolish, splicing in C. elegans (Rushforth and Anderson 1996; Zhang and Blumenthal 1996). Second, the alteration of Cys154 in s1263 results in severe reduction-of-function, but non-null, phenotypes (Table 1). It indicates that a full-length EGF motif containing a mutation in one of the conserved cysteines can have residual activity. This is consistent with previous findings that changing Cys164 to serine, or changing Cys184 to serine, does not completely abolish vulval induction (Hill and Sternberg 1992).

sy53 causes a lethal phenotype as severe as that caused by n1059, but retains some activity in fertility, vulval induction, and male spicule development. It is not clear why a change in the EGF motif would cause total loss of activity in one developmental process but partial loss in others.

The cytoplasmic domain is important for both tissue-general and tissue-specific functions of lin-3: The cytoplasmic domain of LIN-3 (185 residues) is longer than that of most other growth factors that have transmembrane precursors (reviewed by Massagué and Pandielle 1993). Three alleles, n1058, sy52, and s751, have mutations in this domain.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

(A) A summary of phenotypes caused by lin-3 alleles. +, wild-type; (+), (+/−), (−), and −, four levels of activity, each level lower than the preceding one, with (+) closest to wild-type activity. (B) Phenotypic series of lin-3 alleles. The four series were established on the basis of the data in Table 1. Alleles to the left have more wild-type activity than the ones to the right of each row. In the series of viability, n1059 and sy53 were determined to have less activity than the rest of the lethal alleles, since they both cause lethality at L1 stage, while the rest of the lethal alleles cause worms to die at the L2-L3 stages. In the series for vulval induction and spicule development, defects caused by the lethal alleles were assessed when the lethal alleles were in trans to n378. Since sy52 and s751 have the same molecular lesion, only the phenotypes of sy52 were analyzed.

n1058 disrupts a conserved splice donor at the 3′ end of exon I (Figures 2 and 5). It could make a protein that has only the N-terminal 94 residues of the cytoplasmic domain, but could also make a small amount of full-length protein if, as suggested for sy51, splicing is only diminished, but not abolished. Both sy52 and s751, isolated in two laboratories, disrupt the splice acceptor of exon I (Figures 2 and 5). It may make a protein that lacks 92 residues at the N terminus of the cytoplasmic domain. It is also possible that other mutant protein products are made by mRNAs spliced at cryptic splice sites close to the mutated one (Aroianet al. 1993). sy52 and s751 cause tissue-general defects.

The role of lin-3 in animal viability: We tested whether the let-23/let-60 pathway, which mediates vulval induction, also mediates lin-3 function in viability. A gain-of-function (gf) let-60 mutation can rescue the lethality caused by let-60 dominant-negative (dn) mutations and a let-23 null mutation (Aroianet al. 1990; Beitelet al. 1990; Hanet al. 1990). Similarly, a let-23(gf) mutation rescues the lethality caused by a let-23 null allele (Katzet al. 1996). However, let-23(gf) and let-60(gf) failed to rescue the lethality caused by lin-3(lf). In the progeny of let-23(gf) unc-4; lin-3(sy51) dpy-20/+ hermaphrodites, we have almost never seen adult Unc Dpy animals, which should be let-23(gf) unc-4; lin-3(sy51) dpy-20 (n > 500). Instead, we have seen dead larvae. Similar results were obtained when we used lin-3(n1059) in the experiment. We only occasionally saw Dpy animals in the progeny of lin-3(n1059) let-60(gf) dpy-20/lfe-1 unc-24 (n > 500). This failure to suppress lin-3 lethality is unlikely to result from let-23(gf) and let-60(gf) only weakly activating the let-23/let-60 pathway, since both are able to rescue the lethality caused by a let-23 null allele. It is possible that the let-23/let-60 pathway is not sufficient to mediate lin-3 signaling in viability, and a different downstream pathway may be required. We then tested whether lin-1(rf) is able to suppress the lethality caused by lin-3(n1059). A total of 10% of unc-24 lin-3 dpy-20/lin-1 animals segregated recombinant lin-1 unc-24 lin-3 dpy-20 (n = 120), which grew to sterile adults. If lin-1 does not suppress lin-3 lethality, no lin-1 lin-3 animals should grow to adulthood. If lin-1 completely suppresses lin-3 lethality, all lin-1 lin-3 animals should grow to reach adulthood. The number of adult lin-1 lin-3 recombinants is consistent with the null hypothesis that lin-1 lin-3 animals are generated from recombination and that lin-1 completely suppresses lin-3 lethality (P = 0.547, Fisher's exact test); therefore, we concluded that lin-1 almost fully suppresses the lethality of lin-3. Therefore, if there is an alternative pathway mediating the function of lin-3 in viability, this pathway should also repress lin-1. An alternative possibility is that lin-3 mediates animal viability using one of the two previously identified pathways, but there are different regulatory mechanisms for the pathway in different developmental processes (N. Hopper, personal communication).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Molecular lesions of lin-3. Codon changes and changes in the conserved splice sites are listed for all lin-3 alleles except e1417 and sy91, which have no change in the coding sequences and exon-intron junctions. Only exons C to K are shown, since there are no mutations in exons A, B, and L. sy91 has a Tc1 insertion in the intron after exon C2 (see Figures 2 and 3). The downward arrows indicate intron positions, and the upward arrows indicate alternative splice sites. The underlined sequences are amino acids, and the nonunderlined sequences are nucleotides. Uppercase letters represent exon sequences, and lowercase letters represent intron sequences.