A Cuticle Collagen Encoded by the lon-3 Gene May Be a Target of TGF-β Signaling in Determining Caenorhabditis elegans Body Shape
Yo Suzuki, Gail A. Morris, Min Han, William B. Wood

Abstract

The signaling pathway initiated by the TGF-β family member DBL-1 in Caenorhabditis elegans controls body shape in a dose-dependent manner. Loss-of-function (lf) mutations in the dbl-1 gene cause a short, small body (Sma phenotype), whereas overexpression of dbl-1 causes a long body (Lon phenotype). To understand the cellular mechanisms underlying these phenotypes, we have isolated suppressors of the Sma phenotype resulting from a dbl-1(lf) mutation. Two of these suppressors are mutations in the lon-3 gene, of which four additional alleles are known. We show that lon-3 encodes a collagen that is a component of the C. elegans cuticle. Genetic and reporter-gene expression analyses suggest that lon-3 is involved in determination of body shape and is post-transcriptionally regulated by the dbl-1 pathway. These results support the possibility that TGF-β signaling controls C. elegans body shape by regulating cuticle composition.

HOW do animals dictate the correct proportional sizes for their body parts? Although factors such as growth hormone are known to play pivotal roles, genetic regulation of body size and body shape is still poorly understood (reviewed in Conlon and Raff 1999). Recent findings in Drosophila suggest that factors involved in size regulation, for example, insulin-like growth factor signaling, are conserved among metazoans (reviewed in Edgar 1999), validating the utility of studying model organisms for better understanding of this process.

In Caenorhabditis elegans, a TGF-β signaling pathway has been implicated in regulation of body size and body shape. Loss-of-function (lf) mutations in the dbl-1 gene, which encodes the TGF-β family ligand DBL-1, or in genes encoding the receptors and Smad transcription factors of the dbl-1 pathway cause a Sma phenotype (reviewed in Patterson and Padgett 2000). Conversely, hyperactivation of this pathway (e.g., by means of dbl-1 overexpression) causes a Lon phenotype (Moritaet al. 1999; Suzukiet al. 1999). Mutations in three genes, lon-1, lon-2, and lon-3, cause a similar Lon phenotype (Brenner 1974), suggesting that these genes have related functions in body morphology regulation. However, the cellular mechanisms controlled by the dbl-1 pathway are largely unknown.

Other mutations that cause changes in body shape are those in the dpy and sqt genes, many of which have been shown to encode cuticle collagens (e.g., Krameret al. 1988; von Mendeet al. 1988). These mutations cause shorter body length, and a few alleles and allelic combinations result in Lon phenotypes (Kusch and Edgar 1986; reviewed in Kramer 1994). Dpy and Sqt mutant animals appear short and fat but generally no thinner than wild type, in contrast to Sma animals, which are also short but more normally proportioned. Whether Dpy and Sqt cuticle defects are related to changes in activity of the dbl-1 pathway was previously unclear.

Here we present evidence suggesting that dbl-1 regulation of body shape may be partly accomplished by a cellular mechanism that involves regulation of cuticle composition. To identify possible regulatory targets of the dbl-1 pathway, we screened for suppressors of the Sma phenotype caused by a dbl-1 mutation and found two that are alleles of the lon-3 gene. We show that lon-3 encodes a collagen that is a component of the C. elegans cuticle and that it may be regulated post-transcriptionally by the dbl-1 pathway.

MATERIALS AND METHODS

C. elegans strains: All strains were cultured using standard methods (Brenner 1974). The alleles used in these experiments (all in the N2 wild-type background) were as follows, in map order by linkage group (LG). Unless otherwise noted, the references for these alleles are found in Riddle et al. (1997). LGI: cye-1(ku256) (Fay and Han 2000), unc-13(e450); LGII: sma-6(wk7) (Krishnaet al. 1999), dpy-2(e8) and (e1359), sqt-1(sc103) and (sc101); LGIII: lon-1(e185), sma-4(e729), unc-119(ed3); LGIV: dpy-9(e12), him-8(e1489); LGV: unc-46(e177), dbl-1(ev580) and (wk70) (Suzukiet al. 1999), dbl-1(nk3) (obtained from Moritaet al. 1999), ctDf1, sma-1(e30), vab-8(e1017), sqt-3(sc63ts), lon-3(e2175), lon-3(sp5), (sp6), and (sp23) (obtained from A. Leroi, personal communication), lin-25(e1446), him-5(e1467), unc-76(e911); LGX: dpy-7(e88), lon-2(e678). The small deficiency ctDf1 uncovers sma-1, vab-8, and sqt-3 (Manser and Wood 1990). Except as noted above, all alleles were obtained from the Caenorhabditis Genetics Center or from our collection. The dbl-1 overexpression array kuEx136 was generated by injecting a dbl-1 genomic clone, pYS1 (Suzukiet al. 1999), at 100 ng/μl with no other markers and picking Lon progeny. The transmission frequency of this array was ∼100%.

Suppressor screens and complementation tests: dbl-1(ev580) animals (Sma) were mutagenized with EMS using a standard protocol (Brenner 1974). On the basis of the number of F1 progeny, we screened an estimated 14,000 haploid genomes. Suppressed animals, either non-Sma or Lon, were isolated in the F2 generation.

Suppressed animals were crossed with N2 males to generate males heterozygous for the new mutations as well as dbl-1 (ev580); these heterozygotes were crossed with lon-1(e185) or lon-3(e2175) homozygous hermaphrodites. The presence of Lon male progeny indicated noncomplementation. Some of the suppressed animals, upon crossing with N2 males, produced Lon males but no Lon hermaphrodites, indicating that these mutants contain recessive, X-linked lon mutations.

Phenotypic and genetic analysis of lon-3 mutants: Unless indicated otherwise, body size phenotypes were analyzed only by direct observation using a dissecting microscope. Measurements of body length were performed using Scale Master II (Calculated Industries) on Nomarski images captured and printed using Scion Image (Scion, Frederick, MD).

To test if lon-3 mutations are dominant or recessive, N2 males were crossed with lon-3(e2175) hermaphrodites, and these animals were transferred daily; the second plate contained 173 wild-type (WT) males, 180 WT hermaphrodites, and 10 Lon hermaphrodites (presumably self-progeny). WT hermaphrodites from this cross were indistinguishable when compared side by side with N2 wild-type hermaphrodites of similar stages. Similar crosses of N2 males to other lon-3 mutants (hermaphrodite) also produced WT progeny, except one mutant, ct418, which produced only semi-Lon (cross) and Lon (self) progeny. These results indicate that e2175, ct417, sp5, sp6, and sp23 are recessive and that ct418 is semidominant.

RNAi experiments were performed as described (Fireet al. 1998). A T7 polymerase promoter was attached to both ends of the exon 1 sequence using PCR with primers CMo42 and CMo442 (C. Mello, personal communication) and with the template pYSL3C1 (see below). Double-stranded (ds) RNA was prepared using the MEGAscript T7 kit (Ambion, Austin, TX). In addition to the Lon phenotype, lon-3(RNAi) resulted in a low level (∼5%) of late embryonic arrest. We did not carefully characterize this possibly nonspecific effect. The effectiveness of lon-3 RNAi was demonstrated when injection of lon-3 dsRNA abolished the green fluorescent protein (GFP) expression from the translational fusion construct pYSL3G3 (see below), as well as the Rol phenotype caused by the same construct, and caused a Lon phenotype in injected animals.

Cloning of lon-3: Cosmids, C27A7, ZK863, and ZK836, and subclones, pYSL3S1 and pYSL3S2 (see below), were injected at 20 ng/μl into the lon-3(e2175) mutant with 100 ng/μl of pTG96 (a GFP reporter plasmid; Yochemet al. 1998). pYSL3S2 contained only one predicted intact open reading frame (WormBase; http://www.wormbase.org).

The initiator methionine predicted in WormBase appears to be the bona fide initiator on the basis of our finding that five sequenced 5′-RACE products terminated at 24, 35, 36 (two independent PCR products), and 37 bp before the predicted initiator methionine. The exon-intron structure was determined using PCR of reverse transcribed samples.

The lon-3 cDNA sequence has been submitted to GenBank (accession no. AF465981).

Plasmid construction: An 11-kb PstI-SalI fragment from the cosmid ZK863 was cloned into pBluescript II SK+ (Stratagene, La Jolla, CA) to generate pYSL3S1. A 5.5-kb BamHI-XhoI fragment from pYSL3S1 was cloned into pBluescript II SK+ to generate pYSL3S2. This construct contains 2.5 kb of the 5′ sequence, the complete coding sequence including the sole intron, and 2 kb of the 3′ sequence. A 2.5-kb BamHI-HaeIII fragment from pYSL3S2 containing ∼2.5 kb of the lon-3 5′ sequence and the first five codons of the lon-3 coding sequence was cloned into the BamHI and SmaI sites of the GFP vector pPD95.67 (A. Fire, personal communication) to generate the “transcriptional” reporter gene pYSL3G1. The “translational” reporter gene pYSL3G3 had the complete 5′ and coding sequences present in pYSL3S2, followed by a GFP sequence, the last eight codons repeated of the lon-3 coding sequence, the native lon-3 stop codon, and the complete 3′ sequence of pYSL3S2. To generate this plasmid, PCR with primer sequences, TGGTGGGC-GGCGCC (NarI)-GTAAGCACAAGA GAGTCTAC (the end of the lon-3 coding sequence)-ATGAG TAAAGGAGAAGAACTTTTCAC (the beginning of the GFP sequence) and TGGTGGGC-GGCGCC (NarI)-GATCCCCCG GGCATTTGTATAG (the end of the GFP sequence), was used to generate a fragment that had the GFP sequence connected to the last part, from the NarI site to the end, of the lon-3 coding sequence. This fragment flanked by NarI sites was cloned into the unique NarI site of pYSL3S2. For RNAi experiments, an 840-bp PCR fragment flanked by BamHI sites, containing only the complete exon one sequence, was cloned into pBluescript II SK+ to generate pYSL3C1. The portions of these plasmids amplified by PCR were sequenced to confirm that no errors were introduced by PCR.

Lesions of lon-3 alleles: PCR fragments amplified from genomic DNA samples isolated from lon-3 mutants were directly sequenced. Lysis of animals was performed for 2 hr at 60° in 10 mm Tris-HCl (pH 8.3), 50 mm KCl, 2.5 mm MgCl2, 0.45% Tween-20, 0.05% gelatin, and 500 μg/ml proteinase K. Sequencing of the coding region revealed the lesions of sp6 (R114 [CGA] → stop [TGA]), sp23 (W36 [TGG] → stop [TAG]), ct417 (W36 [TGG] → stop [TAG]), and ct418 (G61 [GGA] → E [GAA]). The lesion for sp5 was not found in this region. For e2175, which contained a gross aberration to the lon-3 sequence, a PCR fragment that spanned the lesion was identified and sequenced. e2175 lacked part of the lon-3 sequence (-2001 bp to +388 bp in relation to the initiator ATG codon) and contained part of the transposon Tc5 (1 bp to 2779 bp; see Collins and Anderson 1994) in a left-to-right orientation.

Expression of lon-3: For reporter expression analyses, pYSL-3G1 and pYSL3G3 were injected at 100 and 20 ng/μl, respectively, to unc-119; him-5 animals with 30 ng/μl of pDP#MM 016B (an unc-119 rescuing plasmid; Maduro and Pilgrim 1995). Some of these lon-3::gfp arrays were integrated using 3000 R of gamma irradiation, followed by six backcrosses with unc-119; him-5 and N2. The lon-3::gfp expression patterns did not change before and after integration. The integrated arrays kuIs56 (pYSL3G1) and kuIs55 (pYSL3G3) were used to assess the effects of various mutations on lon-3 expression.

Probably due to the punctate nature of the lon-3 translational reporter-gene expression patterns, we were unable to analyze the pixel intensity of its digitized photographs. To compare the intensity of expression in different mutant backgrounds, we instead relied on “by eye” scoring. Unless otherwise noted, individual L4 animals with the identical Christmas-tree vulval structure were mounted on a slide and photographed with shortest exposure possible to UV to minimize photobleaching of GFP. These photographs were given unlabeled (blind scoring) to a scorer to be evaluated by eye and rated for fluorescence intensity on a scale of one to six (in the case of the translational construct) or of one to five (in the case of the transcriptional construct), where one represents the lowest intensity.

Figure 1.

—Lon phenotypes resulting from a lon-3 mutation and from dbl-1 overexpression. (A) N2 wild-type hermaphrodite. (B) lon-3(ct417) mutant hermaphrodite. (C) Hermaphrodite overexpressing dbl-1 from kuEx136 (see materials and methods). These photographs show adult animals grown for 52 hr posthatching at ∼25°. Bar, 0.1 mm.

For the lon-3 transcriptional reporter gene (pYSL3G1), additional methods were used. Progeny from dbl-1(ev580) heterozygotes that carried an integrated array of this construct were scored for body size and fluorescence intensity under a dissecting microscope equipped with UV epi-illumination. Although some variation in fluorescence was observed in these animals, there was no correlation between the fluorescence intensity and the Sma or WT body size phenotypes of these animals. Furthermore, when WT and Sma progeny were compared quantitatively using Scion Image, no significant difference in brightness was detected (data not shown).

RESULTS

Suppressors of the Sma phenotype caused by a dbl-1(lf) mutation define lon-3: To understand the mechanisms downstream of dbl-1 signaling in body shape regulation, we screened 14,000 haploid genomes for suppressors of the Sma phenotype caused by a dbl-1 mutation (see materials and methods). We isolated 14 mutations, of which 2 that produced Lon phenotypes were alleles of lon-1 and have been reported previously (Moritaet al. 2002; see also Maduziaet al. 2002). The remaining mutations, which produced non-Sma phenotypes, included two alleles of lon-3, which were designated ct417 and ct418 (see materials and methods). In addition to these new lon-3 alleles and the reference allele e2175 (Brenner 1974) used for complementation tests, three other alleles, sp5, sp6, and sp23, have been found in the lon-3 gene (A. Leroi, personal communication).

lon-3 is required for body length restriction: All lon-3 alleles, when homozygous, caused Lon phenotypes similar to each other and to the Lon phenotype of dbl-1 overexpressing animals (Figure 1; the adult column in Table 1). In lon-3 animals Lon phenotypes were not seen in early larval stages but became evident by the adult stage (Table 1). In contrast, dbl-1 overexpressing animals were Lon in the L3 stage as well as in the adult stage (Table 1).

View this table:
TABLE 1

Comparison of body lengths in lon-3 mutant and dbl-1 overexpressing animals

The lon-3(ct418) mutation was semidominant, whereas other alleles were apparently recessive (see materials and methods). RNAi of lon-3 causes the same Lon phenotype in wild-type animals and does not enhance the Lon phenotype in a lon-3 mutant (Table 2), suggesting that the Lon phenotype is the null body shape phenotype of lon-3. No additional phenotypes such as embryonic lethality or male tail phenotypes were observed in lon-3 homozygotes or hemizygotes (data not shown).

lon-3 mutations are only partially epistatic to dbl-1 pathway mutations: Since lon-3 alleles were isolated as suppressors of the Sma phenotype caused by a dbl-1 mutation, lon-3 may function downstream of or in parallel to the dbl-1 pathway. However, animals carrying null mutations for dbl-1 (nk3; Moritaet al. 1999) and lon-3 (ct417; see below for molecular lesion) had an intermediate phenotype (Table 2). One possible interpretation of this result is that lon-3 mediates only a subset of dbl-1 functions.

lon-3 encodes a C. elegans cuticle collagen: lon-3 was previously mapped to LGV (Brenner 1974) and shown to be uncovered by chromosomal deletions ctDf1 (Manser and Wood 1990) and arDf1 (see Sundaram and Greenwald 1993). Using a series of three-factor mapping crosses with genetic markers uncovered by these deletions (Table 3), we mapped lon-3 to a small region between sma-1 and vab-8 (Figure 2). The Lon phenotype of e2175 could be rescued by either of two overlapping cosmids (ZK863 and ZK836) in this region and by subclones containing the overlapping sequence. The only intact predicted open reading frame (ZK836.1) in the smaller of these subclones, pYSL3S2, encodes a protein with characteristics of C. elegans cuticle collagens. On the basis of the positions of cysteines (reviewed in Kramer 1997), the predicted LON-3 protein belongs to the SQT-1 subgroup. The exon-intron structure was determined using PCR of reverse transcribed samples. We did not detect a splice leader, SL1 or SL2, using 5′-RACE (see materials and methods) or RT-PCR with SL-specific primers (data not shown), suggesting that the lon-3 transcript is not transspliced. [While this article was in revision, similar results on cloning and identification of lon-3 were reported (Nystromet al. 2002).]

View this table:
TABLE 2

Body lengths of animals defective in lon-3, dbl-1, or both

View this table:
TABLE 3

Three-factor mapping of lon-3

Molecular lesions were found in five out of six lon-3 alleles, indicating that lon-3 encodes this collagen protein (Figure 3; see materials and methods). sp6, sp23, and ct417 are nonsense mutations upstream of the sequence encoding the Gly-X-Y domain, which is expected to generate the collagen triple helix. Therefore, these alleles are likely to be severe loss-of-function or null mutations. Since ct418 is semidominant, the G-to-E substitution it is predicted to cause may create a dominant negative form of LON-3. The reference allele e2175 lacks 2.4 kb of the lon-3 sequence and has an insertion of part of the transposon Tc5. No lesion in the coding sequence was found for sp5, suggesting that this mutation could be in a lon-3 regulatory sequence.

Reporter-gene analysis suggests that lon-3 encodes a cuticle component: We determined the expression of an integrated reporter gene that contained the entire 2.5-kb 5′ sequence present in the rescuing clone pYSL3S2 and the first five amino acids including the initiator methionine connected to the GFP sequence with an SV40 nuclear localization signal (the transcriptional reporter gene pYSL3G1; see materials and methods). GFP fluorescence was localized in the nuclei of most, if not all, hypodermal cells (Figure 4A). Expression was first detected in the pretzel stage just prior to hatching. The GFP signal gradually intensified toward the young adult stage and then diminished.

To test experimentally the possibility that lon-3 encodes a cuticle component, we examined the expression patterns of another reporter gene that contained the entire coding sequence, in addition to the 5′ and 3′ sequences present in pYSL3S2 (the translational reporter gene pYSL3G3; see materials and methods). GFP fluorescence was faint, but clearly observable on the surface of the larval and adult animal, presumably in the cuticle (Figure 4B). This reporter gene did not rescue the Lon phenotype of lon-3 mutants, but instead caused an adult Rol (Roller) phenotype both in lon-3 mutants and in N2 wild type. These results suggest the cuticular localization of the reporter-gene product.

lon-3 appears to be post-transcriptionally regulated by the dbl-1 pathway: To test the possibility that dbl-1 affects lon-3 expression, we examined the GFP expression from the lon-3 translational reporter gene (pYSL3G3) in various mutant backgrounds. When we placed the integrated array (kuIs55) containing this reporter gene in dbl-1 mutant backgrounds, the GFP expression appeared to be increased (Figure 4, C and D). To more quantitatively assess the effects of dbl-1 and other pathway genes on the GFP expression, we rated the GFP expression intensities by eye using blind scoring (see materials and methods). In these experiments, the GFP expression was upregulated in dbl-1 and sma-4 mutants and was downregulated in dbl-1 overexpressing animals (Table 4; Figure 4, E-G). The Rol phenotypes of these mutant animals were indistinguishable under the dissecting microscope. These animals apparently lacked alae in the adult stage, so that we could not use these structures to count the number of twists of the body. This upregulation of reporter expression in dbl-1 and sma-4 mutants is not likely to result from a higher concentration of the GFP signal simply due to the small body size of dbl-1 mutants, since mutations such as dpy-2(e8), dpy-7(e88) (Table 4), unc-46(e177), and dpy-9(e12) (data not shown), which also cause smaller body size, did not affect GFP expression from pYSL3G3.

In contrast, when we compared GFP expression from the lon-3 transcriptional reporter gene (pYSL3G1) in wild-type and dbl-1(lf) backgrounds, we observed no differences in the intensity of GFP expression, suggesting that lon-3 transcription is not affected by the dbl-1 pathway (Table 5; see materials and methods). These results suggest that the dbl-1 pathway regulates lon-3 expression post-transcriptionally, presumably by modulating processing of lon-3 mRNA, degradation of lon-3 mRNA or the LON-3 protein, or synthesis of the LON-3 protein.

Figure 2.

—Cloning of lon-3. Genetic and physical maps of the lon-3 region are shown (see text). pYSL3S1 and pYSL3S2 are subclones of ZK863 (see materials and methods). Presumably due to the overexpression of lon-3, some rescued animals were shorter than wild type.

Figure 3.

—Structure of the predicted LON-3 protein. Homology blocks (see text) are shown in gray (Kramer 1997). The Gly-X-Y domain with a few gaps (white) is shown in black. Molecular lesions corresponding to five lon-3 alleles are indicated by arrows (see materials and methods). The lon-3 cDNA sequence is found in GenBank (accession no. AF465981).

lon-3 GFP expression is affected by dbl-1 pathway and some cuticle-related mutations, but not by mutations that cause altered ploidy: To understand the relationship between lon-3 and other lon genes, we placed the lon-3 translational reporter gene in various lon mutation backgrounds. A lon-2 mutation downregulated the GFP expression, whereas a lon-1 mutation or the Lon allele (sc101) of sqt-1 (Kramer and Johnson 1993) had no effect (Table 4). This result is interesting because lon-2 is thought to function upstream of dbl-1 (Y. Suzuki and W. B. Wood, unpublished observations; S. Baird, C. Savage-Dunn and R. Padgett, personal communication). From this result, lon-1 and sqt-1 appear to be unrelated or to function downstream of or in parallel to lon-3. lon-1 is a dose-dependent regulator of adult hypodermal ploidy (Moritaet al. 2002). To see if ploidy change has any effect on lon-3 expression, we used a cyclin E mutation that is known to reduce hypodermal ploidy at least in older adults (A. Flemming, A. Leroi and D. Fay, personal communication). We did not see any difference in lon-3 expression in adults of a cye-1; unc-13 strain relative to adults of the unc-13 control (Table 4).

Figure 4.

—Expression patterns of lon-3::gfp reporter transgenes. Integrated lines carrying lon-3::gfp reporter genes were generated as described in materials and methods. (A) L4 hermaphrodite carrying a transcriptional reporter gene with an SV40 nuclear localization signal (pYSL3G1) as an integrated array (kuIs56). GFP is localized to the nuclei of hypodermal cells; many are out of focus. Asterisks indicate nuclei of lateral hypodermal cells (seam cells). (B) Head region of an adult hermaphrodite carrying a translational fusion construct (pYSL3G3) as an integrated array (kuIs55). Note the GFP expression outlining the animal. (C-G) Comparison of GFP expression levels from kuIs55 in dbl-1(+), dbl-1(lf), and dbl-1 overexpression (kuEx136) genetic backgrounds. These animals are at precisely the same stage based on vulval morphology. C and D show dbl-1(+) and dbl-1(wk70) animals. In D, the contrast was adjusted to visualize the low level of GFP expression in the dbl-1(+) animal. E-G are three of the photographs used to generate Table 4, showing animals with close to the average rating of GFP expression among the dbl-1(+), dbl-1(nk3), and kuEx136 groups, respectively. The same exposure setting was used for all the images in E-G with no contrast adjustment. Light spots in E-G are caused by autofluorescence of gut granules and are unrelated to lon-3 expression.

To uncover possible interactions among collagens in the cuticle, we placed the lon-3 translational reporter gene in sqt-1, dpy-2, and dpy-7 mutation backgrounds. The null allele (sc103) of sqt-1 (Kramer and Johnson 1993) slightly enhanced the GFP expression although the Lon allele (sc101) of sqt-1 appeared to have no effect (Table 4). Allele specificity was also observed with dpy-2, where the e8 allele had no effect, while the e1359 allele caused a dramatic reduction in GFP expression. dpy-2(e8) and dpy-7(e88) cause substitutions for Gly in the Gly-X-Y domain (Johnstoneet al. 1992; Levyet al. 1993), whereas the lesion of dpy-2(e1359) is not known. The alteration of GFP expression by collagen mutations supports the hypothesis that lon-3 product is present in the cuticle.

DISCUSSION

Cuticle modeling may be a target of the TGF-β (dbl-1) pathway in C. elegans body shape control: We have identified mutations in lon-3 as suppressors of the Sma phenotype caused by a dbl-1 mutation and shown that lon-3 encodes a C. elegans cuticle-type collagen. On the basis of our studies of its expression patterns, lon-3 appears to be a cuticle component. GFP from the lon-3 transcriptional fusion construct is observed in most if not all hypodermal nuclei, and GFP from the lon-3 translational fusion construct is observed on the surface of the animals, presumably in the cuticle. The translational construct contains additional sequences not present in the transcriptional construct, such as the sole intron and the 3′ sequence. These sequences may have caused differences in transcription of the two reporters and could account for our observation that expression from the transcriptional construct is more robust than that from the translational construct. Results obtained with the translational reporter must be interpreted with care because it does not rescue the Lon phenotype of lon-3 mutants, despite including all the sequences that are present in the rescuing clone.

With this caveat, changes of lon-3 reporter-gene expression patterns in dbl-1 mutants seen with the translational construct, but not with the transcriptional construct, suggest that either processing or degradation of the lon-3 mRNA, or synthesis, processing, or degradation of the LON-3 protein, is regulated by the dbl-1 pathway. This regulation of lon-3 by the dbl-1 pathway may be direct or indirect. We have found that lon-3 expression is affected in an allele-specific manner by sqt-1 and dpy-2. It has been shown recently that lon-3 mutations interact genetically with sqt-1 and rol-6 (Nystromet al. 2002). It is equally plausible that these collagen genes are the direct targets of the dbl-1 pathway and encode binding partners of the LON-3 protein, affecting the stability or incorporation of the LON-3 protein in the cuticle. Body shape phenotypes (e.g., Dpy phenotype) caused by mutations in previously known cuticle component genes appear to be slightly different from the Sma phenotype caused by dbl-1 pathway mutations (see Introduction). Our results provide the first evidence of a link between the cuticle and the dbl-1 pathway.

View this table:
TABLE 4

The intensity of GFP expression from a translational fusion construct

To further evaluate whether lon-3 represents a true downstream target gene, the mechanism responsible for its regulation (e.g., protein degradation, translational control, and mRNA processing) must be clarified. A microarray study has found that a predicted collagenase gene is downregulated in a dbl-1 mutant, suggesting the possibility that LON-3 is controlled by the dbl-1 pathway through this collagenase (Mochiiet al. 1999). Our preliminary observations indicate, however, that silencing of this collagenase gene using RNAi does not cause a Sma phenotype or an increase in reporter expression from the lon-3 translational fusion construct (Y. Suzuki and W. B. Wood, unpublished results). Because some collagens encoded by the dpy genes are presumably involved in functions opposite to those of lon-3, the regulation of collagen turnover may be complex. Testing the effects of inactivating other metalloproteinase genes and characterization of other suppressors isolated in our genetic screen may help to elucidate the mechanisms of lon-3 regulation.

Arguments for a branched dbl-1 pathway with multiple regulatory targets: Earlier studies have shown that the dbl-1 pathway is involved in the control of body shape and cell size (Moritaet al. 1999; Suzukiet al. 1999), as well as the ploidy of normally polyploid hypodermal cells (Flemminget al. 2000), but the relationships between these effects have been unclear. Mutations in dbl-1 pathway genes cause a reduction in body size and hypodermal nuclear ploidy in adults (Flemminget al. 2000), whereas mutations in the lon-1 gene cause an increase in body length and ploidy (Moritaet al. 2002). However, it is unclear whether the effect of these mutations is primary or secondary. For example, it is possible that cuticle defects or other structural changes resulting from these mutations affect cell size, and cell size only secondarily affects ploidy. Consistent with this possibility are the findings that dbl-1 overexpression does not appear to cause increase in ploidy (Nystromet al. 2002) and that only a subset of lon-1 alleles affects ploidy, although all cause Lon phenotypes (Moritaet al. 2002). Morita et al. (2002) have presented evidence that lon-1 is regulated by the dbl-1 pathway.

View this table:
TABLE 5

Intensity of GFP expression from a transcriptional fusion construct

Our results together with the above findings suggest a branched pathway in which lon-3 functions downstream of the dbl-1 pathway as one of perhaps several independent targets of dbl-1 signaling (Figure 5). Another of these targets may be lon-1. This model is based on the following arguments:

  1. lon-3 expression is affected in opposite ways by dbl-1(lf) mutations and dbl-1 overexpression, suggesting that lon-3 does not act by an independent pathway to determine body shape. A lon-2 mutation, like dbl-1 overexpression, decreases lon-3 expression (Table 4), consistent with lon-2 acting as an upstream regulator of the dbl-1 pathway.

  2. However, double mutants carrying both lon-3 and dbl-1 mutations are neither Lon nor Sma, but exhibit an intermediate phenotype, suggesting that lon-3 acts in only one branch of the dbl-1 pathway.

  3. Branching is likely to occur downstream of Smads because both dbl-1 and sma-4, which encodes the common Smad transcription factor of the Smad complexes, affect lon-3 expression patterns similarly (Table 4).

  4. Effects of various mutations on lon-3 expression and ploidy, which may be controlled by lon-1, appear to be independent events, suggesting independent branches of the pathway. Mutations in lon-3 affect body shape but not ploidy (Nystromet al. 2002), and mutations that cause changes in ploidy do not appear to affect lon-3 expression (Table 4). Mutations in lon-1 affect body shape but do not appear to affect lon-3 expression (Table 4). Finally, the effects of lon-1 and lon-3 mutations on body size are additive: lon-1; lon-3 double mutants are longer than either of the single mutants (Nystromet al. 2002). Thus independent downstream branches could include one for lon-3-mediated control of cuticle composition and another for lon-1-mediated control of ploidy (Figure 5).

Figure 5.

—A model for regulation of body size control by the dbl-1 pathway. LON-3 may represent a branch pathway specific to late larval elongation. This regulation is likely to be post-transcriptional. LON-1 controls ploidy of adult hypodermal nuclei and is also a candidate target of the dbl-1 pathway. Additional unknown targets may mediate early larval elongation, which the dbl-1 pathway also appears to control.

A third branch, mediated by unknown proteins, would be consistent with the finding that dbl-1 pathway mutations affect body size from the L2 stage onward, while lon-3 effects are seen only in adults (Tables 1 and 2). This branch may employ lon-1 or perhaps additional collagen genes controlled by the dbl-1 pathway (Figure 5). The collagen genes we and others (Nystromet al. 2002) have shown to interact with lon-3 are candidates for such genes. If they are controlled at the level of transcription, microarray analysis may be a fruitful approach to identifying them among the ∼150 predicted C. elegans collagen genes (reviewed in Johnstone 2000).

Extracellular matrix and growth: Various collagen genes are targets of TGF-β regulation in mammals (reviewed in Massague 1990), although the mammalian role of TGF-β is to activate collagen expression and not to repress it as in the case of lon-3 regulation by dbl-1. At least for lon-3 (other collagen genes have not been tested), dbl-1 does not appear to affect transcription, which is unlike the mechanism of mammalian collagen regulation by TGF-β; hence these interactions may not be conserved.

In mammals, production of collagens signifies a differentiated state for cells, when growth is inhibited. The importance of collagen remodeling for tissue growth has been demonstrated by mice mutated for a matrix metalloproteinase gene. These mice exhibit dwarfism among other tissue remodeling defects (Holmbecket al. 1999). The situation appears to be similar in C. elegans, where LON-3 collagen production imposes restrictions on body elongation. Therefore, C. elegans may be a promising model system for studying mechanisms of growth and remodeling of extracellular matrix, as well as relationships between them.

Acknowledgments

We thank A. Leroi for lon-3 alleles (sp5, sp6, and sp23), A. Fire for GFP vectors, The Caenorhabditis Genetics Center for C. elegans mutant strains, F. Wang and A. Keith for technical assistance, and members of the Boulder/Denver C. elegans community for helpful discussions. This work was supported by National Institutes of Health grant HD-14958 to W.B.W.

Footnotes

  • Communicating editor: B. J. Meyer

  • Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession no. AF465981.

  • Received July 30, 2002.
  • Accepted September 6, 2002.

LITERATURE CITED

View Abstract