Increased or Decreased Levels of Caenorhabditis elegans lon-3, a Gene Encoding a Collagen, Cause Reciprocal Changes in Body Length

Corresponding author: Simon Tuck, Umeå University, Lasarettsområdet, Byggnad 6L, SE-901 87 Umeå, Sweden., simon.tuck{at}ucmp.umu.se (E-mail)

Communicating editor: B. J. MEYER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Body length in C. elegans is regulated by a member of the TGFß family, DBL-1. Loss-of-function mutations in dbl-1, or in genes encoding components of the signaling pathway it activates, cause worms to be shorter than wild type and slightly thinner (Sma). Overexpression of dbl-1 confers the Lon phenotype characterized by an increase in body length. We show here that loss-of-function mutations in dbl-1 and lon-1, respectively, cause a decrease or increase in the ploidy of nuclei in the hypodermal syncytial cell, hyp7. To learn more about the regulation of body length in C. elegans we carried out a genetic screen for new mutations causing a Lon phenotype. We report here the cloning and characterization of lon-3. lon-3 is shown to encode a putative cuticle collagen that is expressed in hypodermal cells. We show that, whereas putative null mutations in lon-3 (or reduction of lon-3 activity by RNAi) causes a Lon phenotype, increasing lon-3 gene copy number causes a marked reduction in body length. Morphometric analyses indicate that the lon-3 loss-of-function phenotype resembles that caused by overexpression of dbl-1. Furthermore, phenotypes caused by defects in dbl-1 or lon-3 expression are in both cases suppressed by a null mutation in sqt-1, a second cuticle collagen gene. However, whereas loss of dbl-1 activity causes a reduction in hypodermal endoreduplication, the reduction in body length associated with overexpression of lon-3 occurs in the absence of defects in hypodermal ploidy.

THE regulation of body length and body size are poorly understood aspects of animal development (CONLON and RAFF 1999 ). The nematode Caenorhabditis elegans represents an excellent model for the study of how these features are controlled because mutations have been isolated that have marked effects on organismal morphology (KRAMER et al. 1988 ). Recent work has shown that wild-type body size in C. elegans requires a highly conserved signal transduction pathway activated by a member of the TGFß superfamily, DBL-1 (SAVAGE et al. 1996 ; MORITA et al. 1999 ; SUZUKI et al. 1999 ). Wild-type C. elegans hermaphrodites grow to a length of just over 1 mm. Worms lacking dbl-1 activity, however, are smaller (Sma): They grow to only about two-thirds the length of wild type and are slightly thinner (MORITA et al. 1999 ; SUZUKI et al. 1999 ). dbl-1 is a dose-dependent regulator of body length: Worms overexpressing dbl-1 can grow to a length of one and one-half times or more that of wild type (they display the Lon phenotype; MORITA et al. 1999 ; SUZUKI et al. 1999 ). DBL-1 is thought to function by activating the type I and type II TGFß receptors SMA-6 and DAF-4 (ESTEVEZ et al. 1993 ; KRISHNA et al. 1999 ). In turn, the receptors are thought to activate a complex of SMAD transcription factors containing the proteins, SMA-2, SMA-3, and SMA-4 (SAVAGE et al. 1996 ). Loss-of-function mutations in genes encoding any components of this pathway cause a Sma phenotype identical to that shown by worms lacking dbl-1 activity (ESTEVEZ et al. 1993 ; SAVAGE et al. 1996 ; KRISHNA et al. 1999 ).

Presently it is not known which genes SMA-2, SMA-3, and SMA-4 regulate in order to affect body length. Differential hybridization analysis with 3390 independent cDNAs has led to the identification of 21 genes whose expression is affected in worms mutant for components of the TGFß pathway regulating body length (MOCHII et al. 1999 ). One of these genes encodes SMA-6. Activation of the pathway thus leads to an increase in the transcription of at least one of the genes in the pathway, suggesting that a positive autoregulatory loop may exist. The functions of the other 20 genes whose transcription is differentially regulated are not presently known. The genes are expressed in a variety of tissues including the intestine, the hypodermis, the head, and the vulva (MOCHII et al. 1999 ).

Worms displaying the Lon or Sma phenotypes caused by defects in TGFß signaling appear to have the same number of somatic cells as wild-type worms (SUZUKI et al. 1999 ). This observation suggests that the pathway may affect body length not by altering cell proliferation but rather by directly or indirectly altering the size or shapes of some or all cells. However, the molecular mechanisms by which the TGFß pathway affects body size and body length are not yet known. Recent work has suggested that mutants lacking TGFß pathway activity may be smaller than wild type in part because of reduced endoreduplication of nuclei in the hypodermal syncytium (FLEMMING et al. 2000 ). When measured in terms of organismal volume, a major period of growth in nematodes occurs during the adult stage (when cell division has ceased). It is therefore thought that final body size is determined to a significant extent by changes in the size of certain cells and not by increases in cell number (FLEMMING et al. 2000 ). The syncytial hypodermal cell hyp7 (and its counterpart in other nematodes) is by far the largest cell in the worm and it has been suggested that final body size in nematodes is largely determined by changes in size of the hypodermis (FLEMMING et al. 2000 ). Nuclei in the syncytial hypoderm in C. elegans and a number of other nematode species have been found to undergo rounds of endoreduplication during the adult stage (FLEMMING et al. 2000 ). Since in some organisms the size of mononucleate cells correlates with ploidy (NURSE 1985 ), it has been suggested that body size in C. elegans (and other nematodes) is determined in part by the degree to which hypodermal nuclei become endoreduplicated. In support of this model, a survey of 12 different nematode species revealed that body size correlated not simply with the number of hypodermal nuclei but with the product of this number and the extent of endoreduplication (FLEMMING et al. 2000 ). Furthermore, in C. elegans the hypodermal nuclei in mutants lacking daf-4 or sma-2 activity were found to have reduced ploidy compared to wild type (FLEMMING et al. 2000 ). These studies, however, have not revealed whether reduced endoreduplication is a consequence of reduced body size or vice versa. It is noteworthy in this regard that young adult sma-2 or daf-4 mutant hermaphrodites have wild-type hypodermal ploidy (A. LEROI, unpublished results), suggesting that the Sma phenotype displayed by worms at earlier larval stages is not caused by defects in hypodermal endoreduplication.

Mutations in genes encoding components of the cuticle, the exoskeleton of the worm, can also affect body length in C. elegans (JOHNSTONE 2000 ). For example, mutations that affect body length have been identified in dpy-2, dpy-7, dpy-10, dpy-13, sqt-1, and sqt-3, all of which encode cuticle collagens (KRAMER et al. 1988 ; VON MENDE et al. 1988 ; JOHNSTONE et al. 1992 ; LEVY et al. 1993 ; VAN DER KEYL et al. 1994 ). Worms carrying certain mutations in these genes are shorter and fatter than wild type: They display the Dpy phenotype. Cuticle collagens are synthesized and secreted by hypodermal cells and they polymerize on the apical surface of the epithelium to form a complex structure consisting of six definable layers (COX et al. 1981B , COX et al. 1981C ; PEIXOTO and DE SOUZA 1992 ). New cuticle is synthesized five times during development, once in the embryo before hatching and then during each of the larval molts. During each molt, different collagen genes are expressed in discrete temporal periods (COX et al. 1981A ; JOHNSTONE and BARRY 1996 ). However, the mechanism by which proteins polymerize to form the ordered layers that make up the cuticle is currently unknown. Some collagens are known to be required for a cuticle of the correct shape to be generated. For example, mutations that are thought to reduce or eliminate activity in dpy-2, dpy-7, dpy-10, and dpy-13 all cause a Dpy phenotype, implying that these genes have specific roles in the formation of the exoskeleton (VON MENDE et al. 1988 ; JOHNSTONE et al. 1992 ; LEVY et al. 1993 ). No evidence exists at the present time, however, that these genes actively regulate body length. The sqt-1 gene, which also encodes a cuticle collagen, has interesting genetic properties in that it can mutate to give alleles that affect the morphology of the worm in different ways (KRAMER et al. 1988 ). Alleles of sqt-1 have been isolated that can cause, respectively, Rol (Roller) or Dpy (Dumpy) phenotypes. In Roller mutants the entire cuticle has a helical twist to it but the worms are neither shorter nor longer than wild type. Worms mutant for Dpy mutations, on the other hand, are shorter and fatter than wild type. Certain heteroallelic combinations of sqt-1 mutations can cause a Lon phenotype (KRAMER et al. 1988 ). sqt-1 does not seem to be a prime regulator of body length, however: Null mutations in sqt-1 cause only a slight decrease in body length. Mutations in sqt-1 that cause Dpy, Rol, or Lon phenotypes are thought to be neomorphic (KRAMER and JOHNSON 1993 ).

Here we describe a study of the role of lon-3 in the control of body length and body size in C. elegans. We have cloned the gene, analyzed the pattern of its expression, and investigated genetic interactions between mutations in lon-3 and in other genes affecting body length. To understand the relationship between the extent of endoreduplication of hypodermal nuclei and body length, we have measured the ploidy of hypodermal nuclei in a variety of Sma and Lon mutants.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Nematode strains and culture conditions:
Maintenance and handling of C. elegans strains were as described by BRENNER 1974 . Experiments were carried out at 20° unless otherwise noted. C. elegans Bristol strain N2 is the wild-type parent for all strains used in this study. All strains were constructed by standard genetic methods. The mutations used are listed below. Mutations are described by BRENNER 1974 unless otherwise stated.

• LG I: dpy-5(e61)

• LG II: rol-6(su1006) (COX et al. 1980 ), rol-6(n1178) and (e187n1268) (PARK and HORVITZ 1986 ), sqt-1(sc103) (KUSCH and EDGAR 1986 ), and sqt-1(sc13) (COX et al. 1980 )

• LG III: daf-7(e1372) (RIDDLE and BRENNER 1978 ), dpy-17 (e164), lon-1(e185), daf-4(m63) (ESTEVEZ et al. 1993 ), ncl-1 (e1865) (HEDGECOCK and WHITE 1985 ), unc-36(e251), dpy-19(e1259), sma-2(e502) (SAVAGE et al. 1996 ), and dpy-18(e304)

• LG IV: dpy-9(e12), dpy-13(e184), dpy-4(e1166), and him-8(e1489) (HODGKIN et al. 1979 )

• LG V: dpy-11(e224), dbl-1(nk3) (previously called kk3; MORITA et al. 1999 ), sma-1(e30), vab-8(e1017) (HEDGECOCK et al. 1987 ), myo-3(st386) (DIBB et al. 1989 ; MARUYAMA et al. 1989 ), lon-3(e2175) (RIDDLE et al. 1997 ), lon-3(sp5, sp6, sp23, sv18) (this study), and unc-42(e270)

• LG X: dpy-8(e130), dpy-7(e88), and dpy-6(e14)

ctDp8 is a chromosomal duplication covering much of the right arm of LGV, including lon-3 (HUNTER and WOOD 1992 ). arDf1 is a deficiency (TUCK and GREENWALD 1995 ). ctIs40 is an integrated array containing multiple copies of dbl-1(+) (SUZUKI et al. 1999 ).

Isolation of new lon-3 alleles:
Wild-type (N2) hermaphrodites were treated with ethyl methanesulfonate (EMS) and their F₂ progeny screened for Lon mutants. From a screen of 50,000 haploid genomes three new lon alleles were isolated, sp5, sp6, and sp23, which mapped to the right arm of chromosome V. All three mutations failed to complement lon-3(e2175) for the Lon phenotype. sv18 was isolated in an unrelated screen in which the mutagen was EMS.

Growth curves:
A total of 20 adult worms were placed onto a plate with OP50 bacteria and allowed to lay 100 eggs. After 12 hr, when the majority of worms had hatched, 20 worms (chosen randomly) were photographed by using a video camera connected to the microscope. Their lengths were calculated from images obtained using the software application Object-Image 1.62. A total of 20 worms (chosen randomly) were photographed every 12 hr until 120 hr after hatching.

Mapping and cloning of lon-3:
We localized lon-3 to within 0.1 map units of myo-3 by three-factor mapping. Cosmids from this region were injected at a concentration of 10 µg/ml into hermaphrodites of the genotype unc-36(e251); lon-3(e2175) together with 50 µg/ml of RIp16 plasmid DNA [which encodes unc-36(+); HERMAN 1995]. C35G11 rescued lon-3(e2175) when injected at a concentration of 2 µg/ml and caused the majority of F₁ worms to be considerably shorter than wild type when injected at a concentration of 10 µg/ml. From one set of injections in which C35G11 was injected at a concentration of 10 µg/ml into lon-3(e2175) hermaphrodites, four transformed lines were generated containing, respectively, the extrachromosomal arrays svEx50, svEx51, svEx52, and svEx53. In three of the lines (those containing svEx50, svEx52, and svEx53) the majority of worms carrying the array were Dpy and all worms were rescued for the Lon phenotype. In the fourth (that containing svEx52), all worms were partially or completely rescued for the Lon phenotype and 5% were Dpy. A 5.5-kb NcoI-to-KpnI fragment from C35G11 that spans the predicted gene, ZK836.1, was subcloned into pBluescript II KS(+) to generate pVB52JN. pVB52JN both rescued lon-3(e2175) (when injected at 2 µg/ml) and caused worms to be shorter than wild type when injected at concentrations of 5 µg/ml or above. The extrachromosomal array, svEx57, was generated by injecting pVB52JN at a concentration of 50 µg/ml into hermaphrodites of the genotype unc-36(e251) together with 50 µg/ml of RIp16.

Determination of sequence changes associated with lon-3 mutant alleles:
To identify mutations associated with lon-3 alleles we first used the method of RNAse cleavage mismatch detection (Ambion, Austin, TX) to determine the regions of the gene in which the changes resided. The regions containing the differences were amplified and sequenced. The sequence changes were confirmed by sequencing two independent PCR products. It proved impossible to generate PCR products from genomic DNA isolated from worms homozygous for lon-3 (e2175). Southern blot analysis indicated that this allele is associated with a DNA rearrangement at the lon-3 locus (data not shown).

lon-3 reporter genes:
To generate pVB54JN, which contains lacZ under the control of lon-3 promoter sequences, pVB52JN was digested with EaeI, blunt ended by filling in using Klenow, and then digested with BamHI. The fragment generated was inserted into pPD95.07 (FIRE et al. 1990 ) digested with SmaI and BamHI. pVB54JN was injected at a concentration of 100 µg/ml into worms of the genotype unc-36(e251) together with 50 µg/ml RIp16 (HERMAN 1995 ) to generate the extrachromosomal arrays svEx77, svEx78, and svEx79. lacZ expression was analyzed by using a ß-galactosidase assay as described elsewhere (FIRE et al. 1990 ). The temporal and spatial patterns of expression were the same for all three arrays. To generate pVB82JN, which contains green fluorescent protein (GFP) under the control of lon-3 promoter sequences, pVB52JN was digested as described for pVB54JN. The fragment generated was inserted into pPD95.67 digested with BamHI and BalI. pVB82JN was injected at a concentration of 50 µg/ml as described for pVB54JN to generate the extrachromosomal array svEx130.

Body length measurements:
A total of 10–20 hermaphrodites of the appropriate genotype were placed onto seeded plates and allowed to lay eggs for 1 hr. The adults were removed and the eggs allowed to develop. Worm lengths were measured 96 hr after hatching with a Leica MZ6 dissecting microscope connected to a digital video camera. To avoid bias, the lengths of all progeny on a given plate were measured. A Zeiss Axioplan 2 microscope was used to analyze worms, and the software application Openlab 2.0.7 (Improvision) was used for all micrographs.

Endoreduplication:
DNA content was determined by microdensitometry as described elsewhere (FLEMMING et al. 2000 ). In many instances, the measurements of ploidy did not fall into multiples of two as might be expected. As discussed in FLEMMING et al. 2000 , this is explained by the fact that different nuclei endoreduplicate to different extents.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

lon-1 is a negative regulator of endoreduplication:
Mutations that reduce or eliminate daf-4 or sma-2 activity cause a reduction in the extent of endoreduplication of hypodermal nuclei (FLEMMING et al. 2000 ). To determine whether dbl-1 is also required for wild-type hypodermal ploidy we measured the extent of endoreduplication in hermaphrodites homozygous for a dbl-1 null mutation, nk3. Results presented in Table 1 show that hypodermal nuclei in dbl-1(0) mutants have an average ploidy considerably less than those in wild-type hermaphrodites. Thus, as for other genes required for wild-type body size, dbl-1 is required for hypodermal nuclei to undergo the appropriate number of rounds of endoreduplication.

Since the Sma phenotype is associated with reduced hypodermal endoreduplication, we investigated whether hypodermal nuclei in worms displaying the Lon phenotype were hyperendoreduplicated. Hermaphrodites homozygous for mutations in lon-1 can be as much as 50% longer than wild type (BRENNER 1974 ). We measured hypodermal ploidy in hermaphrodites homozygous for lon-1(e185), which is thought to be a null mutation (N. UENO, personal communication), and found that the average ploidy was 20% higher than wild-type hermaphrodites (P < 0.0001; Table 1). Thus lon-1 is a negative regulator of both body length and hypodermal endoreduplication.

To determine whether the Lon phenotype was invariably associated with hyperendoreduplication or whether the effect occurred in only certain Lon mutants, we measured hypodermal ploidy in worms overexpressing dbl-1. No significant difference was detected in the extent of endoreduplication in ctIs40 worms (which overexpress dbl-1; MORITA et al. 1999 ; SUZUKI et al. 1999 ) compared to wild type (P = 0.3). Although we cannot conclude definitely that dbl-1 overexpression does not cause an increase in endoreduplication, power calculations indicate that we would have had a 99% chance of detecting a difference in ploidy of 18%, which is comparable to the difference between N2 and lon-1(0). Our results suggest therefore that if dbl-1 can cause an increase in endoreduplication when overexpressed, it does not do so very efficiently.

Isolation and characterization of lon-3 mutant alleles:
To learn more about how body length and body size are regulated in C. elegans we carried out a genetic screen for new mutations causing a Lon phenotype. From a screen of 50,000 haploid genomes, we isolated 18 new Lon alleles. Genetic mapping and complementation tests (MATERIALS AND METHODS) revealed that three of these new mutations were allelic to lon-3. While we were carrying out this screen, another allele of lon-3, sv18, was isolated in an unrelated genetic screen (H. FARES and I. GREENWALD, personal communication). Both the new lon-3 alleles reported here and the previously existing allele, e2175, are recessive to wild type (Table 2 and data not shown). By genetic criteria they appear to reduce or eliminate activity. For example, hermaphrodites homozygous for lon-3(sp23) but carrying a wild-type allele on a chromosomal duplication are non-Lon, whereas hermaphrodites of the genotype lon-3(sp23)/Df are Lon (Table 2).

To determine when lon-3 mutants become longer than wild type, we measured the lengths of lon-3 mutant hermaphrodites at different times after hatching. For comparison, wild-type hermaphrodites were measured under identical growth conditions. Two different lon-3 alleles, lon-3(sp23) and lon-3(e2175), were used in the analysis. Results described below suggest that sp23 is a null allele. Fig 1 shows that lon-3 mutants first begin to express the Lon phenotype 36 hr after hatching. After this time they become progressively longer than wild type until, when they are adults, they are 22% longer than wild type (Fig 1 and Fig 2; Table 3).

View larger version (22K): In this window In a new window Download PPT slide   Figure 1. Growth curves. (A) lon-3(-) denotes lon-3(e2175), lon-3(0) denotes lon-3(sp23), and lon-3(++) denotes unc-36 (e251);svEx57. svEx57 is an extrachromosomal array containing multiple copies of lon-3(+). (B) dbl-1(++) denotes ctIs40, an integrated array containing multiple copies of dbl-1(+) (SUZUKI et al. 1999 ), and dbl-1(0) denotes dbl-1(nk3) (MORITA et al. 1999 ). (C) lon-1(-) denotes lon-1(e185) and lon-3(-) denotes lon-3(e2175). (D) lon-1(-) denotes lon-1(e185) and lon-3(0) denotes lon-3(sp23).

View larger version (40K): In this window In a new window Download PPT slide   Figure 2. Defects in lon-3 expression affect body length. dbl-1(0) denotes dbl-1(nk3) (MORITA et al. 1999 ), lon-3(0) denotes lon-3(sp23), and lon-3(++) denotes worms carrying an extrachromosomal array, svEx57, containing multiple copies of lon-3(+). dbl-1(++) denotes ctIs40, an integrated array containing multiple copies of dbl-1(+) (SUZUKI et al. 1999 ). Bar, 0.2 mm.

To determine whether lon-3 mutant hermaphrodites are longer than wild type because particular regions of the worm are longer or because all organs are proportionately longer, we carried out a morphometric analysis. Results presented in Table 4 show that, in lon-3 mutants, the regions between the pharynx and the anterior bend of the gonad and between the posterior bend of the gonad and the rectum are expanded compared to wild type. The gonad is on average 11% longer than in wild type but constitutes a lower percentage of the total worm length.

To determine whether lon-3 mutations, like the mutation in lon-1, cause an increase in endoreduplication of hypodermal nuclei, we measured hypodermal ploidy in lon-3(sp23) mutant hermaphrodites. Results presented in Table 1 show that loss of lon-3 activity causes an increase of body length without increasing hypodermal ploidy [P = 0.6 for lon-3(sp23) vs. wild type; P = 0.1 for lon-3(e2175) vs. wild type].

Cloning of lon-3:
To investigate further how lon-3 functions we cloned the gene (Fig 3A and MATERIALS AND METHODS). A 5.5-kb fragment from the cosmid C35G11, spanning the predicted gene ZK836.1, rescued the Lon phenotype caused by lon-3(e2175). We found that the lon-3 alleles sp6, sp23, and sv18 are associated with mutations that introduce premature stop codons into ZK836.1, indicating that this predicted gene does indeed correspond to lon-3 (Fig 3B). The splicing pattern predicted by the Genefinder program for ZK836.1 was confirmed by sequencing the cDNA clone cm06a10.

View larger version (32K): In this window In a new window Download PPT slide   Figure 3. lon-3 encodes a collagen. (A) Top, genetical and physical maps of the lon-3 region, bottom, exon/intron structure of lon-3, where solid boxes in pVB52JN represent exons. (B) The primary structure of LON-3 protein. The darkly shaded boxes represent Gly-X-Y repeat domains. The lightly shaded boxes, labeled D to A, represent homology blocks found in many cuticle collagens (KRAMER 1994 ). The sequences contained in the homology blocks have been shown to be important for the processing of collagen chains (YANG and KRAMER 1999 ). Thin vertical lines represent the positions of highly conserved cysteine residues, important for disulfide bonding during cuticle assembly. The positions of stop codons introduced by the mutations sp23, sp6, and sv18 are indicated both by arrows (top) and solid squares below the affected amino acid in the predicted protein sequence of LON-3 (bottom). The Gly-X-Y repeat sequences are underlined, the homology blocks D, C, B, and A are boxed, and the conserved cysteines are in boldface type. (C) The spacing of the cysteine residues in different cuticle collagen families (KRAMER 1994 ). "Amino side" indicates cysteine residues lying amino terminal to the first Gly-X-Y domain. "Carboxyl end" indicates residues lying immediately after the last Gly-X-Y domain. These sequences have been submitted to the GenBank database under accession no. AF262406.

lon-3 is predicted to encode a collagen (Fig 3B). The predicted protein sequence contains a central region containing four closely spaced domains consisting of Gly-X-Y repeat sequences (in which X and Y frequently are prolines), characteristic of collagens in both vertebrates and invertebrates (VAN DER REST and GARRONE 1991 ; KRAMER 1994 ). The primary structure of LON-3 is typical of C. elegans cuticle collagens (KRAMER 1994 ). First, at the amino terminus there are four short sequence motifs named homology blocks D-A that are conserved in most cuticle collagens. Second, the predicted sequence contains a number of cysteine residues that show spacing typical of that seen in cuticle collagens: Three conserved cysteine residues lie immediately N terminal to the first Gly-X-Y domain, two lie immediately after, and two lie after the fourth and last Gly-X-Y domain (Fig 3B and Fig C). C. elegans cuticle collagens have been divided into nine subfamilies on the basis of the precise spacing of the cysteine residues (KRAMER 1997 ). It is thought that collagens belonging to the same subfamily may be able to form heterotrimers (KRAMER 1994 ). The spacing of the cysteines places LON-3 in the SQT-1 subfamily of cuticle collagens (KRAMER et al. 1988 ; KRAMER 1994 ; Fig 3C).

Increasing or decreasing lon-3 activity causes reciprocal changes in body length:
It is known that neomorphic mutations in certain C. elegans cuticle collagen genes can affect the morphology of the worm (KRAMER and JOHNSON 1993 ). In some cases null mutations in these genes have a wild-type phenotype suggesting that, in an otherwise wild-type background, the genes are not necessary for the formation of a cuticle of the correct shape. To help elucidate the nature of the mutations in lon-3 causing a Lon phenotype, we determined the sequence changes associated with three mutant lon-3 alleles. lon-3(sp23) was found to be associated with a G-to-A transition in codon 34 that leads to the generation of a stop codon (Fig 3B). sp23 is therefore predicted to encode a severely truncated protein lacking eight-ninths of the full-length protein. Two other lon-3 alleles were also found to be associated with sequence changes that introduce stop codons into the lon-3 open reading frame (Fig 3B). sp6 is associated with a C-to-T transition that introduces an ochre stop codon into position 112, and sv18 is associated with a C-to-T transition that introduces an ochre stop at position 237. Two alleles of rol-6, e187n1268 and n1178, are predicted to encode truncated proteins similar in length to those predicted to be encoded by lon-3(sp23) and lon-3(sv18), respectively (KRAMER and JOHNSON 1993 ). rol-6(n1178) and rol-6(e187n1268) are both thought to be null alleles (KRAMER and JOHNSON 1993 ). We were unable to identify any lesions in the lon-3 open reading frame associated with sp5. It is possible that the sp5 mutation lies in a site in noncoding DNA that is required for the transcriptional regulation of lon-3.

The nature of the sequence changes associated with the sp6, sp23, and sv18 alleles suggests that these alleles reduce or eliminate gene activity and therefore that the loss-of-function phenotype of lon-3 is Lon. Consistent with this idea, lon-3(RNAi) hermaphrodites were Lon (Fig 4 and data not shown). No C. elegans collagen gene has previously been identified for which the loss-of-function phenotype is Lon. The null phenotypes of cuticle collagen genes that have been characterized genetically to date are Dpy, wild type, lethal, or Tal (KRAMER 1997 ).

View larger version (69K): In this window In a new window Download PPT slide   Figure 4. RNA interference experiment with lon-3 results in worms that are longer than wild type. (Top) The progeny of wild-type worms injected with dsRNA corresponding to lon-3 cDNA (cm06a10) are much longer than wild type (bottom). Bar, 0.2 mm.

In the course of cloning lon-3 we noticed that DNA encoding lon-3 was not only able to rescue the lon-3 mutant phenotype but could also cause worms to be shorter than wild type. For example, whereas injection of 2 µg/ml of C35G11 cosmid DNA into lon-3(e2175) hermaphrodites rescued the Lon phenotype of transformed progeny to wild type, injection of 10 µg/ml caused the F₁ transformants to be shorter than wild type (MATERIALS AND METHODS). To investigate more closely the effect of multiple copies of lon-3 on body length, we generated worms of the genotype unc-36(e251); svEx57. svEx57 is an extrachromosomal array containing lon-3(+) (encoded by the plasmid pVB52JN). Worms carrying the array were Dpy (Fig 2). pVB52JN contains lon-3, 3.8 kb of DNA upstream from lon-3, and part of the gene ZK836.2 (which encodes 2-oxoglutarate dehydrogenase), in whose third intron lon-3 lies. Disruption of the lon-3 open reading frame by insertion of lacZ sequences abolished the ability of the 5.5-kb fragment in pVB52JN to confer the Dpy phenotype on transformed progeny (data not shown). Thus the phenotype observed requires LON-3 protein activity and is not caused by other sequences on the rescuing fragment.

Worms carrying the svEx57 array were first noticeably shorter than wild type during the L3 stage. They remained shorter throughout the period of measurement and as older adults (data not shown). On average, worms carrying the array were 15% shorter than wild type (Table 2; Fig 1 and Fig 2). Many animals, however, were considerably shorter than this: Some were as much as 30% shorter than wild-type hermaphrodites of the same age (data not shown). Animals carrying the array were otherwise wild type and healthy: They did not show defects in locomotion or viability. Furthermore, as described below, the array did not cause synthetic lethality in combination with a variety of other mutations. Together, our results suggest that reducing or eliminating lon-3 activity results in increased length whereas increasing gene activity decreases length. No other C. elegans collagen gene that causes these reciprocal effects on body length has been described previously.

Defects in lon-3 expression have no effect on dauer development or on development of the male tail:
The similarity of the lon-3 loss-of-function phenotype with the phenotype caused by overexpression of dbl-1 suggested to us that the TGFß pathway might regulate LON-3 activity or be regulated by it. For example, one possibility might be that LON-3 functions as a negative regulator of TGFß ligands in C. elegans. To investigate this possibility we first analyzed the effects of defects in lon-3 expression on processes (other than the control of body length) that are known to require TGFß-mediated signaling. Besides regulating body length, dbl-1 is also required for correct dorso-ventral patterning of a group of sensory neurons on the fan, a structure used during mating (MORITA et al. 1999 ; SUZUKI et al. 1999 ). In wild type, three of the sensory rays open out onto the dorsal surface of the fan while four open out onto the ventral surface (SULSTON and HORVITZ 1977 ). In dbl-1 mutants, the rays that normally open out onto the dorsal surface lose their dorsal identity and instead open out onto the ventral surface (SUZUKI et al. 1999 ). Neither lon-3(0) mutant males nor males of the genotype him-8(e1498); svEx57 [which harbor multiple copies of lon-3(+)] showed defects in patterning of the sensory rays (data not shown), suggesting that LON-3 has no effect on DBL-1 activity in ray patterning.

A second TGFß signaling pathway in C. elegans regulates the decision between entering the dauer developmental state and undergoing the reproductive life cycle. Larvae mutant for daf-7, which encodes the ligand activating this pathway, are dauer constitutive: They enter the dauer state under conditions in which wild-type worms remain in the reproductive life cycle, when the population density is low and the food source plentiful (REN et al. 1996 ). Neither lon-3(lf) mutants nor worms of the genotype lon-3(+), svEx57[lon-3(+)], show defects in dauer development. Furthermore, overexpression of lon-3 does not enhance the dauer constitutive phenotype of a temperature-sensitive daf-7 mutation at the semipermissive temperature. Thirty-six percent of both daf-7(e1372ts) (n = 50) mutant hermaphrodites and hermaphrodites of the genotype daf-7(e1372ts); svEx57 (n = 50) become dauers when raised at 20° (the semipermissive temperature).

lon-3 is expressed in hypodermal cells and shows strong genetic interactions with sqt-1 and rol-6:
To determine in which tissues lon-3 is transcribed, we generated worms carrying either a gfp or lacZ reporter gene under the control of lon-3 promoter sequences (MATERIALS AND METHODS). The pattern of gfp expression in hermaphrodites carrying the gfp reporter was dynamic but at all stages restricted to hypodermal cells. Early in the L1 stage, expression was seen in H0, H1, and H2, in the anterior V cells, and in the T cells. Weak expression was also seen in hypodermal nuclei in the head and the tail, including those in hyp5, hyp6, hyp8, hyp9, and hyp10. Expression was not seen at this stage, however, in the P cells or in nuclei in the hyp7 syncytium. After division of H1, both H1.a, a seam cell, and H1.p, which joins hyp7, expressed gfp. Likewise, after division of V1–V4 both the anterior daughters (which join hyp7) and the posterior daughters (which remain seam cells) expressed gfp. This pattern was repeated at each of the larval molts with the result that in adult worms, all descendents of H1, H2, and V1–V4 expressed gfp. At the end of the L1 stage, V5.p could be seen to express gfp but not V5.a, which is a neuroblast. In the P cell lineages, expression was first seen toward the end of the L1 stage in P1.p, P2.p, P9.p, P10.p, and P11.p nuclei. These cells fuse with hyp7 during the L1 stage (SULSTON and HORVITZ 1977 ). Expression of gfp was not seen, on the other hand, in the daughters of P3–P8 or P12 (all of which remain separate from hyp7). By the end of the L1 stage most nuclei in hyp7 expressed gfp, as did those in hyp5, hyp6, hyp8, hyp9, hyp10, and hyp11. After division of P3.p, P4.p, and P8.p during the L3 stage, all six daughters started to express gfp concomitant with their fusion with hyp7. The daughters of P5.p, P6.p, and P7.p, the progenitors of the vulva, remained negative for gfp expression as did all the cells they subsequently gave rise to. In adult worms, no hypodermal nuclei that failed to express gfp could be identified. Conversely, no nonhypodermal nuclei were observed at any stage that expressed the gfp marker. Examples of the expression of the lon-3:: gfp marker are shown in Fig 5.

View larger version (121K): In this window In a new window Download PPT slide   Figure 5. gfp and lacZ reporter genes under the control of lon-3 promoter sequences are expressed in epidermal cells throughout the length of the worm. (A) Bright-field photomicrograph of an adult hermaphrodite carrying multiple copies of pVB54JN, a plasmid containing lacZ under the control of lon-3 promoter sequences. Bar, 0.1 mm. (B) Nomarski differential interference contrast photomicrographs of the same worm at higher magnification. Bar, 0.01 mm. Arrows in B indicate nuclei in the hyp5, hyp6, and hyp7 syncytia. (C and E) Nomarski differential interference contrast photomicrographs of an L4 hermaphrodite carrying multiple copies of pVB-82JN, a plasmid containing gfp under the control of lon-3 promoter sequences. (D and F) Fluorescence micrographs of the same views. Bars in C–F, 0.03 mm. (C and D) Arrows indicate nuclei of seam cells and arrowheads indicate hypodermal nuclei in the hyp7 syncytium. (E and F) Arrows indicate nuclei in hyp10 and hyp11.

Although we did not analyze worms carrying the lon-3::lacZ marker gene in as much detail, the pattern of expression appeared to be the same as that for the gfp reporter. In adult worms carrying the lacZ reporter gene, expression of lacZ was seen in many hypodermal nuclei including those in hyp5, hyp6, hyp7, hyp8, hyp9, hyp10, and hyp11 (Fig 5). Expression of lacZ was first seen during the L1 stage (data not shown) and then persisted through the subsequent larval stages and into the adult stage. The sequence of the predicted LON-3 protein—together with the fact that reporter genes under the control of lon-3 promoter sequences are transcribed in hypodermal cells, cells that synthesize and secrete cuticle—suggested that LON-3 might be a cuticle collagen. Further evidence in favor of this hypothesis is that the plasmids pVB61JN and pVB71JN encoding LON-3::GFP fusion proteins, in which GFP is fused in frame close to the carboxy terminus of LON-3, conferred a Rol phenotype on transformed progeny (Table 5). (Transgenic worms containing these plasmids did not fluoresce green.)

As described above, the predicted LON-3 protein sequence is most similar in sequence to the SQT-1 family of cuticle collagens. Certain alleles of sqt-1 and a gene encoding another member of the family, rol-6, have marked effects on organismal morphology. sqt-1 and rol-6 show strong genetic interactions with one another and it has been suggested that the two proteins might function together in a physical complex (KRAMER et al. 1990 ; KRAMER and JOHNSON 1993 ; YANG and KRAMER 1999 ). To investigate the possibility that LON-3 might function with either ROL-6 or SQT-1, we examined genetic interactions between lon-3 mutations and those in sqt-1 or rol-6. Worms homozygous for the sqt-1 null mutation, sc103, or the rol-6 null mutation, e187n1268, are almost wild type in length, being only slightly Dpy (Table 6). The Lon phenotype caused by lon-3(sp23) was strongly suppressed by sqt-1(0): sqt-1(0); lon-3(0) double mutants were found to be longer than wild type but considerably shorter than lon-3 single mutants (Table 6). Thus sqt-1 is required for the expression of the Lon-3 phenotype. To help determine whether the suppression observed represented a significant genetic interaction or simply represented the independent effects of the two genes, we calculated the expected length of the sqt-1(0); lon-3(0) double mutant assuming simple additivity of the effects of the individual mutations (Table 6). The double mutant was shorter than would be expected if sqt-1 and lon-3 affected length entirely independently of one another (P < 0.0001). Similarly, a rol-6(0) mutation suppressed the Lon phenotype caused by lon-3(0), and the rol-6(0); lon-3(0) double mutant was shorter than expected if the genes acted entirely independently (P < 0.0001; Table 6). Interestingly, sqt-1(0) also suppressed the phenotype caused by overexpression of lon-3: Worms of the genotype sqt-1(0); svEx57 were longer than either sqt-1(0) single mutant worms or worms overexpressing lon-3 (P < 0.0001; Table 6). No sqt-1(0); svEx57 worms displayed the Dpy phenotype. In contrast to sqt-1(0), rol-6(0) did not suppress the phenotype caused by worms overexpressing lon-3 but rather increased the expressivity of the phenotype: rol-6(n1178); svEx57 worms were strongly Dpy (Table 5).

To determine whether the interactions observed above were reciprocal, we examined the ability of lon-3 null mutations to suppress phenotypes caused by neomorphic mutations in sqt-1 or rol-6. We found that lon-3(0) almost completely suppressed the Rol phenotypes caused by sqt-1(sc13) or rol-6(su1006) and the Sqt phenotype caused by sqt-1(sc1) (Table 5). Thus not only does LON-3 require sqt-1 activity to affect body morphology but the reverse is also true: Mutant SQT-1 and ROL-6 proteins require lon-3 activity to confer Rol or Sqt phenotypes.

dbl-1 requires sqt-1 and rol-6 activity to regulate body length:
Since mutations in both lon-3 and dbl-1 can affect body length, and phenotypes caused by defects in lon-3 expression are modified by null mutations in sqt-1 or rol-6, we investigated whether the Lon phenotype caused by overexpression of dbl-1 is also modified by sqt-1 or rol-6 mutations. A sqt-1(0) mutation strongly suppressed the Lon phenotype caused by multiple copies of dbl-1 (Table 6). Worms carrying an integrated array, ctIs40, harboring multiple copies of dbl-1(+) but homozygous for sqt-1(0), were not longer than wild type. The length of the double mutant was considerably less than the length expected if db1-1 and sqt-1 functioned entirely independently (P < 0.0001; Table 6). Similarly, rol-6(0) suppressed the phenotype caused by ctIs40 to wild type (data not shown). Thus functional SQT-1 and ROL-6 proteins are required for dbl-1 to regulate body length.

The fact that both dbl-1 and lon-3 require sqt-1 and rol-6 to affect body length led us to investigate the relationship between lon-3 and dbl-1. Morphological measurements indicate that the phenotype caused by overexpression of dbl-1 is similar to that caused by loss of lon-3 function (SUZUKI et al. 1999 ; Table 4). Qualitatively, the regions that are expanded in worms that overexpress dbl-1 are the same as those in lon-3(0) mutant worms. Quantitatively, in the former, the region between the pharynx and the gonad is expanded to a greater extent, and the gonad spans a commensurately smaller percentage of the total worm length.

To investigate further how body length is regulated in C. elegans, we carried out genetic epistasis tests with a dbl-1 null mutation and mutations in lon-1 or lon-3. The Sma phenotype caused by dbl-1(0) is characterized by both a reduction in length compared to wild type and also a slight reduction in width (MORITA et al. 1999 ; SUZUKI et al. 1999 ). lon-1(e185) partially suppressed one aspect of the dbl-1(0) phenotype: lon-1(e185); dbl-1(0) double-mutant hermaphrodites were longer than dbl-1(0) single mutants [although not as long as lon-1(e185) single mutants (Table 6)]. lon-1(e185) also partially suppressed the endoreduplication defect caused by dbl-1(0): Hypodermal ploidy was slightly increased in the lon-1 (e185); dbl-1(0) double mutant compared to the dbl-1(0) single mutant (Table 1). It is noteworthy, however, that lon-1(e185) did not appreciably suppress the thinness aspect of the dbl-1 phenotype: Double-mutant animals were thinner than both wild type and lon-1(e185) single mutants.

lon-3(0) partially suppressed the "shortness" aspect of the dbl-1(0) Sma phenotype (Table 6 and Fig 2) but did not suppress the thinness caused by dbl-1(0): dbl-1(0) lon-3(0) were thinner even than dbl-1(0) single mutants (Fig 2). In addition, lon-3(0) did not rescue the endoreduplication defect caused by dbl-1(0) (Table 1). The fact that lon-1, but not lon-3 mutations, affect endoreduplication suggests that lon-3 does not function by regulating lon-1. Consistent with this idea, lon-1(e185); lon-3 (0) double-mutant hermaphrodites were longer than either single mutant alone (Table 6).

Results presented in Table 6 show that mutations in lon-1 or lon-3 can partially suppress the Sma phenotype caused by null mutation in daf-4 and a hypomorphic mutation in sma-2.

The lon-3 overexpression phenotype is not caused by indiscriminate inhibition of collagen function:
Thirty-three genes that can mutate to give rise to alleles conferring a Dpy phenotype have previously been identified. Seven of these genes (dpy-2, dpy-7, dpy-10, dpy-13, sqt-1, sqt-3, and rol-6) are known to encode collagens (JOHNSTONE 2000 ), six [dpy-21, dpy-26, dpy-27, dpy-28, sdc-3 (previously dpy-29), and dpy-30] encode genes that are involved in X chromosome dosage compensation (MEYER 2000 ), one (dpy-18) is a prolyl-hydroxylase (HILL et al. 2000 ), one (dpy-23 or apm-2) a clathrin-associated protein (SHIM and LEE 2000 ), and three (dpy-5, dpy-19, and dpy-20) encode novel proteins (CLARK et al. 1995 ; RIDDLE et al. 1997 ; HONIGBERG and KENYON 2000 ). The products encoded by the remainder are not yet known. We reasoned that lon-3 might cause a Dpy phenotype by interfering with the function of one or more of the dpy genes. To test this possibility we examined a panel of Dpy mutants for their genetic interactions with sqt-1(0). As described above, the sqt-1 null mutation sqt-1(sc103) completely suppresses the Dpy phenotype caused by overexpression of lon-3(+) (Table 6). Therefore, if lon-3 functioned by interfering with dpy-17 activity, for example, the phenotype caused by dpy-17 mutations should also be suppressed by sqt-1(0). However, for dpy-17 and all the other genes we tested, sqt-1(0) had the converse effect: It slightly or markedly enhanced the expressivity of the Dpy phenotype. Results presented in Table 7 show that worms homozygous for sqt-1(0) and a mutation in dpy-5, dpy-13, or dpy-4 were not wild type in length; rather, they displayed an extreme Dpy phenotype. sqt-1(0) did not enhance the Dpy phenotype caused by mutations in dpy-6, dpy-7, dpy-8, dpy-9, dpy-11, dpy-14, or dpy-17 to the same extent but in no case was the Dpy phenotype suppressed. These observations suggest that lon-3 overexpression does not cause a Dpy phenotype by interfering with the activity of these genes and, in turn, that lon-3 does not affect length by indiscriminately blocking the function of all cuticle collagens.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We show here that increasing or decreasing lon-3 levels causes reciprocal changes in body length in C. elegans. Loss-of-function mutations in lon-3 cause a 22% increase in body length whereas an increase in lon-3 activity causes worms to be considerably shorter than wild type. We show that lon-3 is predicted to encode a collagen that likely is a component of the cuticle and that lon-3 requires the activity of two cuticle collagen genes, sqt-1 and rol-6, to affect body length. We also demonstrate that defects in lon-3 expression can affect body length independently of their effects on endoreduplication of hypodermal nuclei. Finally, we have shown that null mutations in dbl-1 or lon-1 cause, respectively, a decrease or increase in hypodermal endoreduplication.

lon-3 encodes a collagen that requires sqt-1 activity to function:
The predicted LON-3 protein sequence is most similar to that of the cuticle collagen ROL-6. Furthermore, the spacing of conserved cysteine residues in LON-3 places it in the SQT-1 subfamily of cuticle collagens of which ROL-6 is also a member (KRAMER 1994 ). Besides these three proteins, four other members of the family are encoded in the C. elegans genome (Fig 3C), F54D8.1, W08D2.6, ZK1248.2, and C34D4.15. F54D8.1 may correspond to dpy-17. To date lon-3, rol-6, and sqt-1 (and possibly dpy-17) are the only genes of the family for which mutations have been isolated. W08D2.6, ZK1248.2, and C34D4.15 do not lie in regions to which any of the mutations affecting body morphology map. It appears therefore that mutations in these other genes that affect body morphology are rarer than those in sqt-1, dpy-17, rol-6, or lon-3. lon-3 is the first member of the family for which the loss-of-function phenotype has been shown to be Lon. lon-3 is also the first collagen gene that has been reported to give rise to a Dpy phenotype when present in multiple copies.

It has previously been speculated that cuticle collagens in the same subfamily may function together in the generation of the cuticle (KRAMER et al. 1990 ). Evidence supporting this idea is that sqt-1 and rol-6 show strong genetic interactions with one another: Null mutations in one gene can suppress phenotypes associated with neomorphic mutations in the other (KRAMER and JOHNSON 1993 ). Although no evidence for a direct covalent interaction between ROL-1 and SQT-1 proteins has been forthcoming, it is thought that the two proteins may be able to form noncovalent heterodimers (KRAMER et al. 1990 ; KRAMER and JOHNSON 1993 ). The genetic interactions that we have observed between lon-3 and rol-6 and between lon-3 and sqt-1 are similar in one respect to those reported for rol-6 and sqt-1: A null mutation in lon-3 can suppress phenotypes caused by neomorphic mutations in rol-6 and sqt-1 (Table 5). For example, the Rol phenotype of sqt-1(sc13) is suppressed by lon-3(0). In the case of Rol alleles of sqt-1 and rol-6, the suppression is not simply at the level of the ability of the worms to roll: lon-3(0) suppressed the twisting of the cuticle caused by aberrant SQT-1 or ROL-6 proteins. These observations suggest that these neomorphic forms of SQT-1 and ROL-6 require LON-3 to cause defects in the cuticle. We have not addressed in this article whether LON-3 can bind directly to SQT-1 or ROL-6. It will be interesting in the future to investigate whether the genetic interactions we have observed reflect direct physical interactions between SQT-1 or ROL-6 and LON-3.

The fact that a null mutation in sqt-1 completely suppresses the phenotype caused by multiple copies of lon-3(+) is instructive since it implies that wild-type LON-3 requires wild-type SQT-1 to cause a decrease in body length. Furthermore, our observation reported here that sqt-1(0) does not suppress the Dpy phenotype caused by mutations in dpy-5, dpy-6, dpy-7, dpy-8, dpy-9, dpy-11, dpy-13, dpy-14, dpy-17, dpy-18, dpy-19, or dpy-20 suggests that overexpression of lon-3 does not give rise to a Dpy phenotype by causing nonspecific defects in the generation of the cuticle (for example, by binding to and inactivating many different components of the cuticle). It has previously been reported that sqt-1(0) strongly enhances the Dpy phenotype caused by mutations in the genes dpy-3 and dpy-10 (KUSCH and EDGAR 1986 ; LEVY et al. 1993 ). Thus LON-3 is unlikely to function by interfering with the activities of these genes. Although LON-3 requires SQT-1 to confer a Dpy phenotype, it is not possible to conclude from our results that LON-3 necessarily functions directly with SQT-1. SQT-1 could be required for the expression or function of a protein together with which LON-3 acts.

The identity of LON-3, its pattern of expression, and the fact that lon-3::gfp fusion genes can cause morphological defects suggest that one possible mechanism by which LON-3 might act is to affect directly the elasticity of the cuticle. Confirmation of this model, however, must await the demonstration that LON-3 is indeed a component of the cuticle. Worms homozygous for lon-3 (0) do not appear to be longer than wild-type worms because of an increase in cell number. Although we did not determine the exact numbers, no obvious increase in the number of somatic nuclei compared to wild type was seen in lon-3 mutants stained with 4'6-diamidino-2-phenylindole (J. NYSTRÖM and S. TUCK, unpublished results). We also examined lon-3 mutant worms by Nomarski microscopy but found no extra somatic cells.

dbl-1 and lon-3 could function either in the same pathway or in parallel:
We have not determined in this article whether lon-3 and dbl-1 function in the same pathway or in parallel pathways. It is noteworthy, however, that morphometric analyses indicate that both lon-3(0) mutants and worms overexpressing dbl-1 are longer than wild type largely because two particular regions of the worm—that between the pharynx and the anterior arm of the gonad and that between the posterior arm of the gonad and the rectum—are expanded relative to wild type. Although it is presently not clear how dbl-1 regulates body length, as in the case with lon-3, it is not thought that the number of somatic cells in worms expressing different levels of dbl-1 is different from that in wild type (SUZUKI et al. 1999 ).

The expression of the lon-3::lacZ fusion gene reported here was not affected by mutations that increase or decrease the activity of the TGFß pathway affecting body length (J. NYSTRÖM and S. TUCK, unpublished data). Thus no evidence exists presently to suggest that dbl-1 regulates body length by regulating the transcription of lon-3. However, since LacZ protein perdures, it is possible that the lon-3::lacZ transgene is not a sufficiently sensitive reporter to detect changes in the rate of lon-3 gene transcription caused by mutations in genes in the TGFß pathway. Further work will be needed to address this issue. Another possibility is that dbl-1 does not affect lon-3 transcription but instead regulates length, in part at least by affecting LON-3 protein levels. It is noteworthy in this regard that the TGFß pathway regulating body length regulates the expression of a gene predicted to encode a protein similar to collagenase (MOCHII et al. 1999 ). It will be interesting in the future to determine whether the product of this gene can affect body length, LON-3 protein stability, or both.

It is also possible that dbl-1 does not regulate lon-3 and that the genes affect body length independently. However, our findings that both lon-3 and dbl-1 interact genetically with sqt-1 suggest that if lon-3 and dbl-1 do not regulate one another's activity, they may at least have a common target. Further biochemical work will be required to determine more precisely how the two genes function.

Several observations suggest that if dbl-1 does regulate lon-3, then lon-3 cannot be the only target of the pathway. First, the phenotype caused by overexpression of lon-3 is Dpy rather than Sma. Second, lon-3(0) is not fully epistatic to dbl-1(0). Furthermore, while worms lacking dbl-1 activity show appreciably reduced endoreduplication of hypodermal nuclei, the ploidy of hypodermal nuclei in worms that overexpress lon-3 is only slightly less than wild type (Table 1). This observation suggests that the effect of dbl-1 loss-of-function mutations on the ploidy of hypodermal nuclei is not mediated through LON-3.

The data presented here do not exclude the possibility that LON-3 functions upstream of the TGFß pathway regulating body length. Studies on vertebrates have shown that the activity of some ligands in the TGFß superfamily can be inhibited by decorin, a protein that was first isolated by virtue of its ability to bind to collagen (VOGEL et al. 1984 ; YAMAGUCHI et al. 1990 ). One possibility therefore is that LON-3 is required for the function of a negative regulator of DBL-1. However, the fact that the Sma phenotype caused by dbl-1 null mutants is not fully epistatic to a lon-3 null mutation argues against a model in which lon-3 functions solely upstream of the TGFß pathway, for example, by recruiting an inhibitor of DBL-1. A model for LON-3 as an inhibitor of DBL-1 cannot be excluded, but one in which LON-3 functions exclusively in this way is not consistent with our results: LON-3 may be an inhibitor of DBL-1 but, if so, it must also function either downstream of the pathway or in parallel.

Recent work has shown that dbl-1 negatively regulates the transcription of lon-1 (N. UENO, personal communication). We have shown here that endoreduplication of hypodermal nuclei is increased in a lon-1 mutant. Therefore, the fact that we failed to detect a significant increase in hypodermal ploidy in worms overexpressing dbl-1 is surprising. One possible explanation for this paradox could be that even small amounts of LON-1 can prevent hyperendoreduplication and that the worms used in this study that overexpress dbl-1 do not express high enough levels to reduce LON-1 activity sufficiently to allow extra rounds of endoreduplication to occur. Alternatively, LON-1 might be regulated in more than one way and a second, as yet unidentified, negative regulator of lon-1 might exist that functions in parallel to dbl-1.

Worms that are either mutant for dbl-1 or overexpress the gene appear to have the same number of somatic cells as wild type (SUZUKI et al. 1999 ). Since dbl-1 mutants are both shorter and thinner than wild type, it is thought that at least some cells in dbl-1 mutants are smaller than in wild type. It has been proposed that the decrease in body size seen in mutants defective in TGFß signaling results in part at least from a decrease in the size of the hypodermis caused in turn by reduced ploidy of hypodermal nuclei (FLEMMING et al. 2000 ). These observations raise the question of whether some or all cells are larger in worms that overexpress dbl-1(+) or that are mutant for lon-3. We have made estimates of the volumes of such worms on the basis of images obtained by Nomarski microscopy. Within the limits of the accuracy of this technique, neither lon-3 mutants nor worms that overexpress dbl-1 have volumes that are obviously greater than wild type (J. NYSTRÖM and S. TUCK, unpublished results). Both lon-3 null mutants and worms that overexpress dbl-1 are longer than wild type but also thinner. Since the nuclei in the hypodermis of these worms are not markedly hyperendoreduplicated, our results are consistent with a model in which final body size is determined to a significant extent by the extent of endoreduplication. It will be interesting in the future to test this model rigorously by making accurate measurements of the size of the hypodermis, for example, by electron microscopy. Such measurements would make it possible to determine, first, whether body size is mostly correlated with the size of the hypodermis and, second, whether the size of the hypodermis is determined largely by the degree of endoreduplication of hypodermal nuclei.

A total of 33 genes that can mutate to give rise to alleles conferring Dpy or Sqt phenotypes have been identified in C. elegans. Seven of these (dpy-2, dpy-7, dpy-10, dpy-13, rol-6, sqt-1, and sqt-3) have been shown to encode cuticle collagens (KRAMER et al. 1988 , KRAMER et al. 1990 ; BIRD 1992 ; JOHNSTONE et al. 1992 ; LEVY et al. 1993 ; VAN DER KEYL et al. 1994 ). At least 6 of these (dpy-2, dpy-7, dpy-10, dpy-13, sqt-1, and rol-6) do not cause a Lon phenotype when present in multiple copies (JOHNSTONE et al. 1992 ; YANG and KRAMER 1994 ; GILLEARD et al. 1997 ; J. NYSTRÖM and S. TUCK, unpublished results). (The effect of overexpressing sqt-3 has not been reported.) Thus while these genes may in some cases be required for wild-type body morphology, they do not appear to be able to regulate body length. The C. elegans genome is predicted to encode 150 cuticle collagens. However, the loss-of-function phenotype of the majority of these genes is unlikely to be either Lon or Dpy since screens for mutations conferring these phenotypes appear to be approaching saturation. Multiple alleles of all but two of the dpy genes (dpy-24 and dpy-25) have been isolated. In the screen reported here for new Lon mutations, multiple alleles of lon-1, lon-2, and lon-3 were isolated, together with rare alleles of four other genes, lon-4, lon-5, lon-6, and lon-7 (Z.-Z. SHEN and A. LEROI, unpublished results). In addition, RNAi surveys have yielded just four loci that give a Lon phenotype out of the 4500 currently tested (FRASER et al. 2000 ; MAEDA et al. 2001 ); none appear to be collagens. Just two collagen genes, C31H2.2 and F38B6.5, were found to give rise to a Dpy phenotype (MAEDA et al. 2001 ). C31H2.2 maps very close to dpy-8 and may therefore correspond to this gene.

Thus while we cannot exclude the possibility that other C. elegans cuticle collagens can, like lon-3, cause reciprocal changes in body length when over- or underexpressed, it is unlikely that many have this property. While the C. elegans cuticle is somewhat unusual in that it contains so many collagens (COX et al. 1984 ; KUSCH and EDGAR 1986 ; KRAMER 1994 ), it nevertheless serves as a useful model for the control of morphology (JOHNSTONE 2000 ). Many structures within more complicated organisms are surrounded by collagenous membranes. Perhaps some collagens in these membranes might not simply have a passive role in allowing the membrane to adopt the appropriate shape but may instead actively determine the shape of the structures they surround. It will be interesting in the future to determine how changes in the expression of just one collagen gene in C. elegans can have such pronounced effects on body length.