Genetic Analysis of Caenorhabditis elegans glp-1 Mutants Suggests Receptor Interaction or Competition

Corresponding author: E. Jane Albert Hubbard, New York University, 1009 Silver Center, 100 Washington Sq. E., New York, NY 10003., jane.hubbard{at}nyu.edu (E-mail)

Communicating editor: B. J. MEYER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

glp-1 encodes a member of the highly conserved LIN-12/Notch family of receptors that mediates the mitosis/meiosis decision in the C. elegans germline. We have characterized three mutations that represent a new genetic and phenotypic class of glp-1 mutants, glp-1(Pro). The glp-1(Pro) mutants display gain-of-function germline pattern defects, most notably a proximal proliferation (Pro) phenotype. Each of three glp-1(Pro) alleles encodes a single amino acid change in the extracellular part of the receptor: two in the LIN-12/Notch repeats (LNRs) and one between the LNRs and the transmembrane domain. Unlike other previously described gain-of-function mutations that affect this region of LIN-12/Notch family receptors, the genetic behavior of glp-1(Pro) alleles is not consistent with simple hypermorphic activity. Instead, the mutant phenotype is suppressed by wild-type doses of glp-1. Moreover, a trans-heterozygous combination of two highly penetrant glp-1(Pro) mutations is mutually suppressing. These results lend support to a model for a higher-order receptor complex and/or competition among receptor proteins for limiting factors that are required for proper regulation of receptor activity. Double-mutant analysis with suppressors and enhancers of lin-12 and glp-1 further suggests that the functional defect in glp-1(Pro) mutants occurs prior to or at the level of ligand interaction.

THE GLP-1 protein is a member of the conserved LIN-12/Notch family of receptors. In addition to receptor conservation, many components of the pathway including ligands, proteases, and nuclear components are also highly conserved among metazoans (GREENWALD 1998 ; MUMM and KOPAN 2000 ; BARON et al. 2002 ). The role of LIN-12/Notch-mediated signaling has been analyzed in many systems, and aberrant activity of the receptors and other pathway components has been linked to human disease (JOUTEL and TOURNIER-LASSERVE 1998 ). While much has been learned in recent years regarding the mechanism of LIN-12/Notch-mediated signaling, questions still remain regarding the precise form that the receptor takes at the membrane, how the receptor is modified and processed prior to membrane localization, and how receptor activity is regulated in specific developmental contexts (BARON et al. 2002 ).

Caenorhabditis elegans has two well-characterized receptors in the LIN-12/Notch family that act in binary cell fate decisions: LIN-12 acts in somatic development (GREENWALD et al. 1983 ), and GLP-1 acts in both somatic and germline cell fate decisions (AUSTIN and KIMBLE 1987 ; PRIESS et al. 1987 ; Table 1). These two receptors are functionally interchangeable (FITZGERALD et al. 1993 ) and can utilize the same ligands and effectors (LAMBIE and KIMBLE 1991 ; HENDERSON et al. 1994 ; FITZGERALD and GREENWALD 1995 ; GAO and KIMBLE 1995 ; CHRISTENSEN et al. 1996 ; DOYLE et al. 2000 ; PETCHERSKI and KIMBLE 2000 ).

A general picture for signaling by LIN-12/Notch family receptors has emerged from studies in both vertebrate and invertebrate systems (GREENWALD 1998 ; MUMM and KOPAN 2000 ; WEINMASTER 2000 ; BARON et al. 2002 ). A brief overview follows, with an emphasis on C. elegans (Fig 1). After translation, receptors are transported from the endoplasmic reticulum (ER) to the Golgi, a process that involves the p24 family of proteins (KAISER 2000 ). SEL-9 is a C. elegans p24 protein that is likely involved in quality control at this step (WEN and GREENWALD 1999 ). Once in the Golgi, the receptor is glycosylated; the glycosyltransferase Fringe modifies Notch (HICKS et al. 2000 ; JU et al. 2000 ; MOLONEY et al. 2000 ). The receptor is then cleaved, appearing on the membrane in the form of a heterodimer. This S1 cleavage is ligand independent and is mediated by the furin class of proteases (BLAUMUELLER et al. 1997 ; LOGEAT et al. 1998 ). However, furin-mediated cleavage is not necessary for signaling in all cases examined (BUSH et al. 2001 ; KIDD and LIEBER 2002 ). Although roles for C. elegans Fringe and furin homologs have not been reported, evidence exists for cleavage and glycosylation in this system (CRITTENDEN et al. 1994 ).

View larger version (26K): [in this window] [in a new window] [Download PPT slide]   Figure 1. Composite schematic representation of GLP-1 signaling in the C. elegans germline. Major events of GLP-1 receptor maturation and signaling are represented, including events inferred from work on other organisms. See text for details. Genes used in this study that encode positively acting factors are depicted in green and those that encode negatively acting factors are depicted in red. SEL-10 is depicted in the nucleus on the basis of results obtained with vertebrate Sel-10 (GUPTA-ROSSI et al. 2001 ). The layout of the figure was adapted from BARON et al. 2002 .

Ligand activation by DSL (Delta, Serrate, LAG-2) family ligands leads to a metalloprotease-dependent S2 cleavage, just N-terminal to the transmembrane domain. Specifically, a TNF--converting enzyme metalloprotease is thought to mediate S2 cleavage in mammalian systems, while the ADAM-family metalloprotease Kuzbanian is required for S2 cleavage in Drosophila (PAN and RUBIN 1997 ; BROU et al. 2000 ; MUMM et al. 2000 ; LIEBER et al. 2002 ). In C. elegans, genetic evidence is consistent with the Kuzbanian-related SUP-17 playing a similar role (WEN et al. 1997 ). Subsequent to S2 cleavage, a presenilin-dependent cleavage (S3) occurs within the transmembrane domain of the receptor, detaching the intracellular domain from the membrane and allowing translocation to the nucleus (STRUHL and GREENWALD 1999 ). Two C. elegans presenilins, SEL-12 and HOP-1, are functionally redundant (LEVITAN and GREENWALD 1995 ; LI and GREENWALD 1997 ; WESTLUND et al. 1999 ). Once the intracellular domain of the receptor reaches the nucleus, target gene activity is altered by the action of the intracellular domain in a complex with other proteins, including a CSL family member (CBF1, Suppressor of Hairless, LAG-1). All LIN-12 and GLP-1 signaling analyzed to date is dependent on LAG-1 (CHRISTENSEN et al. 1996 ), although there are reports of CSL-independent Notch signaling in other systems (BRENNAN and GARDNER 2002 ). Finally, receptor signaling can be terminated by ubiquitination. sel-10 encodes a conserved F-box-containing protein that interacts with and facilitates ubiquitination of the intracellular domain of LIN-12/Notch proteins in both C. elegans and mammals (HUBBARD et al. 1997 ; GUPTA-ROSSI et al. 2001 ; OBERG et al. 2001 ; WU et al. 2001 ).

In the C. elegans germline, GLP-1 signaling mediates the mitosis/meiosis decision. GLP-1 activity is associated with mitosis and/or inhibition of meiosis. Loss of glp-1 causes a severe germline proliferation defect and premature entry into meiosis (AUSTIN and KIMBLE 1987 ). One gain-of-function allele, glp-1(oz112gf), causes the opposite germline phenotype: persistent mitosis and failure to enter meiosis (BERRY et al. 1997 ). This gain-of-function phenotype is referred to as tumorous (Tum), and this allele behaves as a genetic hypermorph. For glp-1(oz112gf), lowering the temperature or lowering the dosage of glp-1 does not alter the initial larval pattern of germline development but results in a highly penetrant late-onset Tum phenotype in which the distal mitotic region expands as the animals age (BERRY et al. 1997 ).

We have characterized three gain-of-function glp-1 mutations that cause an array of germline defects consistent with elevated GLP-1 activity but that are phenotypically and genetically distinct from the previously characterized glp-1(Tum) allele. These three alleles display a proximal proliferation (Pro) phenotype that is characterized by ectopic germline proliferation in the proximalmost region of the adult C. elegans germline (SEYDOUX et al. 1990 ; WESTLUND et al. 1997 ). Since all three alleles display similar genetic behavior and each displays the Pro phenotype, we refer to them collectively as glp-1(Pro) alleles. Each glp-1(Pro) allele encodes a single amino acid substitution near proposed sites of extracellular proteolytic cleavage and protein-protein interactions. Although these glp-1(Pro) alleles cause gain-of-function phenotypes, they are genetically distinct from other mutations in LIN-12/Notch family genes that affect the same region of the protein. The alleles are dose dependent for germline pattern defects and, surprisingly, are suppressed by wild-type doses of glp-1. Moreover, the mutant phenotype of a trans-heterozygous combination of the two highly penetrant glp-1(Pro) alleles is suppressed compared to either homozygous mutant. These results support a model in which GLP-1 receptors normally form a dimer or other higher-order multimer and/or they compete for localized interacting factors. Finally, interactions with characterized GLP-1/LIN-12/Notch pathway components and modifiers suggest that these mutant receptors act within the canonical signaling pathway and that the phenotypes they cause are due to defects prior to or at the level of ligand interaction.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Strains and genetic manipulations:
Strains were maintained and constructed using standard genetic techniques (BRENNER 1974 ). All strains were raised at 15° and shifted to 25° for analysis unless otherwise noted. All strains were derived from C. elegans var. Bristol strain N2 and described in BRENNER 1974 or as cited:

• LG I: dpy-5(e61), sup-17(n1258) (TAX et al. 1997 ).

• LG III: dpy-17(e164), ncl-1(e1865) (HEDGECOCK and WHITE 1985 ), unc-36(e251), unc-32(e189), lin-12(n676n930) (SUNDARAM and GREENWALD 1993A ), glp-1(ar202), glp-1(ar218), glp-1(ar224), glp-1(q175, q46, e2141) (AUSTIN and KIMBLE 1987 ), mog-1(q223) (GRAHAM and KIMBLE 1993 ), unc-69(e587) (SIDDIQUI 1990 ).

• LG IV: dpy-13(e184), unc-5(e53) (HEDGECOCK et al. 1990 ), lag-1(q385, q426) (LAMBIE and KIMBLE 1991 ), unc-24(e138) (WATERSTON et al. 1980 ).

• LG V: lag-2(q411, q420) (LAMBIE and KIMBLE 1991 ), dpy-11(e224), sel-9(ar22) (SUNDARAM and GREENWALD 1993B ), sel-10(ar41) (SUNDARAM and GREENWALD 1993B ), him-5(e1490) (HODGKIN et al. 1979 ), ego-3(om40) (QIAO et al. 1995 ), unc-76(e911) (HEDGECOCK et al. 1987 ).

• LG X: sel-12(ar131, ar171) (LEVITAN and GREENWALD 1995 ), unc-1(e719) (PARK and HORVITZ 1986 ).

• Rearrangements and duplications: eT1(III;V) (ROSENBLUTH and BAILLIE 1981 ), nT1[unc-?(n754) let-?] (IV;V) (FERGUSON and HORVITZ 1985 ) [in the derivative of nT1, unc-?(n754) confers a dominant Unc phenotype and let-? confers a recessive lethal phenotype, referred to as "DnT1" (E. FERGUSON, unpublished data)], qDp3 (III;f) (AUSTIN and KIMBLE 1987 ), mnDp68 (X;f) (HERMAN and KARI 1989 ).

Special notes regarding glp-1(Pro) mutants:
All three glp-1(Pro) alleles show extreme sensitivity to marker mutations (e.g., enhancement, suppression, or variation in phenotype). Also, the original glp-1(ar224) strain was markedly slow growing, sickly, and exhibited an incompletely penetrant "small germline" phenotype. After multiple backcrosses, the mutation was recombined onto several marked chromosomes and the health of the strain markedly improved. The ar224 mutation was subsequently reisolated from a healthier marked strain several times independently and the original slow-growing, sickly, and small germline phenotypes reappeared. In heterozygous combinations with marked chromosomes, the phenotypes do not appear. These results suggest that these phenotypes are a property of either ar224 or a closely linked mutation and are dominantly suppressed by (otherwise recessive) markers. We have scored the phenotype of this homozygous mutant in a strain that is m ar224/n ar224, where m and n refer to linked recessive marker mutations.

Special considerations for glp-1(Pro) strain constructions:
In all cases where the resulting strain was expected to be sterile at the restrictive temperature due to a glp-1(Pro) mutation, presumptive homozygous animals were cultured individually at 15°. Some of their progeny were reared at 15° while others were shifted to 25°. Therefore, these strains were verified by their phenotype at 25° and were established from the siblings at 15°. Low penetrance of glp-1(ar218) (Table 2) and the maternal rescue of the glp-1(ar202) and glp-1(ar224) defects (Table 4) ensured unambiguous identification of the homozygous strains carrying these alleles since self-progeny of homozygous mothers display the mutant phenotype whereas homozygous self-progeny of heterozygous mothers do not. Where identical strategies were employed in the construction of strains bearing different glp-1(Pro) mutant alleles, they are described below as glp-1(Pro). Allele designations are as noted above and are omitted from the strain constructions except when necessary to avoid ambiguity. Progeny analysis was used in all cases where self-progeny and cross-progeny were not immediately distinguishable by phenotype. Strains for which construction details are not provided were built using standard methods (BRENNER 1974 ).

Strains relevant to dosage analysis:
dpy-17 ncl-1 unc-36 glp-1(Pro); qDp3: Three steps were employed to construct these strains. First, ncl-1 unc-36 glp-1(q46); qDp3 hermaphrodites were mated with glp-1(Pro) males. Non-Unc cross-progeny hermaphrodites were picked to individual plates and their Ncl, Unc, non-Glp recombinant progeny were picked onto individual plates [presumed genotype: ncl-1 unc-36 glp-1(Pro)/ncl-1 unc-36 glp-1(q46)]. Animals homozygous for the recombinant chromosome were identified in the next generation, and the strain was verified by temperature shift. Second, these ncl-1 unc-36 glp-1(Pro) hermaphrodites were mated with dpy-17unc-32/++ males and the progeny of their non-Unc F₁ hermaphrodites were analyzed to establish a ncl-1 unc-36 glp-1(Pro)/dpy-17 unc-32 strain. Dpy non-Unc recombinant progeny [presumed genotype: dpy-17 ncl-1 unc-36 glp-1(Pro)/dpy-17 unc-32] were picked onto individual plates. Animals homozygous for the recombinant chromosome were identified, and the strain was verified by temperature shift and by Nomarski analysis for the Ncl phenotype. Finally, glp-1(Pro) males were mated with dpy-17 ncl-1 unc-36 glp-1(Pro) hermaphrodites. F₁ males were mated with ncl-1 unc-36 glp-1(q46); qDp3 hermaphrodites. Non-Unc hermaphrodite progeny [presumed genotype: dpy-17 ncl-1 unc-36 glp-1(Pro)/ncl-1 unc-36 glp-1(q46); qDp3 or ncl-1 unc-36 glp-1(q46)/glp-1(Pro); ±qDp3] were picked onto individual plates. Dpy non-Unc hermaphrodite progeny [presumed genotype: dpy-17 ncl-1 unc-36 glp-1(Pro); qDp3] were picked onto individual plates and tested by temperature shift and progeny analysis to establish the strain.

glp-1(Pro) unc-x/unc-y or glp-1(Pro) unc-x/eT1 III; +/eT1 V: unc-y /+ or +/eT1 III; +/eT1 V males were mated with glp-1 (Pro) unc-x hermaphrodites and non-Unc hermaphrodite cross-progeny were picked onto individual plates. F₁ animals of the desired genotype were identified by progeny analysis. Continued linkage of glp-1(Pro) to unc-x was verified by shifting progeny of Unc-x animals.

glp-1(Pro) unc-69/unc-32 glp-1(q175) and glp-1(Pro) unc-69/eT1 III; +/eT1 V: glp-1(Pro) unc-69 hermaphrodites were mated with unc-32 glp-1(q175)/eT1 III; him-5/eT1[him-5] V males. Hermaphrodite cross-progeny [presumed genotype: glp-1(Pro) unc-69/unc-32 glp-1(q175); him-5/+ or glp-1(Pro) unc-69/eT1 III; +/eT1[him-5] V] were picked onto individual plates and the two strains were identified by progeny analysis. To eliminate him-5 from the first strain, 20 non-Unc hermaphrodites were picked from plates segregating Glp, Unc-32, and Unc-69 animals. From each hermaphrodite mother that did not segregate males (presumed genotype +/+ or him-5/+), 12 hermaphrodite progeny were picked onto individual plates, and the self-progeny were inspected for the absence of males from all 12 broods.

Strains relevant to glp-1(Pro)/glp-1(Pro) trans-heterozygote analysis:
glp-1(Pro) unc-x/glp-1(Pro) unc-y: glp-1(Pro) unc-x/++ males were mated with glp-1(Pro) unc-y hermaphrodites and non-Unc F₁ progeny were picked onto separate plates to establish the strain. The germline phenotypes of all three classes (Unc-x, Unc-y, and non-Unc) of F₁ progeny of heterozygous mothers were scored.

Strains relevant to glp-1(Pro); lag double mutants:
glp-1(ar202); lag-1(q385)/DnT1: First, glp-1(ar202); dpy-13/+ males were mated with ego-3 unc-76/DnT1 hermaphrodites. Unc F₁ hermaphrodite progeny were picked onto individual plates and the glp-1(ar202); dpy-13/DnT1 strain was established by progeny analysis for glp-1(ar202) homozygotes that segregated only Dpy non-Unc and Unc non-Dpy animals. Unc hermaphrodites from this strain were then mated with lag-1 (q385)/dpy-13 unc-24 males. Unc non-Dpy hermaphrodites were picked onto individual plates and the final strain was established by progeny analysis for Unc glp-1(ar202) homozygotes.

glp-1(ar202); lag-1(q426)/DnT1: glp-1(ar202); dpy-13/DnT1 hermaphrodites were mated with lag-1(q426)/unc-5 males. Unc non-Dpy F₁ hermaphrodites were picked onto individual plates and the strain was established by progeny analysis for glp-1 (ar202) homozygotes that segregated Unc non-Glp and Glp non-Unc animals.

glp-1(ar202); lag-2(q411)/DnT1: glp-1(ar202); dpy-11/+ males were mated with lag-2(q411)/DnT1 hermaphrodites. Non-Unc male F₁ progeny (glp-1/+; dpy-11/lag-2 or glp-1/+; +/lag-2) were mated with glp-1(ar202); +/DnT1 hermaphrodites. Individual Unc hermaphrodite progeny were picked onto separate plates and selfed. Progeny of individuals that segregated only Unc worms [presumed genotype: glp-1(ar202); lag-2 (q411)/DnT1, glp-1(ar202)/+; lag-2 (q411)/DnT1 or +/+; lag-2 (q411)/DnT1] were picked onto separate plates and their progeny were shifted to the restrictive temperature to identify glp-1(ar202) homozygotes.

unc-32 glp-1(ar202); lag-2(q420): lag-2(q420) males were mated with unc-32 glp-1(ar202) hermaphrodites. Non-Unc hermaphrodite cross-progeny were picked onto individual plates [presumed genotype: unc-32 glp-1(ar202)/+; lag-2(q420)/+]. F₂ animals homozygous for lag-2(q420) were identified by temperature shift. Homozygous glp-1(ar202) animals were identified in the next generation. The resulting strain was sequenced at the lag-2 locus to verify the presence of lag-2(q420).

Double-mutant strains carrying glp-1(Pro) with sel genes and sup-17:
sel-12 double mutants and associated control strains: dpy-17/+ males were mated with sel-12 unc-1; mnDp68 [sel-12(+) unc-1(+)] and unc-1 hermaphrodites. Non-Unc hermaphrodite F₁ progeny were picked onto individual plates. From plates segregating Dpy, Unc, and Dpy Unc hermaphrodites, Dpy Unc hermaphrodites were picked onto individual plates to establish the strains dpy-17; sel-12 unc-1 and dpy-17; unc-1. These hermaphrodites were mated with glp-1(Pro) males. Individual non-Dpy, non-Unc F₁ hermaphrodites were picked and the glp-1(Pro); sel-12 unc-1 and glp-1(Pro); unc-1 strains were established from their F₂ progeny by progeny testing and temperature shift.

sel-10 double-mutant strain and associated control: ncl-1 unc-36 glp-1(ar202) or ncl-1 unc-36 glp-1(ar202); him-5 hermaphrodites were mated with glp-1(ar202) males. F₁ males were mated with unc-32; sel-10 him-5 or unc-32. Non-Unc F₁ hermaphrodites were picked onto individual plates. From plates that segregated both Unc-32 and Unc-36, non-Unc F₂ individuals were picked and their progeny inspected for the presence of males to establish the strains ncl-1 unc-36 glp-1(ar202)/unc-32; sel-10 him-5 and ncl-1 unc-36 glp-1(ar202)/unc-32; him-5.

sup-17; glp-1(ar202): glp-1(ar202) males were mated with dpy-5; glp-1(ar202) hermaphrodites. Non-Dpy cross-progeny males were mated with sup-17; unc-32 hermaphrodites. Non-Unc hermaphrodite cross-progeny were picked onto individual plates and from plates that segregated Dpy animals, non-Dpy non-Unc animals were picked onto individual plates and the strain was established from F₂ progeny that segregated neither Dpy nor Unc.

glp-1(Pro); sel-9: glp-1(Pro); dpy-11/+ males were mated with unc-32 lin-12; sel-9 hermaphrodites and non-Unc F₁ progeny were picked onto individual plates. F₁ progeny that segregated Dpy, Unc, Dpy Unc, and non-Dpy non-Unc progeny were retained. Homozygous sel-9 mutants were identified first by selecting non-Dpy animals that segregated no Dpy progeny but segregated Unc and non-Unc progeny. From their self-progeny, homozygous glp-1 animals were identified by absence of Unc-32 self-progeny.

Isolation and identification of glp-1(Pro) alleles:
Mutageneses were carried out as described (BRENNER 1974 ) except that 25 mM EMS was used to reduce the level of additional second-site mutations that cause fertility defects (in one mutagenesis 50 mM was used). Mutagenized hermaphrodites and their F₁ progeny were raised at 15°, and the self-progeny of individual F₁ hermaphrodites were transferred to 25° as late embryos or L1 larvae while siblings were kept at 15°. F₂ animals were screened first under low magnification for the absence of embryos in the uterus and then under high magnification for a more specific assessment of the fertility defect. Three of the six Pro mutants that displayed an apparently normal somatic gonad mapped to the same genetic interval on linkage group III. For each allele, linkage to III was determined by standard genetic techniques (BRENNER 1974 ) and/or by mapping using polymorphic sequence tagged sites (STS; WILLIAMS et al. 1992 ). STS mapping was also used in some cases to give an indication of position within LG III. Recombinants were isolated from a dpy-17 unc-32/glp-1(Pro) strain; for each allele (n = 9–19 recombinants in each direction) all Dpy non-Unc recombinants were glp-1(Pro)/+ and all Unc non-Dpy recombinants were +/+. Unc non-Mog recombinants were also selected from mog-1 unc-69/glp-1(Pro) and each recombinant was also glp-1(Pro)/+. All three alleles mapped within the same interval on LG III—between unc-32 and mog-1.

The entire glp-1 coding region and all splice junctions were sequenced on ABI 377 and ABI 3700 genetic analyzers. Primers were designed to amplify six large segments of glp-1 genomic DNA from worm lysates. Direct sequencing was performed on the forward strand of these PCR products with nested primers. Where changes were found, the reverse strand was also sequenced to verify the change. A single missense base-pair change was found in each mutant as follows: ar202 G529E (codon change GGG to GAG), ar218 R499Q (codon change CGA to CAA), and ar224 A729T (codon change GCT to ACT).

Synchronization of worms for scoring germline phenotypes:
For phenotypic analysis of all strains the following protocol was used unless noted. Animals were grown at 15° and synchronized using a hatch-off protocol adapted from FRANCIS et al. 1995 . All larvae and adults were washed off mixed-stage nematode growth medium (NGM) plates with M9 buffer (WOOD 1988 ). Plates were inspected under low-power magnification to confirm that only embryos remained on the bacterial lawn. Embryos were then allowed to hatch for 2 hr at 15°. All newly hatched L1 larvae were washed (with M9) into 15-ml Falcon tubes and centrifuged at 2000 rpm for 4 min. Supernatant was removed, and larvae were collected in a glass Pasteur pipette and distributed to 60-mm plates (50 larvae/plate) that had been seeded with OP50 bacteria (WOOD 1988 ). Plates of synchronous larvae were placed at 25°, and the animals were harvested 48 hr later (young adult stage) for fixation and scoring. All data presented here were obtained by examining worms under x400–1000 magnification after fixation and staining with 4',6-diamidino-2-phenylindole (DAPI). In all cases, n is the number of gonad arms scored. In cases where F₁ progeny of a heterozygous strain were scored, 20–30 heterozygous adults were placed on one plate and allowed to lay eggs overnight. F₁ progeny were then synchronized by hatch-off. When necessary, worms of different genotypes were separated as L4s.

Fixation and staining protocols:
Worms were washed off of growth plates with M9 buffer into siliconized 1.5-ml Eppendorf tubes and briefly centrifuged and the supernatant was removed. Animals were then resuspended in 95% ethanol and fixed for 10 min, by which time animals had settled to the bottom of the tube. Ethanol was then removed and a drop of Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, CA) was dropped onto the worms. Worms were then transferred to 5% agar pads using a glass Pasteur pipette and covered for microscopy. These preparations could be kept in the dark at room temperature for up to 4 days without losing resolution.

RNAi methods:
RNAi feeding experiments were performed using the system developed by Timmons and Fire (TIMMONS et al. 2001 ) that employs the HT115(DE3) bacterial strain on NGM plates containing 100 µm/ml ampicillin and 1 mM isopropyl thiogalactoside. Bacteria were grown overnight on LB plates containing 100 µg ampicillin and 50 µg/ml tetracycline. Single colonies were picked and grown overnight in LB with 100 µg/ml ampicillin. Plates were seeded immediately from the overnight culture and were kept at room temperature in the dark for 1–2 days before worms were added. pGC11 contains 720 bp of hop-1 genomic DNA inserted into the L4440 vector. The insert was prepared by PCR (primer 1, TCCCATTCCTAACCGAATTG; primer 2, GAAACTCAGCCAGCCAGAAC) from N2 worms. The PCR product was subsequently digested with XbaI and AccI and ligated into the same sites in the vector. pGC2, a feeding construct containing the lag-1 cDNA, was used in parallel as a positive control for the RNAi conditions. Three L4 worms were placed on each HT115-seeded plate and then transferred after a short time to remove any OP50 bacteria that remained on the worms. Worms were maintained at 15° and then transferred onto fresh RNAi plates each day for 3 days. After each transfer, the plates containing eggs and larvae were placed at 25° while the adults were maintained at 15°. Progeny from the third-day transfer were scored after 2 days at 25°.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Identification of Pro mutants:
We performed a genetic screen for mutations that affect development of the germline. A temperature shift was incorporated into the screening strategy (see MATERIALS AND METHODS) to permit the isolation of mutations in essential genes that also affect gonadogenesis. One class of mutants we identified in this screen displays a Pro phenotype in which proliferative germ nuclei are present in the proximal gonad—between mature gametes and the proximal somatic gonad of the adult (Fig 2; SEYDOUX et al. 1990 ). Normally, the adult proximal germline contains germ cells undergoing gametogenesis and proliferation is restricted to the distalmost part of the adult gonad. Within the first 10,000 haploid genomes screened, >30 mutants were found that appeared to exhibit a Pro phenotype. Of these, 6 Pro mutants exhibited grossly normal somatic gonad development (the somatic gonad in the others appeared abnormal), and these 6 were characterized further. Of the 6, genetic mapping data indicated that three alleles mapped within the same interval (between unc-32 and mog-1) on LG III (see MATERIALS AND METHODS). These three proved to carry mutations in the glp-1 locus (other non-glp-1 Pro mutants will be described elsewhere).

View larger version (60K): [in this window] [in a new window] [Download PPT slide]   Figure 2. Summary of glp-1 mutant germline pattern defects. Yellow indicates zone of mitosis, green indicates transition (leptotene and zygotene) and pachytene of meiosis I, blue denotes late prophase of meiosis I and gametogenesis, gray shows the position of the proximal somatic gonad, and the distal tip cell is drawn in red. (A, top) Nomarski image of a live adult wild-type hermaphrodite. Anterior is to the left and ventral is down. The posterior arm of the gonad is visible in this orientation and is outlined in black. Bar, 50 µm. (A, right) Schematic representation of the four major adult glp-1 germline phenotypes. The white space in the glp-1(null) represents a gonad containing few germ cells, all of which are sperm. (A, left) Photomicrographs of dissected DAPI-stained wild-type and glp-1(Pro) gonads. The asterisk indicates the distal end of each gonad arm. The colored brackets correspond to the stages of germline development depicted in the diagrams on the right. Bar, 50 µm. (B) Schematic representation of minor adult germline pattern defects observed in glp-1(Pro) mutant animals. The minor phenotypes are variable with respect to the size and position of each cluster of nuclei; diagrams depict only overall patterns.

glp-1(Pro) alleles are missense mutations that alter the extracellular domain of the receptor:
Direct sequence analysis revealed that each glp-1(Pro) strain harbored a missense mutation in the glp-1 coding region, just N-terminal to the transmembrane domain (Fig 3). Two of the mutations, ar218 (R499Q) and ar202 (G529E), change a single residue within the LIN-12/Notch repeats (LNR), while the third changes a residue C-terminal to a pair of conserved cysteine residues, ar224 (A729T). The first two mutations affect residues that are conserved among nematode GLP-1 sequences, but not between phyla (RUDEL and KIMBLE 2001 ). Nonetheless, both changes are in regions that are quite well conserved and both are nonconservative substitutions. In particular, ar202 (G529E) may disrupt the structure of the loop between the first two LNRs (ASTER et al. 1999 ). The third mutation, ar224 (A729T), alters a well-conserved amino acid near the putative site for ligand-dependent S2 cleavage (Fig 3; BROU et al. 2000 ; MUMM et al. 2000 ). All of these mutations affect a region of the protein that is thought to negatively regulate receptor activity in the absence of ligand binding (GREENWALD and SEYDOUX 1990 ; LIEBER et al. 1993 ; GREENWALD 1994 ; BRENNAN et al. 1997 ).

View larger version (17K): [in this window] [in a new window] [Download PPT slide]   Figure 3. Schematic diagram of GLP-1 receptor and positions of missense changes encoded by glp-1(Pro) alleles. The positions of conserved motifs are indicated for the EGF-like repeats, LIN-12/Notch repeats (LNR), two cysteine residues (CC), the transmembrane domain (TM), and the cdc10/SWI6/ankyrin repeats (ANK). The corresponding positions of cleavage sites S1 and S2 are also indicated (MUMM and KOPAN 2000 ). Local alignments in the regions affected by the three glp-1(Pro) mutations are given for three related proteins: LIN-12 from C. elegans, Notch from Drosophila melanogaster, and Notch1 from Homo sapiens. Alignment of the LNRs is based on ASTER et al. 1999 . The second alignment begins with the C-terminal cysteine of the conserved pair of cysteines depicted in red.

The glp-1(Pro) mutants display temperature-sensitive germline pattern defects:
GLP-1 activity in the C. elegans germline promotes mitosis and/or inhibits meiosis. glp-1 mutant germline phenotypes are summarized in Fig 2. Null, strong, and partial loss-of-function alleles cause a defect in germline proliferation (Glp) and early entry into meiosis (AUSTIN and KIMBLE 1987 ). A hypermorphic gain-of-function allele, glp-1(oz112gf), can cause a Tum phenotype in which all germ cells remain mitotic and never enter meiosis (BERRY et al. 1997 ).

The early adult Pro phenotype in our glp-1 mutants is characterized by a conspicuous ectopic mass of proliferative germ nuclei in the proximal ovary between gametes and the spermatheca. Excess germline mitosis is consistent with an elevation of GLP-1 activity. In contrast to the Tum phenotype of glp-1(oz112gf), however, the ectopic mitosis of the early adult Pro animals is anatomically limited to the proximal part of the germline, and the distal-to-proximal pattern of germline development appears quite normal (Fig 2). The proximal disruption of germline development in Pro animals is apparent in the late-larval germline, and further studies indicate that this Pro phenotype is due to a delay and mis-positioning of the earliest onset of meiosis in larvae (A. PEPPER, T.-W. LO and E. J. A. HUBBARD, unpublished observations). The glp-1(Pro) mutants are qualitatively different from the Tum mutant glp-1(oz112gf) in several respects. Lowering the temperature or dosage of glp-1(oz112gf) results in a completely penetrant late-onset Tum phenotype (a phenotype shared by glp-1(Pro) mutants; see below), but <1% Pro. Another phenotypic difference between glp-1(Pro) mutants and glp-1(oz112gf) is that the latter display a Multivulva phenotype in addition to the Tum phenotype, whereas the vulva of glp-1(Pro) mutant animals appears normal.

All three new glp-1 alleles, ar202, ar218, and ar224, are temperature sensitive, but they behave differently with respect to temperature dependence and penetrance of the Pro phenotype. In our analysis, we strictly defined the Pro phenotype as the presence of mitotic germ nuclei proximalmost in the early adult germline, proximal to mature gametes, and distal to the spermatheca. Experiments presented here were designed to maximize our ability to score the Pro phenotype since a highly penetrant Pro phenotype has not been previously observed for glp-1. The Pro phenotype is spatially invariant and distinct. A time-course analysis was undertaken to determine the best time to score the Pro phenotype. Early time points underestimated the penetrance of the Pro phenotype since meiotic entry is delayed in animals that display the Pro phenotype (A. PEPPER, T.-W. LO and E. J. A. HUBBARD, unpublished observations). Later time points also underestimated and possibly confounded the scoring of the Pro phenotype since older animals sometimes expelled the proximal mitotic cells. Therefore, we chose 48 hr (that is, 48 hr after synchronized early L1 larvae were shifted to the restrictive temperature) for our analysis. Scoring animals at this time point in the early adult, we observed a very low penetrance in glp-1(ar218), while the other two alleles were highly penetrant for the Pro phenotype at the restrictive temperature (Table 2).

In addition to the Pro phenotype, we observed several non-Pro germline pattern defects in early adult glp-1(Pro) mutants. A slight but highly penetrant extension of the distal mitotic zone beyond the usual 20–25 cell diameters is found in early adults; this extension progresses during adulthood (see below). Two additional classes of germline pattern defects were observed sporadically in early adults and were designated Class A and Class B (Fig 2). A dramatic extension of the distal mitotic zone, together with gametes in the proximalmost part of the germline, defines Class A. Discrete (but variable) patches of meiotic nuclei in the distal germline define Class B. Combinations of "Class A + Class B" and "Pro + Class B" were also observed and are noted in the tables and depicted in Fig 2. The first-described Pro mutants, lin-12(loss-of-function), did not display additional germline pattern defects (SEYDOUX et al. 1990 ). Unlike the glp-1(Pro) mutant phenotype, however, the lin-12(loss-of-function) Pro phenotype is a secondary consequence of defects in the somatic gonad rather than of the germline-autonomous effects of glp-1 mutants, and the activity of glp-1 in the germline likely accounts for the different observations.

Like the Pro phenotype, both classes of non-Pro phenotypes are consistent with elevated glp-1 activity since both exhibit an excess of mitotic nuclei. The genotypes of strains in which non-Pro germline pattern defects appear do not offer a simple interpretation regarding the level of glp-1 activity compared to that of the Pro phenotype. The non-Pro phenotypes appear in the presence of specific marker mutations, in certain genetic combinations, and under certain temperature-shift conditions. They may reflect subtle differences in maternal and zygotic levels of glp-1 activity, as well as distinct late-onset defects (see below). Thus, although the Pro phenotype is the primary and most prevalent phenotype, due to the presence of other defects, the degree to which glp-1 activity is normal or elevated is best assessed by the penetrance of the wild-type germline pattern.

To investigate the possibility of late-onset germline pattern defects, we examined the germlines of late-adult glp-1(ar202) animals. We determined the penetrance of Pro and non-Pro defects and the extent of distal mitotic zone expansion at 60 and 72 hr post-shift at 25°. Our results indicate that the penetrance of the Class B phenotype and the size of the distal mitotic zone increase over time. At 48 hr, the Class B phenotype was not observed (n = 419; Table 2). Of 47 gonad arms scored at 60 hr, 37 displayed a Pro phenotype and 13 of these were Pro + Class B. In addition, 10/47 displayed an apparent Class A phenotype, 4 of which were Class A + Class B. Similar results were obtained at 72 hr. While it is difficult to make firm conclusions regarding the scoring of the Class A phenotype at the later time points (since an early adult Pro animal that later expelled proximal mitotic cells might be indistinguishable from Class A), the increased Class B penetrance suggests that the Class B phenotype is fundamentally a late-onset phenotype that is separable from the Pro phenotype. This possibility may also explain why the two phenotypes are affected differently by genetic marker mutations. Data obtained from the later time points also indicate that the distal mitotic zone expands significantly in the background of all observed pattern defects (Pro and non-Pro) from 48 to 60 hr. We quantitated the extent of the distal mitotic zone by counting the distance in cell diameters from the distal tip to the first full ring of transition nuclei. Whereas the distal mitotic zone of glp-1(ar202) extends an average of 25.8 ± 3.1 (standard deviation) cell diameters from the distal tip at 48 hr (n = 11), the zone extends an average of 102.6 ± 61.0 cell diameters at 60 hr (n = 10). By 72 hr (n = 63 arms), 17% of glp-1(ar202) gonad arms display a completely Tum phenotype, likely the result of further extension of the distal mitotic zone. It is likely that the Class B phenotype is a variant of the distal zone extension phenotype in which mitosis is not maintained in the Class B animals. Thus, our glp-1(Pro) alleles are phenotypically distinct from glp-1(oz112gf) in that they display a highly penetrant Pro phenotype, but are similar to glp-1(oz112gf) in that they display a more or less continuous late-onset extension of the distal mitotic zone (late-onset Tum).

glp-1(Pro) alleles display unusual genetic behavior:
Genetically, the glp-1(oz112gf) mutation behaves in a classic hypermorphic fashion: addition of a wild-type dose of glp-1 to the homozygote (oz112/oz112/+) increases the penetrance of the Tum phenotype. For glp-1(oz112gf), lowering glp-1 dosage or lowering the temperature results in a highly penetrant late-onset Tum phenotype, not in a Pro phenotype. glp-1(oz112gf)/glp-1 (null), however, does cause a very low penetrance (1%) Pro phenotype at 20° (BERRY et al. 1997 ). Thus, one hypothesis is that our glp-1(Pro) mutations are simply weak hypermorphs and that the consequence of increasing GLP-1 activity in the germline is either Pro or Tum, depending on the level of receptor activity. In this case, we would expect the Pro mutant phenotype to be enhanced with increasing doses of glp-1(+) and suppressed by lower doses of glp-1(+).

To test this hypothesis, we performed a dosage analysis, examining the penetrance of mutant germline pattern phenotypes at the 48-hr time point at the restrictive temperature in strains with varying doses of glp-1(Pro) and glp-1(+) (Table 3). We used the glp-1(q175) null allele that encodes a severely truncated product; similar results were obtained with a temperature-sensitive partial loss-of-function allele, glp-1(e2141) (data not shown).

The results of the dosage analysis indicate that the Pro phenotype is highly dosage sensitive and that the alleles are not hypermorphic. A comparison of the penetrance of the wild-type and mutant phenotypes of glp-1(Pro)/glp-1(null) relative to the glp-1(Pro) homozygote reveals that the highly penetrant alleles, ar202 and ar224, are extremely sensitive to dosage (Table 3). Interestingly, the glp-1(Pro)/glp-1(null) animals still displayed a low penetrance of the Pro phenotype while glp-1(Pro)/glp-1(+) animals were completely wild type (Table 3, lines 6, 7, 13, 14). These results suggest that the presence of glp-1(+) interferes with the expression of the Pro phenotype. Since both Pro/null and Pro/+ have one copy of the mutant allele in both the mother and the scored progeny, dosage sensitivity alone does not account for these results. Further analysis with strains bearing one extra copy of glp-1(+) indicates that, rather than enhancing mutant phenotypes, glp-1(+) suppressed glp-1(Pro) mutant phenotypes (Table 3, lines 3, 4, 8, 9, 15, 16). The non-Pro germline pattern defects appeared to be more sensitive than the Pro phenotype to suppression by an extra copy of glp-1(+) in these experiments. Even at lower temperatures, one additional copy of glp-1(+) did not enhance the Pro mutant phenotype (Table 3, lines 10, 11, 17, 18). Taken together, we infer that the antagonism of glp-1(Pro) mutant phenotypes by glp-1(+) is a general property of these alleles. This behavior is not allele specific, since we observed a similar trend in all three alleles.

In summary, (i) for the highly penetrant alleles, the penetrance of the wild-type phenotype was higher in Pro/null than in Pro/Pro, but, surprisingly, lower than in Pro/+, and (ii) Pro/Pro/+ exhibited a higher percentage of wild type than Pro/Pro, but a lower percentage than Pro/+. Thus, although the Pro mutants display a gain-of-function phenotype as indicated by increased and ectopic germline mitosis, the glp-1(Pro) alleles are neither hypermorphic nor neomorphic in the classic sense. Rather, the glp-1(Pro) alleles are dosage sensitive and glp-1(+) appears to compete with glp-1(Pro), lowering or correcting the apparent elevation of glp-1 activity as the dosage of glp-1(+) is increased. This genetic behavior is consistent with competition between glp-1(+) and glp-1(Pro) and is inconsistent with the hypothesis that glp-1(Pro) alleles are simply weak hypermorphs.

Maternal effects of glp-1(Pro):
Previous analysis of glp-1 function indicated that maternal glp-1 is required for embryogenesis and zygotic glp-1 is required for germline development (AUSTIN and KIMBLE 1987 ; PRIESS et al. 1987 ). Our results indicate a role for maternal glp-1 in the germline. Consistent with our observation that the mutant phenotype of glp-1(Pro) alleles is highly sensitive to dosage and is antagonized by glp-1(+), we observed a strong maternal rescue of the glp-1(Pro) mutant phenotype of glp-1(Pro) homozygotes coming from heterozygous mothers (Table 4). The results for glp-1(ar202) indicate that the penetrance of the wild-type germline pattern was higher in ar202/ar202 animals derived from heterozygous (ar202/+) mothers than in animals derived from homozygous (ar202/ar202) mothers (Table 4, lines 1 and 2). We observed a similar phenomenon with glp-1(ar224) (data not shown). We infer that the maternal rescue is likely due to lower maternal dosage of the mutant allele rather than to antagonism by glp-1(+) since the penetrance of the wild-type germline pattern of ar202/ar202 animals derived from ar202/ar202/+ mothers was similar to that of ar202/ar202 animals derived from ar202/ar202 mothers (Table 4, lines 3 and 4). A minor caveat to this interpretation is that we cannot score strains with exactly equivalent cis markers in both mothers and progeny of strains with and without the duplication (Table 4). We also observe a maternal effect based on the maternal presence of glp-1(ar202). Heterozygous animals derived from glp-1(ar202) mothers displayed a mutant phenotype more highly penetrant than that displayed by heterozygous progeny derived from wild-type mothers (Table 4, lines 5 and 6).

Trans-heterozygous combinations of glp-1(Pro) mutants:
Given that the glp-1(Pro) mutants encode receptors with elevated activity, one possibility is that in trans to each other, they would behave as in an allelic series, reflecting their level of activity as homozygotes. Alternatively, they might mutually suppress or enhance in trans, consistent with mutant receptor interaction and/or independent function. To distinguish between these possibilities, we compared the phenotypes of the three possible trans-heterozygous combinations of our glp-1(Pro) mutants with those of three individual homozygous strains (Table 5). Both strong alleles (ar202 or ar224) in trans to ar218 displayed an increased penetrance of the wild-type phenotype, as expected given previous dosage analysis and the nearly wild-type activity of ar218. In contrast, the ar202/ar224 animals exhibited a striking deviation from the expected result based on the behavior of these homozygous mutants. Rather than exhibiting a phenotype at least as strong as the weakest homozygous strain, the ar202/ar224 trans-heterozygote exhibited a dramatic increase in the percentage of wild-type animals, indicating mutual suppression (Table 5). These results are inconsistent with a simple allelic series and are consistent with a physical interaction between receptors encoded by these alleles. Given that each allele is highly dosage sensitive, these data are also consistent with a model of independent mutant receptor activity. Finally, the results suggest that qualitative differences exist between glp-1(ar202) and glp-1(ar224) despite their similar phenotypic and genetic properties.

Genetic interactions of glp-1(ar202) with lag-1 and lag-2:
The glp-1(Pro) phenotype could be the result of locally elevated GLP-1 receptor activity that acts within the known conserved signaling pathway or the result of aberrant activation via some other noncanonical mechanism. In all organisms in which the signaling pathway has been analyzed, LIN-12/Notch receptors are activated by ligands of the DSL family and act in concert with CSL family effectors to alter the transcription of target genes in response to signaling. Notch-mediated signaling that is independent of CSL family effectors has also been reported (SHAWBER et al. 1996 ; NOFZIGER et al. 1999 ; RAMAIN et al. 2001 ; YAMAMOTO et al. 2001 ).

We examined the dependence of the glp-1(ar202) Pro phenotype on the conserved CSL family member, LAG-1. We tested two alleles of lag-1: one apparent null allele, lag-1(q385), which displays a larval lethal "Lin and Glp" (Lag) phenotype, and a second allele, lag-1(q426), which causes a highly penetrant glp-1(loss-of-function)-like germline phenotype (LAMBIE and KIMBLE 1991 ). In both cases, the lag-1 mutant phenotypes were completely epistatic to the Pro phenotype, indicating that expression of the Pro phenotype depends on functional LAG-1 (Table 6).

The LAG-2 ligand is expressed in the distal tip cells and activates GLP-1 in the germline (HENDERSON et al. 1994 ; TAX et al. 1994 ). In addition, LAG-2 is expressed at a very low level during the second larval stage (L2) in two cells in the hermaphrodite proximal somatic gonad, Z1.ppp and Z4.aaa (WILKINSON et al. 1994 ). Here, in the late L2, LAG-2 participates in the anchor cell/ventral uterine precursor cell decision as a ligand for LIN-12, and eventually the cell destined to become the anchor cell expresses a high level of LAG-2. Between the time Z1.ppp and Z4.aaa are born in the late L1 until the very end of the L2, however, these cells are in contact with germ cells. Therefore, one possible explanation for the Pro phenotype is that GLP-1 activity is locally elevated by an inappropriately strong response (hypersensitive receptor) to the low level of LAG-2 ligand produced in the proximal somatic gonad during the L2 stage. Alternatively, the Pro phenotype could be responding to an alternate ligand or it could be ligand independent, consistent with constitutive receptor activity.

To determine whether the glp-1(Pro) mutant phenotype is dependent on LAG-2, we examined the phenotypes of two different glp-1(ar202); lag-2 double-mutant strains (Table 7). First, we tested a null allele, lag-2(q411), that confers a Lag phenotype (LAMBIE and KIMBLE 1991 ). Several individual non-Lag glp-1(ar202); lag-2(q411) double-mutant animals were recovered from glp-1(ar202); lag-2(q411)/+ mothers. These glp-1(ar202); lag-2(q411) animals developed with a Pro germline. This result suggests that the receptor encoded by glp-1(ar202) retains function in the absence of LAG-2. Rescue of the Lag lethal phenotype by glp-1(ar202) was not very efficient (6 animals out of an expected 190), suggesting that the level of constitutive activity of the receptor encoded by glp-1(ar202) is relatively low or that the effect of glp-1(ar202) is tissue specific.

We also examined homozygous double mutants carrying a weaker allele of lag-2, lag-2(q420). lag-2(q420) is the only available lag-2 allele that confers a glp-1(loss-of-function)-like Glp phenotype. Although the single lag-2(q420) mutant displays an incompletely penetrant Glp phenotype, in combination with glp-1(ar202) the Glp phenotype is not observed. This result indicates that the activity of the receptor encoded by glp-1(ar202) is sufficient to bypass a reduction in LAG-2 activity that would normally prevent germline proliferation in 50% of the animals (Table 7) and that both proximal and distal glp-1(ar202) activity is ligand independent. While glp-1(ar202); lag-2(q420) animals did not exhibit a Glp phenotype, the percentage of wild-type animals was elevated in the double mutant compared to glp-1(ar202) alone (Table 7). In summary, glp-1(ar202) can cause a Pro phenotype in the absence of the LAG-2 ligand although the mutant receptor may still be sensitive to the presence of the ligand.

Genetic interactions with other components and modifiers of GLP-1/LIN-12/Notch-mediated signaling:
To better understand the effect of glp-1(Pro) mutations on receptor activity, we examined the phenotype of double-mutant strains carrying the glp-1(Pro) alleles together with mutations previously identified as suppressors or enhancers of lin-12 and glp-1 (SUNDARAM and GREENWALD 1993B ; LEVITAN and GREENWALD 1995 ; TAX et al. 1997 ). The results are presented in Table 8. Interactions with products that affect cleavage or stability of the intracellular domain (sel-12, hop-1, and sel-10; see Fig 1) behaved as expected on the basis of previous results (SUNDARAM and GREENWALD 1993B ; LEVITAN and GREENWALD 1995 ; LI and GREENWALD 1997 ). Specifically, the glp-1(Pro) mutant phenotypes are partially suppressed by loss of function of the presenilins sel-12 and hop-1 and enhanced by sel-10(ar41). Enhancement by sel-10(ar41) overcomes both maternal and zygotic rescue of glp-1(Pro) by reduced dosage and/or glp-1(+) (Table 8).

Modifiers of the pathway that act on the extracellular part of the receptor were of particular interest since glp-1(Pro) mutations alter this part of the protein. One possibility is that the Pro mutations render the receptor susceptible to a low level of constitutive S2 cleavage. If this hypothesis were correct, mutations that reduce the level of S2 cleavage would be effective suppressors. As yet, no C. elegans enzyme is known to be responsible for the S2 cleavage, but in other organisms this cleavage is dependent on metalloproteases. The ADAM-family metalloprotease SUP-17 in C. elegans is homologous with Drosophila Kuzbanian and both are required for signaling (WEN et al. 1997 ). In Drosophila, Kuzbanian is required for ligand-mediated cleavage of the extracellular portion of the receptor (LIEBER et al. 2002 ). We tested interactions between glp-1(ar202) and a reduction-of-function mutation in sup-17, sup-17(n1258) (TAX et al. 1997 ). This mutant displays temperature-sensitive maternal-effect lethality and was originally isolated as a suppressor of a hypermorphic allele of lin-12. We observed dramatic suppression of the glp-1(ar202) Pro phenotype in the double-mutant sup-17(n1258); glp-1(ar202) (Table 8). These results suggest that although genetic interactions with lag-2 indicate little ligand dependence, expression of the glp-1(Pro) phenotype is dependent on SUP-17 function. Moreover, this genetic interaction suggests that the structure/function defect in glp-1(ar202) occurs at or before the level of ligand-dependent cleavage.

Prior to ligand-dependent cleavage, the receptor is transported from the ER to the Golgi, processed and modified in the Golgi, and directed to the extracellular compartment (Fig 1). We investigated the effect of the mutation sel-9(ar22), which likely interferes with proper selection for transport between the ER and Golgi. sel-9 was identified by mutations that suppress phenotypes caused by reduction-of-function mutations in lin-12 and glp-1. In addition, sel-9 mutations enhance phenotypes caused by gain-of-function alleles of lin-12 (SUNDARAM and GREENWALD 1993B ; WEN and GREENWALD 1999 ). We investigated the consequence of a reduction-of-function mutation in sel-9 in combination with glp-1(ar202) and glp-1(ar224). To our surprise, rather than enhancing the glp-1(Pro) phenotype, sel-9(ar22) partially suppressed the Pro phenotype (Table 7). Therefore, the genetic interaction among the Pro phenotypes of glp-1(Pro) mutants differs from lin-12 mutants in their genetic interaction with sel-9. A slight elevation of the penetrance of the non-Pro mutant phenotypes was also observed in the glp-1(Pro); sel-9 double-mutant animals, further supporting the notion that the Pro and non-Pro germline pattern defects exhibited by glp-1(Pro) mutants are distinct.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have isolated three new alleles of glp-1 that differ from previously described glp-1 alleles in several ways. First, they present a Pro phenotype. Second, although the alleles are clearly gain-of-function in character, they are not simple genetic hypermorphs, as indicated by both interactions with glp-1(+) and with each other. The mutant phenotypes are dosage sensitive, glp-1(+) interferes with expression of the mutant phenotype, and trans-heterozygous combinations of highly penetrant alleles mutually suppress. Genetic interactions with the conserved core GLP-1 signaling pathway components LAG-1 and LAG-2 support the idea that the glp-1(Pro) alleles act through the canonical signaling pathway. Interactions between glp-1(Pro) and mutations in genes encoding other components and modifiers of the GLP-1/LIN-12/Notch-mediated signaling pathway indicate that the mutant receptor acts as expected for a ligand-independent, hyperactive receptor. One exception is the interaction with a mutation that likely affects quality control during ER-to-Golgi transport.

Comparison of glp-1(Pro) and glp-1(Tum) gain-of-function alleles:
The glp-1(Pro) mutants described in this article share some similarities with the previously described Tum allele glp-1(oz112gf). Both glp-1(Pro) and glp-1(Tum) mutants (i) are temperature sensitive, exhibiting a more penetrant phenotype at higher temperature; (ii) exhibit a ligand-independent increase in mitotic activity in the germline, including a late-onset extension of the distal germline; and (iii) carry changes in amino acids within the extracellular domain between the epidermal growth factor (EGF)-like repeats and the transmembrane domain of the receptor.

Mutations that cause glp-1 Pro and Tum phenotypes also differ in several important ways. The most obvious difference is the Pro phenotype itself. One particularly puzzling aspect of the Pro phenotype exhibited by glp-1(Pro) alleles is that although glp-1 activity is elevated and largely LAG-2 independent, the activity does not appear to be elevated equally in all germ cells. The general effect of lower dosage of glp-1(Tum) (as in oz112/null) or of more permissive temperature is a highly penetrant late-onset Tum phenotype, not a Pro phenotype. Therefore, meiotic entry from the distal stem-cell population appears to be most sensitive in the Tum class of mutants, whereas an earlier proximal defect results in a Pro phenotype (A. PEPPER, T.-W. LO and E. J. A. HUBBARD, unpublished observations). Genetic analysis shows that the Tum allele glp-1(oz112gf) behaves as a hypermorphic allele, increasing in penetrance with increasing doses of glp-1(+) (BERRY et al. 1997 ), whereas glp-1(Pro) alleles do not. Rather, glp-1(+) competes with glp-1(Pro). Even the phenotypes displayed by the glp-1(Pro) mutants that have a late-onset character may not be hypermorphic since glp-1(Pro)/glp-1(Pro)/+ is not more penetrant for the Class B phenotype than the corresponding glp-1(Pro)/glp-1(Pro) strain (Table 3). We also looked for evidence of distal mitotic zone extension at 48 hr in strains carrying an extra copy of glp-1(+), but no obvious extension was observed. The Pro and Tum glp-1 phenotypic classes are, therefore, genetically separable. This genetic separability suggests that the GLP-1 receptor may undergo different (although overlapping) regulatory interactions within the germline in its role in the meiosis/mitosis decision. In addition, glp-1(Pro) mutants do not display a Multivulva phenotype as does glp-1 (oz112gf). This difference may be due to different levels of GLP-1 activity in the somatic cells of the two mutant classes.

Genetic behavior of glp-1(Pro) alleles suggests multimerization of or competition between GLP-1 receptors:
The glp-1(Pro) lesions cause amino acid changes in the same region as that of other alleles that confer elevated receptor activity. Two glp-1(Pro) mutations map to the LNRs and the third maps just C-terminal to a pair of conserved extracellular cysteine residues. These changes are near putative cleavage sites S1 and S2 (Fig 3). Genetic evidence from many systems indicates that these regions of the protein have a negative regulatory influence on receptor activity, probably preventing receptor activity in the absence of interaction with a ligand (GREENWALD 1994 ). One possibility is that the glp-1(Pro) lesions lead to an alteration in LNR conformation that renders the receptors more susceptible to S2 or S3 cleavage in the absence of ligand activation. Alternatively, the mutant receptors could be less sensitive to a negative regulatory factor. While models invoking an alteration of general susceptibility to activation can account for receptor hyperactivity evident from the glp-1(Pro) mutant phenotypes, they cannot account for the unusual genetic behavior of these alleles.

No other characterized LIN-12/Notch gene family mutations that map within the LNRs or between the LNRs and the transmembrane domain exhibit the same genetic behavior as the glp-1(Pro) mutations [suppression by glp-1(+) dosage and mutual suppression in a trans-heterozygous configuration]. Other alleles that map to this region are hypermorphic (GREENWALD and SEYDOUX 1990 ; LIEBER et al. 1993 ; GREENWALD 1994 ; BERRY et al. 1997 ; BRENNAN et al. 1997 ). Exceptions are loss-of-function alleles glp-1(q158), a point mutation, and glp-1(q172), an in-frame deletion that deletes most of the LNR region (KODOYIANNI et al. 1992 ). Surprisingly, lin-12(n950) encodes the identical amino acid change as glp-1(ar224) (Fig 3), and yet lin-12(n950) behaves genetically as a hypermorph (GREENWALD et al. 1983 ; GREENWALD and SEYDOUX 1990 ). Given that LIN-12 and GLP-1 are functionally interchangeable (FITZGERALD et al. 1993 ), this observation suggests that regulation or local molecular environments differ between LIN-12 and GLP-1 such that the same amino acid change in a conserved residue has different genetic outcomes.

The observation that increasing doses of glp-1(+) can interfere with the glp-1(Pro) mutant phenotypes suggests that either the wild-type receptor binds to and reduces the susceptibility to constitutive activation or the receptors compete for an additional factor or factors that modulate signaling. This, in turn, suggests that similar interactions may normally occur among wild-type receptors. While these are only two of several possible models that could account for our data, they are worth considering in more detail.

The possibility that LIN-12/Notch receptors form a dimeric or a higher multimeric complex that affects signaling has been suggested by previous genetic and biochemical data (GREENWALD and SEYDOUX 1990 ; KOPAN et al. 1996 ; STRUHL and ADACHI 2000 ; SAKAMOTO et al. 2002 ). In the case of lin-12, the data are consistent with receptor dimerization promoting activation (GREENWALD and SEYDOUX 1990 ). Since glp-1(+) dosage suppresses gain-of-function phenotypes of glp-1(Pro) mutants, our results are consistent with a model in which a heterodimeric or heteromultimeric form is resistant to non-ligand-induced activation, whereas the homodimeric (or monomeric) form is more readily activated.

The dosage sensitivity of these alleles suggests the presence of a threshold effect for GLP-1 signaling. At levels under the threshold, little to no signaling occurred, while at levels above the threshold, a high level of signaling occurred. Threshold effects are consistent with the formation of homomultimers in the receptor interaction model. A threshold effect is also consistent with the presence of a feedback mechanism that amplifies the signal once a critical level of signaling is reached. Positive feedback in GLP-1 signaling has been suggested previously (BERRY et al. 1997 ). Feedback and receptor interaction models are not mutually exclusive.

Competition among receptors for a limiting nonreceptor protein could also account for our genetic dosage results. In one such scenario, a positive factor required for non-ligand-mediated activation of glp-1(Pro) receptors would bind to both mutant and wild-type receptors but would bind with greater affinity to wild-type receptors. Thus, insufficient levels of mutant receptor may not allow efficient activation and increasing wild-type dosage would sequester the positive factor from the mutant receptor. The competition model is more difficult to reconcile with mutual suppression observed in animals trans-heterozygous for the highly penetrant glp-1(Pro) alleles. To accommodate a competition model, the two alleles would have to sequester the positive factor from each other without themselves becoming activated to a high level. This possibility, while unlikely, is not completely implausible given the dosage sensitivity of the alleles. A lower level of each mutant receptor in trans-heterozygous animals may be bound to a positive factor but activated below a threshold required for expression of the mutant phenotype.

glp-1(Pro) mutations could alter receptor signaling, trafficking, or both:
The proposed interactions between receptor proteins, either multimerization or competition, may occur at the cell surface as part of the signaling mechanism or, alternatively, they may occur within the cell during receptor maturation. Dimerization or multimerization of G-protein-coupled receptors during receptor transport is well documented and can be required for signaling, quality control, and trafficking (BOUVIER 2001 ).

Genetic interactions between glp-1(Pro) and a sel-9 mutant were unexpected and raise the possibility of an important regulation point at the level o