ABSTRACT
The lin-2 gene is required for the induction of the Caenorhabditis elegans vulva. Vulval development is initiated by a signal from the anchor cell that is transduced by a receptor tyrosine kinase/Ras pathway. We show that lin-2 acts in the vulval precursor cell P6.p, downstream of lin-3 EGF and upstream of let-60 ras, to allow expression of the 1° cell fate. lin-2 encodes a protein of relative molecular mass 109,000 (LIN-2A) with regions of similar-ity to CaM kinase II and membrane-associated guanylate kinases. Mutant lin-2 transgenes designed to lack either protein kinase or guanylate kinase activity are functional, indicating that LIN-2A has a structural rather than an enzymatic role in vulval induction. Most or all identified membrane-associated guanylate kinases are components of cell junctions, including vertebrate tight junctions and arthropod septate junctions in epithelia. Thus, LIN-2A may be a component of the cell junctions of the epithelial vulval precursor cells that is required for signaling by the receptor tyrosine kinase LET-23. We propose that LIN-2A is required for the localization of one or more signal trans-duction proteins (such as LET-23) to either the basal membrane domain or the cell junctions, and that mislocal-ization of signal transduction proteins in lin-2 mutants interferes with vulval induction.
INTRODUCTION
Intercellular signals specify cell fates during the development of the vulva of the C. elegans hermaphrodite (for reviews, see Eisenmann and Kim, 1994; Horvitz and Sternberg, 1991). The vulva develops from three precursor cells that are selected by these signals from a group of six equipotent vulval precursor cells (P3.p to P8.p) during the L3 larval stage (Sulston and Horvitz, 1977). Each vulval precursor cell has the potential to express any one of three discrete cell fates, termed 1°, 2° and 3° (Fig. 1; Sternberg and Horvitz, 1986; Thomas et al., 1990). A signal from the gonadal anchor cell (positioned adjacent and dorsal to P6.p) induces P6.p to adopt the 1° cell fate, which is to generate cells that form the inner parts of the developing vulva (Kimble, 1981; Sternberg and Horvitz, 1986). Subse-quently, P6.p sends a lateral signal that induces each of the adjacent vulval precursor cells (P5.p and P7.p) to adopt the 2° cell fate, which is to generate cells that form the outer parts of the developing vulva (Sternberg, 1988; Simske and Kim, 1995; Koga and Ohshima, 1995). In addition to the lateral signal, some evidence suggests that P5.p and P7.p receive the anchor cell signal, which may assist in the induction of the 2° cell fate (Sternberg, 1988; Thomas et al., 1990; Katz et al., 1995). An inhibitory signal from the hyp7 syncytial epidermis ensures that the remaining vulval precursor cells (P3.p, P4.p and P8.p) are not induced and adopt the 3° cell fate, which is to generate a pair of cells that fuse with hyp7 (Herman and Hedgecock, 1990).
Schematic representation of a longitudinal cross-section at theL2 to early L3 larval stage showing the vulval precursor cells and the anchor cell. It is not known whether a basal lamina separates the anchor cell from P6.p at the time of anchor cell signaling (J. White, personal communication).
Schematic representation of a longitudinal cross-section at theL2 to early L3 larval stage showing the vulval precursor cells and the anchor cell. It is not known whether a basal lamina separates the anchor cell from P6.p at the time of anchor cell signaling (J. White, personal communication).
The anchor cell signal is transduced by a highly conserved receptor tyrosine kinase/Ras signaling pathway (for review, see Eisenmann and Kim, 1994). Loss-of-function mutations in genes in this pathway confer a vulvaless (Vul) phenotype in which all six vulval precursor cells express the 3° cell fate. Three genes are particularly relevant to the work reported here: lin-3 encodes a protein related to epidermal growth factor (EGF) that is likely to be the anchor cell signal (Hill and Sternberg, 1992), let-23 encodes a homolog of the EGF receptor tyrosine kinase that is the putative LIN-3 receptor (Aroian et al., 1990), and let-60 encodes a homolog of Ras that transduces the anchor cell signal (Beitel et al., 1990; Han and Sternberg, 1990).
We are analyzing other genes in the signaling pathway to further elucidate how the anchor cell signal induces P6.p to express the 1° cell fate. We are focusing on genes that are required specifically in the vulval precursor cells and not in other cell types. Some of these may be signaling genes with specialized roles in polarized epithelial cells. The vulval precursor cells are squamous epithelial cells that lie in a row along the ventral midline and make up the ventral epidermis of the L2 and L3 larva (Fig. 1; Sulston and Horvitz, 1977). Their plasma membranes are divided into apical and basal membrane domains by cell junctions that form a lateral seal around the circumference of each cell. The apical membrane domains secrete the components of the cuticle, and the basal membrane domains may secrete components of the basal lamina (White, 1988). By electron microscopy, the cell junction appears as a single darkly staining structure called the belt desmosome that may combine the functions of the tight junction and the adherens junction of vertebrates, separating the apical and basal membrane domains, controlling diffusion across the epithelium in the extracellular space, and modulat-ing intercellular adhesion (White, 1988). The cell junctions form a continuous network connecting the vulval precursor cells to each other along their anterior and posterior edges, and each vulval precursor cell to the surrounding hyp7 syncytial epidermis along its lateral edges.
lin-2 mutations, like mutations in genes in the let-23 receptor tyrosine kinase/let-60 ras signaling pathway, were isolated in screens for vulvaless mutants (Horvitz and Sulston, 1980; Ferguson and Horvitz, 1985). Two observations suggest that lin-2 is required for activation of the receptor tyrosine kinase/Ras pathway specifically in the vulval precursor cells. Firstly, lin-2(n397), which either strongly reduces or elimi-nates gene activity, results in a Vul phenotype in 99% of animals (Ferguson and Horvitz, 1985; see below). Secondly, lin-2 is apparently required only for vulval development, as no lin-2 mutation, including n397, has any obvious effect on other aspects of growth and development (Horvitz and Sulston, 1980; Ferguson and Horvitz, 1985; see below). In contrast, the genes described above in the anchor cell signaling pathway are also required for the development of various other cell types (Ferguson and Horvitz, 1985; Beitel et al., 1990; Han et al., 1990; Aroian and Sternberg, 1991).
lin-2 appears to interact closely with two other genes, lin-7 and lin-10, during vulval induction. Mutations in each of these genes result in essentially the same Vul phenotype, and each gene displays very similar genetic interactions with other vulval signaling genes (e.g. lin-12, lin-15 and lin-31) (Horvitz and Sulston, 1980; Ferguson and Horvitz, 1985; Sternberg and Horvitz, 1989; Kim and Horvitz, 1990; Miller et al., 1993). The similar genetic behaviors of these three genes suggests that they may act at the same step in anchor cell signaling. lin-10 encodes a protein that does not yet have a known biochemical function (Kim and Horvitz, 1990).
We have undertaken a molecular and genetic analysis of lin-2 to understand how it interacts with the anchor cell signaling pathway and to gain insight into the related genetic functions provided by lin-7 and lin-10. We show that lin-2 activity is required in P6.p for the expression of the 1° cell fate, but is not required in either P5.p or P7.p for the expression of the 2° cell fate. lin-2 acts downstream of lin-3 EGF and upstream of let-60 ras at an early step in vulval induction and encodes a protein of relative molecular mass (Mr) 109 ×103(LIN-2A) with simi-larity to Ca2+/calmodulin-dependent protein kinase II (CaM kinase II) and membrane-associated guanylate kinase proteins (MAGUKs). Surprisingly, site-directed mutagenesis experi-ments indicate that neither protein kinase activity nor guanylate kinase activity is required for lin-2 function in vulval induction. Thus, lin-2 does not appear to function as a protein kinase in the transduction of the anchor cell signal, nor as a guanylate kinase that might, for example, modulate the GTPase activity of LET-60 Ras. MAGUK proteins, in addition to having guanylate kinase domains, are peripheral membrane proteins that are typically associated with cell junctions (for review, see Woods and Bryant, 1993). Thus, lin-2 may encode a component of the cell junctions of the vulval precursor cells, and specific cell junction components may be involved in signal transduction. We propose that lin-2 may be required for localization of the LET-23 receptor tyrosine kinase to either the basal membrane domain or the cell junctions of the vulval precursor cells, and that this localization may be necessary for activation by the LIN-3 ligand.
MATERIALS AND METHODS
General methods and strains
C. elegans strains were maintained and handled by standard methods at 20°C (Brenner, 1974). Wild type refers to C. elegans, variety Bristol, strain N2. B0 refers to C. elegans, variety Bergerac. Genetic markers used in this work are listed below and are described by Wood (1988), except as noted.
LGI: unc-29(e1072). LGIII: ncl-1(e1942). LGIV: let-60(n1700gf) (Beitel et al., 1990). LGX: egl-15(n484), sma-5(n678), lin-2(e1309, e1424, e1453) (Horvitz and Sulston, 1980), lin-2(e1437, n105, n167, n305, n380, n397, n670, n674, n768, n1052) (Ferguson and Horvitz, 1985), lin-2(n1389) (S. K. K., unpublished data), lin-2(n1610, n2686, n2687) (gift from S. G. Clark and M. J. Stern), lin-2(ad527) (gift from L. Avery), lin-2(ga59, ga60, ga61) (gift from D. M. Eisenmann), unc-9(e101).
Genetic analyses
The strain MT397, in which 77% of animals are Vul (Ferguson and Horvitz, 1985), carries lin-2(n397) and a suppressor mutation. MT397 was backcrossed to N2 and yielded the strain SD294, which carries lin-2(n397) but not the suppressor and in which 99% (n=280) of animals are Vul.
Standard methods were used to construct double and triple mutant strains. Some let-60(n1700) animals burst at the L4 molt, apparently as a result of the strong Muv phenotype, and consequently produce no progeny. Almost all let-60(n1700); lin-2(n397) animals burst, so it is not possible to propagate a homozygous strain. The genotypes of candidate let-60(n1700); lin-2(n397) homozygotes, segregated from let-60(n1700); lin-2(n397)/unc-9(e101) heterozygotes, were deter-mined by finding animals that carried the lin-2(n397) deletion and lacked the corresponding lin-2(+) DNA using PCR amplification and oligonucleotides specific for a wild-type sequence that is deleted in lin-2(n397) (data not shown).
The syIs1 lin-2(n397) X strain was constructed by picking 30 Vul progeny from syIs1 +/+ lin-2(n397) heterozygotes. Four of these animals were recombinants of genotype syIs1 lin-2(n397)/+ lin-2(n397) because their progeny exhibited the lin-2 Vul phenotype and carried syIs1, which was detected by PCR amplification with oligonu-cleotides specific to the lin-3 transgene. Homozygous syIs1 lin-2(n397) strains were established by cloning progeny, allowing them to self, and choosing those strains in which the lin-3 transgene was segregated to each of 20 individuals.
RFLP mapping
RFLPs were identified by Southern blot analysis of genomic DNAs of the wild strains N2 and B0 digested with EcoRI, using cosmids selected from the C. elegans genome physical map as probes. Cosmids used to identify RFLPs are as follows: C43C10 (gaP1), C02B4 (gaP2), C34E7 (gaP3), C51A5 (gaP4), C56B1 (gaP5), C34E11 (gaP6), C25B2 (gaP7), C23H4 (gaP8), C07H2 (gaP9), C16C5 (gaP10). gaP11 and gaP12 are insertions of the Tc1 trans-posable element in the B0 strain, both of which can be detected using single-copy probes derived from DNA flanking the insertions (S. K. K., unpublished data). The position of gaP11 was determined by hybridizing the gaP11 probe to a grid of yeast artificial chromosome clones that span the C. elegans genome (gift from A. Coulson and J. Sulston); gaP11 is contained on the clones Y16F9, Y45E9 and Y47B2. The position of gaP12 was determined by using the gaP12 probe to isolate genomic lambda clones (MT#2-2, -6 and -7) and placing them on the C. elegans genome physical map.
To map RFLPs located to the left of lin-2, hermaphrodites of an N2-derived strain of genotype egl-15(n484) sma-5(n678) lin-2(e1309) X were crossed to males carrying a B0 X chromosome. Heterozygous cross-progeny were picked, allowed to self, and Egl non-Sma non-Lin and Lin non-Sma non-Egl recombinants were picked. Strains homozygous for each recombinant chromosome were established, and the segregation of RFLPs was scored by Southern blot analysis. The following genetic map order was obtained (brackets indicate number of recombinants obtained between two markers): egl-15 [12] gaP6 [1] sma-5 [1] gaP7 [3] gaP8 [3] gaP1 [0] lin-2. In a second experiment, Lin non-Sma recombinants were picked from the self-progeny of sma-5(n484) lin-2(e1309)/B0 heterozygotes and yielded: sma-5 [11] gaP8 [6] gaP1 [0] lin-2. To map RFLPs located to the right of lin-2, Lin non-Unc recombinants were picked from the self-progeny of lin-2(e1309) unc-9(e101)/B0 heterozygotes. A first experiment yielded: lin-2 [0] gaP1 [0] gaP2 [0] gaP9 [3] gaP5 [1] gaP4 [0] gaP3 [4] gaP11 [2] unc-9. A second experiment yielded: lin-2 [0] gaP1 [3] gaP10 [1] gaP2 [0] gaP9 [10] unc-9.
DNA manipulation
Standard methods were used in molecular biology experiments (Wood, 1988; Sambrook et al., 1989).
lin-2 allele-specific rearrangements were identified and mapped by Southern blot analysis. e1309 and n768 are both complex rearrangements within a 6.7 kb EcoRI fragment at the 5′ end of lin-2. e1424 is a 300 bp deletion that removes a HindIII site within a 2.6 kb KpnI/EcoRI fragment. e1437 carries the same deletion as e1424 and is probably not an independent mutation. We mapped the breakpoints of the n397 deletion, both of which are in sequenced genomic regions (data not shown; H. Wang, personal communica-tion). We used PCR to amplify a 3.9 kb genomic fragment from n397 DNA using oligonucleotide primers located on opposite sides of the deletion (5′CATGTCGCGATTAAGGCTTGTG3′ and 5′AGTCAAGTCCGCAAACACCACG3′), and subcloned the fragment into pCRII (Invitrogen). Comparison of sequence from this fragment with the wild-type sequence showed that the left break-point of the n397 deletion is within an exon (after position 411 in lin-2A), and the right breakpoint is within an intron (which is located before position 1386 in lin-2A). No DNA rearrangements were found by Southern blot analysis using either EcoRI or HindIII in strains containing the lin-2 alleles ad527, e1453, ga59, ga60, ga61, n105, n167, n305, n380, n670, n674, n1052, n1389, n1610, n2686 and n2687 (data not shown).
Two lin-2 cDNA clones (λSK1 and λSK2) were isolated from an N2 mixed-stage cDNA library (Kim and Horvitz, 1990) in λgt10 by hybridization with a 2.6 kb KpnI/EcoRI genomic fragment derived from the region affected by all four lin-2 DNA rearrangements. Additional cDNA clones (including λB2A, λB6A and λB10A) were subsequently isolated from another N2 mixed-stage cDNA library (gift from B. Barstead) in λZAP hybridized with a fragment of λSK1 (positions 739 to 2886 of lin-2A). DNA sequence analysis showed that λSK1, λSK2 and λB6A represent the lin-2A transcript, including sequences from position 163 to 371, and 443 to the 3′ end. λB2A and λB10A define the alternatively spliced lin-2B transcript, which comprises sequence in common with lin-2A (positions 1227 to the 3′ end) and a lin-2B-specific 5′ exon. We characterized the 5′ end of the lin-2A coding region by sequence analysis of 2.3 kb of genomic DNA located immediately 5′ to the lin-2A cDNA clones. This sequence contains open reading frames and consensus splice sites predicting two exons, and the predicted splicing pattern was confirmed by ampli-fication and analysis of RT-PCR products from N2 RNA (data not shown).
To identify the e1453 and n105 mutations, two oligonucleotides (5′ATGAGGCTGTCACAATATCGG3′ and 5′CACCGGCAACAT-GAAGAAAG3′) were used in PCR experiments to amplify a 1439 bp genomic fragment from wild-type, e1453 and n105 strains. This fragment extends from position 2320 to 2937 in lin-2A and encodes all of the guanylate kinase domain. PCR fragments were subcloned into pCRII (Invitrogen), and the exons and flanking intron sequence were determined and compared. For both mutants, three clones derived from two independent PCR reactions were analyzed. No other mutations were found in the lin-2 genomic sequences representing the region common to lin-2A and lin-2B using single-stranded conforma-tion polymorphism (SSCP) gels (FMC) to analyze PCR fragments derived from genomic DNA from wild-type, e1453 and n105 mutants (data not shown).
pKX3 and pKX4 contain 10.9 kb and 9.2 kb KpnI/XhoI genomic fragments, respectively, in the vector pBGST (Spratt et al., 1986; D. Alley, unpublished data). pSK7 contains a 7.0 kb StuI/KpnI genomic fragment in the vector pBluescript KS+. plin-2A contains a 7.7 kb StuI/ClaI genomic fragment joined to a 4.0 kb ClaI fragment from a lin-2 cDNA clone (λSK2) in the vector pBluescript KS+. plin-2Δwas constructed by deleting a 563 bp KpnI/ClaI fragment from plin-2A. Site-directed mutagenesis of plin-2A using mutagenic oligonu-cleotides and unique-site elimination mutagenesis (Pharmacia) were used to generate the following mutant lin-2 minigenes (mutations in parentheses): pCaM1 (E176A), pCaM2 (D146A), pATP (deletion of H783, G784 and G786) and pGMP (R811A, R814A, E817A). To construct pCaM1 and pCaM2, a 563 bp KpnI/ClaI fragment was mutagenized, and to construct pGMP and pATP, a 1.3 kb XhoI fragment was mutagenized. The mutations were verified by DNA sequence analysis, and the mutagenized fragments were subcloned into plin-2A.
Germline transformation
Transformation experiments were conducted by microinjection of DNAs into the germlines of unc-29(e1072); lin-2(e1309) hermaphro-dites using unc-29(+) (cosmid C45D10) as a co-transformation marker, by standard methods (Fire, 1986; Mello et al., 1991). DNA concentrations were 100 μg/ml per clone, except that the C43C10 con-centration was 5 μg/ml. Strains transformed with DNA clones that rescued the lin-2 Vul phenotype exhibited 80-92% non-Vul animals, whereas clones that did not rescue exhibited less than 10% non-Vul animals. For each construct tested, at least three independent trans-genic lines were analyzed.
Genetic mosaic analysis
An unc-29(e1072); ncl-1(e1942); lin-2(e1309) strain was trans-formed with a mixture of DNA clones (50 μg/ml each) containing wild-type copies of unc-29 (cosmid C45D10), ncl-1 (cosmid C33C3) and lin-2 (plin-2A), yielding a non-Unc, non-Ncl, non-Vul strain (SD422) carrying the transgenes as an extrachromosomal array (gaEx62) that was transmitted to approximately 50% of progeny and in which approximately 90% of animals carrying the extrachromo-somal array were non-Vul. Mosaic animals were identified and analyzed as described by L. M. Miller, D. A. Waring and S. K. Kim (unpublished data). Briefly, certain diagnostic cells in L4 animals were viewed by Nomarski microscopy, and the pattern of Ncl cells was used to identify which cell was most likely to have lost the extra-chromosomal array and generated the clone of mutant cells (see Lackner et al., 1994). A loss in EMS could not be distinguished from a loss in MS, and a loss in AB.pr could not be distinguished from a loss in AB.pra. Presence of the extrachromosomal array in the germline was scored by determining whether the array was trans-mitted to progeny. Fifty mosaic animals were obtained by screening about 1500 animals. For the animals described in Table 1, we were usually able to infer the lineage relationship of each vulval cell nucleus without directly observing its pattern of divisions because most of these animals (8 of 9) had the correct number, position and morphology of vulval cell nuclei, and subsequently developed a functional vulva in the adult.
Generation of anti-LIN-2 antiserum and quantitation of LIN-2A expression
To generate an anti-LIN-2 antiserum, a 610 bp EcoRI fragment of λSK1 encoding protein sequence from the LIN-2A/B common region (positions 457-679 of LIN-2A), including the GLGF/DHR domain and a portion of the SH3 domain, was subcloned into the EcoRI site of pATH10. A TrpE/LIN-2 fusion protein was expressed in E. coli, purified by preparative SDS/PAGE, and used to raise a rat antiserum (Josman Labs). To make a reagent for affinity purification, the same lin-2 fragment was subcloned in pTrcHisC, and a 6xHis-tagged fusion protein was expressed in E. coli, purified on a His-Bind column (Novagen), and coupled to a SulfoLink column (Pierce). The antiserum was affinity purified on this column and then further purified by adsorption to a powdered acetone extract of E. coli proteins.
To determine levels of LIN-2 expression, 40-50 L4 and young adult animals were lysed in 20 μl of SDS sample buffer. Proteins were separated by 8% SDS-PAGE, transferred to nitrocellulose, stained with anti-LIN-2 antiserum, and visualized with an HRP-conjugated anti-rat secondary antibody and enhanced chemiluminescence (Amersham). Multiple autoradiographic exposures of each blot were collected, and levels of protein were measured by densitometry (Molecular Dynamics) using non-specific bands for calibration.
RESULTS
lin-2 acts upstream of let-60 ras in vulval induction
The anchor cell induces the development of the vulva via a receptor tyrosine kinase/Ras signal transduction pathway. One class of genes in this pathway (including lin-3, let-23 and let-60) is characterized by a loss-of-function mutant phenotype that includes defects in vulval induction as well as other aspects of growth and development (for review, see Eisenmann and Kim, 1994). We seek to identify new components of this pathway by molecular and genetic analysis of another class of genes (including lin-2) that is characterized by a loss-of-function vulvaless mutant phenotype with no other apparent defects (Fig. 2B) (Horvitz and Sulston, 1980; Ferguson and Horvitz, 1985; see below).
lin-2 acts upstream of let-60 ras. (A-F) Nomarski photomicrographs of lateral views of animals at the L4 larval stage. Vulval tissue generated from 1° and 2° cells (arrows) and hypodermal tissue generated from 3° cells (bars) are indicated. lin-2 refers to lin-2(n397), let-60(d) refers to let-60(n1700gf), and lin-3(d) refers to a chromosomal insertion of multiple copies of a lin-3(+) transgene (syIs1). Anterior is to the left and ventral is down. Scale bar, 10 μm.
lin-2 acts upstream of let-60 ras. (A-F) Nomarski photomicrographs of lateral views of animals at the L4 larval stage. Vulval tissue generated from 1° and 2° cells (arrows) and hypodermal tissue generated from 3° cells (bars) are indicated. lin-2 refers to lin-2(n397), let-60(d) refers to let-60(n1700gf), and lin-3(d) refers to a chromosomal insertion of multiple copies of a lin-3(+) transgene (syIs1). Anterior is to the left and ventral is down. Scale bar, 10 μm.
Two lin-2 alleles (e1309 and n397) were used in the genetic experiments presented here. lin-2(e1309) is a strong loss-of-function allele and was previously reported to produce the strongest Vul phenotype of lin-2 alleles; it confers a Vul phenotype in 93% of animals and is a complex rearrangement at the 5′ end of lin-2 that severely reduces protein expression (Horvitz and Sulston, 1980; Ferguson and Horvitz, 1985; see below). During the course of this work, we found that lin-2(n397) confers a Vul phenotype in 99% of animals (n=280) and is a candidate null allele. It is a 7.5 kb deletion that elim-inates a large fraction of the lin-2 coding sequence (see below). The severity of the n397 Vul phenotype was underestimated in previous reports because the strain that was examined contained a weak suppressor mutation in a second gene (see Materials and Methods).
In order to better understand how lin-2 interacts with the anchor cell signaling pathway, we determined whether lin-2 acts upstream or downstream of lin-3 EGF and let-60 ras. lin-3 EGF encodes the presumed anchor cell signal. syIs1 X is a chromosomal insertion of multiple copies of a lin-3(+) transgene that confers a dominant Muv phenotype presumably due to overexpression of the signal (Fig. 2C) (Hill and Sternberg, 1992; J. Liu and P. Sternberg, personal communi-cation). We found that syIs1 lin-2(n397) double mutants exhibit a Vul phenotype in 88% of animals and a wild-type vulval phenotype in the remaining animals (n=397) (Fig. 2D). This result suggests that lin-2 acts downstream of lin-3, and that lin-2(n397) can block vulval induction caused by overex-pression of lin-3.
let-60(n1700gf) is a gain-of-function ras allele that consti-tutively activates the pathway, resulting in a multivulva (Muv) phenotype in 96% of animals in which all six vulval precursor cells express induced (1° or 2°) cell fates (Fig. 2E) (Beitel et al., 1990). We found that let-60(n1700gf); lin-2(n397) double mutants exhibit a Muv phenotype (Fig. 2F). The double mutants burst as young adults, so a homozygous strain cannot be maintained. We scored multiple let-60(n1700gf); lin-2(n397) homozygous animals that were segregated from let-60(n1700gf); lin-2(n397)/unc-9(e101) heterozygotes (see Materials and Methods). This result suggests that the signaling pathway downstream of let-60 ras and the cellular response to its activation can function in lin-2 mutants. Thus, lin-2 likely acts before let-60 ras, and lin-2 mutations may block activa-tion of LET-60 Ras by the LIN-3 anchor cell signal.
Molecular cloning of lin-2
We cloned lin-2 in order to begin to understand how it functions at the molecular level in anchor cell signaling. Initially, the C. elegans genome project had identified a large set of overlapping cosmid and YAC clones, called a contig, that possibly contained lin-2 (Coulson et al., 1988; A. Coulson, personal communication).
We showed that lin-2 was contained within this contig by using individual cosmid clones to identify restriction fragment length polymorphisms (RFLPs) and then genetically mapping them with respect to lin-2 (Fig. 3A). These experiments showed that lin-2 is located to the right of one RFLP (gaP8) and to the left of another (gaP10) so that lin-2 must be contained within the 350 kb region of cloned DNA between them (Fig. 3A). Furthermore, a third RFLP (gaP1) was never separated from lin-2 in these genetic crosses, indicating that lin-2 is located near gaP1 (within about 0.15 map units, which corresponds to 50 kb on average) (Sulston and Brenner, 1974; Herman, 1988; see Materials and Methods).
Molecular cloning of lin-2. (A) Genetic and physical maps of the lin-2 region of the X chromosome. Twelve RFLPs (gaPs) used to define the location of lin-2 on the physical map are shown. (B) Restriction map of the lin-2 region. The extents of the lin-2A and lin-2B primary transcripts are indicated. A 2.6 kb KpnI/EcoRI genomic fragment (filled box) was used as a probe to obtain lin-2 cDNAs, and a 6.7 kb EcoRI genomic fragment (open box) was used as a probe in a northern blot experiment. e1424 and n397 are deletions (lines with bars); the n397 endpoints are known precisely, and the e1424 endpoints are known to within about 100 bp. e1309 and n768 are rearrangements within a 6.7 kb EcoRI fragment (lines). Clones and their activities in lin-2(e1309) germline transformation experiments are indicated (see Materials and Methods). The ends of C43C10 extend beyond the limits of the figure (arrowheads). R, EcoRI; X, XhoI; K,KpnI. (C) Northern blot analysis of lin-2A mRNA. Poly(A)+ RNA (∼10 μg) from mixed-stage N2 animals was analyzed by northern blotting using the 6.7 kb genomic EcoRI fragment as a probe. Positions and sizes (in kb) of RNA markers (BRL) and the lin-2A band are indicated. (D) Western blot analysis of LIN-2. Whole lysates were prepared from L4 and young adult animals from wild-type (N2), lin-2(e1309) and lin-2(n397) strains. The extracts were analyzed by western blotting using 8% SDS-PAGE and anti-LIN-2 antiserum. The LIN-2A band and the positions and sizes (in Mr ×10−3) of prestained protein markers (BioRad) are indicated. The faint band at ∼Mr 110 ×103(arrow) represents a protein that is slightly larger than LIN-2A and cross-reacts with the anti-LIN-2 antiserum.
Molecular cloning of lin-2. (A) Genetic and physical maps of the lin-2 region of the X chromosome. Twelve RFLPs (gaPs) used to define the location of lin-2 on the physical map are shown. (B) Restriction map of the lin-2 region. The extents of the lin-2A and lin-2B primary transcripts are indicated. A 2.6 kb KpnI/EcoRI genomic fragment (filled box) was used as a probe to obtain lin-2 cDNAs, and a 6.7 kb EcoRI genomic fragment (open box) was used as a probe in a northern blot experiment. e1424 and n397 are deletions (lines with bars); the n397 endpoints are known precisely, and the e1424 endpoints are known to within about 100 bp. e1309 and n768 are rearrangements within a 6.7 kb EcoRI fragment (lines). Clones and their activities in lin-2(e1309) germline transformation experiments are indicated (see Materials and Methods). The ends of C43C10 extend beyond the limits of the figure (arrowheads). R, EcoRI; X, XhoI; K,KpnI. (C) Northern blot analysis of lin-2A mRNA. Poly(A)+ RNA (∼10 μg) from mixed-stage N2 animals was analyzed by northern blotting using the 6.7 kb genomic EcoRI fragment as a probe. Positions and sizes (in kb) of RNA markers (BRL) and the lin-2A band are indicated. (D) Western blot analysis of LIN-2. Whole lysates were prepared from L4 and young adult animals from wild-type (N2), lin-2(e1309) and lin-2(n397) strains. The extracts were analyzed by western blotting using 8% SDS-PAGE and anti-LIN-2 antiserum. The LIN-2A band and the positions and sizes (in Mr ×10−3) of prestained protein markers (BioRad) are indicated. The faint band at ∼Mr 110 ×103(arrow) represents a protein that is slightly larger than LIN-2A and cross-reacts with the anti-LIN-2 antiserum.
The position of lin-2 within the interval between gaP8 and gaP10 was determined by searching for lin-2 mutations that were DNA rearrangements. We assayed genomic DNAs of twenty lin-2 mutants by Southern blot analysis, using cosmids selected from the lin-2 interval as probes (see Materials and Methods). Only one cosmid (C43C10, which detects gaP1) revealed any rearrangements, and four different lin-2 mutants had rearrangements that were clustered within a 15 kb region. Subsequent mapping experiments showed that lin-2(e1424) is a 300 bp deletion, lin-2(n397) is a 7.5 kb deletion, and both lin-2(e1309) and lin-2(n768) are complex rearrangements (Fig. 3B). Most or all of lin-2 is contained in C43C10, as it can rescue the lin-2(e1309) Vul phenotype in germline transfor-mation experiments (Fig. 3B).
Characterization of lin-2 mRNAs
To identify lin-2 mRNAs, we isolated and characterized candidate lin-2 cDNA clones by hybridization with a 2.6 kb genomic fragment affected by all four lin-2 rearrangements (Fig. 3B; see Materials and Methods). As demonstrated below, these cDNA clones represent lin-2 mRNAs. DNA sequence analysis indicated that the clones represent two classes of mRNAs derived by alternative RNA splicing, termed lin-2A and lin-2B, which share a common 3′ end, but have different 5′ ends (Figs 3B, 4). None of our lin-2A cDNA clones rep-resents a full-length mRNA, as none encodes a start methion-ine. The sequence of the predicted amino terminus of LIN-2A was determined by sequencing DNA from the 5′ end of the lin-2 genomic region and using RT-PCR experiments to determine the exon and intron boundaries (data not shown). This analysis indicated that the lin-2A mRNA is at least 3.4 kb long and contains an open reading frame encoding 961 amino acids (Fig. 4). The sequence shown in Fig. 4 is likely to represent most or all of the lin-2A sequence, since Northern blot analysis revealed a single lin-2A mRNA band of 3.4 kb (Fig. 3C). The predicted protein sequence likely represents all of LIN-2A, since the 5′ end of the lin-2A sequence encodes a putative initiator methionine that is closely preceded by an in-frame stop codon (data not shown), and the 3′ end is polyadenylated. lin-2B, which is not necessary for vulval induction (see below), is represented by a cDNA clone that includes a putative initiator methionine, which is preceded by an in-frame stop codon (data not shown), and a polyadenylated 3′ end (Fig. 4). The C. elegans genome project has recently released the genomic sequence of the lin-2 region (cosmid F17E5) (Wilson et al., 1994; EMBL data base accession no. Z50873). This sequence shows how lin-2A and lin-2B are produced by alter-native splicing. The lin-2 gene comprises eleven lin-2A-specific exons, followed by a single lin-2B-specific exon and fourteen exons common to lin-2A and lin-2B.
Nucleotide and predicted protein sequences of lin-2A and lin-2B. Four domains of sequence similarity are underlined and labeled. The 5′ region specific to lin-2A and nucleotide changes in two lin-2 mutant alleles, both of which result in premature stop codons, are indicated. The 5′ specific exon of lin-2B is continuous with the lin-2A/B common region, as indicated. The EMBL database accession numbers for lin-2A and lin-2B are X92564 and X92565, respectively.
Nucleotide and predicted protein sequences of lin-2A and lin-2B. Four domains of sequence similarity are underlined and labeled. The 5′ region specific to lin-2A and nucleotide changes in two lin-2 mutant alleles, both of which result in premature stop codons, are indicated. The 5′ specific exon of lin-2B is continuous with the lin-2A/B common region, as indicated. The EMBL database accession numbers for lin-2A and lin-2B are X92564 and X92565, respectively.
We verified the predicted size of LIN-2A using western blot analysis of protein extracts of wild-type and lin-2 mutant strains. We affinity purified a rat antiserum raised against a LIN-2 fusion protein expressed in E. coli (see Materials and Methods). The predicted Mrs of LIN-2A and LIN-2B are 109 ×103 and 69 ×103, respectively. The anti-LIN-2 antiserum recognizes a putative LIN-2A protein of apparent molecular mass 109 ×103 in extracts from the wild type that is absent in lin-2(n397) extracts and greatly reduced in lin-2(e1309) extracts (Fig. 3D). We did not observe a protein of the size predicted for LIN-2B in these experiments, suggesting that LIN-2A is the primary product of lin-2.
Three lines of evidence indicate that the cDNA clones described above represent lin-2 mRNAs. First, we identified the breakpoints of the 7.5 kb deletion in lin-2(n397) and found that it removes amino acids between positions 137 and 463 in LIN-2A, and also the N-terminal specific portion of LIN-2B. The 5′ end of the deletion is within a lin-2A exon, and the 3′ end is within an intron common to lin-2A and lin-2B. Thus, the deletion is predicted to disrupt the normal RNA splicing patterns of both lin-2A and lin-2B, and thereby interfere with translation of the coding region located 3′ to the deletion (residues 463 to 961 of LIN-2A). Furthermore, expression of the LIN-2A protein is reduced in lin-2(e1309) and absent in lin-2(n397) (Fig. 3D).
Second, we identified two lin-2 point mutations within the region common to lin-2A and lin-2B (Fig. 4). The amber-sup-pressible allele lin-2(e1453) has an amber stop mutation at position 2690 that is predicted to result in a truncation of the C terminus of LIN-2 by 65 amino acids. lin-2(n105) has an ochre stop mutation at position 2818, resulting in a predicted truncation of 22 amino acids.
Third, we constructed a minigene (plin-2A) that expresses the lin-2A cDNA sequence under the control of the lin-2A promoter and showed that it can efficiently rescue the lin-2 Vul phenotype in germline transformation experiments (Fig. 3B). The 5′ untranslated region and the first 5 exons of lin-2A are represented in 7.7 kb of genomic sequence, and the remainder of lin-2A is derived from a cDNA clone. (The construct deletes the lin-2B-specific exon.) We engineered two constructs that should not express lin-2A and showed that they do not rescue the lin-2 Vul phenotype (Fig. 3B). plin-2Δ contains a 563 bp deletion that results in a frameshift within the coding region after amino acid 123 of LIN-2A (see Materials and Methods), and pSK7 contains 7.0 kb of the genomic region but lacks lin-2A cDNA sequences. These results demonstrate that the cDNA clones represent lin-2 mRNAs and that lin-2A is sufficient for vulval induction.
lin-2 acts in the vulval precursor cell P6.p
We analyzed genetic mosaics to determine which tissues require lin-2 activity for vulval induction, since the lin-2 vulvaless phenotype could result from a lack of gene activity in any of the tissues involved in vulval development. For example, lin-2 could be required in the anchor cell to produce the anchor cell signal, in the hyp7 syncytium that surrounds the vulval precursor cells to repress an inhibitory signal, or in the vulval precursor cells to transduce the anchor cell signal.
For this purpose, we constructed a strain in which the lin-2(+) minigene (plin-2A) and the cell lineage marker gene ncl-1(+) were carried on an extrachromosomal array (gaEx62), and the chromosomal copies of these genes were mutant (see Materials and Methods). ncl-1 is an excellent cell lineage marker as the Ncl mutant phenotype (an enlarged nucleolus) can be scored in most cells. Each mosaic animal, in which lin-2(+) and ncl-1(+) were absent in a subset of the cell lineage, was generated by spontaneous loss of this extrachromosomal array at a single cell division during development. We used the Ncl phenotype to determine which cells lacked the array and knowledge of the complete cell lineage to infer at which cell division the array had been lost (Sulston and Horvitz, 1977; Sulston et al., 1983). Subsequently, vulval development in each mosaic animal was scored to determine whether loss of lin-2(+) activity in the clone of mutant cells prevented normal vulval induction.
Mosaic animals lacking lin-2(+) in the six vulval precursor cells (due to a loss of the extrachromosomal array in either AB or AB.p) almost always displayed defects in vulval induction (ten of eleven cases) (Fig. 5). Conversely, animals exhibited normal vulval induction if the array was lost in cells other than the vulval precursor cells, including the anchor cell or most of the hyp7 syncytial epidermis. Specifically, we obtained nine mosaic animals that lacked lin-2(+) in the anchor cell (loss in P1, EMS or MS.p), and four mosaic animals that lacked lin-2(+) in most of the nuclei in the hyp7 syncytium (loss in AB.a). All thirteen of these mosaics exhibited normal vulval induction. These results suggest that lin-2 acts in the vulval precursor cells to transduce the anchor cell signal.
lin-2 mosaic analysis. (A) The cell lineage giving rise to the tissues known to be involved in vulval development is shown. Vertical lines represent single cells, horizontal lines represent cell divisions, and arrows indicate many cell divisions. Each mark represents an individual mosaic animal, and the marks are placed next to the cell that lost the extrachromosomal array based on scoring the Ncl phenotype of individual cells in each mosaic animal. Squares denote non-Lin animals that showed normal vulval induction from three vulval precursor cells when viewed using Nomarski optics in the L4 and normal egg laying in the adult. Circles denote Lin animals that showed either no induction or partial induction in the L4 stage and were egg-laying defective as adults. The incomplete penetrance of the Vul phenotype of lin-2(e1309) probably accounts for one mosaic animal that lacked the array in the Pn.p cells but still generated a normal vulva. (B) Nomarski photomicrographs of the lin-2(e1309) mosaic animal described in Table 1 row 3 at the early fourth larval stage. Two focal planes are shown. Cells derived from Ncl or non-Ncl vulval precursor cells are denoted in white and black, respectively. P5.p, P6.p and P7.p expressed 2°, 1° and 2° cell fates, respectively, and generated a wild-type vulva. P5.p and P7.p lacked the array since their descendants (between white arrowheads) were Ncl. P6.p contained the array since its descendants (between black arrowheads and in inset) were non-Ncl. Scale bar, 10 μm.
lin-2 mosaic analysis. (A) The cell lineage giving rise to the tissues known to be involved in vulval development is shown. Vertical lines represent single cells, horizontal lines represent cell divisions, and arrows indicate many cell divisions. Each mark represents an individual mosaic animal, and the marks are placed next to the cell that lost the extrachromosomal array based on scoring the Ncl phenotype of individual cells in each mosaic animal. Squares denote non-Lin animals that showed normal vulval induction from three vulval precursor cells when viewed using Nomarski optics in the L4 and normal egg laying in the adult. Circles denote Lin animals that showed either no induction or partial induction in the L4 stage and were egg-laying defective as adults. The incomplete penetrance of the Vul phenotype of lin-2(e1309) probably accounts for one mosaic animal that lacked the array in the Pn.p cells but still generated a normal vulva. (B) Nomarski photomicrographs of the lin-2(e1309) mosaic animal described in Table 1 row 3 at the early fourth larval stage. Two focal planes are shown. Cells derived from Ncl or non-Ncl vulval precursor cells are denoted in white and black, respectively. P5.p, P6.p and P7.p expressed 2°, 1° and 2° cell fates, respectively, and generated a wild-type vulva. P5.p and P7.p lacked the array since their descendants (between white arrowheads) were Ncl. P6.p contained the array since its descendants (between black arrowheads and in inset) were non-Ncl. Scale bar, 10 μm.
We also identified mosaic animals in which three vulval precursor cells that contained lin-2(+) were intermixed with three cells that lacked it. The row of vulval precursor cells consists of cells derived from AB.pl intermixed with cells derived from AB.pr (Sulston and Horvitz, 1977). Thus, three of the six vulval precursor cells lacked lin-2(+) when the extra-chromosomal array was lost in either AB.pl or AB.pr (or later in the AB.pl/r lineages). Such mosaic animals almost always expressed a normal 2°-1°-2° pattern of vulval cell fates from three vulval precursor cells (Table 1 and Fig. 5B) and generated a functional vulva (Fig. 5A). In the mosaic animals analyzed, P6.p expressed the 1° cell fate if it contained the extrachromosomal array (six cases), but did not express the 1° cell fate if it lacked the array (three cases) (Table 1). This observation indicates that lin-2(+) is required in P6.p for expression of the 1° cell fate. Furthermore, vulval precursor cells that were adjacent to a 1° cell always expressed the 2° cell fate, even if they lacked lin-2(+) (13 cases). This result indicates that lin-2 is not required for the response to the lateral signal during specification of the 2° cell fate. Finally, the patterns of cell fates expressed in these lin-2 mosaic animals are similar to those in let-23 and lin-7 mosaic animals (Simske and Kim, 1995; Koga and Ohshima, 1995). These mosaic analyses indicate that the vulva may be induced by sequential signals, in which the anchor cell signal induces the 1° cell fate in P6.p, and a lateral signal from P6.p then induces the 2° cell fate in P5.p and P7.p (Table 1 legend). In summary, lin-2 and let-23 receptor are both required for vulval induction by the anchor cell and appear to act in the same cell (P6.p).
lin-2 encodes a protein with similarity to CaM kinase II and MAGUKs
The lin-2A transcript encodes a protein composed of two regions of sequence similarity: an N-terminal region similar to Ca2+/calmodulin-dependent protein kinase II (CaM kinase II) and a C-terminal region similar to membrane-associated guanylate kinases (MAGUKs) (Figs 6, 7). CaM kinase II is a protein serine/threonine kinase present in most cell types that can phosphorylate a broad array of substrates in vitro and becomes activated by binding to calmodulin in response to increased intracellular Ca2+ concentrations (for reviews, see Bronstein et al., 1993; Schulman and Hanson, 1993). Although LIN-2A exhibits significant similarity to CaM kinase II, several differences between these proteins suggest that LIN-2A is unlikely to be an active protein kinase (Fig. 7A). Protein kinases contain eleven conserved subdomains (for reviews, see Hanks et al., 1988; Taylor et al., 1992). LIN-2A is missing a Gly residue, a Lys residue and an Asp residue expected in sub-domains I, II and VII, respectively (Fig. 7A). The latter two residues appear to be essential for protein kinase activity, since both are present in all known protein kinases and mutagenesis of either eliminates the enzymatic activity of a number of protein kinases (Hanks et al., 1988).
LIN-2A is similar to CaM kinase II and MAGUK proteins. Structures of the proteins encoded by lin-2A and lin-2B cDNAs. Regions of similarity to the αsubunit of rat CaM kinase II, yeast guanylate kinase, human p55, Drosophila DlgA, rat PSD-95, human ZO-1 and human ZO-2 (aka X104) (Duclos et al., 1994; Ponting and Phillips, 1995) are shown, and the percentage identity of each domain to LIN-2A is indicated. The C-terminal portion of ZO-1 is not shown (diagonal line).
LIN-2A is similar to CaM kinase II and MAGUK proteins. Structures of the proteins encoded by lin-2A and lin-2B cDNAs. Regions of similarity to the αsubunit of rat CaM kinase II, yeast guanylate kinase, human p55, Drosophila DlgA, rat PSD-95, human ZO-1 and human ZO-2 (aka X104) (Duclos et al., 1994; Ponting and Phillips, 1995) are shown, and the percentage identity of each domain to LIN-2A is indicated. The C-terminal portion of ZO-1 is not shown (diagonal line).
LIN-2A sequence comparisons. Residues identical to LIN-2A are boxed in black. (A)Alignment of amino acid sequences of LIN-2A (7-323) and the αsubunit of rat brain Ca2+/calmodulin-dependent protein kinase II (3-314) (Lin et al., 1987). Protein kinase subdomains are overlined and numbered. Residues conserved in all known protein kinases are indicated in bold text below (Hanks et al., 1988). Four positions at which LIN-2A differs from this consensus (!) and two positions altered by site-directed mutagenesis (*) are marked. The autophosphorylation site conserved in CaM kinase II proteins is missing in LIN-2A (#) (Bronstein et al., 1993; Schulman and Hanson, 1993). The putative calmodulin-binding site (overlined) suggests that LIN-2A could be regulated by binding calmodulin in response to changes in calcium levels.(B) Alignment of the amino acid sequences of GLGF/DHR domains in LIN-2A (545-620), p55 (70-145), PSD-95 repeat 3 (312-387), DlgA repeat 3 (485-560), ZO-1 repeat 3 (411-485) and ZO-2 repeat 3 (511-585). Positions at which the chemical character of residues is conserved in >90% of GLGF/DHR domains are indicated by asterisks (Ponting and Phillips, 1995). (B) Alignment of the amino acid sequences of SH3 domains in LIN-2A (637-716), p55 (162-227), DlgA (604-669), PSD-95 (432-497), ZO-1 (508-571) and ZO-2 (608-668). A general consensus is indicated in bold text below (Musacchio et al., 1994). Highly conserved residues (upper case), less well conserved residues (lower case), and positions that define SH3 subclass 3 sequences (&) are indicated. (D) Alignment of amino acid sequences of LIN-2A (774-895), p55 (282-402), DlgA (770-889), PSD-95 (534-653), ZO-1 (632-719), ZO-2 (730-816) and yeast guanylate kinase (1-143). The anion hole of yeast guanylate kinase, predicted to bind the α and β phosphates of ATP, is underlined. Residues that contact the guanine ring (!), the phosphate of GMP (#), and Mg2+ complexed to ATP (+) in yeast guanylate kinase and residues altered by site-directed mutagenesis (*) are marked.
LIN-2A sequence comparisons. Residues identical to LIN-2A are boxed in black. (A)Alignment of amino acid sequences of LIN-2A (7-323) and the αsubunit of rat brain Ca2+/calmodulin-dependent protein kinase II (3-314) (Lin et al., 1987). Protein kinase subdomains are overlined and numbered. Residues conserved in all known protein kinases are indicated in bold text below (Hanks et al., 1988). Four positions at which LIN-2A differs from this consensus (!) and two positions altered by site-directed mutagenesis (*) are marked. The autophosphorylation site conserved in CaM kinase II proteins is missing in LIN-2A (#) (Bronstein et al., 1993; Schulman and Hanson, 1993). The putative calmodulin-binding site (overlined) suggests that LIN-2A could be regulated by binding calmodulin in response to changes in calcium levels.(B) Alignment of the amino acid sequences of GLGF/DHR domains in LIN-2A (545-620), p55 (70-145), PSD-95 repeat 3 (312-387), DlgA repeat 3 (485-560), ZO-1 repeat 3 (411-485) and ZO-2 repeat 3 (511-585). Positions at which the chemical character of residues is conserved in >90% of GLGF/DHR domains are indicated by asterisks (Ponting and Phillips, 1995). (B) Alignment of the amino acid sequences of SH3 domains in LIN-2A (637-716), p55 (162-227), DlgA (604-669), PSD-95 (432-497), ZO-1 (508-571) and ZO-2 (608-668). A general consensus is indicated in bold text below (Musacchio et al., 1994). Highly conserved residues (upper case), less well conserved residues (lower case), and positions that define SH3 subclass 3 sequences (&) are indicated. (D) Alignment of amino acid sequences of LIN-2A (774-895), p55 (282-402), DlgA (770-889), PSD-95 (534-653), ZO-1 (632-719), ZO-2 (730-816) and yeast guanylate kinase (1-143). The anion hole of yeast guanylate kinase, predicted to bind the α and β phosphates of ATP, is underlined. Residues that contact the guanine ring (!), the phosphate of GMP (#), and Mg2+ complexed to ATP (+) in yeast guanylate kinase and residues altered by site-directed mutagenesis (*) are marked.
The C-terminal region common to LIN-2A and LIN-2B is similar to the MAGUK family of proteins. MAGUKs are typically associated with the inner surface of the plasma membrane at cell junctions (for review, see Woods and Bryant, 1993). The family includes ZO-1, ZO-2, DlgA, PSD-95, SAP97 and p55. ZO-1 and ZO-2 are major components of the vertebrate tight junction (Stevenson et al., 1986; Willott et al., 1993; Gumbiner et al., 1991; Jesaitis and Goodenough, 1994). DlgA is a product of the Drosophila tumor suppressor gene lethal(1)discs-large-1 (dlg) and is localized to the septate junction in imaginal discs and other epithelia (Woods and Bryant, 1991). PSD-95 (also known as SAP90) and SAP97 are proteins associated with mammalian synaptic junctions (Cho et al., 1992; Kistner et al., 1993; Müller et al., 1995). p55 is a peripheral membrane protein in erythrocytes (Ruff et al., 1991).
MAGUK proteins contain three distinct types of domain. First, each contains one to three GLGF/DHR domains (Gly-Leu-Gly-Phe/discs-large homology region; also referred to as PDZ (PSD-95/DlgA/ZO-1) domains) (Cho et al., 1992; Woods and Bryant, 1993; Kornau et al., 1995). This domain can mediate protein-protein binding interactions; a GLGF/DHR domain in FAP-1, which is similar to protein tyrosine phos-phatase-BAS, binds to the C terminus of the Fas receptor in vitro (Sato et al., 1995).
Second, MAGUK proteins contain an SH3 domain, which is a protein-protein interaction domain (for review, see Musacchio et al., 1994). The SH3 domain of LIN-2A has an insertion of 12 residues relative to the SH3 domains of other MAGUKs (Fig. 7C). However, alignment of the LIN-2A sequence with the known structures of the SH3 domains of c-Src and c-Crk suggests that the SH3 domain in LIN-2A may be properly folded, since it occurs in the distal loop of the domain between two β-pleated sheets that hold the domain together (Musacchio et al., 1994).
Third, MAGUKs contain a domain that is similar to the enzyme guanylate kinase, which uses ATP to convert GMP to GDP (for review, see Woods and Bryant, 1993). The sequence of LIN-2A is highly conserved at positions predicted to be essential for guanylate kinase activity (Fig. 7D). X-ray crys-tallographic studies have identified ten residues of the yeast enzyme that bind to GMP (Fig. 7D) (Stehle and Schulz, 1992). At the equivalent positions in LIN-2A, six of these residues are identical and four are conservatively substituted. Furthermore, three highly conserved Gly residues in the putative ATP-binding site of yeast guanylate kinase are conserved in LIN-2A (Fig. 7D).
Neither CaM kinase II nor guanylate kinase activity is required for lin-2 function during vulval induction
One tempting model is that the CaM kinase II domain in LIN-2A acts as a protein kinase involved in transducing the anchor cell signal. However, the fact that LIN-2A is missing key amino acids that are highly conserved in protein kinases makes this model less plausible. To further test this model, we con-structed two mutant lin-2 minigenes that encode proteins that lack additional highly conserved amino acids and thus almost certainly lack protein kinase activity, and then tested them in transformation rescue experiments. First, the sequence Ala/Ser-Pro-Glu in protein kinase subdomain VIII (residues 181-183 of the α subunit of rat CaM kinase II) is implicated in binding to ATP (Hanks et al., 1988; Taylor et al., 1992). The Glu residue is present in all known protein kinases, and muta-genesis of this residue either severely reduces or eliminates the kinase activity of cAMP-dependent protein kinase and v-Src, and results in a strong reduction in C. elegans mek-2 gene activity (Gibbs and Zoller, 1991; Kornfeld et al., 1995; Wu et al., 1995). The protein encoded by the mutant lin-2 minigene pCaM1 should lack protein kinase activity, as the conserved Ala-Pro-Glu sequence in LIN-2A (positions 193-195) is changed to Ala-Pro-Ala. Second, all protein kinases contain an Asp residue in subdomain VI (position 135 of the α subunit of rat CaM kinase II) that is at the center of the catalytic site and implicated in binding Mg2+ complexed to ATP (Hanks et al., 1988). Replacement of this Asp residue with Ala in cAMP-dependent protein kinase abolishes kinase activity (Gibbs and Zoller, 1991). The protein encoded by pCaM2 should also lack protein kinase activity, as this conserved residue in LIN-2A (Asp146) is changed to Ala. Both pCaM1 and pCaM2 rescue the lin-2(e1309) Vul phenotype to a significant extent in germline transformation experiments (Fig. 8A). However, the pCaM minigenes were slightly less efficient than the wild-type minigene (plin-2A); pCaM1 and pCaM2 rescued the lin-2(e1309) Vul phenotype in 77% and 49% of animals, respec-tively.
(A) Germline transformation with mutant lin-2A minigenes. The structure of the plin-2A minigene is shown. Boxes interrupted by lines represent exons in genomic DNA sequence. The amino acid changes in each of the mutant minigenes and their function assayed by rescue of the lin-2(e1309) Vul phenotype are indicated. ‘% rescue’ refers to the fraction of non-Vul animals in the best of at least three transgenic lines. Control lines transgenic for the unc-29(+) co-transformation marker were <10% non-Vul. (B) LIN-2A expression in transgenic strains. Whole lysates were prepared from 40-50 non-Unc-29 L4 to young adult animals and analyzed by western blotting, using 8% SDS-PAGE and anti-LIN-2A antiserum. The positions and sizes (in Mr ×10−3) of prestained protein markers (BioRad) and the LIN-2A band are indicated. LIN-2A expression levels in transgenic strains ranged from 2-to 5-fold higher than the wild-type level.
(A) Germline transformation with mutant lin-2A minigenes. The structure of the plin-2A minigene is shown. Boxes interrupted by lines represent exons in genomic DNA sequence. The amino acid changes in each of the mutant minigenes and their function assayed by rescue of the lin-2(e1309) Vul phenotype are indicated. ‘% rescue’ refers to the fraction of non-Vul animals in the best of at least three transgenic lines. Control lines transgenic for the unc-29(+) co-transformation marker were <10% non-Vul. (B) LIN-2A expression in transgenic strains. Whole lysates were prepared from 40-50 non-Unc-29 L4 to young adult animals and analyzed by western blotting, using 8% SDS-PAGE and anti-LIN-2A antiserum. The positions and sizes (in Mr ×10−3) of prestained protein markers (BioRad) and the LIN-2A band are indicated. LIN-2A expression levels in transgenic strains ranged from 2-to 5-fold higher than the wild-type level.
Another tempting model is that the guanylate kinase domain of LIN-2A metabolizes guanine nucleotides and thereby modulates the GTPase activity of LET-60 Ras. We used the same strategy as above to ask whether guanylate kinase activity is required for lin-2 function during vulval induction. The protein encoded by the mutant minigene pGMP should lack GMP-binding and hence guanylate kinase activity, as three of the five residues predicted to bind the phosphate moiety of GMP (Arg811, Arg814 and Glu817 of LIN-2A) are all replaced with Ala (Stehle and Schulz, 1992). The protein encoded by the mutant minigene pATP should lack ATP-binding and hence guanylate kinase activity, as the putative ATP-binding consensus sequence is disrupted by deletions of three amino acids (residues His783, Gly784 and Gly786 of LIN-2A). (The mutant sequence in pATP resembles the wild-type sequences in DlgA, PSD-95, ZO-1 and ZO-2.) Both pGMP and pATP rescue the lin-2(e1309) Vul phenotype as efficiently as the wild-type minigene (plin-2A) in germline transformation experiments (Fig. 8A).
We determined the level of LIN-2A protein expression in these transgenic strains to exclude the possibility that rescue by the mutant minigenes was due to strong overexpression of weakly active LIN-2 proteins. The transgenic strains shown in Fig. 8A were analyzed by western blotting, staining with anti-LIN-2 antiserum, and quantitating the relative levels of LIN-2A expression by scanning densitometry. We found that the level of LIN-2A expression in the transgenic strains was approximately 2-to 5-fold higher than the wild-type level (Fig. 8B). These slightly increased LIN-2A protein levels should not compensate for greatly reduced or absent enzymatic activities, indicating that vulval induction in these strains is independent of CaM kinase or guanylate kinase activity.
We conclude that neither protein kinase activity nor guanylate kinase activity is required for lin-2 function in vulval development. Thus, LIN-2A may have a structural rather than an enzymatic function in the transduction of the anchor cell signal. This function may be mediated in part by the GLGF/DHR and SH3 protein-protein interaction domains in LIN-2A. The observation that the pCaM1 and pCaM2 minigenes were slightly less efficient than the wild-type minigene in rescuing the lin-2 vulvaless phenotype in trans-formation experiments suggests that there may be a non-enzymatic role for the CaM kinase II domain in vulval devel-opment. A catalytically inactive CaM kinase II domain may be required for LIN-2A structure or stability, or as a protein-protein interaction domain. Finally, the observation that the minigenes pGMP and pATP were just as efficient as the wild-type minigene suggests that not only guanylate kinase activity but even ATP- and GMP-binding to LIN-2A are dispensable for its function during vulval induction.
DISCUSSION
lin-2 is required for normal activation of the let-23 receptor/let-60 ras signaling pathway that induces the expression of the 1° cell fate in C. elegans vulval development. We have shown that lin-2 acts downstream of lin-3 EGF and upstream of let-60 ras, and that lin-2 activity is required in the vulval precursor cell P6.p. lin-2 encodes a protein, LIN-2A, that has an N-terminal region similar to CaM kinase II and a C-terminal region similar to MAGUKs. Neither of these proteins has been implicated previously in receptor tyrosine kinase/Ras signaling pathways. Surprisingly, lin-2 function in vulval induction requires neither the potential CaM kinase II activity nor the potential guanylate kinase activity that were predicted by sequence similarity. As discussed below, the structural role of MAGUK proteins at tight/septate junctions and the role of these junctions in cell polarity suggest how a cell junction component of the vulval precursor cells could be required for activation of the let-23 receptor/let-60 ras signaling pathway. We propose that LIN-2A is required for the localization of signal transduction molecules (such as LET-23 receptor) to either the basal membrane domain or the cell junctions, and that asymmetric localization of signaling molecules is crucial to their signaling functions. Direct evidence in support of this model has recently been obtained in our laboratory (J. S. Simske, personal communication).
Cell signaling in epithelia
Epithelial cells are polarized, partitioned into apical and basal membrane domains by specialized cell junctions (called tight junctions in vertebrates and septate junctions in arthropods) that form lateral belts around the circumference of each cell and a continuous seal between cells. Tight/septate junctions prevent mixing of lipids and membrane-associated proteins between the apical and basal membrane domains of individual cells and provide a regulated barrier to diffusion across epithelia in the extracellular space.
As tight/septate junctions prevent passive diffusion across epithelia, extracellular signals generally contact epithelial cells either apically or basally. Perhaps for this reason, receptors can be distributed asymmetrically on epithelial cells, such that receptor density is highest on the cell surface exposed to the signal. For example, hepatocyte growth factor receptor (also known as scatter factor receptor) and EGF receptor are localized predominantly on the basolateral surface of MDCK cells and bind ligands that are presented basolaterally (Maratos-Flier et al., 1987; Mullin and McGinn, 1987; Crepaldi et al., 1994). In Drosophila, EGF receptor, the receptor tyrosine kinase Sevenless and the receptor Notch are all localized apically in epithelial cells (Banerjee et al., 1987; Tomlinson et al., 1987; Fehon et al., 1991; Zak and Shilo, 1992). These observations suggest that receptor localization may be important for signal transduction in epithelial cells. However, there is as yet no reported evidence of the conse-quences of receptor mislocalization for cell signaling, so the functional significance of receptor localization is unknown.
Several proteins present at the tight/septate junction have been characterized, including the MAGUK proteins ZO-1, ZO-2 and DlgA. ZO-1 and ZO-2 are major structural components of the vertebrate tight junction (also known as the zonula occludens) (Stevenson et al., 1986; Gumbiner et al., 1991; Jesaitis and Goodenough, 1994). ZO-1 and ZO-2 bind to each other and to other components of the tight junction, including the transmembrane protein occludin (Gumbiner et al., 1991; Furuse et al., 1994; Jesaitis and Goodenough, 1994). Drosophila DlgA is localized to the septate junction in imaginal discs and other epithelia (Woods and Bryant, 1991). In dlg mutants, septate junctions are disrupted, epithelial cell polarity is lost, and epithelial cells undergo neoplastic growth, suggesting that septate junctions have an important role in an inhibitory signaling pathway that limits growth (Stewart et al., 1972; Abbott and Natzle, 1992; Woods and Bryant, 1989). The signaling genes that interact with dlg in Drosophila have not yet been identified.
lin-2 function during vulval induction
The most important point of this work is that a MAGUK protein (LIN-2A) is required for normal activation of a receptor tyrosine kinase/Ras signaling pathway in an epi-thelial cell (the vulval precursor cell P6.p). It is presently unclear how LIN-2A functions and how it interacts with this highly conserved signaling pathway during vulval induction. However, previous work on MAGUK proteins and the asym-metric localization of transmembrane receptors in epithelial cells together provide a framework for the following model. In P6.p, the MAGUK protein LIN-2A may be associated with the cell junction, and the LET-23 receptor may be localized either to the basal membrane domain facing the anchor cell or at the cell junction. LIN-2A may be required either to establish or maintain the localization of LET-23, and this localization may be important for activation of LET-23 by the LIN-3 ligand. For example, receptor localization may increase receptor density and promote receptor dimerization, it may allow the receptor to interact with downstream signal trans-duction molecules that are also localized, or it may simply prevent LET-23 receptor from appearing on the apical surface where it would be inactive (because LIN-3 ligand is presum-ably excluded from this surface). In lin-2 mutants, LET-23 would be mislocalized on vulval precursor cells and so would fail to be activated by LIN-3.
lin-2 mutations do not eliminate the polarity of the vulval precursor cells; the general morphology of the cells (viewed by Nomarski microscopy) and the appearance of their cell junctions (viewed by immunocytochemical staining with an antibody that recognizes the belt desmosome (MH27; Francis and Waterston, 1991) appear the same in lin-2 mutants and wild type (Fig. 2B; J. S. Simske, personal communication). lin-2 differs from dlg in this regard, as strong loss-of-function dlg mutations disrupt the integrity of the septate junction and destroy cell polarity in Drosophila epithelia (Stewart et al., 1972; Abbott and Natzle, 1992; Woods and Bryant, 1989). Thus, the signaling defect in lin-2 mutants is not an indirect consequence of a general loss of cell polarity, but more likely reflects a specific effect of LIN-2A on the activity of signaling molecules such as the LET-23 receptor.
The receptor localization model proposes a structural role for LIN-2A in intercellular signaling that depends on protein-protein interactions. MAGUK proteins are known to interact with other proteins, and they have several protein-binding domains. ZO-1 binds to both ZO-2 and occludin (Gumbiner et al., 1991; Furuse et al., 1994), and both p55 and a human DlgA homolog bind band 4.1 in vitro (Marfatia et al., 1994; Lue et al., 1994). These interactions may be mediated by the SH3 domain and the GLGF/DHR domains of MAGUKs, since both types of domain are known to mediate protein binding (Cicchetti et al., 1992; Ren et al., 1993; Sato et al., 1995). Fur-thermore, even the protein kinase and guanylate kinase domains in LIN-2A may have important roles in protein binding, since they have been conserved in evolution, but their enzymatic activities do not appear to be required for vulval induction.
We have considered other models to explain the lin-2 vulvaless phenotype. The presence of a calmodulin-binding motif in LIN-2A suggests that vulval induction could be regulated by intracellular Ca2+ levels, which could modulate binding of calmodulin and thereby regulate LIN-2A activity (Schulman and Hanson, 1993). In addition, the observation that ZO-1 may be a tyrosine kinase substrate (Kurihara et al., 1995; van Itallie et al., 1995) suggests that LIN-2A could be phosphorylated by LET-23 receptor tyrosine kinase. Both of these observations might suggest that lin-2 acts in a signal transduction pathway parallel to the let-60 ras pathway, and that activation of both pathways is required for expression of the 1° cell fate. However, our genetic results are inconsistent with parallel pathway models. Such models predict that acti-vation of the Ras pathway should not compensate for a defect in a parallel pathway in a lin-2 mutant, yet we observed that a let-60(gf) mutation completely suppresses the lin-2 vulvaless phenotype (Fig. 2). Instead, calmodulin binding and tyrosine phosphorylation may be mechanisms for modulating the activity of LIN-2A in localizing signal transduction proteins.
Results that directly support the receptor localization model have recently been obtained. A LET-23 reporter protein is localized to the cell junctions of the vulval precursor cells in the wild type but is mislocalized in lin-2 mutants (J. S. Simske, personal communication). Furthermore, overexpression of wild-type let-23 suppresses the Vul phenotype of lin-2(e1309), suggesting that increased receptor density can compensate for receptor mislocalization (J. S. Simske, personal communica-tion).
Implications for other signaling systems
MAGUK proteins are conserved in animals from diverse phyla, suggesting that they have similar functions in all metazoans. We have uncovered a genetic interaction between the LIN-2A MAGUK protein and a receptor tyrosine kinase/Ras signaling pathway in C. elegans. Other MAGUKs may also have signaling functions in receptor tyrosine kinase pathways, and perhaps other intercellular signaling pathways, in all animals. Thus, understanding how LIN-2A functions in C. elegans vulval induction will be important for understand-ing how MAGUK proteins function in signaling pathways in other organisms, and vice versa. We have presented a model for LIN-2A function based on localization of LET-23 receptor that can now be tested in C. elegans.
ACKNOWLEDGEMENTS
The receptor localization model was proposed and evidence in support of it has been obtained by Jeff Simske in our laboratory. We thank Dave Eisenmann, Mark Lackner, Jeff Simske and Patrick Tan for critical reading of the manuscript and all members of the lab for productive discussions. Some nematode strains were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources. We thank Scott Clark, Michael Stern, Leon Avery and Dave Eisenmann for providing strains carrying lin-2 mutant alleles, and Alan Coulson for providing cosmid clones from the C. elegans genome physical map. Finally, we thank former Secretary of State George P. Schultz for technical assistance in obtaining lin-2 transformants. This work was supported by a post-doctoral fellowship from the Cancer Fund of the Damon Runyon-Walter Winchell Foundation to R. H., a postdoctoral fellowship from the Swiss National Foundation to A. F. H., and research grants from the Lucille P. Markey Charitable Trust, the Searle Scholars Program and the National Institutes of Health to S. K. K.