A pair as a minimum: The two fibroblast growth factors of the nematode Caenorhabditis elegans

Authors


Abstract

Fibroblast growth factors (FGFs) regulate many important developmental and homeostatic physiological events. The FGF superfamily contains several families. In this review, we present recent findings on the two FGFs of the nematode Caenorhabditis elegans from both functional and phylogenic points of view. C. elegans has a single FGFR (EGL-15) with two functionally exclusive isoforms, and two FGFs (LET-756 and EGL-17), which play distinct roles: an essential function for the former, and guidance of the migrating sex myoblasts for the latter. Regulation of homeostasis by control of the fluid balance could be the basis for the essential function of LET-756. Phylogenetic and functional studies suggest that LET-756, like vertebrate FGF9, -16, and -20, belongs to the FGF9 family, whereas EGL-17, like vertebrate FGF8, -17, and -18, could be included in the FGF8 family. Developmental Dynamics 232:247‒255, 2005. © 2004 Wiley-Liss, Inc.

INTRODUCTION

Fibroblast growth factors (FGFs) are peptide regulatory molecules that ensure several functions in development and adult life, including patterning, cell–cell communication, cell migration, and tissue homeostasis. Once secreted in the extracellular compartment, their effects are mediated by high affinity transmembrane tyrosine kinase receptors (FGFR) in an autocrine or paracrine manner. There are four FGFRs in mammals, each with different isoforms. The prototypic FGFR has three immunoglobulin-like domains (IgL) in its extracellular region and a tyrosine kinase domain in its intracellular region. Additional or complementary routes of FGF action have been described, such as interaction with heparan sulfate proteoglycans (HSPG) that serve as coreceptors in FGF/FGFR interactions, and intracellular trafficking including nuclear targeting. The FGF superfamily contains at least 22 members in mammals, distributed in several families (Ornitz and Itoh, 2001; Fig. 1). The four members of one of these families (FGF11 family), the fibroblast homologous factors (FHF), do not interact with FGFRs and remain intracellular (Shoorlemmer and Goldfarb, 2001). Each of the other FGFs interacts with several FGFRs, and reciprocally, each FGFR binds several FGFs (Ornitz et al., 1996). Thus, it is difficult to define the network of FGF signaling in mammals. In nonvertebrates, only three FGFs have been identified in Drosophila (BNL, pyramus and thisbe), and two in the nematode Caenorhabditis elegans, EGL-17 (Burdine et al., 1997) and LET-756 (Roubin et al., 1999). C. briggsae has two FGFs related to EGL-17 and LET-756, respectively. There is only one identified FGFR in C. elegans, EGL-15, among the 40 tyrosine kinase receptors of this animal (Popovici et al., 1999). In vertebrates, the high number of FGF members, their large range of activities and complex signaling have made difficult the precise understanding of FGF functions. The nematode worm C. elegans is a model system with reduced cellular and anatomical complexity compared with mammals. The complexity of the FGF/FGFR network is also much reduced in this animal, providing for different insights. Here, we describe the recent findings on the FGF/FGFR system in C. elegans.

Figure 1.

A: Nonrooted tree representation of phylogenetic relations in the fibroblast growth factor (FGF) superfamily. The FGF tree includes FGF core sequences from Homo sapiens, Mus musculus FGF15, Ciona intestinalis (asterisks), Drosophila melanogaster (BNL, PYR, THS), and Caenorhabditis elegans (EGL-17, LET-756). The seven FGF families are labeled according to the founding member of the family and are shown in different colors. Some bootstrap values are indicated (significance is above 700). Our hypothesis is that a first expansion, prior the separation of protostomians and deuterostomians, brought the number of FGF genes to six or seven. Losses occurred in protostomians, whereas a second expansion, along the 2R theory, occurred in the vertebrate lineage. According to this hypothesis EGL-17 belongs to the FGF8 family and LET-756 to the FGF9 family (inclusion is supported by sequence alignments or functional experiments but not by significant bootstrap values). B: Chromosomal localization of FGF and FGFR genes in three species. A similar clustering of genes from FGF8 family and FGFR in C. elegans and H. sapiens species might be purely coincidental or might indicate that a common ancestor region to C. elegans and vertebrates contained an FGF and an FGFR gene and was conserved en bloc during bilaterian evolution. Pyramus (pyr) and thisbe (ths) are two ligands for htl/FGFR. SLIT and UNC5 genes are linked to genes of the FGF8 family in the human (FGF8/17/18) and Drosophila (pyr, ths) genome. The vab-15, Dr, and MSX metaHOX genes (Coulier et al., 2000) are shown as another example of family. This conserved synteny might be a remnant of an ancestor locus controlling migration in an early metazoan. The drawing is not to scale, but the exact location on chromosome is indicated below each locus (http://www.ncbi.nih.gov/entrez/, http://www.ensembl.org/Homo_sapiens/). SLIT family: slt-1, sli, SLIT.

C. ELEGANS FGFs FROM A PHYLOGENETIC POINT OF VIEW: A SEVEN-FAMILY GAME

A series of large-scale duplication events is thought to have occurred at the base of vertebrate ancestry. The “2R hypothesis” (Ohno, 1970), as it is now commonly referred to, assumes that two rounds of large-scale duplication in a chordate ancestor led to a large increase in the number of genes (Popovici et al., 2001; Wolfe, 2001; Leveugle et al., 2003). The C. elegans FGFR EGL-15 is an ortholog of vertebrate FGFRs (FGFR1 to -4), and expansion of the FGFR family fits well with the 2R hypothesis (Coulier et al., 1997). In contrast, based simply on phylogenetic analyses, it is more difficult to draw a clear picture of FGF gene evolution (Ornitz and Itoh, 2001; Itoh and Ornitz, 2004). Vertebrate and nonvertebrate FGFs share sequence similarities in a core region. Based on these core sequences, a phylogenetic tree of FGFs can be constructed (Fig. 1A). The FGF superfamily can be separated in seven major families. As in a recent review (Itoh and Ornitz, 2004), these families hereafter will be designated by the name of their founding member.

Could LET-756 or EGL-17 be assigned to a particular FGF family? By using an approach called “functional phylogeny,” we have shown that LET-756 could be related to the FGF9/16/20 family (FGF9 family) because the core of FGF9, -16, or -20, introduced in a LET-756 backbone, could rescue the let-756 mutant (Popovici et al., 2004). Although bootstrap values are just below the limit of significance, the inclusion of EGL-17 in the FGF8/17/18 family (FGF8 family) is tempting. This reasoning is reinforced by the recent findings of two new FGF genes in Drosophila, pyramus, and thisbe, which resemble FGF8 upon sequence alignment (Gryzik and Muller, 2004; Stathopoulos et al., 2004). We thus favor the constitution of FGF8 family with mammalian FGF8/17/18, Drosophila pyramus and thisbe, and C. elegans EGL-17; work on the small ascidian Ciona intestinalis, a chordate nonvertebrate, has also identified a potential member of this family (Satou et al., 2002). Thus, we suggest that two families of FGF genes, FGF8 and FGF9 families, are represented in the present-day C. elegans genome and that the other FGF families, an ancestor of which existed in the common protostomian/deuterostomian ancestor, may have been lost in the nematode lineage.

These phylogenetic considerations may be further supported by a possible conservation of synteny between genes encoding ligands and receptors in C. elegans and humans. In humans, FGF genes of the FGF8 family belong to the same regions of paralogy (paralogon 4/5/8/10) as the FGFR genes (Fig. 1B). In C. elegans, egl-17 and egl-15 are on the same chromosome. This finding may suggest that they all derive en bloc from a common ancestor region that contained an FGF (the ancestor of the FGF8 family) and an FGFR gene (however, in Drosophila FGF8-like genes pyramus and thisbe have been separated from the FGFR receptor genes). This hypothesis is reinforced by the fact that other genes, such as the ones encoding SLIT and UNC5 proteins, map close to members of the FGF8 family (Fig. 1B).

EGL-17/EGL-15(5A) AND LET-756/EGL-15(5B): TWO MODULES WITH DISTINCT FUNCTIONS

The let-756 gene, located on chromosome III (Fig. 1B), encodes an FGF of 425 amino acid residues and a predicted molecular mass of ∼50 kDa (Fig. 2A). LET-756 has a conserved core region characteristic of all FGFs, three putative bipartite nuclear localization signals (NLS), a region of similarity with lamin and a stretch of polyglutamine in its C-terminal part but has no recognizable N-terminal hydrophobic signal sequence (Roubin et al., 1999). Several let-756 mutants have been characterized. In a severely affected strain (s2887), a loss-of-function let-756 allele causes developmental arrest early in the larval stages; in a partial loss-of-function allele (s2613), a mutation introduces a stop codon at the R318 position (hence, the designation R318 mutant), which truncates the C-terminal quarter of the protein (which contains the lamin-like region and the Q-rich stretch; Fig. 2A). This mutation allows a few worms to develop up to the adult stage. The egl-17 gene, located on chromosome X, encodes a protein of 216 amino acid residues, and a predicted molecular mass of ∼25 kDa. EGL-17 has a conserved core region. It has a potential signal sequence detectable by PSORT (http://psort.nibb.ac.jp/). Several egl-17 mutants have been described (Burdine et al., 1997). The works of M. Stern's laboratory, ours and others (Bülow et al., 2004) have established that egl-17 and let-756 genes have distinct patterns of expression (Fig. 2B).

Figure 2.

A: Structure of Caenorhabditis elegans fibroblast growth factors (FGFs) and human FGF6 (to show an FGF with a bona fide signal peptide) and FGF9. NLS, nuclear localization signal; Q-rich, stretch rich in glutamine residues. The EFISIA motif serves as a noncleaved export signal. B: Expression territories of egl-17 (red) and let-756 (green) genes.

C. elegans FGFs interact with a single FGFR, EGL-15 (De Vore et al., 1995). The egl-15 gene, located on chromosome X, encodes two receptor isoforms, EGL-15(5A) and EGL-15(5B). They result from alternative splicing of exon 5 encoding a specific insert between IgL1 and IgL2 (Goodman et al., 2003). From genetic data, Goodman and coworkers concluded that this insert could be a site of binding for EGL-17 and LET-756 (Goodman et al., 2003): EGL-15(5A) is, thus, the receptor for EGL-17, and EGL-15(5B) that of LET-756 (Fig. 3A). Specific egl-15 partial loss-of-function mutants exhibit defects in the migration of a pair of cells known as the sex myoblasts (SM), which lead to animals that are unable to lay eggs properly (egg-laying defective or Egl; DeVore et al., 1995; Borland et al., 2001). Complete loss-of-function is lethal. Signaling through EGL-15 may be affected by mutations in downstream or regulatory components of the FGFR signaling pathway and lead to a variety of defects (Borland et al., 2001). Thus, mutations of the EGL-15 associated CLR-1 phosphatase enhance FGFR signaling; they are characterized by the accumulation of fluid in the pseudocoelomic cavity (Kokel et al., 1998). Reciprocally, the attenuation of EGL-15 signaling, for example after mutation of SEM-5/GRB2 or LET-341/SOS downstream adaptor substrates, leads to suppression of a Clr (clear) phenotype (i.e., to a Soc phenotype), to a scrawny body phenotype (Scr), or to developmental arrest (a Let phenotype).

Figure 3.

A: Schematic representation of the fibroblast growth factor (FGF) ligand and receptor system in Caenorhabditis elegans (adapted from Goodman et al., 2003). The two FGF ligands, EGL-17 and LET-756, each bind to a specific isoform of the FGFR/EGL-15 receptor generated by alternative splicing of the egl-15 mRNA. Specificity is associated with exon 5-encoded insert (pink/yellow). B: Migration of sex myoblasts. From an initial posterior position in L1 stage larvae (a), hermaphrodite sex myoblasts (SMs), which express EGL-15/FGFR (blue), migrate anteriorly during the L2 stage. In the adult, they have become sex muscles around the vulva and gonad center (b) under stimulation exerted by EGL-17/FGF (red arrow) as a gonad- and vulva-dependent secreted chemoattracting signal (Burdine et al., 1998; Branda and Stern, 2000; Goodman et al., 2003) and build the uterine and vulval musculature. Mutations in egl-15 and egl-17 prevent correct positioning of SMs and egg-laying defects. DTC, distal tip cells of the gonad at the end of their migration in the adult. C: Image of a L1 LET-756 null mutant [dpy-17 (e164) let-756 (s2887) unc-32 (e189)] to illustrate the essential function of LET-756/FGF. Large vacuoles are present in both the head and the body.

At first examination, there seems to be little redundancy between the two FGF ligands. Like some loss-of-function egl-15 mutants, egl-17 mutants have an egg-laying defect due to abnormal SM migration (Burdine et al., 1997). In contrast, complete loss-of-function of let-756 results in lethal arrest (Fig. 3C), and partial loss-of-function mutants are small to lethal but SM migration is not affected. There appears to be no synthetic phenotype in egl-17/let-756 double mutants (Roubin et al., 1999; Huang and Stern, 2004). Thus, the two ligands seem to mediate different functions. This specificity is in a large part determined by the specific receptor isoform. Goodman et al. (2003) have shown that the EGL-17/EGL-15(5A) module is responsible for mediating SM chemoattraction, whereas the LET-756/EGL-15(5B) module carries an essential function (Fig. 3B,C). These conclusions are drawn only from genetic studies, and no biochemical experiments have yet been performed to show that FGF ligands can actually bind to and activate specific EGL-15 isoforms.

FGF/FGFR SIGNALING IN C. ELEGANS

EGL-15 signaling pathway has been relatively well defined by using genetic screens (Borland et al., 2001, for review). Clr and Soc mutants have been instrumental in identifying proteins associated with EGL-15 signaling. Several of these proteins are conserved in mammalian FGFR signaling. This conservation is the case for proteins of the RAS-MAPK cascade, including GRB2/SEM-5, SOS/LET-341, RAS/LET-60, RAF/LIN-45, MAPK/MPK-1 (Fig. 4). This cascade had been implicated early in the control of SM migration (Sundaram et al., 1996; Chen et al., 1997). However, other proteins are unique to the nematode or to mammals (Kokel et al., 1998; Borland et al., 2001; Schutzman et al., 2001); this is the case of phospholipase C-γ and FRS2/SNT1, which are among the best-characterized substrates of FGFR in mammals (Goldfarb, 2001), but which have not been identified in FGFR signaling pathway in C. elegans. Reciprocally, membrane-associated CLR-1 phosphatase regulates negatively EGL-15 but has no identified equivalent in mammals (Kokel et al., 1998). SOC-1 is a multisubstrate adaptor protein similar to mammalian GAB1 and Drosophila DOS but appears to be relatively more specific to FGFR signaling (Schutzman et al., 2001). PTP-2 is similar to Drosophila corkscrew and mammalian phosphatase SHP2. SOC-1 and PTP-2 act together to regulate EGL-15 signaling. SOC-2/SUR-8 is a protein with leucine-rich repeats that is also involved in the positive regulation of EGL-15 (Selfors et al., 1998). It is supposed to enhance the activity of the cascade by direct interaction with LET-60/RAS (Sieburth et al., 1998; Li et al., 2000).

Figure 4.

Schematic representation of fibroblast growth factor (FGF)/FGF receptor (FGFR) signaling. Proteins found in both Caenorhabditis elegans and mammals are indicated with their two names (the worm name is underlined). FGFs (yellow) interact with their cognate receptor (blue) in the presence of heparan sulfate proteoglycans (HSPG). Downstream signal is relayed by a series of adaptors and enzymes and reaches the nucleus to trigger transcription. Regulators of this pathway, such as CLR-1 (negative regulation) and PTP-2 (positive regulation) phosphatases, are indicated to the right. In addition, FGF can reach the nucleus by alternative routes from the surface or from the cytoplasm.

Thus, both a well-conserved, nonspecific downstream RAS-MAPK–positive effector module and a distinct, not so well-conserved, positive (SOC-2, PTP-2) and negative (CLR-1) regulatory module collaborate to ensure proper EGL-15 signaling.

DEVELOPMENTAL ROLES OF FGF/FGFR IN THE NEMATODE: CONTROL OF CELL MIGRATION AND AXON GUIDANCE

EGL-17/EGL-15(5A) and migration/guidance of sex myoblasts. EGL-17 was first described as a factor that specifically controls SM migration in C. elegans (Stern and Horvitz, 1991; Burdine et al., 1997). The SMs begin as a pair of muscle precursor cells born at the posterior of L1 stage larvae and, in the hermaphrodite, migrate anteriorly to functional positions flanking the central gonad and developing vulva, where they differentiate into uterine and vulval muscles required for egg laying (Fig. 3B). Proper migration of the hermaphrodite SMs is necessary for egg-laying proficiency (Branda and Stern, 2000). Mutations of EGL-17 or EGL-15 (special class of egl mutations) cause SMs to be severely posteriorly displaced. In a model described by Burdine et al. (1998), EGL-17, which is expressed in the developing vulva and gonad (Branda and Stern, 2000; Inoue et al., 2002), serves as a gonad-dependent attractant signal to correctly position SMs, which express EGL-15(5A), and counteracts a nonidentified repulsion system. Several potential transcription factors, such as LIN-39/HOX, which control egl-17 gene expression in vulval cells, have been identified (Cui and Han, 2003).

The pattern of expression of EGL-17 has not been completely and precisely established, and it is possible that this FGF plays a role in other areas of the nematode body. Nevertheless, even if it is not immediately visible from the mutant phenotype—because of a subtle modulatory role or the existence of functional redundancy with other factors—EGL-17 probably plays other roles in cell migration. The distal tip cells (DTCs) of the hermaphrodite gonad lead the extensions of two U-shaped tubular arms during the third and fourth stages of larval development. Migration of DTC from ventral midbody to their dorsal position is regulated by several coordinated guidance signals (Branda and Stern, 1999; Lehmann, 2001, for reviews). The unc-52 gene, which encodes the basement membrane HSPG perlecan, plays a role in growth factor signaling, including EGL-17. In a genetically compromised unc-52 mutant organism, EGL-17's role on DTC migration can be revealed (Merz et al., 2003). The L1-like adhesion molecule LAD-1, which is expressed to the plasma membrane at sites of cell–cell contacts in all cells, is phosphorylated by EGL-15 (Chen et al., 2001). It is possible that FGFR, under EGL-17 stimulation, transitorily releases cell–cell contacts at the tip of the gonad to allow migration of DTC.

LET-756/EGL-15 and migration/guidance of neurons. Recent works have shed further light on FGF functions in C. elegans. In addition to SMs, EGL-15 is expressed in the hypodermis (Huang and Stern, 2004). This expression and correct FGFR signaling are required for proper axon outgrowth at the ventral midline (Bülow et al., 2004). Both axon extension and axon guidance are affected in case of EGL-15 signaling defects. For outgrowth, axons use a special substratum made by hypodermis and muscle at the midline of the ventral cord. LET-756 is produced by the muscle cells (Fig. 5A). Hypodermally expressed EGL-15(5B) is activated by LET-756 secreted by the adjacent muscle cells. Thus, whereas in vertebrates, FGFR acts autonomously in the growth cones of neurons, in C. elegans, the FGF/FGFR system is associated with two types of cells to provide the substratum for axon outgrowth and acts nonautonomously and indirectly. During development, the LET-756/EGL-15(5B) system acts through kinase activation and downstream signaling for neuron growth and positioning (Fig. 5B). In addition, the EGL-15(5A) isoform regulates the maintenance of axon position of the several ventral cord axons (Bülow et al., 2004). This function of FGF may be coordinated with that of other growth factors, for example, transforming growth factor beta (TGFβ; Colavita et al., 1998).

Figure 5.

A: Schematic representation of a transverse section of adult Caenorhabditis elegans showing the ventral cord structure (modified from Bülow et al., 2004). The hypodermis expresses EGL-15 FGF receptor (blue) and muscles express FGF/LET-756 (green). In the embryo, guidance is dependent on muscle/hypodermis substratum. In the adult, axon maintenance function is provided by EGL-15 ectodomain. B: LET-756::GFP is found in body muscle cells (top) and in various neurons, i.e., a ventral cord motor neuron and its associated commissures, which innervate body muscles (bottom).

A more global view. The two processes of cell migration and axon growth cone guidance are highly related, and the same mechanisms and molecules participate in their temporal and spatial regulation. This finding seems to be a general phenomenon in bilaterian biology. In C. elegans, it was noted many years ago (Hedgecock et al., 1987) and is well-documented (Blelloch et al., 1999; Branda and Stern, 1999, Montell, 1999; Lehmann, 2001, for reviews). Thus, for example, the disintegrin and metalloprotease UNC-71 is required for both processes (Huang et al., 2003). Mutations of unc-73 or ina-1 functions lead to defects in cell migration and axon organization (Baum and Garriga, 1997; Desai et al., 1988). In C. elegans, cell migration and axon pathfinding may have been organized around a common theme and a limited number of molecules, involving secreted signals (growth factors, receptor ectodomains, chemotactic molecules) and remodeling and adhesion proteins. Among the latter, the four major systems of guidance used in other animals—that is, netrins (UNC-6) and netrin receptors (UNC-5, UNC-40), SLIT (SLT-1) and SLIT receptors ROBO (SAX-3), semaphorins and receptors (neuropilins and plexins), ephrins and receptors (VAB-1)—play a major role (Blelloch et al., 1999; Branda and Stern, 1999, Montell, 1999, Lehmann, 2001, for reviews). Mutations of the CLR-1 phosphatase influence both FGFR and netrin signaling (Chang et al., 2004); whether EGL-15 plays a role in netrin-mediated guidance pathways remains to be established. Similarly, HSPGs are required for the activity of both FGFR and SLIT/ROBO (Merz et al., 2003; Steigemann et al., 2004). It should be interesting to determine whether FGFR signaling could influence one or several of the above-mentioned guidance systems more directly, in a manner similar to vascular endothelial growth factor receptors, which form complexes with neuropilins. We surmise that, even if the mutant phenotypes do not indicate their participation directly, the two FGFs of C. elegans are attractants and/or repellents in several migration systems of the nematode. As in vertebrates (Tickle, 2003), C. elegans FGFs act in concert with other gradients of growth factors, namely TGFβ/BMPs and WNTs, to regulate organogenesis, and anteroposterior and dorsoventral positioning. As an example, EGL-17, TGFβ growth factors DBL-1 and UNC-129, and EGL-20/WNT, regulate DTC migration (Merz et al., 2003); another example concerns male spicule morphogenesis, which requires both FGF and TGFβ signaling (Jiang and Sternberg, 1999). C. elegans may be a good model to examine the complex phenomenon of cell migration, and this interplay of factors and receptors because there are only a limited number of cells that move significantly during development, and these migrations are invariant from animal to animal. Genetic screens aiming at double defects of axon guidance and cell migration may provide interesting information.

PHYSIOLOGICAL ROLES OF FGF/FGFR IN THE NEMATODE: REGULATION OF TISSUE HOMEOSTASIS

Increased EGL-15 signaling leads to fluid accumulation and confers a Clr phenotype, whereas decreased EGL-15 signaling results in a Soc, Scr, or Let phenotype. Huang and Stern (2004) have studied the molecular basis of the Clr phenotype, in which animals display accumulation of clear fluid within the pseudocoelomic space upon mutation in the phosphatase-encoding clr-1 gene. EGL-15 is expressed in the hypodermis and may regulate fluid balance upon LET-756 paracrine stimulation (Fig. 5A). The RAS/MAPK cascade is involved in this activity as MAPK/MPK-1 carries a Soc function that counteracts CLR-1 activity. Regulation of LET-756 may constitute a fluid sensing mechanism in the hypodermis. Fluid balance would be ensured by regulating the permeability of the hypodermis and/or the activity of ion channels. We have also shown that LET-756 is expressed in the CAN neurons (Fig. 5B), which control the excretory canal. In the worm, the latter is by itself responsible for the regulation of osmolarity and allows the traffic of nutrients and growth factors from one site to another.

Another piece of work showed that EGL-15 activation is involved in protein degradation in C. elegans muscle (Szewczyk and Jacobson, 2003). Proteolysis in muscle is induced in times of need by several extracellular stimuli. In C. elegans, muscle degradation is dependent on the RAS pathway. EGL-15 activation by either LET-756 or EGL-17 triggers the RAS-MAPK pathway and consequently body-wall and vulval muscle proteolysis. The endogenous proteins that are degraded in response to FGFR activation have not been identified. FGFR and RAS signaling may also stimulate protein synthesis. A balance between FGFR-induced protein degradation and FGFR-induced protein synthesis may regulate differentiation and tissue remodeling. This balance may also be coupled to the migratory processes described above. Here also, FGFR-induced processes may be coordinated with other growth factor signalings.

Finally, EGL-15 phosphorylates LAD-1, and thus may regulate tissue morphogenesis and maintain tissue integrity by modulating the strength of cell–cell junctions in polarized epithelia and axon–body-wall muscles (Chen et al., 2001).

Thus, FGFs and EGL-15 signal both in muscle and nonmuscle (hypodermis) cells to regulate tissue homeostasis. In mammals, roles for FGF in tissue homeostasis have been demonstrated. Two families, FGF1 and FGF19 families, which are not represented in C. elegans, are particularly involved (Zhou et al., 1998; Shimada et al., 2001; Tomlinson et al., 2002), but other families contribute too. This finding suggests that regulation of tissue homeostasis was a primordial function of the FGFs and remains fundamental, whatever the organism, even if some degree of specialization is developed in species with a high number of FGFs.

MECHANISMS OF SECRETION OF C. ELEGANS FGFs

The majority of FGFs contain classic cleavable N-terminal signal sequences for secretion by the endoplasmic reticulum (ER)/Golgi pathway. However, FGF1 family members and FHFs lack signal sequences. FHFs remain intracellular, whereas FGF1 and FGF2 are excreted by undefined mechanisms. FGF9, FGF16, and FGF20 are efficiently secreted in the ER/Golgi-dependent pathway because of the presence of a high hydrophobic N-terminal region and a six amino acid-long region (EFISIA) located within the core (Miyakawa et al., 1999; Revest et al., 2000). We have shown recently that LET-756, which like FGF9, FGF16, and FGF20 does not have a typical N-terminal signal sequence, is secreted by a Golgi-associated mechanism dependent on the presence of the same motif of six residues as for the FGF9 family; this motif (Fig. 2A) is also essential for the rescue activity of the let-756 loss-of-function mutant (Popovici et al., 2004).

Like some mammalian FGFs (e.g., FGF1 family), LET-756 is also found in specific subareas of the nucleus. The loss-of-function R318 mutant displays a different nuclear sublocalization from the wild-type LET-756 (our unpublished observations). However, so far, the respective roles of secreted and nuclear LET-756 pools are not known. It is possible that LET-756 has subsequently acquired a C-terminal tail that allows a nuclear function to compensate for the extensive loss of FGF members during evolution of the protostomian lineage. With that extension, LET-756 might recapitulate functions ensured in mammals by FGF1 family members.

The mechanism of EGL-17 secretion has also been documented. Proper EGL-17 secretion requires the product of the dab-1 gene (Kamikura and Cooper, 2003). DAB-1 is a cytoplasmic adaptor protein with a PTB domain. It regulates EGL-17 secretion by binding to transmembrane lipoprotein receptors LRP-1 and LRP-2 present in vesicles in the developing vulva. EGL-17 binds to the extracellular domains of the lipoprotein receptors, whereas DAB-1 binds to their tails. It is still not entirely clear how the DAB-1/LRP complex coordinates EGL-17 transport in the secretory pathway. DAB-1 may help in the sorting of LRPs and, consequently, of bound EGL-17 into nascent export vesicles. The DAB-1/LRP complex may also participate in the transport from the trans-Golgi network to the basolateral surface of vulval cells. Perhaps this interaction between an FGF and a lipoprotein receptor should be investigated further for two reasons. First, it may apply to other FGFs. Second, there might be more to it than simple transport. Indeed, dickkopf secreted factors bind lipoprotein receptor-related LRP6, a component of the WNT receptor complex, and modulate the WNT pathway (see Kawano and Krypta, 2003, for review). It is not known whether EGL-17 has also an intracellular function.

FURTHER SPECULATIONS AND PERSPECTIVES

We hypothesize that, after series of gene duplications and losses during evolution, two FGFs have been selected and maintained in the present-day C. elegans to ensure all the functions attributed to these factors in this animal. These functions seem to operate at two levels, a developmental level in which FGFs are required for migration and positioning of cells such as myoblasts and neurons, and a physiological level in the larvae and adult in which FGFs ensure tissue homeostasis and organ maintenance. The identification of families allows for direct comparison of the role of FGF across species, for example between members of FGF8 (FGF8, Ciona FGF8/17/18, PYR/THS, EGL-17) and FGF9 (FGF9, LET-756) families. It can be noted that knock-outs of the Fgf8 and Fgf9 genes are both lethal in the mouse, whereas many other Fgf are dispensable; however, this may not be relevant because at least one FGF from each family (except FGF1 family and perhaps FGF11 family) is associated with a lethal phenotype after gene knock-out. FGF9 family members are strongly involved in cell migration during development of different species (Burdine et al., 1997; Sun et al., 1999; Gryzik and Muller, 2004; Stathopoulos et al., 2004). FGF8 is a major determinant in several developmental areas (Lewandoski et al., 2000). Maintenance functions in various tissues may also be a common theme for several FGFs, especially for FGF8 family, in different species. In keeping with this hypothesis, only two families of FGFs—one identical, FGF8 family (with a gene cis-duplication creating two members), the other different—have been conserved in Drosophila. We surmise that PYR/THS and EGL-17 might be functionally interchangeable. The alternative hypothesis is that LET-756 and EGL-17 are not related to any families but to the few FGFs present in the ancestor of both protostomians and deuterostomians. A first expansion to six FGFs (as found in present-day Ciona) would then have taken place after the separation of protostomians and chordates, followed by a second expansion in the vertebrate lineage along the 2R hypothesis. Resemblance of EGL-17 to the FGF8 family, and of LET-756 to the FGF9 family, would then be the result of convergence.

In mammals, FGFs from different families, for example FGF4 and FGF8, cooperate in limb development (Sun et al., 2002). This finding may not be the case in C. elegans, perhaps because migratory and maintenance processes may not be so complex as to need several rounds of tuning; moreover, budding and branching, which may require complex cocktails of factors in mammals, are not so prominent in nematode development. However, it is possible that not all the functions of the two FGFs have been revealed yet. Exciting prospects for the future are to understand how FGF signals might be coordinated with each other and with other molecules, such as signaling factors, guidance systems and HSPGs, and how they can regulate the dorsal–ventral and anterior–posterior axis of the animal.

Understanding the role and regulation of FGFs and FGFRs may help design new therapies and fight developmental diseases and cancer. In humans, mutations of FGFs and FGFRs are involved in more than a dozen congenital diseases and in various cancers (Coumoul and Deng, 2003 for review). Many effectors and regulators of the FGF/FGFR signaling cascade have been identified. An interesting area of research could be dedicated to the study of the mechanisms of FGF transport and trafficking through the cell compartments and toward the exterior. This discovery could lead to the identification of new regulators and may provide opportunities to act upon FGF signaling in a different way, rather than tampering with FGFR and FGFR downstream effectors.

FGFs that lack classic signal peptide can remain in the cytoplasm or be addressed to the nucleus, either directly or with an FGFR (Schmahl et al., 2004). We would like to propose the following hypothesis: for an FGF to remain intracellular, its mechanism of secretion needs to be modified or even abolished. This can be achieved by the absence of signal peptide (as in the FGF1 and FGF11 families) or use of a nonclassical noncleaved sequence (as in the FGF9 family). During evolution, an FGF intracellular function may have been dissociated from secretion (as in the FGF11 family) or coupled with a nonclassical export (as in the FGF1 and FGF9 families) to become optimal. The modification, which may occur in homologous proteins (FGFs of the same family) or be a result of convergent evolution, may result in the attenuation of the secretion or in the fixation of new binding partners. This hypothesis is supported by some of our preliminary experiments; we have found that a classic peptide signal hooked to LET-756 prevents its nuclear addressing. In a similar way, a nonclassic transmembrane domain may allow nuclear translocation of an FGF receptor (Myers et al., 2003). This may apply to other families of bifunctional proteins and become a general model of molecular adaptation. For example, we have found the same motif as in FGF9 family members in dickkopf proteins.

Finally, if our current knowledge of FGF functions in C. elegans is almost exclusively based on genetic studies, and overall still relatively limited, it could rapidly benefit from other approaches, especially if combined (Tewari et al., 2004). Thus, the C. elegans model could contribute to this wonderful field of FGF, which is at the crossroads of evolution, development, signaling, trafficking, and physiology.

Acknowledgements

We thank C. Mawas, J. Reboul, and M. J. Stern for critical reading of the manuscript. Work on this topic in our laboratory is supported currently by INSERM and Institut Paoli-Calmettes.

Ancillary