Functional phylogeny relates LET-756 to fibroblast growth factor 9.

Fibroblast growth factors (FGFs) are secreted regulatory proteins involved in various developmental processes. In vertebrates, the FGF superfamily comprises 22 members. In non-vertebrates, six FGF genes have been identified in Ciona intestinalis, three in Drosophila melanogaster, and two (let-756 and egl-17) in Caenorhabditis elegans. The core of LET-756 shares a 30-50% sequence identity with the various members of the superfamily. The relationships between vertebrate and non-vertebrate FGFs are not clear. We made chimeric FGFs by replacing the core region of LET-756 by the cores of various mammalian, fly, and worm FGFs. LET-756 deleted in its core region was no longer able to rescue the lethal phenotype of a let-756 null mutant, and only chimeras containing the cores of FGFs 9, 16, and 20 showed rescue capacity. This core contains an internal motif of six amino acid residues (EFISIA) whose deletion or mutation abolished both the rescue activity and FGF secretion in the supernatant of transfected COS-1 cells. Chimera containing the core of C. intestinalis FGF9/16/20, a potential ortholog of FGF9 lacking the complete EFISIA motif, was not able to rescue the lethal phenotype or be secreted. However, the introduction of the EFISIA motif restored both activities. The data show that the EFISIA motif in the core of LET-756 is essential for its biological activity and that FGFs 9, 16, and 20, which contain that motif, are functionally close to LET-756 and may be evolutionary related. This non-classical mode of secretion using an internal motif is conserved throughout evolution.

Fibroblast growth factors (FGF) 1 are secreted regulatory proteins. In vertebrates, the FGF superfamily comprises 22 members involved in various developmental processes including the formation of mesoderm during gastrulation, patterning during early post-implantation, and development of various tissues such as ear, limb, hair, and nervous and skeletal systems (1). This role has been established through a series of studies on animal models and identification of human heredi-tary skeletal disorders that are associated with mutations of the tyrosine kinase FGF receptors (2,3). In non-vertebrates, only a few FGF genes have been identified and characterized. To speak only of fully sequenced genomes, three Fgf genes are known in Drosophila melanogaster (4,5), two in Caenorhabditis elegans (let-756 and egl-17) (6,7), and six in the small marine ascidian Ciona intestinalis (8,9). 2 Phylogenetic analysis allows classification of FGFs in several families that separated early during the evolution of bilaterians (8, 10 -12). In the human genome, the FGF families are included into distinct sets of paralogous chromosomal regions (13). This classification is associated with the structural and functional properties of FGFs. For example, the FGF homologous factors (FHF), which in contrast to all other FGFs, exert their function by ways independent of the FGF receptors (14,15) group in a separate family. However, this analysis is not conclusive for the non-vertebrate FGFs and does not allow us to establish a direct phylogenic relationship between human, fly, and worm FGFs. To understand the importance of FGF expansion and the phylogenetic and/or functional relationships that may exist between vertebrate and non-vertebrate FGFs, we studied C. elegans LET-756/FGF as a model. The purpose was to assess whether this worm FGF could be related to any particular FGF family despite the limitation of the phylogenetic analysis.
The boundaries of core regions were determined by SMART (16) (smart.embl-heidelberg.de/) and then aligned using ClustalX (17) and by the human eye (see supplementary data for the alignments of the various FGFs). Phylogenetic trees were constructed using both maximum parsimony and bootstrapped neighbor-joining techniques. Branch support for the best-fitting tree was assessed using 1000 bootstrap replicates. Alignments, bootstrap analysis, and neighborhood-joining trees were carried out at default parameters.
Eukaryotic Cells and C. elegans Cells Expression Vectors-The strategy for cloning eukaryotic and C. elegans expression vectors is depicted in Fig. 1A. Additional details are given as supplementary material, and the sequences of the primers used to amplify the various FGF fragments are available upon request.
The coding region of let-756, FGF-E, and FGF9 was amplified with F1-R1 primers (containing BamHI and BglII restriction sites) and inserted in the BglII/BamHI sites of pEGFP-N2-eukaryotic cell expression vector (underlined restriction sites shown in Fig. 1 are introduced by mutagenesis). The cores of various FGFs are amplified with F2-R2 primers (using as template the FGF coding region inserted in a cloning vector) and introduced in the plet containing the BssHII and SpeI restriction sites. The cDNA fragments containing the N-terminal core or core C-terminal coding regions are amplified with F3-R2 and F2-R3 primers and inserted by using NheI/SpeI and BssHII/XmaI restriction sites, respectively. The mutagenized core regions made in the eukaryotic cell expression vector are transferred in the C. elegans expression vector by using the PvuI/SacI restriction sites.
The chimeric constructs allowing the expression in mammalian cells are designated P CMV ::let-756[FGFn]let-756::gfp in which in the bracketed region is the core-coding region and n stands for different FGFs as follows: mammalian FGF2/3/5-9/12/16/20/21; C. intestinalis Fgf8/17/18, Fgf9/16/ 20, and Fgf11/12/13/14 (referred to as Fgf-D, Fgf-E, and Fgf-F, respectively); C. elegans let-756 and egl-17; and D. melanogaster bnl. The C. elegans expression vectors are designated accordingly with the exception that P let-756 for the let-756 promoter replaces P CMV . Other denominations are as follows. The constructs containing the N-terminal and core-coding regions of FGF9 fused to the C-terminal coding region of let-756 is P CMV ::FGF9[FGF9]let-756::gfp, and the constructs containing N-terminal coding region of let-756 fused to the core and C-terminal coding regions of FGF9 is P CMV ::let-756[FGF9]FGF9::gfp. Mutations or deletions in the FGF molecule are indicated as exposants. As an example, the constructs deleted for the EFISIA-coding region are referred as P CMV ::let-756 ⌬EFVSVA ::gfp or P CMV ::let-756[FGF9 ⌬EFVSVA ]let-756::gfp.
Nematode Culture and Transformation Rescue Experiments-C. elegans Bristol (N2) strain nematodes were cultured using standard techniques (18). let-756-rescuing activity was assayed by injecting tester DNA at 50 ng/l into gravid s2887 strain hermaphrodites, establishing transformed lines and scoring the Dpy progeny for a viable phenotype. The s2887 allele used in these experiments was generated on a dpy-17(e164) unc-32(e189) chromosome after UV irradiation (19) in which sDp3, a portion of chromosome III, balances the lethal phenotype of the s2887 strain. The tester DNA was co-injected with the pPD95.86 vector (a gift of A. Fire) in which GFP is driven by the myo-3 promoter, such that Unc F1 positive for GFP was isolated and the rescue activity was tested for the presence in the progeny of live Dpy animals expressing GFP. Conversely, the absence of rescue activity was assessed by the isolation of Dpy GFP-positive animals (containing the myo-3 marker associated with the tester FGF), which fail to develop to the fertile stage and thus are unable to establish a strain. Homozygous (free of sDp3) Dpy GFP-positive animals were further genotyped for the presence of the inversion characteristic of the s2887 strain by PCR on a single worm. The following primer pairs were used: (i) one in the first exon and one in intron 3, which do not amplify if animals are free of the sDp3 or do not result from sDp3 recombination and (ii) one in the first exon and one in the R13A5 cosmid, which detect the inversion occurring in let-756. The sequences of the primers can be obtained upon request. We also determined that, in the progeny of the rescued Dpy, some animals die because of the absence of the array. Conversely, the absence of rescue activity was assessed by the isolation of Dpy GFP-positive animals (containing the myo-3 marker associated with the tester FGF), which fail to develop to the fertile stage and thus are unable to establish a strain. The tester DNA was also co-injected with the pRF4 plasmid in wild type worms, and transformants were selected for their roller phenotype.
Microscopy-F1 GFP-expressing nematodes were selected under a stereomicroscope (Leica MZ6) equipped with a GFP fluorescence module. When necessary, individual nematodes were picked from plates onto 2% agar pads containing 10 l of 1 mM levamisole as an anesthetic and observed with a Leica TCS NT confocal microscope. Transfected COS-1 were grown on coverslips, fixed with 3.7% paraformaldehyde, and mounted in Dako. In some experiments, fixed cells were incubated with anti-giantin antibodies (a gift of H. F. Hauri, Biozentrum) followed by Texas Red-coupled secondary antibodies.
Western Blot-To assay secreted proteins, a 48-h conditioned medium from COS-1-transfected cells were collected, cleared by centrifugation, and immunoprecipitated with polyclonal rabbit anti-GFP antibodies (Abcam, Cambridge, United Kingdom) and protein A-Sepharose beads (Amersham Biosciences). To study cell-associated proteins, lysates from transfected cells were, depending on the experiment, either immunoprecipitated or boiled for 5 min in 2% SDS. Total cell lysates and anti-GFP immunoprecipitated samples were analyzed by SDS-PAGE and, after blotting, revealed by anti-mouse anti-GFP antibody (Roche Applied Science). The signal was detected using HRP-conjugated anti-mouse IgG and ECL kit (Pierce, Rockford, IL). Brefeldin A was used at a 2.5 g/ml concentration, and tunicamycin was used at 5 g/ml.

Identification of FGF Gene
Families-FGF sequences share identity and similarity in a central core region of the molecule (10). We constructed a phylogenetic tree of the FGF core sequences (Fig. 2). The tree included core sequences from human, C. intestinalis, D. melanogaster, and C. elegans FGFs. We added the FGF15 mouse sequence, because it is not present in the human genome, whereas reciprocally, FGF19 is not found in the mouse. The phylogenetic tree was constructed using the distance matrix (Blosum 30 matrix) and neighbor-joining algorithms implemented in ClustalW. A total of 1000 bootstrapped replicates were run. The tree showed that the FGF superfamily could be divided into seven evolutionary divergent families: During evolution, the formation of the FGF families may have occurred before and/or after the separation of protostomians from deuterostomians. If they formed before this separation, a non-vertebrate FGF should belong to a given family and be the respective ortholog of its mammalian members. If they formed after this separation, non-vertebrate FGFs should be orthologous to all of the deuterostomian FGFs. Documenting this difference may help to understand the FGF features and functions and the reason for FGF gene expansion.
To address this question, we studied the LET-756 molecule. We previously characterized the let-756 locus (7). 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 Arg-318 position leading to a truncated protein that allows a few worms to develop up to the adult stage (see Fig. 1B). LET-756 shares features with FGFs of different families, such as expression in cytoplasm and nucleus of muscles and neurons (Fig. 3), because of the presence of numerous nuclear localization signals (NLS) and the absence of a classical leader but the presence of a central hydrophobic sequence (see Figs. 1B and 4A). Interestingly, it is not produced by hypodermal cells where it interacts with its receptor EGL-15 to regulate fluid balance (20) and axonal outgrowth (21), indicating that it is secreted in vivo despite the absence of a classical leader sequence. Thus, based on phylogeny as well as on other features, it was impossible to associate LET-756 with any family.
Replacement of the Core of LET-756 by the Core of FGF9, FGF16, or FGF20 Confers Rescuing Activity-The let-756 gene has rescuing activity when expressed under the control of its own promoter in the strain carrying the s2887 null allele (7). This activity is conserved when LET-756 is tagged with GFP in the C-terminal region. In contrast, a construct deleted of the core region of LET-756 could not rescue the let-756(s2887) mutants (Table I). To determine whether LET-756 is functionally closer to any of the FGF families, we replaced the LET-756 core sequence with that of other FGFs including one from each human family (FGF2, FGF3, FGF6, FGF7, FGF8, FGF9, FGF12, and FGF21) and with that of Drosophila BNL and C. elegans EGL-17. The sequence alignments are shown in Fig.  4A. Only the core of FGF9 replaced efficiently the LET-756 core for rescuing ability (Table I).
Family E Groups FGF9, FGF16, and FGF20 -Similar to the FGF9 core, the cores of FGF16 and FGF20, introduced in the LET-756 background, rescued the let-756 phenotype (s2887). Because FGF-E from C. intestinalis is supposed to be orthologous to FGF9 (8), we also tested the P let-756 ::let-756[Fgf-E]let-756::gfp chimera. This chimera was unable to rescue the lethal let-756 phenotype as did two other chimeras containing cores of Ciona FGF-D and FGF-F. In control experiments, wild-type worms were transformed with FGF9, FGF12, and FGF16 chimeras (Fig.  3, b-d). GFP expression was detected in neurons and muscle cells, indicating that the lack of rescuing activity was not linked to a defect in protein expression.
Residues in the Core Region Are Necessary but Not Sufficient for Rescuing Activity-These results suggest that LET-756 could resemble FGF family E members FGF9, FGF16, and FGF20. We looked for what LET-756 and this family have in common and do not share with the other families. An examination of the alignment of the various FGFs pointed to the presence of a highly conserved motif (shaded in Fig. 4A) located in the core region. It was only present in LET-756 (EFVSVA: amino acids 116 -120), FGF9, (EFISIA), FGF16 (EFISLA), and FGF20 (EFISVA). This sequence is slightly different in Ciona FGF-E, because only four of six conserved amino acids are present (EFISTG).
The EFISIA/EFVSVA sequence is a hydrophobic stretch within the LET-756 and FGF9 molecules (Fig. 4B). To test for the importance of this motif on the function of LET-756, we made a construct where let-756 under the control of its own promoter was deleted in the sequence encoding these six residues (P let-756 ::let-756 ⌬EFVSVA ::gfp construct) and tested the rescuing activity in let-756(s2887) animals. The construct deleted of the sequence was devoid of rescuing activity (Table I). No other mutation in the core impaired the rescue of the lethal phenotype. As an example, LET-756 deleted from the NLS4 -6 region was still able to rescue. When the region coding for the EFISIA motif was deleted from the core region of the LET-756/ FGF9 chimera (P let-756 ::let-756[FGF9 ⌬EFISIA ]let-756::gfp construct), the rescue was abolished.
To confirm the importance of this region in the function of LET-756, we exchanged the corresponding regions in LET-756 and Ciona FGF-E. When the EFISIA stretch replaced EFISTG in the FGF-E chimera (P let-756 ::let-756[Fgf-E EFISIA ]let-756::gfp construct), the rescuing activity was restored. Conversely, the replacement of EFVSVA by EFISTG in LET-756 (P let-756 ::let-756 EFIST ::gfp) abolished the rescuing activity. However, the role of the EFISIA motif worked only in the context of the FGF-E family. Appending an EFISIA motif to the LET-756/ FGF2 chimera (P let-756 ::let-756[FGF2 EFISIA ]let-756::gfp construct) did not confer rescuing activity (Table I). Thus, an EFISIA motif was necessary for the FGFs of the E family but was not sufficient for the other FGFs.
The EFISIA Motif Is Essential for Secretion-To understand how this motif influences the rescue activity, we tested its role in secretion. COS-1 cells were transfected with constructs including or not including the EFISIA motif. The P CMV ::let-756::gfp construct expressed GFP in the nucleus as speckles. A faint staining of the Golgi apparatus was also observed (Fig. 5a), which colocalized with the Golgi marker giantin (Fig. 5d). The fusion protein corresponding to the partial loss-of-function allele (R318Stop) was found in the nucleoli as well as in the Golgi (Fig.  5, c and f). The protein deleted of the EFISIA motif exhibited the same speckle staining as the wild-type molecule but no staining of the Golgi apparatus (Fig. 5, b and e). FGF9 was mostly found in the Golgi apparatus (Fig. 5g), in agreement with previous findings (22). By contrast, the various mammalian FGF chimeras, including the FGF9 chimera, were localized at speckles in the nucleus (Fig. 5, h, k, and l). Replacing the C-terminal region of LET-756 by that of FGF9 abolished this localization (Fig. 5i), whereas replacing the N-terminal region of LET-756 by that of FGF9 was without effect on the chimera localization (Fig. 5j), pointing out the crucial role of the C-terminal region of LET-756 for nuclear sublocalization.
Immunoblots were done with supernatants and cell lysates from transfected cells. LET-756 was efficiently secreted (Fig. 6A). This secretion was dependent on the Golgi-associated system, because it was inhibited by the addition of brefeldin A (Fig. 6A). The majority of the released protein was glycosylated as shown by the addition of tunicamycin. As N-linked glycosylation only occurs in the endoplasmic reticulum, this is evidence for export of LET-756 through the secretory pathway. Several other GFP fusion molecules of smaller molecular mass were also detected, which probably resulted from endogenous proteolysis.
In contrast, the absence of the EFVSVA motif impaired secretion (LET-756 ⌬EFVSVA) . This was not the case for the absence of other regions (LET-756 ⌬NLS4 -6 , LET R318 ) (Fig. 6B). The core of FGF-E with its unconserved EFISTG motif was not secreted. In contrast, changing EFISTG in EFISIA restored secretion (Fig. 6D) as it had restored rescuing activity (Table I). The protein encoded by the FGF2 chimera, appended or not with the EFISIA motif, which as seen above did not have rescuing activity (Table I), was not secreted (Fig. 6D, lanes 12  versus 11). Thus, the EFISIA motif is necessary but not sufficient by itself for either secretion or rescuing capacity.
Because the hydrophobicity of the EFISIA motif has been involved in its capability to allow secretion of FGF9 and FGF16 (22)(23)(24), we introduced various point mutations in the LET-756 EFVSVA motif to change its hydrophobicity (Fig. 4B). Neither change of those amino acids in Lys or Gly, which makes molecules less hydrophobic, nor in Ala, which increases hydrophobicity, allowed the resulting molecules to be secreted (Fig. 6C, lane 6) indicating that hydrophobicity alone is not a key determinant of LET-756 activity. The conformation of the molecule, which is possibly induced by the primary sequence of this region, might be important for secretion.
Role of Sequences Outside the Core-Finally, we tested whether FGF9 per se could confer rescue. We made a construct (P let-756 ::FGF9::gfp) where the expression of FGF9 was directed by the let-756 promoter and tested for the rescue of the let-756(s2887) animals. FGF9 by itself was unable to rescue the lethal phenotype. This indicates that at least another region of the LET-756 molecule is necessary to ensure the full biological function of this C. elegans FGF, in agreement with the truncation of the LET-756 C-terminal region (LET R318 ) that induces a partial loss-of-function. The construct (P let-756 ::FGF9[FGF9]let-756::gfp), in which the 3Ј region of let-756 is present, rescued the null mutant in contrast to the P let-756 ::let-756[FGF9]FGF9::gfp construct in which the 3Ј end is that of FGF9. Both constructs encoded proteins capable of being secreted when transfected into COS-1 cells (Fig. 6D). This shows again the importance of the EFISIA motif for secretion but not for full biological activity. As expected, the secretion of the FGF9 chimera deleted from the EFISIA motif was impaired (Fig. 6C, lane 8) and unable to exert rescue ability when injected in the null mutant. Thus, LET-756 has both similarity to FGF-E family members and specific features associated with its C-terminal region.

DISCUSSION
From a functional point of view, it is hard to draw parallels among FGF activities in different species. For example, LET-756 shares nuclear localization with FGF1, FGF2, FGF3, and FHFs (for review see Ref. 24) and muscle expression with FGF5, FGF6, and FGF9 (24 -26). From an evolutionary point of view, two striking features are associated with the FGF superfamily. First, there has been an important expansion of the number of FGF genes in vertebrates. Second, there is no clear orthology relationship between mammalian and non-vertebrate FGFs. This expansion could be explained by the series of duplications that have accompanied vertebrate evolution, but the absence of direct orthology makes difficult to reconstitute the different steps. This prompted us to study the interchangeability of FGFs from various species with LET-756 as a "functional phylogenic" approach. We have shown that LET-756 is related to FGF family E. This has two broad implications: 1) biology and 2) evolution.
Ϫ ϩ Biology-Among the chimeric constructs tested, only those containing the core of FGF9, FGF16, or FGF20, which belong to the same FGF family (FGF-E family), were capable of a functional substitution. Most FGFs, such as FGF3, FGF6, FGF8, FGF10, FGF17, FGF18, and FGF22, contain a signal sequence for cell export. However, FGF3 is not secreted and is retained in the Golgi complex. FGF1 and FGF2, which do not contain such sequence, are released from cells by a mechanism inde-pendent of the ER/Golgi secretory pathway. FGF9, FGF16, and FGF20 do not have a classical signal sequence. They are efficiently secreted in an ER/Golgi-dependent pathway because of the presence of a high hydrophobic N-terminal region and a six amino acid-long region located within the core (22)(23)(24). We have shown here that LET-756, which like FGF9, FGF16, and FGF20 does not have a typical N-terminal signal sequence, is efficiently secreted by a Golgi-associated mechanism depend- FIG. 4. Structural features of FGFs. A, alignment of core region sequences from FGFs used for chimeric constructs (hN and hC for N-and C-terminal ␣-helixes, respectively, and b1-b12 for the 12 ␤-sheets in the core regions. The core region used for chimeric constructs is delineated by stars, the highly conserved amino acids are in red, and the conservative substitutions and the conserved amino acids of the EFISIA region are in green and blue, respectively. B, hydrophobicity profiles of the core region of wild-type LET-756, human FGF9, FGF2, and FGF-E. The hydrophobicity profile of wild-type LET-756 protein is in blue, that of LET-756 swapped with the corresponding region of FGF-E is in yellow, and those of mutated EFVSVA in EAVAVA (the presence of alanine residues makes the molecule even more hydrophobic), EFESEA, or KFKSKA (which make the molecule less hydrophobic) are in magenta, green, and red, respectively. ent on the presence of the same six core residues as for the FGF-E family that are also essential for rescue activity. Only core sequences from members of the FGF-E family that have this sequence are able to rescue the worm null mutant. However, inserting this motif in the cores of FGFs from other families is not sufficient to confer rescuing ability. C. intestinalis FGF9/16/20 (designated FGF-E in this paper) is supposed to be orthologous to mammalian FGF9, FGF16, and FGF20 (8). It contains an incomplete EFISIA motif that does not allow secretion and rescue in the LET-756 background. Replacement of EFISTG by EFISIA is sufficient to confer both rescuing activity and secretion to the protein. Here we have shown that the atypical mechanism of secretion of LET-756 and FGF-E members that utilizes that internal hydrophobic motif is conserved in different species. Such atypical signal peptides have been found in some other proteins in vertebrates but remain rare (for review see Refs. [22][23][24]. It is the first time that such a mechanism is demonstrated in the nematode. In present-day Ciona FGF-E EFISTG, it is not efficient for export, which uses a classical peptide signal. For FGF9, in addition to the EFISIA motif, the N-terminal portion of the molecule is also important for secretion, whereas it is not the case for LET-756. It is possible that the routing of the proteins associated with this mode of export is different from those using a classical peptide sequence. In this case, interacting proteins could be specific of this pathway but they remain unidentified. Work is in progress to determine whether secretion of LET-756 and subsequent interaction with its membrane FGF receptor, EGL-15 5B (28), is the only activating process or whether the nuclear pool of LET-756 exerts any other biological activity.
There may have been a limited conservation of function within the FGF superfamily during evolution. The three mammalian FGFs of the E family and Ciona FGF-E share a common function in the nervous system (9,27,29,30). FGF9 directs embryogenesis of several organs, including the reproductive and excretory system (31). LET-756 is expressed in the CAN cells, which regulates excretion in the worm. It acts upon EGL-15 in the hypodermis to regulate the balance of fluids (20). Thus, a control of the excretory system might be a primordial function of this family. Beside this function, a role for LET-756 in axonal outgrowth has also been reported recently (21). LET-756 expressed in muscle cells allows outgrowth of the axons on the surface made by EGL-15B expressing hypodermal cells. In conclusion, the physiology of LET-756 and FGF-E members may be compared directly within this family, both in terms of trafficking and of cellular targets.
Evolution-The structural and functional features shared by LET-756 and FGFs from family E may be due to a common ancestry or to the convergent acquisition of a secretory capacity. As previously suggested from sequence alignments, (6) the other C. elegans FGF, EGL-17, might belong to FGF family D, although we did not find again the real phylogenic relationship because the bootstrap value was not significant. The common . d-f, giantin immunostaining (red). GFP is found in the Golgi apparatus of wild-type (a and d) and R318Stop (c and f) but not if the protein is deleted in the EFISIA motif (b and e). As opposed to LET-756 (a), FGF9 (g) is found exclusively associated to the Golgi apparatus. Replacing the core of LET-756 by the core of any FGF (h, k, and l) restitutes a wild-type phenotype to the chimera. The chimera containing the C-terminal region of FGF9 is uniformly expressed in the nucleus (i), whereas the chimera containing the N-terminal region of FGF9 is expressed as speckles (j). ancestry of each of the two C. elegans FGF with a given family would signify that a first expansion of the FGF genes has taken place early before the protostomian/deuterostomian split, but gene losses have been important in the protostomian branch, and that only two FGF families, E and D, are represented in the present-day nematode. However, LET-756, in addition to a core region that is functionally close to the FGFs of family E, has evolved to acquire a unique C-terminal region and new functions that come with it.