The prototypical Fgf genes contain three coding exons (Figure 1), with exon 1 containing the initiation methionine, but several Fgf genes (for example, Fgf2 and Fgf3) have additional 5' transcribed sequence that initiates from upstream CUG codons [1,2]. The size of the coding portion of Fgf genes ranges from under 5 kb (in Fgf3 and Fgf4) to over 100 kb (in Fgf12). In several Fgf subfamilies, exon 1 is subdivided into between two and four alternatively spliced sub-exons (denoted 1A-1D in the case of Fgf8). In these Fgf genes, a single initiation codon (ATG) in exon 1A is used. This gene organization is conserved in humans, mouse and zebrafish, but its functional consequences are poorly understood. Other subfamilies of Fgfs (such as Fgf11-14) have alternative amino termini, which result from the use of alternative 5' exons. It is not known whether a common 5' untranslated exon splices to these exons or whether alternative promoter and regulatory sequences are used.

Most Fgf genes are found scattered throughout the genome. In human, 22 FGF genes have been identified and the chromosomal locations of all except FGF16 are known (Table 1) [3,4,5,6,7]. Several human FGF genes are clustered within the genome. FGF3, FGF4 and FGF19 are located on chromosome 11q13 and are separated by only 40 and 10 kb, respectively; FGF6 and FGF23 are located within 55 kb on chromosome 12p13; and FGF17 and FGF20 map to chromosome 8p21-p22. These gene locations indicate that the FGF gene family was generated both by gene and chromosomal duplication and translocation during evolution. Interestingly, a transcriptionally active portion of human FGF7, located on chromosome 15q13-q22, has been amplified to about 16 copies, which are dispersed throughout the human genome [8].

In the mouse, there are at least 22 Fgf genes [3,9], and the locations of 16 have been identified (Table 1). Many of the mouse Fgf genes are scattered throughout the genome, but as in the human, Fgf3, Fgf4 and Fgf19 are closely linked (within 80 kb on chromosome 7F) and Fgf6 and Fgf23 are closely linked on chromosome 6F3-G1.

Fgfs have been identified in both invertebrates and vertebrates [3]. Interestingly, an Fgf-like gene is also encoded in the nuclear polyhedrosis virus genome [10]. Fgf-like sequences have not been found in unicellular organisms such as Escherichia coli and Saccharomyces cerevisiae. Although the Drosophila and Caenorhabditis elegans genomes have been sequenced, only one Fgf gene (branchless) has been identified in Drosophila [11] and two (egl-17 and let-756) have been identified in C. elegans [12,13], in contrast to the large number of Fgf genes identified in vertebrates. The evolutionary relationship between invertebrate and vertebrate Fgfs is shown in Figure 2a.

The Fgf gene expansion has been hypothesized to be coincident with a phase of global gene duplications that took place during the period leading to the emergence of vertebrates [14]. Across species, most orthologous FGF proteins are highly conserved and share greater than 90% amino-acid sequence identity (except human FGF15 and mouse Fgf19; see below). To date, four Fgfs (Fgf3, 8, 17 and 18) have been identified in zebrafish, seven (Fgf3, Fgf(i), Fgf(ii), Fgf8, 9 and 20) in Xenopus (Fgf(i) and Fgf(ii) are most closely related to Fgf4 and Fgf6 [15]) and seven (Fgf2, 4, 8, 12, 14, 18 and 19) in chicken [3].

The apparent evolutionary relationships of the 22 known human FGFs are shown in Figure 2b. Vertebrate FGFs can be classified into several subgroups or subfamilies. Members of a subgroup of FGFs share increased sequence similarity and biochemical and developmental properties. For example, members of the FGF8 subfamily (FGF8, FGF17, and FGF18) have 70-80% amino acid sequence identity, similar receptor-binding properties and some overlapping sites of expression (for example, the midbrain-hindbrain junction) [16,17]. Members of FGF subgroups are not closely linked in the genome, however, indicating that the subfamilies were generated by gene-translocation or by genome-duplication events, not by local duplication events.

Human FGF15 and mouse Fgf19 have not been identified. Human FGF19 is evolutionarily most closely related to mouse Fgf15 (51% amino acid identity; Figure 2b) [18] and both the human FGF19 and mouse Fgf15 genes are closely linked to the human and mouse Fgf3 and Fgf4 genes on orthologous regions of human chromosome 11q13 and mouse chromosome 7F (N.I., unpublished observations). These findings indicate that human FGF19 may be the human ortholog of mouse Fgf15. Because all other Fgf orthologs share greater than 90% amino acid identity, it remains possible that the true orthologs of these genes have not been identified, have been lost or have diverged during vertebrate evolution.

The expression patterns of FGFs (see above) suggest that they have important roles in development. FGFs often signal directionally and reciprocally across epithelial-mesenchymal boundaries [36]. The integrity of these signaling pathways requires extremely tight regulation of FGF activity and receptor specificity. For example, in vertebrate limb development, mesenchymally expressed Fgf10 in the lateral-plate mesoderm induces the formation of the overlying apical ectodermal ridge; the ridge subsequently expresses Fgf8, which signals back to the underlying mesoderm [37]. This directional signaling initiates feedback loops and, along with other signaling molecules, regulates the outgrowth and patterning of the limb. Importantly, the differential expression of the alternative splice forms of the receptors in the apical ectodermal ridge and underlying mesoderm is such as to limit or prevent autocrine signaling within a given compartment.

Studies of the biochemical activities of FGFs have focused on the specificity of interactions between FGFs and FGFRs, on factors that affect the stability of FGFs and on the composition and mechanism of the active FGF-FGFR signaling complex.

Many members of the Fgf family have been disrupted by homologous recombination in mice. The phenotypes range from very early embryonic lethality to subtle phenotypes in adult mice. The major phenotypes observed in Fgf knockout mice are shown in Table 2. Because FGFs within a subfamily have similar receptor-binding properties and overlapping patterns of expression, functional redundancy is likely to occur. This has been demonstrated for Fgf17 and Fgf8, which cooperate to regulate neuroepithelial proliferation in the midbrain-hindbrain junction [17]. In the case of Fgf knockouts resulting in early lethality, other functions later in development will need to be addressed by constructing conditional alleles that can be targeted at specific times and places in development. For example, Fgf8^-/- mice die by embryonic day 9.5 [67]. A conditional allele for Fgf8 targeted to the apical ectodermal ridge has been used to demonstrate an essential role for Fgf8 in early limb development [68,69].

Several mutations in Fgf genes have been identified in C. elegans, Drosophila, zebrafish, mouse and human. The C. elegans gene egl-17 is required for sex myoblast migration [12], and a null allele of let-756 causes developmental arrest of the early larva [13]. The Drosophila branchless gene is required for tracheal branching and cell migration [11]. In zebrafish, acerebellar (ace) embryos lack the cerebellum and the midbrain-hindbrain boundary organizer. The ace gene encodes the zebrafish homolog of Fgf8 [70]. Interestingly, zebrafish aussicht mutant embryos, which overexpress Fgf8, also have defects in development of the central nervous system [71].

In the mouse, the angora mutation, which affects hair growth, was found to be allelic with Fgf5 [72]. A mouse mutant with a Crouzon-syndrome-like craniofacial dysmorphology phenotype was found to result from an insertional mutation in the Fgf3/Fgf4 locus [73]. Recently, positional cloning of the autosomal dominant hypophosphataemic rickets gene identified missense mutations in human FGF23 [74]. A recent paper demonstrates that this disease is caused by a gain-of-function mutation [75]. The chromosomal location (Xq26) and tissue-specific expression pattern of Fgf13 (also called Fhf2) suggests that it may be a candidate gene for Borjeson-Forssman-Lehmann syndrome, an X-linked mental retardation syndrome [76].

FGFs have been intensely studied for nearly 30 years. Most of the early work focused on the mechanisms that regulate stability, secretion, export and interactions with heparin and on the mechanisms and consequences of signal transduction in various types of cells. More recent work has focused on the mechanisms regulating receptor specificity and receptor activation, the structure of the FGF-FGFR-HS complex, and the identification of new members of the FGF family. Functional studies have begun to address the role of FGFs in cell biology, development and physiology. Initial studies focused on the regulation of cell proliferation, migration and differentiation; more recent work has addressed the negative effect of FGFs and FGFRs on proliferation of some cell types, which was surprising as FGFs were thought to promote proliferation. In vitro studies have now been complemented by gene targeting in mice. The knockout approach has been fairly successful in identifying primary phenotypes but will be challenged by the need to address redundancy amongst the 22 FGFs and to study their developmental and physiological functions after the point of lethality of the null allele.

A major unresolved question concerns the mechanism(s) regulating FGF activity in vivo in the presence of cell-surface and extracellular-matrix HSPG. Current hypotheses predict that tissue-specific heparan fragments of defined sequence (and particularly of defined sulfation pattern) will differentially regulate FGFs by controlling their diffusion in the extracellular matrix and their ability to activate specific receptors [77]. These issues will be resolved by determining the sequence of tissue-specific HS and by demonstrating whether specific HS sequences can modulate the binding specificity of FGFs beyond that determined by the specific FGFR and its alternative splice form in the presence of heparin.

A second area of research will aim to elucidate the developmental roles of all the FGFs, first alone and then in various combinations. This will include determining whether a single FGF with a defined developmental function interacts with one or multiple FGFRs. A third major frontier will be to elucidate the physiological roles of FGFs that are expressed in adult tissues. This will again involve testing combinations of FGFs in cases in which knockouts are viable and designing conditional alleles in cases of embryonic lethality. Major areas being considered include neuronal and cardiovascular physiology, neuronal regeneration and homeostasis and tissue repair.

The last major frontier will be to elucidate the primary roles of FGFs in genetic diseases and cancer. Several FGFs were initially cloned from human and animal tumors. Future work will be required to determine whether FGF activation is itself an etiological agent in primary human tumors or whether it is a progression factor in the pathogenesis of cancer. As functional roles for FGFs are elucidated in embryonic development, it is expected that various human birth defects and genetic diseases will be attributed to mutations in Fgf genes. These studies will probably lead to the development of pharmacogenetic agents to treat these diseases. Because a large number of skeletal diseases are caused by mutations in Fgfr genes, it is anticipated that mutations in some Fgf genes will also be involved in skeletal pathology.