Expression of the fibroblast growth factor-5 gene in the mouse embryo

Fibroblast growth factors (FGFs) are structurally related mitogenic proteins encoded by at least seven distinct genes in mammals (Goldfarb, 1990). In addition to their growth-promoting activities towards a broad spectrum of cell types, FGFs are suspected of playing various roles in the control of cellular differentiation (reviewed in (Burgess and Maciag, 1989)). In vitro, the prototypic FGFs (acidic and basic FGF) can promote neuronal survival and neurite outgrowth (Hatten et al. 1988; Lipton et al. 1988; Morrison et al. 1986; Unsicker et al. 1987; Walicke et al. 1986; Walicke and Baird, 1988), commit sympathoadrenal progenitor cells to the sympathetic neural lineage (Birren and Anderson, 1990), regulate the development of skeletal muscle precursor cells (Lathrop et al. 1985; Seed and Hauschka, 1988), and induce mesodermal differentiation in amphibian blastula ectoderm (Kimelman and Kirschner, 1987; Slack et al. 1987). FGF in vivo implants can promote angiogensis (Folkman and Klagsbrun, 1987), promote the formation of new nerve fiber tracts (Thompson et al. 1989) and prevent retrograde degeneration of neurons in the central nervous system (Anderson et al. 1988). Consistent with their potential roles in development, several FGF genes are expressed at various stages of embryogenesis (reviewed in (Goldfarb, 1990)).

FGF-5 was first identified as the product of a human proto-oncogene detected by DNA transfection assays (Zhan et al. 1988). The growth factor is a secreted glycoprotein (Bates et al. 1991) that is mitogenic towards fibroblasts in vitro (Zhan et al. 1988). As a step towards understanding the native functions of this growth factor in vivo, we have studied the profile of mouse Fgf-5 gene expression in adult tissues and during embryogenesis. We have previously reported the presence of Fgf-5 RNA at low levels in the adult central nervous system and its localization within certain neurons (Haub et al. 1990). In this paper, we present an analysis of Fgf-5 gene expression in the mouse embryo, as detected by in situ hybridization.

Enzymes for nucleic acid biochemistry were purchased from New England Biolabs, Boehringer Mannheim and Stratagene. Radiolabelled nucleotides were purchase from DuPont/New England Nuclear.

A 2.2 kbp cDNA clone was isolated by screening a mouse embryonic cDNA library (Clonetech) with a murine Fgf-5 third exon fragment (Haub et al. 1990). The cDNA clone extends from approximately the 50th codon in the Fgf-5 open reading frame to near the 3’ end of the message. RNA was isolated from cells following guanidine isothiocyanate extraction (Chirgwin et al. 1979). CCE embryonic stem (ES) cells were provided by Dr E. Robertson and were maintained in the undifferentiated state by passage on STO fibroblast feeder layers, or were used to make embryoid bodies by standard procedure (Martin and Evans, 1975). egg cylinders were dissected from their decidua and teased free of ectopiacental cone, which is reddish in appearance. E. coli RNA was added as carrier during egg cylinder RNA purification. To quantitate egg cylinder RNA, one tenth of recovered material was analyzed by RNA ‘northern’ filter blotting (Thomas, 1980) in comparison to a 10–1000 nanagram range of ES RNA, using rat glyceraldehyde phosphate dehydrogenase (GAPDH) cDNA (Fort et al. 1985) as hybridization probe. DNA probes were ³²P-radiolabelled by random hexamer-primed DNA synthesis (Feinberg and Vogelstein, 1983).

Radiolabelled RNA riboprobes were synthesized in standard reactions (Melton et al. 1984) using SP6, T3, or T7 RNA polymerase, depending upon vector DNA template. Uridine(α³⁵S-thio)triphosphate and uridine[<r¹²P]triphosphate were incorporated to specific activities of 7 × 10^s and 2×10⁹ctsmin^-1 μg^-1, respectively. The PstI–SacI fragment of Fgf-5 DNA (Haub et al. 1990) (see Fig. 1) was used as template for antisense- and sense-strand riboprobes (antiPS and sensePS), and the PstI–PstI cDNA fragment was used to make antisense probe (antiPP) to corroborate hybridization data. A plasmid containing 120 base pairs of non-coding first exon sequence of the murine α-cardiac actin gene (Sassoon et al. 1988) and pINT2fg, containing part of the murine int-2 gene (Wilkinson et al. 1988), were also used as vectors for riboprobe synthesis.

Embryos (⁠⁠, and daily thereafter through ⁠) were dissected from pregnant MFI mice, fixed overnight at 4°C in phosphate-buffered saline plus 4 % paraformaldehyde, and embedded in paraffin. to embryos were left in their decidua. blastocysts were flushed from uteri, prefixed and loaded into dissected ampulae of pseudopregnant females before processing. Serial microtome sections (7–8 microns thick) were deparaffinized and hybridized essentially according to the procedure of Wilkinson (Wilkinson et al. 1987). The only modifications were in the content of the hybridization mix, which was prepared as previously described (Haub et al. 1990), and contained 3 nanograms probe in 0.2 ml per slide. After posthybridization RNAase treatment and washes (Wilkinson et al. 1987), slides were dipped in NTB2 emulsion (E. Kodak) diluted 1:1 with 2% glycerol, and autoradiographed (14–20 days for ³²P, 6 weeks for ³⁵S). Exposed slides were developed in D19 (Kodak), stained with hematoxylin and eosin, and mounted with Permount (Fisher Scientific). Bright- and darkfield images were photographed from a Nikon Microphot-FXA microscope. All embryonic stages were exhaustively analyzed; multiple embryos were hybridized with both sense and antisense Fgf-5 riboprobes, using a complete series of sections for each probe for embryos, and alternating groups of three sections for both probes with older, larger embryos. Anatomical assignments were made with the aid of several texts on embryonic anatomy (Gasser, 1975; Rugh, 1968; Theiler, 1989).

Camera-lucida drawings on transparencies were made from bright- and dark-field images of thirty hybridized sections spaced 96 microns apart across an embryo. The drawings were manually aligned, and then traced into a Sun View computer using CARP software (Biographies, Inc.). Superimposed objects were ‘built’ from the contours of whole embryo, dorsal aorta and arches, primary head and cardinal trunk veins, Fgf-5 signal in myotomes, and Fgf-5 signal in somatic mesoderm, and visualized in various combinations for direct photography from the computer’s screen.

In order to localize Fgf-5 expression during embryogenesis, deparaffinized sections of prefixed mouse embryos at various stages of development were hybridized in situ, using radiolabelled antisense and sense Fgf-5 RNA probes. Two different antisense probes, transcribed from nonoverlapping segments of Fgf-5 cDNA (Fig. 1), detected identical profiles of gene expression. Hybridization to deparaffinized sections gives signals 3–4 fold weaker than to post-fixed, fresh frozen sections (our unpublished observations); in fact, our previous in situ analysis of Fgf-5 expression in adult mouse brain required the use of fresh frozen sections to detect the very low abundance of Fgf-5 RNA in this tissue (Haub et al. 1990). However, morphology in sections of fresh frozen embryos is very poor, mandating our use of paraffin-embedded material in these studies. With this technique, we have detected seven distinct sites of Fgf-5 RNA expression in the mouse between embryonic days to post-conceptus¹.

Embryonic Fgf-5 expression first occurs soon after uterine implantation. As can be seen in Fig. 2, there is no detectable expression in preimplantation blastocysts (panels E,F). Fgf-5 RNA was detected in embryos at the next stage examined, (panels A,B), while no signal was seen with the negative control (sense-strand) probe (panels C,D). The failure to detect Fgf-5 expression in blastocysts (embryonic day ⁠), cannot be attributed to the small size of the blastocyst; expression was readily observed in tangential embryonic slices containing very few cells (data not shown). The postimplantation induction of expression is consistent with the observed rapid induction of Fgf-5 mRNA levels when embryonal carcinoma cells or totipotent embryonic stem (ES) cells differentiate as embryoid bodies in vitro ((Hebert et al. 1990), and see below).

Fgf-5 expression in embryos is restricted to the embryonic ectoderm and adjacent visceral endoderm (Fig. 2, panels G,H). By contrast, extraembryonic ectoderm its flanking visceral endoderm and other extraembryonic tissues are negative. This same expression profile is evident through the onset of gastrulation at (data not shown). While expression in the embryonic ectoderm of the egg cylinder persists at E7, newly formed mesoderm is negative (panels K,L). A cross-section through the E7 egg cylinder illustrates Fgf-5 RNA expression throughout the epiblast, without substantial lateral/medial or anterior/ posterior variation (panels I,J). The embryo shows weaker Fgf-5 RNA signals, and expression is undetectable in the E8 embryo.

The pattern of early embryonic expression for the Fgf-5 gene contrasts with that for another gene in the FGF family, int-2. int-2 is expressed in parietal extraembryonic endoderm and also in newly formed mesoderm leaving the primitive streak, commencing at (Wilkinson et al. 1988). By hybridizing adjacent embryo sections with Fgf-5 and int-2 antisense RNA probes, we have observed Fgf-5 expression in embryonic ectoderm adjacent to int-2 expression in mesoderm (data not shown).

Expression of Fgf-5 RNA in later embryogenesis Between embryonic days and ⁠, six additional sites of Fgf-5 expression are evident. Two of these sites are in derivatives of lateral mesoderm, two others are skeletal muscle precursors derived from paraxial mesoderm, and the other two sites are limb mesenchyme and a cranial ganglion (see below). On embryonic day ⁠, no sites of Fgf-5 expression are identifiable in situ. Since Fgf-5 RNA can still be detected by northern blot analysis at this developmental stage (Hebert et al. 1990), our in situ hybridization assays are failing to detect very low levels of expression in unidentified tissues. We estimate such expression to be <10 copies mRNA/cell (discussed later).

On embryonic day ⁠, the Fgf-5 gene is expressed in one small region of splanchnic mesoderm ventral to the portion of the foregut bearing the hepatic bud (Fig. 3, panels A–C). Expression in this mesenchyme is less readily detectable on (data not shown), and undetectable thereafter. The onset of expression is approximately coincident with that of rapid liver growth, which proceeds by the invasion of hepatic cords from the hepatic bud into the underlying mesenchyme. Whether the apparent fall-off of splanchnic mesodermal Fgf-5 expression from to represents down-regulation or merely the dilution of signal due to mesenchyme thinning and intermingling with hepatic cells cannot be ascertained.

Fgf-5 RNA is expressed in a region of somatic mesoderm between and ⁠. On embryonic day ⁠, this region extends rostrocaudally from the level of the newly formed sixth aortic arch to approximately the level of the liver primordium. Fig. 3 shows this expression in a transverse section through a embryo (panels F,G) and a parasagittal section through a embryo (panels H,I). Fgf-5 expression also occurs in myotomes at these times (see below). The somatic mesoderm signal is readily distinguished from that of skeletal muscle precursors by hybridization of adjacent sections with a probe for α-cardiac actin mRNA (Sassoon et al. 1988), which is expressed in striated muscle cells and their precursors. This is illustrated in Fig. 3, panel J, where the actin probe detects myotomes and cardiac muscle in the embryo, but does not hybridize to somatic mesoderm.

In order to illustrate the overall region of somatic mesoderm expressing Fgf-5 RNA in the embryo and the proximity of this region to vascular structures, the data in thirty transverse sections spaced 96 microns apart were used for three-dimensional computer modelling. CARP software (Computer-Aided Reconstruction Package, Biographies Inc.) was used to generate models of the entire embryo, dorsal aorta and aortic arches, primary head veins and cardinal trunk veins, and sites of somatic mesoderm and myotome Fgf-5 expression (Fig. 4). The broad anterior regions of Fgf- 5-positive somatic mesoderm (panels A–C) lie both lateral and just caudal to the newly formed sixth aortic arch. It is worth noting that this approximate region of expression sees the enlargement of the sixth arch and the emergence of a caudally projecting vascular branch (the future pulmonary artery) during the ensuing 36 h of development. The caudal ‘tails’ of expression in the embryo run adjacent to the paired cardinal veins (panels B,C).

Skeletal muscles of the trunk and limbs are descendants of the somites, which arise by segmentation of paraxial mesoderm in an anterior-to-posterior sequence beginning at ⁠. Each maturation phase of anterior somites precedes that of posterior ones owing to their different times of birth. The precursor cells for limb muscle emigrate from the newly formed myotome derivative of the somite and enter limb buds shortly after their formation on (forelimb) and E10 (hindlimb) (Sassoon et al. 1989). The trunk muscle precursor cells remain in the segmented myotomes and commence migration on to ⁠.

Fgf-5 RNA is expressed in myotomes starting on (Fig. 3, panels D–G). This induction occurs later than the morphological appearance of myotomes and their expression of α-cardiac actin mRNA ((Sassoon et al. 1989), and our observations). On embryonic day ⁠, only myotomes anterior to the middle of the hindlimb bud are Fgf-5 positive, as shown in the 3-D model (Fig. 4, panel A). More caudal myotomes also express Fgf-5 RNA a day later, reflecting their delayed development, but tail region myotomes never express Fgf-5 (data not shown).

Myotomal cells continue to express Fgf-5 RNA on (Fig. 3, panels H–J) and (data not shown) during their ventral and lateral migration. Expression in trunk muscle precursor cells is barely evident in embryos and is undetectable by ⁠. By contrast to trunk muscle differentation, precursors of limb muscles, identifiable by α-cardiac actin RNA expression (see below), do not express Fgf-5 RNA at any stage in their development.

Cranial skeletal muscles (tongue, mastication, extrinsic ocular and facial) all derive from paraxial mesoderm in birds (Noden, 1988), and similar origins in mammals are suspected. A prominent site of Fgf-5 RNA expression in the head of the embryo (Fig. 5, panel F) corresponds to the mastication muscle precursor cells. Hybridization of a nearly adjacent section with the a-cardiac actin probe (Fig. 5, panels D,E) reveals expression in mastication muscles as well as in the extrinsic ocular muscles. Facial and tongue muscle lineages, as recognized by α-cardiac actin hybridization, are also Fgf-5 negative (data not shown). The strong Fgf-5 mastication muscle signal persists on ⁠, but is completely absent 24 h later (data not shown). This site of Fgf-5 RNA expression originates on embryonic day as bilateral patches, each situated near the juncture of the maxillary process and the mandibular arch (Fig. 5, panels A,B). Curiously, a-cardiac actin RNA expression is very weak at this site (Fig. 5, panel C), and only becomes prominent on (data not shown).

These cranial expression data are open to radically different interpretations. Mastication muscle precursor cells may strongly express Fgf-5 RNA in advance of substantial a-cardiac actin RNA expression. This would represent the opposite temporal order of gene expression from that which occurs in myotomes. Alternatively, the cranial Fgf-5 RNA expression at may be in mesenchyme of neural crest or lateral mesoderm origin which neighbors or is intermingled with muscle cell precursors.

Fgf-5 RNA has been detected in developing limbs from embryonic days through ⁠. Expression is limited to a patch of cells near the base of each limb, and while expression is readily detected in hindlimbs, it is far less evident in forelimbs. Panels A and B of Fig. 6 shows the expression patch in a transverse section through the hindlimb. In the embryo, parasagittal sections, which effectively represent cross-sections through the base of the limb, show that the mesenchymal region of Fgf-5 expression (panel E) is ventral to the femur (panel D), which is undergoing chondrification. The expression site is distinct from the various developing muscle groups, which are visualized in adjacent sections by their expression of a-cardiac actin RNA (panel E). The tracing in panel G illustrates the spatial relationship of Fgf-5 positive mesenchyme to developing muscle and cartilage. The weaker Fgf-5 RNA signal in forelimbs is positioned similarly to that in hindlimbs (data not shown).

Fgf-5 RNA has been detected in the acoustic branch of the eighth cranial ganglion within the inner ear on embryonic days through (Fig. 6, panels G-J). All other cranial ganglia lack detectable Fgf-5 RNA, including the vestibular branch of the eighth ganglion (Fig. 6, panels G,H) and the trigeminal fifth ganglion (Fig. 5, panels D, F). The cochlea is another developmental site where Fgf-5 and int-2 RNAs are expressed in neighboring cells, int-2 is expressed in portions of the cochlear sensory epithelium and underlying mesenchyme (Wilkinson et al. 1989), which are innervated by projections from acoustical neurons.

The seven sites of Fgf-5 embryonic expression are catalogued in Table 1.

The measurement of intracellular Fgf-5 RNA concentration is only possible in tissues where dissection allows isolation of expressing cells to a known degree of purity. The one site amenable to RNA quantitation is the postimplantation egg cylinder. On embryonic day ⁠, embryonic ectoderm accounts for ∼70% of the cells in the embryonic compartment (embryonic ectoderm, mesoderm, flanking visceral endoderm) (Snow, 1977), and this compartment is larger than the extraembryonic compartment of the egg cylinder. Hence, the embryonic ectodermal cells expressing Fgf-5 RNA constitutes 40–50 % of the cells in the egg cylinder.

Thirty egg cylinders (embryonic day ⁠) were dissected from decidua, teased free of ectopiacental cone tissue and dissolved in guanidinium isothiocyanate for the isolation of RNA. An aliquot of the RNA was analyzed by northern blotting for GAPDH mRNA, a constitutively expressed message, as compared to GAPDH mRNA levels in known amounts of ES cell RNA. The analysis demonstrated a recovery of one microgram total RNA from the egg cylinders (data not shown). The egg cylinder RNA was then analyzed for Fgf-5 mRNA by northern blot hybridization, and the signal compared to those obtained for several quantities of Fgf-5 cDNA. As shown in Fig. 7, one microgram RNA from 30 egg cylinders gave a 2.9 kb Fgf-5 RNA signal (lane e) comparable to 2 picograms 2.2 kb Fgf-5 cDNA (lane b). From this, we calculate that Fgf-5 mRNA represents 2.5×10⁻ of the total egg cylinder RNA. Since the embryonic compartment of each egg cylinder contains ∼3000 cells (Snow, 1977), the estimated level of egg cylinder Fgf-5 expression is 30–40 copies/embryonic ectodermal cell.

The Fgf-5 RNA expression level in the egg cylinder (lane e) is similar to that in ES cells allowed to differentiate into simple embryoid bodies following three days in suspension (lane g), and is dramatically higher than the level expressed in undifferentiated ES cells, which is undetectable at the exposure length shown (lane f). These simple embryoid bodies have morphological features comparable to the egg cylinder prior to gastrulation (Martin et al. 1977). Hence, the induction of Fgf-5 expression in embryoid bodies quantitatively and temporally parallels the Fgf-5 gene’s in vivo induction postimplantation.

The intensity of the Fgf-5 in situ hybridization signal in egg cylinders was consistently equal to or stronger than those seen locally at the later sites of Fgf-5 gene expression. We estimate local expression in later embryogenesis to be at 10–30 copies mRNA/cell. Due to the constraints of signal versus background, we do not feel that our assay conditions could detect <10 copies RNA/cell. Hence, sites of very weak expression could be missed in our analysis. This presumably explains our failure to define the sites of Fgf-5 expression in the day embryo. Additionally, our preliminary in situ analysis of embryos, using fresh frozen tissue for higher sensitivity, has revealed weak sites of Fgf-5 expression (dorsal root ganglia, myenteric ganglia) in addition to the more prominent hindlimb, mastication muscle and acoustic ganglion signals described earlier.

Embryonic expression of the Fgf-5 gene is regulated as a function of time, tissue type and position within tissue. This lattermost characteristic has defined subdivisions of splanchnic mesoderm, somatic mesoderm and limb mesenchyme that were not morphologically evident nor previously appreciated by other molecular criteria. The spatial restriction of expression may, in part, be governed by combinations of transcription factors, such as homeobox proteins, which themselves are expressed within spatial boundaries. For example, the expression boundaries of various homeobox genes in vertebrate limbs define a ‘grid’ along both anterior– posterior and proximal–distal axes (Smith et al. 1989), and this grid may help dictate Fgf-5 expression in limb mesenchyme. Another manifestation of spatially restricted expression is the presence of Fgf-5 RNA in some skeletal muscle groups and cranial ganglia, but not in others.

The structurally related int-2 gene is also expressed in a complex spatiotemporal pattern, although the sites of expression differ from those for the Fgf-5 gene (Wilkinson et al. 1989; Wilkinson et al. 1988). At certain stages in development, Fgf-5 and int-2 are expressed in close proximity: Fgf-5 in visceral endoderm versus int-2 in parietal endoderm, Fgf-5 in embryonic ectoderm versus int-2 in embryonic mesoderm, and Fgf-5 in acoustic ganglion versus int-2 in cochlear sensory epithelium. Since both of these growth factors are secreted proteins (Bates et al. 1991; Dixon et al. 1989), we suspect that FGF receptors expressed in these regions differentially react with or respond to these two FGFs. While the known spectra of in vitro biological activities are similar among FGFs, there are contrasts (Finch et al. 1989; Lipton et al. 1988; Valles et al. 1990) which are presumably mediated by differences in ligand–receptor interactions.

FGF-5 is likely to mediate a diverse set of events during embryogenesis. The Fgf-5 expression profile along with known biological effects of other FGFs in in vitro and in vivo assays allow for speculation regarding FGF-5’s native functions. Such conjecture merely serves as a guide for further studies.

Uterine implantation of the blastocyst is followed by rapid growth of the egg cylinder and rapid growth of the uterine deciduum. Fgf-5 expressed in embryonic ectoderm postimplantation could serve as an autocrine or paracrine factor to mediate these growth events. Several FGFs have been shown to possess mesoderminducing activity when assayed on cultures of amphibian animal cap ectoderm (Kimelman and Kirschner, 1987; Paterno et al. 1989; Slack et al. 1987). The expression of Fgf-5 RNA at the onset of mouse gastrulation suggests that this growth factor may contribute to mesoderm formation, although two facts argue against a direct role for FGF-5 in this process: (1) the onset of gene expression far precedes the start of gastrulation ⁠, and (2) Fgf-5 RNA is not spatially restricted within the embryonic ectoderm. Future localization of Fgf-5 protein and receptors in the developing egg cylinder is of obvious importance.

Acidic and basic fibroblast growth factors are mitogenic for vascular endothelial cells in vitro and are angiogenic when applied to tissues in vivo (Burgess and Maciag, 1989). FGF-5 is most likely an endothelial cell mitogen, since conditioned medium containing secreted FGF-5 will stimulate endothelial cell growth (Zhan et al. 1988). FGF-5 might act as an angiogenic factor in lateral mesoderm, which becomes more highly vascularized than dorsal mesoderm (Le Douarin, 1975; Sherer, 1975). In splanchnic mesoderm, local Fgf-5 expression might promote vascularization required for the induction of neighboring hepatic cord proliferation (Sherer, 1975). In somatic mesoderm, local Fgf-5 expression could contribute to the ongoing remodelling of the arterial vasculature (Rugh, 1968).

FGFs have well-documented effects upon the differentiation of skeletal muscle progenitor cells. Fibroblast growth factors can inhibit the terminal differentiation of myoblasts, as monitored by biochemical markers and by myotube formation (Lathrop et al. 1985; Seed and Hauschka, 1988). Fgf-5 expression commences in myo tomes on well after their formation and commitment to muscle lineage, as gauged by the onset of a-cardiac actin and MyoD gene expression (Sassoon et al. 1989). Fgf-5 expression diminishes by ⁠, before the first appearance of myotubes on E15 (Rugh, 1968). FGF-5 could inhibit the terminal differentiation of myotomal myoblasts during their migration through the trunk. Limb muscle development differs from that of trunk muscle by several molecular criteria. The onset of α-cardiac actin and MyoD expression in limb muscle precursors occurs later in embryogenesis (Sassoon et al. 1989), although appearance of myofibrils occurs at similar times for limb and trunk muscle (Rugh, 1968); i.e. the later stages of muscle differentiation transpire within a more compressed time frame in limb versus trunk. A lack of Fgf-5 expression in limb muscle myoblasts might account for their more rapid terminal differentiation.

In vitro culture of dissociated chicken limb bud cells has revealed an FGF-dependent subcomponent of myoblasts which require treatment with basic FGF in order ultimately to form colonies of terminally differentiated muscle (Seed and Hauschka, 1988). It is possible that Fgf-5 expression in limb mesenchyme acts to induce the development of FGF-dependent limb myoblasts. Furthermore, if the mandibular site of Fgf-5 expression consists of mesenchymal as opposed to myoblastic cells, such expression could act to induce local myoblasts to develop into the mastication muscle groups.

Several lines of experimentation are required to elucidate FGF-5’s roles in embryogenesis. FGF’s receptors and their sites of expression need to be identified. Alternatively spliced transcripts of the FLG and BEK genes encode a family of related FGF receptors (reviewed in (Goldfarb, 1990)), but their affinities toward FGF-5 await characterization. A more precise cellular assignment of Fgf-5 expression is clearly desirable, particularly as it relates to mesenchymal versus myoblastic gene expression in the mandibular region. Transgenes linking an enzymatic reporter to Fgf-5 transcriptional regulatory elements may provide such insights. Lastly, the use of in vivo gene targetting to disrupt FGF-5 production partially or completely may generate informative embryonic phenotypes.

We thank Beverly Drucker for help in several experiments, and Matthew Buttrick for assisting in paraffin embedding and sectioning. We are most grateful to David Sassoon (Boston University) and to Godon Peters (Imperial Cancer Research Fund, London) for α-cardiac actin and int-2 clones, respectively, to Elizabeth Robertson for providing ES cells, to Francoise Poirier for instruction in egg cylinder dissection, to James Lee for help with blastocyst preparation, to Connie Cepko (Harvard Medical School) for access to her SunView/ CARP 3-D imaging system, and to Lisa Jepeol and Elizabeth Ryder for instruction in its use. We also thank Vassilus Pachnis, Jane Dodd, Taube Rothman, David Sassoon, Elizabeth Taparowsky (Purdue University) and Cliff Tabin (Harvard Medical School) for helpful discussions. This work was supported by PHS research grant CA48054. O.H. is supported by NIH Training Grant DK07328.