Main

FGF-2 (or basic FGF) is a potent mitogenic and angiogenic peptide, and one of a family of nine related FGF molecules expressed differentially during rodent embryonic development(1–3). Four high affinity FGFR have been identified, designated FGFR1-4, which are all capable of binding FGF-2(4, 5). FGF and their receptors have been linked with specific morphogenic events in early embryogenesis; FGF-4 being necessary for postimplantational development in mouse and for limb pattern formation(6, 7), FGF-2 and FGFR1 for mesodermal patterning in the amphibian embryo(8, 9), FGF-3 controlling formation of the inner ear in mice(10), and FGFR2 mediating effects of FGF on murine embryonic lung branching(11).

In the rat we showed that the expression of FGF-2 persisted in many tissues during late fetal life, and that immunoreactive FGF-2 was associated with extracellular matrix(12). The high affinity of FGF-2 and other FGF for heparan sulfate proteoglycans on the cell surface and within extracellular matrix(1, 13) could provide an extracellular storage depot for the growth factors, whose availability to high affinity receptors may depend on proteolytic modification of the matrix during the tissue remodeling of fetal development. Because the association of FGF-2 with heparan sulfate can enhance recognition and signaling by high affinity receptors(14), a tissue co-localization of extracellular FGF-2 and plasma membrane-associated receptor may indicate important anatomical sites of FGF action. Little information exists as to where FGF-2 and its receptors are expressed during human fetal development. In this study, we analyzed the anatomical distribution of FGF-2 and FGFR1 mRNA and peptides in a variety of tissues from the 12- to 16-wk human fetus. Evidence is presented for a widespread and coordinated expression of both growth factor and receptor, and for the storage of FGF-2 within extracellular matrix.

METHODS

Collection and preparation of tissues. Tissues were obtained from human fetuses of 12- to 16-wk gestation and collected within 30 min of prostaglandin-induced abortion at the Northern General Hospital, Sheffield, England. Approval of the local ethical committee was obtained. The weight, length, and sex of each fetus were recorded. Tissue samples (0.5 cm3) of skeletal muscle (quadriceps femoris), heart, abdominal skin, epiphyseal growth plate (proximal tibia), brain (cerebral cortex), lung, pancreas, stomach, ileum, kidney, adrenal, thyroid, and liver were fixed by immersion in 2% (wt/vol) paraformaldehyde and 2% (vol/vol) glutaraldehyde in 67 mM PBS (pH 7.4) at 4°C for 18 to 24 h. After two rinses in 0.01 M PBS (pH 7.4; Sigma Chemical Co., St. Louis, MO), each of 24 h, tissues were dehydrated in ascending ethanol series (70, 90, and 100%) and xylene before embedding in paraffin. Tissue sections (8 μm) were mounted on poly-L-lysine-coated glass slides and stored at 4°C. Tissue samples prepared from four to eight separate fetuses were analyzed by immunohistochemistry or in situ hybridization.

Immunocytochemistry for FGF-2 and FGFR1. The technique used was the avidin-biotin peroxidase method(15) adapted for the visualization of FGF-2 as reported previously(13). Tissue sections were rehydrated in descending ethanol series (100, 90, and 70%) into PBS, and incubated in 0.5% (vol/vol) hydrogen peroxide in methanol for 30 min to eliminate endogenous peroxidase activity. After washing in PBS, sections were exposed to 1.5% (vol/vol) normal goat serum in PBS to block nonspecific binding. For the detection of FGF-2 immunoreactivity, a rabbit primary antiserum against amino acids 1-24 of bovine FGF-2 (1-146) was used. A purified IgG antibody fraction was obtained by precipitation of the antiserum with 30% (wt/vol) ammonium sulfate and passage over a protein A-Sepharose affinity column (Pharmacia Biotech Inc., Piscataway, NJ). The antibody demonstrates less than 1% cross-reactivity with FGF-1, FGF-3 (int-2), FGF-4(hst/ks), FGF-5, FGF-6, and FGF-7. Cross-reactivity with FGF-8 and FGF-9 is not known. However, each of these peptides shows low sequence homology with the peptide fragment of FGF-2 used as the antigen(12). For the immunolocalization of FGFR1, a rabbit polyclonal antibody against a peptide encoding the carboxy terminus domain of human FGFR1 (amino acids 806-822) was affinity purified by passage over a column of FGFR1 peptide coupled to Affi-Gel 102 (Pharmacia). This antibody recognizes chicken (cek-1), human and rat FGFR1, but not FGFR2, R3, or R4. The slides were incubated with 2.5-10 μg/mL of either rabbit anti-FGF-2 or anti-FGFR1 antibody diluted in 0.01 M PBS (pH 7.5) containing 5% (wt/vol) BSA, 0.01% (wt/vol) sodium azide, and 0.3% (vol/vol) Triton (100 μL per slide) for 24h at 4°C in a humidified chamber. After washing in PBS, biotinylated goat anti-rabbit immunoglobulin (1:500 dilution; Vector Laboratories, Burlingame, CA) was added to each slide (100 μL) for 1.5 h at 23°C. After a further wash, the sections were incubated with avidin and biotinylated peroxidase (Vector) for 1 h at 23°C, washed with 10 mL of PBS, and then with 10 mL of 50 mM Tris buffer (pH 7.5). Immunoreactivity was visualized using 0.5% (wt/vol) freshly prepared 3,3′-diaminobenzidine tetrahydrochloride (Aldrich Chemical Co., Milwaukee, WI; 1.89 mM containing 0.03% (vol/vol) hydrogen peroxide). The reaction was quenched by washing in excess of 50 mM Tris buffer. The sections were lightly counterstained with hematoxylin, dehydrated in an ascending ethanol series and xylene, and mounted under glass coverslips before examination by light microscopy.

Specificity of staining was verified by several criteria. Immunostaining was absent when 1) nonimmune, protein A-Sepharose-purified rabbit IgG was substituted for the primary antibody; 2) the primary antibody against FGF-2 was preincubated for 24 h at 4°C with 240 nM recombinant human FGF-2 (Upstate Biotechnology, Inc., Lake Placid, NY) in a siliconized, polypropylene tube before application to the slide; 3) the eluant from the FGFR1 affinity column used to purify the antibody was substituted for the primary antibody; or 4) the anti-FGFR1 antiserum was similarly preincubated with an excess of antigen.

In situ hybridization. For the preparation of the human FGF-2 cRNA, a 0.46-kb cDNA encoding human FGF-2 was used. The coding region of a synthetic human FGF-2 gene was inserted into pBluescript SK+ or KS+ (Stratagene, San Diego, CA) and linearized with HindII.In vitro transcription was performed using T7 RNA polymerase(Boerhinger Mannheim, Indianapolis, IN) and 35S-UTP (Amersham International, Mississauga, ON, Canada). For the preparation of the human FGFR1 cRNA, a full-length cDNA for the FGFR1 receptor was cut atBgl II sites to yield a 1.6-kb fragment coding for the extracellular domain, and this fragment was inserted into pBluescript SK+ or KS+ (Stratagene). 35S-UTP-labeled cRNA (1 × 107 cpm/mL) were synthesized in a transcriptional run-off reaction using T7 polymerase (Boerhinger Mannheim) to yield labeled sense and antisense RNA, as described by us previously(16). Briefly, the tissues were prepared as described for the immunohistochemical studies. Paraffin sections were mounted on poly-L-lysine-coated slides, digested with proteinase K (10 mg/mL; 37°C for 30 min), acetylated for 10 min, rinsed in 2 × SSPE, dehydrated through a graded series of ethanol washes, and then air-dried for 2 h before hybridization. Hybridization was performed at 55°C overnight in 10 mM Tris, pH 8.0, containing 50% formamide, 0.3 M NaCl, 1 mM EDTA, 0.05% tRNA, 10 mM dithiothreitol, 1 × Denhardt's solution, and 10% dextran sulfate. After hybridization, sections were treated with ribonuclease A (25 mg/mL; 37°C for 30 min) and washed in 0.1 × SSPE, 1 mM dithiothreitol at 65°C. Dehydrated slides were exposed to βmax film(Amersham) for 5 d. For microscopic analysis, slides were coated with Kodak NTB-2 liquid autoradiographic emulsion, and exposed at 4°C for 2 wk. They were developed in Kodak D-19 for 3.5 min, rinsed, and fixed. After washing in distilled water the sections were counterstained with hematoxylin and eosin and analyzed under bright and darkfield microscopy.

RESULTS

Both FGF-2 and its receptor, FGFR1, are widely expressed and distributed amongst human fetal tissues in the early second trimester. FGF-2 immunoreactivity is generally strong and is often associated with basement membranes underlying epithelial cell layers. mRNA for FGF-2 is found in both mesenchymal and epithelial cell populations. FGFR1 mRNA and peptide are co-localized within the same cell types of both mesenchymal and epithelial origin, often in juxtaposition to cell populations expressing abundant FGF-2, or to likely sites of extracellular storage of FGF-2. No differences in FGF-2 or FGFR1 distribution or sites of expression were found between male and female fetuses for the tissues studied. In situ hybridization using sense strand cRNA, and preabsorbtion of the antisera against FGF-2 and FGFR1 with excess homologous ligands before immunocytochemistry resulted in low, nonspecific signal in all tissues studied.

Musculoskeletal tissues. Hybridization signal for FGF-2 mRNA is detected in differentiated skeletal muscle fibers, as well as in isolated satellite cells (Fig. 1A). Hybridization with a sense strand cRNA shows low hybridization signal, confirming the specificity of hybridization to FGF-2 mRNA (Fig. 1B). Immunolocalization studies confirm the presence of FGF-2 peptide associated with the plasma membrane of muscle cells (Fig. 2A). Little sarcoplasmic staining for FGF-2 is seen in skeletal muscle fibers. Although most nuclei are immunonegative, a minority show a presence of FGF-2, which is most apparent for the satellite cell population of fibroblastic cells with elongated nuclei(Fig. 2A). In situ hybridization shows that mRNA encoding FGFR1 is also present in differentiated muscle fibers(Fig. 3A), and co-localizes with immunoreactivity for FGFR1 protein in the sarcoplasm (Fig. 4A). Many nuclei also show positive immunoreactivity for FGFR1. No hybridization signal is seen when tissue sections are incubated with a sense strand cRNA encoding FGFR1(Fig. 3B). FGF-2 mRNA and FGF-2 immunoreactive peptide are present in the resting and proliferative chondrocytes of the epiphyseal growth plate (Fig. 5), and in the cells of the perichondrium. FGF-2 immunoreactivity is observed in both the cytoplasmic and nuclear compartments of these cells. Resting and proliferative chondrocytes, and perichondrial cells, also contain FGFR1 mRNA and peptide. FGF-2 and FGFR1 mRNA and peptides are absent from the differentiated, hypertrophic chondrocytes.

Figure 1
figure 1

Localization of FGF-2 mRNA within histologic sections of human fetal tissues using in situ hybridization. Hybridization signal with antisense 35S-cRNA (left- and right-hand columns), or radiolabeled sense probe (centre column) for some tissues, is shown for representative sections by bright field microscopy. Representative views are shown for skeletal muscle (A, B; arrows indicate signal within the sarcoplasm), small intestine (C; arrow indicates signal in the submucosa), heart (D, E; arrow indicates signal in the cardiocytes), lung (F), skin (G, H; arrow indicates signal within the dermis), and kidney (I; arrows indicate signal within the tubular epithelium and glomerular cells). The horizontal bar represents 5 μm.

Full size image
Figure 2
figure 2

Localization of FGF-2 immunoreactivity within histologic sections of human fetal tissues using immunohistochemistry. Representative views are shown for skeletal muscle (A, arrows indicate staining on the sarcolemma; B, primary antiserum preabsorbed with FGF-2), heart (C), vein (D, arrow indicates endothelial cell staining), intestine (E, arrow indicates staining associated with mucosal basement membrane), kidney (F, arrows indicate strong staining associated with basement membranes), skin[G, epidermis (e), germinal zone (gz), dermis(d)], lung (H, arrows indicate staining within basement membranes of the alveoli), and cerebral cortex (I). The horizontal bar represents 5 μm in E and G, and 8 μm inA-D, F, H, and I.

Full size image
Figure 3
figure 3

Localization of FGFR1 mRNA within histologic sections of human fetal tissues using in situ hybridization. Hybridization signal with antisense 35S-cRNA (left and right hand columns), or radiolabeled sense probe (center column) for the some tissues, is shown for representative sections by bright field microscopy. Representative views are shown for skeletal muscle (A, B; arrows indicate signal within the sarcoplasm), small intestine (C; arrow indicates signal in the mucosa), heart (D, E; arrow indicates signal in some cardiocytes), lung (F; arrows indicate abundant signal in the alveolar epithelia), skin (G, H; arrows indicate signal within the germinal layer of the epidermis), and kidney (I; arrows indicate signal within the tubular epithelium). The horizontal bar represents 5 μm.

Full size image
Figure 4
figure 4

Localization of FGFR1 immunoreactivity within histologic sections of human fetal tissues using immunohistochemistry. Representative views are shown for skeletal muscle (A, arrows indicate staining in the sarcoplasm; B, primary antiserum substituted with eluant from the FGFR1 affinity column), heart (C), vein (D, arrow indicates endothelial cell staining), intestine(E, arrow indicates staining associated with mucosal cell membranes and nuclei), kidney (F, arrows indicate strong staining associated with the tubular epithelia and glomeruli), skin [G, epidermis(e), hair follicle (f), germinal zone (gz)], lung (H), and cerebral cortex (I). The horizontal bar represents 8 μm in A-C, E-I, and 5 μm in D.

Full size image
Figure 5
figure 5

Localization of FGF-2 immunoreactivity within histologic sections of human fetal epiphyseal growth plate cartilage using immunohistochemistry. (A) proliferative zone [perichondrium(p), germinal zone (gz)]; (B) hypertrophic chondrocytes. The horizontal bar represents 5 μm.

Full size image

Heart. FGF-2 mRNA is associated with cardiocytes(Fig. 1D), whereas the FGF-2 immunoreactivity in the heart localizes to the extracellular matrix and/or cardiocyte plasma membranes(Fig. 2C). Only weak immunoreactivity for FGF-2 is found in the cytoplasm of cardiocytes, and no nuclear staining is observed. FGFR1 immunoreactivity in heart is generally weak, and is restricted to the cardiocyte plasma membranes in most cells (Fig. 4C). However, a minority of cardiocytes show a more intense staining for FGFR1 within the cytoplasm. Similarly, although FGFR1 mRNA is barely detectable in many cardiocytes after in situ hybridization, a minority of cells show a dense hybridization signal (Fig. 3D).

Blood vessels. Immunoreactive FGF-2 and FGFR1 are strongly localized to the endothelium of arterial and venous blood vessels in all of the organs studied (Figs. 2D and4D). Small capillaries also show immunostaining for FGF-2 and FGFR1 in endothelial and perivascular cells (not shown). Arterial smooth muscle cells also show cytoplasmic immunoreactivity for FGF-2. Both FGF-2 and FGFR1 mRNA are located in vascular endothelial cells and associated smooth muscle (not shown).

Gastrointestinal tract. Little FGF-2 mRNA is detected byin situ hybridization within the mucosal components of the stomach or small intestine, but the hybridization signal is diffusely distributed within the submucosal stromal tissues (Fig. 1C). In contrast, an intense immunoreactive staining for FGF-2 is localized to the basement membranes underlying the mucosal epithelia of the gut, associated with the extracellular matrix of the mucosal epithelial cells and within capillary endothelial cells of the submucosa (Fig. 2E). A similar pattern of staining for FGF-2 is seen in the submucosal tissue underlying both the crypt regions and the tips of the villi. No nuclear localization of FGF-2 is found in either mucosal or submucosal cell types. FGF-2 immunoreactivity and mRNA are also present in the serosa, but are less apparent in smooth muscle layers. mRNA for FGFR1 is juxtapositioned to that for FGF-2, being mainly located in the mucosal layers of the stomach and intestine, and within the crypts of the small intestine (Fig. 3C). Mucosal epithelial cells also contain immunoreactivity for FGFR1; this being associated with the plasma membranes and within the nuclei of approximately half of the columnar epithelial cells (Fig. 4E). Goblet cells are immunonegative. Immunoreactivity for FGFR1 is also found in the vascular endothelial cells of capillaries permeating the submucosal areas, and although sparse in the serosa, staining is found in the intestinal smooth muscle.

Lung. FGF-2 mRNA is expressed within the epithelial layer of the developing airways, as well as in the surrounding interstitial connective tissue (Fig. 1F). FGF-2 immunoreactivity is particularly strong within the basement membranes of the airways, but can also be visualized at the intercellular junctions of the alveolar epithelial cells and the connective tissue (Fig. 2H). No nuclear staining for FGF-2 is seen in the epithelial cell populations, but some stromal cells are positive. The epithelial cells of the airways contain immunoreactive FGFR1 and mRNA (Figs. 4H and3F, respectively). FGFR1 protein is localized to both the plasma membranes and nuclei of alveolar epithelial cells. Additionally, both FGFR1 mRNA and immunoreactivity are widely distributed within the interstitial connective tissue, and staining is often associated with stromal cell nuclei.

Kidney. mRNA for FGF-2 is widely expressed within the kidney, located within the renal tubular epithelia, in the mesangial cells of the glomeruli, and within the surrounding stromal tissues (Fig. 1I). Immunoreactive FGF-2 is localized to both the stroma and renal tubules. In the stroma, FGF-2 immunoreactivity is associated with the intercellular matrix, but not with cell nuclei or cytoplasm(Fig. 2F), and is equally distributed throughout the cortex and medulla. Intense immunoreactivity is found on the basement membranes of the collecting, distal, and proximal tubules, whereas the tubular epithelial cells in all regions are immunonegative. FGF-2 immunoreactivity is also seen on the basement membranes of the Bowman's capsules, whereas mesangial cells are largely unstained. FGFR1 mRNA is present in the tubular epithelial cells with an intense signal in the comma and S-shaped bodies located in the outer cortex. Hybridization signal is seen in the vascular tissue and mesangial cells of the glomeruli, but less is found in the surrounding stroma (Fig. 3I). FGFR1 immunoreactivity is juxtaposed to that for FGF-2 in the renal tubules, being intense within the cytoplasm of the tubular epithelium, especially in the comma and S-shaped bodies (Fig. 4F). Staining for FGFR1 is less intense in the epithelial cells of the collecting ducts, and in the ureteric bud. Staining for FGFR1 is also evident on the plasma membranes and nuclei of stromal tissue, especially in the outer rim of undifferentiated blastemic mesenchyme, in some mesangial cells, and within the vascular endothelium supplying the glomeruli.

Skin. mRNA for FGF-2 is widely distributed within the connective tissue of the dermis, but is less dense in the epidermal layers(Fig. 1G). FGF-2 is immunolocalized to the basement membranes underlying the epithelial cells of the stratum germinativum of the epidermis, and is also located within the extracellular matrix between cells of the dermis (Fig. 2G). Staining for FGF-2 is also associated with the surfaces of the outermost keratinocytes of the epidermis, but little nuclear localization of FGF-2 is seen in any cell population in skin. mRNA and immunoreactivity for FGFR1 are present mostly in the stratum germinativum of the epidermis, especially associated with presumptive follicles. This is adjacent to FGF-2 within the underlying basement membrane(Figs. 3G and4G). A sparce hybridization signal intensity for FGFR1 mRNA is seen within the dermis, and FGFR1 immunoreactivity is similarly low. FGFR1 immunostaining co-localizes with FGF-2 in the outer cell layer of the epidermis, and FGFR1 mRNA can also be visualized in these cells.

Brain. The brain cortex contains both immunoreactive FGF-2 and FGFR1 which are widely distributed but not homogeneous, the intensity of the staining being greatest in areas that are abundant in neurons(Figs. 2I and4I). Many cell nuclei are immunopositive. In the inner layers of the cortex, the FGF-2 and receptor immunoreactivity is mainly associated with fibers and with basement membranes of the capillary endothelia. Immunoreactive FGFR1 is also localized to the perikarion of scattered cortical neurons. FGF-2 and FGFR1 mRNA hybridization signals are found throughout the cortex associated with the neuronal cell population (not shown).

Other tissues. Within the fetal adrenal gland, FGF-2 mRNA and immunoreactivity are absent from the definitive cortical zone, but are localized within the cytoplasm of many cells within the fetal cortical zone. FGFR1 mRNA and peptide are detected in subpopulations of cells throughout the definitive cortex, as well as the fetal cortex. In the thymus FGFR1 immunoreactivity is seen within the cytoplasm of specific populations of rounded thymic cells. Their nuclei are immunoreactive. In contrast, FGF-2 immunoreactivity is limited to the thymic capsular cells, and the underlying basement membrane. In the fetal pancreas, FGF-2 immunoreactivity is limited to the intercellular matrix of the exocrine glandular tissue and the islets of Langerhans. FGF-2 mRNA hybridization signal is sparce throughout the pancreas. FGFR1 immunoreactivity is found within the cytoplasm of exocrine secretory cells and islets. Within the developing thyroid gland, FGF-2 immunoreactivity is present within the basement membranes surrounding follicles, whereas FGFR1 immunoreactivity is detected on the plasma membranes of follicular epithelial cells, and within the capillary endothelial cells of the interfollicular stroma. In liver, both immunoreactive FGF-2 and FGFR1 are present in vascular endothelial cells and some peri-sinusoidal cells, but no immunoreactivity is associated with the parenchymal or erythropoietic cell populations.

DISCUSSION

These results demonstrate that FGF-2 and one of its receptors, FGFR1, are widely expressed within human fetal tissues early in the second trimester. There is little previous information concerning the expression of FGF-2 in the human fetus. Yeh and Osathanondh(17) detected mRNA encoding FGF-2 in fetal ovaries and uteri between 10 and 22 wk of gestation, together with mRNA for FGFR1, using the polymerase chain reaction amplification. FGF-2 mRNA was also detected in the human fetal adrenal gland in early second trimester, and bioassayable FGF-2-like activity was found in adrenal extracts(18). mRNA encoding the four FGFR forms, FGFR1-4, was quantified in various human fetal tissues of 17-18 wk of gestation using Northern blot hybridization by Partanen et al.(19, 20). Abundant FGFR1 mRNA was found by these authors in the epiphyseal growth plate, skin, pancreas, and kidney; with lesser amounts in neural tissues, lung, and intestine, and no detectable mRNA in liver. Examination of the other FGFR forms showed a limited distribution of FGFR2 mRNA in skin, choroid plexus, and neural tissues; an abundance of FGFR4 in adrenal, kidney, intestine, pancreas, and lung; and FGFR3 in skeletal tissues, kidney, intestine, lung, and brain. FGFR isoforms therefore showed distinct but overlapping anatomical sites of expression in the human fetus; and FGFR1-4 are all possible signaling receptors for FGF-2(4, 5). The present study is the first to examine the cellular patterns of expression of FGF-2 and FGFR1 mRNA in the human fetus, and to compare these patterns with the cellular locations of the translated proteins.

A number of general observations can be made concerning the localization of FGF-2 in human development. First, in many tissues FGF-2 peptide was localized to cell membranes and extracellular matrix, as in striated muscle and heart, or basement membranes as seen in the lung, kidney and gastro-intestinal tract. Despite FGF-2 having no signal peptide sequence in its translated form, it would appear capable of leaving the cell. Using cell lines transfected with FGF-2 cDNA, novel secretory mechanisms independent of the Golgi and endoplasmic reticulum have been described(21, 22). The presence of FGF-2 immunoreactivity on cell membranes and extracellular matrix, accompanied by only slight cytoplasmic staining, suggests that matrix molecules may sequester and store FGF-2. Sulfated glycosaminoglycans such as heparin sulfate bind FGF-2 with high affinity and protect the growth factor from proteolytic degradation(13). Heparan sulfate may also enhance FGF biologic activity by presenting a favorable conformational complex to the high affinity receptor(14), and may participate in the cellular internalization of FGF-2(23). The intense immunoreactivity for FGF-2 found on some basement membranes suggests that much of the peptide may be unavailable to receptors on adjacent cell populations. Selective degradation by proteolytic enzymes may be a necessary step to liberate soluble FGF:glycosaminoglycan complexes, and could be a fundamental mechanism of regulating FGF-2 availability during fetal tissue growth and remodeling. The cellular patterns of FGF-2 immunoreactivity seen here in the human fetus closely agree with those reported by us previously in the rat(12).

In some tissues, such as striated and cardiac muscle, as well as cartilage and vascular endothelium, there is a concurrent cellular expression of both FGF-2 and FGFR1, suggesting autocrine or paracrine modes of action. In other tissues, there is a clear anatomical distinction between cell populations expressing either FGF-2 or the receptor. For instance, in skin, FGF-2 is predominantly expressed and located in the upper dermis, with FGFR1 in the adjacent epidermis. In the intestine FGF-2 mRNA and peptide are mainly in the submucosa and FGFR1 is located in the overlying mucosa. These observations suggest a mesenchymal:epithelial paracrine action for the growth factor. Within the kidney and lung, FGF-2 mRNA is situated in both epithelial and mesenchymal cell types. However, the FGF-2 immunoreactivity is not intense in the tubular or alveolar epithelia suggesting a rapid sequestration by the surrounding basement membranes. In contrast, FGFR1 peptide is found in cells of both the renal tubules and lung epithelium. Anatomical specificity of FGF-2 presence and action appears to occur both at the level of differential mRNA expression for this growth factor and its receptor, and by a tissue compartmentalization of FGF-2 peptide. However, it must also be considered that FGF-2 is able to interact with the four high affinity receptor species, FGFR1-4(4, 5), and that the cellular distribution of FGFR2-4 in the human fetus is unknown. We have previously reported the existence of circulating FGF-2 in the human fetal circulation(24), raising the possibility of a complimentary endocrine mode of action. The tissues which contribute FGF-2 to the fetal circulation are not known, but are likely to include placenta, which is rich in FGF-2 mRNA(25), and the vascular endothelium which was shown here to possess intense immunoreactivity for FGF-2.

The mitogenic actions of FGF-2 have been shown to be mediated by binding to the high affinity plasma membrane receptors leading to the activation of second messengers which include the ras GTP binding protein and the mitogen-activated protein kinase (MAPK) cascade(26). However, FGF-2 has also been shown to be translocated to the nucleus in several cell types including chondrocytes and vascular endothelial cells(27, 28), to specifically bind to DNA(29) and to activate the transcription of ribosomal genes(30). This suggests that certain nonmitogenic actions of FGF-2 may be mediated by direct nuclear actions. For instance, a transient nuclear localization of FGF-2 has been seen in presumptive mesoderm during germ layer formation in the Xenopus embryo(31). Nuclear import of FGF-2 may involve the internalization of the ligand from the cell surface, or may use endogenous, intracellular growth factor. When isolated chick cardiocytes were transfected with cDNA encoding FGF-2 molecules of different molecular sizes a cellular compartmentalization resulted, with the 18-kD molecule appearing within the cytoplasm but extended molecules up to 24 kD, which derive from alternative translation initiation codons, being preferentially localized to the nucleus(32). In this study we noted a nuclear presence of immunoreactive FGF-2 in striated muscle, in brain, in chondrocytes and in mesenchymal tissue within the lung. This may indicate that nuclear compartmentalization may contribute to specific temporal effects of FGF-2 in the differentiation of muscle and cartilage, the branching of the pulmonary airways, and the maturation of neuroblast and glioblasts within brain as discussed below. Immunoreactivity for FGFR1 was also associated with the nuclei in skeletal muscle, chondrocytes, brain and epithelial cells of the intestine and lung. A juxtanuclear localization of FGFR1 has been reported previously in 3T3 cells transfected with the human receptor cDNA by Prudovsky et al.(33), and we have recently described how an extracellular domain fragment of FGFR1 may act as an intracellular carrier protein for FGF-2 which delivers the growth factor to the nuclear membrane in isolated fetal chondrocytes(34).

The distribution of FGF-2 and FGFR1 mRNA and peptides in human fetal tissues support many of the actions proposed for FGF-2 as a result of studiesin vitro. The intense localization of both FGF-2 and FGFR1 immunoreactivity in the vascular endothelium is consistent with a potent mitogenic action of FGF-2 on isolated endothelial cells, and its angiogenic actions in vivo(1). The presence of FGF-2 and its receptor within striated muscle and heart supports a suggested role in myoblast and cardiocyte proliferation and the progression of differentiation. It has been suggested that exogenous FGF-2 can suppress the transcription of at least two myogenic regulatory genes, myogenin and MyoD1, in rodent myoblast cell lines(35, 36). However, other FGF isomers are also relevant to muscle development; FGF-4 is expressed in embryonic mouse muscle immediately before differentiation(37), whereas FGF-5 is expressed in myotome in the trunk immediately after muscle cell differentiation(38). The expression of FGF-5 in these cell lineages during migration into the limb buds persists. Thus, different FGF species may coordinate the amplification of myoblast populations, their migration schedules, and their eventual differentiation. FGF species acting via FGFR1 have also been shown to regulate cardiac myocyte growth, and the tubular stages of cardiogenesis in the chick embryo(39). The expression of FGF-2 and FGFR1 within chondrocytes of the epiphyseal growth plate supports the demonstrated role of FGF-2 as an autocrine mitogen in growth plate chondrocytes isolated from the fetal sheep(40). Receptor species other than FGFR1 are also likely to mediate FGF-2 signaling within the growth plate, because achondroplasia in man has been shown to involve specific mutations within the transmembrane region of FGFR3(41).

The presence of FGF-2 and its receptor in human fetal lung is consistent with a previous report of immunoreactive FGF-2 in fetal rat lung(42). A morphogenic role for the FGF family in lung development has been shown in studies of functional ablation of the FGFR2 gene(11). In homozygous mice, no branching of the central airway occurred, and animals died at birth without functional lungs. The localization of FGF-2 mRNA in the adrenal gland has been investigated using isolated cortical cells from the human fetus(18, 43). Exogenous FGF-2 has been shown to be mitogenic for cortical cells, whereas levels of endogenous FGF-2 mRNA are increased in response to ACTH. Within the CNS, FGF-2 has been shown to stimulate neurite outgrowth, survival and differentiated function of embryonic neurons(44–46). FGF-2 has also been shown to regulate the growth and differentiation of astrocytes and oligodendrocytes(47, 48).

In summary, the widespread but anatomically specific expression of FGF-2 and FGFR1 in the human fetus early in the second trimester suggests a prolonged role in tissue growth and maturation well beyond embryogenesis.