Differentiating and mature neurons express the acidic fibroblast growth factor gene during chick neural development

Polypeptide growth factors of the fibroblast growth factor (FGF) family have recently been implicated in the regulation of important embryonic processes such as mesoderm induction, angiogenesis and neuronal survival (Folkman and Klagsbrun, 1987; Woodland, 1989; Barde, 1989; Risau, 1990). Acidic and basic FGF are the best-studied growth factors of the FGF family. Both polypeptides are potent mitogens for a wide variety of mesodermal and ectodermal cells (for reviews see: Burgess and Maciag, 1989; Rifkin and Moscatelli, 1989; Folkman and Klagsbrun, 1987; Thomas and Gimenez-Gallego, 1986; Risau, 1990). Like other multifunctional mitogens, they have differing biological effects on different cell types. For example, bFGF was found to stimulate the growth of many cell types but inhibited the growth of Ewing sarcoma cells (Schwei-gerer et al. 1987). Both factors have been isolated mainly from adult brain tissue but are also present in the embryonic brain (Risau et al. 1988) as well as in other adult and embryonic tissues (Folkman and Klagsbrun, 1987; Risau, 1990). Since these factors at very low concentrations stimulated endothelial cell proliferation, migration and invasion in vitro and also angiogenesis in model systems in vivo, they seemed to be the prototype of angiogenic growth factors (Mosca-telli et al. 1986; Rifkin and Moscatelli, 1989; Montesano et al. 1986; Risau, 1990). Furthermore, a number of reports have indicated that FGFs (basic FGF was used in most cases) can have neurotrophic activity in vitro (Walicke et al. 1986; Morrison et al. 1986) and in vivo (Sievers et al. 1987; Anderson et al. 1988).

Our interest in the FGFs emanated from our attempts to characterize embryonic angiogenesis factors. We have isolated soluble growth and angiogenic factors from the embryonic brain (Risau, 1986) which were subsequently identified as acidic and basic FGF (Risau et al. 1988). However, these factors are also present in the adult brain in which angiogenesis does not occur and cells have a low turnover under physiological conditions. Therefore, their activity as growth factors in the adult brain may be regulated by as yet unknown mechanisms. Our knowledge about the possible regulation of FGF activity would be much enhanced if the expression of these growth factors could be correlated with cellular proliferation or differentiation in vivo. It is thus essential to investigate the developmental pattern of gene expression and to identify the cell types that express the genes. Previously, those studies were hampered by the fact that the FGF mRNA seemed to be unstable and expressed at a very low level (Abraham et al. 1986a; Jaye et al. 1986). For example, bFGF cDNA clones were found to be present in libraries at very low levels when compared to the quantity of the growth factor in the tissues. Furthermore, several of those clones represented unspliced transcripts suggesting that the cytoplasmic mRNA is unstable and that the protein is stored in tissues (Abraham et al. 1986b). Thus, the cellular origin and regulation of FGF gene expression remained unknown.

Here we report the cloning of the chick aFGF gene and its expression during embryonic brain development using a sensitive hybridization technique based on single-stranded DNA probes. By using this method and an optimized in situ hybridization technique, we have identified aFGF gene expression in neurons. The observation that the mRNA level dramatically increased during neuronal development suggest that aFGF may have a function in later stages of the development of the nervous system after neuronal proliferation has ceased.

Fertilized eggs were incubated for the desired period of time, and embryos were staged according to Hamburger and Hamilton, (1951). Organs were removed, frozen immediately in liquid nitrogen and stored at −80°C.

Genomic DNA was isolated according to Hermann and Frischauf, (1987). 10 μg samples were digested with various restriction enzymes, electrophoresed in 0.6% agarose gels, transferred in 0.4 N NaOH to Genescreen Plus membrane (DuPont) and hybridized overnight in a solution containing 5×SSPE, 5×Denhardt’s, 0.5% SDS at 65°C with l·3×10⁶cts min⁻¹ ml⁻¹ of ³²P-labelled human aFGF cDNA (Jaye et al. 1986). Blots were washed in 0.3×SSPE, 0.5% SDS at 65 °C and exposed at −70 °C to Fuji films for up to 6 days.

Total cytoplasmic RNA was isolated according to the method of Chomczynski and Sacchi (1987). Enrichment for poly(A)⁺-containing fractions was achieved by oligo(dT) column chromatography (Aviv and Leder, 1972). 5 μg aliquots of poly(A)⁺ RNA were electrophoresed in agarose gels containing 0.66 M formaldehyde and transferred to Hybond N membrane (Amersham) according to the manufacturer’s instructions. Hybridizations were performed overnight in 5×SSPE, 5×Denhardt’s, 0.5% SDS at 65°C with l–3× 10⁶ cts min⁻¹ ml⁻¹ of ³²P-labelled single-stranded DNA followed by high-stringency washes at 65 °C in 0.l×SSPE, 0.5% SDS. Filters were autoradiographed at −70°C on Fuji or Kodak films.

The complete nucleotide sequence of the genomic EcoRI-Xhol fragment and the chick aFGF cDNA clone was derived from systematic deletion subclones produced by the method of Lin et al. (1985). Sequencing was done on alkali-denatured plasmid templates using the Sequenase System (USB). The cDNA sequence was determined on both strands, the genomic clone was sequenced on one strand.

A partial genomic library was contructed after electroeluting (Biotrap, Schleicher and Schitll) the size fraction around 6.5 kb of an electrophoresed genomic EcoRI digest. The fragments were cloned into the lambda-gt 10 vector and 120.000 phages were screened using the human cDNA probe labelled by random hexanucleotide priming (Boehringer). One clone (GAF 1) out of 6 positives was chosen for further analysis. The inserted DNA fragment of GAF 1 as well as various subfragments were introduced into the Bluescript vector (Stratagene). A cDNA library from E15 chick brain poly(A)⁺ RNA was prepared using a Pharmacia cDNA synthesis kit. 500.000 primary plaques were analysed for aFGF containing sequences using the genomic Sacll–BomHI fragment as a hybridization probe. The insert of the positive clone CAF 1 was subcloned into the Bluescript Vector.

RNA transcribed from linearized plasmid templates containing the chick cDNA clone served as a template for reverse transcription reactions to generate single-stranded DNA. 1 μg of either sense or antisense RNA was incubated with 100 units MMLV reverse transcriptase (BRL) for 1 h at 37°C in 20 μl of a buffer containing 1.8i.u. pl^-1 RNAase inhibitor (Pharmacia), 2.5 mg ml⁻¹ random hexanucleotides (Boehringer), 100 μg ml⁻¹ actinomycin D (Boehringer), 0.5 mM dCTP, 0.5 mM d TIP, 0.5 mM dGTP, 50 mM Tris-HCl (pH 8.3), 75 mM KC1, 3mM MgCl₂, 10mM DTT and 100 μCi ³⁵S-dATP (for in situ hybridization) or 100 μCi ³²P-dATP (for northern hybridization). Reactions were stopped and the RNA template was hydrolyzed by incubation in 0.2 N NaOH at 65°C for 10min. After neutralization with 1 vol 0.4N HCL, 1/50 vol 1 M Tris–HCl (pH 7.5), unincorporated nucleotides were removed by Biogel P10 spun column chromatography. The probe was precipitated with ethanol and redissolved in the desired volume of hybridization buffer.

Unfixed chick brains or whole embryos were embedded in Tissue Tek (Miles), frozen and sectioned on a Leitz Cryostat. 10 pm sections were mounted on organosilane-treated slides (Uhl, 1986), dried at 55°C and postfixed in 4% paraformaldehyde in PBS for 15 min at room temperature. After two washes in PBS for 5min each, sections were dehydrated in increasing ethanol concentrations and stored desiccated at −80 °C.

Sections were brought to room temperature, treated with 40 μg ml⁻¹ predigested Pronase (Sigma) in 50 mM Tris-HCl, 5mM EDTA, pH8 for 10 min, incubated for 30 s in 0.2% glycine in PBS, washed in PBS for 1 min and fixed in 4% paraformaldehyde in PBS for 15 min at room temperature. Acetylation was performed by immersion in fresh 100 mM ethanolamine and 25 mM acetic anhydride. After rinsing in PBS for 5 min, each section was covered with 25 pl hybridization buffer containing 2xSSPE, 2 mg ml−¹ tRNA, 2 mg ml−¹ BSA, 50% formamide, 400 pg ml^-1 polyadenylic acid, 10 mM DTT, 100 μM SdATP, 500 μg ml^-1 heparin and prehybridized for Ih at 37°C. For hybridization, sections were incubated with l–2×10⁶ctsmin^-i of ³⁵S-labelled single-stranded antisense DNA in the above buffer for 6h at 37°C. On each slide a control section was hybridized with labelled sense DNA under identical conditions. Sections were then washed overnight in 2×SSPE, 50% formamide, 20 MM β-mercapto-ethanol and dehydrated. Autoradiography was performed using Kodak NTB2 emulsion. Exposure times varied from 8 days to 30 days. Finally, sections were stained with 1 % fast green, 0.5% cresyl violet or with 0.02% toluidine blue and mounted. Conditions employed for in situ hybridization were optimized using a random hexanucleotide primed chicken actin probe (Kost et al. 1983). Variables that were found significantly to affect the signal-to-noise ratio were protease treatment, temperature of the hybridization and washing steps. In contrast, pretreatment of sections with HCI and/or hot SSC did not increase the hybridization signal and were therefore omitted. Hybridization times of 6h were found to be sufficient to detect even the low abundant FGF message and, in addition, led to reduced background levels.

Human aFGF cDNA hybridized to a single 6.5 kb EcoRI fragment of chick genomic DNA detected under high-stringency conditions (Fig. 1A). A partial genomic library containing DNA fragments of corresponding size was constructed in a lambda gtlO vector and screened using the human cDNA probe. Several clones were identified. One of these, GAF1 (Fig. IB), was further analyzed. Restriction and Southern analysis showed that only the 1 kb EcoRI-XhoI fragment indicated in Fig. IB contained sequences homologous to human aFGF. The nucleotide sequence of this fragment (Fig. 2) revealed an open reading frame with high homology to the N-terminal sequence of human acidic FGF. The predicted amino acid sequence matched our previously published N-terminal amino acid sequence (Risau et al. 1988). The exon-intron structure of the presumptive first exon of the human aFGF gene was also conserved in the chick genome (Fig. 2). Typical eucaryotic promotor elements (CAAT and TATA; boxed in Fig. 2) were present in the 5′ region of the ATG codon. Thus, we believe that this clone represents part of the chick aFGF gene.

The SaclI-ZfamHI fragment of clone GAF1 (Fig. 2) was used to isolate a corresponding cDNA clone from a cDNA library prepared from embryonic day 15 chick brain mRNA (preliminary northern hybridizations had shown that aFGF mRNA was expressed at this developmental stage, see Fig. 4). Only one phage plaque out of 500000 gave a positive hybridization signal. The nucleotide sequence of this clone (CAF1) is shown in Fig. 2. At its 5′ end, the first 42 nucleotides diverged from the genomic sequence. We cannot exclude at present that this is due to a cloning artefact. Translation is probably initiated by the ATG (underlined in Fig. 2) because an in-frame stop codon (TGA; also conserved in the human sequence) is present preceding this ATG. The 1419 nucleotides of the CAF1 clone constitute the entire coding region and a long 3′ noncoding region. Since longer transcripts are detected in northern blots (Fig. 4), CAF1 does not represent a full-length cDNA clone. There are 16, 19 and 16 predicted amino acid exchanges (out of 155) in the chick sequence compared to the human, bovine and rat sequence, respectively (Fig. 3). Therefore, this gene is highly conserved during evolution.

Chick aFGF cDNA labelled by nick translation or hexanucleotide primers resulted in very weak hybridization signals in northern blots of chick brain poly(A)⁺ RNA. This is consistent with previous observations indicating a low level of expression of this gene (Jaye et al. 1986). The more sensitive riboprobes were found to hybridize unspecifically to 28S ribosomal RNA. Therefore, we used single-stranded DNA probes generated by reverse transcription using hexanucleotide primers and in vitro synthesized RNA as a template. The method allows the detection of less than 1 pg of aFGF mRNA as determined by dot–blot analysis of poly(A)⁺RNA (not shown).

Fig. 4 shows that a 1.7 kb mRNA hybridizing to the chick aFGF cDNA is present in the embryonic chick brain. It was already detectable after 3.5 days of embryonic development and dramatically increased relative to actin mRNA (not shown). High levels of mRNA were found in the adult brain. Upon longer exposure times (Fig. 4) at least three additional mRNA species of about 2.7, 3.2 and 3.4 kb could be detected. It is presently unknown whether the additional transcripts represent precursors.

Initial attempts using established protocols for in situ hybridization of Drosophila or mouse tissue sections were not successful. By using antisense cDNA probes of high specific activity and optimizing hybridization conditions, specific hybridization patterns as compared to sense cDNA probes were observed (Figs 6–10). In sagittal sections of the adult chick brain, as schematically indicated in Fig. 5, the highest level of aFGF mRNA was found in the Purkinje cells of the cerebellum (Fig. 6A,B), deep cerebellar neurons (dentate nucleus; Fig. 6C) and neurons of the brainstem (Fig. 6D). No specific hybridization was detectable in neurons of the molecular layer and granular layer. A lower level of aFGF expression was found throughout the telencephalon (Figs6B; 8A,B), particularly in a dorsorostral region which overlaps with the hyperstriatum accessorium (van Tienhoven and Juhasz, 1962). Figs 7A–D show higher magnifications of particular regions indicated in Fig. 6 to reveal the hybridization signals at the cellular level. A high density of grains is found over the large neuronal perikarya whereas cells of adjacent blood vessels were essentially devoid of grains. The majority of glial cells appear to be unlabeled. To examine the anatomical regions of aFGF expression in more detail hybridizations were performed on frontal serial sections of the adult brain (Fig. 8). The results are consistent with the observation that aFGF mRNA is present in the majority of telencephalic neurons of the adult brain.

During brain development, aFGF mRNA is present in deep cerebellar neurons of the dentate nucleus at embryonic day 15 whereas much weaker signals are detected in the molecular layer (Fig. 9A). At this developmental stage, the Purkinje cells are present within this layer and probably give rise to this signal. A comparison of identified brain regions like the hyperstriatum accessorium region (referred to above) revealed developmental changes in aFGF gene expression during neuronal development. At embryonic day 12, the pattern of aFGF expression in this region is already similar to that observed in the adult (Fig. 6B) whereas at day 9 expression was restricted to neurons localized at the periphery (Fig. 9C).

As illustrated in Figs 6-9 endothelial cells and smooth muscle cells of blood vessels (cortical capillaries, veins and pial vessels; Figs 6, 7) as well as ependymal cells (of the lateral ventricles; Fig. 8A) and choroid plexus epithelial and endothelial cells (Fig. 8B) did not express aFGF mRNA at a detectable level.

In situ hybridizations of earlier developmental stages (days 6 and 9) were performed on sections of entire embryos. We therefore could also examine aFGF expression outside the central nervous system. Our observations suggest that at these stages aFGF mRNA is predominantly, if not exclusively, expressed in the central and peripheral nervous system. We also observed significant hybridization associated with other organs (e.g. stomach, gut). Preliminary analyses indicate that this is due to aFGF gene expression in nerve cell complexes that innervate these organs (e.g. ganglion coeliacum, ganglion mesentericum). As an example for the expression in the peripheral nervous system cervical dorsal root ganglia (DRG) are shown in Fig. 10. At embryonic day 6, no specific expression was observed (Fig. 10A). Expression commenced at day 9 and increased until day 12 (Fig. 10C). The grains were concentrated over the large cell bodies of neurons (Fig. 11A,B). The increase in expression in peripheral neurons is consistent with the changes observed in the central nervous system.

Acidic and basic FGF isolated from tissues or cells are potent mitogens for a wide variety of cells. They have been implicated in a number of important physiological processes such as angiogenesis, mesoderm induction, neuronal survival and wound healing (Folkman and Klagsbrun, 1987; Woodland, 1989; Barde, 1989; Risau, 1990). Expression of bFGF mRNA has been detected in the Xenopus oocyte and embryo when mesoderm induction occurs (Kimelman and Kirschner, 1987). Endothelial cel] growth factor activity and angiogenesis-inducing activity have been characterized from tumor and embryonic tissues and identified as bFGF and aFGF (Shing et al. 1984; Libermann et al. 1987; Folkman and Klagsbrun, 1987; Risau et al. 1988). The FGFs also supported the in vitro survival of embryonic neurons (Walicke et al. 1986; Morrison et al. 1986). These results supported the view that the FGFs may have a role as growth and inducing factors. However, direct evidence for such a role is lacking because the biological roles of the FGFs have been determined mainly using in vitro assays. Only a few experiments have been performed to try to interfere with proposed functions of the FGFs in vivo.

In vivo studies intended to interfere with FGF functions can only be reasonably well designed if the cellular origin and level of expression are known. This has been impossible in the past, probably due to the low abundance of FGF mRNAs. Our results presented here demonstrate that the mRNA level of the 1.7 kb aFGF transcript increased during embryonic development and was high in the adult brain. These results were consistent with our previous results showing an increase in mitogenic activity during embryonic chick brain development (Risau, 1986). Furthermore, our in situ hybridization results reveal these developmental changes at the cellular level. Cortical and peripheral neurons were found to have very little aFGF mRNA in early stages of neurogenesis but the level of expression increased dramatically in later stages of neuronal development. Before embryonic day 9, the level of aFGF gene expression in the central nervous system was very low as determined by northern blots. At day 6, we did observe a low level of specific in situ hybridization in the anterior part of the telencephalon (not shown). Therefore, we can conclude that the level of aFGF mRNA is low between 3 and 9 days of embryonic brain development and that there is no switch in the cell type that expresses the gene during development. Studies are in progress using polymerase chain reaction techniques to determine the level of aFGF gene expression in earlier stages of embryonic development. In any case, the pattern of aFGF gene expression does not correlate with endothelial cell proliferation and angiogenesis in the brain because these processes do occur in the embryonic but not adult brain. However, it is possible that other mechanisms regulate or inhibit aFGF activity during later stages of development (see below). Our results indicate that the abundance of aFGF in the brain is due to local synthesis rather than to transport of the protein via blood or neuronal processes.

In the chick CNS, neurons are postmitotic after 4 to 9 days of embryonic development (Fujita, 1964; Wilson, 1978; Jacobson, 1978). The production of Purkinje cells in the cerebellum ceases already after 2 days while dorsal root ganglion neurons are postmitotic after 7 days (Jacobson, 1978; Hanaway, 1967; Hamburger and Levi-Montalcini, 1949). Our data on the temporal and spatial pattern of aFGF expression indicate that aFGF gene expression is initiated in postmitotic neurons and raise the possibility that aFGF has a role in neuronal function such as determination of neurotransmitter phenotype, axonal growth or synaptogenesis after neuroblast proliferation has ceased. Another member of the family of FGF-related genes, int-2, was localized to mouse Purkinje cells. Unlike the aFGF gene int-2 gene expression appeared to be restricted to early stages of Purkinje cell differentiation (Wilkinson et al. 1989).

It is noteworthy that not all adult neurons expressed the gene. This is most strikingly seen in the cerebellum where neurons of the molecular and granular layers are negative. Further work is necessary to analyze aFGF gene expression in specific sets of neurons. Interestingly, in vitro experiments indicate that FGFs can exert their neurotrophic activity on neurons from different brain regions (hippocampus: Walicke et al. 1986; cerebral cortex: Morrison et al. 1986; Walicke, 1988; striatum, septum and thalamus: Walicke, 1988; cerebellum: Hatten et al. 1988; ciliary ganglia and spinal cord: Unsicker et al. 1987; retina: Lipton et al. 1988) which is consistent with the widespread expression of aFGF in the nervous system.

Cells of the vascular wall (e.g. endothelial cells and smooth muscle cells), ependymal cells and most cortical glial cells did not express detectable amounts of aFGF mRNA. Interestingly, aFGF expression has been found in cultured smooth muscle cells and glial cells (Winkles el al. 1987; Ferrara et al. 1988). In the brain in vivo, neither cell type was found to express this gene suggesting that aFGF mRNA is upregulated upon in vitro culture conditions.

Recently, immunolocalization studies have demonstrated that aFGF protein is present in neurons (Huang et al. 1987; and our unpublished results). bFGF has also been localized in neurons by immunohistology (Janet et al. 1987) and in situ hybridization (Emoto et al. 1989). It will be interesting to investigate whether aFGF and bFGF are coexpressed in certain neurons.

A major problem concerning the physiological role of the aFGF and bFGF proteins is their lack of a hydrophobic signal sequence required for the secretion via the classical secretory pathway. The lack of a hydrophobic signal sequence is shared by human interleukin-1, which is approximately 30% homologous to aFGF (Abraham et al. 1986a; Gimenez-Gallego et al. 1985), and the platelet-derived endothelial cell growth factor (PD–ECGF) which is not significantly homologous (Ishikawa el al. 1989). In contrast to the FGFs, interleukin-1 has been found to be released by cultured cells (Hazuda et al. 1988). Thus, it is possible that the FGFs may be sequestered inside the cell and may not have direct access to target cells (Abraham et al. 1986a; Jaye et al. 1986). Alternatively, they may be released by other mechanisms, or may only be released after cell death following tissue injury (wounding, tumors, inflammation, ischemic damage) or programmed cell death. Programmed neuronal cell death indeed occurs during the development of the nervous system and may lead to the release of aFGF. It may then act as an angiogenic factor and/or neurotrophic factor. Alternatively, aFGF may act as a growth factor for other cells, particularly glial cells which are known to be mitogeni-cally stimulated by aFGF (Pettmann et al. 1985). However, this activity as well as the angiogenic activity have to be controlled in the adult brain to prevent aberrant cell proliferation. It is conceivable that in the adult the FGFs may act as general ‘rescue’ factors (including angiogenic and neurotrophic activity) released by damaged cells after lesioning as has been discussed previously (Barde, 1989) or may act as a physiological factor if released in limiting .amounts by intact cells.

We thank Herbert JSckle and his group for support in the initial stages of this project and Adriano Aguzzi. Yves-Alain Barde, Georg Breier, Bastian Hengerer, Georg Kreutzberg, Hermann Rohrer, Hans Thoenen and Erwin Wagner for helpful suggestions and critically reading the manuscript. We also thank Michael Jaye and Thomas Maciag for the human aFGF cDNA clone. This work was supported in part by the Bundesministerium fiir Forschung und Technologie.