Stimulatory effects of basic fibroblast growth factor on DNA synthesis in the human embryonic cornea

Little is known about how growth factors influence pivotal events during embryogenesis. Since the mam malian fetus is secluded in the womb, it is unavailable for physical examination and experimental manipu lation. Thus at present, the environment in which the fetus lives and the requirements of its constitutive tissues for growth and differentiation are poorly under stood (Engstrom and Heath, 1988; Hill et al. 1987). However, evidence has recently been accumulating that polypeptide growth factors might be central to these processes by virtue of their pleiotropic effects on growth, differentiation and related processes.

Development of the mammalian eye is a complex process of regulatory interactions between a wide variety of cell types of disparate embryological origins, and little is known of the factors that affect the rate of cell division, migration and differentiation during its ontogeny (Hay, 1980). The corneal endothelium plays a central role in development and turgidity of the cornea by pumping water from the corneal stroma to the aqueous humour. Corneal development begins in the human when the lens detaches from the overlying ectoderm at approximately day 32. Six days later a thin primitive stroma forms between the corneal epithelium and the lens, and this is followed by three successive waves of mesenchymal cell migration between the lens and the ectodermal stroma to form the inner endo thelial cell monolayer.

Whilst the adult corneal endothelial cells in some species are able to regenerate after wounding by cell proliferation, others, notably primates, cannot, sugges ting that differences arise during development in the environment of the cornea or in the responsiveness of the cells themselves (van Horn and Hyndiuk, 1975). Several factors that might limit endothelial cell prolifer ation have been proposed on the basis of cell culture experiments, particularly the fibroblast growth factor family (see Baird and Walicke, 1989, for review). However, much of this experimental data was obtained from nonprimate adult corneas, which might have limited relevance to the problem we wish to address. We present evidence that basic fibroblast growth factor is not only present in the human fetal cornea, but that the cells of the endothelial monolayer respond to bFGF in a concentration dependent manner.

The primary material used in this study was received from 10—12 week old human embryos obtained from elective therapeutic terminations. No apparently malformed material was used as judged by morphological examination. The fetal age post-fertilization was in each case estimated according to Shi et al. (1985). All material was processed within 2h after surgery. The fetal specimens delivered in collection vessels were initially diluted with an approximately equal volume of dextrose saline (4 % dextrose (w/v), in 0.18 % (w/v) aqueous NaCl, purchased from Steriflex PLC Nottingham, UK.) and sieved through a domestic plastic sieve with a hole size of lmmxlmm, after which blood and debris were washed through the sieve with two rinses of approximately 200 ml of phosphate-buffered saline (PBS) lacking calcium and mag nesium ions at pH7.3 (solution A of Dulbecco and Vogt, 1954) and obtained from Oxoid Ltd. UK). The contents of the sieve were then shaken and rinsed into a 37 cm × 27 cm tray containing approximately 1000 ml of PBS. Despite several transfers from one vessel to another intact organs were frequently found. We collected 84 intact eye globes from 129 samples containing embryonic material.

In one set of experiments we combined intact organs from several embryos. We also subjected thoroughly rinsed eye globes from which the connective tissue had been removed, to microdissection. A knife with a half circle cutting tip was used to penetrate the sclera at the superior limbus. The incision was then completed to separate the cornea (anterior eye) from the remainder of the eye globe (posterior eye). The organs were quickly homogenised by an ultra turrax high speed homogeniser in aqueous 4M guanidinium thiocyanate (Fluka) with 5 mM sodium citrate, 0. 1 M β-mercaptoethanol and 0.5 % sarkosyl. Dissolved samples were either processed immediately or stored at –20°C. Samples were layered over a 0.5 ml cushion of aqueous 5.7M CsCl with 0.1M EDTA (diaminoethane tetraacetic acid) at pH 8 in 3 ml capacity tubes (Beckman). They were centrifuged in a TLA 100 rotor at 100000 revs min⁻¹ for 20 h at 15°C. The pellets were dissolved in 100 μl of aqueous 10 mM Tris base with 1 mm EDTA at pH 7.4 (TE buffer). Next they were ethanol precipitated twice, redissolved again in 100 μl TE buffer. Polyadenylylated RNA was subsequently extracted by the use of oligo (dT)-cellulose as described in detail by Maniatis et al. (1982).

The cDNA probe for human basic FGF (Abraham et al. 1986) was donated by Dr Judith Abraham, California Biotech nology Inc. USA, and the probe for murine glyceraldehyde 3 phosphate dehydrogenase by Dr Peter Curtis, Wistar Insti tute, Philadelphia, USA. The probes were labelled with ³²P-dATP (NEN, DuPont) by random hexanucleotide priming as described by Feinberg and Vogelstein, (1984).

4–10 μg of polyadenylylated RNA was denatured for 15 min at 60 °C. The 20 μl final sample volume contained 20 mM MOPS (3-(N-morpholino) propane sulphonic acid, 5MM sodium acetate, ImM EDTApH7.0 (1×MOPS with 50% deionized formamide and 5% (v/v) formaldehyde. Before gel loading, 2μl of 6× loading buffer was added. This contained aqueous 15 % (w/v) Ficoll 400, 0.25 % (w/v) bromophenol blue, and 0.25 % (w/v) xylene cyanol. The samples were run on a 1 % (w/v) denaturing agarose gel containing 1–MOPS buffer and 0.23 M formadehyde. Gels sized 11×14 cm were run for 6–10 h at 20mA and the buffer was 1×MOPS with 23.5% formal dehyde (v/v). RNA was transferred to nitrocellulose filters (Schleicher and Scull, BA 85, 0.45 pore size), by blotting overnight in 20×SSC (l×SSC=aq. 0.15M NaCl, 0.015M sodium citrate, pH7). The blots were washed in 1 mM EDTA at pH 7.0. air dried, baked for 4h at 80°C and prehybridised overnight with 250 μg ml⁻¹ of sonicated and denatured salmon sperm BNA in 5xSSC, 50 mM phosphate buffer at pH 6.8, 5xDenhardt’s solution, 0.1 % (w/v) sodium dodecyl sulphate and 50% deionised formamide. They were hybridised with the labelled probes for at least 48 h at 42°C with a final radioactive concentration of l–3×10⁶ctsmin ⁻¹ ml⁻¹. The hybridising buffer was 5xSSC, 25 mM phosphate buffer, pH6.8, 2.5×Denhardt’s solution, 0.1% SDS, 50% forma mide, 50μml“’ sonicated salmon sperm DNA. The probe was alkali denatured before adding and filters were washed at high stringency with a final wash of 20 min at 55°C in O.lxSSC. Radioactivity was detected with preflashed X-Ray film (Kodak X-Omat S, Laskey and Mills, 1977).

For histology, the organs were fixed between 1 and 2 h after fetal aspiration. Routine histology was performed on formol saline-fixed material, which was subsequently embedded in paraffin and cut into 5 gm thick sections. The sections were stained in haematoxylin/eosin and examined and photo graphed through a Leitz inverted microscope with an attached camera system. For antigen localisation, fetal organs were snap frozen in liquid nitrogen and cut with cryostat into 10 gm thick sections. These sections were attached to Hendley Essex multispot glass slides precoated with 1 % (w/v) gelatin and 0.1% (w/v) chrome alum, in Analar water, air dried and stored at —70°C until use.

A polyclonal antiserum raised against a fusion protein be tween β-galactosidase and human basic fibroblast growth factor was a kind gift from Dr Judith Abraham, California Biotechnology Inc. USA. Slides with sections of human eyes were rinsed briefly in PBS at room temperature. To determine whether bFGF protein was present, sections were incubated with a blocking solution of 0.1% bovine serum albumin for 30 min at room temperature and then with several different dilutions of bFGF antiserum. The optimal dilution used was 1:1000 incubated for Ih. Detection was carried out using a second layer of goat anti-rabbit antibody (Miles, UK) diluted 1:50 conjugated to Texas Red. To ensure that our first antibody was specific for bFGF it was preincubated with recombinant bFGF. This treatment eliminated binding to the sections. The slides were examined in a Zeiss photomicro scope using epifluoresence illumination. To produce photo graphs Kodak Ektachrome EES 800 ASA films were used.

Basic fibroblast growth factor was purchased from R&D systems Inc, via British Biotechnology (UK). Alpha modified Eagles medium (Alpha-MEM) (Morton, 1970) was obtained as dry powder from Flow laboratories (UK) and made up according to the manufacturers instructions. All tissue culture plastics were obtained from NUNC (Denmark) through Gibco (UK). Fetal calf serum was purchased from Sera Lab. (UK) and trypsin from DIFCO (UK).

Thoroughly rinsed eye bulbs from which the connective tissue had been removed were subjected to microdissection. A knife with a half circle cutting tip was used to penetrate the sclera at the superior limbus (Fig. 1). The completed incision was approximately 120° in circumference around the cornea. Using this limited wounding procedure, the eye globe was maintained in a reasonably intact state and there was free passage between the external environment and the anterior chamber. Furthermore, this procedure maintained an intact anatomy of the bulb which facilitated the histological orien tation of the corneal layers. The eye bulbs were thereafter subjected to organ culture in 1cm² dishes of NUNC 24-well plates containing 1.5 ml of alpha-MEM supplemented with 50μgml⁻¹ of streptomycin sulphate, 50 units ml⁻¹ of benzyl penicillin, as well as growth factors or fetal calf serum as described in figure legends. The penetrated eye bulbs were placed with the incised area up (Fig. 1). The organ cultures were incubated in a humidified 5% CC>2/95% (v/v) air mixture at 37°C for 24 h.

At the end of the culture period, each eye globe was fixed in formal saline for at least 24 h. The preparations were then dehydrated through a graded series of alcohols, passed through toluene and embedded in paraffin wax (melting point 56°C). The exact orientation of the cornea was noted, and then each paraffin block was cut into 5 /un thick sections using a Leitz microtome. The sections were then either stained in haematoxylin/eosin or processed for autoradiography as described in Hyldahl et al. (1986). All slides were examined and photographed in a Leitz inverted microscope with an attached camera system using Ilford HP4 400ASA film.

DNA synthesis was assayed by labelling the eye globes in organ culture with 50μCi [³H]thymidine (Amersham, 56 Ci mmol⁻¹) per ml medium for 24h prior to fixation. After processing and sectioning as described above, non-incorporated radioactive thymidine was removed by treating the slides with ice-cold 10% (w/v) trichloroacetic acid for 10 min. The preparations were then washed in tap water for 10 min and finally air dried in a dust-free desiccator. To coat the slides, equal volumes of emulsion (Ilford, K2) and AnalR water were thoroughly mixed at 45°C. The slides were coated in a glass slide dipping device and left to dry on a vertical rack. The autoradiographs were dried overnight before being placed in a sandwich box containing desiccant at 4°C. The slides were stored for 4–8 weeks. Prior to development, the autoradio graphs were allowed to equilibrate to room temperature for 2–3 h. The preparations were then developed (8 min Ilford phenisol developer diluted 1:4 (v/v) with AnalR water) fixed (10 min Ilford IF23 paper fixer diluted 1:1 (v/v) with AnalR water) washed extensively in running water, air dried and finally stained in Giemsa. Only sections from eye globes in which a) the anatomy of the eye was intact (Hyldahl, 1986), b) the endothelial cell monolayer was undamaged and c) the endothelial cell layer, the corneal stroma and the posterior eye contained clearly labelled cells were included in this study. On average, one globe out of six fulfilled these criteria. The proportion of labelled cells, in the endothelial cell monolayer was determined by counting the percentage labelled cells in each section available in each experiment. All figures were based on counting at least 200 cells.

In the first set of experiments, eye globes were incised and placed in alpha-MEM with 10 % fetal calf serum for 24 h. Entry into S phase was assayed in the different corneal cell layers by labelling the entire eye globe with [³H]thymidine. Cross sections from the eye globes were subjected to autoradiography so that the proportion of each corneal cell type that had entered S-phase during the experimental period could be determined. Fig. 2 shows sections from an eye globe maintained for 24 h in organ culture. It was found that all three corneal cell layers contained labelled as well as unlabelled cells. If the autoradiographs were stained in Giemsa, the pro portion of [³H]thymidine labelled cells could be deter mined by light microscopy.

Table 1 demonstrates the effects of macromolecular supplements on DNA synthesis in human embryonic corneal endothelial cells in organ culture. The labelling index of the endothelial monolayer after 24 h mainten ance in alpha-MEM without serum was 16 %, whereas the labelling index after 24 h exposure to 10% serum was 31%. A similar high labelling index (27%) was observed if the eye globes were exposed to a medium supplemented with 20 ng basic fibroblast growth factor per ml alone. The response of the corneal endothelial cells was thereafter examined in greater detail. As shown in Fig. 3, the percentage labelled cells was maximal at 2–20ng bFGFml⁻¹ (25.9–27. 1% labelled cells) and thereafter declined. When 2μgml bFGF was added, no significant stimulatory effect was achieved.

It is of interest to examine whether the bFGF-derived effects on endothelial cell proliferation are paralleled by activation of the bFGF gene in vivo. Human fetal eyes taken from first trimester terminations were pro cessed as described in Materials and methods and polyadenylylated mRNA prepared from pooled intact eyes. When this RNA was hybridised with the bFGF cDNA, we observed one transcript of 7.5 kb (Fig. 4A). The size of bFGF mRNA detected is similar to that observed in cells from human embryonic tumours grown in tissue culture (Schofield el al. 1990).

We also examined whether mRNA for bFGF could be detected in any other organ in the first trimester human embryo. It appears that FGF transcripts are absent or below the detection level in yolk sack, central nervous system, kidney, adrenals and gut (Fig. 4A). Nor could we detect any RNA species that hybridised with our bFGF probe in any other human fetal RNA examined (data not shown).

We thereafter attempted to study the spatial distri bution of bFGF transcripts in the human fetal eye. Intact eye globes were dissected into anterior and posterior halves as described in Materials and methods, the anterior containing the whole cornea and limbus and the posterior the remainder of the eye. Poly(A)⁺ RNA was prepared from both segments and the bFGF gene expression examined. Northern blot analysis re vealed that the gene is transcribed in both the anterior and posterior portions of the eye (Fig. 4B). The bFGF mRNA was, however, found to be more abundant in the posterior eye. In both cases a single RNA species of 7.5 kb was detected. The accuracy of the quantitation was controlled by stripping and reprobing the filter with a murine glyceraldehyde 3 phosphate dehydrogenase probe.

Our finding that the human embryonic eye contains transcripts for bFGF made it pertinent to examine whether these transcripts were translated into protein. We found that anti-basic FGF antiserum stained the cells of the corneal endothelium most strongly (Fig. 5). The distribution of fluorescence suggesting that the epitope was distributed around the periphery of the cells, either on the membrane or in the intracellular matrix. No well-defined basement membrane has formed by this stage of development (9-10 weeks). However, a second site of localisation was found at the basement membrane of the lens epithelium, where in contrast localisation was mainly to the extracellular matrix. In the inner optic cup, the base of the cell bodies of the retinal neurons also stained strongly. In each case, preincubation with the target peptide sub stantially reduced the level of immunofluorescence, and the second antibody showed no reaction under these conditions.

This study has used a recently devised procedure to study the growth phenotype of human corneal endo thelial cells in organ culture. By maintaining the cells on their natural basement membranes and at physiological cell density, we have for the first time assessed their normal requirements for peptide growth factors. This cannot be achieved in other systems so far described because adult endothelial cells are contact inhibited and proliferation of cells in situ will not occur without wounding (e.g. Goldminz et al. 1979). It is of consider able interest why human corneal endothelial cells fail to proliferate in response to injury, in contrast to the readily induced cell divisions that occur in endothelial cells of other species. This has been taken to suggest that provision of some essential rate limiting factor is lacking in the wounded human cornea. It is therefore important to establish which factors are required to support proliferation of human corneal cells during development, a period of natural, rapid cell growth and migration. (Waring et al. 1982; van Horn and Hyndiuk, 1975).

We have found, unexpectedly, that the human cor neal endothelium responds mitogenically to bFGF in situ. This property has also been reported for cultures of bovine corneal endothelial cells on semi-defined sub strates (Giguere el al. 1982; Gospodarowicz and Green berg, 1979).

The mitogenic profile for bFGF shows a marked con centration dependence, those higher than 20 ng ml⁻¹ being less effective than lower concentrations. This profile is strongly reminiscent of the response seen to basic FGF by McAvoy and Chamberlain (1989) who showed that bFGF stimulates proliferation in rat lens epithelial expiants in a concentration-dependent man ner. However, once the maximum effect was reached at Ing bFGFml⁻¹, the effects on cell proliferation were reduced and other effects stimulated. The half maximal activities for proliferation, migration and lens fibre differentiation occurring at different concentrations of growth factor. This underlies the concentration-depen dent pleiotropic effects of bFGF, and suggests that similar processes, especially migration of endothelial cells, might be active in the cornea. This is currently the focus of further investigation.

This study suggests that the reason why human corneal endothelial cells do not regenerate after wounding might be a rate-limiting concentration of bFGF. Never theless cells in the developing endothelium do replicate and we were able to detect immunoreactive bFGF in the extracellular matrix of the cornea lens and retina, suggesting that during the fetal period bFGF is avail able to the cells of the endothelium. We did not find significant quantities of material in the retinal capillary endothelium as recently reported by Hanneken et al. (1989) in the bovine eye, and this may represent either interspecific differences or differences in the maturity of the tissue examined. Basic FGF has a strong affinity for the basement membranes of the eye, predominantly the proteoheparan sulphate component (Jeanny et al. 1987) as well as to other basement membranes (Folkman et al. 1987). Thus the availability of bFGF could be regulated locally at the target cell level.

The eye seems to be the major site of expression of the basic FGF gene as under identical conditions we failed to detect bFGF mRNA in polyadenylated RNA from other human fetal tissues. This is in keeping with the purification of apparently eye-specific factors pre viously, all of which were identified finally as either members of the heparin-binding growth factor subfam ily or bFGF (see Lobb et al. 1986). The presence of mRNA in both areas of the eye shown to contain immunoreactive material strongly suggests that local production of bFGF occurs in the cornea, and the presence of peptide does not indicate accumulation by binding from another source. Whether this local pro duction represents an autocrine or a paracrine system is outside the resolution of this study.

We propose that local stimulation of corneal endo thelial cell growth during their development might facilitate the integrated morphogenesis of the cornea, which requires the mitotic and locomotory cooperation of cells of different embryological origins to interact to form a unique optically functional and structural unit (Hay, 1980).

The authors are greatly indebted to Professor C.F. Graham FRS for generous support throughout this study. We also thank Dr ludith Abraham for provision of both the bFGF cDNA and antiserum. The financial support of the Swedish Medical Research Council, Cancerfonden, BarnCancerfon-den, Kronprinsessan Margaretas fond for synskadade, and the Cancer Research Campaign of Great Britain is much appreciated. Collaboration was facilitated by an ACE award from the CIBA foundation.