Culture medium conditioned by P19 embryonal carcinoma (EC) cells contains potent mitogenic activity which is markedly potentiated when the medium is conditioned in the presence of heparin. Fractionation of P19 medium conditioned in the presence of heparin reveals the existence of two biochemically distinct growth factor species both of which exhibit high affinity for immobilised heparin and significant activity as amphibian mesoderm-inducing agents. One of the species is recovered as a single polypeptide of apparent Mr = 15 000. This molecule is immunologically related to the protein product of the human K-FGF proto-oncogene. Transcripts derived from the murine K-FGF gene are also expressed by both differentiated and undifferentiated EC cells and embryonic stem cells. The second heparin-binding growth factor is recovered as a complex of four polypeptides, the largest of which has an apparent Mr = 17 000. This agent is immunologically and biochemically distinct from both acidic and basic fibroblast growth factor as well as K-FGF, and represents the predominant mitogenic activity in EC-cell-conditioned medium.

It has been well established in the last few years that murine embryo-derived embryonal carcinoma (EC) cells express specific growth and differentiation regulatory factors (reviewed Heath & Rees; 1985, Rizzino, 1987; Heath & Smith, 1988; Mummery et al. 1989). The identity of these agents is of interest since they are strong candidates for important regulatory factors in the control of cell proliferation and differentiation in early mammalian development.

The principal mitogenic activity expressed by EC cells is characterised by a strong affinity for immobilised heparin. Heath and Isacke (1984) purified a growth factor from PC13 EC-cell-conditioned medium, embryonal carcinoma derived growth factor (ECDGF), which was purified as a,Mr = 17 500 polypeptide with potent mitogenic effects on a variety of cell types in vitro. The expression of ECDGF was developmentally regulated, since the activity was no longer present in conditioned medium when the cells were induced to differentiate in vitro by exposure to retinoic acid (RA) (Isacke & Deller, 1983). Further optimisation of ECDGF purification techniques revealed that it exhibited strong affinity to immobilised heparin (Heath, 1987) and was therefore a candidate member of the heparin-binding family of growth factors (HBGFs), of which the prototype members are basic (bFGF) and acidic (aFGF) fibroblast growth factor (reviewed Gospodarawicz et al. 1987). Van Veggel et al. (1987) and Van Zoelen et al. (1989) have also reported the presence of a heparin-binding growth factor activity in EC-cell-conditioned medium which strongly resembles ECDGF in biochemical and biological properties. Further important evidence for a functional relationship between ECDGF and HBGFs came from the finding that, like aFGF and bFGF, ECDGF was a potent inducer of mesodermal differentiation in isolated animal pole expiants of Xenopus laevis embryos (Slack et al. 1987). This property has been found to be a distinctive feature of other members of the HBGF family (Paterno et al. 1989).

Direct structural characterisation of ECDGF by amino acid sequencing has, however, proved problematic, largely due to the difficulties in obtaining sufficient quantities of the factor from PC13 EC-cell-conditioned medium. Several lines of evidence have shown, however, that ECDGF differs significantly in structure from either aFGF or bFGF (reviewed Heath & Smith, 1988). These include failure to react with antibodies directed against either aFGF or bFGF, distinct tryptic peptide maps and, perhaps most significantly, the fact that the majority of ECDGF activity is found in a soluble form in conditioned medium, whereas aFGF and bFGF are generally found to be insoluble and tightly associated with their cellular sources (Jaye et al. 1988). It seemed probable, therefore, that ECDGF was similar, but not identical, to aFGF and bFGF and contained sequences that enabled the ECDGF protein to be processed through cellular secretory pathways. The existence of HBGF-like factors, which are secreted into the medium, has been confirmed by the identification of genes that encode proteins that are related by sequence to aFGF and bFGF, but which contain functional secretory signal sequences. In particular, a transforming oncogene has been isolated from both a human Kaposi’s sarcoma (Ks,Delli-Bovi et al. 1987) and a human stomach tumour (hst,Taira et al. 1987) by standard NIH/3T3 transfection techniques which encodes a Mr = 22 000 predicted primary translation product with significant sequence relationship to both aFGF and bFGF. This factor (hereafter termed K-FGF) has been shown to be both secreted and biologically active as a mitogen (Delli-Bovi et al. 1988) and a mesoderm-inducing factor (Paterno et al. 1989) when expressed under the control of heterologous promoters in transfected cells or synthesised in vitro.

The existence of these new, secreted members of the heparin-binding growth factor family, combined with a number of technical improvements in the purification of mitogens from EC-cell-conditioned medium prompted a re-examination of the identity of the HBGFs expressed by embryonal carcinoma and embryonic stem cells. Here we report that P19 EC cells in fact secrete two, seemingly distinct, heparin-binding growth factors. One of these factors is only revealed when conditioned medium is prepared in the presence of heparin and appears to be a proteolytically processed form of the murine K-FGF gene. The second (predominant) activity appears unrelated by immunological criteria to either aFGF, bFGF or K-FGF.

Cells, cell culture and biological assays

Murine P19 EC cells, CPI ES cells and C3H10T1/2 fibroblasts were maintained as described (Heath, 1987; Rudnicki & McBurney, 1987; Smith et al. 1988). PC13 EC cells were induced to differentiate by exposure to retinoic acid (Heath, 1987, and ES cells by withdrawal of DIA/LIF (Smith et al. 1988). Mitogenic activity was assayed by induction of [3H]thy-midine incorporation into quiescent C3H10T1/2 cells followed by liquid scintillation counting (Heath, 1987).

For the production of conditioned medium, 6000cm2 cell factories (Nunc) were inoculated with 2×107 P19 EC cells in 1·5 litres per factory DME:F12 (50:50 vol: vol) supplemented with 5 % newborn calf serum (selected batches). After culture at 37 °C for two days the medium was changed to 2 litres per factory serum-free medium ECM (Heath & Deller, 1983) supplemented with 10μgml-1 heparin (BDH). The conditioned medium was collected after a further two days culture at 37 °C and pumped through a 2μm filter to remove particulate matter prior to processing.

Protein purification

P19-conditioned medium (typically 20 litres) was fractionated according to the procedures described by Heath (1987) up to elution from heparin-affinity columns. Two changes to the published protocol were made; Tween-20 was omitted from the affinity chromatography buffers and heparin-affigel (BioRad) was employed for heparin-affinity chromatography.

Following elution from heparin–affigel, active fractions were pooled, diluted twofold with 10mM-sodium phosphate buffer (pH 7·0) and applied to a TSKgel heparin-5PW HPLC heparin-affinity column at a flow rate of 1·5 ml min-1. After sample application, the column was washed for 5 mins with buffer A (10mM-sodium phosphate/0·5M-NaCl, pH7·0) at a flow rate of 1·5 ml min-’ and eluted with a linear gradient from buffer A to buffer B (10mM-sodium phosphate/1·5M-NaCl, pH7 0) over 60min at a flow rate of l·5ml min-1. Fractions were collected at 1 min intervals. Absorbance was monitored at 220 nm and corrected by subtraction of absorbance data from an equivalent blank gradient run immediately before the sample.

Active fractions from HPLC affinity chromatography were pooled and applied to Brownlee RP-300 microbore reversephase guard cartridge (2 mm diam: 3 cm length) at a flow rate of 1ml min-1, the column was washed with 4ml 0·1% trifluoracetic acid (TFA) in water/20% acetonitrile (ACN) at a flow rate of 1 ml min- ‘, and the column eluted with a linear gradient from 20% ACN to 60% ACN in 0T% TFA over 40 min at a flow rate of 100 μl min-1. Absorbance was monitored at 220 nm and corrected by subtraction of absorbance data from an equivalent blank gradient run immediately before the sample. Fractions were collected at 1 min intervals and neutralised by the addition of 10 μl lM-ammonium carbonate to each fraction. Protein fractions were analysed by SDS–PAGE on a Phastgel system (Pharmacia) using resolving gels containing 20% acrylamide. Silver staining was performed according to the manufacturer’s instructions. Protein was quantified by amino acid analysis, assuming a relative molecular mass of 16000 for pool A material and 17500 for pool B.

Immunoblotting

100 ng of pool A and pool B polypeptides recovered after the microbore rpHPLC step, and 100 ng each of bovine brain derived aFGF and bFGF were subjected to electrophoresis in polyacrylamide gels containing SDS and electrophoretically transferred to nitrocellulose.

The filters were reacted with rabbit antibodies directed against recombinant human K-FGF (gift of Dr D. Rogers, Genetics Institute), bovine aFGF or bovine bFGF (Gift of Dr L. Gillespie, ICRF Developmental Biology Unit, Oxford). Bound antibodies were localised by subsequent reaction of the filter with biotin-conjugated goat anti-rabbit Ig, followed by avidin-conjugated alkaline phosphatase and staining for alkaline phosphatase activity according to the manufacturer’s instructions (Vectastain, Burlingame CA, USA).

Mesoderm induction in Xenopus embryos

Isolated animal pole expiants were tested for induction of differentiation as described by Slack et al. (1987). Differentiation was scored by observation of ‘elongation’ (Godsave et al. 1988) in experimental expiants, and confirmed by histological examination of sectioned samples. Biological activity was quantified by end-point dilution analysis according to Godsave et al. (1988).

Isolation of murine K-FGF genomic DNA and RNase protection mapping

A 1·7 kb fragment spanning exon 2 and 992 bp of exon 3 of the K-FGF gene was isolated by Hind III digestion of a Cosmid (COS16, gift of Drs G. Peters and C. Dickson, 1CRF, St Bartholomew’s Hospital, London, Brookes et al. 1989) containing the entire K-FGF gene-coding sequence. The fragment was subcloned into the expression vector pGem-3Z, linearised by digestion with EcoRI and used for the synthesis of cRNA in vitro using SP6 RNA polymerase. RNase protection of RNAs was performed according to Nomura et al. (1989). Total cellular RNA and poly(A)+ RNA was isolated from EC and ES cells according to Edwards et al. (1987).

Heparin potentiation of P19 EC-derived mitogenic activity

In the course of attempts to improve the yield of mitogenic activity in EC-cell-conditioned medium, two significant technical improvements were made.

Firstly, screening a variety of EC cell lines for secreted mitogenic activity revealed that, by dilution endpoint analysis of serum-free conditioned medium, P19 EC expressed approximately twofold more activity than PC13 EC (data not shown). P19 EC cells also proved in practice to be more amenable than PC13 EC to large-scale serum-free culture, which allowed larger volumes of conditioned medium to be reliably produced for purification. These findings prompted us to focus on P19 EC for subsequent characterisation of secreted mitogens.

Secondly, following the report of Delli-Bovi et al. (1988) that K-FGF protein was stabilised by the presence of heparin, we investigated the results of exposing EC cells to heparin during the course of conditioning on subsequent recovery of mitogenic activity. P19 EC cells were grown in ECM serum-free defined medium (Heath & Deller, 1983) which lacks exogenous growth factors, and the effects of adding heparin (10μg ml-1) during the conditioning period were tested by examining the induction of DNA synthesis in murine C3H/10T1/2 fibroblasts by [3H]thymidine incorporation marked increase in mitogenic activity, measured by the concentration of conditioned medium required to induce half-maximal [3H]thymidine incorporation, was observed. Significant mitogenic activity could still be observed at a 1/200 dilution of medium derived by conditioning in the presence of heparin. The addition of heparin to medium conditioned by P19 EC cells in the absence of heparin had no effect (data not shown) on its mitogenic activity and heparin alone had no mitogenic effect (Fig. 1). No significant mitogenic activity (data not shown) could be found associated with P19 EC cell extracts (in contrast to the report of Rizzino et al. 1988) suggesting that the principal mitogen(s) expressed by EC cells are efficiently secreted into the culture medium. P19 EC cells threfore express a major mitogenic activity, which is either stabilized or induced in the presence of heparin.

Fig. 1.

Potentiation of P19 conditioned medium mitogenicity by heparin. P19 ECM serum-free medium was conditioned in the presence or absence of 10 μg ml-1 heparin for 24 h and tested for the ability to induce DNA synthesis in 10T1/2 mouse fibroblasts by [3H]thymidine incorporation.

Fig. 1.

Potentiation of P19 conditioned medium mitogenicity by heparin. P19 ECM serum-free medium was conditioned in the presence or absence of 10 μg ml-1 heparin for 24 h and tested for the ability to induce DNA synthesis in 10T1/2 mouse fibroblasts by [3H]thymidine incorporation.

Purification of P19 EC-cell-derived heparin-binding growth factors

P19 EC-cell-conditioned ECM serum-free defined medium supplemented with heparin (10 μg ml−1) (P19 CM/Hep) was fractionated by the procedures developed for the purification of PC13 EC cell-derived ECDGF (Heath, 1987). This comprises batch absorption of medium on CM-sephadex, followed by high-salt elution onto phenyl sepharose, elution from phenyl sepharose with ethylene glycol, and affinity chromatography on heparin-affigel, eluting the biological activity with 1 M-NaCl. Mitogenic activity was measured by induction of [3H]thymidine incorporation into C3H/10T1/2 cells.

In accord with previous observations on PC13 EC-cell-conditioned medium (Heath & Isacke, 1984), all the detectable mitogenic activity present in P19 CM/Hep bound to CM-sephadex, was eluted by 1 M-NaCl and bound to phenyl sepharose in 50mM-phos-phate buffer (pH 7·0) in the presence of CM-sephadex eluent containing 1 M-NaCl and the 150 mM-NaCl/50mM-sodium phosphate (pH7·0) wash. A significant fraction of the total mitogenic activity (hereafter termed pool A) was eluted from phenyl sepharose by 10% ethylene glycol, 50mM-sodium phosphate (pH7·0) (Fig. 2). This activity is not observed when PC19 EC or PC13 EC cell medium conditioned in the absence of heparin is fractionated under identical conditions (Heath, 1987). The remaining activity in P19 CM/Hep was eluted from the phenyl sepharose column with 50% ethylene glycol/50mM-sodium phosphate/ 150mM-NaCl (pH 7·0) (Fig. 2) as previously reported for the major activity in PC13 EC-cell-conditioned medium (Heath, 1987). This activity is hereafter termed pool B. Pool A and pool B were separately applied to heparin–affigel affinity columns which were washed with 50mM-sodium phosphate buffer/0·5 M-NaCl (pH 7·0) and eluted with 50mM-sodium phosphate/1 M-NaCl (pH7·0). In both cases, the majority of the mitogenic activity bound to the heparin affinity column and was eluted with 1 M-NaCl (Fig. 3A,B). In the case of pool B material, a minor (approximately 5% of total) peak of mitogenic activity was recovered in the 0·5 M-NaCl wash. This activity is consistent with the expected chromatographic behaviour of the PDGF-like activity previously reported in EC-cell-conditioned medium (Rizzino & Bowen-Pope, 1985).

Fig. 2.

Fractionation of P19 CM/Hep by chromatography on phenyl sepharose. Activity eluted from CM-sephadex by 1 M-NaCl was applied directly to a phenyl sepharose column (5 ml bed volume) which was washed with 50mM-phosphate buffer (pH 7·0) and eluted successively with 10% ethylene glycol (fractions 1·9) and 50 % ethylene glycol (fractions 10·20) in phosphate buffer. 5 ml fractions were collected. Mitogenic activity in 10 μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells.

Fig. 2.

Fractionation of P19 CM/Hep by chromatography on phenyl sepharose. Activity eluted from CM-sephadex by 1 M-NaCl was applied directly to a phenyl sepharose column (5 ml bed volume) which was washed with 50mM-phosphate buffer (pH 7·0) and eluted successively with 10% ethylene glycol (fractions 1·9) and 50 % ethylene glycol (fractions 10·20) in phosphate buffer. 5 ml fractions were collected. Mitogenic activity in 10 μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells.

Fig. 3.

Heparin–affigel affinity chromatography of mitogenic activity eluted from phenyl sepharose by 10 % ethylene glycol (Fig. 3A) and 50 % ethylene glycol (Fig. 3B). After loading, the columns were washed with phosphate buffer containing 0·5 M-NaCl (fractions 1–9) and 1 M-NaCl (fractions 10-20). Mitogenic activity in 10 μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells.

Fig. 3.

Heparin–affigel affinity chromatography of mitogenic activity eluted from phenyl sepharose by 10 % ethylene glycol (Fig. 3A) and 50 % ethylene glycol (Fig. 3B). After loading, the columns were washed with phosphate buffer containing 0·5 M-NaCl (fractions 1–9) and 1 M-NaCl (fractions 10-20). Mitogenic activity in 10 μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells.

From these results, it would seem that there are two biochemically distinct heparin-binding growth factor activities present in P19CM/Hep; ‘pool B’ resembles ECDGF in its purification properties. ‘Pool A’ activity appears when medium is conditioned in the presence of heparin and exhibits lower affinity for phenyl sepharose since it can be eluted by 10% ethylene glycol. Pool A and pool B do, however, exhibit similar affinity for heparin–affigel, judged by their elution characteristics.

Pool A and pool B activities from the soft-gel heparin-affinity column were further fractionated by application to an HPLC heparin-affinity column followed by gradient elution with NaCl (Fig. 4A,B). Pool A material eluted as a discrete peak of biological activity coincident with a UV absorption peak centred at 27min (nominally 0·97 M-NaCl). The biological activity in pool B eluted as a broad peak of activity associated with a UV absorption peak centred at 20 min (nominally 0·8 M-NaCl). The activities present in pool A and pool B therefore exhibit high, but slightly different, affinities for immobilised heparin.

Fig. 4.

High-pressure heparin-affinity chromatography of mitogenic activity eluted from heparin–affigel. Mitogenic activity (blocked columns) in 10 μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells. Figure 4A shows separation of pool A material and Fig. 4B separation of pool B material.

Fig. 4.

High-pressure heparin-affinity chromatography of mitogenic activity eluted from heparin–affigel. Mitogenic activity (blocked columns) in 10 μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells. Figure 4A shows separation of pool A material and Fig. 4B separation of pool B material.

The peak activities in pool A and pool B recovered from high-pressure heparin-affinity chromatography were further analysed by microbore reverse-phase high-pressure liquid chromatography. About 90% of the biological activity was lost at this stage from both pools. This phenomenon is well documented in the case of aFGF or bFGF (e.g. Esch et al. 1985) and is thought to arise from denaturation and loss of biological activity upon exposure to acidic solvents. Pool A activity was recovered associated with a single, symmetrical and discrete UV absorption peak at a nominal 39 % acetonitrile (Fig. 5A), and pool B with a major, but broader UV absorption peak at a nominal 42 % ACN (Fig. 5B). Pool A and pool B therefore correspond to different biochemical species of heparin-binding growth factors based on their biochemical characteristics and chromatographic behaviour.

Fig. 5.

High-pressure microbore reverse-phase chromatography of mitogenic activity recovered after high-pressure heparin-affinity chromatography. Mitogenic activity (blocked columns) in 10μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells. Figure 5A shows separation of pool A material and Fig. 5B separation of pool B material. Data capture was terminated after 40 min in the fractionation of pool B.

Fig. 5.

High-pressure microbore reverse-phase chromatography of mitogenic activity recovered after high-pressure heparin-affinity chromatography. Mitogenic activity (blocked columns) in 10μl samples of each fraction was measured by induction of DNA synthesis in 10T1/2 cells. Figure 5A shows separation of pool A material and Fig. 5B separation of pool B material. Data capture was terminated after 40 min in the fractionation of pool B.

Accurate determination of the overall recoveries of these individual factors is precluded by the complexity of mitogenic activities found in P19CM/Hep, the inability to accurately measure the low protein concentrations during the later low-pressure chromatography stages and by the acid inactivation phenomenon described above. However, in several equivalent experiments, between 0·41 and 0·53 μg litre-1 of conditioned medium was found in the pool A material and 2·3 and 3·6μg litre-1 of conditioned medium in the pool B material after the microbore reverse-phase HPLC step. These quantities can, however, account for the majority of the fibroblast mitogenic activity found in P19 CM/Hep.

Peak fractions containing mitogenic activity from pool A and pool B after the rpHPLC step were subjected to SDS–PAGE followed by silver staining (Fig. 6). Pool B material contained a single protein species of apparent Mr = 15 000. Pool A contained four species of apparent Mr = 17000, 15 000, 14000 and 12000. It has not proved possible to separate these species further by rpHPLC.

Fig. 6.

SDS–PAGE of peak fractions recovered from reverse-phase fractionation. 0· μl of each fraction was mixed with an equal volume of double strength sample buffer prior to elecrophoresis. Lanes 1 & 2 are fractions 26 and 27 from pool A and lanes 3, 4 & 5 fractions 28, 29 and 30 from pool B. M = 100 ng each size markers (Mr = 14000, 21000, 24000, 29000, 35 000 and 62000 in ascending order).

Fig. 6.

SDS–PAGE of peak fractions recovered from reverse-phase fractionation. 0· μl of each fraction was mixed with an equal volume of double strength sample buffer prior to elecrophoresis. Lanes 1 & 2 are fractions 26 and 27 from pool A and lanes 3, 4 & 5 fractions 28, 29 and 30 from pool B. M = 100 ng each size markers (Mr = 14000, 21000, 24000, 29000, 35 000 and 62000 in ascending order).

Immunoblotting of Pl9-clerived heparin-binding growth factors

The relationship between the two species of heparin-binding growth factors recovered from P19CM/Hep and the already characterised agents aFGF, bFGF and K-FGF was analysed by immunoblotting the purified preparations with antibodies directed against aFGF, bFGF and K-FGF. Approximately 100 ng of pool A and pool B heparin-binding growth factors, as well as 10 ng aFGF and bFGF standards, were subjected to SDS–PAGE, transferred to nitrocellulose, and tested for reactivity with antibodies directed against either bovine bFGF (Fig. 7A,D), bovine aFGF (Fig. 7B) or human ks gene-derived K-FGF (Fig. 7C).

Fig. 7.

Immunoblotting of biologically active fractions from reverse-phase chromatography. 100 ng of pool A (track 2 A–D) and pool B (track 3 A–D) and 10 ng each of purified bovine brain-derived acidic FGF (track 4, B & D) or basic FGF (track 5, B & D) were subjected to SDS–PAGE, transferred to nitrocellulose and blotted with antibodies directed against acidic FGF, basic FGF or K-FGF as indicated. Track 1 is blank in all cases.

Fig. 7.

Immunoblotting of biologically active fractions from reverse-phase chromatography. 100 ng of pool A (track 2 A–D) and pool B (track 3 A–D) and 10 ng each of purified bovine brain-derived acidic FGF (track 4, B & D) or basic FGF (track 5, B & D) were subjected to SDS–PAGE, transferred to nitrocellulose and blotted with antibodies directed against acidic FGF, basic FGF or K-FGF as indicated. Track 1 is blank in all cases.

Staining conditions for the anti-aFGF and anti-bFGF antibodies were chosen such that 5 ng of immunoreactive protein could be detected. Pool B proteins did not react with either anti-aFGF, anti-bFGF or anti-K-FGF. Pool A material exhibited clear reactivity with anti-K-FGF, with a single reactive species of apparent Mr = 15 000, but did not react with antibodies directed against either aFGF or bFGF. No cross-reaction of the anti-K-FGF antibody was observed with the aFGF or bFGF controls in separate experiments (not shown). Some cross-reactivity of anti-aFGF with bFGF was, however, detected (compare Fig. 7B and 7D).

We conclude that pool A represents a Mr = 15 000 form of a heparin-binding growth factor which is immunologically related to human K-FGF. Pool B represents a distinct secreted heparin-binding growth factor species, which is not identical to either K-FGF, aFGF or bFGF.

Mesoderm induction properties of P19-d.erived heparin-binding growth factors

The ability to induce mesodermal differentiation in isolated animal pole expiants of Xenopus laevis embryos is a characteristic property of all heparin-binding growth factors tested to date (Slack et al. 1987; Paterno et al. 1989). We accordingly tested both pool A and pool B heparin-binding growth factors from P19 CM for their mesoderm induction properties. Both pool B and the K-FGF-like pool A factor induced mesodermal differentiation in embryo expiants. Pool A had an activity of 2·5 mesoderm-inducing units ng-1 and pool B 3·25 mesoderm-inducing units ng-1 as measured by endpoint dilution assay (Godsave et al. 1988). The relative potencies of these preparations in this assay was paralleled by their mitogenic activity in the C3H/10T1/2 asaay (data not shown). Although both fractions were tested for mesoderm induction after the rpHPLC step, the proportion of responding expiants was similar in potency to bFGF or K-FGF synthesised in vitro and greater than that found for the recombinant version of another member of the FGF family, Int-2 (Paterno et al. 1989).

We conclude that both heparin-binding growth factor species present in P19CM/Hep are active as mesoderminducing agents in Xenopus laevis embryos, a property previously reported for recombinant human K-FGF (Paterno et al. 1989) and PC13-derived ECDGF, as well as aFGF and bFGF (Slack et al. 1987). These observations provide additional support for identity of the pool A factor with K-FGF.

Expression of K-FGF-derived transcripts in murine EC and ES cells

The structural characterisation of the pool A HBGF protein described above suggested that it might represent the murine homologue of the human K-FGF protein. However, the apparent molecular mass of the protein recovered (15 000) is less than that reported for human K-FGF expressed in vitro under the control of a heterologous promoter (Delli-Bovi et al. 1988) or synthesised in vitro (Paterno et al. 1989). This could mean that either the K-FGF gene is expressed in EC cells and subjected to further proteolytic processing than in other cell types, or that the growth factor purified from P19CM/Hep is derived from the transcripts of a gene distinct from K-FGF but with which a common epi-tope(s) is shared. This issue was investigated by examining the expression of murine K-FGF transcripts in EC cells by RNase protection of a homologous probe, which enables the detection of transcripts of defined sequence identity in cellular RNAs. A subsidiary benefit of this approach is the ability to examine the expression of K-FGF transcripts in murine ES cells (which cannot be conveniently examined for the expression of HBGFs by the protein purification approach described above) and thus determine the extent to which K-FGF expression is a characteristic property of cells derived directly from pluripotential embryonic ectoderm by different routes.

A 1·7 kb HindIII fragment (Fig. 8) was subcloned from a cosmid (COS16) containing the mouse K-FGF gene into a plasmid vector containing the SP6 promoter and radiolabelled antisense RNA transcripts synthesised in vitro. This fragment spans all of exon 2 of the K-FGF gene and part of exon 3 as well as the intervening intron. Cellular transcripts with complete sequence identity over the length of exon 2 should yield a predicted protected RNA fragment of 103 nucleotides. Additional larger protected species would also be anticipated corresponding to exon 3 and RNA splicing intermediates. RNase protection of poly(A)+ and total cytoplasmic RNA from P19 and P13 EC cells and CPI ES cells yielded a protected fragment of 103 bp corresponding to exon 2 as well as larger species corresponding to predicted splicing intermediates (Fig. 9). Identical protected fragments were observed in RNA derived from differentiated ES cells produced by withdrawal of DIA/LIF, and PC13 END cells produced by retinoic acid treatment (Fig. 9). No protected fragments were found in control 3T3 cell RNA. On the basis of these observations, we conclude that both P19 EC, PC13 and CPI ES cells express transcripts with sequence identity to murine K-FGF. These transcripts persist in the immediate differentiated progeny of both EC and ES stem cells.

Fig. 8.

Restriction map of murine K-FGF gene (data taken from Brookes et al. 1989) showing origination of 1·7 kb HindIII / HindIII fragment used for RNase protection studies. Boxes indicate exons and hatched boxes indicate protein coding sequences. H = HindIII restriction sites, X = XbaI restriction site and E = EcoRI restriction sites.

Fig. 8.

Restriction map of murine K-FGF gene (data taken from Brookes et al. 1989) showing origination of 1·7 kb HindIII / HindIII fragment used for RNase protection studies. Boxes indicate exons and hatched boxes indicate protein coding sequences. H = HindIII restriction sites, X = XbaI restriction site and E = EcoRI restriction sites.

Fig. 9.

RNase protection mapping of murine K-FGF gene transcripts in murine EC and ES cells and their differentiated derivatives. Track A: size markers, track B: tRNA control, track C: CPI ES cell poly(A)+ RNA, track D: total RNA from CPI ES cells propagated in BRL cell conditioned medium, track E total RNA from CPI cells maintained in purified recombinant DIA/L1F, Track F total RNA from differentiated CPI ES cells derived by DIA/LIF withdrawal, track G: PC13 EC cell poly(A)+ RNA, track H: total RNA from PC13 EC, track 1: total RNA from PC13 END cells, track J: total RNA from P19 EC cells, track K: total RNA from Swiss 3T3 cells.

Fig. 9.

RNase protection mapping of murine K-FGF gene transcripts in murine EC and ES cells and their differentiated derivatives. Track A: size markers, track B: tRNA control, track C: CPI ES cell poly(A)+ RNA, track D: total RNA from CPI ES cells propagated in BRL cell conditioned medium, track E total RNA from CPI cells maintained in purified recombinant DIA/L1F, Track F total RNA from differentiated CPI ES cells derived by DIA/LIF withdrawal, track G: PC13 EC cell poly(A)+ RNA, track H: total RNA from PC13 EC, track 1: total RNA from PC13 END cells, track J: total RNA from P19 EC cells, track K: total RNA from Swiss 3T3 cells.

Here we report that P19 EC cells express potent mitogenic and mesoderm-inducing growth factors which are secreted into the culture medium. The mitogenic activity in P19-conditioned medium is markedly potentiated by exposure of the cells to heparin during the conditioning process. Fractionation of P19-conditioned medium reveals that these activities are predominantly due to the existence of two biochemically and immunologically distinct growth factor species, which exhibit high affinity for immobilised heparin.

One of these growth factors is only observed when the medium is conditioned in the presence of heparin and proves to be immunologically related to the protein product of the human K-FGF proto-oncogene. Furthermore, we have also shown that P19 EC cells express transcripts with sequence identity to the murine homologue of the human K-FGF gene. This provides very strong circumstantial evidence for identity between the Mr = 15 000 growth factor and the protein product of the K-FGF transcripts. Nevertheless, formal proof of this issue will require the complete amino sequence determination of the Mr = 15 000 protein. The putative K-FGF protein isolated from P19 EC cells has an apparent Mr = 15 000 whereas the predicted secreted form of the human K-FGF has an apparent Mr = 22000 and that of murine K-FGF 22000 (Delli-Bovi et al. 1987, G. Peters and C. Dickson, personal communication). Furthermore, K-FGF synthesised in vitro in either COS cells (Delli-Bovi et al. 1988) or reticulocyte lysates programmed with ks-derived cRNA is of Mr = 22000 by SDS – PAGE (Paterno et al. 1989). However, the predicted protein sequence of murine K-FGF (Brookes et al. 1989) contains three pairs of basic amino acids (positions 67/68,70/71 and 185/184). A protein of the approximate size of the molecule described here could in principle be produced by removal of the secretory signal sequence and proteolytic cleavage at the COOH-terminal site, if EC cells, unlike COS cells, expressed functional K-FGF proteolytic processing enzymes. It is important to note, however, that these paired basic amino acid residues are preserved in the predicted protein product of the K-FGF-related FGF-6 gene (Maries el al. 1989), and it is not currently known whether the anti-K-FGF antibody used in these studies would react with the putative FGF-6 protein product. The nucleotide sequence divergence between K-FGF and FGF-6 is sufficiently great to preclude the possibility of detecting FGF-6 transcripts in our RNase protection experiments.

The RNase protection analysis reveals two further aspects of K-FGF expression. Firstly, murine ES ceils express equivalent amounts of these transcripts, showing that cells that retain the pluripotential characteristics of normal pluripotent embryonic stem cells, and whose differentiation is suppressed by the action of DIA/LIF (Smith et al. 1988) also express K-FGF-related growth factors. This suggests that the expression of K-FGF by P19 and PC13 EC cells is not an adventitious phenomenon but is, in fact, a property common to many stem cell lines derived from pluripotential embryonic ectoderm by different means. Moreover, we have (unpublished observations) been unable to detect rearrangements or amplification of the K-FGF gene (which are associated with K-FGF activation in adult tumours) in EC or ES cells. K-FGF expression may therefore represent a feature of the normal embryonic origins of EC and ES cell lines, and is not obviously related to their ability to differentiate in vitro or developmental potential.

The second feature of these results is that the immediate differentiated progeny of both P19 and CPI ES cells also express K-FGF transcripts. K-FGF expression is not, therefore, a diagnostic feature of EC or ES cells and K-FGF cannot be considered to be a ‘stemcell-specific’ growth factor. The presence of K-FGF gene transcripts in differentiated, nonmalignant cells also shows that its expression does not necessarily result in growth-factor-independent proliferation or a transformed phenotype. This expression of K-FGF transcripts by differentiated EC or ES cells was somewhat unexpected in view of the disappearance of secreted potent mitogenic activities associated with PC13 EC cell differentiation (Isacke & Deller, 1983). It is important to note, however, that the K-FGF-like growth factor was only detected in EC-cell-conditioned medium in the presence of heparin and would not have been observed in these earlier experiments.

This expression of K-FGF-like growth factor and K-FGF-derived transcripts by cell types equivalent to pluripotential stem cells of the normal embryo and their early differentiated derivatives leads to the prediction that the K-FGF gene is expressed in early postimplantation stages of normal mouse development. This does not necessarily imply, however, that the protein product of the K-FGF gene is present in sufficient quantity to elicit a biological effect. Indeed the in vitro requirement for heparin to stabilise extracellular K-FGF protein suggests that the delivery of a K-FGF-derived signal in vivo may depend upon, and be modulated by, the local presence of appropriate oligosaccharides.

The second growth factor expressed by P19 EC cells is recovered as a complex of four species, the largest of which has an apparent Mr = 17 000, which proves to be biochemically and immunologically distinct from aFGF, bFGF and K-FGF. This agent represents, on the basis of quantities recovered, the predominant fibroblast mitogen in P19 CM/Hep. The exact relationship between the four protein species recovered after microbore rpHPLC is not currently clear. Some of these individual species could represent biologically inactive contaminants. However, both aFGF and bFGF can be recovered in multiple molecular forms from natural sources (Esch et al. 1985), and, as discussed above, this may also be the case for K-FGF. In addition, during purification, the biological activity associated with this factor exhibits significantly broader elution characteristics than any individual protein species within the complex. We accordingly favour the view that these individual species represent differentially processed forms of a precursor. This may be confirmed by amino acid sequence analysis of each member of the complex.

It is clear, nevertheless, that this growth factor species exhibits all the features now associated with other members of the heparin-binding growth factor family; it exhibits a high affinity for immobilised heparin and has both mitogenic and amphibian mesoderm-inducing properties. It might be concluded that this material is equivalent to ECDGF, previously identified in P13 EC-cell-conditioned medium (Heath & Isacke, 1984). Indeed, the behaviour of this agent during purification is almost identical to ECDGF, and the largest molecular species recovered is very similar to ECDGF in molecular mass. ECDGF was also found to induce mesoderm in amphibians (Slack et al. 1987). However, no evidence for the existence of lower molecular weight forms of PC13-derived ECDGF has been found. This could be either due to a difference between P19 and PC13 in their processing capacities or to the presence of heparin in the current experiments which might protect lower molecular weight species from degradation. The relationship between this factor and more recently identified members of the FGF family of growth factors remains to be determined. It would appear unlikely that this factor represents the protein product of the int-2 proto-oncogene which is known to be expressed at low levels in EC cells (Smith et al. 1988) since it is a more potent mesoderm-induction agent than the int-2 gene protein product even after partial acid inactivation and synthetic int-2 does not bind avidly to immobilised heparin (Paterno et al. 1989). It is conceivable that this factor may be derived from either the FGF-5 (Zhan et al. 1988) or FGF-6 (Maries et al. 1989) genes. This issue may again be resolved by amino acid sequencing and the ability to recover nanomolar quantities of the factor from P19 CM/Hep will be useful in this regard.

Finally, a number of general issues, raised by the results reported here, may be considered. Firstly, what is the point of expressing two growth factors with apparently overlapping biological functions from a single cell type? Whilst this may be a manifestation of some aberrant feature of P19 cells, we favour the possibility discussed elsewhere (Heath & Smith, 1989) that significant quantitative differences between the two agents may exist, such as their net hydrophobicity or sensitivity to heparin-induced stabilisation, which leads to different biological roles in vivo. Secondly, why are these particular agents expressed by EC and ES cells? The demonstrated role of FGF-like growth factors in amphibian mesoderm induction clearly suggests that these agents may be involved in inductive or cell specification processes in early mammalian development, although the identity of such processes is not, at the moment, clear. There is fairly good circumstantial evidence in amphibians that the amphibian homologue of bFGF may be the ‘natural’ mesoderm-inducing agent (Slack & Isaacs, 1989). The expression of FGF-related genes (but not aFGF or bFGF) in the early stages of mouse development may therefore, as argued above, reflect the particular requirements for delivery of inductive signals in a very different anatomical and physiological situation.

We are grateful to Gordon Peters and Clive Dickson for the donation of the COS16 mouse genomic clone and sharing of unpublished data, David Rogers for helpful suggestions and provision of the anti-K-FGF antibody, Tony Willis for amino acid analysis, Jonathan Slack for the Xenopus embryos, Laura Gillespie for anti-FGF antibodies and Maureen Mee for skilled technical support. The research was supported by the Cancer Research Campaign.

Brookes
,
S.
,
Smith
,
R.
,
Thurlow
,
J.
,
Dickson
,
C.
&
Peters
,
G.
(
1989
).
The mouse homologue of hst/k-FGF: sequence, genome organisation and location relative to int-2
.
Nucleic Acids Res.(In the press)
.
Delli Bovi
,
P.
,
Curatola
,
A.
,
Kern
,
F.
,
Greco
,
A.
,
Ittman
,
M.
&
Basilico
,
C.
(
1987
).
An oncogene isolated by transfection of Kaposis’s sarcoma DNA encodes a growth factor that is a member of the FGF family
.
Cell
50
,
729
737
.
Delli-Bovi
,
P.
,
Curatola
,
A.-M.
,
Newmark
,
K.
,
Sato
,
Y.
,
Moscatelli
,
D.
,
Hewick
,
R.
,
Rifkin
,
D.
&
Basillico
,
C.
(
1988
).
Processing, secretion and biological properties of a novel growth factor of the fibroblast growth factor family with oncogenic potential
.
Mol. Cell Biol
.
8
,
2933
2941
.
Edwards
,
D.
,
Murphy
,
G.
,
Reynolds
,
J.
,
Docherty
,
A.
,
Angel
,
P.
&
Heath
,
J.
(
1987
).
TGF-/J modulates the expression of collagenase and TIMP
.
EMBO J
.
6
,
1899
1902
.
Esch
,
F.
,
Baird
,
A.
,
Ling
,
N.
,
Ueno
,
N.
,
Hill
,
F.
,
Geurov
,
L.
,
Kleeper
,
R.
,
Gospodarawicz
,
D.
,
Bohlen
,
P.
&
Guillemin
,
R.
(
1985
).
Primary structure of bovine pituitary basic fibroblast growth factor (FGF) and comparison of N-terminal sequence with the sequence of bovine brain acidic FGF
.
Proc, natn Acad. Sci. U.S.A
.
81
,
6963
6967
.
Godsave
,
S. F.
,
Isaacs
,
H. V.
&
Slack
,
J M. W.
(
1988
).
Mesoderm inducing factors: a small class of molecules
.
Development
102
,
555
556
.
Gospodarawicz
,
D.
,
Ferrara
.
N.
,
Schweigerer
,
L.
&
Neufeld
,
G.
(
1987
).
Structural characterisation and biological functions of fibroblast growth factor
.
Endocrine Rev
.
8
,
95
114
.
Heath
,
J. K.
(
1987
).
Experimental analysis of teratocarcinoma cell multiplication and purification of embryonal carcinoma-derived growth factor
.
In Teratocarcinomas and Embryonic Stem Cells: a Practical Approach
(Ed.
E.
Robertson
).
IRL press
,
Oxford
, pp.
183
206
.
Heath
,
J. K.
&
Deller
,
M. J.
(
1983
).
Serum free culture of PC13 murine embryonal carcinoma cells
.
J. Cell Physiol
.
115
,
225
230
.
Heath
,
J. K.
&
Isacke
,
C. M.
(
1984
).
Embryonal carcinoma derived growth Factor
.
EMBO J
.
3
,
2597
2962
.
Heath
,
J. K.
&
Rees
,
A. R.
(
1985
).
Growth factors in mammalian embryogenesis
.
In Growth Factors in Biology and Medicine
(ed.
D.
Evered
and
M.
Stoker
)
Ciba symp.
Vol.
116
, pp.
1
22
.
Heath
,
J. K.
&
Smith
,
A. G.
(
1988
).
Regulatory factors of embryonic stem cells
.
J. Cell Sci. Supp
10
,
257
266
.
Heath
,
J. K.
&
Smith
,
A. G.
(
1989
).
Growth factors in embryogenesis
.
British Medical Bulletin
45
(
2
),
319
336
.
Isacke
,
C. M.
&
Deller
,
M. J.
(
1983
).
Teratocarcinoma cells exhibit growth cooperativity in vitro
.
J. Cell Physiol
.
117
,
407
414
.
Jaye
,
M.
,
Lyall
,
R.
,
Mudd
,
R.
&
Schlessinger
,
J.
(
1988
).
Expression of acidic fibroblast growth factor cDNA confers growth advantage and tumourgenesis to swiss 3T3 cells
.
EMBO J
.
7
,
963
970
.
Marics
,
L
,
Adelaide
,
J.
,
Raybaud
,
F.
,
Mattei
,
M.-G.
,
Coulier
,
F.
,
Planche
,
J.
,
Delapeyrier
,
O.
&
Birnbaum
,
D.
(
1989
).
Characterisation of HST-related FGF.6 gene, a new member of the fibroblast growth factor gene family
.
Oncogene
4
,
335
340
.
Mummery
,
C.
,
Van Den Eijnden-Van Raaj
,
J.
,
Feuen
,
A.
,
Tsung
,
H.C.
&
Kruijer
,
W.
(
1989
).
Regulation of growth factors and their receptors in early murine embryogenesis
.
In Cell to Cell Signals in Mammalian Development
. (Eds
De Laat
,
S.
,
Bluemink
,
J.
and
Mummery
,
C.
)
NATO ASI series H. Springer-Verlag Berlin and Heidelberg. Vol H26
, pp.
231
246
.
Nomura
,
S.
,
Hogan
,
B. L. M.
,
Wills
,
A. J.
&
Heath
,
J. K.
&
Edwards
,
D. R.
(
1989
).
Developmental expression of tissue inhibitor of metalloproteinase (TIMP) RNA
.
Development
105
,
575
583
.
Paterno
,
G. D.
,
Gillespie
,
L.
,
Dixon
,
M.
,
Slack
,
J. M. W.
&
Heath
,
J. K.
(
1989
).
Mesoderm-inducing properties of Int-2 and kFGF: two oncogene encoded growth factors related to FGF
.
Development
106
,
79
84
.
Rizzino
,
A.
(
1987
).
Defining the roles of growth factors during early mammalian development
.
In The Mammalian Preimplantation Embryo: Regulation of Growth and Differentiation in vitro
(ed.
B. D.
Bavister
).
Plenum press
,
New York and London
pp.
151
174
.
Rizzino
,
A.
&
Bowen-Pope
,
D. F.
(
1985
).
Production of PDGF-like growth factors by embryonal carcinoma cells and binding of PDGF to their endoderm-like differentiated cells
.
Devi Biol
.
110
,
15
22
.
Rizzino
,
A.
,
Kusynski
,
C.
,
Ruff
,
E.
&
Iesman
,
J.
(
1988
).
Production and utilisation of growth factors related to FGF by embryonal carcinoma cells and their differentiated cells
.
Devi Biol
.
129
,
61
71
.
Rudnicki
,
M.
&
McBurney
,
M. W.
(
1987
).
Cell culture methods and induction of differentiation of embryonal carcinoma cell lines
.
In Teratocarcinomas and Embryonic Stem Cells: a Practical Approach
(Ed.
E.
Robertson
).
IRL press
,
Oxford
, pp.
19
50
.
Slack
,
J.
,
Darlington
.
B.
,
Heath
,
J.
&
Godsave
,
S.
(
1987
).
Mesoderm induction in early xenopus embryos by heparin-binding growth factors
.
Nature
326
,
197
200
.
Slack
,
J. M. W.
&
Isaacs
,
H. V.
(
1989
).
Presence of basic fibroblast growth factor in the early Xenopus embryos
.
Development
105
,
147
154
.
Smith
,
A. G.
,
Heath
,
J. K.
,
Donaldson
,
D. D.
,
Wong
,
G. G.
,
Moreau
,
J.
,
Stahl
,
M.
&
Rogers
,
D.
(
1988
).
Inhibition of pluripotential embryonic stem cell differentiation by purified polypeptides
.
Nature
336
,
688
690
.
Smith
,
R.
,
Peters
,
G.
&
Dickson
,
C.
(
1988
).
Multiple RNAs expressed from the int-2 gene in mouse embryonal carcinoma cell lines encode a protein with homology to fibroblast growth factor
.
EMBO J
.
7
,
1013
1022
.
Taira
,
M.
,
Yoshida
,
T.
,
Miyagawa
,
K.
,
Sakamoto
,
H.
,
Terada
,
M.
&
Sigmura
,
T.
(
1987
).
cDNA sequence of human transforming gene hst and identification of the coding sequence required for transforming activity
.
Proc. natn. Acad. Sci. U.S.A
.
84
,
2980
2984
.
Van Veggel
,
J.
,
Oostwards
,
T.
.
Delaat
,
S.
&
Vanzoelen
,
E.
(
1987
).
PC13 embryonal carcinomas cells express a heparin binding growth factor
.
Expl Cell Res
.
169
,
280
286
.
Van Zoelen
,
E. J.
,
Ward-Van Oostwaard
,
T.
,
Nieuwland
,
R.
,
Van Der Burg
,
B.
,
Van Den Eijden-Van Raaj
,
A.
,
Mummery
,
C.
&
Delaat
,
S.
(
1989
).
Identification and characterisation of polypeptide growth factors secreted by murine embryonal carcinoma cells
.
Devi Biol
.
133
,
272
283
.
Zhan
,
X.
,
Bates
,
B.
,
Hu
,
X.
&
Goldfarb
,
M.
(
1988
).
The human FGF-5 oncogene encodes a novel protein related to fibroblast growth factors
.
Mol. Cell Biol
.
8
,
3487
3495
.