Developmentally regulated expression of insulin-like growth factors by differentiated murine teratocarcinomas and extraembryonic mesoderm

The insulin-like growth factors (IGF) I and II are single-chain polypeptide hormones, which have structural similarities to insulin (Rindernecht & Humbel, 1978a,h). IGFs exert a number of biological effects upon cells in culture including the induction of cellular DNA synthesis and cell multiplication, and stimulation of sugar and amino acid transport (reviewed by Rechler et al. 1981). The levels of circulating IGF-I in plasma are regulated by growth hormone and IGF-I is consequently implicated as a mediating agent in the effects of growth hormone on skeletal growth during puberty (reviewed by Van Wyk et al. 1974). Serum IGF-II concentrations are not regulated by growth hormone but are elevated in the foetus (D’Ercole & Underwood, 1980; Moses et al. 1981), leading to the suggestion that IGF-II may regulate foetal growth in vivo (D’Ercole et al. 1980). There is, in fact, evidence that IGF-II-like molecules are synthesized by multiple tissues in the foetus, including liver (Rechler et al. 1979), fibroblasts (Adams et al. 1983), myoblasts (Hill et al. 1984), kidney, heart, lung, limb bud mesenchyme, intestine and brain (D’Ercole et al. 1980). More recently Scott et al. (1985) have shown that IGF-II transcripts are widespread in first trimester human foetal tissues. These findings have consequently led to the proposition that IGF-II acts as a paracrine growth factor in vivo where local secretion results in the proliferation of neighbouring cell types (D’Ercole et al. 1980, Hill et al. 1984).

The biological effects of IGFs are exerted through association with specific plasma membrane IGF receptors. Two distinct species of IGF receptor have been demonstrated by crosslinking radiolabelled IGFs to plasma membranes. The type 1 IGF receptor structurally resembles the insulin receptor, and comprises a tetramer of two α chains (relative molecular mass M_r 130 000) and two β chains (M_t 60000) linked by disulphide bonds (Masague & Czech, 1982). The type 1 IGF receptor binds IGF-I and IGF-II with high affinity, and insulin with lower but significant affinity. The type 2 IGF-receptor is a single-chain molecule of apparent relative molecular mass 260000 (Masague et al. 1981) which binds both IGF-I and IGF-II with high affinity, but has no significant affinity for insulin (Masague & Czech, 1982).

That IGF-II-like molecules may be expressed and function, in embryonic stages of mammalian development, arises from the finding that murine embryo-derived embryonal carcinoma (EC) cells express receptors for, and respond to, exogenous IGF-II in vitro (Nagarajan et al. 1982; Stern & Heath, 1983; Heath & Deller, 1983; Heath & Rees, 1985). Here we report that the induction of PC13EC cell differentiation in vitro by retinoic acid is accompanied by a drop in the number of available plasma membrane IGF receptor sites and in the developmentally regulated expression of molecules which, in structure and activity, resemble IGF-II. The use of antibodies directed against rat IGF-II reveals that multiple forms of IGF are expressed in vitro and that IGF-like molecules are synthesized in the embryo by extraembryonic tissues of mesodermal derivation.

The origin and preparation of cell culture materials has been previously described (Heath & Deller, 1983). The Buffalo rat liver (BRL) cell line was obtained from the cell bank, Sir William Dunn School of Pathology, University of Oxford, and cultured as described (Heath & Deller, 1983). The maintenance and retinoic-acid-induced differentiation of PC13EC clone 1A4 cells was as described by Heath & Deller (1983).

PC13END-cell-conditioned medium was prepared from PC13END cells generated by RA exposure of PC13EC cells in 175 cm² culture flasks. Following 5 days’ exposure to differentiation-inducing conditions PC13END cell monolayers (approximately 10⁷ cells per flask) were washed extensively with basal medium and then cultured for 2 days in 100 ml ECM serum-free supplemented medium (Heath & Deller, 1983) per flask at 37 °C in an atmosphere of 5 % CO₂ in air. Conditioned medium was decanted, centrifuged at 9000g for 30min and concentrated approximately 150-fold in a stirred cell ultrafiltration device (Amicon, UK) equipped with YM2 membranes.

BRL-cell-conditioned medium was prepared from cells cultured in 1030 cm² glass roller culture vessels in 200 ml DME: F12 (50:50 vol: vol) supplemented with 5 % calf serum (Gibco) at a density of approximately 10⁶ cells per vessel. The cells were cultured with continuous rotation at 37°C for 3 days in an atmosphere of 5% CO₂ in air. The medium was changed to 200 ml DME:F12 (50:50 vol: vol) supplemented with iron-conjugated transferrin at 5 μg ml⁻¹ and cultured as above. The medium was removed after 24 h and replaced with 200 ml of fresh DME:F12: transferrin, the first serum-free medium collection was discarded and the medium was subsequently collected and replaced every 2–3 days with fresh serum-free medium as above. Media collections could be made for up to 14 days before overt cell degeneration became evident. Conditioned medium (approximately 101 per batch) was centrifuged at 9000g for 30 min to remove cells and debris and concentrated 50-fold in a hollow fibre filtration apparatus (DC2 Amicon, UK, HIP2-43 cartridges).

IGF-I (Amersham, UK), IGF-II, and experimental fractions were iodinated by the iodogen method. Freeze-dried peptides were dissolved in 10 μl 0·1 % TFA for 20 min at room temperature and diluted to 100 μl by addition of 0·2M-phosphate buffer pH7·8. The dissolved peptides were added to a vial precoated with 10pg iodogen (Pierce, UK) and the reaction initiated by addition of 0·5–1 mCi ¹²⁵I (Amersham, UK). The mixture was incubated for 20 min at room temperature and then terminated by addition of KI and tyrosine to a final concentration of 0·1 M. Following a further 5 min incubation the mixture was diluted to 2 ml with 0·1 % TFA and applied to a Sep-pak ODS cartridge (Waters, UK). The cartridge was washed with 10 ml 0·1 % TFA in water, and iodinated peptides eluted with two 1 ml portions of 50 % acetonitrile/0-1 % TFA in water. Specific activities thus obtained ranged from 46 000–104 000 cts min⁻¹ ng⁻¹.

IGF radioreceptor assays were performed on cell monolayers essentially as described by Heath et al. (1981). PC13EC and PC13END cells were plated into gelatin-coated 3·5 cm diameter dishes at a density of 10⁵ cells/dish (EC) or 2× 10⁵ cells/dish (END) in 4 ml F12: DME supplemented with 5% foetal calf serum. After 24h culture the medium was changed to 4ml ECM serum-free supplemented media (Heath & Deller, 1983) and the cell monolayers used for binding assays 24 h later. PYS-2 cells were plated at a density of 5x10⁴ cells per well in 24-well cluster plates in F12:DME 5% calf serum and used for radioligand binding assays 48 h later. PC13 membranes were prepared by homogenization of PC13EC cells in lOmM-tris, 1mM-EDTA, 10% sucrose, pH7·2, followed by centrifugation at 3000g for 30 min at 4°C to remove nuclei. A crude membrane fraction was prepared by centrifugation of the supernatant at 100000g for 1 h, the membrane pellet was resuspended in PBS and rehomogenized. Membranes were stored in liquid nitrogen before use. For crosslinking studies, membranes were incubated at 250 μg ml⁻¹ in binding buffer (Earles balanced salt solution, 20mM-Hepes, 0-5% bovine serum albumin pH 7·2) with ¹²⁵I-IGF-I to a final concentration of 1 nM, with or without 1 μg unlabelled IGF-I. Following 2h incubation at 4°C the membranes were pelleted by centrifugation at 15000g for 10 min, resuspended in PBS and bound IGF-crosslinked by addition of disuccinimidyl suberate (Pierce, UK, freshly dissolved in dimethyl sulphoxide) to a final concentration of 0·25 MM or 0·5 MM. Following 15 min incubation on ice the membranes were repelleted as above and dissolved in 50 μl SDS sample buffer with or without 50 IHM-dithiothreitol, separated by SDS–PAGE (5 % acrylamide resolving gels, Laemmli, 1970), dried, and exposed to Fuji-RX X-ray film at −70°C in the presence of intensifying screens.

Rat IGF-II was purified by a modification of the methods described by Marquardt et al. (1981). Concentrated BRL-conditioned media were dialysed at 4°C for 48 h against 0·1 M-acetic acid (201, two changes) in ‘Spectrapor’ dialysis bags (nominal cut-off 3000 daltons, Spectrum Medical Industries, USA). Acid-insoluble material was removed by centrifugation (100000g for 30min). The acid-soluble portion was freeze-dried, reconstituted in 1 M-acetic acid (1ml for every 400ml of the original concentrated medium), and subjected to gel filtration at 4°C on a 2·5 cm diameter × 90 cm length column of Biogel P-60 (200–400 mesh, BioRad, UK) equilibrated in 1 M-acetic acid. The column was eluted at a flow rate of 10 ml h⁻¹ collecting 10 ml fractions. 1ml samples from column fractions were freeze-dried and reconstituted in 1ml phosphate-buffered saline (PBS) for the determination of growth-promoting activity and protein concentration (method of Read & Northcote, 1982). IGF activity was assayed by incorporation of trichloracetic-acid-insoluble [³H]thymidine into quiescent NRK 49F fibroblasts as described (Marquardt et al. 1981) or by inhibition of ¹²⁵I-IGF-binding to PYS cells. Three zones of mitogenic (or IGF competing) activity were routinely observed. The latest eluting fractions containing mitogenic activity (corresponding to molecular mass about 7000 daltons) were pooled, freeze-dried, reconstituted in 0-01 M-trifluoroacetic acid (TFA, solvent A, sequencer grade, Rathburn Chemicals, UK) and subjected to reverse-phase high-pressure liquid chromatography (rpHPLC) on an Altex model 344 gradient liquid chromatography apparatus with a C3 30nm pore size support (Ultrapore RSPC, Beckman, UK). Absorbed material was eluted by application of a gradient of acetonitrile (HPLC grade S, Rathburn, UK), 0·01 M-TFA (solvent B) at a flow rate of 1 ml min⁻¹. Sample elution was monitored with a Beckman (UK) model 160 fixed wavelength u.v. detector equipped with 214 nm filters. A single peak of mitogenic activity eluted at a nominal concentration of 28% solvent B. Peak activities were concentrated by solvent evaporation and subjected to size exclusion chromatography on a TSK-SW2000 (7 mm × 60 cm) column in 10% acetonitrile/0·1 % TFA, in water at a flow rate of 0·5 ml min⁻¹. Two peaks of activity were obtained with M_r= 10000 and 7000.

Concentrated PC13END-cell-conditioned medium was adjusted to 1 M-acetic acid, centrifuged for 30 min at 100000 g to remove acid-insoluble material and fractionated by Biogel P-60 chromatography and rpHPLC as above.

Freeze-dried fractions were reconstituted in phosphate-buffered saline. Crosslinking reactions took place in 200 μl phosphate-buffered saline/1 % (by weight) bovine serum albumin at a final sample protein concentration of 50 μg ml⁻¹. The reaction was initiated by addition of ¹²⁵I-rat IGF-I to a final concentration of 0·5 nM in the presence or absence of 500 ng unlabelled IGF-I, and incubated for 2h at room temperature. Crosslinking was achieved by addition of DSS to a final concentration of 0·2 MM followed by incubation at 4°C for 15 min. The crosslinking reaction was terminated by addition of double-strength SDS sample buffer and boiling for 5 min followed by electrophoresis in 12·5 % polyacrylamide gels containing SDS.

IGF-II peptide (M_r = 7000 prepared as above) was coupled to keyhole limpet haemocyanin (KLH) before immunization. 5 mg KLH (Calbiochem-Behring) was dissolved in 450 μl 0·1 M-NaHCO₃ (pH 9·0) and reacted with 50 μl of 2·5% glutaraldehyde (BDH, UK) at room temperature with continuous agitation. Free glutaraldehyde was removed after 2h incubation by desalting on a 0·9 cm diameterxllcm column of Biogel P-2 (200–400 mesh, Bio-Rad) equilibrated in 0·lM-NaHCO₃ (pH9·0). The glutaraldehyde-activated KLH was mixed with 250 μg of IGF-II peptide and incubated overnight at 4°C with constant agitation. The conjugation mixture (2 ml) was dialysed for 24 h at 4°C against 51 of 0·1 M-acetic acid in a 3000 M_r cut-off ‘Spectrapor’ dialysis bag, divided into four 500 μl portions and freeze-dried. No attempt was made to quantify the extent of conjugation. The freeze-dried portions were reconstituted in 200 μl of PBS and emulsified with Freunds adjuvant. Male New Zealand white rabbits were injected subcutaneously on days 0 (complete adjuvant), 14, 28 and 38 (incomplete adjuvant). Blood was taken on day 44 and at intervals thereafter. IgG fractions were prepared from immune serum by 40 % ammonium sulphate precipitation, followed by dialysis of the precipitate against 20mM-sodium phosphate (pH 7·0) and chromatography on a 2·5 cm diameterx 10 cm column DEAE-trisacryl (LKB, UK) equilibrated in the same buffer. Unabsorbed fractions containing immune IgG were dialysed against PBS containing 0·01 % sodium azide and concentrated by ultrafiltration to a final concentration of 2 mg ml⁻¹. Antibody species directed against KLH were removed by passing the immune IgG fractions through a 5 ml column of KLH-Affigel (Bio-Rad, approx. 5 mg KLH ml⁻¹ Affigel-10 prepared according to the manufacturers’ recommendations) equilibrated in phosphate-buffered saline (PBS) and collecting the unabsorbed material. The specificity of KLH-absorbed anti-IGF-II antisera was tested by titration in a solid-phase ELISA assay against a variety of target peptides, including: porcine insulin; desoctapeptide insulin; glucagon; epidermal growth factor; nerve growth factor; platelet-derived growth factor; ovine growth hormone; bovine prolactin; human placental lactogen; and four different preparations of rat IGF-II, two prepared in this laboratory, a commercial preparation (rat multiplication-stimulating activity, Collaborative Research) and a preparation of rat IGF-II generously provided by Dr M. Czech (Dept of Biochemistry, University of Massachussets Health Centre, Worcester, MA, USA). Significant reactivity was only observed with the preparations of IGF-II. Significant binding of the immune IgG preparations to IGF-II peptides was observed at IgG concentrations above 20 ng ml⁻¹.

Embryonic tissues were obtained from natural matings of C3HXC57/B16 Fj mice. Embryos were staged and dissected as described by Shi & Heath (1984).

Embryonic tissues were metabolically labelled with [³⁵S]methionine in conditions described previously (Shi & Heath, 1984). PC13EC, differentiated PC13EC cells (derived by retinoic acid treatment) and BRL cells were plated at a density of 10⁶ cells in a 25 cm² tissue culture flask in DME: F12 (50:50) supplemented with FCS (5 % by volume). The cells were incubated overnight at 37°C in 5 % CO₂ in air. The cell monolayers were then washed once with PBS and the medium changed to 5 ml methionine-free minimal essential medium (Gibco, UK) supplemented with FCS (0·5% by volume) and 50 μCi ml⁻¹ [³⁵S]methionine (specific activity 2000 mCi mmol⁻¹, Amersham, UK), followed by culture for 16–18h in the conditions described above. Culture supernatants were centrifuged at 15000g for 15 min and divided into 1ml or 0·5 ml samples which were either used for immunoprecipitation immediately or stored frozen at −70 °C for up to 2 weeks before use. No differences were noted between fresh or frozen material.

The procedures for immunoprecipitation and electrophoretic analysis of secreted products in labelled culture media were as described (Shi & Heath, 1984). Immune IgG was used at a concentration of 200 μg ml⁻¹.

Samples resolved by discontinuous electrophoresis in polyacrylamide gels containing sodium lauryl sulphate (buffer system of Laemmli, 1970) were electrophoretically transferred (Towbin et al. 1979) to nitrocellulose paper (Sartorius, UK, type SM) and reacted with immune IgG and alkaline-phosphatase-conjugated sheep anti-rabbit IgG (Serotec, UK, diluted 1/1000) as described by Blake et al. (1984).

Acid-extracted BRL-conditioned medium was subjected to SDS–PAGE (resolving gel 15 % acrylamide) in multiple sample slots (approx. 250 μg of protein per slot). Paired sample lanes were excised from the gel after the bromophenol dye front had migrated to approx. 1 cm from the end of the gel, sliced into 3 mm sections, and then cut in halves representing each sample lane. Each section from one of the paired lanes was stored at −70°C in SDS sample buffer and the sections from the other lane were equilibrated with SDS sample buffer and electrophoresed on a second 15 % SDS–PAGE gel followed by transfer to nitrocellulose paper and reaction with rabbit anti-IGF-II IgG to locate sections containing immunoreactive material. The corresponding sections from the paired sample lane were then subjected to partial proteolysis with Staphylococcus protease V8 (Worthington, UK, lug/slot) and electrophoresis according to Cleveland et al. (1977). Immunoreactive peptides were detected by immunoblotting as above.

The expression of IGF receptors by teratocarcinoma cells was examined by testing the ability of ¹²⁵I-IGF-I to bind specifically to PC13EC cells and their differentiated derivatives. Significant specific saturable binding of IGF-I to EC cells and their differentiated derivatives was observed. Replotting the data according to the method of Scatchard (1949) permitted an estimation of IGF receptor number and affinity (Fig. 1); PC13EC cells express approximately 80 000 sites (K_d 0·98 nM) per cell. These findings confirm the earlier reports of IGF receptor expression by F9EC (Nagarajan et al. 1982) and PC13EC (Stern & Heath, 1983) employing IGF-II as the radiolabelled ligand. Upon RA-induced differentiation the number of sites fell by about 60 % (to 20 000 sites per cell) whilst the apparent affinity of IGF-I for PC13END cell receptor sites remained similar (K_d = 0·83 MM). Since the number of contaminating residual EC cells found in PC13END cell preparations under our differentiation-inducing conditions is very low (less than 0·01 %, Heath & Deller, 1983) it is probable that this represents a genuine decline in the apparent number of unoccupied IGF receptor sites available after RA-induced differentiation.

IGF-I and IGF-II have been found to react with two different receptor species on a variety of cultured cell types when analysed by crosslinking labelled ligand to membrane receptor sites (Masague & Czech, 1982). The identity of IGF receptors expressed by PC13EC cells was determined by allowing radiolabelled IGF-I to react with PC13EC cell membranes followed by crosslinking with disuccinimidyl suberate (DSS, Pilch & Czech, 1980) and electrophoresis in polyacrylamide gels containing SDS (SDS–PAGE). Electrophoresis under nonreducing conditions revealed a number of specifically labelled molecular species ranging from M_r= 280000 to M_r = 330000 (Fig. 2). Electrophoresis under reducing conditions yielded two molecular species; one of M_r = 260000 which corresponds to the reported size of the type 2 IGF receptor, and a second of M_r = 130000 corresponding to the α chain of the type I (insulin-like receptor). In common with other studies (e.g. Masague & Czech, 1982) the β chain of the type 1 receptor, which does not interact directly with the ligand, would not be detected under the conditions employed. PC13EC cells therefore express both type 1 and type 2 IGF receptors.

We considered the possibility that the apparent fall in the number of IGF receptor sites expressed by PC13EC cells upon RA-induced differentiation was due to the developmentally regulated endogenous expression, by PC13END cells, of molecules which inhibited the binding of exogenous IGF-I either by reaction with IGF-I itself, or with the corresponding plasma membrane receptors. Serum-free ECM media conditioned by PC13END cells were concentrated by ultrafiltration, solubilized in 1 M-acetic acid and subjected to gel filtration on Biogel P-60. Individual fractions were tested for the ability to inhibit the binding of IGF-I to PYS-2 parietal yolk sac cells. Two zones of inhibitory activity were observed; zone A eluting near the ovalbumin size marker and zone B eluting around and after the lysozyme size marker (Fig. 3). No IGF-I-binding inhibition activity was recovered by similar analysis of either PC13EC cell conditioned or unconditioned ECM media (not shown).

The existence of two similar classes of IGF-inhibitory activity has been reported in the analysis of media conditioned by other IGF-secreting cells (Marquardt et al. 1981; Anderson et al. 1984) and tissues (Rechler et al. 1979). The high molecular weight material has been ascribed to the presence of a specific soluble IGF-binding protein ( Knauer et al. 1981; Knauer & Smith, 1982). We examined the possibility that zone A contained binding protein activity by testing the ability of zone A fractions to bind IGF-I directly. Zone A fractions were pooled, freeze-dried, reconstituted in PBS/BSA, incubated with ¹²⁵I-IGF-I in the presence or absence of excess unlabelled IGF-I and then subjected to crosslinking with DSS, followed by SDS–PAGE. A single species of approximate M_r = 40000 was specifically labelled in the presence of ¹²⁵I-IGF-I; this species was not labelled in the presence of 500-fold excess unlabelled IGF-I or in the absence of zone A material or crosslinking reagent (Fig. 4). We conclude that the zone A contains a species which binds directly to IGF-I of approximate M_r= 35 000 (derived by subtracting the molecular weight of IGF-I from 40000 and assuming that one molecule of IGF-I binds one binding protein molecule.

Zone B fractions were pooled and further fractionated by reverse-phase high-pressure liquid chromatography in a solvent of 0·1 % TFA in water, eluting with a linear gradient of acetonitrile/0·1 % TFA (Fig. 5). Individual fractions were tested for the ability to inhibit IGF-I binding to PYS-2 cells and stimulate DNA synthesis in quiescent NRK rat fibroblasts by [³H]thymidine incorporation. A major peak of IGF-binding inhibition, coincident with a peak of DNA-synthesis-promoting activity and a u.v. absorption peak, was observed which eluted at a similar nominal acetonitrile concentration to authentic rat IGF-II (26% acetonitrile). A second minor peak with mitogenic activity and IGF-inhibitory activity eluted about 8 min later. This finding suggested that the IGF-I-binding inhibition activity present following reverse phase chromatography separation was due to the presence of an activity which, in structure and function, resembles IGF-II. Samples of the peak fractions from reverse-phase separations were freeze-dried, radiolabelled with ¹²⁵I, and tested for their ability to bind to PYS-2 cells. Specific binding of PC13END-cell-derived IGF-like material was observed (Table 1), which could be inhibited by unlabelled END-cell-derived IGF, IGF-I, IGF-II, but not EGF, or embryonal carcinoma derived growth factor (ECDGF).

We conclude that PC13END-cell-conditioned medium contains a developmentally regulated IGF-like activity based on its ability to interact specifically with IGF receptors and stimulate DNA synthesis in quiescent NR K fibroblasts.

Although the presence of IGF-competing activity in conditioned medium provides circumstantial evidence for IGF synthesis, it would be advantageous to have independent confirmation of IGF-II expression by metabolic labelling of PC13END-cell-derived IGF.

We accordingly raised polyclonal antibodies to BRL-cell-derived rat IGF-II, conjugated to KLH carrier protein. Immune IgGs from two rabbits were analysed in detail with essentially similar results. BRL cells were metabolically labelled with [³⁵S]methionine and immunoprecipitated with rabbit anti-IGF-II IgG. Three radiolabelled species (Fig. 6A) of apparent M_r= 35000, 18000 and 16000 were observed. In some experiments a fourth species of apparent M_r = 14000 (doublet) was also observed. Immunoprecipitation of all these species was inhibited in the presence of IGF-II. Similar species (including the M_r = 14000 form in this instance) were also observed by immunoblotting acid-extracted BRL-conditioned media (Fig. 6B) with our rabbit anti-IGF-II demonstrating that these forms were recognized directly by the antibody and were not the result of coprecipitation. Immunoprecipitation of [³⁵S]methionine-labelled PC13EC and PC13END cell culture supernatant resulted in the specific precipitation of the M_r= 35 000 and M_r= 14 000 doublet and minor bands of 18 000 and 16 000 species from PC13END but not PC13EC cells (Fig. 7). We conclude that our rabbit anti-IGF-II IgG recognizes molecules which are expressed by BRL cells and PC13END cells but not PC13EC cells and whose expression is developmentally regulated during teratocarcinoma differentiation.

The 18000, 16000 and 14000 forms observed in these experiments probably represent precursor forms of an IGF-II prohormone since our antibodies were raised against size exclusion chromatography purified M_r = 7000 rat IGF-II. Sequence analysis of rat IGF-II cDNAs (Bell et al. 1984; Dull et al. 1984; Whitfield et al. 1984; Soares et al. 1985) have revealed that sequences encoding the 7000 dalton form of IGF-II are embedded in a longer open reading frame with the capacity to code for a 156 amino acid precursor whose predicted amino sequence contains a number of paired basic amino acid residues forming potential proteolytic processing sites. The expression of these proposed IGF-II prohormone forms by BRL cells has been established by Acquaviva et al. (1983) and Yang et al. (1985) who showed that immunoprecipitation of either radiolabelled BRL cell extracts, or in vitro translation reactions primed with BRL mRNA, with antibodies directed against IGF-II yielded a variety of secreted IGF species of approximate relative molecular masses 19000, 15000 (doublet), 10000, 8000 and 7000 derived from an M_r = 20000 primary intracellular IGF-II translation product. It is important to note that in our studies radiolabelling with [³⁵S] methionine would only be expected to detect proforms of IGF-II since the first methionine residue occurs at amino acid 117 in the predicted rat IGF-II amino acid sequence (assuming sequence conservation between rat and mouse IGF genes).

The prominent M_r = 35 000 species observed here is not readily accommodated within the predicted amino acid sequence of pro-IGF-II. The immunoblot analysis (Fig. 6B) shows, however, that the 35000 species is directly recognized by antibodies present in our rabbit anti-IGF-II IgG preparations. One-dimensional peptide mapping of the 35 K and 19 K species present in BRL-conditioned media did not reveal the presence of shared immunoreactive peptides, suggesting that the 35 K species was not directly related to the other forms (Fig. 8). The 35 K species observed in our experments is, however, similar to the 33 K species identified in immunoprecipitates of BRL-conditioned media by Yang et al. (1985) as the soluble IGF-binding protein. A 35K IGF-binding protein has also been identified by crosslinking studies in liver perfusates by Schwander et al. (1984) and D’Ercole & Wilkins (1984). We have confirmed the identification of the immunoreactive 35 K species as the IGF-binding protein by its purification from BRL conditioned medium (J. K. Heath & B. J. Smith, in preparation). TV-terminal amino acid sequence analysis of the purified protein does not reveal any sequence homology with either pro-IGF-II or the predicted translation products of the long open reading frame described in the 5′ extension of the pro-IGF-II cDNA clone described by Dull et al. (1985). It is probable therefore that the IGF-binding protein is encoded by a separate gene from IGF-II.

It is not clear why immunization with purifed IGF-II should yield antibodies which react with the binding protein. Although Yang et al. (1985) ascribed the presence of antibodies directed against the binding protein to immunization with impure preparations of IGF, we consider this explanation unlikely in this case since the preparations of IGF-II used for immunization were subjected to high-resolution SEC as the final stage of purification. It may be significant, however, that immunoprecipitation of the M_r = 35 000 species is inhibited by IGF-II (Fig. 6) suggesting that the antibodies recognize sites which interact with IGF-II.

On the basis of these considerations we concluded that our antibodies recognized prohormone forms of IGF and the soluble IGF-binding protein.

The availability of specific antibodies that recognize murine IGF-II and the IGF-binding protein allowed us to analyse the expression of IGF-like molecules by tissues of the early postimplantation mouse embryo, which would be difficult to study by direct methods. Extraembryonic tissues of the 9·5 days post coitum (d.p.c.) embryo were explanted in culture and metabolically radiolabelled with [³⁵S]methionine. Radiolabelled culture supernatants were immunoprecipitated with rabbit anti-IGF-II. Specific coexpression of the four IGF-related protein species was observed in amnion and extraembryonic yolk sac mesoderm (Fig. 9). However, no evidence was found for expression of IGF-like molecules or binding protein by either parietal endoderm or visceral endoderm (Fig. 9). The results suggest that IGF-like molecules and their cognate binding proteins are specifically expressed in vitro by early postimplantation tissues of extraembryonic mesodermal origin.

Here we report that PC13EC cells, and their retinoic-acid-induced differentiated progeny PC13END express specific high-affinity receptors for insulin-like growth factors. These receptors probably mediate the effects of physiological concentrations of IGFs on EC cell viability, multiplication and differentiation (Heath & Deller, 1983; Heath & Rees, 1985). Crosslinking studies show that PC13EC cells express both type 1 and type 2 IGF receptors. Since type 1 IGF receptors, unlike type 2, can bind insulin (Masague & Czech, 1982) and PC13EC cells do not express detectable insulin receptors (Heath et al. 1981), the reported (Heath & Deller, 1983) effects of relatively high concentrations of insulin and insulin analogues on EC cell survival may be mediated through these type 1 receptors.

The apparent number of IGF receptors falls by about 60 % upon RA-induced differentiation, and this is accompanied by the expression of both IGF-like molecules and soluble IGF-binding proteins; PC13END cells are also responsive to exogenous IGF-II. By analogy with similar findings in other cellular systems it is reasonable to suppose that these two phenomena are linked, in that the endogenous production of IGF-like molecules results in the occupation of a fraction of PC13END cell IGF receptors, resulting in an apparent decrease in the number of sites available to bind exogenous IGF. This type of phenomenon has often been linked with malignant cell behaviour (reviewed by Burk, 1980). It is of interest, therefore, that PC13END cells are nontumourigenic in vivo and exhibit a finite proliferative lifespan in vitro (reviewed by Heath, 1983). ‘Autocrine’ cell proliferation may therefore be a feature of certain normal embryonic cell types as well as malignant cells. The autocrine synthesis of IGF-like molecules by PC13END cells may also provide an explanation for the observed dependency of PC13END cell multiplication on cell density (Heath & Rees, 1985) in the absence of exogenous growth factors. Since insulin has been reported to increase the affinity of IGF for the type 2 IGF receptor, at least in the case of adipocytes (Oppenheimer et al. 1983), the mitogenic effects of low concentrations of insulin on PC13END cells (Heath et al. 1981) may occur indirectly as a result of increased insulin-induced occupancy of PC13END cell IGF receptors by endogenous IGF.

The expression of IGF receptors by PC13EC cells, and the synthesis of IGF-like molecules by their immediate differentiated progeny points to the existence of a feedback relationship whereby PC13END cells secrete factors which maintain proliferation of their undifferentiated parents (reviewed by Heath & Rees, 1985). Experimental support for this notion comes from the work of Isacke & Deller (1983) who showed that PC13END cell feeders could substitute for exogenous serum-derived factors to support PC13EC cell multiplication in vitro. (An opposite feedforward growth dependency also occurs in this system through the mitogenic action of PC13 embryonal carcinoma derived growth factor on PC13END cells (Heath & Isacke, 1984).) Together these considerations suggest that a sophisticated interactive network of growth and differentiation control pathways exists in murine teratocarcinomas in which IGF-like molecules play a key role, acting simultaneously as ‘autocrine’ and paracrine growth factors.

Further characterization of secreted forms of IGF-like molecules by immunoprecipitation of metabolically radiolabelled molecules from culture supernatants with antibodies directed against rat IGF-II revealed multiple molecular species consistent with the proposed production of IGF by proteolytic processing of a longer precursor. The sizes of the secreted pro-IGF forms in the present study are generally in accord with those observed for IGF-II secreted by BRL cells by Yang et al. (1985) and in our own study. We have also observed, however, an immunoreactive putative M_r= 14000 form, which is particularly prominent in immunoprecipitates of embryonic tissues. This may simply reflect quantitative, or qualitative, differences between IGF processing in different tissues or technical differences between the two studies. However, it is important to note firstly that cDNA sequence analysis of the human IGF-I gene (Jansen et al. 1983) suggests that it is also initially synthesized as a longer prohormone, and secondly4hat the possible existence of additional related IGF-like genes cannot as yet be eliminated. These considerations do not allow us at present definitively to identify the species, recognized by antibodies directed against rat IGF-II, and secreted by PC13END cells and embryonic tissues as IGF-II, IGF-I or other IGF-like molecules. Furthermore, the existence of differences in 5’ noncoding sequences from separate IGF-II cDNA clones (Soares et al. 1985), combined with the complexity of human IGF-II transcript sizes observed in normal tissues (Soares et al. 1985; Scott et al. 1985) suggest that post-transcriptional processing of the IGF-II transcript may occur prior to translation. It may be possible, therefore, to generate multiple molecular forms of IGF-II by a variety of different mechanisms.

The expression of IGF and IGF receptors by embryo-derived teratocarcinomas leads to the supposition that IGF is expressed in the early postimplantation mouse embryo. Immunoprecipitation of radiolabelled culture supernatants from 9·5 d.p.c. extraembryonic tissues has shown the amnion and extraembryonic yolk sac mesoderm, both mesodermal derivatives of the primitive ectoderm, to be the principal extraembryonic sites of IGF and IGF-binding-protein synthesis. This expression of IGF by the extraembryonic membranes may account for the presence of relatively high levels of both IGF and IGF-binding-protein activity in foetal amniotic fluid (Bala et al. 1978; Chochinov et al. 1978) and underlines the proposed function of these tissues in supporting embryonic growth in utero (Heath & Shi, 1984). It is, furthermore, important to note that extraembryonic mesodermal cells, like PC13END cells, not only secrete IGF-like molecules but are responsive to exogenous IGF-II in vitro (Heath & Rees, 1985). Together these findings demonstrate that, as is the case for PC13END cells, the multiplication of extraembryonic mesodermal cells is, at least in part, controlled by an ‘autocrine’ mechanism, involving IGF-like molecules.

The widespread expression of IGF-II in later development by foetal organs (Scott et al. 1985), many of which are initially derived from two component epithelial/mesenchymal structures analogous to the yolk sac, may mean that, as in the yolk sac, IGF-II is expressed by the mesodermal or mesenchymal cells in these organs and that the type of interactive growth control system we have described in murine teratocarcinomas may be recapitulated in later development.

Finally, our observations show that: (i) specific growth factors are expressed during the early embryonic stages of mammalian development, (ii) the multiplication of cells in the mammalian embryo may be regulated by both autocrine and paracrine action of these growth factors, and (iii) IGFs may have an important regulatory function in embryonic as well as later foetal and postnatal development.

We thank Judy Morton, Hilary Oakley and John Eady for their assistance with this project, and Michael Czech, Mathew Rechler and Chris Graham for helpful suggestions and discussion. This work was funded by grants from the Cancer Research Campaign, the Royal Society and the Medical Research Council. JH acknowledges the support of an MRC Senior Fellowship and W-KS the support of the Chinese Academy of Sciences, the Henry Lester Fund and the British Universities China Committee.