ABSTRACT
EGF receptors are expressed on most fetal and adult cells but their precise roles are not well known. We previously reported that, in P19 embryonal carcinoma cells, the expression of kinase-negative EGFR inhibits retinoic acid (RA)-induced differentiation to nervous tissue, suggesting that EGFR plays a role in differentiation (J.-X. Wu and E. D. Adamson (1993) Dev. Biol. 159, 208-222). Embryo stem (ES) cells differentiate into a wide range of tissue types after the removal of the cytokine LIF from the culture medium. We demonstrate here that the induction of some early markers of differentiation, tissue-type plasminogen activator (tPA), AFP and keratins 8 and 19 is inhibited, whilst brachyury and myosin are increased, in clones containing kinase-negative mutant EGFR. After an extended period of differentiation, the cell types present in mutant and control cultures differed. Mutant clones produced frequent cardiac and skeletal muscle as the predominant differentiated cell types in vitro; other cells types were sparse or absent. Teratocarcinomas formed by EGFR-Δkinase-expressing ES cells contained frequent skeletal and cardiac muscle as well as apoptotic nuclei, while normal ES cells produced no detectable muscle and less apoptoses. Since mutant differentiated cultures had slower growth rates and increased levels of cell death, we concluded that: (1) inactive EGFR does not allow some cell types to survive and/or proliferate; (2) tissues that do not require EGFR for their survival, development or function predominate in long-term mutant cultures; (3) EGFR activity is not necessary for cardiac and skeletal muscle or endoderm formation and (4) Impaired survival of EGF-dependent lineages leads to preferential selection of muscle in differentiating ES cells.
INTRODUCTION
The proto-oncogene epidermal growth factor receptor (EGFR, ErbB1, HER-1) is expressed by most cell types of the body with some exceptions including parietal endoderm (Adamson and Meek, 1984) and mature skeletal muscle (Lim and Hauschka, 1984a,b). The function of the receptor varies according to the cell type and its degree of differentiation. Expression of the gene starts as early as the 4-to 8-cell preim-plantation embryo and increases rapidly (Wiley et al., 1992) on trophectoderm cells where receptors change gradually from an apical to a basal position in the completed epithelium (Dardik et al., 1992). Its presence on the trophectoderm apical surface suggests, together with data from other laboratories, that this receptor may play a role in the implantation of the embryo. Clearly, a main role for the ligands that bind the receptor to create a signal, is the stimulation of DNA synthesis, cell growth and mitosis. However, a large number of pleiotropic effects have been observed after ligand stimulation of the EGFR, indicating the complexity of the signal pathways and hence the diversity of the effects and roles of the receptor.
In order to simplify the analysis of the expression and roles of the EGFR during development, we have used embryonal carcinoma (EC) cells as models. P19 EC cells can differentiate into three main directions depending on the retinoic acid (RA) concentration. After aggregate culture in the presence of 0.5 μM RA and subsequent outgrowth on adhesive plastic, P19 cells differentiate into neurons and glial-type cells marked by the intermediate filament markers, neurofilament protein NF-1 and glial fibrillary acidic protein (GFAP), respectively. We have demonstrated that the expression of EGFR mRNA and protein are induced at least 10-to 20-fold during RA-induced differentiation from very low levels found in the undifferentiated stem cells (Joh et al., 1992). In fact, we have been unable to detect EGF binding in P19 or any EC line tested except F9 cells (Adamson and Hogan, 1984). The low level of receptor protein present in undifferentiated cells appears to be intracellular while differentiated cells express cell surface EGFR (Weller et al., 1987).
Previously, we introduced an expression vector into P19 cells that encodes the extracellular and transmembrane domains of EGFR but is deficient in all but 43 amino acids of the intracellular component of the polypeptide. We have shown that the truncated kinase-deficient form of the protein disrupts the activity of the endogenous EGFR by heterodimerization to produce an inactive form. The expression of the truncated EGFR, driven by the CMV promoter and enhancer, was activated after the differentiation of the EC cells. At the same time as endogenous EGFR was produced, the truncated version was also synthesized and we demonstrated that it inhibited differentiation and the production of nerve cells (Wu and Adamson, 1993). This was partially accomplished by the increased rate of cell death (2-to 6-fold) in mutant cultures compared to controls during the 6 days of differentiation (our unpublished results). We concluded that EGFR plays a role in the differentiation of wild-type P19 cells to nervous tissues and, by extrapolation, in neural development in vivo.
To determine other roles and activities of the EGFR, we are using totipotent ES cells that can give rise to chimeric mice after their introduction into a host blastocyst and transfer into a recipient female. ES cells are able to respond to all the signals that occur in vivo during development, differentiation and organogenesis. These cells are also able to take part in gonadal development and incorporate into the germ line of chimaeras. However, in culture, developmental cues are quite limited or absent and ES cell differentiation is a random and incomplete process leading to a mixture of cell types in a process that cannot yet be directed to one or another particular cell type (but see Bain et al., 1995). The tissues produced frequently are visceral endoderm in the early stages, parietal endoderm which persists, cardiac, skeletal and smooth muscle, fibroblasts and various epithelial cell types in the later stages. The most obvious cell type is the cardiac myocyte since groups of these beat rhythmically in the dish. ES cells give rise to teratocarcinomas after subcutaneous injection into syngeneic adult hosts. Keratin pearls, epidermal epithelia, secretory and ciliated epithelia, pseudostratified epithelia and occasionally a little muscle can be observed in tumours.
Using an improved dominant negative (kinase-minus) EGFR expression vector, we have now transfected the E14 ES cell line and derived several mutant clones resistant to G418 by the cotransfection of the bacterial neomycin-resistance plasmid, pMCneo. Control clones were transfected with the empty cassette and were also selected with G418. We measured several parameters of differentiation and came to the conclusion that ES cell differentiation is modified by the expression of the mutant EGFR since the cell types remaining after selective elimination appeared to be those that did not require the activity of the EGFR for their production and/or activity. Muscle was the principal differentiated tissue produced in mutant clones. The selective death and slower growth rates of EGFR-dependent cells could account for abnormal proportions of each cell type during differentiation.
MATERIALS AND METHODS
Cells and their culture
E14 embryo stem cells (Handyside et al., 1989) were cultured on gelatin-coated plastic dishes without feeder cells in Dulbecco’s modified Eagles Medium (DMEM) with high glucose and sodium pyruvate (10 mM) and glutamine (2 mM) in the presence of 15% bovine fetal serum (BFS), 27 μM 2-mercaptoethanol, 1000 U/ml LIF, 0.5 u/ml penicillin and 0.5 mg/ml streptomycin. For differentiation, cells were trypsinized and seeded at one million/10 ml in 10 mm Petri dishes for 5 days (or unless otherwise indicated) before reseeding in a tissue-culture dish for the indicated number of days. The medium used to stimulate differentiation was the same ES medium without LIF. In some cases, all-trans retinoic acid (0.6 μM RA) was used as an additional inducer of differentiation.
Tumour analyses
Teratocarcinoma formation was compared between the two cell types by subcutaneous injection of 107 cells into each of two sites in nu/nu mice. The tumours that developed (in two weeks) were weighed and were processed for histological analyses by standard procedures by staining with haematoxylin and eosin. Formaldehyde-fixed tissue was examined for apoptosis (see below). Frozen sections were also analyzed by immunocytochemical procedures (see below).
Expression vector construction and electroporation of ES cells
The expression vector for truncated EGF-receptor contains an EcoRI fragment (2.3 kb) from the cDNA (Joh et al., 1992) inserted into the CXN2 vector (Niwa et al., 1991). The promoter is β-actin (1.7 kb) with an enhancer of 200 bp derived from CMV and with β-globin poly(A) sequences (0.9 kb) downstream. This 4.9 kb fragment was cloned into the unique BamHI site of vector NNTG1 (Fig. 1). NNTG1 is derived from the human keratin gene and encodes 2.3 kb 5′ sequences, which were inserted upstream of the EGFR vector, and 3.5 kb 3′ sequences downstream (Neznanov et al., 1993). The role of the flanking K18 sequences is to insulate the vector sequences from inactivation after insertion into DNA and to endow dose dependency to concatamer inserts (Neznanov et al., 1993). A control vector lacking the EGFR insert was prepared similarly. For electroporation, the completed plasmids, K18-CXN-mEGFR (25 μg), K18-CXN (25 μg), and pMCNeo (5 μg), were linearized with NotI, NotI and BglI, respectively. DNA (20 μg vector plus 2 μg pMCNeo) was added to cells (8×106) in single-cell suspensions in 0.7 ml culture medium in the cuvette. The mixture was electroporated with 275 V for 5 mS and allowed to stand on ice for 10 minutes. Each reaction mix was plated into two 100 mm gelatinized plates. The selective drug G418 (400 μg/ml) was added after 24 hours. After 9 days most cells had died and clones were visible. About 20 clones were isolated from each dish after 13 to 15 days. Mutant clones are designated Km, control clones are Kc. The inhibitory activity of our truncated EGF receptor construct was demonstrated earlier (Wu and Adamson, 1993). The construct was used successfully, to inhibit EGF receptor function in two cell types by Moshier et al. (1995) and by Huang et al. (1996).
The plasmid construct used to introduce the truncated EGFR gene. K18, cassette derived from the human keratin 18 gene that insulates the internal sequences from inactivation and renders site-independent and dose-dependent insertion of vectors in genomic DNA (Neznanov et al., 1993; Thorey et al., 1993). TM, transmembrane domain of the mouse EGFR cDNA in a total of 2.3 kb. The expression of the kinase-negative EGFR is driven by the β-actin promoter and the CMV enhancer (Niwa et al., 1991). B, BamI; E, EcoRI; X, XhoI.
The plasmid construct used to introduce the truncated EGFR gene. K18, cassette derived from the human keratin 18 gene that insulates the internal sequences from inactivation and renders site-independent and dose-dependent insertion of vectors in genomic DNA (Neznanov et al., 1993; Thorey et al., 1993). TM, transmembrane domain of the mouse EGFR cDNA in a total of 2.3 kb. The expression of the kinase-negative EGFR is driven by the β-actin promoter and the CMV enhancer (Niwa et al., 1991). B, BamI; E, EcoRI; X, XhoI.
Antibodies
We have previously described the rabbit polyclonal antibody to purified EGF receptor from mouse liver (Weller et al., 1987; Wu and Adamson, 1993). TROMA-1 and TROMA-3 are rat monoclonal antibodies to keratins 8 and 19 (Brûlet et al., 1980; Kemler et al., 1981a,b) kindly provided by Dr R. Kemler. Hybridoma supernatant to chicken (cardiac and skeletal) myosin heavy chain (MF 20) was obtained from the Developmental Studies Hybridoma Bank maintained by the Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, MD and the Dept of Bio-logical Sciences, University of Iowa, Iowa City, IA under contract from NICHD.
Immunofluorescent staining
ES cells were seeded onto gelatinized glass coverslips and induced to differentiate for up to 6 days in medium containing RA. The cells were fixed in methanol at −20 °C for 5 minutes and stored desiccated and frozen until used. The cells were stained with undiluted TROMA-3 and TROMA-1 monoclonal antibody supernatants for 1 hour, washed and then reacted with fluorescein-labeled goat anti-rat IgG. The coverslips were mounted in 90% glycerol in PBS and observed using a Nikon epifluorescence Biophot microscope fitted with automatic exposure meter and camera. Frozen sections of teratocarcinomas were similarly processed for indirect immunofluorescence analysis for myosin H-chain using M20 MCAb.
In situ apoptosis by terminal transferase dUTP nick end-labelling (TUNEL)
The procedure described by the manufacturers (ApoTag™Plus, Oncor, Gaithersburg, MD) was used to detect the presence of apoptotic nuclei in paraffin wax embedded tumor tissues. Briefly, this consisted of labelling new 3′-OH ends generated by DNA fragmentation. Digoxigenin-labelled dUTP was incorporated into the fragmented DNA by terminal deoxynucleotide transferase, and detected by peroxidase-labelled anti-digoxigenin and substrate for peroxidase. The sections were counterstained with haematoxylin. A brown coloured nuclear stain indicates the cells undergoing apoptosis; the nucleus becomes contracted during this process. Sections were photographed at 400× on a Leitz Dialux 22 microscope.
Immunoprecipitation
ES clones in 35 mm dishes containing 105 cells were metabolically labelled with [35S]methionine and cysteine (TranSlabel, ICN, Irvine, CA) for 2 hours. Cells were lysed in RIPA buffer containing protease inhibitors and aliquots were taken for total radioactive protein measurement as previously described (Wu and Adamson, 1993). Samples containing equal amounts of radioactive protein were analyzed on 6% polyacrylamide gels containing SDS. After fixing, staining and drying the gels, fluorographic records were made on X-Omat film (Kodak).
Northern blotting
Cells in two 100 mm dishes were harvested at each of the indicated days of differentiation and total RNA was extracted using guanidium thiocyanate as described (Chomczynski and Sacchi, 1987). 30 μg RNA from each sample was electrophoresed and blotted using standard procedures (Sambrook et al., 1989). A [32P]dCTP-labeled cDNA encoding tissue-type plasminogen activator (tPA) (Strickland et al., 1980) was used as a probe (a kind gift from S. Strickland). Other probes were for α-fetoprotein (AFP), a marker of visceral endoderm (PstI fragment of pBR322.AFP2, obtained from Dr S. Tilghman); Brachyury, a marker for pre-mesodermal tissues (EcoRI fragment of pSK75 kindly supplied by Dr B. G. Herrmann, Wikinson et al., 1990), ErbB2 (fragment from pSV2neuN kindly supplied by Dr R. A. Weinberg, Bargmann et al., 1986); and ErbB3 (EcoRI-HindIII fragment from pTZ19U/ErbB3, kindly supplied by Dr G. Plowman). A probe that detects L32 mRNA, a ribosomal protein gene (Dudov and Perry, 1984), or mouse β-actin (cDNA fragment from pHβA-1 obtained from Dr L. Kedes, Ponte et al., 1984) were used as controls for equal loading of RNA in gel slots.
Immunoblotting (Western)
Cells were lysed in Laemmli sample buffer and the optical density at 280 nm measured to estimate the concentration of protein. Equal amounts of protein in each sample were analyzed on 6% polyacrylamide gels containing SDS, electrotransferred on to PVDF membranes (Immobilon, Millipore Corporation, Bedford, MA) and myosin was visualized using monoclonal antibodies (MF 20) followed by peroxidase-labeled anti-mouse IgG antibodies and the ECL kit as described by the manufacturer (Amersham Corporation, UK).
RESULTS
Evidence for the expression of the truncated EGFR in transfected clones
ES14 cells were electroporated in the presence of the linearized plasmid expression vector for truncated mouse EGF receptor (pK18-CXN-mEGFR) together with a neomycin-resistance plasmid. At least ten mutant and ten control clones were selected in medium containing 400 μg/ml G418 and these were expanded and frozen as stocks. Three of each were chosen for the analysis of expression of EGFR protein. The mutant clones were selected on the basis of their positive staining in immunofluorescence tests using rabbit anti-EGFR antibody as described in the Methods section (data not shown). To confirm that the clones selected expressed full-length and truncated EGFR proteins, we metabolically labeled the cultures and immunprecipitated EGFR protein from lysates using the same antibody. Wild-type EGFR protein (170×103Mr) is not detectable in undifferentiated ES cells but appears during differentiation. Fig. 2A shows that the truncated 120×103Mr EGFR was expressed in mutant ES clones and it increased 1.4-fold during RA-stimulated or spontaneous differentiation indicating the increased efficiency of the β-actin promoter in differentiated cells. Mutant EGF receptor protein was expressed from 5-fold to 20-fold more than endogenous wild-type receptor protein in mutant clones.
Biosynthesis of EGF receptor polypeptides in ES clones. Cells were metabolically labeled with [35S]methionine as described in the Methods section. Immunoprecipitation with rabbit EGF-receptor antisera (A,C) or with non-immune sera (B,D) were analyzed on 6% PAGE-SDS gels, which were then processed for autoradiography. (A,B) The results for three mutant clones analyzed on the day after the start of differentiation indicated above. Differentiation was induced by the addition of retinoic acid (RA) to aggregates in Petri dishes followed by culture for 5 days or differentiation occurred spontaneously (sp) by the removal of LIF from the medium. (C,D) The results for control clones. Only the mutant clones in A express the 120×103Mr exogenous EGF-receptor polypeptide while very low levels of the endogenous normal receptor protein was detected. In contrast, the control clones synthesized high levels of full length EGF-receptor in late differentiation shown in C.
Biosynthesis of EGF receptor polypeptides in ES clones. Cells were metabolically labeled with [35S]methionine as described in the Methods section. Immunoprecipitation with rabbit EGF-receptor antisera (A,C) or with non-immune sera (B,D) were analyzed on 6% PAGE-SDS gels, which were then processed for autoradiography. (A,B) The results for three mutant clones analyzed on the day after the start of differentiation indicated above. Differentiation was induced by the addition of retinoic acid (RA) to aggregates in Petri dishes followed by culture for 5 days or differentiation occurred spontaneously (sp) by the removal of LIF from the medium. (C,D) The results for control clones. Only the mutant clones in A express the 120×103Mr exogenous EGF-receptor polypeptide while very low levels of the endogenous normal receptor protein was detected. In contrast, the control clones synthesized high levels of full length EGF-receptor in late differentiation shown in C.
During differentiation a clear and notable difference was observed between the control and mutant cultures. The level of the wild-type EGFR was low (Fig. 2A, lanes 6-16) or absent (see clone Km25, Fig. 2A, lanes 1-5) in the mutant cultures while it was clearly increasing starting at 3 days after the start of RA-induced differentiation and reaching readily detectable levels that were maximal at about 9 days during control cell differentiation (Fig. 2C, lanes 1-18). The low production of EGFR protein in mutant clones was surprising since dominant negative mutated receptor molecules are not known to inhibit the synthesis of the wild-type receptor protein. Two possible explanations are: (1) that more rapid degradation of the wildtype form occurs after the formation of inactive dimers with the truncated polypeptide; (2) that cell death occurs in cells that express high levels of mutant receptor. The results described below support the second possibility.
Morphological appearance and growth of mutant cell lines
There was no detectable difference in the morphology of mutant ES cells (Km clones) compared to control cells (Kc clones). They grew at the same rates and were tightly clustered small cells when undifferentiated. This was expected because wild-type ES cells, similar to most EC cells, do not express surface EGFR and do not respond to EGF (Adamson and Hogan, 1984) and therefore they are not expected to behave differently in the presence of mutant receptor protein. After the removal of LIF, when endogenous EGFR synthesis is induced, the ES cell aggregates differentiated to give similar outer layers of endoderm in all clones. In other words, embryoid body formation appeared normal.
Tumour formation by ES cell clones
ES cells are tumourigenic when placed in subcutaneous locations in athymic or syngeneic animals. The tumours produced (teratocarcinomas) usually contain a wider range of recognizable differentiated cell types than during differentiation in vitro. Control and mutant cells (ten million) were injected into nu/nu mice to determine if the growth rate or the differentiated tissues produced in the tumours differed. Tumours grew at similar rates in both cell types to achieve a large mass after 2 weeks. Tumours produced from mutant clones were slightly smaller on average but this was not statistically significant because of the wide range in sizes (Table 1). Histological analyses (Fig. 3) revealed that there were differences in the predominant tissues appearing in the control and mutant tumours. The control cells produced tumours with more frequent and larger areas of highly differentiated tissues including cartilage (Fig. 3C). Epithelioid organs such as ciliated secretory epithelia (bronchial-like), cysts and glands lined by epithelial cells (Fig. 3D) were present. These tissues occurred in tumours from only one (Km 207) out of three mutant ES clones tested and was much less frequent, in smaller amounts and less-well-differentiated (Fig. 3A,B).
Tumour sections stained with Hematoxylin and Eosin. (A) Km25 and (B) Km 207; sections from mutant (mut) tumours display large areas of fibrous-like tissue, heavily infiltrated with blood cells. (C,D) Kc101; control tumours (con) have large areas of carlilage (c) and epithelia (e) some were columnar and resembled neuroepithelium (ne). Bar, 50 μm.
Tumour sections stained with Hematoxylin and Eosin. (A) Km25 and (B) Km 207; sections from mutant (mut) tumours display large areas of fibrous-like tissue, heavily infiltrated with blood cells. (C,D) Kc101; control tumours (con) have large areas of carlilage (c) and epithelia (e) some were columnar and resembled neuroepithelium (ne). Bar, 50 μm.
Further observations using higher magnification of H&E-stained sections revealed a striking difference in the composition of mutant tumours. Striated muscle was evident in many areas of the Km tumours: both skeletal muscle (Fig. 4B) and cardiac muscle (Fig. 4C) were found frequently while control tumours (Kc) had very little. Control tumours contained a predominance of epithelia with little if any muscle (Fig. 4C). Frozen sections of similar teratocarcinoma tumours examined by indirect immunofluorescence with an antibody to myosin heavy chain, confirmed this observation (data not shown).
High-power magnification of teratocarcinoma sections stained with H & E. (A) Control tumours (Kc) contained largely epithelial (arrows), fibroblasts and smooth muscle-like cells. (B) Mutant ES cell tumours were rich in striated myotubes (thick arrows) in which the thin and thick banding pattern typical of skeletal muscle was seen (small arrows). (C) Cardiac muscle (thick arrows) was also observed more frequently in Km tumours. Bar in A indicates 50 μm in all panels.
High-power magnification of teratocarcinoma sections stained with H & E. (A) Control tumours (Kc) contained largely epithelial (arrows), fibroblasts and smooth muscle-like cells. (B) Mutant ES cell tumours were rich in striated myotubes (thick arrows) in which the thin and thick banding pattern typical of skeletal muscle was seen (small arrows). (C) Cardiac muscle (thick arrows) was also observed more frequently in Km tumours. Bar in A indicates 50 μm in all panels.
In situ apoptosis tests (TUNEL assays) revealed differences in the frequency of apoptotic figures in tumours, with the highest level in the Km-derived cell tumours (Fig. 5, Mut) compared to control cell tumours (Kc 107, Fig. 5, Con). Km 207 tumours had the highest density of apoptosis of the Km tumours and this correlated with the presence of some differentiated tissues. Cell death in Kc tumours was present but occurred at a lower frequency. This suggested that rates of cell death may differ between the two cell types during differentiation and this could account for the observations described above.
In situ TUNEL assays for apoptotic cells. (A) Little apoptosis is seen (arrows) in Kc102 tumours (con). (B) Section of a Km207 (mut) teratocarcinoma showing numerous brown-stained apoptotic nuclei, singly and in groups largely in epithelial cells. Clusters of undifferentiated ES cells are seen in the upper half. Bar, 50 μm.
In situ TUNEL assays for apoptotic cells. (A) Little apoptosis is seen (arrows) in Kc102 tumours (con). (B) Section of a Km207 (mut) teratocarcinoma showing numerous brown-stained apoptotic nuclei, singly and in groups largely in epithelial cells. Clusters of undifferentiated ES cells are seen in the upper half. Bar, 50 μm.
Cell death in vitro
ES cells in aggregate cultures were analyzed for cell death and growth rates during differentiation induced by the withdrawal of LIF. Each experiment was performed at least twice with similar results and two clones of each type were examined. We analyzed growth rates (Fig. 6A) (cells not stained by Trypan blue) in cultures during 7 days of differentiation. This type of analysis in ES cells is subject to underestimates of the number of live cells because the disaggregation of cell clumps is incomplete and cells can be injured by extended enzyme treatments. Assuming that a similar error is applicable to all cell populations, the growth rates were about three-fold lower in mutant clones. The numbers of dead cells that accumulated in the medium in 5 days consistently showed that cell death was 2-fold more frequent in mutant clones compared with Kc clones (Fig. 6B).
Growth and death in ES cell cultures after the removal of LIF. (A) The number of cells, both adherent and suspended, that were Trypan blue negative was recorded over 7 days of differentiation. Cells were counted on the day indicated and normalized to the exact number (10,000) of cells seeded Triplicate samples, repeated once were used to generate averages and standard deviatons. See inset for cell types. (B) The accumulated dead cells after 5 days of culture (± s.d.).
Growth and death in ES cell cultures after the removal of LIF. (A) The number of cells, both adherent and suspended, that were Trypan blue negative was recorded over 7 days of differentiation. Cells were counted on the day indicated and normalized to the exact number (10,000) of cells seeded Triplicate samples, repeated once were used to generate averages and standard deviatons. See inset for cell types. (B) The accumulated dead cells after 5 days of culture (± s.d.).
Early markers of differentiation
Early stages of the differentiation of EC cells are often marked by the appearance of the secreted protease, tissue-type plasminogen activator (tPA), produced by parietal endoderm and other cells and by the presence of intermediate filament proteins, TROMA 1, TROMA 2 and TROMA 3 (K8, K18 and K19), also characteristic early markers of differentiation. We tested for these markers in differentiating ES clones in total RNA. In all three mutant clones examined, there was no evidence for tPA gene activation which is expected to occur on the 3rd day in control and wild-type clones (Fig. 7, lane 8). The approximately equal loading of the gel was ascertained by stripping the blot and reprobing for β-actin (Fig. 7, lower panel). A positive signal for tPA mRNA in RNA extracted from control cells at 3 days of differentiation was a consistent finding and occurred in other control clones in addition to those shown in Fig. 7.
Northern blot analysis of tissue-type plasminogen activator (tPA) expression in differentiating ES clones. Top panel shows that only control clone Kc18 started to express tPA on the 3rd day (lane 8) while mutant clones were not expressing detectable tPA at this time. To indicate that the gels were loaded with similar amounts of total (30 μg) RNA, the same blot was stripped and analyzed for the expression of β-actin.
Northern blot analysis of tissue-type plasminogen activator (tPA) expression in differentiating ES clones. Top panel shows that only control clone Kc18 started to express tPA on the 3rd day (lane 8) while mutant clones were not expressing detectable tPA at this time. To indicate that the gels were loaded with similar amounts of total (30 μg) RNA, the same blot was stripped and analyzed for the expression of β-actin.
Similarly, control clones produced outgrowths that expressed TROMA-1 and TROMA-3 cytoskeletal proteins one day earlier than mutant clones, appearing at 6 days in controls but not in mutant clones (Fig. 8). These markers are typical of parietal endoderm cells and all clones eventually produced similar staining levels, showing no difference in TROMA expression at later stages of differentiation. We concluded that the formation or activity of the primitive endoderm cells that give rise to visceral and parietal endoderm was delayed by the presence of mutant receptors. This conclusion was confirmed by northern blots probed with a cDNA to AFP, a visceral endoderm-specific marker. Fig. 9 shows that the control clone, Kc107, started to produce AFP after 5 days of aggregate culture in the absence of LIF (Fig. 9, lane 5). By 5 + 5 days in culture, a large amount of AFP mRNA was detected (Fig. 9, lane 6) whereas it took 10 days culture of mutant cells (Km27) before a low level of AFP was detected (Fig. 9, lane 3). Another early marker of pre-mesoderm tissues in mouse embryogenesis, Brachyury (Wilkinson et al., 1990), gave the opposite result. Brachyury (By) mRNA was strongly expressed in mutant differentiating ES Km27 cells by day 5 (Fig. 9, middle panel, lane 3) whilst control clones expressed barely detectable levels at all three stages examined. This result appears to indicate that muscle or other mesodermal tissues might be more predominant in mutant ES cell cultures.
Immunofluorescent analysis for the expression of keratin 19 (TROMA 3) in parietal endoderm cells produced from ES differentiation. Cultures were fixed on the 6th day of differentiation and stained with rat monoclonal antibody to TROMA 3. (A) Mutant clones (mut) were negative in this assay while the edges of colonies of B,C, control clones (con), were positive. Bar, 20 μm.
Immunofluorescent analysis for the expression of keratin 19 (TROMA 3) in parietal endoderm cells produced from ES differentiation. Cultures were fixed on the 6th day of differentiation and stained with rat monoclonal antibody to TROMA 3. (A) Mutant clones (mut) were negative in this assay while the edges of colonies of B,C, control clones (con), were positive. Bar, 20 μm.
Northern blot analysis of 20 μg total RNA extracted from mutant (Km) and control (Kc) ES cell clones. Cells were stimulated to differentiate in aggregate cultures by the removal of LIF from the medium for 5 days or for 5 days followed by 5d in tissue culture dishes. Labeled probes for marker gene products, α-fetoprotein (AFP), brachyury (By) and ribosomal protein (L32) were used. AFP production is delayed in mutant clones, while brachyury is elevated.
Northern blot analysis of 20 μg total RNA extracted from mutant (Km) and control (Kc) ES cell clones. Cells were stimulated to differentiate in aggregate cultures by the removal of LIF from the medium for 5 days or for 5 days followed by 5d in tissue culture dishes. Labeled probes for marker gene products, α-fetoprotein (AFP), brachyury (By) and ribosomal protein (L32) were used. AFP production is delayed in mutant clones, while brachyury is elevated.
Morphological appearance of differentiated cultures
The spontaneous differentiation of E14 ES cells after the removal of LIF is a slow process occurring over at least 14 days when beating cardiac muscle is observed as a dominant feature. Later cultures become enriched in matrix-containing areas of parietal endoderm cells and by frequent areas of myotubes. Dishes containing mutant cell lines (Fig. 10A) always contained fewer cells compared to control or wildtype cells (Fig. 10C). The other distinct difference was the earlier and increased appearance of actively beating cardiac-type muscle cells in the mutant cultures compared to control cultures after prolonged spontaneous differentiation (in the absence of LIF and RA). Fig. 10 illustrates some features of the differences observed in the phase-contrast microscope in mutant and control cultures. Skeletal muscle was observed in both types of cultures whether spontaneously or RA induced but there was always more in the mutant cultures in late differentiation stages. The most prominent feature of the mutant cell cultures were the frequent large colonies of small densely packed cells that resembled ES cells that had failed to differentiate (Fig. 10A,E,F).
Phase-contrast micrographs to compare the appearance of mutant (m) and control (c) clones in culture after various days of differentiation. (A,C) RA-induced differentiation for 4+9 days, endoderm and fibroblast-like cells predominate. (B,D) After 4+23 days of differentiation (RA-induced), mutant cultures are sparse (B) while control cultures (D) contain more fibroblast-like cells. (E,F) Mutant clones on day 5+59 of spontaneous differentiation, showing large areas of ES-like (es) cells and fusing myoblasts (mu). (G,H) Control cells on day 5+59 of spontaneous differentiation have areas with thyroid-like cells (th) and cartilage-like cells (ca). (I,J) Control cultures on day 5+53 of RA-induced differentiation have numerous areas of complexly organized structures that are absent from mutant cultures. Bar, 50 μm.
Phase-contrast micrographs to compare the appearance of mutant (m) and control (c) clones in culture after various days of differentiation. (A,C) RA-induced differentiation for 4+9 days, endoderm and fibroblast-like cells predominate. (B,D) After 4+23 days of differentiation (RA-induced), mutant cultures are sparse (B) while control cultures (D) contain more fibroblast-like cells. (E,F) Mutant clones on day 5+59 of spontaneous differentiation, showing large areas of ES-like (es) cells and fusing myoblasts (mu). (G,H) Control cells on day 5+59 of spontaneous differentiation have areas with thyroid-like cells (th) and cartilage-like cells (ca). (I,J) Control cultures on day 5+53 of RA-induced differentiation have numerous areas of complexly organized structures that are absent from mutant cultures. Bar, 50 μm.
Biochemical evidence for the altered course of late differentiation in mutant clones
We described above that the appearance and levels of mRNA for tPA in mutant clones was delayed for this early marker of differentiation. In contrast, the appearance of the brachyury marker that precedes mesodermal differentiation was markedly enhanced. As a later marker of mesodermal-type differentiation, the induction of the striated muscle myosin heavychain gene indicates the appearance of muscle cells in cultures. When RA was used to induce differentiation, the process was faster. In the experiment shown in Fig. 11, RA-stimulated differentiation is shown on the left and spontaneous differentiation on the right. In this experiment, we used serum-free medium during the later stages of culture to remove cells that were serum-dependent. The mutant cultures at harvesting were rich in beating cardiac cells. In support of this observation, we found that only the mutant cells and not the control or wildtype cells expressed high levels of the myosin gene after prolonged culture. Clone Km27 gave the strongest RA-stimulated myosin gene expression (Fig. 11, lane 11) while Km25 and Km207 produced highest levels of cardiac muscle during spontaneous differentiation (Fig. 11, lanes 14 and 16, respectively). We concluded that the presence of the truncated EGF receptor altered the outcome of the differentiation of ES cells in that muscle formation was frequent and this occurred at the expense of epithelium formation.
Immunoblot to demonstrate the presence of the myosin heavy chain in ES cell differentiated cultures after various days of differentiation. Lanes 1-6, control clones (Kc 106 and Kc107) after 0 and 15 days of RA-induced differentiation of which 6 days were in serum-free medium (SFM) and 26 days, the last 17days being SFM. Mutant clones in lanes 7-12 (Km 25, Km 207, Km 27) were harvested at the same times as control cultures as indicated above the panel. Lanes 13-16, various mutant clone cultures were harvested after spontaneous differentiation. Equal amounts of protein were analyzed by 6% SDS-PAGE, blotted and detected with a monoclonal antibody to chicken muscle myosin heavy chain. The myosin polypeptide gives a signal at about 200×103Mr which is stronger in mutant clones than in control or wild-type cultures (wt). Marker proteins (×10−3Mr) migrated as shown on the left.
Immunoblot to demonstrate the presence of the myosin heavy chain in ES cell differentiated cultures after various days of differentiation. Lanes 1-6, control clones (Kc 106 and Kc107) after 0 and 15 days of RA-induced differentiation of which 6 days were in serum-free medium (SFM) and 26 days, the last 17days being SFM. Mutant clones in lanes 7-12 (Km 25, Km 207, Km 27) were harvested at the same times as control cultures as indicated above the panel. Lanes 13-16, various mutant clone cultures were harvested after spontaneous differentiation. Equal amounts of protein were analyzed by 6% SDS-PAGE, blotted and detected with a monoclonal antibody to chicken muscle myosin heavy chain. The myosin polypeptide gives a signal at about 200×103Mr which is stronger in mutant clones than in control or wild-type cultures (wt). Marker proteins (×10−3Mr) migrated as shown on the left.
DISCUSSION
In this study of the effect of the over-expression of a truncated EGFR, we used a construct driven by the β-actin promoter together with the CMV enhancer (Niwa et al., 1991). It was predicted that this construct (Fig. 1) would be active in undifferentiated ES cells in contrast to the CMV promoter used in our previous study (Wu and Adamson, 1993) and this proved to be the case. The 120×103Mr truncated protein expressed, however, was the same from both constructs and was detected in immunofluorescence assays as exogenous receptor protein on the surface of the mutant clones derived by G418 selection (data not shown). There was no apparent effect of the expression of the mutant receptor in ES cells before differentiation. We conclude that the growth, morphology, survival and phenotype of the mouse ES cell does not depend on the EGFR. This is not surprising since the low level of receptor protein detected in ES and EC cells appears to be unreceptive since the application of EGF does not elicit a response (Weller et al., 1987).
The observed differences between mutant EGFR and control clones indicate that the expression of the kinase-negative mutant receptor affects the differentiation of ES cells. Differentiation is retarded as early as the 3rd day. The early marker, tPA, does not appear on the third day of differentiation, AFP does not appear on the 5th day of differentiation and TROMA 1 and TROMA 3 (keratins 8 and 18) do not appear on the 6th day of differentiation in mutant clones as they do in control clones. EGF is known to stimulate the production of tPA and urokinase in squamous cells (Niedbala et al., 1990) and this could explain the slow production of tPA in differentiating mutant ES clones.
Several effects of the lack of active EGFR in mutant cells likely contribute to the retardation of differentiation and to the reduced expression of endogenous EGFR. Normally, as ES cells differentiate in the absence of LIF, EGF (and other) receptors appear and cells are able to produce a proliferative signal to the nucleus in response to growth factors in the serumrich medium. Cells that lack this response are less able to survive and also grow at slower rates (Fig. 6A). Cell death assays support this hypothesis, because mutant cells produce 2-fold more dead cells in the medium than control cultures during the first 5 days of differentiation (Fig. 6B). Apoptotic nuclei seen in EC and ES cell aggregates during cystic embryoid body formation, is part of embryonic programmed cell death (Coucouvanis and Martin, 1995). At later stages of differentiation, we were able to show that apoptosis is more frequent in mutant cells, using the TUNEL procedure in teratocarcinomas (Fig. 5). The prevention of apoptosis by EGF has been previously described for ovarian granulosa cells (Tilly et al., 1992) and in kidney development (Koseki et al., 1992; Coles et al., 1993). Death of cells with inactive EGF-receptors is one possible mechanism for the low expression of endogenous EGF-receptors in differentiated mutant ES cultures. Slower proliferation rates will compound this effect so that the population of mutant cells have lower levels of normal EGFR. The net result of negative selection of certain cell types during ES cell differentiation is the skewing of differentiation pathways caused by the limited capacity of the surviving EGF-receptor-poor cells.
The normal response to EGFR signalling is the gradual accumulation of transcription factors that (a) ensure survival, (b) maintain cell cycling and (c) commit the cells towards differentiation. In the absence of the signals provided by EGFR, the pathways of differentiation are altered. Fig. 12 summarizes the results obtained and the conclusions that we made. We have shown that very low levels of 170×103Mr receptor protein are produced by mutant clones (Fig. 2) compared to control clones especially during RA-induced differentiation. This results in the increased survival of EGFR-independent cells such as those that express brachyury (Fig. 9) followed later by the production of higher levels of muscle myosin in mutant cultures (Fig. 11). EGFR-dependent cell types such as cartilage and epithelial tissues are scarce in mutant cultures and tumours (Figs 3,4) compared to controls. We conclude that the predominant surviving cells are those that need little or no EGFR to differentiate and function. Cardiac and skeletal muscle are such tissues. In the case of skeletal muscle, it has been documented that, although EGFR protein is present and functional on myoblast cells during the proliferation stages, the receptors are not essential to myoblast cell proliferation and they become inactive after fusion of the cells into myotubes (Olwin and Hauschka, 1988; Lim and Hauschka, 1994). Our results indicate that cardiac muscle can be added to the list of tissues that do not require EGF-receptor expression to proliferate, to differentiate or to function in contractility.
Summary of ES cell differentiation. LIF, Leukemia Inhibitory Factor; AFP, α-fetoprotein; Tr1, TROMA-1; Tr3, TROMA-3; tPA, tissue-type plasminogen activator; By, brachyury; MHC, myosin heavy chain.
EGF-receptor protein expression is lost by the ES cell adapted to in vitro culture but is present in the inner cell mass (ICM) in the blastocyst in vivo (Wiley et al., 1992; Dardik et al., 1992). Indeed, the function of the receptor has been shown to be essential to the implantation of the blastocyst and the survival of the ICM in vivo in some strains of the EGFR-null mouse (Threadgill et al., 1995). Microinjection of the mutant EGFR construct used here into >200 zygotes (C57Bl/6×SJL F2 hybrids) and subsequent transfer of embryos into the uteri of recipient animals yielded only 21 pups and none were positive for the transgene (data not shown). Therefore, in vivo, the mutant construct in preimplantation embryos appears to produce a lethal behaviour, much like the CF-1 EGFR-null embryos of Threadgill and colleagues.
How does the exogenous mutant EGF-receptor behave in a cell with normal endogenous receptor? Previous studies (Boni-Schnetzler and Pilch, 1987; Cochet et al., 1988) have shown that the activity of the wild-type EGF-receptor is mediated by ligand-stimulated receptor dimerization and that stimulation of the tyrosine kinase activity leads to the pleiotropic responses of the cell. Several groups have described the dominant-negative effect of a mutant receptor dimerizing with and inactivating signal transduction from the wild-type receptor. Recently, evidence was presented for some types of signalling to proceed from normal receptor molecules even in the presence of kinase-negative mutant molecules (Honegger et al., 1987; Kashles et al., 1991). It is possible that a mutation is less dominant if the dimerizing receptor molecules are from different species as in the cited studies, or varies according to the mutation site and its extent within the kinase domain. Another consideration must be given to the possibility of the diversity of the fates of the dimers and how this is affected by the ratio of the two partners. In the ES cells described here, the expression of the truncated receptor (amino acids 1-689) is always in excess of the endogenous wild-type receptor molecule. We know that the exogenous receptor is expressed on the cell surface and that differentiated P19 cells expressing the mutant receptors are incapable of forming tyrosine-phosphorylatable dimers in response to EGF while control cells can do so (Wu and Adamson, 1993). Therefore, it is likely that in ES cells the excessive levels of truncated receptors will also inactivate the endogenous receptors, although this could not be demonstrated because of the paucity of normal EGF-receptor protein in these cells. We do not know the precise fate of the EGF-receptor molecules in mutant clones but we presume that heterodimers of wild-type and mutant proteins will form as well as homodimers of mutant receptors.
The truncated receptor molecule contains the sequences 647 to 688 (including the transmembrane domain), shown to be needed for internalization and turnover (Wiley et al., 1991 and H. S. Wiley, personal communication) but it does not have the more carboxy-terminal sequences needed for lysosome targeting (French et al., 1994) and mitogenic signalling. We believe that the truncated receptor will go to the default pathway and be recycled to the surface to be effective again in preventing the responses of the wild-type receptor. This type of receptor lacks SH2 domains and is unlikely to be able to act as an absorber of signalling components needed for other growth factor pathways. There is a possibility, however, that the truncated mutant EGF-receptor may form dimers with other members of this receptor family, mErbB-2, mErbB-3 and mErbB-4. It is possible that the protein products of c-ErbB-2 and c-ErbB-3 have functions that overlap with those of the EGFR that may compensate for the loss of the EGFR protein. However, we have been unable to detect any ErbB-2 or Erb-B3 transcripts in differentiating ES cells by northern blotting (data not shown) and so this seems unlikely. In any case, truncated EGF receptor should also inactivate other ErbB receptors by heterodimerization.
The results of this study indicate that the presence of mutant EGF-receptor is injurious to the survival, growth and differentiation of ES cells. In the beginning, the expansion of selected populations of committed cells might be compromised as in inner cell mass cells in EGFR null mutant mice (Threadgill et al., 1995) while later, the course of differentiation is altered. The end result of the activities of the dominant-negative EGF-receptor in ES cells is differentiation to lineages that do not require EGFR for their production. We have identified cardiac muscle as one of these while skeletal muscle was already known not to require EGF receptors (Olwin and Hauschka, 1988; Lim and Hauschka, 1994). We were unable to detect any nerve cell differentiation in the E14 ES cell control clones and so the formation of this tissue type in the presence of mutant receptors could not be tested. In P19 cells, however, RA-induced differentiation to nerve and glial cells was inhibited by the dominant-negative mutant receptor protein (Wu and Adamson, 1993) and supports the idea that EGF-receptors are required for differentiation to nervous tissues.
The results of this study also serve to underline some differences between ES cells in culture and ICM cells in the blas-tocyst and emphasize the need for in vivo studies to determine how the maternal environment and the time and spatial demands of the developmental process in vivo are regulated by the activities of the EGF-receptor. The ES cell, however, provides a simplified model system in which to test the cellular and physiological effects of mutant genes in differentiating embryonic cells in vitro.
ACKNOWLEDGEMENTS
We thank Dr S. Krajewsky for the TUNEL assays and Mr. M. Hasham for photographic services. Excellent technical help was provided by W. Matheny and D. Okamura. We thank Drs R-P Huang, C. Niemeyer and D. Mercola for critical comments on the manuscript and Dr Mercola for histological services and analyses. The β-actin expression vector (pCXN2, Niwa et al., 1991) was kindly provided by Dr K. Ozato. This work was supported by grants from the Public Health Service, CA 28427 and P30 CA 30199.