Teratocarcinomas form a subset of germ cell tumors in which a wide array of differentiated tissues and pluripotent embryonal carcinoma (EC) stem cells are mixed together. These tumors appear to recapitulate the processes of early embryogenesis, albeit in a disorganized way.1, 2 It is widely recognized that embryonal carcinoma (EC) cells are the stem cells of teratocarcinomas, and studies in the laboratory mouse have shown that EC cells are closely related to embryonic stem (ES) cells derived from the blastocyst shortly after fertilization.3–5 Recently, this conclusion has been confirmed in humans with the demonstration that EC cells derived from human testicular teratocarcinomas and ES cells isolated from early human embryos, produced by in vitro fertilization, share common features.6, 7 At the same time, human ES and EC cells differ from their murine counterparts, suggesting that results with the laboratory mouse cannot necessarily be extrapolated directly to human development.8
While teratocarcinomas present a particularly dramatic form of germ cell tumor, some such tumors contain only EC cells without obvious signs of somatic differentiation. It is also evident that the differentiated tissues of teratocarcinomas are typically non-malignant, or have a limited growth capacity, and the malignancy of these tumors is conferred primarily by the EC cell component.9, 10 Since the differentiation of EC cells appears to result in a loss of malignancy, it is easy to imagine that mutations that limit the capacity of EC cells to differentiate could provide a selective advantage for the development of more aggressive cancer. Cell differentiation is a feature of many other cancers, though perhaps in less dramatic form, so that knowledge of the mechanisms by which stem cells can lose their capacity for differentiation is a key element in understanding tumor progression.
A number of cell lines have been derived from human testicular germ cell tumors and express features typical of human EC cells.11 A few of these EC cell lines are able to undergo extensive differentiation. For example, NTERA2, a derivative of TERA2, forms well differentiated tumors in nude mice.12 In culture, NTERA2 cells also differentiate extensively in response to retinoic acid, yielding neurons and other cell types that have a limited life span.13–15 However, many other established human EC cell lines do not show a significant capacity for differentiation and do not differentiate in response to retinoic acid.16 For example, 2102Ep is a well-characterized, prototypical human EC cell line that forms xenograft tumors comprising only pure embryonal carcinoma with no other differentiated cell types in athymic (nu/nu) mice.17, 18 In culture, 2102Ep cells do not differentiate when exposed to retinoic acid,16 although they undergo a rather limited form of morphological differentiation with some antigenic changes when grown at low cell density.18
Among EC cell lines derived from teratocarcinomas of the laboratory mouse, some similarly retain a capacity for differentiation, while others do not.19 A number of studies have been made of hybrids produced by the fusion of murine EC cells and differentiated somatic cells taken from adult mice, or from established murine cell lines (e.g., references 20–26). In some cases these hybrids retained an EC cell phenotype; in others it was extinguished, depending upon the particular cell combination used. Several authors also reported that hybrids of the nullipotent murine EC cell line F9 and either pluripotent EC cells or somatic cells were able to differentiate extensively, indicating that the nullipotency of F9 cells is due to the loss of function of a key gene, or genes, perhaps by mutations that could be complemented by wild-type alleles introduced into the hybrids from the pluripotent EC cell or the non-EC cell parent.27–29 A similar observation was made in the case of PCC4 by Andrews and Goodfellow.23
To test whether nullipotent human EC cells fail to differentiate because of such a loss of function, or because they have acquired a mutation that actively inhibits differentiation, we have now investigated the properties of hybrids of the 2 human EC cell lines, 2102Ep and NTERA2, both derived from testicular germ cell tumors. The results suggest that the failure of 2102Ep cells to respond to retinoic acid,10 and their failure to differentiate in xenografts17, 18 is also due to the loss of function of a key gene(s) rather than the gain of function of an inhibitor of differentiation. On the other hand, the hybrid cells do not appear to differentiate into neurons, as do the parent NTERA2 cells,12 suggesting that 2102Ep cells may also have acquired inhibitory mutations relevant to this specific pathway of differentiation.
MATERIAL AND METHODS
Cultures of NTERA2 cl.D112, 30 and 2102Ep cl.4D318 human EC cells were maintained at high cell densities (>5×106 cells per 75 cm2 tissue culture flask) and 37°C in Dulbecco's Modified Eagles Medium (DMEM) (high glucose formulation) (GIBCO Life Technologies, UK), supplemented with 10% fetal calf serum (FCS), under a humidified atmosphere of 10% CO2 in air, as previously described. NTERA2 cells were harvested for subculture by scraping with glass beads;12 2102Ep cells were harvested using 0.25% trypsin in calcium and magnesium-free Dulbecco's phosphate-buffered saline (PBS), containing 1 mM EDTA (trypsin:EDTA).13 To induce differentiation, cells were harvested by using trypsin:EDTA and seeded at 106 cells per 75 cm2 tissue culture flask in medium containing 10−5 M all-trans-retinoic acid (Eastman-Kodak, Rochester, NY).13
Selection of a 6-thioguanine (6TG), neomycin-resistant subline of NTERA2 (NTERA2 cl.D1.TG11.nR, abbreviated TG11nr)
NTERA2cl.D1 cells were plated at 107 cells per 100 mm tissue culture dish in medium containing 1μg/ml 6-TG (Sigma-Aldrich, Poole, Dorset). After 2 weeks, surviving colonies were picked, expanded and tested for sensitivity to HAT (hypoxanthine, aminopterin and thymidine) selective medium (Sigma-Aldrich, Poole, Dorset). Two sublines sensitive to HAT were selected, NTERA2 cl.D1.TG5 and NTERA2 cl.D1.TG11, which closely resembled the parental NTERA2 line in morphology and expression of surface antigens typical of human EC. NTERA2 cl.D1.TG11 cells were subsequently transfected with a plasmid, containing the bacterial neomycin-resistance gene under the control of the SV40 immediate early promoter, pBKCMV-phagemid (Stratagene Europe, Amsterdam). Briefly, 10 μg purified pBKCMV DNA were incubated with 100 μg LipofectAMINE (GIBCO Life Technologies, UK) in 2 ml Optimem-I reduced serum medium (GIBCO Life Technologies, UK) at room temperature for 30 min. The resulting DNA:liposome complex was mixed with 2 × 107 cells in 10 ml Optimem, and incubated at 37°C for 4 hr, with gentle agitation. Subsequently, the cells were plated at 2.5 × 106 per 100 mm tissue culture dish in DMEM, to which G418 (Geneticin®, GIBCO Life Technologies, UK) (400 μg per ml) was added after 24 hr. After 2 weeks, surviving colonies were picked, expanded and maintained in medium containing 400 μg per ml G418. From these, 1 subline (NTERA2 cl.D1.TG11.nR., abbreviated TG11.nR) that closely resembled the original NTERA2 cells in morphology, growth pattern and surface antigen phenotype was chosen for use in further experiments.
Parental cells were harvested from stock cultures using trypsin:EDTA, and washed in serum-free DMEM containing 10 mM HEPES (DMEM-H). Approximately, equal numbers of the 2 parental cells (4 × 106) were mixed and pelleted by centrifugation. One ml of pre-warmed (37°C) 50% w/v polyethylene glycol 1500 (PEG 1500) in 75 mM HEPES (Boehringer Mannheim UK, Lewes, East Sussex) was added over 1 min to the pellet of mixed cells which were agitated gently to ensure suspension. After incubating for a further minute at 37°C, 1 ml of pre-warmed DMEM-H was added over 1 min, followed by 8 ml of DMEM-H over a further 3 min, with gentle agitation. The cells were then pelleted by centrifugation, resuspended and distributed between the wells of a 12-well tissue culture plate in DMEM supplemented with HAT (Sigma-Aldrich, Poole, Dorset), 400 μg/ml G418 and 20% FCS. After 2 weeks, surviving cells from a single well were harvested and replated into a 75 cm2 flask, also in DMEM supplemented with HAT and G418. Subsequently, single, independent colonies were picked and expanded for analysis.
To obtain chromosome preparations for analysis, colcemid (GIBCO Life Technologies, UK) was added to subconfluent cultures for a period of 5–7 hr (final concentration 0.2 μ/ml). The cells were then harvested by using trypsin:EDTA, exposed to hypotonic treatment with 0.0325M KC1 for 10 min and resuspended in a fixative (methanol:acetic acid, 3:1). Metaphase spreads were prepared according to a routine laboratory protocol, and chromosomes were G-banded after 10 min incubation at 80°C.31 From each cell line, 10–20 metaphases were counted and at least 4 G-banded metaphases were analyzed. Images were captured digitally using the CytoVision System (version 4.1) (Applied Imaging, Newcastle, UK).
SCID mice were injected subcutaneously with approximately 107 cells in 0.1 ml PBS. The mice were monitored weekly and when tumors were evident (approximately 1 cm diameter), they were killed and the tumors excised. Pieces of tumor were fixed in Bouin's fixative for histology.
Cell surface antigen expression was determined by indirect immunofluorescence, assayed as previously described,13, 15, 17, 18 using a Coulter Epics Elite, ESP flow cytofluorimeter. The following monoclonal antibodies to selected cell surface antigens were used: MC631, recognizing Stage Specific Embryonic Antigen-3 (SSEA3);32 MC813-70, recognizing SSEA433 and A2B5, recognizing ganglioside GT3.15, 34 Antibody from the parent myeloma cell line P3X63Ag8 was routinely used as a negative control.18 Expression of neurofilament proteins was detected by immunofluorescence staining of acetone-fixed cells as previously described,13 using monoclonal antibody 7H11 that recognizes the 200 kDa neurofilament-protein.35
Total RNA was isolated from selected cells using Tri-reagent (Sigma Aldrich, Poole, Dorset, UK). DNA contamination was removed with DNA-free (Ambion, Austin, TX). Each (5 μg) RNA sample was denatured and annealed with 100 pmol of oligo (dT) primer, and then reverse transcribed for 2 hr at 37°C in a 40 μl reaction mixture containing 200 units of M-MLV RT (Promega), 1.25 mM dNTPs and reaction buffer (final concentrations: 75 mM KCl, 50 mM Tris-HCl, 3 mM MgCl2 and 10 mM DTT).
Primers were designed against genes of interest using the PRIMERSELECT package from the DNASTAR (London, UK) suite of programs. PCR was performed using 0.25 μl of the RT reaction mix, added to a 50 μl reaction volume containing 15 pmol of each primer, 100 μM dNTPs, 2.5 mM MgCl2, 2 units of Taq polymerase (Promega) and reaction buffer (final concentrations: 50 mM KCl, 10 mM Tris-HCl and 0.1% Triton X-100). Thermal cycling was carried out according to the protocol: [(95°C 1 minute, TA 1 minute, 72°C 1 minute) ×35 cycles; 72°C 15 minutes ×1 cycle], where TA = 66°C for neuroD1, 58°C for RARβ, 60°C for β-actin, Wnt13 and RARγ and 65°C for collagen II and Mal. The primer sequences used are shown in Table I.
Table I. Forward (f) and Reverse (r) Primers Used for PCR Analysis of Gene Expression During Differentiation of EC Cell Hybrids
Primer sequence (5′ to 3′)
Expected product size (bp)
Genebank accession number
Primer location (bp)
Collagen II f
Collagen II r
Cells of the 6TG-resistant, neomycin-resistant derivative of the NTERA2 cl.D1 human EC line (TG11.nR) all died when cultured in HAT medium, with no evidence of reversion (reversion frequency ≫ 10−7). Similarly, 2102Ep cl.4D3 cells all died when cultured in the presence of 400 μg/ml G418. From a fusion between TG11.nR and 2102Ep, 4 colonies resistant to both G418 and HAT were obtained and so were considered putative hybrids. Of these, 2 (C4 and C17) proved to have karyotypes with chromosome numbers similar to those of NTERA2 and 2102Ep, and were not studied further (data not shown). However, 2 others (C7 and C10) contained approximately a double complement of chromosomes and were subject to detailed analysis.
Metaphase spreads of TG11.nR cells (not shown) had a mode of 59 chromosomes, with a range of 57–60 chromosomes. No normal sex chromosomes were evident. TERA2 (from which this cell was originally derived) showed a similar modal chromosome number, but all metaphase spreads contained a single normal X and Y chromosome, as previously reported.12 TG11.nR exhibited trisomy for chromosomes 2, 3, 7, 9, 12, 18, 20 and 21 in at least a proportion of cells, most of which were noted in TERA2. In addition, a number of chromosomal abnormalities present in differing proportions of cells resulted in partial trisomy or tetrasomy for several other chromosomal regions. Of particular note, the abnormalities included an isochromosome 12p, known to be associated with tumors of this type. Other abnormalities included at least 3 rearrangements, involving chromosome 1, and additional material on the short arms of chromosomes 8, 10, 11 and 19 and on the long arms of chromosome 13. Rearrangements of chromosome 1 and the derivative chromosome 13 have been described previously,12 but chromosomal evolution was clearly evident. Five to 8 markers of unknown chromosomal origin were also noted.
2102Ep cells exhibited a modal, 58, XX karyotype with a range of 56–59 chromosomes. All of the metaphase spreads examined contained 2 normal X chromosomes as previously reported36 but no Y-chromosome. Three copies of an isochromosome 12p were present in most cells. Other chromosomal rearrangements included a der(7)t(7;9)(q22; q13) and additional material present on the short arms of chromosomes 8, 14 and 15 and on the long arm of chromosome 16. Extra copies of the short arms of chromosome 1 and 3 were also present.
Metaphase spreads of C10 contained a range of 94–112 chromosomes, with a modal number of 100, whereas those of C7 had a modal chromosome number of 106 with a range of 86–114. Marker chromosomes from TG11.nR and 2102Ep were evident in both C7 (Fig. 1) and C10 (Fig. 2), confirming that they were indeed hybrids derived by fusion of those 2 parental cells. Some additional marker chromosomes that were not present in either parent line were also seen, suggesting continued karyotypic evolution in the hybrids.
When cultured at high densities, 2102Ep cells18 tend to grow in tight uniform clusters and to present the typical morphology of human EC cells, with pale nuclei, prominent nucleoli and little cytoplasm (Fig. 3a). Cells with a similar morphology could also be found in cultures of TG11.nR, as with the parental NTERA2 cells,12 though often the cells in the clusters were less tightly packed, and more than a single nucleolus was often apparent (Fig. 3b). The latter, less homogeneous morphology is most likely a consequence of the NTERA2 and TG11.nR EC cells, tending to undergo limited spontaneous differentiation, even at high cell densities. High density cultures of both C7 and C10 also contained mostly cells that were morphologically consistent with an EC phenotype, though their overall appearance was more closely similar to TG11.nR than 2102Ep cultures (Fig. 3c,d).
Like NTERA2 cells, TG11.nR cells differentiated extensively in response to retinoic acid, so that few, if any, morphologically identifiable EC cells were present in cultures a week after initial exposure and cells resembling neurons were evident after 2–3 weeks (Fig. 3e). By contrast, as anticipated from previous studies,16 2102Ep cells did not differentiate morphologically and continued to grow similarly in the presence or absence of retinoic acid (not shown). However, retinoic acid induced both C7 and C10 cells to undergo extensive morphological differentiation, like TG11.nR cells though neuronal-like cells were not observed (Fig. 3f,g).
To further test the phenotype of the hybrid cells before and after retinoic acid induction, the expression patterns of 2 characteristic EC cell marker antigens, SSEA3 and SSEA4, were analyzed by immunofluorescence (Fig. 4). Indeed, both C7 and C10 cells expressed SSEA3 and SSEA4, like 2102Ep and TG11.nR EC cells, though the percentage of SSEA3(+) cells was somewhat lower in the hybrids. Such a result would be consistent with the occurrence of partial spontaneous differentiation, and we have previously noted rather greater variability in expression of SSEA3 by NTERA2 cells, in contrast to 2102Ep, presumably reflecting the same phenomenon.11 Following 6 days exposure to retinoic acid, the expression of both antigens was markedly diminished in TG11.nR, C7 and C10, but not in 2102Ep. At the same time, the expression of the glycolipid antigen, ganglioside GT3, detected by antibody A2B5,15, 34 was induced in TG11.nR, C7 and C10, but not in 2102Ep. After extended exposure to retinoic acid (30 days), most TG11.nR cells remained SSEA3(−) and SSEA4(−), and many continued to express A2B5. By contrast, many cells expressing SSEA3 and SSEA4 reappeared in C7 and C10 cultures though, at least in C10, cells continuing to express the A2B5 antigen also persisted. Thus, a proportion of the hybrid cells either failed to differentiate, or reverted to an EC phenotype, and overgrew the differentiated cells during extended culture. Such extended culture of 2102Ep cells in retinoic acid was not possible since, in the absence of differentiation, the EC cells rapidly overgrew the culture.
Neural differentiation is a marked feature of NTERA2 differentiation,12 and the presence of neurons in retinoic acid induced cultures of TG11.nR cells after 2–3 weeks, was confirmed by immunofluorescence staining with antibodies to neurofilament proteins (Fig. 5a). By contrast, we did not observe neurons in either hybrid line, although rare cells in retinoic acid-treated cultures of C10 did express cytoplasmic filaments that stained with antibody to the high molecular weight neurofilament protein but such cells did not exhibit a neuronal morphology (Fig. 5b). Such non-neuronal neurofilament-positive cells are also observed in NTERA2 cultures.15
To examine further the ability of the hybrids to differentiate, approximately 107 cells of each were injected subcutaneously into each of 2 SCID mice. No tumors were obtained with C7, but one of the 2 mice injected with C10 cells developed a xenograft tumor. Histological examination of this C10 xenograft tumor revealed extensive differentiation (Fig. 6a,b) similar to that previously seen in NTERA2 xenografts,12 but not in 2102Ep xenografts.17, 18 Of particular note were glandular structures, which are generally characteristic of NTERA2 xenografts, but we also found extensive nodules of cartilage, which we had not previously noted in NTERA2 tumors. However, examination of a TG11.nR xenograft also revealed a single cartilage nodule, as well as the characteristic glandular structures of NTERA2 xenografts (Fig. 6c).
To confirm that the hybrid cells retained the capacity of the NTERA2 cells to differentiate, we used RT.PCR to examine the expression of several developmentally-regulated genes in 2102Ep, TG11.nR and C10 cells before and after retinoic acid induction (Fig. 7). Thus, Wnt13, which is induced by retinoic acid in NTERA2 cells,37 was similarly induced in TG11.nR and C10 cells, but not in 2102Ep. However, Mal and neuroD1, which are induced in NTERA2 cell38,39 and were also induced in TG11.nR, were not induced in either 2102Ep or C10 cells. Since both genes are characteristic of the neural differentiation seen in NTERA2, their absence in C10 correlates with an absence of overt morphologically identifiable neurons after retinoic acid induction. On the other hand, we also examined the expression of collagen II, as a marker of cartilage differentiation.40 In this case, no expression was seen in 2102Ep cultures, but it was induced in cultures of both NTERA2 and TG11.nR after induction with retinoic acid, suggesting that cartilage differentiation may be induced by retinoic acid in these cells in vitro, as well as by growth in xenografts. However, it was notable that collagen II was also readily detected in the untreated C10 cells, as well as after retinoic acid induction, suggesting not only that the hybrids differ from the 2102Ep parental cells in this respect, but that such differentiation occurs spontaneously without retinoic acid exposure necessary for such differentiation in NTERA2 and TG11.nR cells. Finally, we examined the expression pattern of a gene, RARβ, that is directly regulated by retinoic acid, and also RARγ, another member of the retinoic acid receptor family. RARβ was induced in TG11.nR and C10 cells by exposure to retinoic acid and, as previously reported,41 it was also induced in 2102Ep even though those cells do not otherwise differentiate in response to retinoic acid. RARγ, which has been reported to mediate the differentiation response of human EC cells to retinoic acid, was readily detected in all cells before and after retinoic acid induction. Thus, the failure of 2102Ep cells to differentiate in response to retinoic acid is evidently not due simply to an absence of a retinoic acid response system.
In addition to resistance to selective conditions (HAT and G418) that killed both parental lines, the simultaneous presence of rearranged marker chromosomes derived from each parent confirmed that C7 and C10 were indeed hybrids of 2102Ep and TG11.nR EC cells. As might be anticipated, these hybrids retained an EC phenotype, as indicated by morphology and surface antigen phenotype, despite the presence of close to 100 chromosomes. It has been proposed that human germ cell tumors arise from a tetraploid precursor and progress from a transformed germ cell, or seminoma, phenotype to an EC cell phenotype by the loss of chromosomes so that, whereas seminoma cells typically contain a 3n-4n DNA complement, EC cells typically contain an approximately 3n DNA complement.42 However, presumably this transformation in phenotype requires the loss of specific genes during tumor progression, so that merely doubling the DNA content in the hybrid is not sufficient to reverse their EC phenotype.
It is further evident that both C7 and C10 cells can differentiate in culture in response to retinoic acid, and C10 can differentiate in xenografts. Thus, the failure of 2102Ep EC cells to differentiate appears to be a recessive trait, most easily explained by the loss of function of a key gene, or genes, that can be complemented by a wild-type allele(s) retained in the genome of pluripotent TG11.nR EC cells. The combined modal chromosome number of 2102Ep and TG11.nR cells was 117, compared with the 106 and 100 observed in C7 and C10 hybrid cells, respectively. Thus, in both hybrids, the loss of a small number of chromosomes was evident, as is generally common in intra-specific hybrids, and so we cannot definitively exclude the possibility that a mutant gene actively inhibiting differentiation of 2102Ep cells was lost in each case. However, the loss of chromosomes in such hybrids is likely to be random and there was no selection for retention of a multipotent phenotype; indeed, any unintentional selection would more likely have been in favor of a nullipotent phenotype, since such cells would most likely be easier to grow. The conclusion that the loss of an ability to differentiate is consequent upon the loss of a key gene function is also consistent with the enhanced pluripotency of those mouse EC x somatic cell hybrids that retain an EC cell phenotype.23, 27–29 Furthermore, a hybrid of 2102Ep human EC cells and the mouse EC cell line PCC4 also retained the differentiation potential of the mouse EC parent, though in this case considerable loss of the human chromosomes from 2102Ep occurred.43
The nature of the function that is required for pluripotency and lost in 2102Ep cells is not known. An obvious possibility would be that the cells have lost the ability to respond to retinoic acid, perhaps by not expressing a retinoic acid receptor. Several genes encode retinoic acid receptors, including those of the RAR and RXR families, but the precise role of the different members of these families in mediating retinoic acid responsiveness during embryogenesis remains unclear, though substantial redundancy is evident.44 In the case of human EC cells, a particular role for RARγ in mediating retinoic acid-induced differentiation has been posited, and reduced levels of expression in 2102Ep compared with NTERA2 EC cells have been reported.45, 46 Kitareewan et al.47 also reported that RARγ selective agonists, but not RARα, RARβ or RXR agonists were able to induce the differentiation of NTERA2 cells. On the other hand, we have readily detected, by RT.PCR, RARγ expression in our cultures of 2102Ep, as well as TG11.nR cells (Fig. 7). However, perhaps more telling, is our confirmation (Fig. 7) that RARβ, which subject to direct induction by retinoic acid through a cis-acting retinoic acid response element,48, 49 is inducible in 2102Ep cells as previously reported by Miller et al.41 as well as in TG11.nR and the C10 hybrid cells. Therefore, 2102Ep cells clearly retain a retinoic acid responsive system. It is also notable that the C10 hybrid and TG11.nR cells, but not 2102Ep, differentiate in xenograft tumors, a process that does not obviously involve retinoic acid, while in culture the hybrids seemed prone to some degree of spontaneous differentiation in the absence of retinoic acid, for example, indicated by the expression of collagen II. Thus, it seems likely that the block to differentiation of 2102Ep cells lies downstream to the action of retinoic acid, perhaps at the level of HOX gene induction. Previously, we have noted that the HOX genes which are strongly expressed during retinoic acid induction of NTERA2 EC cells, are not induced by retinoic acid in 2102Ep cells.50 Although the regulatory elements of some HOX genes that are induced early contain retinoic acid response elements, the extended time course of most HOX gene activation in NTERA2 cells51 suggests that their regulation involves mechanisms subsequent to direct retinoic acid activation.
Despite the ability of the hybrids to differentiate, it is also evident that they do not behave identically to the TG11.nR parental cells. This was most marked by the apparent absence of neural differentiation. Few, if any, morphologically identifiable neurons were seen in retinoic acid-induced C10 cultures, in contrast to TG11.nR, and the rare neurofilament-positive cells observed did not exhibit a neuronal morphology; such cells are also seen in NTERA2 and TG11.nR cultures but their identity is obscure.15 Genes associated with neural differentiation, neuroD139 and Mal38 were not induced. Of these, neuroD1 is particularly striking as a marker for committed post-mitotic neuroblasts, the precursors of the mature neurons during both embryogenesis52 and the differentiation of NTERA2 EC cells in culture.39 Thus, such neuroblasts are apparently absent from differentiating C10 cultures, and the block to neural differentiation in these cells must be at a relatively early stage. These results suggest that the 2102Ep parental cells contribute an inhibitory mutation, specifically active during neural differentiation, to the hybrids. Such a mutation could lead either to a gain-of-function, or to a failure to downregulate a gene that normally restricts differentiation of committed neural stem cells, for example, a member of the Notch family.53, 54
On the other hand, cartilage differentiation was especially evident in the C10 xenograft and also in culture, as evidenced by the expression of collagen II. We had not previously noted cartilage differentiation in NTERA2,12 but some cartilage differentiation was seen in TG11.nR xenografts, and collagen II was detected in retinoic acid induced NTERA2 and TG11.nR cultures. In fact, although collagen II is strongly induced during the early stages of cartilage differentiation during embryogenesis, its expression is not restricted to cartilage, and it is also expressed by various other mesenchymal tissues, as well as in some parts of the developing nervous system.40 In the light of the clear absence of neural differentiation in C10 cells, it seems unlikely to be a marker of neurogenesis in this case, but it is consistent with mesenchymal differentiation which has been previously noted in NTERA2.12 It certainly appeared that this pathway was enhanced in C10 in comparison to the TG11.nR parent cells, perhaps by complementation with the 2102Ep genome, or perhaps as a compensation for the lack of neural differentiation. However, definitive quantitative comparisons of the extent of differentiation are difficult, and such a conclusion based upon subjective assessment can only be tentative.
A further striking difference between differentiation of the hybrids and the parental TG11.nR cells, was the re-appearance of significant numbers of EC cells after 2–3 weeks exposure to retinoic acid. Indeed, some EC cells always do reappear in retinoic acid-induced, differentiated NTERA2 cultures,13 but their numbers are generally very small. Since many of the differentiated cells either do not divide or divide only slowly, even a small increase in the number of “escaper” or revertant EC cells could lead to a dramatic increase in the rate at which such cells overgrow the differentiated cells. We have not made a detailed study of the EC cells that grow out from long term retinoic acid cultures of C7 and C10. One possibility is that, through karyotypic instability, they represent cells that have lost a key “differentiation-permissive” gene originally derived from TG11.nR, in which case detailed study of such “escapers” might indicate the chromosomal location of such a gene. Alternatively, there might be more complex mechanisms involving interaction of the 2102Ep and TG11.nR genomes that alter the probability with which differentiating cells revert to a stem cell phenotype.
In summary, it is evident that 2102Ep EC cells fail to differentiate for complex reasons, involving both the loss of function of key genes required generally for differentiation, and the altered function of genes that regulate differentiation along specific pathways, notably the neural pathway. Although the nature of these mutations is currently unknown, their elucidation may provide important clues to the role of loss of differentiation capacity in tumor progression and the role of such genes in cell differentiation during development.