Reversible and irreversible effects of retinol upon the phenotypic properties of embryonal carcinoma cells

Since embryonal carcinoma (EC) cells are considered to be pluripotent, they can theoretically differentiate into a spectrum of cell types. Accordingly, when one assesses the state of differentiation of cells in EC cultures, it would seem prudent to use several general discriminatory markers. Among these are alterations in morphology, elevated levels of secretion of plasminogen activator (e.g. Sherman et al. 1976), cell-surface deposition of intercellular matrix proteins such as fibronectin and laminin (e.g. Zetter & Martin, 1978; Wolfe, Mautner, Hogan & Tilly, 1979; Knowles et al. 1980), appearance of cytoskeletal elements such as keratin (e.g. Jones-Villeneuve, McBurney, Rogers & Kalnins, 1982), and disappearance of EC-specific surface antigens including SSEA-1 (Solter & Knowles, 1978). In studies with several EC lines and differentiation-defective EC cell mutants derived from them (McCue, Matthaei, Taketo & Sherman, 1983), we have shown that these parameters are collectively useful in attempts to evaluate differentiation, even though they do not all invariably change when differentiation is induced. According to classical concepts of differentiation, it is also to be expected that such alterations in phenotype are permanent, i.e. they should not be reversed after removal of the inducer (see Sherman, 1986a).

Among known inducers of differentiation of EC cells, retinoids, especially retinoic acid, are particularly active (reviewed recently by Jetten, 1986). Retinol is less potent than retinoic acid in inducing differentiation of EC cells: quantitative differences have been evidenced among various lines (Eglitis & Sherman, 1983) and even within a given line, F9, when studied by different laboratories (Strickland & Mahdavi, 1978; Eglitis & Sherman, 1983; Jetten & DeLuca, 1983). Furthermore, one EC line, PCC4.azalR, differentiates in response to retinoic acid but fails to do so when treated with retinol at non-toxic concentrations (Jetten & Jetten, 1979; Eglitis & Sherman, 1983). Nulli-SCCl is an EC line which appears to respond, albeit poorly, in response to retinol (Eglitis & Sherman, 1983). We have derived from Nulli-SCCl a clonal line, which we have called NR1. As described below, this cell line responds in an uncharacteristic manner to retinol. The behaviour of these cells raises some fundamental questions about the criteria used to assess differentiation of EC cells.

The Nulli-SCCl line was described initially by Martin & Evans (1975) and was obtained from Dr Gail Martin. Culture of these cells in serum- and antibiotic-supplemented Dulbecco’s modified Eagle’s medium in gelatin-coated dishes is described by McCue et al. (1983). The NR1 line was derived from Nulli-SCCl cells after treatment with N-methyl-N′-nitroW-nitroso-guanidine as described by Schindler et al. (1981). Retinol (0·25μgml⁻¹, or 8·7×10⁻⁷M) was added 24 h after cells were treated with mutagen and seeded at clonal density. NR1 cultures were maintained in the same way as Nulli-SCCl cultures. The origin and maintenance of F9 EC cells have been described elsewhere (Eglitis & Sherman, 1983). For growth curves, cells were plated in gelatin-coated 6 cm culture dishes at densities of 1–2×10⁵ cells.

When cells were treated with all-trans-retinol (8·7×10⁻⁷M) or all-tram-retinoic acid (10⁻⁶M) the retinoids were added in ethanol such that the concentration of vehicle did not exceed 0-1 %. It has been shown previously (Jetten & Jetten, 1979) that this concentration of vehicle has no detectable effect on EC cells. Retinoids were stored as 10⁻²M (retinoic acid) or 8·7×10⁻³M (retinol) stock solutions and kept for up to two weeks at –70 °C. HPLC analyses have indicated that the retinoids retain their integrity under these conditions (M. L. Gubler, personal communication). Cultures were maintained in a room equipped with amber lights.

The proportion of cells with non-EC morphology was estimated by examining several fields in two or more dishes with a phase-contrast microscope. (Cells with a typical EC morphology are small and rounded with indistinct boundaries and high nuclear/cytoplasmic ratios; see, e.g. Figs 3A, 4A.) For measurements of secretion of plasminogen activator, culture media were assayed for plasminogen-dependent degradation of ¹²⁵I-fibrin (described in detail by Eglitis & Sherman, 1983). Surface-associated fibronectin was detected by immunofluorescence with rabbit antihuman fibronectin (BRL, Bethesda, MD: diluted 1:5) and loss of SSEA-1 antigen by immunofluorescence with monoclonal antibody (diluted 1:100 or 1:200) kindly supplied by Drs D. Solter and B. Knowles (see Solter & Knowles, 1978). Cells were also challenged with rabbit antisera against laminin and the cytoskeletal component keratin. These antisera were obtained from Drs R. Oshima and H. Sun and diluted at 1:5 and 1:30, respectively. In all instances, fluorescein-conjugated second antibody (diluted 1:20) was used. Appropriate non-immune sera were used on control coverslips. The immunofluorescence procedure used was the same as that described by Eglitis & Sherman (1983) except that in the present study, cells were cultured on glass coverslips and fixed for 6 min with ice-cold acetone (for SSEA-1), methanol (for fibronectin or laminin) or with methanol for 6min followed by acetone for 2 min (for keratin).

Cells for tumour analyses were cultured for the indicated time periods in the absence or continuous presence of retinol at 8·7×10⁻⁷M, with passage as the cells approached confluence. Cells were harvested by trypsinization, washed and resuspended in phosphate-buffered saline. After counting, 5×10⁵,10⁶ or 2×10⁶ cells were each injected s.c. into the left flank of strain 129 mice. Mice injected with control or retinol-treated cells at each inoculation level were matched for age and sex. Mice were inspected at least three times per week and the earliest time of appearance of a palpable tumour was noted.

After Nulli-SCCl cells were treated with the mutagen N-methyl-N′-nitro-N-nitrosoguanidine and exposed continuously to retinol at 8·7×10⁻⁷M, a clonal line, NR1, was isolated. Unlike the parental cells, NR1 cells were not growth-inhibited by retinol, as reflected by typical growth curves illustrated in Fig. 1A,B. However, the density of NR1 cells at confluence was lower in the presence of retinol than in its absence (Fig. 1C).

Table 1 illustrates that Nulli-SCCl cells secreted 30-fold increased levels of plasminogen activator following 8 days of exposure to retinol, with only a modest increase after 4 days, in confirmation of an earlier report (Eglitis & Sherman, 1983). Also as demonstrated previously (McCue et al. 1983), the same cells responded readily to a similar dosage of retinoic acid within 4 days. The response of NR1 cells to retinol was much less marked than that of Nulli-SCCl cells: levels of plasminogen activator secretion were elevated only five-fold after 8 days (Table 1). On the other hand, secretion levels increased more than 50-fold when NR1 cells were exposed to retinoic acid for 4 days.

In a separate experiment involving long-term treatment with retinol, Nulli-SCCl cells secreted maximal amounts of plasminogen activator after 14 days of treatment, with the level falling dramatically by 28 days (Fig. 2A). NR1 cells consistently failed to secrete markedly elevated levels of plasminogen activator (Fig. 2B).

Fig. 2A reveals that although Nulli-SCCl cells produced a large increase in plasminogen activator secretion in response to retinol, the estimated percentage of cells showing obvious morphological alteration remained relatively low throughout the test period. By contrast, a large percentage of NR1 cells underwent morphological change despite the fact that there was little elevation of secretion of plasminogen activator (Fig. 2B). Figs 3–5 document the change in appearance of NR1 cells in response to retinol. Figs 3A and 4A illustrate that in control cultures the cells have a typical EC morphology. After only 2 days of exposure to retinol, although the individual cells did not look different, a subtle alteration appeared in the contours of the cell clusters: instead of being relatively smooth and rounded, they became more irregular and pointed (cf. Fig. 3A,B). By 4 days of treatment, some of the cells individually became more stellate. This property continued until, by 9 days, most of the cells in the culture possessed an irregular appearance (Fig. 4B); at this time the cells were also less likely to form tight clusters in which cell boundaries were obscured, as is typical of EC cells (e.g. Fig. 4A). At higher densities, retinol-treated NR1 cells tended to stream, as do fibroblasts (Fig. 5A). The cells retained their atypical appearance for at least 28 days of treatment with retinol (Fig. 5C).

When retinol was removed from NR1 cultures previously treated for up to two weeks with the retinoid, there was a rapid and remarkable reversal of morphology: within 2 days, most of the cell clusters became smooth and non-streaming once again (Figs 4C, 5B); individually, the cells lost their irregular appearance and became tightly adherent. It is only after cells had been treated with retinol for extended time periods that they failed to undergo this morphological reversal within two days of withdrawal of retinol (e.g. Fig. 5D).

The reversibility of the morphological change of NR1 cells in response to retinol appears to be atypical. It was not, for example, observed when the cells were exposed to retinoic acid; in this case, the stellate, irregular appearance of the cells (not markedly different from that observed when the cells were treated with retinol; cf. Fig. 4B,D) failed to reverse when that effector was removed (Fig. 4E).

Fig. 6 illustrates that those cells in Nulli-SCCl cultures which underwent morphological change in response to retinol retained their altered morphology after the retinol had been removed. Similarly, F9 cells, which we have reported to undergo extensive alteration in morphology (as well as other parameters) in response to retinol (Eglitis & Sherman, 1983), did not readily reverse their appearance following removal of that retinoid. The same irreversibility in appearance has been found (data not shown) for another cell line, OC15-S1, that differentiates readily in response to retinol (Eglitis & Sherman, 1983).

NR1 cells can be maintained in a proliferative state in the presence of retinol for long periods of time. To determine whether they possessed the tumourigenic capacity of EC cells (as opposed to differentiated derivatives which tend to be benign), NR1 cells maintained in the presence of retinol for 120 days were injected s.c. into six mice (twice each at 5×10⁵,10⁶ and 2×10⁶ cells). The average time for tumour appearance in the six mice was 15·7 days versus 15·2 days for tumours derived from untreated NR1 cells. In similar experiments, tumour incidence of Nulli-SCCl cells was also 100% after exposure to retinol for 37 days, though the average time for appearance of the tumours, 18·0 days, was twice as long as that observed from control cells (9·0 days).

We tested NR1 and Nulli-SCCl cells for the presence of surface-associated SSEA-1 antigen (Table 2). In untreated cultures, a higher proportion of NR1 than parental Nulli-SCCl cells contained detectable levels of the antigen. With both cell lines, there was a reduction by more than half in the percentage of reactive cells following exposure to retinoic acid for 7 days. Following the same interval of treatment with retinol, Nulli-SCCl cells showed a more modest reduction in SSEA-1 display; reactivity with NR1 cells was also clearly diminished relative to untreated cells, although the decrease was not so great as for Nulli-SCCl cells (Table 2).

NR1, Nulli-SCCl and F9 cells were also tested for surface-associated laminin and fibronectin reactivity. Since these antigens were arrayed as networks over the monolayer, it was not possible to unequivocally score individual cells for reactivity. However, observed changes in the overall extent of fluorescence are recorded in Table 3. In the absence of effectors, there were either low levels of reactivity or no reactivity of the monolayers with anti-laminin or anti-fibronectin serum (e.g. Fig. 7A). Following treatment with retinol for 7 days, there was no evidence of increased reactivity with anti-laminin serum in any of the cultures. F9 cells showed clearly increased reactivity with anti-laminin serum following exposure with retinoic acid for as few as 5 days, whereas the results with the other lines were less obvious and variable under similar conditions.

There were increases in reactivity with anti-fibronectin serum when NR1, Nulli-SCCl and F9 cultures were treated with retinoic acid for 5 or 7 days. There were also obvious increases in surface-associated fibronectin following treatment for 7 days with retinol; this increase was most dramatic with NR1 cells (see, e.g. Fig. 7B). When NR1 cells were treated with retinol for 5 days and then placed in medium without added retinol for 2 days (a protocol which leads to the reappearance of a typical embryonal carcinoma morphology), fibronectin levels were judged to be reduced relative to those observed in continuously treated cultures, but still elevated relative to those of untreated cultures (not shown).

Immunofluorescence analyses with antiserum against the cytoskeletal component keratin indicated no reactivity in untreated cultures but obvious display of antigen when F9, Nulli-SCCl or NR1 cells were treated with retinoic acid for seven days. Only F9 cells showed evidence of keratin expression following exposure to retinol for 7 days and under these conditions only 5–10% of the cells were scored as positive (as opposed to more than half the cells following treatment with retinoic acid).

In order to carry out immunofluorescence analyses, we plated our cells on gelatin-coated glass coverslips instead of gelatin-coated plastic dishes. NR1 cells adhered poorly to glass, even after gelatin treatment (Fig. 8A). The same was found to be true for Nulli-SCCl cells (unpublished observations). It was noted, however, that in NR1 cultures treated with retinol or retinoic acid, the ability of the cells to adhere to the coverslips was markedly improved (Fig. 8B,C). It was also obvious that under these conditions many retinoic acid-treated NR1 cells possessed a non-embryonal carcinoma morphology whereas almost all of the cells exposed to retinol retained a morphology typical of embryonal carcinoma cells.

The studies described with Nulli-SCCl and F9 cells are consistent with earlier reports (Eglitis & Sherman, 1983) that these cells differentiate in the presence of retinol although the latter cells are more responsive than the former. Nulli-SCCl cells secrete increased levels of plasminogen activator during continuous exposure to retinol for 7 to 14 days, but levels appear to fall thereafter. A similar acquired refractoriness is observed with F9 cells (Sherman, 1986a) and the cause appears to be epigenetic rather than genetic in nature (manuscript in preparation). Nevertheless, the delay in Nulli-SCCl tumour appearance after a lengthy exposure of the cells to retinol suggests that at least some of the cells possess reduced tumourigenic potential under these circumstances.

NR1 cells undergo several changes in response to retinol: their morphology is altered, they show modest increases in secreted levels of plasminogen activator and some surface properties are modified: display of SSEA-1 is diminished, fibronectin levels are elevated and adhesiveness to the substratum (gelatin-coated glass or plastic) is strengthened. All of these criteria are consistent with the view that NR1 cells differentiate in response to retinol. Yet, the cells appear to revert to a typical EC phenotype following removal of the retinol. The reversal of morphology is rapid and affects the vast majority of cells in the culture. This argues against the possibility that the cultures contain a small population of cells refractory to retinol which eventually overgrow the differentiated progeny following removal of the retinoid. Thus, insofar as one accepts the view that differentiation is an irreversible event, one is led to conclude that retinol fails to elicit differentiation of NR1 cells. The suggestion that even very long exposure periods to retinol are without apparent effect upon the tumourigenicity of NR1 cells supports this view since EC cells induced to differentiate in vitro commonly exhibit reduced tumourigenic potential (e.g. Sherman, Matthaei & Schindler, 1981). At present, we have no information about the nature of the mutation(s) in NR1 cells which influenced their response to retinol.

If NR1 cells do not undergo permanent phenotypic change in response to retinol, what is the significance of the several reversible alterations which are observed? Two likely possibilities are that (a) NR1 cells fail to progress beyond a precommitment stage of differentiation in response to retinol or (b) retinol triggers a series of changes in NR1 cells which are not directly related to differentiation even though they might commonly accompany retinoid-induced differentiation of other EC cells. The former possibility is consistent with the view expressed by Johnson, Handyside & Braude (1977) that differentiation in early mammalian embryogenesis proceeds as a series of events beginning prior to cell commitment and that such commitment is attained only after the cascade of molecular reactions reaches an irreversible stage. Insofar as the loss of ability to grow at clonal density reflects commitment of EC cells (see Sherman, 1986a), the studies of Rayner & Graham (1982), Rodrigues, Balicki, Newrock & Muckherjee (1985) and Sherman et al. (1985) suggest that most cells in PC13, F9 and Nulli-SCCl cultures, respectively, become committed to differentiate in response to retinoic acid within 24 h, with at least some cells appearing to be irreversibly affected within 6 h. Therefore, the precommitment period when these EC cells are treated with RA is relatively short. On the other hand, Ogiso, Kume, Nishimune & Matsushiro (1982) have claimed that cells from EC line 311 pass through a 48 h reversible phase of phenotypic change in response to retinoic acid since longer exposures were required before observed alterations in surface properties were retained following removal of the retinoid. In our studies, we have found evidence of reversibility of phenotype even after several days of exposure of NR1 cells to retinol. Since NR1 cells had been exposed to mutagen, it might be argued that they are genetically blocked from progressing beyond a precommitment stage of differentiation. However, this is not a universal property of these cells since they appear to undergo characteristic and irreversible phenotypic changes in response to retinoic acid.

There is substantial precedence for retinoid-induced alterations in phenotype and growth properties of non-EC cells that appear to be independent of differentiation (see Lotan, 1981 and Sherman, 1986b for reviews). Most pertinent to these studies are observations that cells treated with retinoids secrete modestly elevated levels of plasminogen activator (Schroder, Chou & Black, 1980), display increased amounts of surface-associated fibronectin (Jetten, Jetten, Shapiro & Poon, 1979; Bolmer & Wolf, 1982; Keeble & Maden, 1984) and become more adherent to the substratum (Jetten et al. 1979) without evidence of alteration of their state of differentiation. NR1 cells might be undergoing the same kind of changes even though their ability to differentiate in response to retinol has been extinguished. Such observations might imply that retinoids can independently influence the expression of more than one set of genes, not all of which lead to irreversible phenotypic change. However, such a conclusion would be premature for NR1 cells since it remains to be demonstrated that any of the retinol-induced alterations are dependent upon changes in gene expression. In fact, in the studies of Bolmer & Wolf (1982) retinoid-induced increases of cell-surface-associated fibronectin did not require elevated levels of transcription since this change was observed following enucleation. It is therefore conceivable that retinoid-induced changes which result in differentiation require alterations in gene expression (perhaps mediated by retinoid-binding proteins, which can translocate their ligands to the nuclear compartment; reviewed by Sherman, 1986b), whereas a series of reversible changes in cell appearance and behaviour can result from retinoid modifications of membrane properties, possibly through direct interaction with membrane components (see, e.g. Jetten, DeLuca & Meeks, 1982). The latter action could readily explain retinol alteration of the adhesiveness of NR1 cells to the substratum and to each other. This, in turn, could influence cell morphology as well as the antigenicity of surface-associated components.

There are long-standing arguments in the literature as to the site of action of retinoids (see Sherman, 1986b). If the preceding proposal is correct, then the pleiotropic effects of retinoids could result from different types of action at different cellular locations. How such retinoid actions relate to their physiological effects remains to be established. For example, it is possible that under non-pharmacological circumstances retinoids might only be presented to cells complexed to a plasma retinol-binding protein and not as free ligand (see Sherman, 1986b). Thus, retinoid interactions with membranes in vivo and in vitro (where retinoid levels might be substantially in molar excess of plasma retinol-binding protein) might be very different. We are currently evaluating the influence of dietary retinol and retinoic acid on differentiation of cells from several EC lines in vivo.