New Research In
Physical Sciences
Social Sciences
Featured Portals
Articles by Topic
Biological Sciences
Featured Portals
Articles by Topic
- Agricultural Sciences
- Anthropology
- Applied Biological Sciences
- Biochemistry
- Biophysics and Computational Biology
- Cell Biology
- Developmental Biology
- Ecology
- Environmental Sciences
- Evolution
- Genetics
- Immunology and Inflammation
- Medical Sciences
- Microbiology
- Neuroscience
- Pharmacology
- Physiology
- Plant Biology
- Population Biology
- Psychological and Cognitive Sciences
- Sustainability Science
- Systems Biology
Phenotypic selection as the biological mode of epigenetic conversion and reversion in cell transformation
Contributed by Harry Rubin, December 6, 2017 (sent for review October 16, 2017; reviewed by Allen Mayer and Robert H. Rice)

Significance
Single exposure of a certain mouse cell line to carcinogens causes delayed neoplastic transformation of most, if not all its cells. In the absence of carcinogens, the extended maintenance at high density of another cell line or of cells recently explanted from mice leads to increased saturation density, neoplastically transformed foci in culture, and tumor formation in mice. These properties are reversed to normal by frequent passages at low cell density in high serum concentrations. Such population-wide dynamics are incompatible with somatic mutation but are characteristic of epigenetic behavior. They represent phenotypic selection of cells from a normal population which progresses either to cancer or to the reversion of transformed cells to pretreatment phenotype, depending on differing microenvironments.
Abstract
Exposure of certain cell lines to methylcholanthrene, X-rays, or physiological growth constraint leads to preneoplastic transformation in all or most of the treated cells. After attaining confluence, a fraction in those cells progress to full transformation, as evidenced by their ability to form discrete foci distinguishable from the surrounding cells by virtue of their higher density. Transformation induced by suspension in agar, an even stronger growth-selective condition than confluence, is reminiscent of all but the final differentiated stage of a normal developmental process, epithelial–mesenchymal transition. Changes associated with transformation are not restricted to focus-forming cells, as the permissiveness for focus formation provided by confluent cells surrounding transformed foci is greater than that of nonselected cells. The neoplastic process can also be reversed in culture. Transformed cells passaged at low density in high serum revert to normal morphology and growth behavior in vitro and lose the capacity for tumor formation in vivo. We propose that transformation and its reversal are driven by a process of phenotypic selection that involves entire heterogeneous populations of cells responding to microenvironmental changes. Because of the involvement of whole cell populations, we view this process as fundamentally adaptive and epigenetic in nature.
In his classic 1969 treatise, cancer pathologist Foulds (1) defined neoplasia as “a developmental process akin to normal development in some respects, but differing from it in important particulars.” Foulds viewed neoplastic development as an epigenetic process for which a complete understanding was possible only through careful examination of the behavior of cells in their tissue and organismal contexts. It was later stated by Robert Weinberg that a new era began in cancer research with the development of an assay for Rous sarcoma virus in cell culture, which indeed opened the field to genomic approaches (2, 3). Nevertheless, Weinberg (4) wrote that “endless complexity” evident after many further years of genomic research in cancer, had created a situation in which “we can’t really assimilate and interpret most of the data that we accumulate.” Foulds (1) anticipated Weinberg’s view by observing that, despite the accumulation of vast amounts of experimental data, “The truth is that we already have more ‘facts’ than anyone knows what to do with.” Adding to this skepticism, James Watson, a founding father of molecular biology, said that attempts to identify genes that cause cancer have been “remarkably unhelpful” in contributing to a unified understanding of oncogenesis (5).
The inordinate complexity that arose from attempts to trace the origin of cancer to the mutation of thousands of genes recommended an alternative, more holistic approach (4). Ordinarily that would mean a turn to epigenetics as stressed by Foulds (1). However, the term “epigenetics” is understood by modern molecular biologists largely in biochemical terms––methylation of DNA and histones as well as acetylation of histones, changes in chromatin states, and associated enzymes––most of which are involved in the regulation of gene expression.
The evidence supporting Foulds’ view is derived from functional experiments using living cells rather than from the identification of molecular components of the cells, as intimated by Weinberg (4). To some extent this started with the elucidation of the initiation–promotion paradigm in skin chemical carcinogenesis in vivo. Promotion, which ends in the overt manifestation of tumors, was considered nonmutagenic in nature and dependent on a selection process that required extended exposure to chemical promoters (6). In the following paragraphs we review a key series of experiments in cell culture that demonstrate the involvement of entire cell populations in a constant process of phenotypic selection that are rate-limiting but reversible steps in cell transformation. The implications of these experiments for our understanding of oncogenesis in a general sense are also discussed.
Number of Cells That Undergo Conversion to the Transformed State
Quantitative Effects of Methylcholanthrene on the Process of Transformation.
An established line of aneuploid mouse fibroblasts was developed from a clone of inbred C3H adult ventral prostate cells with ∼5 × 105 cells per confluent culture in 60-mm plastic Petri dishes (7). Exposure to methylcholanthrene (MCA) resulted in the production of multilayered foci of transformed cells. An experiment was subsequently devised to distinguish whether the cells that produced foci in response to MCA were newly transformed or underwent selection from preexisting transformed cells in the presence of MCA (8). Single cells on small coverglass chips were treated with MCA and then grown to confluence under optimal concentrations of MCA. One hundred percent of the cultures derived from individual cells under optimal conditions produced transformed foci, in contrast to cultures derived from single cells but not treated with MCA, none of which produced foci (9). Thus, MCA induced the transformation of the entire population of previously nontransformed cells.
To determine whether all the cells in a single focus would produce transformed foci, the cells from a single long-term focus were recloned. Expansion of every recloned population generated cultures consisting of multilayered, randomly arranged cells, indicating that all cells from MCA-transformed foci were capable of reinitiating foci.
When single cells were cloned before MCA treatment and recloned after the treatment, all the recloned cells produced transformed colonies, but the beginning of transformation varied from 21–77 d. Of 76 mice injected with cells from clones derived from individual, single MCA-treated cells, 73 developed fibrosarcomas. This confirms that the cells derived from piled-up colonies are malignant. It also suggests that transformation is driven by common events rather than by rare genetic changes.
Quantitative Effects of X-Ray Irradiation on the Process of Transformation.
An experiment was done to establish the relationship between exposure to X-rays and transformation of the C3H 10T1/2 cloned cell line (10). Several hundred cells per dish were irradiated with 400 rad and grown to confluence over a 10- to 14-d period. The cells were resuspended and diluted in 10-fold steps from 1:10 to 1:10,000, along with some cultures that were not resuspended. All the cultures were grown to confluence, where they were maintained for almost 6 wk to allow the development of transformed foci. The total number of foci formed per dish was constant, despite having been seeded at densities that varied over an ∼1,000-fold range. The investigators concluded that exposure to X-rays produced a functional heritable change, which was expressed as an increase in the probability of focus formation at confluence, in virtually every cell. The observations suggest that few, if any, of the clones were transformed as a direct consequence of the X-ray exposure and challenge the hypothesis that transformed foci are the clonal products of occasional cells that sustained X-ray–induced mutational change. At least two steps appear to be involved: (i) X-ray irradiation, which induces altered cell function in many or all treated cells and is transmitted to the progeny of the surviving cells, and (ii) overt transformation, which occurs after the cells have achieved confluence. The growth rate of cells is markedly reduced at confluence, suggesting that the driving force of transformation is not mutation in growing cells but a selective process acting on the resting, heterogeneous population that promotes their progressive transformation. Since this involves a considerable fraction of the population, we refer to it as “phenotypic selection.” The likelihood of this occurrence is assessed in the following experiments.
Selective Effects of Growth Restraint on the Process of Transformation.
A third case dealt with the number of cells that express transformation solely in response to changes in the conditions of growth. Both the NIH 3T3 and Balb 3T3 lines of mouse embryo fibroblasts were grown in MCDB 402, a medium optimized for their clonal growth in minimal serum (11). The NIH 3T3 cells produced transformed foci at confluence within 2–3 wk in 5% and 10% calf serum (CS) but not in 2% CS (12). Thus, 2% CS was chosen as the baseline medium for assaying transformed focus-formers. When the cells were grown in 10% CS, the number of foci peaked at least 5 d after the cells reached their saturation density, suggesting that transformation occurred when net growth ceased. When the cells were transferred to medium containing 2% CS following transformation in 10% CS, they had gained the ability to express foci in 2% CS. These results suggested that foci arose by phenotypic selection rather than by mutation that depends on cell replication.
After repeated passage of NIH 3T3 cultures at low densities (105 cells per 60-mm dish) in 10% CS, the cells expressed a few foci when assayed for 2 wk in 2% CS. This prompted the question of whether more foci would form if given more time for their development. This question could not be addressed simply by extending the incubation period, because any foci that formed might spread to cover much of the area of the culture dish. Transformed cells could also detach from the original foci to initiate new foci by reattaching at a distance from their site of origin. These problems were averted by growing cells in multiwell plates, which simulated partitioned culture dishes. All the wells in a given plate were assayed for varied stages of focus formation at consecutive intervals for as long as 14 wk (13). As before, most of the transformations were observed long after confluence was achieved in the multiwells, indicating that transformation was more likely to occur in nondividing than in dividing cells. The results showed a trend toward transformation of most, if not all, the cells if those that produced light, small, incipient foci were included. Hence, transformation based on phenotypic selection in crowded cultures supported the results obtained from MCA and X-ray treatment: A high proportion of the cells exposed to selective conditions initiate the transformation process.
Dynamics of Pre- and Posttransformed Cells.
An experiment was designed to explore the dynamics of the selective process at confluence using variations in serum concentration (2%, 5%, and 10% CS), time (2 and 3 wk), and three to four serial postconfluent passages of four NIH 3T3 lineages (14). The cells established contact with each other within 2 d and were confluent within 3–5 d. These manipulations were followed by several rounds of selection, all in a single concentration of serum (2%) and duration (2 wk) for each of the initial combinations (14, 15). An overview of the results most relevant to the dynamics of the process begins with the linear relationship between serum concentration and saturation density in the first round, which increases slightly between weeks 2 and 3 (Table S1). This relation continues in the subsequent rounds under constant conditions. The relatively high frequency, regularity, and uniformity of increases in saturation density are characteristic of an epigenetic process, which, as noted above, was the most plausible basis for incipient neoplasia (1). The continued increases in saturation density with successive postconfluent passages can be understood by the high passage densities, which allow additive degrees of selection. The greater the postconfluent density, the sooner are the onset and degree of transformation (Fig. S1). In several cases in which one to three very small foci appeared in one round of selection, a large number of dense foci were invariably observed in the next round. In some cases, light and dark foci appear at the same time. Their occurrence is far too frequent to be caused by rare mutations or aneuploidy. If a large number of distinct foci appeared in one round, the next round would invariably give uncountable, or even confluent, foci.
Sequential Stages of Progression After Very Strong Selection During Suspension in Agar.
The Balb/3T3 line of mouse embryo cells was derived by the same methodology as the NIH 3T3 line. It was used to determine cell morphology and growth rates after a single round of selection by confluence consisting of overnight incubation at very high density on a coverslip followed by 2 wk in agar suspension with 10% CS (16). The cloning efficiency in agar was 0.05% in colonies ≤0.2 mm and 0.005% in transformed colonies (≥0.5 mm). Seven of the former and two of the latter were isolated from the agar, plated in plastic culture dishes, and observed during exponential growth. Control cells that had not been seeded in agar were plated in parallel. The control cells consisted of flat, isometric epithelioid cells. Two of the clones from the small (nontransformed) agar colonies looked more like the control cells than the other colonies, but close inspection revealed that they were different in appearance from the controls and from each other. Most of the remaining small agar clones had different mixtures of fibroblastic and flatter morphologies on plastic, although all were distinctly different in appearance from the control. The two large, transformed agar clones had clumps of round cells but differed in the size of the clumps. Clonal growth rates on plastic and in agar differed from colony to colony, with the large agar clones growing much faster than the others on both substrata. Taken together, these observations suggest a sequence of progression from the clonal through the varied mixed fibroblastic to the roundish, transformed colonies. With exception of the transformed clones, the differences in morphology were later identified as similar to the epithelial–mesenchymal transition found in normal embryonic development (17). The fact that the single-step outcome of selection was variation in morphology from that of the epithelioid control supports an epigenetic or phenotypic rather than a mutational–chromosomal origin of the process. Upon injection into mice, only the transformed colonies produced tumors, both of which were malignant fibrosarcomas.
Cellular Microenvironment of Transformed Foci.
As shown above, carcinogenic treatment of cultures with chemical or physical agents alters all cells, increasing susceptibility to transformation of a few of them after they become confluent (9, 10). The same effect occurs when the cells are selected by contact inhibition, which increases the saturation densities of whole populations, leading ultimately to the development of transformed foci (14, 15). This suggests that the field surrounding individual foci, which also arises through repeated rounds of selection with increases of saturation density, may be more permissive to transformation than cells in their first round of contact inhibition. To test this hypothesis, a culture seeded with 105 NIH 3T3 cells that had undergone three rounds of selection at high density was compared with a culture that had been diluted 10-fold, leaving 104 cells that were combined with 105 nontransformed cells that had undergone no selection (18). The former had 40 very large, dense foci with a background of many small, light foci (Fig. 1). The mixed culture had only four small, dense foci with nearly invisible foci in the background. It was apparent that the nontransformed cell background had shrunk the size of the dense foci and almost eliminated the light foci. It was also clear that the nontransformed cells that had undergone three rounds of selection were more permissive for the development of large, dense foci than a background of nonselected cells. This was another sign that cells that had been subjected to selection without obvious transformation had undergone a pretransformed stage that was permissive for further transformation. The results again supported the conclusion that all cells exposed to selective treatment were on the epigenetic road to transformation. They also suggested a relation between increased saturation density and susceptibility to transformation.
Cellular microenvironment of transformed foci. A culture seeded with 105 cells that had undergone three rounds (3°) of selection at high density (Left) was compared with one that had been diluted 10-fold, leaving 104 cells combined with 105 nontransformed cells (Right). The cultures were assayed in 2% CS and then were fixed and stained at 14 d. Reproduced from ref. 18.
Reversion of Transformation by Maximizing Growth at Low Cell Density in High CS
Reversion of Polyoma-Transformed Cells.
Golden hamster embryo cells were transformed by infection with Polyoma virus and selected by growth in agar. The transformed cells lacked contact inhibition and had the capacity to form colonies in agar and tumors in hamsters (19, 20). They could not grow on a layer of gluteraldehyde-fixed normal cells, whereas normal cells did have that capacity. After eight passages at low density, the initially transformed cells reversed those growth capacities. They also developed a low saturation density and could no longer form colonies in agar or tumors in hamsters. The high frequency of reversion suggested that the expression of some, and perhaps all, properties characteristic of transformed cells is regulated by one or more balanced control mechanisms that can be switched from one pathway to another. Surprisingly, the transformed cells were diploid, and the revertant, phenotypically normal cells were aneuploid (21, 22). It was proposed that the formation of aneuploid nontransformed cells results from the transformation of binucleate polyploid cells, which later lose chromosomes and stabilize the reverted nontransformed state. Importantly, the revertant nontransformed cells retained the Polyoma virus genome (19, 23), suggesting that reversal was not simply a process of eliminating viral genes.
Reversion of Cells Transformed by X-Ray Irradiation to the Parental Nontransformed State by Passage at Low Cell Density.
C3H mouse embryo cells were of the same origin as those used in the population-wide transformation by MCA (9) and X-ray (10). As in the latter case, they were transformed by X-ray irradiation, exhibiting a decrease in sensitivity to contact inhibition, a high saturation density, a change from a flat, epithelioid to a multilayered fibroblastic morphology, and an ability to form colonies in agar and malignant tumors in mice (24). Serial passage of the cells at low density resulted in a stable reversion to the phenotype of the parental nontransformed cells. Passage at high cell density, however, resulted in a back reversion to the transformed phenotype. Passage at an extremely low (clonal) density caused reversion to the parental phenotype at the first passage, but back reversion to the transformed phenotype no longer occurred, even at high cell density. The results indicated that the reversion to the parental phenotype was due to a population-wide change rather than to selection of a preexisting population of the parental phenotype.
Conversion of NIH 3T3 Cells to the Transformed State by Growth at Low Density in 2% CS and Its Reversion in 10% CS.
NIH 3T3 cells produced transformed foci by passage at low cell density in 2% CS (25). Cells isolated from such a transformed focus continued their transformed state, although with large quantitative fluctuations during frequent passages in 2% CS at low cell density (Fig. 2) (26). However, they gradually reverted to the nontransformed state by frequent passages in 10% CS at low cell density. All the cultures maintained in a low CS concentration produced fibrosarcomas when injected into athymic mice (Table 1). Those switched to high CS for six passages reduced transformation by half but continued to produce tumors in all injected mice. However, they lost all their capacity to produce foci and tumors by passage 34 in high CS.
Conversion of NIH 3T3 cells by growth at low cell density in 2% CS and reversion of the cells in 10% CS. Cells produced transformed foci by repeated passage at low density in 2% CS. Transformed cells were isolated from a transformed focus and passaged three times per week at low cell density in 2% CS (○, upper curve). At the indicated passages 100 transformed cells were mixed with 105 nontransformed cells to assay for focus formation in 2% CS. Transformed cells were also passaged in 10% CS and assayed in 2% CS (●, lower curve) (26).
Effects of culture conditions on focus formation and tumorigenicity of transformed cells
To gain a more elaborate picture of the dynamics of reversion, 500 focus-forming cells were mixed with 5 × 104 nontransformed cells. The mixture was passaged at low density in 2% and 10% CS biweekly for 32 d and was assayed for focus formation at each passage (27). At the same time unmixed cultures in 2% and 10% CS were passaged to determine their separate growth rates and were used to predict the ratio of focus-formers to nonfocus-formers (Fig. 3). The relative number of focus-formers on day 0 (500 cells out of 50,500 total cells) is plotted as 1.0 on the ordinate. The upper line in Fig. 3 shows the anticipated increase in focus-formers in 2% CS, and the lower line shows the anticipated decline of focus-formers in 10% CS. Low serum concentrations can increase the frequency of focus-formers by decreasing the proliferation of nonfocus-forming cells relative to focus-formers. However, the reproductive selection cannot completely account for the focus-enhancing effects of low serum. Evidence for this point is drawn from focus-forming assays in 2% CS for a clone of cells maintained in 10% CS before assay. Maintenance of these same cells in 2% CS for as little as 2 d before assay caused these consistent nonfocus-formers to make foci. It was concluded that the basis for this progression to the focus-forming phenotype is adaptation to the metabolically constraining 2% CS environment. It was proposed that this adaptation results from progressive selection of physiological states. In contrast, focus-formers maintained similar rates of growth in 2% or 10% CS. The frequency of focus-forming clones was highly heterogeneous, thereby allowing a large target for phenotypic selection. There is also heterogeneity in the responsiveness of cells with the focus-forming phenotype to the confluent state as well as in their morphology (27, 28).
Predicted changes in percentage of focus-forming (FF) cells as a function of serum concentration and time for mixtures of FF and nonfocus-forming (NFF) cells. Five hundred FF cells and 5 × 104 NFF cells were mixed together on day 0. One mixture was maintained in 2% CS, another in 10% CS. Based on separate growth rates of FF and NFF cells in 2 and 10% CS the expected number of FF cells, relative to NFF cells, was predicted and is graphed in this figure. The relative number of FF cells (500 cells out of 50,500 total cells) on day 0 is plotted as 1 on the ordinate. The upper line shows the anticipated increase in FF cells, relative to NFF cells, for mixtures maintained in 2% CS. The lower line shows the anticipated decline in FF cells for mixtures maintained in 10% CS. Reproduced from ref. 27, with permission from Oxford University Press.
The studies reviewed to this point suggest that transformation in cell culture is an adaptive response to altered microenvironmental conditions. Adaptive responses in cell populations are characterized by their dependence on condition, the involvement of most or all cells in the population, and phenotypic reversibility. A study was undertaken to directly address the issue of the adaptive nature of cell transformation in culture (29). Cells passaged three time per week for 10 wk at low densities in 10% CS achieved a saturation density in 2% CS of ∼8 × 105 per 60-mm culture dish. The same cells passaged for 8 wk in 2% CS reached a density of ∼1.5 × 106 cells per dish, an increase of almost twofold. The enduring nature of this change was evident in the fact that a third group passaged first in 2% CS for 5 wk and then in 10% CS for 3 wk achieved the same saturation density in 2% CS as the cells passaged in 2% CS for 8 wk, although the logarithmic growth rate flagged to some degree. Changes in focus-forming ability were correlated with the changes in saturation densities achieved after the various treatment regimens (29). Virtually no foci appeared in cultures of nontransformed cells passaged in 10% CS (Table 2). Foci began to appear in the first measurement at 2 wk in cultures passaged in 2% CS and increased to a maximum of 0.65% at 5 wk. Focus formations in cultures passaged in 2% CS for 5 wk and then shifted to 10% CS decreased to 0.21% at 6 wk and to 0.017% at 8 wk, 35-fold less than in cultures maintained continuously in 2%. There also was a decrease in the size and staining density of the cells in the remaining foci (Fig. 4), indicating that phenotypic selection was characterized by a gradual decrease in the degree of transformation in the remaining focus-formers rather than merely a genetic selection against transformed cells. This finding complements the finding of early onset of transformation with switching from 10 to 2% CS (27), demonstrating phenotypic rather than genotypic modes of action in both transformation and its reversal.
Focus formation in 2% CS by cells frequently passaged in 10% CS, in 2% CS, or switched from 2% CS to 10% CS at week 5
Reduction in size and density of foci in cultures shifted from 2% CS to 10% CS at 5 wk and retained there for 3 wk. (A) Culture maintained in 10% CS for 8 wk. (B) Culture maintained in 2% CS for 8 wk. (C) Culture maintained in 2% CS for 5 wk and then switched to 10% CS for 3 wk (29).
It is worthwhile to consider the different conditions under which the experiments depicted in Figs. 2 and 3 were conducted. Fig. 2 utilized transformed cells that were derived from a large transformed focus. They were maintained at frequent passages at low cell density in 2% and 10% C and were assayed for transformed foci at every passage by adding 100 transformed cells to 105 nontransformed cells. The number of focus-forming cells in 2% CS fluctuated between 20% and 60%, indicating a high degree of instability. Those in 10% CS reverted sharply to zero in 18 passages and remained there through passage 34, indicating these cells have a high sensitivity in for reversion at very low cell density.
In Fig. 3, where the number of transformed cells constituted only 1% of the total population, and the passages were made at relatively high total cell numbers, there was a uniform, gradual 20-fold increase in the number of focus-forming cells in 2% CS and a gradual decrease in 10% CS. Not only were the increases in 2% CS more uniform than those shown in Fig. 2, but the declines in 10% CS also flattened out at 1/30th of the original value, without approaching zero. This would agree with the requirement of low cell density to obtain full reversion (20, 24). It is suggested that the combination of the 100:1 ratio of nontransformed to transformed cells, which would allow multiple contacts of the nontransformed cells with individual transformed cells and conditioning of the medium by the high concentration of the nontransformed cells, would stabilize the transformed cells. Another point of interest from Fig. 3 is that combination of the 20-fold increase in focus formation in 2% CS and the 30-fold decline in 10% CS results in a 600-fold divergence in which both curves are without fluctuation. This high rate of reversion supports an epigenetic origin of transformation.
The Dynamics of Phenotypic Selection in Adaptation and Deadaptation to Extreme Differences of Serum Concentration
So far we have examined the effects of 2% and 10% CS concentrations on transformation. These observations led to a detailed examination of the early and late effects of the switch in CS concentration on cell growth rates. Cells maintained in 10% CS were adapted to growth in 0.25% CS by stepwise reduction of CS (30). The adapted cells initially multiplied almost as quickly in 0.25% CS as they did in 10% CS. Upon return to 10% CS, their loss of the capacity to multiply in 0.25% was monitored in successive passages (Fig. 5). By the second passage in 10% CS there was a marked reduction in growth rate in 0.25% CS for the first 4 d, followed by an increase between days 4 and 6, when a reduced saturation density was reached (30). This pattern continued with successive reductions in the early 0.25% CS growth rate at passages 4, 7, and 10, followed by small increases in the secondary growth rate, which exhibited no saturation density. In fact, at passages 7 and 10 there was almost no growth up to day 4, followed by a much reduced rate to about 8 d, then a continuously increasing but suboptimal rate after that time, although a saturation density was never achieved. The control, which had been continuously grown in 10% CS, exhibited no growth at all in 0.25% CS throughout the experiment. The overall result was that there was a continuous loss of growth rate in 0.25% CS of the entire population in a complex manner when cells adapted to 0.25% CS were shifted to 10% CS. This is another example of population-wide effects, as seen in the transformation reversal upon passages at low cell density in 10% CS.
Deadaptation in 10% CS of cells previously adapted to growth in 0.25% CS. Cells adapted to growth in 0.25% CS were deadapted by 0 (+), 2 (◯), 4 (●), 7 (▲), and 10 (∆) serial passages in 10% CS, followed by their growth in 0.25% CS. The latter were also measured in cells that had been maintained exclusively in 10% CS (□) (30).
Discussion
The concept of somatic mutation as the cause of cancer has dominated cancer research for over half a century. It received a large boost with the advent of monolayer cell culture as applied to the biology of cell transformation by the Rous sarcoma virus (2⇓–4, 31). However, starting with a single transforming gene of the Rous virus, then increasing to five or six mutated genes in human cancer, then to thousands in later research, the picture became exceedingly complex. According to Weinberg, “It was realized that a complex system can only be understood if all of its moving parts are analyzed in one sweeping overview. Such holistic analyses should ideally describe the complex reality of actual biological systems, including that of cancer cells” (4).
The analysis presented here, which basically measures the number of cells started on the road to carcinogenesis by treatment with carcinogens or by physiological induction through cellular growth constraint, makes no assumption regarding the necessity for particular genetic or biochemical pathways. As the data have shown, all the cells subjected to treatment sustained changes that led to potential transformation in all progeny (9, 10, 13). Since somatic mutation leads to rare transforming events, the potential transformation of the entire cell population in itself rules out a genetic origin. This conclusion is reinforced by the finding that entire populations of transformed cells are normalized by passage at low cell densities in relatively high serum concentrations (19, 20, 24, 26, 27).
The exclusion of a genetic origin of conversion and reversion and the lack of a biological theory to explain the early results described here may account for the rarity of follow-up studies. However, after many years of close analysis by Elsasser, first as a condensed outline (32) and later elaborated in booklet form (33), a holistic theory was developed. Elsasser’s (33) basic assumption was that “an organism is a source … of causal chains which cannot be traced beyond a terminal point because they are lost in the unfathomable complexity of the organism.” He used mathematical combinatorics to develop a hypothetical model of the number of possible states that can be derived from the presumption that 106–108 of the 1012 atoms (mainly C, H, O, and N) per cell are actively involved in metabolic transformations that take place at any given moment. That number of the metabolically active states is “very immense,” defined as much greater than 101,000. The essential value of this exercise in combinatorics is conceptual. It requires a holistic model of the cell, an alternative to the mechanistic model, thereby pointing from mechanistic causality toward the condition of indeterminacy. While compatible with the laws of quantum mechanics, the immensity of states obviates the need to assume that the laws of physics uniquely determine the transition from one state to another. Indeterminacy gives rise to an independent or semi-independent order of succession of states that occur in organisms, thus requiring a holistic rather than mechanistic approach (33). Phenotypic selection as the biological mode of epigenetic transformation and its reversal is the biological correlate of the holistic approach.
Support for Elsasser’s views came from Frederick Seitz, a founder of modern solid-state physics with an ardent interest in biology, who had been President of the National Academy of Science and spent a decade (1969–1978) as President of the Rockefeller University, which was fundamentally a biomedical research institute. He was in continuous contact with many of the most creatively active individuals in molecular and cell biology and was impressed with their ingenuity (34). However, he was struck by the comparative rigidity of their molecular concepts and their enormous confidence that reductionism would lead to an understanding of all aspects of living systems. Flying in the face of these attitudes was the fact that the picture of such systems was becoming ever more complex with each new major phase of development. Seitz felt that the outlook of the molecular biologists was reminiscent of the attitudes of some 19th-century physicists who believed that the universe was gigantic clockwork governed by the laws of classical physics.
While musing on the situation in biology, he came upon Elsasser’s work, which he considered a profound analysis of the status of biological systems in the physical world. He felt that the biological community had, to a substantial degree, lost sight of the forest for the trees and would continue to do so until it was forced to reexamine its own foundations, either through the appearance of obvious paradoxes or because it became enmeshed in unresolvable complexity. When that time inevitably arises, Seitz was certain that the profoundness of Elsasser’s work would be understood and would form a significant part of the cornerstone of the understanding of living systems of the biological community.
In the past two decades there has been intense activity in identifying the biochemical components of epigenetics and their role in human carcinogenesis (35). A recent example is the progression of pancreatic cancer to distant metastasis (36). During that progression heterogeneous subclonal populations emerge that drive primary tumor growth, regional spread, and distant metastasis. In fact, there is evidence that tumor progenitor genes are epigenetically disrupted at the earliest stages of malignancies, even before mutations, and thus cause altered differentiation throughout tumor evolution (37). The genetics of metastasis is very much like that of the primary tumor, which raised the possibility that epigenetic processes operate during metastasis, just as they do during the earlier stages of oncogenesis. There was large-scale reprogramming of chromatin modifications during the evolution of distant metastases. Changes were targeted to thousands of large chromatin domains across the genome that collectively specified malignant traits. Distant metastasis coevolved with dependence on the oxidative branch of a metabolic pathway, and its inhibition selectively reversed reprogrammed chromatin, malignant gene expression programs, and tumorigenesis. These findings suggest that linked epigenetic programs are selected for tumorigenic fitness during the evolution of metastasis.
The cell-culture system of transformation described earlier is simpler, more dynamic, and quantitative (38) than identification of the biochemical components of epigenetics. A combination of both methodologies would allow identification of the epigenetic components of the preneoplastic state in the form of increased saturation density as well as the progressive stages of transformation. That holds true for the reversion of transformation as well and could be worked out for metastasis after the injection of transformed cells into experimental animals. Given the immense complexity of components, it could not predict the complete outcome of changing conditions, but it might give information useful for prevention and treatment of cancer.
We acknowledge that cells contain numerous chemical mechanisms that are nearly homogeneous and are thus characterizable in physicochemical terms. Such mechanistic approaches have preoccupied most current biological research on cells. However, a cell is also a system of intrinsic complexity in which homogeneous and heterogeneous aspects are mixed in such a way that they cannot be fully separated from each other. Such complexity is magnified enormously at the level of cell populations.
The results presented here are characteristic of epigenetics as conceived by Aristotle as the alternative to preformation in embryonic development. Cell phenotypes result from the combination of genetic and epigenetic processes as influenced by input from the microenvironment. This plays out as phenotypic selection in the cell populations described here whereby growth constraints exerted on phenotypically heterogeneous cells result in the progressive conversion to the neoplastic state (14, 15, 25). Reversion also results from phenotypic selection but does so under conditions of maximum growth rate at low density.
The relevance of the results in cell culture to neoplastic development in vivo is illustrated by Farber's characterization of the development of resistant hepatocyte nodules, an essential step in liver carcinogenesis in response to carcinogenic treatment (39). Farber suggests that these nodules represent a form of physiological adaptation to a severely toxic condition resulting from carcinogenic exposures. He showed unequivocally that this remodeling is due to the redifferentiation of nodule hepatocytes, since 95% of the hepatic nodules disappear and only 5% progress to form nodules within nodules. As the great majority of nodules spontaneously change their cell structure, organization, architecture, and biochemical pattern, their development is physiological, i.e., programmed into the genome. Farber thus perceived the nodule as a physiological response to environmental perturbations rather than an aberration or a mutation. It is therefore not surprising that he borrowed the term “progressive state selection” (40) from experiments in cell culture (29), nor is it surprising that the term “physiological adaptation” can be applied to both the hepatic nodule in vivo and to cell transformation in vitro. Phenotypic selection, as envisioned here, is likely to be the mechanism by which physiological adaptation and progressive state selection lead to cell transformation in vivo and in vitro.
The most conclusive evidence for the epigenetic origin of cancer in vivo was obtained in the development of malignant mouse teratocarcinoma cells and their reversion to normal tissues (41). A solid teratoma developed from a 6-d embryo, which had been placed under a testis capsule, where it became disorganized, formed an ascites teratoma, and metastasized to the kidney. When the primary tumor was minced and transplanted intraperitoneally, it became an ascites tumor consisting of “embryoid bodies” of yolk sac “rinds” and teratocarcinoma “cores.” The cores underwent almost 200 transplant generations entirely in vivo over 8 y. Five cells from the malignant cores were injected into blastocysts, which were transferred to the uterus of a pseudopregnant foster mother, and mated to a vasectomized male. Pregnancy followed and live mice were born, some of which exhibited mosaicism in a variety of tissues that functioned normally and synthesized specific products. At maturity a mosaic male was found to produce functionally normal sperm derived from the teratocarcinoma cells. The teratocarcinoma cells, as well as the normal cells derived from them, had a normal chromosome number of 40, in contrast to the aneuploidy of in vitro passages.
The results present an unequivocal example of a nonmutational basis for transformation to malignancy, and of reversal to normalcy. The origin of this tumor from a disorganized embryo suggests that malignancies of some other, more specialized stem cells might arise comparably through tissue disorganization, leading to epigenetic, developmental aberrations of gene expression, rather than changes in gene structure. This emphasizes the importance of the cellular microenvironment in determining the normal and neoplastic phenotype of cells.
Arguably the most significant basic cancer researcher of the last century was Peyton Rous, the discoverer of the first solid cancer virus (42) and codiscoverer of the fundamental processes of initiation and promotion in carcinogenesis (6). His midcentury remarks on the cause of cancer are strikingly relevant to the overall impetus of this paper, e.g., “Most serious of all the results of the somatic mutation hypothesis has been its effect on research workers. It acts as a tranquilizer on those who believe in it, and this at a time when every worker should feel goaded now and then by his ignorance of what cancer is” (43).
Acknowledgments
Dorothy M. Rubin helped in every phase of preparation of this manuscript.
Footnotes
- ↵1To whom correspondence should be addressed. Email: hrubin{at}berkeley.edu.
↵2Present address: Human Health Assessment Branch, Department of Pesticide Regulation, California Environmental Protection Agency, Sacramento, CA 95812-4015.
Author contributions: H.R. designed research; H.R. performed research; and H.R. and A.L.R. wrote the paper.
Reviewers: A.M., retired; and R.H.R., University of California, Davis.
The authors declare no conflict of interest.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1717299115/-/DCSupplemental.
Published under the PNAS license.
References
- ↵
- Foulds L
- ↵
- ↵
- Weinberg RA
- ↵
- ↵
- Apple S
- ↵
- ↵
- ↵
- ↵
- Mondal S,
- Heidelberger C
- ↵
- Kennedy AR,
- Fox M,
- Murphy G,
- Little JB
- ↵
- ↵
- Rubin H,
- Xu K
- ↵
- Chow M,
- Yao A,
- Rubin H
- ↵
- Rubin H
- ↵
- Rubin H
- ↵
- Rubin H
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Rubin H
- ↵
- Rubin AL,
- Arnstein P,
- Rubin H
- ↵
- ↵
- Grundel R,
- Rubin H
- ↵
- Rubin AL,
- Yao A,
- Rubin H
- ↵
- Yao A,
- Huang W,
- Rubin H
- ↵
- ↵
- Elsasser WM
- ↵
- Elsasser WM
- ↵
- Seitz F
- ↵
- ↵
- ↵
- ↵
- Rubin H
- ↵
- ↵
- Farber E,
- Rubin H
- ↵
- Mintz B,
- Illmensee K
- ↵
- ↵
Citation Manager Formats
Sign up for Article Alerts
Article Classifications
- Biological Sciences
- Medical Sciences
Jump to section
- Article
- Abstract
- Number of Cells That Undergo Conversion to the Transformed State
- Reversion of Transformation by Maximizing Growth at Low Cell Density in High CS
- The Dynamics of Phenotypic Selection in Adaptation and Deadaptation to Extreme Differences of Serum Concentration
- Discussion
- Acknowledgments
- Footnotes
- References
- Figures & SI
- Info & Metrics