6. Spontaneous Transformation in Established Cell Lines
A problem in using newly explanted or primary cultures of cells in studying spontaneous neoplastic transformation is that the growth properties of the cells are in a constant state of decline. There is an increasing rate of death with each passage, and the growth rate of the survivors steadily declines. Cells of rodent species go through a crisis period of minimal cell growth before a new cell type emerges that can grow indefinitely in culture, but normal human and chicken cells eventually expire. It is possible to study spontaneous transformation systematically in long-established cell lines of mouse embryo fibroblasts, such as the NIH 3T3 line. These can be kept in a nontransformed state, capable of forming a flat monolayer of regularly arranged cells at confluence, if they are subcultured every few days at low population density. If they are kept under the growth constraint of confluence for 2 weeks or longer, foci of transformed cells begin to appear. These foci are just like those produced in C3H10T1/2 cells after exposure to carcinogenic PAH, consisting of closely packed, multilayered criss-crossing cells. The number and size of the transformed foci depends on the length of the incubation period and on the concentration of animal serum in the medium. If a low concentration of serum is used, the confluent cells have to be subcultured and grown again to confluence for foci to appear. Each cell from a transformed focus can initiate a new focus if subcultured on a background of nontransformed cells. The continued capacity for nontransformed cells to undergo transformation depends on the type of serum and the frequency of subculture. After many low-density passages in medium containing a high concentration of calf serum, followed by growth to confluence in a low concentration of serum, the foci produced at first are small and light-staining, but they become larger and darker staining on further passages. Cells from the larger foci produce sarcomas within a few weeks when inoculated subcutaneously into immune-deficient mice. There are many types of morphologically distinct foci in a spontaneously transformed, established cell line, just as there are in primary cultures of baby mouse skin, indicating that there are many pathways to transformation in both cases. Unlike the foci from the primary cultures, however, those from the NIH 3T3 cells are tumorigenic. There is also evidence that incipient, or unapparent, changes are occurring in the cells before they produce morphologically distinct foci (11). The progressive development of the foci has parallels to the progressively malignant stages of naturally occurring cancer in animals and humans.
The characteristic of cells in culture that correlates most closely with tumorigenesis in animals is known as anchorage independence , and it signifies the capacity to multiply while suspended in a soft agar or methylcellulose. Such growth in suspension demonstrates an escape from the requirement of normal cells to attach to and spread on a solid substratum in order to multiply. Oddly enough, repeated seeding of large numbers of nontransformed cells in soft agar suspension, which inhibits their growth, sometimes results in a transformation that permits a cell to multiply in suspension; this does not happen when the cells are growing at a maximal rate attached to the surface of a culture dish. Each of the transformed clones isolated from the agar is morphologically unique, just as each human cancer is unique. Inoculation of large numbers of cells from an anchorage-independent colony into immune-deficient mice results in a rapidly growing tumor within 2 to 3 weeks, but smaller numbers of inoculated cells may produce a disproportionate delay before a tumor appears. Cells from the early tumors grow as well as the original cell culture line when placed in cell culture, but very few cells from the delayed tumors grow in culture, and most of these grow at a low rate. This indicates that the cells have undergone a selection for ability to grow in the mouse during their long incubation and thereby have reduced their capacity for growth in culture. Furthermore, there is great heterogeneity in capacity to produce colonies in agar suspension among individual cells from the delayed tumors. Such variation indicates that a continuous, progressive evolution is occurring in the delayed tumor for improved capacity for continuous growth in the mouse. The great complexity of these growth dynamics has parallels in the great diversity of genetic lesions in human cancers, as will be discussed in the next section.
An unexpected feature of the work with spontaneous transformation is that the procedure of long-term confluence that evokes it also results in some cell death in the confluent culture and in a heritable reduction in growth rate at low population density of the surviving population that later produces transformed foci. It will be recalled that mouse cells transformed by treatment with carcinogenic PAHs also multiply at a lower rate at low density than the parental nontransformed cells. This appears to be true of most, but not all, clones of transformed cells, whether induced by chemicals or spontaneously. These are the same cells that continue to multiply and form transformed foci at high population density when the surrounding nontransformed cells have ceased net multiplication. The reduction in growth rate at low density of the cells is heritable, indicating that genetic damage has occurred (12). It occurs both in transformed and nontransformed cells after prolonged incubation at confluence, indicating that the genetic damage is widespread in the population, but the extent of growth reduction is highly variable, and only certain damage results in transformation. In another cell system, a wide variety of mutagenic chemicals cause a heritable reduction in growth rate of the mutated cells, and the slowdown is related to chromosomal rearrangements and deletions. Because most of these compounds are also carcinogenic, the implication is that neoplastic transformation is frequently the result of chromosomal changes rather than point mutations.
The chromosomal changes, however, are not necessarily visible at the level of the light microscope. For example, two fibroblast cell lines were isolated from a Chinese hamster embryo. Both lines have a stable diploid mode of chromosomes that are apparently normal in appearance. By the criteria of growth in suspension, colony morphology, and tumorigenicity, one of the lines is transformed and the other is not. On prolonged exposure to the cytotoxic drug methotrexate (see Aminopterin, Methotrexate, Trimethoprim, and Folic Acid), the transformed line develops resistance three times faster than the nontransformed line. A major part of the resistance to methotrexate is the result of gene amplification (increase in gene copy number) of the gene that codes for the enzyme blocked by methotrexate. In this treatment, large-scale changes in chromosome composition are seen. The implication is that chromosomal changes occur much more readily in neoplastic than in normal cells, with the further implication that those changes underlie the progressive increases in malignant behavior (invasiveness, metastasis) that occur in tumors.
The same biochemical effect that is induced by treating cells with methotrexate can be achieved by depriving the cells of the vitamin folic acid, which is the substrate of the enzyme blocked by methotrexate. Either treatment causes single- and double-strand breaks in DNA. Folic acid deficiency also increases the metastatic capacity of cancer cells and causes a heritable decrease in the growth rate of cells. It has been suggested that dietary deficiency in folic acid may contribute to the onset of cancer in humans. It may be recalled that a deficiency of choline or methionine results in cancer in experimental animals. Therefore, a part of the damaging and transforming effect of prolonged incubation of cells at the high population density of confluence could be due to depletion of vitamins or amino acids from the medium. Heritable reduction in growth rate of cells is also a hallmark of cell aging in animals and could be the result of accumulated chromosomal damage below the resolution of the light microscope. The relation between the genetic damage, age, and neoplastic transformation will be discussed below.
7. Summary of Findings in Experimental Cancer and Cell Transformation
Injection of a single RNA tumor virus particle can initiate a tumor in animals or transformation in culture. Molecular analysis has shown that a single transforming gene in the viral RNA is the responsible agent. However, such transforming viruses are of little importance in nature, and the tumors they produce in adult animals usually regress. Cellular counterparts of viral transforming genes do not transform cells unless associated with strong retroviral promoters. Transformation initiated spontaneously or by physical and chemical carcinogens is a much more complex and multistaged affair than viral infection, and in many cases it involves large-scale changes at the chromosomal level. A single initiating treatment of cultured cells with a carcinogenic PAH or X-rays genetically alters most or all of the cells in a population in a manner that increases the probability of transformation in their descendants, but the transformation itself only occurs in a small fraction of descendants of each altered cell. The population-wide alteration induced by carcinogens was at first thought by many to be epigenetic or physiological in nature. However, it has been shown to be heritable, and there is evidenc that treatments such as X-irradiation, chemical carcinogens, or prolonged periods of crowding and nutritional deficiency cause genetic damage that apparently differs from cell to cell, but genetically destabilizes most of them. There is a general association of transformation with chromosome changes in these treatments. Some of the carcinogens interact directly with DNA, but many do not, indicating that the genetic changes in the latter case are indirect. Such indirect effects have been called epigenetic or dysgenetic, but the ultimate effect is on DNA structure. Many of the indirectly acting agents are clastogenic or disruptive at the chromosomal level, causing recombination, deletion, or asymmetric distribution of chromosomes at mitosis. Such changes are not detected in the Salmonella test for mutagenesis, which depends on local base substitution or frameshifts in DNA. They can, however, be detected in animal cell culture. There is great diversity in the morphology of transformation, indicating a very large variety of transforming genetic changes. But there are many more chromosomal changes that do not transform cells, so the transforming changes can be considered a selection from among many possibilities. This indicates that it may not be possible to predict which chromosomal changes will cause transformation. This is all the more true because the sensitivity of cells to transformation varies with the initial state of the cells and the methods and materials used in subculturing the cells. Cells thawed from different frozen vials of the same stock of tumor cells will exhibit different growth properties, especially when those properties are repeatedly measured in serial subcultures over an extended period of time. The unpredictable details of the long-term growth behavior of cells, combined with its general dependence on the conditions of culture, are not dissimilar from the extreme heterogeneity and unpredictability of tumors in animals, and tumors are in stark contrast with the relative constancy of function and behavior of normal tissues over extended periods of time. In vitro transformation reveals great sensitivity to environmental conditions, such as cell density, serum type, and frequency of subculture, which therefore reflect an epigenetic or conditional aspect of transformation.
8. Neoplastic Transformation in Humans
The concerted application of molecular genetics to human cancer began with the transformation of NIH 3T3 cells in culture by DNA from a long-term cell population that had been cultured from a human bladder cancer and with the demonstration that the transforming gene was a mutant form of the cellular ras gene. (See section on Viral transformation .) Given the common occurrence of genetic variation in cell culture, especially that of human tumors, there is reason to question the conclusion that the mutation caused, or even existed in, the tumor. Those doubts are reinforced by the finding that the gene has to recombine with a strong mouse promoter in order to transform the NIH 3T3 cell. Nevertheless, the finding set off attempts to detect the altered gene in other human cancers. The mutation was reported in the benign tumors (adenomas) that precede colon cancers. Other genetic changes were found in early stages of colon cancer, and it was concluded that five to seven mutations were required to produce advanced stages of cancer. However, it was then found that more than 20% of the alleles of heterozygous genetic loci were lost in the average sporadic colon cancer, which casts doubt on the estimates of the number of mutations needed to produce the cancer. Indeed, the estimates of the number of altered genes in the average case of sporadic colon cancer is now estimated at greater than 25% of the entire genome (13), which implies many thousands of genetic changes during development of the cancer. It should be kept in mind, however, that ras mutations found in mutagen-induced rat mammary tumors actually arose from preexisting ras mutants. That raises the question of how many of the mutations reported in tumors arose before the tumor, and what role, if any, they had in development of the tumor. Another finding that raises specific questions about the role of ras mutations is that N-ras gene function of mice can be knocked out entirely in the earliest stages of development without affecting either the growth or development to adulthood. As methods for detecting genetic changes improve, the number of such changes in tumors increases. When the technique of comparative genomic hybridization was introduced into studies of human mammary cancer, 21 new chromosomal subregions of gene amplification were added to the five already known. The same methodology applied to prostate cancer detected about 30% of the chromosomal regions with a significant increase or decrease in gene copy number in the average case, and some of the cancers had such alterations in more than 50% of the sites (14). These changes involve large chunks of chromosomes, indicating rearrangements and large deletions or amplifications. Although some particular sites were altered in a high percentage of the prostate cancers, no two cancers had the same distribution of altered sites. This confirms what had been reported earlier, that the karyotype of glial tumors was different in every case and, furthermore, that it rapidly changed when the cells were placed in culture. The very large number of copy number changes in genes of prostate cancer seems surprising in view of evidence that nearly 50% of metastatic prostate cancers have a diploid DNA content. The diploid DNA content in so many advanced prostate cancers shows that equal proportions of the genome can be lost or gained, resulting in an overall balance of genetic material. The large number of genetic alterations in cancers indicate great instability in the tumors, but it also creates problems in determining which of the alterations bear a causal relation to development of the tumor, and how many are required to produce its progressive growth. It also raises questions about the source of such instability. One possibility is that an epigenetic change in methylation might lead to defects in chromosome disjunction and distribution at mitosis. Marked decreases are found in methylation among each of four genes tested in benign and malignant colonic neoplasms. Because the hypomethylation is prevalent in the benign tumors, the great majority of which have been reported to have no genetic alterations, it is possible that the altered methylation patterns drive the chromosomal changes. In support of such a possibility is the finding that experimentally induced hypomethylation induces transformation in a diploid line of Chinese hamster cells at a high frequency and that the transformation is in every case associated with DNA hypomethylation and the presence of an extra copy of part of a single chromosome. Hence, there may also be a relation between methylation and fragility of chromosome sites, which determines the likelihood of chromosome breaks in particular tissues.
A percentage of human tumors occur in dominantly inherited patterns in families. About 15% of colorectal cancers fall into this pattern (12). Hereditary nonpolyposis colorectal cancer (HNPCC) is one such familial cancer. It is related to germ-line mutations of mismatch repair genes. This results in changes in short repeated sequences of DNA (microsatellites) in colorectal cancers. The question arose whether similar changes occur in normal tissues of HNPCC patients who had few tumors. Using special amplification methods, it was found that the normal tissues of these patients had a high incidence of these mutations. This shows that the existence of many mutations in cells of tissues throughout the body is not sufficient to initiate cancer, nor is the presence of the mutations in the tumor proof of their causative role. This moved a prominent geneticist to recommend a more rigorous look at the evidence that mutation induction constitutes a rate-limiting step in carcinogenesis (15). Other factors that contribute to carcinogenesis will now be considered.
9. The Role of the Tissue Environment in the Origin of Tumors
Teratocarcinoma is a malignant tumor derived from germ cells and consists of many different cell types. It can be induced experimentally in mice by transplanting 6-day-old mouse embryos under the testis capsule of adult male mice. The normal diploid tumors can be grown in culture and will produce malignant growth when a single cell is inoculated under the skin of mice. When the same cells are inoculated into the very early embryo (blastocyst stage) of mice, normal development occurs, and redifferentiated teratocarcinoma cells contribute to normal development of many tissues, including some never seen in the original tumors (16). They also contribute to the germ line, which produces normal progeny mice from gametes of the teratocarcinoma line. The authors of this work concluded that the conversion to malignancy in all likelihood did not involve mutational events and was completely reversible to normal function and appearance. While it has not been proven unequivocally that there were no mutations involved in causing the teratocarcinoma, it is apparent that the local environment was the determining factor in the origin and in the reversal of this malignant behavior.
A related situation exists in the newt Triturus cristatus . Inoculation of carcinogenic PAHs in nonregenerative regions of the newt induced metastasizing tumors that killed the host. Tumors induced in regenerating regions healed spontaneously and differentiated to normal tissues. The malignant behavior of these tumors, therefore, is dependent on the lack of regenerative potential of the surrounding tissue.
There is evidence that the local environment of tissues changes with age. For example, there is a 50fold increase in the activation of the "inactive" X-chromosome in the liver of aging female mice, and there is a sixfold decrease in the transcription of a certain globulin gene in liver cells. The expression of genes is related to their state of methylation, which may serve as a surveillance mechanism for chromosome loss. The 5-methylcytosine content of tissues decreases with age in mammals, including humans, and the rate of loss of these methylated nucleotides correlates inversely with age in two rodent species that differ in lifespan by a factor of two. This suggests a tie-in between aging, tissue organization, methylation, and chromosomal alterations underlying cancer. The incidence of cancer in humans increases sharply with age (17), and the susceptibility of tissues to carcinogenesis also increases with age. If lung tissue of mice is damaged by X-irradiation or certain cytotoxic chemicals, there is a large increase in the number of metastases that occur in the lung when metastatic cells are inoculated intravenously. Inoculation of rat liver cancer cells into the liver of old rats is much more likely to produce a progressively growing tumor there than if the cells are inoculated into the liver of young mice. In contrast, the liver cancer cells inoculated into the liver of young mice are much more likely to differentiate into normal liver cells. It is apparent that the local environment of the liver in old mice is more favorable for tumor growth than the liver of young mice.
The tissue environment in which human cancers develop appears to be different from that of normal tissues. In cancer of the bladder, there is a gradient of biochemical and cytological abnormality extending for some distance from the edge of the tumor. Loss of alleles of some genes occurs in morphologically normal tissue adjacent to breast cancers. Cancer of the esophagus in patients with predisposing conditions is preceded by chromosome changes in large areas of the esophagus in which cancer later arises. The transitional mucosa immediately adjacent to colorectal cancers contains a variety of biological abnormalities but no evidence of genetic change. There is a 15-fold increase in transcripts of the methyltransferase enzyme catalyzing DNA methylation in normal colonic mucosa from patients with either polyps or cancer of the colon. It is generally believed that tumors progress by a sequence of genetic changes. As each new mutation appears that selectively favors the mutated cell, the microenvironment of altered cells in which the newly mutated cell is growing may be more favorable for further tumor development than normal tissue. These are all factors that must be taken into consideration in understanding the growth of tumors, and particularly their high degree of genetic instability. A comparable instability is seen when normal rodent cells are dissociated from one another and grown in monolayer culture. Human cells under the same conditions remain largely diploid but lose reiterated sequences of DNA, including telomeric DNA. Chromosomal instability persists after the cells have established themselves as a cell line, as indicated by the report that no two cells of a line of rat hepatoma cells were found to be karyotypically identical. Hence, it is possible that the high degree of genetic instability found in human cancer is due not only to the intrinsic instability of the cells, but to the loss of normal organization within and adjacent to the tumor.
10. Overview
There are many ways of producing neoplastic transformation: with high efficiency by RNA tumor viruses, with low efficiency by carcinogens, and by dependence on the environment, as in prolonged crowding in certain cell lines in culture. The resultant transformed cells are usually genetically altered, as indicated by the irreversibility of the altered state and the demonstration of altered gene and chromosome composition of the cells (except for exceptional cases like teratocarcinoma, which exhibits in its very origin and reversibility the importance of the surroundings in which cells grow). The idea that transformation could result from mutation in one or a few genes arose first from transformation by RNA tumor viruses and seemed to receive confirmation by transforming especially sensitive NIH 3T3 cells with an altered ras gene isolated from an established cell line that originated from a human bladder cancer. Attempts to generate the same kind of transformation with other genes from tumors were unsuccessful. It was then shown that the transformation by the ras gene depended on recombination with strong promoter elements, with no indication that it had such a relation in the original tumor. Furthermore, the NIH 3T3 target cells were shown to transform by themselves without transfection, by long-term incubation under growth constraining conditions. There is also the likelihood that the ras mutation in chemically induced tumors of rats occurred during normal development.
Despite these reservations, the idea grew that cellular genes with nucleotide sequences like those of RNA tumor viruses were potential "oncogenes" if activated (mutated). It was then found that some genetic loci that were heterozygous in normal tissues of an individual with cancer were homozygous in the tumor. These "loss of heterozygosity" (LOH) genes were then considered to be tumor suppressors in the heterozygous state in normal tissue, although there was no direct evidence for such a role aside from the clonal occurrence of the homozygous state in tumors. There were mathematical arguments that the development of solid cancers required five or more steps; and mutational analysis, particularly of colorectal cancers, was taken as support of the argument. As methods detecting genetic changes were refined, however, more and more of them were found in cancers. Chromosome counting evolved to banding of chromosomes, methods were developed for allelotyping DNA, and comparative genomic hybridization could detect changes in gene copy number in any region of the genome. These methods then detected changes in some common cancers at an average of 25% to 30% of the chromosomal sites per case, and in more than 50% of the sites in a few cases. This indicated a great instability in the cancers, with alterations in thousands of genes. It is well known from classic genetics that the expression of any multigenic phenomenon is very dependent on the genotypic milieu, so that a given mutation may be deleterious in one genetic milieu and advantageous in another. Thus, the combination of mutagenic changes in genotypic milieus that are different in every human, plus the sensitivity of multigenic phenotypes to the surrounding environment, account for the difficulty in predicting the likelihood of nonfamilial or sporadic cancers or their outcome once they appear. Such a high degree of complexity and the problems of establishing causal chains in organisms were anticipated in the theoretical work of Walter Elsasser (18). Even where there is a dominant germ-line mutation that favors development of cancer with a probability approaching unity, the time of onset cannot be predicted, and only a very small fraction of the cells, all of which carry the mutation, become transformed; to do so, additional mutations are required, but they can be found in normal tissue as well. To achieve a better understanding of cancer, it will be necessary to take into account the genome of the transformed cell, the state of the surrounding tissue, the age of the organism, its diet, and the environment in which it lives.
The reader should also consult the entries on Antioncogenes, Cell cycle, Contact inhibition, Oncogenes, Protooncogenes, ras genes, Rous sarcoma virus, Somatic mutation, Tumor necrosis factor, Tumor promoters, and Tumor suppressor genes for further information on neoplastic transformation.
