The Story of 3T3 Cells: A Voyage of Discovery Without an Itinerary
In the 1950s, Dulbecco and Vogt published their important papers showing that the properties of fibroblasts recently removed from the animal and cultivated for short periods (cell strains) could be transformed by infection with polyoma virus, an oncogenic virus. When I began to work on the subject, I had no trouble in reproducing the essentials of their experiments, but I was troubled by the difficulties of using primary or secondary cultures as the target for the virus because all such cultures were heterogeneous and the frequency of viral transformation was variable. I felt that one needed a more homogeneous kind of culture that was readily reproducible. Of course the cells would have to be susceptible to the virus and the transformants would have to be easily scored.
It seemed that only an immortalized (established) cell line capable of indefinite propagation could meet the requirements. Many such cell lines had been made previously, but none had the desired properties. The population density to which these cell lines could grow in culture was limited only by exhaustion of nutrients or acidification of the medium by metabolic products, because the cells had no means of controlling their growth: they could not adopt a reversible resting state, from which growth could be reinitiated by a suitable stimulus, and so could not be used to study the growth-promoting effects of oncogenic viruses. When injected into animals, most of these cell lines were able to grow as tumors. The BHK 21 line (MacPherson and Stoker, 1962) possessed little growth control, but could be used to detect transformation by polyoma virus, as its cellular morphology was altered by the virus.
Origin of 3T3 Cells
In 1961 a young medical student, George Todaro, asked to spend a term working in my laboratory at New York University Medical School. His stay was extended by a summer, then by several more terms and finally by a postdoctoral stay of several years.
We began cultivation of mouse embryo fibroblasts with a view to establishing an immortalized cell line suitable as a target for viral transformation. At that time, it was believed that mammalian cells became immortalized in culture only rarely and that it was impossible to predict when such an event might occur or under what conditions. In order to avoid haphazard conditions of cultivation, I thought it necessary to keep both the inoculation density and the transfer interval constant during repeated subcultivation, because those two variables might influence the ability of the cells to become immortalized; in addition, knowledge of the correct conditions might make it possible to develop immortalized cell lines reproducibly.
We settled on inoculation densities of 3, 6, or 12 x 105 cells per 20 cm2 dish and a transfer interval of 3 or 6 days. After a period of declining growth rate that lasted for 10-20 cell generations, during which the doubling time of the murine fibroblasts increased to as much as 100 hours, we were pleasantly surprised to find that the growth rate began to increase and in 9 of 11 cultures carried under several conditions, there evolved cell lines with doubling times of 15-24 hours (Todaro and Greenl, 1963). The ease of evolution of murine fibroblasts into immortalized lines has since been repeated in many laboratories but fibroblasts of some species, especially the human, immortalize extremely rarely. The reason for this important difference is still being studied.
We were surprised a second time to discover that although immortalization occurred under most of the culture conditions we used, the properties of the resulting cell lines depended on the inoculation density and the transfer interval. The line that emerged from cells regularly inoculated with the highest inoculation density and subcultured every 3 days (3T12) grew to the highest saturation density (about 350,000 per cm2). Cells regularly subcultured with half this inoculation density gave rise to a line (3T6) with a somewhat lower saturation density than 3T12. But the most interesting line resulted from serial subcultivation using the smallest inoculum. This line (3T3) grew as vigorously as the others as long as the cells were sparse, but arrested its growth sharply and entered a stable resting state when the cells became confluent at a saturation density of 50,000 cells/cm2, only one sixth that of secondary cultures of strains of mouse fibroblasts. When transferred with dilution, the 3T3 cells resumed exponential growth, and again reached saturation density at 50,000 cells/cm2. These experiments showed that murine fibroblasts maintained at low population density during the period of their immortalization evolved into a cell line which, when allowed to become confluent, entered a reversibly resting state at a very low population density.
This property, which could be described as highly developed density-dependent inhibition of growth, was unprecedented among previously established cell lines, most of which could grow to saturation densities 10-30 fold higher than 3T3. The exceptional behavior of 3T3 was evidently due to culture conditions ensuring the absence of selection for variant cells with the ability to continue growing at high density; without these conditions, cells of the emerging 3T3 phenotype would have been selectively eliminated.
The 3T3 line, unlike nearly all immortalized cell lines known previously, or others that we developed in the same experiments, did not give rise to tumors when injected into mice. Yet 3T3 cells were not normal either, since they had undergone many chromosomal alterations and become aneuploid. The behavior of 3T3 cells drew a clear distinction between the ability to grow indefinitely (immortalization) and the ability to form tumors (oncogenic transformation) (see GENES 2000: 29.2 Tumor cells are immortalized and transformed). There is now a great deal of evidence that telomere shortening, which can lead to chromosomal rearrangements, is an important cause of finite lifetime in cultured cells and must be overcome before the cells become immortalized (see Great Experiment: Immortalizing Human Cells with Telomerase, and GENES 2000: 29.22 Telomere shortening causes cell senescence).
The arrest of growth of 3T3 cells at low saturation density was not simply the result of exhaustion of some metabolite from the culture medium or the production of inhibitors. When cells that had entered the resting state were dissociated with trypsin and replated with dilution in the very same medium that was harvested from the resting culture, the cells had no difficulty in resuming multiplication. Similarly, if the monolayer was wounded by removal of 3T3 cells (Todaro et al., 1967), the wound was quickly repaired by cells that migrated into the wound and proliferated until the bare area was covered with cells ( Figure 1).
An important determinant of the saturation density was the serum concentration of the medium. But the property of arresting its growth at such a low density in the standard medium containing 10% serum made the 3T3 line well-suited for use as a target of transformation by oncogenic viruses such as polyoma and SV40 (Todaro and Green, 1964; Todaro and Green, 1964). The transformants could easily be detected by release of the cells from growth inhibition and the formation of dense, multilayered colonies that could readily be scored ( Figure 2). The transformants, when injected into mice, gave rise to tumors.
While work on transformation of 3T3 cells by DNA viruses continued in numerous laboratories, this cell line was relatively insensitive to the RNA viruses of the murine sarcoma group. Using the same protocol that had led to the development of the 3T3 line, it was possible to evolve new 3T3-like cell lines from fibroblasts of Balb-c and NIH Swiss mice (Aaronson and Todaro, 1968; Jainchill et al., 1969). These lines were less sensitive to density-dependent inhibition, but were more sensitive to transformation by murine sarcoma viruses than the original 3T3, which was of random-bred Swiss origin.
When cellular oncogenes were discovered, a suitable system for cell-based assay was lacking. NIH 3T3, which was highly susceptible to transfection (Smotkin et al., 1975), was found to be the best line for transformation by cellular oncogenes (Shih et al., 1979) and the use of this cell line for scoring transformation promoted vigorous development of the field (see Great Experiment: The Discovery of Oncogenes in Human Tumors) (Krontiris and Cooper, 1981; Perucho et al., 1981; Weinberg, 1982).
3T3 Cells Undergo Adipose Differentiation
In 1962, shortly after we obtained the 3T3 line, we noticed that when the cultures remained in a stationary state for a time, there developed occasional foci of cells containing what appeared to be cytoplasmic lipid. When the cells were subcultivated and resumed growth, the accumulated lipid disappeared. At that time, I did not think that the phenomenon was of sufficient interest to merit study. But about 10 years later, while I was at MIT, I noticed that the frequency of cells accumulating lipid seemed to be increasing in the 3T3 cultures that we were cultivating at the time. When we isolated a number of clones, I was surprised to find that when allowed to reach a resting state, they differed remarkably in their tendency to accumulate lipid. In some clones, one could see numerous cells containing lipid, while in others there were very few. This meant that the tendency to accumulate lipid was clonally and possibly genetically determined, but none of 18 clones isolated lacked entirely the ability to accumulate lipid.
This led me to an entirely different field of research: the significance of the lipid accumulation. It soon became evident that although 3T3 cells are fibroblasts (they synthesize fibrous collagen of types 1 and 3 and hyaluronic acid), they also possess a latent program of differentiation that, when activated, converts them to adipocytes (Green and Meuth, 1974). In the earliest stages of the differentiation, when the cells accumulated enough triglyceride to be stainable ( Figure 3, part a), they were seen to contain highly extended processes. These processes were gradually withdrawn, as the cells became more nearly spherical ( Figure 3, parts b-d), and disappeared completely by the time the cells acquired the morphology of young adipose cells ( Figure 4). During their differentiation, the cells accumulated not only triglyceride, but all of the enzymes and other proteins responsible for the synthesis and degradation of triglyceride and the hormonal regulation of lipid accumulation.
Still, even after I had published 7 papers on the subject between 1974 and 1978, the idea that adipose differentiation could occur in cultured fibroblasts continued to meet with skepticism. At a symposium I could hear murmurs. Only when we were able to show that preadipose 3T3 cells injected into athymic mice developed into mature fat pads (Green, 1979) was the concept generally accepted. Since then, a great deal of research has been carried out on the regulation of adipogenesis using 3T3 cells (Rosen et al., 2000).
In retrospect, it is possible to explain why the 3T3 line possessed the ability to undergo adipose conversion. The cells of late mouse fetuses, from which 3T3 cells were derived, are likely to contain preadipose cells, since the adipose tissues of the mouse develop early after birth. Under ordinary conditions of cultivation, in which the cells are not kept continuously in exponential growth, preadipose cells would likely differentiate into adipose cells. This can be commonly seen in primary or secondary mouse fetal fibroblast cultures allowed to become dense. Since maturing adipose cells lose the ability to multiply, any preadipose cells undergoing differentiation would tend to be eliminated from the population. It seems likely that the same culture conditions that made the 3T3 line valuable as a target for viral oncogenesis also made it preserve the program for adipose differentiation.
While a great many papers have since been published on the adipose differentiation of 3T3 cells, one important aspect of this process has received little or no attention. It was possible to apply powerful selection for preadipose 3T3 cells because it was so easy to identify a single triglyceride-accumulating cell in a culture containing 106 cells, using a simple inverted microscope, with no chemical enhancement of the phenotype. Using this selection, we showed that starting with a clone having minimal susceptibility to adipose differentiation, we could, within 2 steps of subcultivation and selection, obtain a subclone with the highest susceptibility we had ever seen (Green and Kehinde, 1976). This clone (3T3- F442A) is now the most commonly used for studies of adipose differentiation.
Evidently, 3T3 cells are subject to evolution in a direction that enhances their character as preadipocytes. This change is a stable hereditary property, because clones selected for increased susceptibility can be maintained over many serial subcultures and cell generations, if suitable precautions are taken to prevent counter-selection. What had been selected for can be regarded as an improvement of the 3T3 cell as a stem cell for adipose differentiation. Whether this took place through changes in a regulatory gene has never been established.
The Cultivation Of Keratinocytes
Prior to 1974, there had been many attempts to grow human epidermal cells in culture. The growth obtained was very limited and insufficient to permit satisfactory subcultivation. Essentially no basic or applied work could be done on cultured human keratinocytes.
In 1974, Jim Rheinwald, at that time a graduate student, was working on a mouse teratoma, a germ line tumor able, while growing as a transplanted tumor, to differentiate into a number of somatic tissues. When he put cells disaggregated from a tumor into culture, colonies of different appearance arose, including an unusual-looking epithelial cell type, together with a background of teratomal fibroblasts. Without these fibroblasts, the epithelial cell type could grow only slowly, but when lethally irradiated 3T3 cells were added, the epithelial cells grew beautifully, while growth of the teratomal fibroblasts was suppressed. The 3T3 cells thus substituted for the fibroblasts of the teratoma culture in supporting the growth of the epithelial cells. This made it possible to isolate clones of the epithelial cells and study their behavior. It soon became obvious that some of these clones were keratinocytes, the principal cell type of all stratified squamous epithelia, including the epidermis (Rheinwald and Green, 1975).
We then asked whether normal human diploid keratinocytes of the skin could grow under these conditions. They could (Rheinwald and Green, 1975), and diploid keratinocytes became a cultivable cell type ( Figure 5). With subsequent improvements in cultivation, keratinocytes became the most cultivable of human diploid cell types, first by the criterion of replicative lifespan in culture, and later by the demonstration of retention of clonal types retaining high growth potential and stem cell character (Barrandon and Green, 1987; Kobayashi, Rochat, and Barrandon, 1993; Pellegrini et al., 1999; Pellegrini, 1999; Rochat, Kobayashi, and Barrandon, 1994). The use of such cultures became important in subsequent studies of the keratins(Cooper, Schermer, and Sun, 1985; Moll et al., 1982; Wu et al., 1982) and their numerous disease-producing mutations(Fuchs, 1992), the junctional proteins and their numerous mutations, and the cross-linked envelopes of terminal differentiation.
Having at hand a method of cultivation which, if properly carried out, could generate vast amounts of epithelium whose basal layer contained cells with stem cell character, it was natural to wonder whether practical use could be made of the cultures for the treatment of injury or disease. First a method had to be developed to detach, as a coherent sheet, the epithelium made in culture by the fusion of adjacent colonies and the simultaneous elimination of nearly all the 3T3 cells (Green, Kehinde, and Thomas, 1979). The keratinocytes in such a sheet do not have the regular organization characteristic of cells in the epidermis but they know what to do when they find themselves in the right situation: applied to the surface of athymic mice (Banks-Schlegel and Green, 1980) and later of humans (O'Connor and J. B. Mulliken, 1981), they regenerated beautifully organized epidermis within a week. Regeneration of the necessary anchoring systems that attach the epidermis to the deeper tissues was considerably slower (Compton, 1989). It was soon found possible, by applying autologous cultures, to regenerate epidermis on humans who had lost over 90% of their epidermis through burns(Gallico et al., 1984; Green, 1991). Since that time, good clinical results have been achieved consistently at the Military Burn Hospital at Percy, near Paris (Carsin et al., 2000), and by a network of burn centers organized by Dr. Michele de Luca around his laboratory in Rome. A similar procedure has been used to restore corneal epithelium (Pellegrini et al., 1999; Pellegrini, 1997). Recent improvements and simplification of the method of preparing the cultures for grafting are likely to expand the use of cultured autologous keratinocytes in the treatment of disease (Pellegrini, 1999; Ronfard et al., 2000).
All successful use of keratinocytes for these purposes that I am aware of have used 3T3 cells for cultivation. The molecular basis for the ability of 3T3 cells to support the multiplication of keratinocytes has been difficult to clarify. 3T3 cells secrete into the medium products that aid multiplication of some keratinocytes, but we were unable to purify these products. Furthermore, secretion of soluble products does not account for the entire effect of the 3T3 cells. Perhaps because the 3T3 cells deposit insoluble products on the surface upon which the keratinocytes grow, the two cell types must be present in close proximity. Finally, in addition to effects on keratinocyte growth rate, 3T3 support is essential for preservation of keratinocyte stem cells (holocolones), without which grafts cannot be expected to produce durable regeneration of epithelium. As markers for these stem cells are now available, it seems likely that study of interaction of 3T3 cells and keratinocytes will advance our understanding of the properties of the stem cells.
I have briefly described some of my research on the development and study of the 3T3 line. I have necessarily omitted reference to most of the vast number of publications resulting from the use of 3T3 by others (in excess of 19,000 citations). My own work on this subject was not the result of a planned course of action, but rather grew out of increasing familiarity with the material and what could be done with it.
The editor (a distinguished scientist) of the journal to which I submitted the first article on the 3T3 line declined to publish it because (as he wrote) it would be of no interest to the journal’s readership. I took this as an encouraging sign of unperceived merit.
Howard Green is the George Higginson Professor of Cell Biology at Harvard Medical School. Dr. Green received his M.D. degree from the University of Toronto in 1947. He began his academic work at New York University School of Medicine and advanced from Instructor to Professor and Chairman of the Department of Cell Biology during his years there from 1954 to 1970. From 1970 to 1980 he was Professor of Cell Biology at Massachusetts Institute of Technology. He moved to Harvard Medical School in 1980 and served as Chairman of the Department of Cellular and Molecular Physiology from then until 1993. Dr. Green is a member of the U. S. National Academy of Sciences and the Institut de France.