�9 1989 by The Humana Press Inc. All rights of any nature whatsoever reserved. 0163--4992/89/1512--0015502.00

DNA Methylation and Epigenetic Mechanisms

ROBIN HOLLIDAY

Genetics Division, National Institute for Medical Research, Mill Hill, London NW7 1AA, UK; and Present address: CSIRO, Laboratory

for Molecular Biology, PO Box 184, North Ryde, Sydney, NSW 2113, Australia

Received February 18, 1988; Accepted September 3, 1988

ABSTRACT

Genes are essential for the transmission of genetic information from generation to generation, and this mechanism of inheritance is fully understood. Genes are also essential for unfolding the genetic program for development, but the rules governing this process are obscure. Epigenetics comprises the study of the switching on and off of genes during development, the segregation of gene activities following somatic cell division, and the stable inheritance of a given spectrum of gene activities in specific cells. Some of these processes may be explained by DNA modification, particularly changes in the pattern of DNA methylation and the heritability of that pattern. There is strong evidence that DNA methylation plays an important role in the control of gene activity in cultured mammalian cells, and the prop- erties of a CHO mutant strain affected in DNA methylation are de- scribed. Human diploid cells progressively lose cytosine methylation during serial subculture, and this may be related to their in vitro sene- scence. There is also evidence that DNA modifications can be inherited through the germ line. Classical genetics is based on the study of all types of change in DNA base sequence, but the rules governing the activity of genes by epigenetic mechanisms are necessarily different. Their elucidation will depend both on a theoretical framework for development and on experimental studies at the molecular, chromo- somal, and cellular levels.

Index Entries: Epigenetics; DNA methylation; 5-azacytidine.

Cell Biophysics 15 Vol. 15, 1989

16 Holliday

INTRODUCTION

The rules governing the transmission of genes from generation to generation have been fully revealed by classical genetics. In higher orga- nisms genes are also essential for unfolding the program of development from the fertilized egg to the adult, but the strategy of the genes in con- trolling this process is very poorly understood. Waddington (1) coined the term epigenetics for the study of the temporal and spatial control of gene activities during development. The program depends on the segre- gation of gene activities and the somatic inheritance of given patterns of gene activities in differentiated cells capable of division, as well as in cells committed to differentiation at some later time, such as stem line situa- tions.

It is widely believed that specific protein-DNA interactions are re- sponsible for the control of gene expression during development, but it is not clear whether DNA sequences per se are responsible for the necessary controls. The DNA methylation hypothesis proposes that additional spe- cificity is provided by the imposition of particular patterns of cytosine methylation, which can be faithfully transmitted through the activity of DNA maintenance enzymes, and which are changed in particular cell lineages by switching mechanisms (2,3). There is considerable evidence that gene expression is related to DNA methylation, and that the pattern of cytosine methylation is heritable (4-6).

REACTIVATION OF SILENT GENES

Many enzyme deficient rodent cell lines can be strongly reactivated to wild type after treatment with 5-azacytidine (5AC), which is a potent de- methylating agent. Holliday (6) lists 17 such cases. The phenotypes, which at first sight appear to be a result of a mutation in a structural gene, are in all probability a result of the epigenetic inactivation of these genes by DNA methylation. Appropriate cell lines provide excellent material for understanding the rules governing the switching off and the reactivation of genes by methylation and demethylation. So far, it seems likely that pseudo-diploid cell lines such as CHO, can inactivate one or both copies of certain genes, provided that this does not cause a selected disadvantage during the growth of cell cultures. The situation where one gene copy is active and the other is inactive can explain the "functional hemizygosity" originally proposed by Siminovitch (6-8). Mutagen treatments can alter the base sequence of the active gene to produce a mutant phenotype. The second silent copy of the gene has a wild type sequence and this can be reactivated by 5AC. However, this general model has not yet been rigor- ously tested by current molecular procedures.

Cell Biophysics Vol. 15. 1989

DNA Methylation and Epigenetics 17

A CHO MUTANT AFFECTED IN DNA METHYLATION

Gene reactivation provides a means of isolating DNA methylation mutants (9). The rationale is based on the assumption that the inactive state of a gene is maintained by DNA maintenance methylase activity and that a defective methylase will result in the high frequency of spontaneous reactivation. Since the enzyme deficiency might well be lethal, it is neces- sary to search for a temperature sensitive defect that reactives a given gene at 39, but not at 34~ A thymidine kinase deficient strain (TK-) of CHO is available that is strongly reactivable by 5AC and TK § revertants are able to grow in Littlefield's HAT selection medium. This strain was treated with the potent mutagen ethyl methane sulphonate and then grown in bromodeoxyuridine to select against any TK § cells. Individual colonies were transferred to wells incubated at 39~ and subsequently in HAT medium at 34~ Duplicate wells were kept throughout at 34~ in normal medium. Among 7000 isolates tested, two produced an elevated frequency of TK § colonies and one of these was temperature sensitive. This isolate was designated tsm. The transfer of tsm cells to 39~ for 24 h followed by selection in HAT medium resulted in a thousand fold increase in reactivation, compared with a millionfold increase after 5 AC treatment. These CHO cells are also metaUothionein (MT) deficient, and, therefore, cadmium sensitive (Cds), and this gene is also reactivated by 5AC to cad- mium resistance at high frequency. The tsm mutant was shown to produce cadmium resistant revertants, at approximately a thousandfold the nor- mal frequency, after 24 h incubation at 39~

Molecular studies with probes for the TK and MT genes have also been carried out that show that induced reactivation by AC and sponta- neous reactivation in the tsrn strain is associated with the demethylation of cytosine at specific sites (9). However, the mutant is not lethal at 39~ nor is there any discernible decrease in total 5-methyl cytosine (5mC) in the DNA, as determined by HPLC chromatography. It appears that the strain is defective in some aspect of methylation control, but probably does not have a temperature sensitive DNA maintenance methylase.

The CHO system is being exploited in the search for further DNA methylase mutations. As well as TK- and Cd s phenotypes, it is also known that X-ray sensitive strains (xrs) are reactivable by SAC (7). CHO cells also require proline for growth and a high frequency of pro § cells is induced by 5AC (10). In addition, a number of HPRT- and APRT- strains (deficient in hypoxanthine and adenine phosphoribosyl transferase, respectively) have been isolated by serial selection in increasing levels of 6 thioguanine and 2,6-diaminopurine without mutagen treatment. It is hoped that such enzyme deficient isolates will arise by de novo methylation and can also be used in gene reactivation experiments. The existence of several reactivable genes in a single CHO strain should make it easier to identify new mu- tants incapable of maintaining DNA methylation.

Cell Biophysics Vol. 15, 1989

18 Holliday

DNA METHYLATION IN DIPLOID CELLS

Immortalized cell lines maintain a constant level of DNA methylation, whereas diploid cells progressively lose 5mC during routine subculture. This has been shown for primary cultures of mouse, hamster, and human fibroblasts (11,12). Moreover, the rate of loss correlates with the in vitro lifespan of these cells, which suggests that the loss of methylation may be an important component of their in vitro aging. If so, 5AC treatments that are known to lower the total level of 5mC should have a strong effect in reducing final lifespan. This has been confirmed for human diploid fibro- blast strain MRC-5 (12-14). This significance of these observations is that a single treatment of 5AC.or 5-azadeoxycytidine does not have any dis- cernible effect on the phenotype once the cells have recovered from the treatment. Nevertheless, a memory of the treatment is retained and is manifested by premature senescence later on. It is also significant that Hayflick demonstrated many years ago, that in vitro lifespan depended on the number of cell divisions rather than from chronological time (15). The loss of methylation would be expected to occur from inefficient main- tenance activity on the hemimethylated DNA molecules at the replication fork. It is possible that the loss of too many methyl residues would have progressively severe effects on the control of gene activity. Thus, these epigenetic errors may contribute to the aging process (6,14). Recently, it has been shown that a gene on the inactive X chromosome is reactivated in liver cells during the aging of mice (16,17).

EPIGENETIC CHANGES IN THE GERM LINE

Epigenetic controls are normally discussed in relation to the behavior of somatic cells in higher organisms, but recently it has become clear that germ line cells are also subject to controls, probably based on DNA methyl- ation. According to Mendelian inheritance, maternal and paternal chromo- somes should be equivalent, but this is now known not to be true for certain insects and mammals. This phenomenon is known as chromosome imprinting, which is in some way imposed on germ line chromosomes and is heritable, but is reversed in the next generation. Recently evidence has been obtained that chromosome imprinting is based on differences in DNA methylation (18-20). Other nonMendelian anomalies in human chromosome inheritance, such as the fragile X syndrome, may be related to the failure to reactivate part of the inactive X chromosome prior to the production of germ cells (21), and there is strong evidence that X chromo- some inactivation is related to DNA methylation (22). Another example comes from the studies of maize. An inactive form of the activator (Ac) transposable element has been shown to be methylated in CG rich regions, whereas the active form of the element is demethylated (23,24). These

Cell Biophysics Vol. 15, 1989

DNA Methylation and Epigenetics 19

differences can be transmitted through the germ line as well as in somatic tissue.

CONCLUSIONS

Heritable epigenetic variants occur in mammalian cell lines, and there is strong evidence that the inactivation of genes is associated with cyto- sine methylation at the specific sites, and that the reactivation of these genes is a result of demethylation. It has been proposed that heritable dif- ferences in gene activity that are not a result of changes in DNA sequence should be referred to as "epimuta t ions" to distinguish them from classi- cal mutations (6, 7). The rules for forward and reverse epimutation will be different from those of normal mutation, but so far little quantitative in- formation is available. However, the low level of spontaneous reversion of certain enzyme deficient epimutants makes it possible to isolate strains with defects in DNA methylation. Classical mutations cannot be recog- nized by repair enzymes once they are fixed in the genome, whereas this may not be true for epimutants. It has been proposed that one of the func- tions of recombination at meiosis is to detect and repair deficiencies in methylation at important regulatory sites (6,25).

Diploid cells progressively lose 5mC during serial subculture, where permanent lines are in a steady state and retain a constant level of 5mC. It is possible that such cells can replace any lost 5mC residues by de novo methylation. It is well known that diploid rodent fibroblasts become im- mortalized and transformed at fairly high frequencies, whereas human cells are very refractory to transformation (6,14). Mutation rates are similar in rodent and human cells, and it is possible that epigenetic changes occur much more frequently in rodent cells. These may contribute to transfor- mation of the normal phenotype.

Methods, such as genomic sequencing, are available to examine dif- ferences in methylation in control regions of DNA in different cells and tissues: As more and more genes are identified that may have important roles in the control of development of differentiation, it should be possi- ble to apply specific tests of the methylation hypothesis. However, in the longer term, a conceptual or theoretical framework will be necessary to help unders tand the strategy of genes in unfolding the program of devel- opment . The bewildering complexity of this process will only be under- stood by a combination of molecular, cellular, genetic, and developmental studies within such a theoretical framework (26).

REFERENCES

1. Waddington, C. H. (1953), Symp. Soc. Exp. Biol. 7, 186. 2. Holliday, R. and Pugh, J. E. (1975), Science 187, 226.

CeU Biophysics Vol. 15. 1989

20 Holliday

3. Riggs, A. D. (1975), Cytogenet. Ceil Genet. 14, 9. 4. Razin, A., Cedar, H., and Riggs, A. D. (1984), DNA Methylation: Biochemistry

and Biological Significance, Springer Verlag, NY. 5. Adams, R. L. P. and Burdon, R. H. (1985), The Molecular Biology of DNA

methylation, Springer Verlag, NY. 6. Holliday, R. (1987), Science 238, 163. 7. Jeggo, P. A. and Holliday, R. (1986), Mol. Cell. Biol. 6, 2944. 8. Siminovitch, L. (1976), Cell 7, 1. 9. Gounari, F., Banks, G. R., Khazaie, K., Jeggo, P. A., and Holliday, R. (1987),

Genes and Dev. 1, 899. 10. Harris, M. (1984), Somatic Cell. Mol. Genet. 10, 615. 11. Wilson, V. L. and Jones, P. A. (1983), Science 220, 1055. 12. Fairweather, S., Fox, M., and Margison, P. (1987), Exp. Cell Res. 168, 153. 13. Holliday, R. (1986), Exp. Cell Res. 166, 543. 14. Holliday, R. (1985), in Molecular Biology of Ageing. Woodhead, A. D., Blackett,

A. D., and Hollaender, A., eds., pp. 269-283, Plenum, NY. 15. Hayflick, L. (1965), Exp. Cell Res. 37, 614. 16. Wareham, K. A., Lyon, M. F., Glenister, P. H., and Williams, E. D. (1987),

Nature 327, 725. 17. Holliday, R. (1987), Nature 327, 661. 18. Reik, W., Collick, A., Norris, M. L., Barton, S. C., and Surani, M. A. (1987),

Nature 328, 248. 19. Sapienza, C., Peterson, A. C., Rossant, J., and Balling, R. (1987), Nature

328, 251. 20. Swain, J. L., Stewart, T. A., and Leder, P. (1987), Cell 50, 719. 21. Laird, C. (1987), Genetics 117, 587. 22. Monk, M. (1986), BioEssays 4, 204. 23. Schwartz, D. and Dennis, E. (1986), Mol. Gen. Genet. 205, 476. 24. Chomet, P. S., Wessler, S., and Dellaporta, S. L. (1987), EMBO J. 6, 295. 25. Holliday, R. (1988), The Evolution of Sex, Michod, R. E. and Levin, B. R.,

eds., pp. 45-55, Sinauer, MA. 26. Holliday, R. (1988), J. Theoret. Biol. 132, 253-262.

Cell Biophysics Vol. 15, 1989