L E T T E R S . . . .

Does DNA methylation control transposition of selfish elements in the germline? The provocative article by Yoder et aL ~ on DNA methylation proposes that the control of intragenomic parasites is the primary function of DNA methylation in mammalian cells. The authors argue that other functions, most notably selective activation of gene expression during development, are either secondary effects, or are illusory. There are several reasons to question their analysis, not least the evidence that transposition in germ cells and early embryos is not controlled by methylation.

The idea that selective loss of DNA methylation controls activation of certain genes has been a driving force behind the DNA methylation field for many years (e.g. Ref. 2). As Yoder et aL point out, the evidence for operation of this mechanism in vivo remains tantalizingly indecisive, although they do not comment on the suggestive evidence that is available3.+*. Before dismissing the possibility out of hand, it is wise to recall that there are very many genes that have promoters within the methylated majority of the genome, and our knowledge of the regulation of all but a handful of these is, to say the least, sketchy. Given that methylation can repress transcription, and that evolution is blindly opportunistic, it seems premature to conclude tl~at no gene will make use of DNA methylation during the silent phase of its expression programme.

The advent of genomic sequencing on a large scale has shown graphically the extent to which the human and other genomes are dominated I)y transposable elements of one sort or another (e.g. Ref. 5). The great majority of these are relics of once-active elements and, therefore, pose no threat except, perhaps, as sources of illegitimate recombination ¢,. Altogether, CpGs in Alu dements have been estimated to account for atx)ut a third of potential methylation sites in human DNA (Ref. 7). The analysis of CCGG sequences by Y(mter et aL extends this u~ other families of elements in the human. genome and shows that the bulk of

such sites are in the inactive remnants of transposition. This is not surprising. It has been known for many years that less than 5% of human DNA enc~Ktes proteins of the organism. Because much of the remainder consists of degenerate copies of transposable elements, it is to be expected that most DNA methylation in the genome will be in the elements.

Relics of transposons might account for much of the methyI-CpG in the genome, but these are clearly not the only places where methylation is found. Genes belonging to the host genome are also methylated. In fact, it is difficult to find DNA sequences that are methylation-free in the mammalian genome. C"~ islands constitute a discrete class of nonmethylated .~quences and, often, transcriptionally active regions of the genome are hypomethylated, The impression

]: ~olo:6~ Za~E on: o f t h e g ~ r ~ [in 6 i:

: s i ~n i fi eani:l y li m i te d: b y D I~lA::rn ~hylatioh.'.'.: i :~.,i:::

one gets is not that specific sequences attract methylation, hut that the methylated state is a default conditio#, which affects all regions of the genome except those where specific protection against the DNA methyltransferase occurs. To sustain the view that methylatiorl is targeted at transposable elements, it is necessary, to explain why exons of genes, as well as other r~gions where there is no evidence of transposon-like sequences, are nevertheless methylated.

The central thesis of the Yoder etal. article is that mammalian transposons are effectively controlled by DNA methylation. This idea can be seriously criticized, as the data argue for the optx~site conclusion. Colonization of the genome by transposable elements must Ilave occurred in germ cells, or in totipotent cells that can become

TIG DI.:CEMBER 1997 VOL. 13 No. 12

germ cells. Therefore, it is here that one would expect transcription of active elements, which are potential intermediates in transposition, to be repressed. 111e data show that transcripts of a number of elements are, in fact, abundant at precisely the developmental stages where one would expect restraint by DNA methylation to be important. For example, both the RNA and proteins of intracistemal A particles are present throughout the earliest stages of mouse embryogenesis, as well as in testis and in ovary'Z~0. Similarly, the intact endogenous retrovims-like element VL30 is actively transcribed in early mouse embryos L*, as are intact LINEd elements in mouse testis 1-~ and human germ cells13.

Not only are the transposons highly transcribed, they can also be hypomethylated at these crucial stages. Thus, the subset of human Ah~ elements that are still capable of transposition are unmethylated in male germ cell DNA (Refs 7, 14). Intracistemal A particle genomes are also unmethytated in mouse oocytes~5, and there is evidence for hypo-methylation in testis. Mouse L1 elements too are hypomethylated during male and female gametogenesisn& The presence or absence of methylation in totipotent cells of the early embryo has been tested at few elements, but the indications from other work are that methylatlon is immaterial at these stages r . Absence of DNA methyltransferase from early embryonic cells has not so far led to any obvious phenotype TM, and two of the proteins that mediate the repressive effect of DNA methvlation on transcription, MeCP1 and MeCP2, are absent in embryonic cells to.2°. Thus even if tuethylation were to be present in these cells, their ability to interpret it as a signal to repress is qu~tionable.

Altogether the evidence casts serious douht on the idea that colonization of the germline genome has been significantly limittxt by DNA methyhtion. Indeed, on suta,eying the debris of transposition in the human genome, one has the impremion that whatever mechanism was responsible for putting a brake on transposon prolifenition did not work particularly well. In this connection it is worth remembering

f that the host genome is not the only party whose future depends on limiting transposition. The successful transposon is one that exercises 'self control', as aggressive elements would doubtless die out with their damaged host genomes. Because of this constraint (one that applies much less to viruses, which can escape the crippled host with impunity), it is conceivable that transposons rely on intrinsic and best mechanisms to ensure that their movement in the gemdine is rare. DNA methylation, it seems, is unlikely to be involved.

The foregoing discussion has centred on germ cells and totipotent embryonic cells because it is there that the danger of transposition seems most acute. The situation in somatic cells of the organism is different, as here the data are compatible with a role for methylation in repressing transcription of the elements. Active Alu elements and intaacistemal A particles, for eyample, are heavily methylated in all tested somatic cell types 7,10.14. Why should the organism fail to repress parasitic elements in the potentially immortal gerrrdine, but do so efficiendy in transient somatic cells? It is possible that suppressien of mutagenesis is the advantage, though this seems unlikely given the extremely low frequency of such events in germ cells, where transcription is file.

L E T T E R S

It is worth considering that there may be an alternative benefit of transposon silencing in somatic cells. The dependence of differentiated cells on tight programming of active and inactive genes may be disturbed by the presence of transcripts from the large number of transposable elements in the cell. Silencing of this and other 'transcriptional noise' (see Refs 22-24) may yet prove to be an important function of DNA methylation in somatic cells.

References 1 Y<nler, J.A, Walsh, CP. and

Bestor, T.H. (1997) Trends Genet. 13,, 335-340

2 Razin, A. and Riggs, A.D. (1979) Science 210, (~4--610

3 Pan,ush, Z., Keshet, I., Yisraeli, J. and Cedar, H. (1990) Cell 63, I229--1237

4 Lichtenstein. M., Keini. G., C:.:dar, H. and Bergman, Y. (1994) Cell76, 913-923

$ Martin-Gallardo, X. eta!. (1992) Nat. Genet. 1.3A-39

6 Kricker, M.C.. Drake. J.W. and Radman, M. (1992) Proc. Natl. Acad. Sci. {L S. A. 89, 1075-1079

7 Hellmann-Blumberg, U., McCarthy Hinz, M F., Gatewc×nt, J.M. and Schmid, C.W. (1993) Mol. Cell. Biol. 13, 4523--4530

8 Selker. E.U. (1990) Trends Biocbem. Sci. 15, 103-t07

9 Poznanski, A.A. and Calarco. P.G. (1991) Dev. Biol. 143, 271-281

10 Dupres.soir. A. and Heidmann. T.

(1996) MoL Cell. BioL 16, 4495--4505

I1 Norton, J.D. and Hogan, B.L (1988) Dev. Biol. 125, 226--228

12 Branciforte, D. and Martin, S.L. (1994) Moi. Cell. Biol. 14, 2584-2592.

13 Singer, M.F. ,'t ai. { 1993) Gene 135, 183--188

14 K~x:hanek, S., Renz, D. and Doerfler, W. (1993) EMBOJ. 12, 1141-1151

15 Sanfotd, j.P., Clark, H.J., Chapman, V.M. and Rossant, J. (1987) Genes Dev. 1, 1039-1046

16 Monk. M., Ikmbelik, M. and Lehnert, S. (1987) Development 99, 371-382

17 JaenLsch, R. (1997) Trends Genet. 13, 323-329

18 Beard, C., Li, E. and Jaeni~'h, R. (1995) Genes Dev. 9, 232%2334

19 Meehan, R.R. etal.(1989) Ce1158, 499-5O7

20 Meehan, R.R., Lewis. J.D. and Bird, A.P. (1992) Nucleic Acids Res. 20, 508%5092

21 Boyes, J. and Bird, A. q1991) Cet164, 1123--1134

22 Bird, A.P. (1987) Trend~ Genet. 3, 342-347

23 Bestor, T.H. (1990) Philos. Tra~. R. Sr~.'. London Ser. B326, 179--187

24 Bird, A.P. (t995) Trends Genet. 11, 94-100

AOmaN BmD a.bird@ed.ac.uk

hzstitute of Cell and Molecular B~Iqw l 'nit~rsitr of Edinburgh, FMinbu~b, I:K EH9 3JR.

Reply We found that the large majority of 5-methylcytosine in the mammalian genome is in transposable elements, and proposed that the primary function of 5-methylcytosine is the transcriptional repression of parasitic sequence elements, which represent more than a third of human DNA and whose potential activity poses a constant threat to the integrity of the genome n. The acthvar.ion of transposon promoters can direc*dy interfere with the regulated expression of local genes and can lead to insertion mutations

at new sites. Bird uses indirect arguments to challenge the host-defense theory and to support

: : " : : . w e p r e d i c t : t h a : t : :: :: -demethylat ion wil l greatly

• increase levels of t ransposon ~:transcripts in germ cells and : ear ly embrYOS ~ "

his 'transcriptional noise' idea 2, but none of his objections compromise our conclusions.

TIG DECEMBER 1997 VOL. 13 NO. 12

Copyright 0 1997 gl~r Sclence ild. All figh~.,~ ~e:verred. 01 @I-9525/97/$17.C4] 470 Pll~ S01 f~l-9525(97JO'l~l i-5

Reversible DNA methylation was long ago proposed to be involved in the regulation of development3:. In the intervening decades, remarkably little direct evidence for this mechanism has appeared, while the evidence for an essential role of allele-specific promoter methylation in genomic imprinting and X inactivation has become very strong 5.6. If promoter methylation does regulate gene expression during cellular differentiation it must represent a vertebrate-specific adaptatf~ofi, because most metazoan ,apecies do not have 5-methylcytosine in their DNA (Ref. 7). Gene control during development tends to rely