Cytosine methylation and mammalian development

Abstract

Programmed methylation and demethylation of regulatory sequences has been proposed to play a central role in vertebrate development. We report here that the methylation status of the 5' regions of a panel of tissue-specific genes could not be correlated with expression in tissues of fetal and newborn mice. Genes reported to be regulated by reversible methylation were not expressed ectopically or precociously in Dnmt1-deficient mouse embryos under conditions where demethylation caused biallelic expression of imprinted genes and activated transcription of endogenous retroviruses of the IAP class. These and other data suggest that the numerous published expression-methylation correlations may have described not a cause but a consequence of transcriptional activation. A model is proposed under which cytosine methylation represents a biochemical specialization of large genomes that participates in specialized biological functions such as allele-specific gene expression and the heritable transcriptional silencing of parasitic sequence elements, whereas cellular differentiation is controlled by conserved regulatory networks that do not depend on covalent modification of the genome.

Figure 1

Methylation analysis of the 5′ regions of tissue-specific genes. Diamonds above the heavy line indicate HpaII sites for all but E, in which HhaI sites (5′-GCGC-3′) are denoted. The amount of fill in the diamond symbol represents average methylation level in nonexpressing tissues. Tick marks below the heavy line indicate the positions of each CpG dinucleotide. Probes and predicted fragment sizes are shown at the bottom of each panel, and transcription start sites are indicated by arrows. Lanes C contained DNA not cleaved with methylation-sensitive enzymes; lanes M contained DNA digested with MspI (absent in E, which contained HhaI digests). DNA of sperm is completely methylated at tested sites in A, B, G, and H and completely unmethylated in C, D, E, and F. Except for Lep in D, none of the latter genes show detectable methylation in somatic tissues. Prf1 in I is completely unmethylated in Prf1-expressing cytotoxic T lymphocytes and largely unmethylated in lung, liver, and whole E14.5 embryos, none of which express appreciable levels of Prf1. Methylation levels of non-CpG island genes are generally higher in brain and whole E11.5 embryos than in E14.5 embryos and adult organs. (E11.5 and E14.5) Whole mouse embryos at 11.5 and 14.5 days postcoitum; (Ki) kidney; (Br) brain; (Liv) liver; (Spl) spleen; (Mu) skeletal muscle from body wall; (Sp) sperm; (Ht) heart; (CTL) cytotoxic T cells purified by flow sorting mouse lymphocytes stained with antibody to CD8.

Figure 2

CpG densities predict methylation status of 5′ regions of tissue-specific genes in sperm DNA. The 5′ regions of the genes in Fig. 1 were analyzed for CpG density, G+C content, and methylation status. (•) Methylated sequences; (○) unmethylated sequences. Notice that sequences with higher observed/expected values for CpG densities tend to be unmethylated in sperm DNA. CpG densities (left) are a more accurate predictor of methylation status in sperm DNA than are G+C contents (right).

Figure 3

Normal expression of tissue-specific genes in DNA methyltransferase-deficient mouse embryos. (A) The genes of interest were largely unmethylated in control samples and underwent additional demethylation in Dnmt1-deficient mouse embryos at day E9.5. (B) A single RNA blot was hybridized consecutively with the eight probes shown. H19 and IAP retroviral transcripts can be seen to increase in the mutants; Igf2 declines and Dnmt1 mRNA is barely detectable. Col1a1, Acta1, and Hbb mRNAs are largely unaffected or reduced in amount in the mutant embryos, although all three genes had been reported to be repressed by cytosine methylation.