Most of the 5-methylcytosine in mammalian DNA resides in transposons, which are specialized intragenomic parasites that represent at least 35% of the genome. Transposon promoters are inactive when methylated and, over time, C-->T transition mutations at methylated sites destroy many transposons. Apart from that subset of genes subject to X inactivation and genomic imprinting, no cellular gene in a non-expressing tissue has been proven to be methylated in a pattern that prevents transcription. It has become increasingly difficult to hold that reversible promoter methylation is commonly involved in developmental gene control; instead, suppression of parasitic sequence elements appears to be the primary function of cytosine methylation, with crucial secondary roles in allele-specific gene expression as seen in X inactivation and genomic imprinting.

The sequence and the structure of DNA methyltransferase-2 (Dnmt2) bear close affinities to authentic DNA cytosine methyltransferases. A combined genetic and biochemical approach revealed that human DNMT2 did not methylate DNA but instead methylated a small RNA; mass spectrometry showed that this RNA is aspartic acid transfer RNA (tRNA ^Asp ) and that DNMT2 specifically methylated cytosine 38 in the anticodon loop. The function of DNMT2 is highly conserved, and human DNMT2 protein restored methylation in vitro to tRNA ^Asp from Dnmt2-deficient strains of mouse, Arabidopsis thaliana, and Drosophila melanogaster in a manner that was dependent on preexisting patterns of modified nucleosides. Indirect sequence recognition is also a feature of eukaryotic DNA methyltransferases, which may have arisen from a Dnmt2-like RNA methyltransferase.

Ingo Braasch, John Postlethwait and colleagues report the genome of the spotted gar (Lepisosteus oculatus), whose lineage diverged from teleosts before genome duplication. Their data provide insights into the evolution of genes involved in immunity, mineralization and development and facilitate the comparison of cis-regulatory elements between teleosts and humans. To connect human biology to fish biomedical models, we sequenced the genome of spotted gar (Lepisosteus oculatus), whose lineage diverged from teleosts before teleost genome duplication (TGD). The slowly evolving gar genome has conserved in content and size many entire chromosomes from bony vertebrate ancestors. Gar bridges teleosts to tetrapods by illuminating the evolution of immunity, mineralization and development (mediated, for example, by Hox, ParaHox and microRNA genes). Numerous conserved noncoding elements (CNEs; often cis regulatory) undetectable in direct human-teleost comparisons become apparent using gar: functional studies uncovered conserved roles for such cryptic CNEs, facilitating annotation of sequences identified in human genome-wide association studies. Transcriptomic analyses showed that the sums of expression domains and expression levels for duplicated teleost genes often approximate the patterns and levels of expression for gar genes, consistent with subfunctionalization. The gar genome provides a resource for understanding evolution after genome duplication, the origin of vertebrate genomes and the function of human regulatory sequences.

The mechanisms that establish and maintain methylation patterns in the mammalian genome are very poorly understood, even though perturbations of methylation patterns lead to a loss of genomic imprinting, ectopic X chromosome inactivation, and death of mammalian embryos. A family of sequence-specific DNA methyltransferases has been proposed to be responsible for the wave of de novo methylation that occurs in the early embryo, although no such enzyme has been identified. A universal mechanism-based probe for DNA (cytosine-5)-methyltransferases was used to screen tissues and cell types known to be active in de novo methylation for new species of DNA methyltransferase. All identifiable de novo methyltransferase activity was found to reside in Dnmt1. As this enzyme is the predominant de novo methyltransferase at all developmental stages inspected, it does not fit the definition of maintenance methyltransferase or hemimethylase. Recent genetic data indicate that de novo methylation of retroviral DNA in embryonic stem cells is likely to involve one or more additional DNA methyltransferases. Such enzymes were not detected and are either present in very small amounts or are very different from Dnmt1. A new method was developed and used to determine the sequence specificity of intact Dnmt1 in whole-cell lysates. Specificity was found to be confined to the sequence 5'-CpG-3'; there was little dependence on sequence context or density of CpG dinucleotides. These data suggest that any sequence-specific de novo methylation mediated by Dnmt1 is either under the control of regulatory factors that interact with Dnmt1, or is cued by alternative secondary structures in DNA.

Trace levels of 5-methylcytosine persist in the DNA of mouse embryonic stem cells that are homozygous for null mutations in Dnmt1 , the gene for the one previously recognized metazoan DNA methyltransferase. This residual 5-methylcytosine may be the product of a candidate second DNA methyltransferase, Dnmt2, that has now been identified in human and mouse. Dnmt2 contains all the sequence motifs diagnostic of DNA (cytosine-5)-methyltransferases but appears to lack the large N-terminal regulatory domain common to other eukaryotic methyltransferases. Dnmt2 is more similar to a putative DNA methyltransferase of the fission yeast Schizosaccharomyces pombe than to Dnmt1. Dnmt2 produces multiple mRNA species that are present at low levels in all tissues of human and mouse and is not restricted to those cell types known to be active in de novo methylation. The human DNMT2 gene was mapped to chromosome 10p12-10p14 in a panel of radiation hybrids. Dnmt2 is a candidate for the activity that methylates newly integrated retroviral DNA and maintains trace levels of 5-methylcytosine in the DNA of embryonic stem cells homozygous for null mutations in Dnmt1.