During development, a small but significant number of CGIs become methylated and stably silenced [27
]. This process of developmentally programmed CGI methylation has been best characterized in genomic imprinting and X chromosome inactivation. Here, using a restriction enzyme–based MCAM approach, we found that non-CGI and sparse-CGI promoters were more susceptible to methylation than dense-CGI promoters, in agreement with a very recent report using an antibody approach to compare methylation profiles between three classes of promoters [28
]. Although most dense-CGI promoters remain free of methylation, we found a small exceptional class of such promoters (4.0%) that become methylated in normal somatic tissues and are not associated with X-chromosome or imprinted genes. By detailed characterization of a subset of such genes, we found dense promoter CGI methylation and gene silencing in most normal somatic tissues except germ-line cells. Using RT-PCR and data from published microarray experiments, we confirmed tissue-specific silencing for this class of genes. Further, we showed that inhibition of methylation reactivates expression in these genes. Our results suggest that DNA methylation plays an important role in silencing germ-cell specific genes in somatic tissues. A previous analysis of methylation using RLGS (restriction landmark genome scanning) identified 150 regions (including promoters, exons, and introns) as TDMs (tissue-specific differentially methylated regions) in C57BL/6J mice [29
]. By comparing 14 of these mouse TDMs with the human genome, six showed human homologs, and five had conserved tissue-specific methylation and expression, being preferentially expressed in testis [30
]. Our results indicate that this pattern affects a relatively large number of genes. On the other hand, other studies failed to identify many dense promoter CpG islands hypermethylated in normal tissues, suggesting that the class of genes we describe here is unique.
It will be important to determine how methylation at these promoter-associated CGIs is established and maintained. One possibility is that methylation is instructed by local sequence features. By comparing DNA sequence flanking the center of CGIs, we identified five sequence motifs significantly enriched in methylated promoter CGIs relative to the bulk of unmethylated CGIs. Although these motifs do not match known transcription factor binding sites, all of them have significant overlap with Alu,
a family of SINEs (short interspersed elements). Alu
repeats are rich in CpG dinucleotides and are common targets for DNA methylation. About one-third of methylated CpGs in the genome are located within Alu
repeats have been proposed as methylation centers for neighboring genes [31
] and we hypothesize that Alu
-regulatory elements may play a role in establishment and/or maintenance of tissue-specific methylation. Experimental approaches such as transfection and transgenic studies will be needed to test this model. Interestingly, while Alu
repeats are almost completely methylated in most tissues, some, particularly young Alu
repeats, are almost completely unmethylated in germ line cells [32
], similar to the genes described here. It remains unclear why tissue-specific gene silencing by DNA methylation is relatively rare, since many genes showing tissue-specific expression are not methylated. What is most exceptional about the genes we describe here is their restricted expression in germ cells. Considering that testis is an immune-privileged site, it is possible that some of these genes, if expressed in somatic tissues, could trigger autoimmune phenomena, justifying the need for a strong mechanism to maintain silencing. In this respect, it is also interesting that we observed hypomethylation of these genes in several cancers. Although the causes and biological effects of cancer-linked hypomethylation remain unclear, such hypomethylation can lead to gene expression that induces an immune response [34
]. The patterns of methylation and gene expression we observed for this class of genes suggest that some may well be cancer-testis antigens.
In summary, we have identified a group of non-X–linked bona fide promoter CpG islands that are densely methylated in normal somatic tissues, escape methylation in germ line cells, and for which DNA methylation is a primary mechanism of tissue-specific gene silencing.