The major observation reported here is that ICF cells, despite being hypomethylated at gene- and L1-associated CpG islands on the inactive X chromosome, exhibit the same histone modification patterns as inactive Xs in normal cells. In addition, we show that cells mutant for MeCP2, a methyl DNA binding protein, also exhibit normal histone modification patterns on the inactive X. These results imply that DNA methylation and/or this methyl DNA binding protein are not critical for determining histone modification patterns on the inactive X chromosomes.
Two major questions can be raised about our results: (1) is the sensitivity of the cytological histone modification assay too low to detect active-type histone modifications on the ICF inactive X? and (2) is the extent of methylation on the ICF inactive X greater than is suggested by CpG island and LINE-1 methylation patterns?
The cytological results imply that most of the genes on the inactive X in ICF cells are subject to inactivation, a conclusion supported by our allele-specific expression analyses of individual genes, such as AR, in ICF cells ( and data not shown). For genes subject to X inactivation in ICF cells, we expect histone modifications at the gene level to be similar to those detected cytologically at the chromosome level, and this is what we have shown here for the AR gene. For genes that escape X inactivation in ICF cells, we expect their histone modification patterns to be those of expressed genes, and one of us (RSH) has previously reported this to be the case for the SYBL1 gene ( and data not shown). We did not detect these active patterns cytologically, suggesting that there are no large blocks of genes escaping inactivation in ICF cells except at the Xp pseudoautosomal region, which normally contains escaping genes (Fig. 2).
Methylation levels at inactive X-linked CpG islands in ICF cells are decreased by an average of 89% from normal as determined by bisulfite analyses at the G6PD, FMR1, and SYBL1 loci, and many of the cloned alleles analyzed were completely unmethylated like active X alleles . It is possible that DNA methylation at other CpG-rich regions (e.g., Alu and LINE-1 elements) on the X chromosome might be differentially methylated and play a role in the X chromosome histone modification pathway. One of us (RSH) has recently shown that LINE-1 elements are hypermethylated on both active and inactive X chromosomes in normal cells but, surprisingly, they are hypomethylated on the inactive X and hypermethylated on the active X in ICF cells . These results argue against a role for LINE-1 methylation in histone modification on the inactive X chromosome. A more complete DNA methylation analysis of the ICF and normal inactive Xs needs to be done, however, because other widespread sequences may be hypermethylated on the ICF inactive X that could direct histone modifications.
Because we know that promoter methylation is important in gene expression, it seems reasonable that if DNA methylation were directly involved in the histone modification pathway, CpG island methylation would play a critical role. Further support for this idea comes from the fact that histone modifications distinguishing active and inactive X-linked genes are concentrated in promoter regions [28, 32]. In fact, Rougeulle et al.  propose that the promoter-restricted histone modification seen at X-linked loci may be unique to monoallelically-expressed genes and provide them with an epigenetic mark.
That DNA methylation is not critical to the developmental appearance of histone modifications is further supported by recent murine studies showing that differential histone modification of the Xs in early development precede differential developmental appearance of DNA methylation [33, 34]. The fact that DNA methylation does not appear to be critical to the development of histone modifications in X-linked gene expression should not be confused with a more important role for DNA methylation in maintaining repression of X-linked genes. Some years ago we showed that the earliest events in reactivating inactive X-linked genes were hemidemethylation followed by chromatin hypersensitivity, and then transcription factor binding and transcription [35, 36]. More extensive studies have recently pointed to a similar conclusion [11, 37, 38]. Thus, DNA methylation appears to play a dominant role in maintaining repression, even though it is a late event in establishing silent chromatin.
We can also consider the implication of this work for the proposed role of methyl-CpG binding proteins in the histone modification pathway. Our ICF cell data and the results from murine developmental studies, showing that histone modification of X-linked genes precedes DNA methylation, argue against such a role for the X chromosome. A role for methyl-DNA binding proteins in the histone modification pathway is supported by studies with Rett syndrome cells where a methyl-DNA binding protein, MeCP2, is mutated. In both humans and mice with Rett syndrome mutations, general hyperacetylation of histones was reported, albeit at different sites. In human cell lines, H4 was hyperacetylated preferentially at K16 , while in mouse mutant tissues hyperacetylation was reported specifically at H3K9 [10, 13]. In our work, however, we saw no major effect of two different MECP2 mutations on inactive X histone modification. The recent discovery that LINE-1 elements on the inactive X are methylated by a methyltransferase distinct from the one that carries out the same modification on the active X raises the possibility that the inactive X could have its own modification rules . We must consider, therefore, the possibility that the inactive X chromosome does not utilize methyl-DNA binding proteins in the histone modification pathway. Such a possibility would fit with the failure to detect protein footprints at promoters on the inactive X chromosome, whereas they are readily detectable on the active X [36, 39–42]. It should be noted that only a small fraction of possible histone modifications have been elucidated at this time, and it is possible that histone modification on the inactive X that depends on methyl-DNA binding protein(s) will be found in the future.
Finally, we would like to comment on the implication of this study regarding the inactive X silencing complex. Systems controlling gene expression tend to be multilayered and the X inactivation system is no exception. We know that silencing on the inactive X involves XIST RNA, DNA methylation, histone modification patterns, chromatin sensitivity, and delayed replication. It is our opinion that these factors tend to act in a more or less independent manner, as we have suggested several times in the past [23, 43–46]. For example, promoter demethylation of inactive X-linked genes, as occurs in ICF cases, does not necessarily lead to reactivation; markedly advanced replication time must also be present for reactivation to take place . The present study would appear to add further support to this idea.