High-Resolution Profiling of Histone Methylations in the Human Genome

Abstract

Histone modifications are implicated in influencing gene expression. We have generated high-resolution maps for the genome-wide distribution of 20 histone lysine and arginine methylations as well as histone variant H2A.Z, RNA polymerase II, and the insulator binding protein CTCF across the human genome using the Solexa 1G sequencing technology. Typical patterns of histone methylations exhibited at promoters, insulators, enhancers, and transcribed regions are identified. The monomethylations of H3K27, H3K9, H4K20, H3K79, and H2BK5 are all linked to gene activation, whereas trimethylations of H3K27, H3K9, and H3K79 are linked to repression. H2A.Z associates with functional regulatory elements, and CTCF marks boundaries of histone methylation domains. Chromosome banding patterns are correlated with unique patterns of histone modifications. Chromosome breakpoints detected in T cell cancers frequently reside in chromatin regions associated with H3K4 methylations. Our data provide new insights into the function of histone methylation and chromatin organization in genome function.

Application of machine learning methods to histone methylation ChIP-Seq data reveals H4R3me2 globally represses gene expression

“…The inputs to our analysis are the read enrichment profiles across the human genome generated by Barski et al [14]. In Figure two of their paper, they also calculated composite plots where they stratified gene expression by quartiles, aligned the transcription start sites (TSS) of all the genes and calculated the normalized read counts as a function of position relative to the TSS.…”

“…A major problem with this method is that many marks do not have promoter/5' end biased enrichment patterns. A striking example is H3K36me3, which has increasing enrichment along the gene body, which peaks near the 3' end of genes [14]. Yu et al [28] calculate correlation coefficients between their 5' end biased mark enrichment estimates and gene expression levels and find little to no correlation between H3K36me3 and gene expression.…”

“…Yu et al [28] calculate correlation coefficients between their 5' end biased mark enrichment estimates and gene expression levels and find little to no correlation between H3K36me3 and gene expression. This is an unexpected result as H3K36me3 has been characterized as an activating mark in a number of biochemical studies [7,8], and its levels have been shown to have a strong positive correlation with gene expression [14]. This discrepancy is likely due to the 5' bias of their amplitude estimation method.…”

“…To help address these questions we applied two machine learning methods, Stepwise Multilinear Regression and Multivariate Adaptive Regression Splines (MARS) [13], to genome-wide ChIP-Seq maps of 20 histone lysine and arginine methylations and histone variant H2A.Z in CD4 + T-cells [14]. We hypothesize that inclusion of two (bivalent) and three (trivalent) interacting cross-terms in the model can reveal (1) putative cross-regulation or multivalent relationships between histone modifications and (2) a global view of the epigenetic regulatory network.…”

“…We were able to identify a number of highly significant multivalent terms, suggesting that multivalency and cross talk among histone modifications may be widespread. However, we were surprised that both models predicted H4R3me2--shown to be repressive in a number of experimental studies [15-28]--to be among the most repressive histone methylations given that its ChIP-Seq enrichment levels showed no response to increasing gene expression [14]. …”