|
Resolution: standard / high Figure 4.
Expression patterns in normal tissues explain differential susceptibility to hypermethylation
in cancer. (a) Consistently hypermethylated genes are more tissue specific than variably hypermethylated
genes. Shown are histograms of tissue-specificity scores (as Figure 2b) observed at
hypermethylation prone genes that were consistently or variably methylated in different
tumor types. Differences between gene sets were tested using Wilcoxon rank sum tests
(*** P < 0.001, ** P < 0.01 and * P < 0.05). (b) Variably hypermethylated genes with differential susceptibility in breast cancer
are differentially expressed in normal breast tissue. Shown are boxplots of the relative
level of expression in different cells from normal breast found at VM genes that are
either frequently or never hypermethylated in breast tumors [85]. Differences between cellular fractions were tested using Wilcoxon rank sum tests.
Lum = luminal epithelial cells, Lum Pro = luminal progenitor cells, Bas = basal myoepithelial
cells, Stroma = breast stromal cells. (c) Variably hypermethylated genes that are prone to hypermethylation in tumors are
repressed in the corresponding normal tissue. Shown are boxplots of the expression
levels measured for VM genes with different susceptibility in individual tumor types
in the corresponding normal tissues. Res = never hypermethylated in tumors, Prone
= frequently hypermethylated in tumors. Differences between gene groups were tested
using Wilcoxon rank sum tests. (d) Repressed genes are more prone to hypermethylation than active genes in colorectal
cancer. Shown are heatmaps of the methylation levels of CGI promoter genes that are
unmethylated in normal colon tissue and are either activated (left) or repressed (right)
in normal colon as compared to normal liver. The 356 repressed genes are methylated
to a significantly higher level than the 1,465 active genes (one-sided Wilcoxon rank
sum test P = 1.6x10-7). CGI, CpG island; VM, variably methylated.
Sproul et al. Genome Biology 2012 13:R84 doi:10.1186/gb-2012-13-10-r84 |