NEW YORK (GenomeWeb News) – Cancer coincides with a widespread loss of epigenetic regulation affecting large chunks of DNA in the genome, according to a study in the early, online version of Nature Genetics yesterday.
Using custom microarrays, American researchers assessed methylation patterns in five types of cancer, focusing on regions of the genome that were shown to be differentially methylated in cancer in the past. The team found that cancer genomes show dramatically different methylation patterns compared to corresponding normal tissues, including a lack of defined methylation boundaries around so-called CpG islands where cytosine and guanine nucleotides frequently neighbor one another.
In addition, they reported, the cancers showed dramatic variability in their methylation levels, along with changes in some parts of the genome that are known to be differentially methylated in other tissue types or in undifferentiated cells.
Moreover, their bisulfite sequencing studies of colon cancer, pre-cancerous polyps, and normal colon tissue uncovered large stretches of DNA that were differentially methylated in the cancer samples, leading to altered expression of some cell cycle and cell matrix-related genes in these regions.
"Epigenetics, specifically DNA methylation, is losing its regulation in cancer and we think that that's helping cancer thrive," co-first author Winston Timp, a post-doctoral fellow in Andrew Feinberg's lab at Johns Hopkins University's Center for Epigenetics, told GenomeWeb Daily News.
"It may be a very early event which acts in conjunction with mutations to cause cancer," he added. "And it seems to happen in all types of cancer."
In 2009, Feinberg, Johns Hopkins biostatistics and epigenetics researcher Rafael Irizarry, and colleagues published a Nature Genetics study describing methylation differences in colon cancers compared to normal colon tissue. In particular, they found cancer-specific differences not necessarily within CpG islands, but more often on the "shores" of these islands.
In addition, many of the same regions that were differentially methylated in the colon cancers correspond to those that tend to be differentially methylated in various tissue types or in undifferentiated cells.
Consequently, the team decided to look at several cancer types in the current study, Timp explained. "If these differences seem to control differentiation state, so to speak, maybe they'll show differences in other cancers too."
Using Illumina custom bead arrays, the team first tested 122 colon, lung, breast, and thyroid cancers and Wilms' tumors, a childhood kidney cancer. They also tested 30 pre-malignant samples, along with 141 matched control samples, focusing on 384 sites in 151 cancer-specific differentially DNA-methylated regions detected in colon cancer in the past.
Indeed, researchers reported, cancer and normal samples from each tissue type had very distinct methylation patterns. While both normal tissues and cancers tended to fall in distinct clusters based on their methylation patterns, methylation in the cancer samples was far more variable.
"This seems to show a loss of control — a loss of regulation — of methylation in cancer compared to [matched normal tissue]," Timp said. "The five different normal tissues also cluster out very well from each other … but all the cancers are much more variable."
To explore the methylation patterns across the genome in cancer cells, meanwhile, the researchers did shotgun, whole-genome bisulfite sequencing of three colorectal tumor samples — along with matched normal colon tissue and two pre-cancerous adenomatous polyp samples — using the ABI SOLiD platform.
From this sequence data, the team found altered methylation in large chunks of the cancer genomes. These blocks frequently had lower methylation levels than normal colon tissue, though, again, methylation patterns were far more variable in the cancer samples.
Consistent with these methylation changes, the expression of genes within these hypo-methylated blocks of the cancer genomes typically showed higher but more variable expression.
"We think that what's happening is a loss of [methylation] control rather than a concerted shift," Timp explained.
In addition, he noted, some of these differentially methylated regions in cancer appear to coincide with regions previously reported to be partially methylated in stem cells or other tissue types, while others overlapped with regions known to have chromatin alteration or epigenetic marks related to lamina function.
"We can say with some certainty that these areas are important for both differentiation and cancer," Timp said. "We would propose that maybe there's a link here and maybe cancer is losing regulation of these areas and reverting to a more primitive or less controlled state."
Although more research is needed to determine whether that is the case or whether there is some other explanation for the importance of the areas in both cancer and tissue development or differentiation, those involved in the study argue that their findings may ultimately lead to improved tests or treatments for cancer.
"Maybe the big lesson learned from our observation of this universal [epigenetic] chaos is that we may need to think not so much about just killing cancer cells, but also about ways of helping cancer cells figure out how to be what they're supposed to be, and re-educate them so they can stay truer to their normal identities," Feinberg said in a statement.