NEW YORK (GenomeWeb News) – In cancer cells, all sorts of functions go awry. In addition to genetic changes, many times there are also epigenetic changes.
DNA methylation typically regulates gene expression, but in cancer that, and other epigenetic controllers, don't act quite as they are supposed to. In many cancers, there is global hypomethylation of DNA, though there is also, paradoxically, hypermethylation at some CpG island promoters, especially in the promoters of tumor suppressor genes.
"Overall, tumors have lost DNA methylation," explained Manel Esteller, a group leader at the Bellvitge Biomedical Research Institute in Barcelona, Spain. "This loss occurs mainly in repetitive sequences," he said, but there are subsets of the tumor genome that become methylated. "These are very important mechanisms to inactive tumor suppressor genes like p16, BRCA1, MGMT, et cetera."
New tools and approaches are now allowing researchers to more fully characterize and explore the epigenetic changes that occur in cancer cells that help those cells develop and spread, and large-scale efforts are compiling such changes. With this data, researchers are beginning to tease apart the pathways that are influenced by epigenetic changes in a number of cancers. This could lead to epigenetic biomarkers of both cancer development and progression as well as identify possible targets for cancer therapeutics.
New Tools of Interrogation
A number of new techniques have been developed to help researchers interrogate the epigenome and deconstruct its role in the development and progression of cancer. Bisulfite sequencings still reigns as the gold standard for uncovering cytosine methylation, though a number of other approaches like ChIP-seq, FAIRE-seq, MethylC-seq and even newer tools like NOME-seq are turning to next-generation sequencing to enable the study of the methylome.
Ben Berman, an assistant professor at the University of Southern California, noted that The Cancer Genome Atlas is sequencing tumors using whole-genome bisulfite sequencing. Employing this method researchers were able to confirm that cancer genomes contain patterned regions of hypomethylation and hypermethyaltion, something Berman said has been suspected since the 1980s.
"But it wasn't really until we were able to have this very high-resolution sequencing assay that we were able to [see] that there are very highly patterned regions that have that hypomethylation, and [we're] understanding now that they correspond to nuclear territories and coincide with the timing of replication, with mitosis," he said.
Such findings, he said, show the power of these new techniques.
Approaches like ChIP-seq, too, are allowing researchers to get a global view of the methylome. Mario Suva, a research fellow at Harvard Medical School in Bradley Bernstein's lab, is studying cellular heterogeneity in glioblastoma and how it correlates to epigenetic changes. His work relies mainly on ChIP-seq.
"It's an excessively powerful tool," he said. "It's really used as a fingerprint for cellular identity, so before these large-scale epigenomic tools were available, cell types were defined on functional properties, but now that we can really interrogate on a genome-wide scale all the regulatory elements present in a given cell, it can really be used as a fingerprint for cellular identity."
Using ChIP-seq, Suva and his colleagues have been able to identify roles for transcription factors in subpopulations of glioblastomas.
Berman and his colleagues, though, have also developed a new approach called NOME-seq that combines a GpC methyltransferase and next-generation sequencing to produce high-resolution nucleosome positioning and endogenous methylation information, as they reported in Genome Research in December.
"The only technique that [had] existed to look at nucleosome remodeling and nucleosome positioning patterns which occur at all these regulatory sites was MNase sequencing," he said. But that approach requires deep — and therefore expensive — sequencing coverage, and because of its specific cuts sites it can be difficult to tell composition effects and true nucleosome occupancy apart.
NOME-seq, he said, lets researchers get both nucleosome positioning information and methylation data from low-coverage reads. "We can read a very accurate picture of these nucleosomal footprints and on the same molecule, on the same sequence read, we can also read off the DNA methylation information," he said. He added that the three states — inactive, active, and poised — could be distinguished.
Further, he said that allelic affects may also be uncovered and could be used to identify subpopulations of cancers cells.
Marking the Spot
The epigenomic patterns found in tumors may reflect not only where the tumor originated from, but also disease prognosis. Primary and metastatic tumors differ, and some of those differences may reflect changes to the cancer cells that allowed them to seek out a new environment, yet similarities to their origins remain.
"We know that the epigenome can change in a very plastic way especially when epigenetic changes are involved in adaptation to microenvironment and the generation of metastasis," Esteller said.
He and his team examined the DNA methylation profiles of 1,628 human samples, including 429 samples from normal tissues, 1,054 tumor samples, and 150 samples from people with other, non-cancer diseases like Alzheimer's disease or an autoimmune disorder.
After examining the samples using the Illumina GoldenGate assay and bisulfite sequencing of 1,505 CpG sites, the researchers found that the disease or normal tissue samples had differential methylation patterns — with the cancer samples exhibiting the most profound changes. Those samples, the researchers reported in Genome Research last year, gained promoter CpG methylation and lost it in non-CpG promoters.
Further, they were able to use the patterns they saw in different tissue types to trace back the primary sites of some cancers of unknown primary origin. For 42 CUP tumors they analyzed, they were able to assign 29 of them to a site of origin.
Epigenetic changes, too, permit cancers to metastasize. For example, Esteller's team reported last year in the Journal of Pathology that the cadherin-11 promoter CpG island undergoes hypermethylation in cell lines derived from primary and metastatic samples collected from the same patient. This change in methylation, the team reports, led to the silencing of cadherin-11 expression, which they suspect enables the cells to loosen from their contacts with other cells and spread.
Potentially, such changes could be used as biomarkers that gauge prognosis and, possibly, drug response. Esteller noted that his team is working to develop such predictors.
Having complete epigenomes from normal, tumor, and other disease tissues will be a boon, he said, for such efforts. "This is going to be important and a lot of work for many people because after the 2000 sequencing of the human genome, many questions were opened and some of them can be solved thanks to these epigenome projects," he said.
And such large-scale efforts are underway to map the epigenetic profiles found in cancer cells as well as in normal tissue.
USC's Berman is part of the epigenetic characterization effort with TCGA, and he said they've sussed out patterns of methylation changes in samples they've analyzed so far.
"We've generated a DNA methylation profile on now thousands of tumors, and we've been able to find patterns using that in ways that wouldn't have been possible a few years ago," he said. "We're now sequencing tumors using whole-genome bisulfite sequencing and with the increased resolution of that approach, we were able to identify these kinds of global patterns of hypomethylation."
For example, in a Nature Genetics paper published last fall, TCGA researchers combined both genetic and epigenetic analyses of breast cancer and found that breast tumors largely fell into four different classes based on their genetic and epigenetic changes.
While groups like TCGA are making progress examining epigenetic changes in cancer cells, Berman said there are still challenges to determining methylation patterns of tumors. The most notable one being tumor purity. Tumors, he noted, contain a hodgepodge of cells, including stem-like cancer cells, more differentiated cells, and more. However, he noted that researchers like Scott Carter at the Broad Institute are developing methods to help estimate the purity of tumor samples.
Despite these challenges, databases can house a lot of useful epigenetic and epigenomic information. Still, he said, such datasets are a mainly untapped resource for many researchers. "The majority of people who do those [DNA methylation] studies really neglect to use all this kind of treasure trove of information that's out there," he added.
Part of the issue is that some databases could be difficult for non-bioinformaticians to use, Berman said. He and his team are working on developing tools to enable others to take advantages of what large datasets offer. "I think it would be really worth it for those large, concerted efforts to develop tools that smaller groups can use," he said.
It also has to come back to the biology, added the University of Cambridge's Tony Kouzarides. "I think you really need to do biology in order to understand how the modifications make a difference," he said. "Just cataloging them is not enough. … You end up with correlations, and they may not be significant."
Drugging the Epigenome
Epigenetic changes that occur in cancer cells also offer intriguing possibilities for cancer drugs. Indeed, there currently are a handful of FDA-approved epigenetic drugs, mainly to treat leukemia and lymphoma. There's azacidine, which inhibits DNA methylation, and is approved to treat myelodysplastic syndromes, as well as the histone deacetylase inhibitor vorinostat that is approved to treat cutaneous T-cell lymphoma that hasn't responded well to other therapies. And even more therapies are under development.
"The idea is to change the cellular state, and even despite the fact that you still carry the genetic mutation, it will have a different or no effect anymore because of the alterations," Harvard's Suva said. "In leukemia, it has been used with success … but there are many other cancers in which that could be applied."
While many of the drugs approved to date focus on altering DNA methylation, there are a number of other epigenetic marks that could be targets. Cambridge's Kouzarides, for example, has examined inhibitors targeting BET proteins, which read histone modifications and associate with the chromatin based on what they encounter there. They also, based on that recognition, recruit a set of proteins to that site that have been implicated in a type of leukemia, which Kouzarides' group studied using a proteomic approach. Those proteins appear to recruit MLL fusions that bring about leukemia.
But by inhibiting BET proteins, the cancer could be targeted, and Kouzarides' team turned to a BET inhibitor called I-BET151, which had been developed as an anti-inflammatory agent.
"The MLL translocations were very sensitive to this drug," Kouzarides said. "Essentially, if you added it to cells from patients and cell lines derived from patients that had this translocation in leukemia, you killed the cells." His team reported these results in Nature in 2011. They further worked out that one of the genes involved in the translocation is an anti-apoptotic gene, so repressing it induces cell death.
This drug, Kouzarides added, has been taken into phase I clinical trials by GlaxoSmithKline. But they are not the only biopharma targeting epigenetic marks. "Now a number of companies are developing inhibitors against other epigenetic targets with the same view that they may be relevant for cancer, and these will be taken forward into the clinic for sure," Kouzarides said.
One drug alone, though, might not be enough as tumor cells adapt and become resistant to treatments. Instead, Kouzarides said that this BET inhibitor and other drugs would likely be most effective when used in combination with other anti-cancer therapies. But what combinations will be the most effective remains to be worked out.
In considering such combinations, Kouzarides said researchers need to better understand the pathways and the biology of the cancers being targeted, to tease out the best targets and develop new drugs to treat both the original change and any resistance that arises.
An Assist from Regenerative Medicine
Epigenetic researchers may also be able to glean insight into factors that may be important in oncogenesis by learning from regenerative medicine studies, particularly the reprogramming of cells.
A number of transcription factors and chromatin regulators mediate the transition from, say, a fibroblast cell back to a stem cell of some sort. "We really think there is an analogy between the two processes," Suva said. He added that "many of the barriers to reprogramming are actually tumor-suppressor genes." He and his colleagues recently discussed this connection in a review in Science in March.
Both fields, Suva said, can feed into each other. Work coming out of regenerative medicine can help to identify the factors that are needed for a cell to change its cellular identity, as it does during oncogenesis. "Then we can study those players in the setting of a given genetic mutation and really learn a lot [about] how epigenetic state and genetic mutation cooperate to give a cancer cell its full phenotype," he said.
For example, there is an overlap between the transcription factors that are expressed in brain tumors and those that are needed to generate neuronal stem cells from differentiated fibroblasts.
"We think that the presence of those transcription factors are actually due to the fact that cancer cells have shared properties with the normal stem cells, and it's the presence of those transcription factors that provide these cancer cells with a certain number of properties," Suva said. "The cellular identity cooperates with the genetic mutations to provide cancer cells with the full spectrum of properties."
He and his colleagues in the Bernstein lab at Harvard are applying such an approach to study heterogeneity in glioblastoma. These tumors contain cells that are more stem cell-like and others that appear to be more differentiated, and they are trying to work out the epigenetic factors that influence those different degrees of differentiation within the same tumor.
"Epigenetics definitely adds a layer of complexity," Suva said. "It gives life to the genome and I think it can contribute to our understanding of cancer in multiple ways."