By Monica Heger
Researchers at Harvard University have optimized a method of targeted bisulfite sequencing on Illumina's Genome Analyzer to analyze DNA methylation in clinical tissue samples.
Previously, they used the method — called reduced representation bisulfite sequencing — on mouse embryonic stem cell lines (see In Sequence 11/20/2007), but in a study published in this week's Nature Methods, they demonstrated their technique using only 30 nanograms of DNA from human formalin-fixed tissue samples from a patient with colon cancer.
By comparison, other DNA methylation analysis methods, such as methylated DNA immunoprecipitation profiles generated using next-generation sequencing (MeDIP-seq), methyl-binding domain–sequencing (MBD-seq), and comprehensive high-throughput arrays for relative methylation (CHARM) require several micrograms of genomic DNA — an amount that is "infeasible for many clinical samples such as tumors obtained by laser capture microdissection or rare stem cell populations," the authors noted in the paper.
"There hasn't been a study looking at DNA methylation in these formalin-fixed samples, and that's important because there are hundreds of millions of these banked samples," said Alexander Meissner, assistant professor of stem cell and regenerative biology at Harvard University and senior author of the paper.
This technique shows that researchers can go back and analyze those samples as well as study new samples, including those from disease cohorts or case studies, he added. And, because the required sample size is so small, "you could run any sample that someone could isolate from a patient," Meissner said.
Meissner and his colleagues used restriction enzymes to select for fragments between 40 and 220 base pairs, fragment sizes known to contain cytosine phosphate guanine (CpG) islands and promoter regions — areas where DNA methylation occurs. They sequenced treated and untreated samples with Illumina's Genome Analyzer at about a 10-fold coverage to detect the methylated regions.
They tested the technique on freshly frozen samples as well as formalin-fixed samples that had been stored since 2001. In order to avoid degraded DNA from the FFPE samples, they first selected DNA fragments with more than 500 base pairs and then used the restriction enzymes to cut those fragments into their target sizes.
"It's a very practical method for clinical applications," said Huidong Shi, associate professor of molecular oncology at the Medical College of Georgia, who is developing methods for mapping the cancer methylome under the NIH Epigenomics Roadmap program. "It will probably be better than many of the microarray methods using the same enzymes," he added. "You have a lot more CpG sites. It's much better resolution and you get more information" compared to microarray analysis.
He said his one concern is with the complexity of the data analysis. The percentage of accepted reads varied greatly, from 16 percent to 53 percent, he said. "Whether that is because of the mapping program or because of sequencing error isn't clear," Shi said, adding that the variable mapping reads could potentially make the results difficult to reproduce.
Despite this, Shi said the method is straightforward and could prove to be extremely useful for studying specific regions of the genome and analyzing clinical samples.
The method is still slightly more expensive than microarray analysis, said Meissner, but is considerably less expensive than sequencing entire methylomes, which he said costs around $100,000. Microarray analysis costs between $350 and $600 and Meissner's approach costs between $700 and $1,200 per lane on Illumina's platform, but he estimates that within six months it will cost half that amount.
Meissner said that he is now collaborating with the Memorial Sloan Kettering Cancer Institute to study leukemia and plans to engage other ongoing cancer projects. "In addition to cancer, it will be useful for studying basic developments and for looking at epigenetics during differentiation and development," he added.