By Monica Heger
Researchers from Washington University School of Medicine have developed a sample-prep technique they call bisulfite patch PCR that enables multiplexed sequencing of promoter methylation. In a study published last week in Genome Research, they demonstrated the technique on colorectal and breast cancer samples, analyzing 94 genes from 48 different samples in one run of Roche's 454 GS FLX.
Rob Mitra, senior author of the study and assistant professor of genetics at Wash U, said that the team plans to use the technique to study methylation in colorectal, breast, and endometrial cancers. Short term, he said the method can be used to validate findings from whole-genome sequencing studies of a small number of samples. The long-term goal, he said, is to use it to determine how methylation differs among cell types in mixed tissue samples.
To detect methylation, the standard procedure is to first bisulfite-convert the genome, which changes cytosines to thymines, but leaves methylated cytosines intact. In this study, the researchers first used an oligo hybridization- and ligation-based method to select 94 targeted gene promoters — 42 genes associated with colon cancer, 44 associated with breast cancer, four associated with both, and four controls. The targeted regions ranged in length from 125 base pairs to 581 base pairs, totaling 25.4 kilobase pairs. After selecting the regions, they then bisulfite-converted the DNA, amplified it, and attached a sample-specific barcode.
Other approaches exist for examining methylation across the whole-genome or a very large set of loci. But these methods are cost-prohibitive to do on a large number of samples. Additionally, there are also methods to amplify a few loci individually, barcode, and then bisulfite sequence, but these are labor and time intensive, and also require large amounts of patient DNA. This method was designed to fill the gap, targeting several hundred loci, so that it could still be performed on many different samples.
They performed the method using 250 nanograms of starting DNA from 12 tumor/normal matched pairs from colon tissue, and 12 tumor/normal matched pairs from breast tissue. The resulting amplicons were then pooled and sequenced on one lane of the 454.
They obtained 97,115 reads, 90 percent of which mapped to one of the targeted promoters, and the average read length was 228 base pairs.
All of their 94 targets were successfully amplified, though there was a range in the number of reads obtained for each region, with longer amplicons being less abundant. The authors concluded that the length bias was introduced by the bisulfite treatment rather than the multiplexed PCR.
The researchers found that about half of the cancer-related genes were unmethylated, one-third were methylated in both cancer and normal tissue, and the remaining nine showed tumor-specific methylation.
Five promoters were hypermethylated in 25 percent to 75 percent of both breast and colon cancer tumors. Two of the promoters had been identified as sites of hypermethylation in other studies, while three were novel findings. The finding suggests that the tumors share similar molecular defects, and could also suggest that methylation leads to an inactivation of the gene, and is a key step in tumorigenesis.
"Methylation is correlated with reduced gene expression, but we haven't shown yet that they are inactivated," Mitra said. "The findings suggests that this is another way that tumor suppressor genes can get inactivated, and it's consistent with what other groups have shown as well," although they will have to go back and validate that those genes are actually inactivated.
While the results are preliminary, the authors reported that tumor-specific methylation can be a useful biomarker, and the frequency of methylation found at the targets "approaches the significance of even the most common genetic mutations," they reported.
Chia Lin Wei, senior group leader at the Genome Institute of Singapore, who has used sequencing to study epigenetics in cancer, said the approach was unique, and could have applications on clinical samples. "The multiplex nature of the PCR step definitely allows them to screen multiple samples either for the clinic or for research," she said.
Huidong Shi, a molecular oncologist at the Medical College of Georgia, who is working on mapping the cancer methylome with next-generation sequencing, said that the method would be useful, and he thought it would be straightforward to scale up. "There are more and more [cancer-associated] genes being discovered by microarrays and whole-genome sequencing, and this should be a very straightforward way to validate them in tumor samples," he said.
Mitra said the team is now working on making several improvements to the method. First he said they would like to reduce the required amount of starting DNA to under one nanogram, which he said could be done by doing the bisulfite treatment more gently, and increasing the amount of amplification. He also wants to increase the number of loci that can be targeted at once and increase the flexibility of the method. The current method relies on restriction enzymes to target the loci, which limits the loci that can be targeted and also the length of the resulting amplicons. Being able to make all amplicons the same length would allow them to be amplified uniformly. He said the team is looking at a number of different techniques, including using a combination of different restriction enzymes, and also at using a PCR amplification step to define fragments.
Mitra also said that being able to create shorter amplicons would make the method more amenable to other sequencing platforms. Currently, the amplicons are all over 100 base pairs, with some around 500 base pairs, so are more suited to 454's longer read lengths. Wei also agreed that the method as is would likely be limited to use on the 454 because of the long amplicon lengths.
Aside from optimizing the protocol the team is also collaborating with another group at Wash U, led by Paul Goodfellow, to use the protocol on endometrial cancer samples. "Methylation is known to play a role in endometrial cancer," said Mitra.
He said Goodfellow's team has around three hundred banked samples from endometrial cancer patients. First, they plan to use a broad approach, like targeted sequencing, on a small number of samples to identify genes of interest. Then they will use the bisulfite patch PCR on all the samples to validate the findings, said Mitra.
"I think the method could be broadly adopted and that it could be a good complement to other techniques that are out there, like targeted sequencing or whole-genome sequencing" said Mitra.