Following up on a whole-genome sequencing strategy for detecting copy number alterations indicative of cancer in patients' blood, Dennis Lo's team at the Chinese University of Hong Kong has shown that whole-genome bisulfite sequencing of plasma can detect hypomethylation, which can be used as a biomarker for circulating tumor DNA.
In a proof-of-principle study published this week in the Proceedings of the National Academy of Sciences, the team found that whole-genome bisulfite sequencing of plasma DNA could detect hypomethylation with a sensitivity of 64 percent and specificity of 94 percent. Additionally, the researchers found that when they increased sequencing depth, they could concurrently detect copy number alterations. By designing algorithms that either required just one of the markers or both of the markers, they could either make the test more sensitive or more specific.
According to Lo, while further validation of the method is required, it could eventually have applications for detecting cancer early in individuals at high risk of the disease or for monitoring cancer patients' response to treatment.
Lo told Clinical Sequencing News that his team is interested in commercializing the technique over the next three to five years, and as such will "carry out further technological development and validation."
Last year, the group published a proof-of-principle study in Clinical Chemistry on a shotgun sequencing method to detect cancer-associated copy number alterations and single point mutations from a blood sample (CSN 10/17/2012).
The recent PNAS study uses a similar method, except with whole-genome bisulfite sequencing to detect methylation. Lo and colleagues first bisulfite-converted and sequenced plasma DNA from 26 hepatocellular carcinoma patients and 32 healthy subjects, obtaining around 93 million aligned reads per case on one lane of the Illumina HiSeq 2000.
Next, they used a binning strategy similar to their previously published technique in Clinical Chemistry to first establish a baseline variation of methylation in the healthy subjects. They established hypomethylation to be three standard deviations or more below the mean.
Lo told CSN that they chose to use hypomethylation as a marker for cancer because they wanted to use a "molecular change that is prevalent across the genome."
Additionally, the team discovered that hypomethylation could be detected at a lower sequencing depth than copy number alterations. With negligible impacts to specificity and sensitivity, the team could reduce sequencing depth from 93 million aligned reads per sample to 10 million aligned reads per sample. This would allow the assay to be performed for around $1,000, Lo said. At that sequencing depth, sensitivity of the assay was 68 percent, while specificity was 94 percent.
In order to try and boost specificity and sensitivity, the team combined hypomethylation analysis with copy number alteration analysis.
They designed two different algorithms — one of which required both hypomethylation and copy number alteration to be present for a positive call, and another that required either of the two markers to be present.
Requiring both hypomethylation and copy number alteration resulted in sensitivity and specificity of 94 percent and 88 percent respectively, while requiring just one or the other led to 68 percent sensitivity and 94 percent specificity.
Going forward, Lo said that the two algorithms could have different functions.
For instance, in individuals at a high risk of cancer, such as those with cancer-predisposing mutations or patients who have been treated for a previous cancer, greater sensitivity may be desired in order to reduce the number of false negatives, while in a patient at low risk, the algorithm that requires both hypomethylation and copy number alteration to call a sample positive will help reduce the number of false positives.
Lo said that while the sensitivity of the test may appear low, "one has to bear in mind that prior to the advent of genome-wide plasma DNA sequencing, there was no single molecular cancer test that [could] be used to screen for any cancer."
Additionally, the test can be "used for virtually all cancer types," he added. For instance, he said the technology's initial use would likely be in individuals at risk for certain cancer types — "hepatitis B carriers for hepatocellular carcinoma; smokers for lung cancer; and individuals with cancer-predisposing mutations, [for example], those for colon cancer and breast cancer."
The technique could potentially be feasible for monitoring residual disease in patients after treatment. In the study, the team evaluated two of the HCC patients with the technique at multiple time points after their tumors were surgically removed.
In one of the patients, the sequencing method found that both hypomethylation and copy number alteration were lower three days post-surgery. Two months after surgery, while hypomethylation and copy number alterations were still lower than they had been prior to surgery, they were elevated above the established normal baseline. The patient was later found to have multiple tumor deposits in the portion of the liver not removed and died of metastatic disease eight months after surgery.
In the second patient, hypomethylation and copy number alteration were both undetectable three months after surgery and remained undetectable one year later. The patient remained in remission 20 months post-surgery.
Muhammed Murtaza at Cancer Research UK, who has helped develop a targeted sequencing method called TAm-seq to detect mutations in circulating tumor DNA, (CSN 6/6/2012), said that the study is promising and "demonstrates how to use genome-wide bisulfite sequencing for early cancer detection and is perhaps an approach for monitoring minimal residual disease."
However, he said that before it's used in clinical applications, the method would need to be validated in larger cohorts and the individuals would need to be monitored over a longer period of time to demonstrate the clinical validity of using hypomethylation as a biomarker.
Additionally, Lo said that the use of single-molecule sequencing technologies such as Pacific Biosciences' RS would increase the performance of the technique since PacBio sequencing is able to detect methylation without bisulfite treatment or amplification, both of which can introduce errors.
Additionally, added Murtaza, not having to do bisulfite treatment prior to sequencing DNA would also enable less starting material.
Going forward, Lo said that he is planning to do further studies in larger cohorts to validate the method and he is also looking at testing multiplex strategies in order to reduce the cost.