Skip to main content
Premium Trial:

Request an Annual Quote

UW Team Combines CRISPR With Duplex Sequencing to Detect Low-Frequency Mutations


SAN FRANCISCO (GenomeWeb) – Researchers from the University of Washington have developed a targeted sequencing method that couples the CRISPR/Cas9 system for target selection and a previously described method known as duplex sequencing, which improves the accuracy of next-generation sequencing.

The researchers are now using the method, which was described recently in a publication on the BioRxiv preprint server, in their translational oncology and forensics work and startup TwinStrand Biosciences, which spun out of UW earlier this year, has licensed it.

The CRISPR-duplex sequencing method builds off the duplex sequencing technique to analyze low-frequency mutations that the UW researchers first described in 2012. Duplex sequencing involves tagging both strands of DNA with random, but complementary, double-stranded primers. The molecules are then amplified, sequenced, and grouped by their tags. Errors that are introduced from PCR or sequencing can be filtered out by looking for discrepancies among molecules with the same tag. Two rounds of filtering are performed, one for each of the complementary tags.

The team has since used the method to show that it can identify very low-frequency mutations to the cancer gene TP53 in ovarian cancer patients.

However, the problem with duplex sequencing is that it is not very efficient and requires a lot of input DNA, according to Scott Kennedy and Rosa Ana Risques, co-lead authors of the BioRxiv paper and both researchers in the Department of Pathology at UW.

Since duplex sequencing involves double-stranded barcodes, hybrid capture must be used as opposed to amplicon sequencing, and hybrid capture applied to small targets only yields 5 percent to 10 percent on-target reads. As such, larger DNA inputs are needed to ensure that the target is captured. But that can be problematic, particularly for clinical applications, and especially in oncology, where the sample is often limited, Risques said.

Looking to improve on the duplex sequencing method, the researchers turned to the CRISPR/Cas9 system, which can be used to fragment DNA at specified locations to select targets that can then be analyzed via next-generation sequencing. Other groups have already demonstrated that CRISPR/Cas9 can be used to select targets, such as short tandem repeats and disease-related genes.

What's different about the UW researchers' method is that it combines the CRISPR/Cas9 system for targeting a specific region with duplex sequencing in order to increase the accuracy of NGS and detect very low-frequency mutations.

In the BioRxiv study, the UW researchers designed guide RNAs to target the exons of the TP53 gene. "You cut out exactly what you want at a predefined size," Kennedy said. That eliminates the random fragmentation step, which can lead to lots of short DNA molecules that end up being preferentially amplified. Then SPRI beads are used to remove the uncut pieces of DNA. The DNA is then A-tailed, duplex sequencing adaptors are ligated, and the molecules are captured by biotinylated DNA probes. In the study, the researchers designed probes to target the TP53 exons, amplified them, and sequenced the targets on the Illumina MiSeq instrument.

The researchers tested DNA inputs of 25, 100, and 250 nanograms, comparing the CRISPR-DS method with standard duplex sequencing for the same sample. With 25 nanograms and one round of hybridization capture, the CRISPR-DS method resulted in 99 percent of raw reads being on target, compared to just over 5 percent for standard duplex sequencing. The latter increased to 99 percent after a second round of hybridization capture.

Next, the researchers analyzed four peritoneal fluid samples collected during surgery from women with ovarian cancer, comparing CRISPR-DS with standard duplex sequencing. Previous testing with standard duplex sequencing had identified TP53 mutations in the samples.

The researchers performed CRISPR-DS using just 100 nanograms of starting DNA, approximately 30 to 100 times less than they needed for standard duplex sequencing, and achieved comparable sequencing depths on all samples. They also found TP53 mutations in all four samples at comparable frequencies. The team then confirmed the performance on a separate set of 13 bladder tissue samples.

Gaetan Burgio, a researcher at the Australian National University who uses CRISPR/Cas9 to study infection and antibiotic resistance, said that the method was "very interesting" and seemed promising, but that he would like to see it replicated on other genes aside from just the TP53 gene, though he doesn't think there are inherent problems preventing it from being used for other genes.

Risques said that she plans to switch to the CRISPR-DS method in her research to evaluate TP53 mutations in ovarian cancer patients. Previously, she had been using around 1 microgram of starting DNA to look for low-frequency mutations in that gene, but now, she can use just 100 nanograms of DNA without sacrificing accuracy.

Identifying low-frequency mutations in TP53 could potentially help detect ovarian cancer earlier, Risques said. In her previous work, she identified TP53 mutations in the peritoneal fluid of both healthy patients as well as those with ovarian cancer. The mutations were present higher frequencies in those with ovarian cancer and she is now working to figure out whether there is a certain frequency at which it's possible to classify someone at high risk for developing ovarian cancer.

Being able to detect ovarian cancer early is important because when detected early, it can be cured by surgery in more than 90 percent of cases. However, most cases are typically detected when the disease is advanced and has spread. TP53 is "one of the main driver genes" for ovarian cancer, Risques said, and is mutated in more than 96 percent of high-grade serous ovarian cancer cases, "so it's a perfect target."

Meanwhile, Kennedy is focused on using the method in forensics applications to evaluate STR loci. "There's a limited amount of DNA in many forensics scenarios and I wanted to boost the efficiency of duplex sequencing," he said.

Jesse Salk, chief scientific officer at TwinStrand Bio and a co-author of the paper, said that TwinStrand has licensed the CRISPR-DS method and continues to collaborate with both Risques and Kennedy.

"From a commercial perspective, it's really valuable because one challenge of duplex sequencing is the efficient conversion of DNA molecules into sequence, which makes it an expensive process and has limited the size of the genomic targets we can look at," Salk said. As such, the firm is developing the CRISPR-DS method, as well as "a variety of other combinations with and new approaches to" duplex sequencing.

The company also recently received a $2.6 million three-year Small Business Innovation Research grant from the National Institutes of Health to develop an assay for the early detection of ovarian cancer, he said. 

According to TwinStrand's grant, the firm is working to develop and launch an early detection test within three years for women at risk of developing ovarian cancer. It will analyze DNA extracted from uterine lavage, a minimally invasive procedure through which fluid is collected from the uterus.

Salk declined to say whether a commercial assay would be based on the CRISPR-DS method or another form of duplex sequencing. "We're evaluating which of our specific products are most likely to benefit from that variation of duplex sequencing," he said.

In addition, Salk said that TwinStrand is evaluating how it could use CRISPR-DS for other applications, including detection of resistance to targeted oncology drugs as well as detection of chemotoxicity. The company presented an abstract at last month's Environmental Mutagenesis & Genomics Society conference on work it has done in collaboration with Amgen and BioReliance to identify chemically induced mutations in transgenic mice using duplex sequencing.

The firm also plans to pursue minimal residual disease detection in leukemia, as well as forensics applications.

"These are all fundamentally areas where we need to solve the needle-in-a-haystack problem," Salk said.