A Stanford University team has developed a sequencing scheme for finding individualized mutation signatures across swaths of sequence in tumor and circulating tumor DNA (ctDNA), while maintaining the sensitivity and quantification potential of more targeted approaches.
"In looking at existing [ctDNA sequencing] techniques, there are some that are very sensitive, but that only really look at one mutation at a time, and there are some that are very broad, like whole-genome or whole-exome [sequencing], that are not very sensitive," Ash Alizadeh, a stem cell biology, regenerative medicine, oncology, and hematology researcher at Stanford, told In Sequence.
"We wanted to sort of marry the two and take the best of both of those ideas to come up with a method that would be as sensitive as the most sensitive method published previously and be broadly applicable off the shelf to greater than 90 or 95 percent of patients with a given disease," Alizadeh explained.
To that end, Alizadeh and fellow Stanford researcher Maximilian Diehn led a team that came up with a sequencing approach known as "cancer personalized profiling by deep sequencing," or CAPP-Seq.
As they reported in Nature Medicine this week, the researchers' CAPP-Seq scheme is designed to track down patient-specific mutation patterns across far-reaching but targeted parts of the tumor genome by using existing genome or exome sequence data and a bioinformatics-based 'selector' to define parts of the genome prone to recurrent mutation in a particular cancer type.
Those regions are then captured and prepared using optimized library prep methods prior to a deep sequencing step designed to pick up even low frequency mutations in targeted regions of the genome. Because each tumor patient typically has distinct combinations of mutations across these targeted regions, the study's authors noted, it becomes possible to define personalized mutation markers for individual patients using a standardized set of reagents.
"What allowed us to accomplish our goals were improvements and innovations on the bioinformatics side — both in what parts of the genome to sequence and how to sensitively and specifically detect the presence of minor allele variants in the plasma," Diehn told IS.
Those bioinformatics tools were coupled with tweaks on the molecular biology side aimed at improving library preparation efficiency to capture as many DNA molecules as possible, he noted.
Preliminary experiments described in the new Nature Medicine paper suggest that CAPP-Seq may ultimately be useful for picking up patient-specific mutations using ctDNA alone, which would make it possible to follow individuals' treatment response, resistance, and recurrence from blood samples.
Even so, most of the proof-of-principle CAPP-Seq experiments described in the study — which focused on non-small-cell lung cancer — involved profiling of tumor samples prior to searching for corresponding mutations in patient blood samples.
By analyzing mutation database information and whole-exome sequence data generated for the Cancer Genome Atlas, the study's authors narrowed in on roughly 125,000 bases of sequence that were particularly prone to mutations in non-small-cell lung tumors.
Mutation patterns in the selector regions made it possible to routinely detect tumor DNA in blood samples from those with stage II to stage IV disease, the study's authors found. For those with stage I disease, they detected ctDNA by CAPP-Seq around half the time and with 96 percent specificity.
Diehn noted that CAPP-Seq should be relatively simple and affordable to apply to large sample sets from various cancers by using the specific set of oligos for each tumor type. "The approach is generalizable," he said, noting that "we did not want to have something where you have to redesign for every patient."
"In order to be able to use this in a clinical setting, one needs to have a tool that, off the shelf, works for every patient with that disease," Alizadeh added. "Otherwise, it's too much patient-specific optimization."
Along with anticipated patient monitoring applications, the researchers expressed enthusiasm about eventually using CAPP-Seq for blood-based early detection of cancer, though they conceded that that would require tweaks to the method, which currently relies on knowledge of the tumor type involved.
If and when that does become possible, Diehn noted that the mutation pattern detected in an individual's blood might make it possible to work backwards to get hints about the tumor type involved and its potential genetic vulnerabilities.
The existing CAPP-Seq method has been optimized for Illumina sequencing protocols and tested using that company's HiSeq 2000 and MiSeq instruments, though it's expected to be compatible with other sequencing chemistries as well.
For their study of non-small-cell lung cancer, for instance, the researchers turned to the Catalogue of Somatic Mutations and available exome sequence data for more than 400 non-small-cell lung cancer patients assessed through TCGA, using its selector tool to focus in on mutation-prone sequences in that cancer type.
The selector, which is based on an iterative algorithm approach, is agnostic to the names of the recurrently mutated genes. In the current study, for example, it highlighted regions containing both unknown genes as well as sequences representing genes implicated in non-small-cell lung cancer development and/or treatment response.
After folding in information on breakpoints in two non-small-cell lung cell cancer lines, the researchers settled on roughly 125,000 bases of sequence for targeting in patient samples. Those sequences spanned more than 500 exons, 139 recurrently mutated genes, and 13 introns.
From there, the group used custom NimbleGen hybrid selection methods and biotinylated oligos to nab such sequences from non-small-cell lung cancer cell lines and from blood and primary tumor samples from 13 individuals with the disease. The captured DNA was subsequently prepared using a Kapa Biosystems library prep kit with slight modifications added to it, including bead-based clean up steps.
Each of the libraries was then sequenced to around 10,000-fold average coverage by Illumina paired-end sequencing, as were libraries made using DNA in blood samples from five cancer-free control individuals.
To define each patient's personalized mutation set, the team assessed the resulting sequence data with another custom bioinformatics tool.
After determining that the method could pick up the presence of somatic mutations in a linear manner in those samples and in samples with pre-determined non-small-cell lung cancer DNA inputs, the researchers applied it to tumor and blood samples from 17 non-small-cell lung cancer patients — a group that included heavy smokers, light smokers, and never smokers.
An analysis of CAPP-Seq data generated for those patients' tumors uncovered six somatic mutations per tumor, on average, providing clues about corresponding mutations in ctDNA.
"We can, in effect, have a personalized biomarker for every patient — that patient's unique mutation signature in their tumor is what we look for in the blood," Alizadeh said.
When they applied the method to ctDNA in corresponding blood samples from 13 individuals with non-small-cell lung cancer and five unaffected controls, the investigators saw that CAPP-Seq could detect tumor DNA in the blood of stage II to stage IV patients with high sensitivity, picking up patient-specific mutations in plasma from 100 percent of patients from those disease stages.
Alterations identified by CAPP-Seq turned up in ctDNA around half the time in stage I non-small-cell lung cancer patients, researchers reported, a decline in sensitivity that the team attributed to lower ctDNA levels in the blood.
Indeed, the researchers noted that levels of ctDNA detected by CAPP-Seq appear to coincide with the size of the tumor shedding DNA into the bloodstream, as assessed by standard imaging approaches.
Follow-up experiments indicated that the newly described sequencing method is sensitive enough to see mutations present at frequencies of 0.1 percent or higher across the targeted selector region.
In general, the cost of applying CAPP-Seq is expected to vary depending on the type of cancer involved and the size of the region that's captured and sequenced from the tumor or ctDNA genome.
"If one has a tumor [type] where we need twice as big a capture space to get enough mutations for an average patient, that would double the cost," Diehn explained. "But we think, really, that all the common cancers should be achievable in the hundreds of dollars per sample range."
In the case of non-small-cell lung cancer, for example, the researchers spent roughly $200 per sample on reagents and sequencing.
Based on their results to date, the study's authors are optimistic that CAPP-Seq could eventually replace or augment expensive imaging methods such as PET CT scans that are used to monitor cancer response or recurrence.
"Having a test that may be able to do the same thing — to tell you whether the cancer is growing or coming back — in the hundreds of dollars range, we think, is potentially one of the applications of this clinically in the near future," Diehn said.
The selector approach behind CAPP-Seq is expected to be applicable to a wide range of common cancer types, given the extensive tumor profiling done by large efforts such as TCGA, the International Cancer Genome Atlas, and other groups.
For rarer tumor types, Diehn noted that researchers may need to do an initial exome sequencing study prior to applying the CAPP-Seq selector analysis and targeted capture steps of the method.
The team is continuing to pursue strategies for making CAPP-Seq more sensitive to mutations shed into the blood by early stage tumors. In particular, it is considering ways of speeding up and optimizing library prep methods since the method's sensitivity is limited, in part, by PCR errors introduced during DNA amplification.
The group also plans to further refine its statistical algorithms with an eye to calling ever more rare mutations above the background sequencing noise.
In the meantime, the researchers have started collaborating with investigators at Stanford Hospital's pathology lab and elsewhere to begin incorporating CAPP-Seq into prospective clinical trials. To that end, they are also looking at how well mutation patterns detected directly from ctDNA by CAPP-Seq correspond to those found using tumor tissue.