NEW YORK (GenomeWeb) – A public-private team from the US and UK has successfully used exome and targeted amplicon sequencing of tissue biopsies and circulating tumor DNA to characterize the genomic architecture of a breast cancer patient's tumor and track its clonal evolution over time in response to various therapies.
Though the work still needs to be replicated in larger cohorts, it identified specific groups of somatic mutations that might be ideal to hunt for when using ctDNA to monitor tumor burden, guide treatment, or follow therapeutic response in cancer patients. However, the research results also underscored some of the current potential pitfalls of using ctDNA in clinical applications.
Muhammed Murtaza, assistant professor and co-director of the Center for Noninvasive Diagnostics at the Translational Genomics Research Institute, presented the results of the study in a presentation sponsored by RainDance Technologies — whose droplet-based targeted sequencing platform was integral to the effort — earlier this month at the Association for Molecular Pathology annual meeting in Austin, Texas. Results from the study were concurrently published in Nature Communications.
Numerous studies have now been published on the potential use of ctDNA in patient blood as a way to diagnose and monitor solid tumors. However, in building the evidence base for this application, studies examining concordance between tumor and plasma samples have generally compared individual mutations, and have focused on single tumor biopsies, neglecting to address the issue of clonal heterogeneity.
In order to address this, Murtaza and colleagues from the Cancer Research UK Cambridge Institute, the University of Cambridge, Illumina, and elsewhere began working together a few years ago to apply whole-exome sequencing to ctDNA from cancer patients to elucidate acquired therapeutic resistance. The first results of their effort were published in Nature in 2013 and covered by GenomeWeb.
"The 2013 paper was actually proof of principle for exome sequencing of plasma DNA, showing that it's possible … to follow clonal evolution over time," Murtaza, who joined TGen from the Cambridge institutes during the course of the study, told GenomeWeb. "What that paper assumed was that there would be a fair representation of all the different tumor samples in plasma. What we've done in the 2015 paper is tried to address that assumption with some empirical data."
In the new study, Murtaza and colleagues analyzed eight tumor biopsy and nine plasma samples from a patient with estrogen receptor-positive, human epidermal growth factor receptor 2-positive metastatic breast cancer treated with sequential targeted therapies — tamoxifen and trastuzumab, followed by lapatinib — over three years.
They performed whole-exome sequencing of tumor biopsy samples using whole-genome libraries prepared with Rubicon Genomics' ThruPLEX-FD and analyzed on Illumina's HiSeq 2500 platform, as well as targeted amplicon sequencing on plasma DNA samples using RainDance's ThunderBolts Cancer Panel coupled with Illumina's MiSeq.
Murtaza noted that the depth of sequencing coverage afforded by RainDance's targeted sequencing technology was a key aspect of the study.
"In this particular study, we used whole-exome sequencing of several tumor and plasma samples to really get the breadth of mutations across all the different samples from this patient," he said. "And once we had that set of mutations, we used RainDance's kind of open-source method … to deep sequence a custom set of PCR amplicons that were designed for this particular patient. That allowed us to really get depth of sequencing, [and] get a fairly accurate assessment of changes in allelic fraction, as well as validate these mutations that we had called from the exome sequencing data."
The sequencing efforts uncovered 207 functional mutations, from which the group identified eight major mutation clusters based on variation in their allelic fractions as determined using Bayesian clustering enabled by PyClone, an informatics package several of the researchers developed for mutation analysis in tumor biopsies and in serially transplanted tumor xenografts.
"One of the observations from this detailed study of a single patient is that the concordance of somatic mutations between ctDNA and tumor biopsies can be qualified based on where each mutation resides in a cancer's phylogeny," Murtaza explained.
"When we … validated 207 mutations across [the tumor samples] and deep sequenced them with targeted sequencing, we saw mutations fell into different groups," he added. "We saw some mutations that were present and at high abundance across all tumor samples, so presumably these mutations occurred earlier in the cancer's phylogeny." The group dubbed these stem mutations, or truncal mutations.
Another group of mutations had a high abundance and were easily detectable in all metastatic tumor samples, but not so much in primary tumor or lymph node biopsies. "These are what we're calling the metastatic clade mutations," Murtaza said. "These were presumably the most recent, common ancestor of the metastatic tumor sample, if you will."
And a third group of mutations, called private mutations, was comparatively diverse because they were essentially spread out among each of the tumor biopsies.
Overall the researchers were able to demonstrate that ctDNA harbors somatic mutations that reflect the size and activity of distinct tumor sub-clones. Further, they showed that analyzing these mutations "reflects the clonal hierarchy determined from multiregional tumor sequencing and tracks different treatment responses across metastases," they wrote in their paper.
One of the study's key findings was that the so-called truncal mutations are likely the best candidates for monitoring tumor burden, as they are highest in circulating levels and least likely to drop out during follow-up.
"When you assess concordance between ctDNA and tumor samples for a particular mutation, lack of concordance could mean a number of different things, including either ctDNA levels are too low, or you are looking at a mutation that does not represent the complete tumor," Murtaza said. "I think that is one key aspect that needs to be kept in mind going forward. Based on these results, we would suggest that if you are following tumor burden, you'd want to pick a mutation that is truncal or stem for the tumor you're looking at, because that really represents the tumor burden and how it's changing."
Another note of caution from the study is that ctDNA analysis can pick up mutations that are not necessarily representative at all of solid tumors, potentially throwing clinical investigators off the scent. For instance, the researchers identified 11 high-confidence SNVs in plasma that were not detectable at greater than 2 percent allelic fraction in any of the analyzed tumor biopsies.
Among these was a hotspot actionable mutation in PIK3CA, which was identified in plasma with an allelic fraction of 3.5 percent during treatment with trastuzumab and tamoxifen, but dropped to about 1 percent and then became undetectable after lapatinib therapy started.
"Here's an actionable mutation that we picked up in plasma … but it's lower than the overall circulating abundance of stem mutations," Murtaza said. "And this mutation is something that is recurrently seen in breast cancer. Here is an actionable mutation, but clearly there has been heterogeneity for this mutation in the patient. So, what could possibly have been a driver mutation in breast cancer seems to be fairly heterogeneous and did not really play a significant role."
Murtaza summarized, "I hope this paper makes people realize that a lot of work still needs to be done in this area before heading to the clinic."