With a proof-of-concept study in hand showing that it's possible to detect cancer-related chromosomal rearrangements and copy number glitches directly from blood samples, a Johns Hopkins University-led team is considering the most appropriate clinical application for the approach given the current cost of sequencing.
"We're very excited about the next step," co-first author Rebecca Leary, a post-doctoral researcher at the Johns Hopkins Kimmel Cancer Center's Ludwig Center for Cancer Genetics and Therapeutics, told Clinical Sequencing News. "We'll need to bring this into larger clinical trials to really figure out the best application."
Leary did not disclose details about the nature or status of potential clinical trials involving the approach. But she noted that as sequencing costs continue to drop, it should be feasible to think about using the blood-based detection method for increasingly broader clinical applications.
As they reported in Science Translational Medicine last week, the researchers generated low-coverage genome sequence data on cell-free DNA in blood samples from 10 individuals with late-stage colon or breast cancers and 10 unaffected controls.
For each of the patients, analyses of the sequence read data uncovered chromosomal rearrangements and/or copy number changes in circulating tumor DNA, which made up between 1.4 and 47.9 percent of the cell-free DNA in patients' blood samples. Such chromosomal alterations were missing from healthy control blood samples.
"This whole-genome, unbiased approach is a really great way to use the phenomenon of chromosome instability as a marker of disease," Leary said.
"Our group is engaged in a number of these whole-genome analyses of several tumor types," she explained, "and what's remarkable is that in every tumor type we examine, we see these structural alterations, this chromosomal instability, as a hallmark of human cancer."
Since it's anticipated that early-stage cancers might release only very small amounts of DNA in the bloodstream, though, researchers suspect that they may need to get extremely deep coverage of cell-free circulating DNA to routinely diagnose and/or characterize cancers using the blood-based method. And that could prove prohibitively expensive — at least for the time being.
"Sensitivity is dependent on the amount of sequence data obtained," Leary explained, "so as sequencing costs continue to decrease, this approach could be used as a diagnostic test in the future."
In the nearer term, those involved in the effort are hopeful that they'll be able to use cell-free tumor DNA sequencing to track treatment response, detect residual or recurrent disease, or perhaps do blood-based biopsies in circumstances where tumor material is unavailable.
While he called it a "really great proof-of-concept paper," Harvard Medical School's Peter Park, who was not involved in the study, argued that even more sequence depth may be also be needed to make more robust structural variation calls.
"The fundamental limitation is that if the fraction of tumor DNA is very small, then you have to sequence a lot to learn about CNVs," he told CSN. "It will be expensive to get the level of resolution that we'd really like.
"But in due time, if the cost goes down far enough, this would really be a good strategy," Park continued. "The fact that you don't have to get the [tumor] tissue is huge, because that's always a major problem."
For his part, the Institute of Cancer Research's Johann de Bono said that while results from the study are somewhat preliminary, they support data from other labs suggesting that tumor DNA in the blood can be used for doing liquid tumor biopsies in a relatively non-invasive way.
"It is going to be a very important way forward to deliver cancer medicine and cancer research, because of the ease with which we can access this tumor DNA," de Bono said.
Several teams have been exploring strategies for using circulating tumor DNA as a window for finding and following cancer-related alterations within specific genes, such as a group at Cancer Research UK that developed a technique dubbed TAm-Seq (CSN 6/6/2012). And others, such as Dennis Lo's group, have sought to follow alterations across the entire cancer genome (CSN 10/17/2012).
For the most part, though, existing approaches have hinged on having some prior knowledge of the genetic glitches present in the tumor itself. In a 2010 Science Translational Medicine study, for instance, Leary and her Johns Hopkins co-authors described a method called personalized analysis of rearranged ends, or PARE, which can be used to find known tumor-related rearrangements in cell-free circulating tumor DNA (IS 2/23/2010).
A blood-based approach similar to PARE was used to track treatment outcomes and residual disease over time in four neuroblastoma patients for a Nature Genetics study published this week. In that paper, researchers from Johns Hopkins, the Children's Hospital of Philadelphia, and elsewhere tested patient blood samples for personalized biomarkers originally found in tumor samples.
For their latest Science Translational Medicine study, the researchers showed that they can also find chromosomal rearrangements and other chromosome arm changes in tumor DNA directly from blood samples without first consulting the tumor sequence for guidance.
"Our approach is unbiased," Leary said. "We don't know what we're looking for when we get our patient plasma — we don't know where our rearrangements are and we don't know which chromosome arms are altered."
Putting it to the Test
The team tested its sequencing-based approach using blood samples from three individuals with breast cancer, seven individuals with colorectal cancer, and 10 unaffected controls.
These samples were prepared using a strategy tailored to both the anticipated fragment size of cell-free tumor DNA, which tends to be less than 200 base pairs long, and to low total amounts of DNA.
From there, researchers generated one to two lanes of sequence data per paired-end library on the Illumina HiSeq 2000. Once these reads were mapped to the human reference genome, the team was able to see potential rearrangements in places where paired-end reads mapped aberrantly.
"We have very stringent criteria for enriching our output candidates as true somatic events," Leary noted. "We require, for rearrangements, that those rearrangements be a minimum size or involve two different chromosomes."
"For copy number alterations, we're looking at gains and losses of whole chromosomal arms, which we tend not to see in the general population," she added.
That copy number analysis relied on the same read data. But there, the researchers used single tags from each paired-end read in a strategy similar to digital karyotyping, looking at how many tags mapped to a given stretch of the genome relative to the number expected to map there based on the overall number of tags sequenced.
In determining how many tags were predicted to map to each region of the genome, Leary explained, researchers took into account a range of factors, including GC bias in reads from different parts of the genome, known copy number polymorphisms, and so on.
Using SNP chip data from three-dozen colorectal cancers and 45 breast cancers, the researchers got a sense of the frequency with which whole chromosome arms are gained or lost in these cancers — information that they used to come up with an aneuploidy score for identifying cancers in cell-free plasma DNA.
Though these analyses did not unearth chromosomal alterations in plasma samples from 10 healthy individuals, chromosomal rearrangements, gains, and losses did turn up in blood samples from each of the patient blood samples. Circulating DNA from every cancer case reached the team's cutoff for chromosomal aneuploidy, and chromosomal rearrangements were detected in nine cases.
All of the rearrangements detected in patient plasma samples were confirmed in matched tumor samples, when available, though Leary noted that the team's bioinformatics approaches were not optimized to do reciprocal analyses, meaning there may be additional chromosomal alterations in the tumor.
It's All About Context
So far the team has only tested its approach using the Illumina platform. Since the method hinges on the detection of sequence tags, Leary said, it is theoretically compatible with other platforms as well, though the throughput would likely need to be on par with the HiSeq.
"There's no minimum sequence depth that we need," she said, "but the sensitivity of the test really depends on sequencing depth."
From one lane of HiSeq data, for example, the researchers estimated that they are able to detect chromosome arm alterations with 90 percent sensitivity and 99 percent specificity, assuming 0.75 percent circulating tumor DNA in a patient's blood sample.
"For things that could be inferred from chromosome arm-level data, this is certainly useful," Harvard's Park said. "It really depends on the context."
For instance, in a study published in Genome Research last month, Park and his colleagues used existing copy number data to look for recurrent chromosomal alterations in more than 8,200 human cancer genomes. The study pointed to some recurrent changes that occur across cancer lineages, along with other profiles that seem to be tumor-type specific. That hints at the possibility of using information on chromosomal alterations in tumor DNA at the time of diagnosis to gain clues about the type of cancer involved.
"If you know certain chromosome arms that are deleted, that could point to a particular tumor type," Park said. "That correlation is relatively easy," he added, though he cautioned that such a strategy likely wouldn't provide information on the cancer subtype or genes affected.
Nevertheless, Park noted that there are instances in which tumors have very little in the way of chromosomal aberrations. And for those tumors, cancer detection based on chromosomal alterations would presumably be trickier.
Moreover, in situations where researchers are interested in focusing on particular genes to find treatment-related mutations, finer scale information on the tumor genome and far greater sequencing depths would likely be needed, he argued. "I think greater sequencing depth is needed to get more reliable gene-specific information."
For the current study, researchers did not strive for the sequence-level resolution needed to see point mutations, small insertions and deletions, or subtle copy number shifts.
Even so, the method did pick up amplifications of known cancer genes such as ERBB2 and CDK6. And there are hints that it could be used for finding even finer scale changes.
"Our approach could be modified slightly and applied to identify focal amplifications and homozygous deletions," Leary noted, "especially in genomic regions containing known driver genes."
Narrowing in on smaller stretches of sequences would likely require some tweaks to the team's bioinformatics approach, she said, noting that the noise tends to increase when looking at these smaller windows.
Meanwhile, the amount of sequencing needed to achieve the types of tumor genome coverage needed for more refined analyses is expected to depend on the amount of DNA that a tumor releases into the blood. And researchers are still working to get a handle on the factors influencing tumor DNA concentrations in the blood.
In the current study, for example, the proportion of tumor DNA in blood plasma varied dramatically from one individual to the next, even though all of the patients had late-stage breast or colon cancers.
"There needs to be a lot more work done in terms of looking at the spectrum of patients with different amounts of circulating tumor DNA, both within a single stage and between stages," Leary said.
Indeed, that variability in circulating cell-free DNA concentrations is something that de Bono's UK team has started exploring in more detail. In a study published in PLoS ONE last month, the group used Sequenom's MassArray system to test blood samples from more than 100 cancer patients, looking at how concentrations of tumor DNA in circulating plasma samples varied with cancer type and patient outcomes.
De Bono said that he expects circulating tumor DNA will be clinically useful, not only for tracking disease progression, treatment response, and recurrence but also for diagnosing cancers. And along with the information this material provides about alterations present in the tumor genome, he added, there are hints that the amount of tumor DNA that's made it's way into patient blood may have prognostic value, too.
"There are some cancers where there seem to be lower levels of plasma DNA — or the plasma DNA levels are prognostic of survival — and I suspect that that might be another way where this could be very useful in the clinic," de Bono said.