NEW YORK – Researchers at Memorial Sloan Kettering Cancer Center shared new data from the center's liquid biopsy efforts this week, using a broad gene panel and deep sequencing to demonstrate the value of tumor-normal sequencing and to clarify lingering questions about differences seen between plasma and tumor tissue DNA.
In a paper published in Nature Medicine on Monday, the MSK authors wrote that using a 2 Mb, 508-gene panel and sequencing to more than 60,000x depth, they found that the vast majority of cell-free DNA mutations in both non-cancer controls and cancer patients had features consistent with clonal hematopoiesis, a process whereby white blood cells progressively accumulate somatic alterations without necessarily producing a hematological condition or malignancy.
Such variants may incorrectly be called tumor mutations by liquid biopsy pipelines that don't sequence and compare cell-free tumor DNA against DNA from circulating white blood cells. This is of concern, the authors argued, considering tumor-only methods are employed widely, including in commercial liquid biopsy tests.
"Without taking into account the results of WBC sequencing, cfDNA [results] might be misleading, given that some CH mutations affecting cancer genes may be interpreted as tumor-derived mutations (for example, TP53 mutations)," the group wrote.
Although bioinformatic methods exist to filter variant calls that are likely derived from CH, they have a long way to go, said Pedram Razavi and Jorge Reis-Filho, MSK physicians and co-authors of the new study.
"We are not there yet," Razavi said. "Maybe when we have hundreds of thousands of cfDNA sequencing paired with white cells, then we can probably come up with the algorithms that call them."
"But we [see clearly in our study that] clonal hematopoiesis can randomly involve other genes that are not just canonical genes. So it's very, very difficult to develop AI algorithms to distinguish [them]," he said.
"We are not saying that's impossible," Reis-Filho added. "But … at present, the methods to mitigate the potential confounders coming from clonal hematopoiesis [don't seem to be] mature enough."
MSK has been offering a clinical liquid biopsy assay, called MSK-ACCESS, since receiving NY State approval in June. Since then the center has analyzed over 350 patient samples, including about 200 lung cancer patients, with performance exceeding expectations, according to Marc Ladanyi, chief of MSK's Molecular Diagnostics Service.
The ACCESS test covers a much narrower 129-gene region than the experimental panel used in the study this week but does incorporate some of the same methodologies, including matched tumor and normal sequencing, as well as so-called duplex unique molecular indexing, an error-correction method to distinguish PCR errors from true mutations.
Both MSK-ACCESS and the research team's 508-gene panel also involve ultra-high depth sequencing. But where the clinical test averages around 15,000x, the researchers were sequencing beyond 60,000x for their study.
"It's a genomic footprint over five times bigger, and raw sequencing over three times deeper," Reis-Filho said.
The data published in the study are an extension of efforts MSK has been involved in for several years with cancer early detection firm Grail, trying to push the limits of sequencing depth and coverage to explore ways to make liquid biopsy sensitive and specific enough for application in cancer screening.
Much of Grail's focus, as well as many of its competitors', has since shifted to genome-wide epigenomic signals, gleaned from methylation patterns, chromatin remodeling, DNA fragmentation patterns, and other measurements.
But Razavi and Reis-Filho said that while mutation detection may also still play a role there, it is increasingly being tapped for disease monitoring.
"With assays being commercially available, we can now send multiple DNA tests during the lifetime of the patient … [and] this type of assay, with low error rates and high genomic footprint, can give you a very good picture of clonal evolution of the disease and development of mechanisms of resistance under therapeutic pressure," Reis-Filho said.
Of course, cost presents an issue. High-intensity sequencing as demonstrated in the study is much more expensive than targeted, shallower assays. But the researchers argued that developments in how Medicare views sequencing tests may open possibilities to prove the value for longitudinal use even of high-intensity approaches like theirs.
"Medicare now considering disease progression … is a big one," Reis-Filho said. "That's one huge step forward, where at least after each line of treatment, we can repeat testing."
"And I think when we [start to prove] the value of doing this type of monitoring, the clinicians and the insurance companies and everyone else will embrace strategies to do it even more often," he added.
In the new study, the MSK team analyzed samples from 124 patients with metastatic cancer and 47 controls without cancer.
Among other takeaways, the group said they found clonal hematopoiesis to be both more prevalent as a biological phenomenon and more problematic as a potential technical pitfall than previously recognized.
The number of probable CH variants detected using their high-intensity sequencing protocol was higher in cancer patients than the number of tumor-derived variants. It was also high overall for both cancer patients and normal controls, significantly outstripping what has been seen in previous studies. Mutations appeared in 93.6 percent of the white blood cells from individuals without cancer and 99.1 percent of those with cancer.
"What we've realized ... is that if you exclude those cases with a hypermutated phenotype ... most of the mutations that we see in plasma don't come from the cancer cells," Reis-Filho said.
Being able to identify this higher rate of CH is a result of sequencing both cfDNA and WBC samples with the same ultra-deep threshold, the authors wrote, which allowed them to distinguish between CH and tumor-derived mutations with much more precision than possible in other studies.
"What we are seeing is what we call very subclonal clonal hematopoiesis … occurring at very very low levels in the bone marrow," said Razavi.
"But if you don't do the white cells ... you would call these low-level variants as tumor derived," he said. "It's going to be very difficult to know they are not ... if you don't do the white cells to the same depth."
Though genes harboring WBC-matched somatic mutations in the study were largely canonical, some were not. And the majority of the variants were private to individual patients, suggesting that accounting for them in clinical liquid biopsy assays may require a tumor-normal matched pipeline, especially for assays that hope to read out biomarkers like tumor mutational burden that require calculation of overall mutation rates.
"Most clinical tests don't sequence such a large panel," Razavi said. But the field seems to be moving in that direction. Guardant Health, for example, offers a 500-gene research assay that does not analyze white blood cells in parallel. And sequencing firm Personalis has said it is developing a whole-exome liquid biopsy test that it intends to commercialize next year.
The broader you go, Razavi said, the more important it becomes to sequence white blood cells, "because the chances of finding false positives increases as you increase the size of the panel."
That said, he added, even with the commercially available smaller panels that don't cover the canonical CH genes, it is still possible for CH variants to be called as tumor derived.
"We have a lot of evidence now that genes such as TP53, KRAS, NF1, and others can be involved," he said, which are often part of smaller clinical assays.
Razavi also argued that the findings help quash earlier worries that high-depth and breadth liquid biopsy can't be done without a correspondingly high false-positive rate.
"Several papers [published previously] have assumed that those mutations that you do not find in cancer tissue but are readily available in the plasma were technical noise. But our paper shows, no, that's not true. [T]hose mutations are [coming from] white blood cells," he said.
"How could people not find them before? Because they had never sequenced the white blood cells to the depth we did," he said.
Razavi and Reis-Filho said that MSK is already exploring using higher-intensity, high-sensitivity methods to track disease evolution in advanced cancer patients, with some kind of longitudinal testing happening in most of the trials the center is involved in at this point.
Aside from implications for cancer patients, a better understanding of clonal hematopoiesis prevalence and patterns, as emerging from the team's work, is also exciting for future work, Reis-Filho added.
"A lot of what we know about CH comes from relatively low-depth sequencing done on hundreds of thousands of individuals, which found this to be prevalent in about 25 percent of people over age 60. But that's definitely an underestimation, based on what we are seeing," he said.
"If you sequence the white cells to 70,000x on a 20-something-year-old, and we have younger people in this cohort, they have CH. So now the question is, what are the patterns of CH that are important, and what are the genes? What are the co-mutations, what are the levels of CH, what patterns are the important ones?"
For example, he suggested, CH may represent ageing phenotypes or biological ageing processes that don't match chronological age.