Circulating tumor DNA identified by fragment size

Circulating tumor DNA could be effectively isolated from plasma by focusing on a particular range of fragment sizes, which paves the way for noninvasive genomic analysis of tumor DNA, new research suggests.

In a study of 344 plasma samples from 200 patients with 18 cancer types and 65 samples from healthy controls, DNA fragment length could be used to distinguish circulating tumor DNA (ctDNA) from other cell-free DNA (cfDNA), investigators reported in Science Translational Medicine.

“We hypothesized that we could improve the sensitivity for noninvasive cancer genomics by selective sequencing of ctDNA fragments and by leveraging differences in the biology that determine DNA fragmentation,” wrote Florent Mouliere, PhD, from the Cancer Research UK Cambridge Institute, and coauthors.

Cell-free plasma fragments are often cleaved at around 167 base pairs in length and differences in length between circulating fetal and maternal DNA are already used for noninvasive prenatal diagnosis. However, the authors said that only a few studies, with conflicting results, have looked at the size distribution of tumor-derived cfDNA.

The study used two approaches to determining the size profile of mutant ctDNA. The first looked at tumor and nontumor cfDNA in mice with human ovarian cancer xenografts and the second approach used deep sequencing in 19 cancer patients. This revealed that tumor-derived cfDNA was most commonly found in fragments between 90-150 base pairs or 250-320 base pairs in size.

The researchers also noted that mutant circulating tumor DNA was generally more fragmented than nonmutant cfDNA and that patients with untreated advanced cancer showed consistently shorter lengths of mutant DNA.

The next question was whether size selection and other biological properties – such as somatic alterations – of the cfDNA could be used to enhance detection of ctDNA via machine learning technology.

Two models, designed to distinguish between healthy and cancerous samples, were developed using 153 samples, then validated on two datasets of 94 and 83 samples.

One of these models correctly classified cancerous samples in 94% of samples from patients with cancers known to have high levels of ctDNA – colorectal, cholangiocarcinoma, ovarian, breast, and melanoma – and in 65% of samples from low-ctDNA cancers – pancreatic, renal, and glioma.

Another model focused just on fragmentation patterns and was still able to distinguish cancer samples from those of healthy controls, although with slightly reduced area under the curve.

“Our results indicate that exploiting fundamental properties of cfDNA with fragment-specific analyses can allow more sensitive evaluation of ctDNA,” the authors wrote. “We identified features that could determine the presence and amount of ctDNA in plasma samples, without a prior knowledge of somatic aberrations.”

The authors pointed out that size selection of DNA fragments was relatively simple and cheap, and was also compatible with other genome-wide and targeted genomic analyses, “greatly increasing the potential value and utility of liquid biopsies as well as the cost-effectiveness of cfDNA sequencing.”

However, they cautioned that their catalogue had focused solely on double-stranded DNA and was subject to potential biases from the DNA extraction and sequencing methods they used in the study. They also commented that other biological effects could help refine the analysis of ctDNA.

“Other bodily fluids [urine, cerebrospinal fluid, and saliva], different nucleic acids and structures, altered mechanisms of release into circulation, or sample processing methods could exhibit varying fragment size signatures and could offer additional exploitable biological patterns for selective sequencing,” they wrote.

The study was supported by the University of Cambridge, Cancer Research UK, and the Engineering and Physical Sciences Research Council. Research supporting the study was also funded by the European Research Council, the National Institute for Health Research Cambridge, National Cancer Research Network, Cambridge Experimental Cancer Medicine Centre, Hutchison Whampoa, Target Ovarian Cancer, the Medical Research Council, and AstraZeneca. Three authors are cofounders, shareholders, and officers/consultants in a company specializing in ctDNA analysis. One author declared research funding and advisory board fees from private industry. Seven authors are listed on related patents.

SOURCE: Mouliere F et al. Sci Transl Med. 2018 Nov 7. doi: 10.1126/scitranslmed.aat4921.
 

Circulating tumor DNA could be effectively isolated from plasma by focusing on a particular range of fragment sizes, which paves the way for noninvasive genomic analysis of tumor DNA, new research suggests.

In a study of 344 plasma samples from 200 patients with 18 cancer types and 65 samples from healthy controls, DNA fragment length could be used to distinguish circulating tumor DNA (ctDNA) from other cell-free DNA (cfDNA), investigators reported in Science Translational Medicine.

“We hypothesized that we could improve the sensitivity for noninvasive cancer genomics by selective sequencing of ctDNA fragments and by leveraging differences in the biology that determine DNA fragmentation,” wrote Florent Mouliere, PhD, from the Cancer Research UK Cambridge Institute, and coauthors.

Cell-free plasma fragments are often cleaved at around 167 base pairs in length and differences in length between circulating fetal and maternal DNA are already used for noninvasive prenatal diagnosis. However, the authors said that only a few studies, with conflicting results, have looked at the size distribution of tumor-derived cfDNA.

The study used two approaches to determining the size profile of mutant ctDNA. The first looked at tumor and nontumor cfDNA in mice with human ovarian cancer xenografts and the second approach used deep sequencing in 19 cancer patients. This revealed that tumor-derived cfDNA was most commonly found in fragments between 90-150 base pairs or 250-320 base pairs in size.

The researchers also noted that mutant circulating tumor DNA was generally more fragmented than nonmutant cfDNA and that patients with untreated advanced cancer showed consistently shorter lengths of mutant DNA.

The next question was whether size selection and other biological properties – such as somatic alterations – of the cfDNA could be used to enhance detection of ctDNA via machine learning technology.

Two models, designed to distinguish between healthy and cancerous samples, were developed using 153 samples, then validated on two datasets of 94 and 83 samples.

One of these models correctly classified cancerous samples in 94% of samples from patients with cancers known to have high levels of ctDNA – colorectal, cholangiocarcinoma, ovarian, breast, and melanoma – and in 65% of samples from low-ctDNA cancers – pancreatic, renal, and glioma.

Another model focused just on fragmentation patterns and was still able to distinguish cancer samples from those of healthy controls, although with slightly reduced area under the curve.

“Our results indicate that exploiting fundamental properties of cfDNA with fragment-specific analyses can allow more sensitive evaluation of ctDNA,” the authors wrote. “We identified features that could determine the presence and amount of ctDNA in plasma samples, without a prior knowledge of somatic aberrations.”

The authors pointed out that size selection of DNA fragments was relatively simple and cheap, and was also compatible with other genome-wide and targeted genomic analyses, “greatly increasing the potential value and utility of liquid biopsies as well as the cost-effectiveness of cfDNA sequencing.”

However, they cautioned that their catalogue had focused solely on double-stranded DNA and was subject to potential biases from the DNA extraction and sequencing methods they used in the study. They also commented that other biological effects could help refine the analysis of ctDNA.

“Other bodily fluids [urine, cerebrospinal fluid, and saliva], different nucleic acids and structures, altered mechanisms of release into circulation, or sample processing methods could exhibit varying fragment size signatures and could offer additional exploitable biological patterns for selective sequencing,” they wrote.

The study was supported by the University of Cambridge, Cancer Research UK, and the Engineering and Physical Sciences Research Council. Research supporting the study was also funded by the European Research Council, the National Institute for Health Research Cambridge, National Cancer Research Network, Cambridge Experimental Cancer Medicine Centre, Hutchison Whampoa, Target Ovarian Cancer, the Medical Research Council, and AstraZeneca. Three authors are cofounders, shareholders, and officers/consultants in a company specializing in ctDNA analysis. One author declared research funding and advisory board fees from private industry. Seven authors are listed on related patents.

SOURCE: Mouliere F et al. Sci Transl Med. 2018 Nov 7. doi: 10.1126/scitranslmed.aat4921.
 

Circulating tumor DNA could be effectively isolated from plasma by focusing on a particular range of fragment sizes, which paves the way for noninvasive genomic analysis of tumor DNA, new research suggests.

In a study of 344 plasma samples from 200 patients with 18 cancer types and 65 samples from healthy controls, DNA fragment length could be used to distinguish circulating tumor DNA (ctDNA) from other cell-free DNA (cfDNA), investigators reported in Science Translational Medicine.

“We hypothesized that we could improve the sensitivity for noninvasive cancer genomics by selective sequencing of ctDNA fragments and by leveraging differences in the biology that determine DNA fragmentation,” wrote Florent Mouliere, PhD, from the Cancer Research UK Cambridge Institute, and coauthors.

Cell-free plasma fragments are often cleaved at around 167 base pairs in length and differences in length between circulating fetal and maternal DNA are already used for noninvasive prenatal diagnosis. However, the authors said that only a few studies, with conflicting results, have looked at the size distribution of tumor-derived cfDNA.

The study used two approaches to determining the size profile of mutant ctDNA. The first looked at tumor and nontumor cfDNA in mice with human ovarian cancer xenografts and the second approach used deep sequencing in 19 cancer patients. This revealed that tumor-derived cfDNA was most commonly found in fragments between 90-150 base pairs or 250-320 base pairs in size.

The researchers also noted that mutant circulating tumor DNA was generally more fragmented than nonmutant cfDNA and that patients with untreated advanced cancer showed consistently shorter lengths of mutant DNA.

The next question was whether size selection and other biological properties – such as somatic alterations – of the cfDNA could be used to enhance detection of ctDNA via machine learning technology.

Two models, designed to distinguish between healthy and cancerous samples, were developed using 153 samples, then validated on two datasets of 94 and 83 samples.

One of these models correctly classified cancerous samples in 94% of samples from patients with cancers known to have high levels of ctDNA – colorectal, cholangiocarcinoma, ovarian, breast, and melanoma – and in 65% of samples from low-ctDNA cancers – pancreatic, renal, and glioma.

Another model focused just on fragmentation patterns and was still able to distinguish cancer samples from those of healthy controls, although with slightly reduced area under the curve.

“Our results indicate that exploiting fundamental properties of cfDNA with fragment-specific analyses can allow more sensitive evaluation of ctDNA,” the authors wrote. “We identified features that could determine the presence and amount of ctDNA in plasma samples, without a prior knowledge of somatic aberrations.”

The authors pointed out that size selection of DNA fragments was relatively simple and cheap, and was also compatible with other genome-wide and targeted genomic analyses, “greatly increasing the potential value and utility of liquid biopsies as well as the cost-effectiveness of cfDNA sequencing.”

However, they cautioned that their catalogue had focused solely on double-stranded DNA and was subject to potential biases from the DNA extraction and sequencing methods they used in the study. They also commented that other biological effects could help refine the analysis of ctDNA.

“Other bodily fluids [urine, cerebrospinal fluid, and saliva], different nucleic acids and structures, altered mechanisms of release into circulation, or sample processing methods could exhibit varying fragment size signatures and could offer additional exploitable biological patterns for selective sequencing,” they wrote.

The study was supported by the University of Cambridge, Cancer Research UK, and the Engineering and Physical Sciences Research Council. Research supporting the study was also funded by the European Research Council, the National Institute for Health Research Cambridge, National Cancer Research Network, Cambridge Experimental Cancer Medicine Centre, Hutchison Whampoa, Target Ovarian Cancer, the Medical Research Council, and AstraZeneca. Three authors are cofounders, shareholders, and officers/consultants in a company specializing in ctDNA analysis. One author declared research funding and advisory board fees from private industry. Seven authors are listed on related patents.

SOURCE: Mouliere F et al. Sci Transl Med. 2018 Nov 7. doi: 10.1126/scitranslmed.aat4921.