Plasma DNA for early cancer detection – opportunities and challenges

KEYWORDS:

• Liquid Biopsy
• plasma DNA tissue mapping
• epigenome
• copy number aberrations
• cancer screening

Plasma DNA is an aggregate of DNA materials released from different cells and tissues into the circulation, including cancer cells if present. Analysis of tumor-derived DNA in circulation, or circulating tumor DNA (ctDNA), forms the basis of plasma DNA-based cancer diagnostics. In patients with an established diagnosis of cancer, the analysis of ctDNA allows genomic profiling of the tumor in a noninvasive manner and provides prognostic and predictive [1] information. Recently, as demonstrated in the model of nasopharyngeal cancer (NPC), ctDNA analysis could potentially be extended to screening early cancer in asymptomatic individuals [2]. Circulating Epstein-Barr virus (EBV) DNA has long been used as a surrogate marker for NPC as virtually all NPCs in endemic regions harbor the EBV genome [3]. In the study, over 20,000 asymptomatic men were prospectively screened by real-time PCR analysis for plasma EBV DNA. Among the total of 34 cases of NPC identified, 24 (70%) of them had early-stage (stage I and II) diseases. This stage distribution was remarkable as only around 30% of NPC patients would present with early-stage disease without screening according to the local cancer registry. Because of the shifting to earlier stages, patients identified by screening had much superior progression-free survival compared with those in historical cohorts with a hazard ratio of only 0.1. This study not only demonstrated the feasibility of using plasma EBV DNA for screening NPC but more importantly provided insight on the amount of DNA would be released from tumors into the circulation in patients with very early cancers. This information is important for the future development of plasma DNA-based tests for the screening of other non-viral-associated cancers.

For the screening of early asymptomatic cancers, a high sensitivity is required for the screening test to be clinically useful. There are various biological and technical considerations for a plasma DNA-based cancer screening test to attain a high sensitivity. Biologically, it has been shown that the concentration of ctDNA in plasma correlates with tumor burden [4]. Therefore, the expectedly low concentration of ctDNA in pre-symptomatic cancer patients poses a challenge to the sensitivity of plasma DNA-based screening tests. To enhance the detection sensitivity, multiple genetic mutations could be used simultaneously as surrogate markers of cancer. Indeed, based on the model of NPC [2], it was estimated that analysis of 500 targets per cancer cell genome (50 EBV genomes per cell and 10 repeats of BamH1-W region as PCR targets per genome) could provide a similarly high level of sensitivity as using plasma EBV DNA for screening NPC. For the detection of multiple genetic mutations in plasma DNA, there are two approaches, including a targeted approach on mutational hotspots, for example cancer gene mutations, and a genomewide non-targeted approach. The former approach is limited by a lack of common hotspot mutations for most cancer types [5]. To overcome this problem, a large genomic region, for example whole exomes or even whole genome, needs to be searched for not only ‘driver mutations’ but also ‘passenger mutations’ [5].

However, a genomewide search of mutations among a background of plasma DNA which is mostly derived from non-malignant tissues is technically challenging. Existing sequencing technologies would introduce errors at a rate of about 0.1%. Therefore, the signal-to-noise ratio of the mutation would be very low when the fractional concentration of the tumoral DNA is less than 1% and there is less than 1 mutation per kb. The strategy of adding unique molecular identifiers (UMI) to each template molecule [6] was introduced to tackle this problem. In this method, each template DNA was tagged with a unique identifier of 6–8 random nucleotides, i.e. UMI, so that different template molecules would have a different UMI. PCR copies of each template DNA molecule would share the same UMI. Errors generated during later cycles of PCR amplification or during sequencing can be identified by making a consensus call for all sequenced molecules with the same UMI. In a recent report by Cohen et al. [7], 1005 patients with non-metastatic cancers of 8 different types and 812 healthy controls were analyzed for cancer-associated mutations and proteins in circulation. Mutations in 16 driver genes were detected via a 61-amplicon-based sequencing assay with the incorporation of the UMI technology. The reported sensitivities of the combined detection of cancer-associated mutations and proteins were 43% for stage I, 73% for stage II and 78% for stage III cancers of all 8 types. The relatively low sensitivity for early-stage cancers once again reflects the challenges of ctDNA analysis for cancer screening.

Another issue of using plasma DNA analysis for cancer screening is the specificity of plasma DNA mutations for malignancies. Clonal hematopoiesis has been postulated to be one source of mutations in plasma DNA of apparently healthy individuals. It could be observed in up to 10% of people older than 65 years old [8]. Although clonal hematopoiesis predicts the future risk of hematological malignancies, most subjects with this age-related hematopoietic mosaicism do not develop into cancers in their lifetime. Sequencing of the white blood cells from the same individual may therefore be used for minimizing the false-positive screening results associated with mutations arising from clonal hematopoiesis.

Copy number aberration is another common feature of various types of cancers, which could be detected through ctDNA analysis and serve as cancer-specific signatures. Shot-gun sequencing approach established for detecting fetal chromosomal aneuploidies in maternal plasma can be adapted for the detection of cancer-associated copy number aberrations in the plasma of cancer patients [9]. Chromosomal regions that are deleted and amplified in tumor tissues would be under- and over-represented, respectively, in plasma DNA. However, in contrast to the common fetal aneuploidies involving a whole chromosome, cancer-associated copy number aberrations are usually subchromosomal and highly variable across different patients. Taking into account the low concentration of ctDNA in early cancer patients, much higher sequencing depths of plasma DNA and more complex bioinformatics algorithms would be required if copy number aberration is used for cancer screening [9].

In addition to genetic changes, cancer-associated aberrant methylation can also be used as tumor markers. Similar to the detection of somatic mutations, the search for aberrant methylation can be performed on a single-target or genomewide level. The detection of methylated SEPT9 for colorectal cancer is one of the most extensively studied DNA methylation tumor markers [10]. As reviewed in a recent meta-analysis, its sensitivity for detecting stage I disease is only around 50% in all studies [11]. These figures further highlight the challenge of using a single-target for detection of early cancers and emphasize the need to increase the number of methylation targets. The observation that cancers are generally hypomethylated across the genome suggests the potential of genomewide methylation profiling of plasma DNA and identification of multiple methylation markers for cancer detection [12]. Further clinical potential of scanning a multitude of cancer-specific methylation markers was also illustrated in a few case-control studies with promising results [13,14].

While the analysis of the fore-mentioned cancer-associated changes could potentially be used as a generic tumor marker for the detection of a wide range of cancers, the lack of tissue specificity of these markers could create uncertainties on the appropriate follow-up investigations in the screening context. In this regard, it has been shown that through the profiling of the methylome of plasma DNA and comparison to tissue-specific methylation patterns, the proportional contribution of different tissues to plasma DNA can be determined [15]. In cancer patients with observable copy number aberrations in plasma DNA, the differential contribution of DNA from amplified and deleted regions can be used to identify the potential source of the tumoral DNA.

In addition to exploring the clinical applications of ctDNA, increasing efforts have been paid to investigate its biology. The fragmentation pattern of ctDNA is one of the intensely studied areas. Plasma DNA exhibits a characteristic nucleosomal size distribution pattern with a most dominant peak at 166 bp [16]. Remarkably, the size distribution of plasma DNA derived from cancer cells is shorter than those derived from non-malignant cells [17]. This better understanding of the physical characteristics of circulating DNA has led to new diagnostic utilities. For example, in the prospective NPC screening study, distinct size profiles of EBV DNA could be observed in the NPC patients and non-NPC subjects with false-positive results. Through size analysis, the false-positive rate could be reduced by 80% [18]. The tissue-specific DNA fragmentation pattern has also been shown to be useful for inferring the origin of plasma DNA [19]. An increased contribution of plasma DNA with nucleosome footprint corresponding to the affected tissue could be observed in some cancer patients [19].

Plasma DNA analysis has shown promise for early cancer detection. Over the past few years, we have witnessed important progress in our understanding of the plasma DNA biology which helped uncover cancer-specific signatures. Also, technology advancement has been translated into improvement in assay performance. The NPC screening study with plasma EBV DNA has shed light on future prospective studies using plasma DNA analysis for screening of other non-viral-associated cancer types in pre-symptomatic individuals [20]. We look forward to these studies which will evaluate its clinical role for early detection in different cancer types.

Declaration of interest

K.C.A.C. holds equity and receives research funding from Grail/Cirina. K.C.A.C. served as a director and held equity of Xcelom. K.C.A.C. serves as a director and holds equity of DRA. K.C.A.C. and W.K.J.L. are consultants to Grail. K.C.A.C. receives royalties from Illumina, Sequenom, Grail/Cirina, Xcelom and DRA. K.C.A.C. and W.K.J.L. have patents/patent applications in the area of molecular diagnostics using circulating nucleic acids. K.C.A.C. received speaker honorarium from BioRad and Astra Zeneca. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewers disclosure

A reviewer on this manuscript has disclosed that they provide independent technical consulting for the company Freenome (San Fran., USA). The peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Additional information

Funding

This work was supported by the Research Grants Council of the Hong Kong SAR Government under the Theme-based research scheme (T12–401/16–W).