Early detection and monitoring of cancer in liquid biopsy: advances and challenges

KEYWORDS:

• Cancer monitoring
• circulating tumor DNA
• early detection
• liquid biopsy
• next-generation sequencing

1. Introduction

Cancer is a serious health issue, which is one of the leading causes of death in this century. Cancer at the early stage is more likely to be curable, therefore early detection is an effective way of decreasing cancer death. For example, annual screening with fecal occult blood tests in a high-risk population over 50 years old aims to achieve the early detection of colorectal cancer (CRC), which substantially decreases CRC mortality by 16% [1]. The early detection of pancreatic cancer is even more urgent. As the main pathological type of pancreatic cancer, the 5-year survival rate of patients with pancreatic ductal adenocarcinoma is less than 5% because it is rarely diagnosed at early stage [2,3]. On the other hand, cancer monitoring is also essential for the development of adjuvant treatment after surgery and the evaluation of minimal residual disease. However, the repeated sampling of tissue specimens for histological detection is difficult to be achieved in clinical practice. Liquid biopsies provide an alternative non-invasive method for the detection and monitoring of cancer.

2. Next-generation sequencing in circulating marker detection

An effective circulating marker should be highly sensitive and specific. Next-generation sequencing (NGS) depicts a comprehensive profile for circulating markers. The most commonly used NGS strategies for early cancer detection and monitoring are ultra-deep sequencing for hotspot mutations in cancer driver genes [4] and bisulfite sequencing for methylation haplotype blocks (MHBs) [5].

2.1. Deep sequencing in circulating tumor DNA (ctDNA)

ctDNA is a tumor-derived marker that reflects cancer genetics and tumor burden. Ultra-deep sequencing for ctDNA detects the hotspot mutations or targeted regions of interest based on results from either whole-genome sequencing or whole-exome sequencing, which can probably detect low-frequency mutations. In a study for hepatocellular carcinoma, researchers used targeted sequencing with about 5,500× in coverage successfully found the driver mutations with allele frequency less than 0.1% in plasma samples, and these low-frequency mutations were confirmed by droplet digital PCR (ddPCR) [6]. Alternatively, Beads, Emulsions, Amplification, and Magnetics (BEAMing) can play the same role as ultra-deep sequencing in the low-frequency somatic mutations detection in liquid biopsy as multiplexed analysis, and the detection of RAS gene mutations based on BEAMing for CRC has been approved as a commercial kit for in vitro diagnostics [7]. However, there are some obstacles to apply ctDNA in cancer detection and monitoring. The variable fraction of ctDNA is difficult to be accurately detected under the high background of circulating DNA from non-tumor tissue or blood cells. The ratio of ctDNA in circulating DNA varies from 0.01% to 93%, so more sensitive and comprehensive detections are critical for profiling ctDNA markers [8].

2.2. Methylation haplotype in cancer detection and monitoring

MHBs method takes into account the co-methylation pattern of adjacent CpG sites as a genomic feature, and it has the potential for the detection of low-frequency alleles through established algorithms to identify and eliminate technical errors. For whole-genome bisulfite sequencing, co-methylation patterns not only increase the resolution of methylation positions at low sequencing coverage, eg. 5x coverage [9], but also make it possible to trace the original tumor of ctDNA shedding based on tissue-specific methylation patterns. Even using 1x to 3x coverage sequencing data, the early stage liver cancer could be detected based on liver-cancer-specific methylation haplotypes of ctDNA with 100% specificity and 94.8% sensitivity [10]. Moreover, a study identified about 2% tissue-specific blocks in all of 147,888 MHBs after feature selection, which demonstrated the potential of Pan-Cancer detection using cancer-specific methylation haplotypes in ctDNA [5].

2.3. Challenges of circulating marker detection in liquid biopsy

The goal of NGS in non-invasive early cancer detection and monitoring is to identify as many meaningful signals as possible from background or noise by optimizing the experimental design and analysis pipeline, while reasonable filtering criteria of mutations help improve the signal-to-noise ratio and low-frequency mutation detection. In the detection of low-frequency ctDNA mutations, one of the important steps is to exclude non-germline mutations found in white blood cells, which is useful to eliminate noise caused by clonal hematopoiesis [11]. Besides, the impact of different experimental operations on repeatability of NGS results is another concern. Different processes for the collection and preparation of liquid biopsies will affect circulating marker detection, and standardized experimental procedures are necessary [12].

3. Multi-omics strategy facilitates Pan-Cancer early detection and monitoring

3.1. The application of multi-omics strategy

Somatic mutations and epigenetic variants of circulating DNA are two main types of markers for cancer early detection and monitoring. Except for ctDNA, protein is another important circulating marker of cancers. A combination of KRAS mutations and protein markers in plasma samples had 64% sensitivity for stage I and II pancreatic cancer detection with 99.5% specificity [13]. Meanwhile, CancerSEEK is a test combining the detection of driver gene mutations and circulating proteins, which had 69% to 98% sensitivity in cancer detection for different types of cancer, and the specificity of detection was over 99% [4]. This is definitely a successful example of multi-omics applications in non-invasive cancer detection. Additionally, researches on circulating RNA, such as mRNA, miRNA, lncRNA and so forth, in early cancer detection and monitoring are increasing. In particular, plasma miRNA is an effective tool for post-surgery metastatic prediction in specific types of cancer [14]. Multi-dimensional modules for combining DNA methylation, gene expression and transcriptional regulation by miRNA have been reported in the previous study for ovarian cancer detection, which demonstrated a comprehensive cancer detection based on both epigenetic and transcriptional profiles [15]. To sum up, a single type of circulating marker is hard to fully reflect details of tumorigenesis and cancer treatment, therefore multi-omics detection is the trend in this field.

For cancer detection, many technologies such as NGS and microarray are high-throughput detections. NGS is not only useful for quantification, but also for finding novel alternations and transcripts. While PCR-based technologies, real-time quantitative PCR and ddPCR provide sensitive and robust detection for circulating nucleic acids. However, taking the throughput into account, PCR-based technologies are mainly used to validate markers selected by NGS data or to apply in the detection of small groups with 4 to 6 markers. Additionally, protein assays reflect proteomics information, which plays an irreplaceable role in early cancer detection [4]. Recently, some novel technologies have emerged in circulating cancer marker detection, such as electric field-induced release and measurement (EFIRM) [16]. For EFIRM, hybridized molecules of ctDNA templates, precoated capture probes, and detector probes generate current signals under specific reaction with chemicals and electrical field effects, and this assay detected two EGFR mutations in plasma samples from stage I and II lung cancer patients with 92% and 77% sensitivity and over 90% specificity.

3.2. The promising Pan-Cancer detection

ctDNA has potential application in Pan-Cancer detection. A small panel of driver genes about 81 kb in region size was successfully used to detect 4 types of common cancers including colorectal, breast, lung, and ovarian cancer, and somatic mutations were detected in about 60% plasma samples from stage I or stage II cancer patients [17]. Meanwhile, the promoter methylation levels of APC, FOXA1, and RASSF1A are useful for Pan-Cancer early detection with over 70% sensitivity and specificity [18]. Notably, different methylation levels of ctDNA are also an indicator to discriminate pancreatic cancer from pancreatic diseases, eg. pancreatitis [19]. Besides, different cohort studies confirmed the diagnostic and prognostic value of circulating miRNA that has a potential to be Pan-Cancerous marker for non-invasive cancer detection [20]. Tumor-educated platelet (TEP) is another interesting topic in Pan-Cancer detection. Combining with RNA sequencing, TEP showed reliable accuracy for cancer detection and primary tumor identification [21].

The foundation of Pan-Cancer assays for different molecules is similar. On the one hand, a group of driver gene mutations or cancer-specific methylation patterns is used as ctDNA markers for cancer detection and monitoring that could be typical Pan-Cancer assays. On the other hand, RNA molecules are universally expressed in different types of cancer but not in normal tissues is promising for Pan-Cancer detection. Additionally, each cancer type has specific cancer type-associated molecular profile, and they can be integrated into a panel of markers for Pan-Cancer detection. Pan-Cancer assays using liquid biopsy are useful for screening cancer patients in populations and are suitable for regular examinations to achieve cancer early detection.

4. Perspectives

Liquid biopsy enables repetitive cancer detection and dynamic monitoring. However, minute amounts of circulating tumor markers and high background signals released by normal tissues or blood cells are challenges we need to overcome. Careful choice of somatic cancer mutations from background signals is one of the key issues in marker selection. It was reported that JAK2, TP53, and KRAS mutations were detected in peripheral blood cells from non-small cell lung cancer patients [22]. This phenomenon once again reminds us of the impact of clonal hematopoiesis on driver gene mutation detection, and filtering mutations found in positive controls (e.g. white blood cells from the patient) from candidate driver gene mutations is a viable way to eliminate the background. Besides, comparing driver gene mutations found in your sequencing data with hotspot somatic mutations in COSMIC database, selecting overlapped mutations as markers probably enhance the reliability of cancer detection and monitoring. Additionally, it is noteworthy that alternations of circulating markers reflect the physiological changes of tumor in many aspects. Not only tumor burden or tumor size, the tumor metabolism under chemotherapy and prognosis are also closely correlated with circulating markers [23,24].

Tumor mutation burden (TMB) and circulating tumor cell (CTC) were expected to assist cancer immunotherapy treatment prediction and cancer progression monitoring. TMB describes the mutation level of tumor cells, and usually, it was measured by whole-exome sequencing. In sequencing data, oncogenic drivers and germline alternations could be excluded from TMB measurement, and the number of somatic mutations per megabase (mutations/Mb) produces a normalized form of TMB that can be compared between samples [25]. Not limited to tissue samples, a recent study reported that TMB can predict the benefits of cancer patients under immunotherapy in plasma samples [26]. For CTC into clinical practice, the combination of CTC level and cancer imaging detection achieves early response assessment for chemotherapy in metastatic CRC patients, which provides a good prediction in progression-free survival for cancer patients [27]. Moreover, we are the first group to detect tissue-specific CTC in CRC patients [28,29]. Notably, in vitro CTC culture and detection is promising for precision medicine and cancer progression monitoring for individuals, which also helps us understand CTC from multi-omics perspective [30]. PD-L1 expressing CTC has been found in different types of cancer, e.g. breast cancer [31], head and neck cancer [32], and it may also be a circulating marker for predicting immunotherapy treatment. At the same time, whether and how PD-L1 expressing CTC involved in cancer metastasis and immune evasion is worthy of further study.

Finally, circulating markers in urine and saliva also demonstrate the potential for early cancer detection. The combination of methylated TWIST1 and NID2 was sufficient for detecting primary bladder cancer with over 90% sensitivity and specificity [33]. In the previous study, the marker panel of salivary RNA includes 3 mRNA and 2 miRNA markers were proved useful for gastric cancer detection [34]. Additionally, exosomal markers are another hot topic in non-invasive cancer detection. Specific types of exosome and cargoes of exosome are likely to be used for cancer detection. For example, Glypican-1 positive exosome was a tool for identifying the different stages of pancreatic cancer from benign diseases [35]. Exosome is an ideal resource for finding circulating cancer markers. It enriches molecules in plasma samples, such as miRNA [36].

In summary, early detection is important to reduce cancer mortality, and effective monitoring assists in tracking the relapse of cancer and guides decisions for cancer treatment. Liquid biopsy is useful in early cancer detection and monitoring because it makes regular examinations or cancer screening in populations more flexible. In a previous clinical trial, 3.9% of occult cancer patients were detected [37]. Based on current studies, although the accuracy of liquid biopsy for stage I and II cancer detection still needs to be improved [13,17], the indistinctive mutation rates of ctDNA in early and late stage cancers imply the feasibility of early cancer detection in liquid biopsy [24]. Additionally, liquid biopsy also facilitates cancer progression and treatment monitoring in follow-up visits. Cancer detection in liquid biopsy achieves repetitive sampling that is limited in histological examination, and it provides a more economical and practical non-invasive detection compared with current imaging based methods such as computed tomography [38]. Furthermore, liquid biopsy also helps in dissecting the tumor heterogeneity and may get a more comprehensive picture about the genetic and proteomic alternations of the tumor [39,40]. Even though there are still many challenges in this area (such as the trace amount of circulating markers impede accurate cancer detection), the careful selection of candidate markers, the development of better analytical modules and more sensitive molecular technologies may contribute to early cancer detection and monitoring.

Declaration of interest

The authors have no 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. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewers Disclosure

Peer reviewers on this manuscript have no relevant financial relationships or otherwise to disclose.

Additional information

Funding

This paper was not funded.