Overview of the role of liquid biopsy in cancer management

Highlights

• •LB is a solid, non-invasive tool in precision oncology.
• •In localized cancer, LB can serve as a tool to detect minimal residual disease.
• •In metastatic setting, LB can detect early tumor response and resistance mechanisms.

Abstract

With the emergence of novel targeted therapeutic options in early-stage and advanced-stage malignancies, researchers have shifted their focus on developing personalized treatment plans through molecular profiling. Circulating tumor DNA (ctDNA) is a cell-free DNA (ctDNA) fragment, originating from tumor cells, and circulating in the bloodstream as well as biological fluids. Over the past decade, many techniques were developed for liquid biopsies through next-generation sequencing. This alternative non-invasive biopsy offers several advantages in various types of tumors over traditional tissue biopsy. The process of liquid biopsy is considered minimally invasive and therefore easily repeatable when needed, providing a more dynamic analysis of the tumor cells. Moreover, it has an advantage in patients with tumors that are not candidates for tissue sampling. Besides, it offers a deeper understanding of tumor burden as well as treatment response, thereby enhancing the detection of minimal residual disease and therapeutic guidance for personalized medicine. Despite its many advantages, ctDNA and liquid biopsy do have some limitations.

This paper discusses the basis of ctDNA and the current data available on the subject, as well as its clinical utility. We also reflect on the limitations of using ctDNA in addition to its future perspectives in clinical oncology and precision medicine.

Introduction

Historically, Mandel and Metais were the first to describe in 1948 the presence of DNA in the blood circulation of humans outside the cells, which they called cell-free DNA (ctDNA) [1]. Nearly 30 years after, Leon et al. found significantly higher levels of this type of DNA in the serum of metastatic cancer patients, which decrease after response to treatment [2]. Then in 1989, Stroun et al. [3] reported that a fraction of ctDNA is derived from cancer cells, thus it is called circulating tumor DNA (ctDNA). ctDNA is a short fragment (100 bp) mainly released from cells located in the tumor bed [4,5]. It is believed that tumor-associated macrophages play an important role in the digestion and fragmentation of tumor-cell-DNA into smaller fragments followed by their release in the bloodstream [6].

With the successful sequencing of the human genome followed by the emergence of new sequencing technologies, the detection and sequencing of ctDNA became readily available in the real-life setting emerging as a promising non-invasive technique frequently used in modern oncology.

Pre-analytical phase

The pre-analytical phase includes blood collection, sample processing, and DNA isolation. Each of these steps may affect the final ctDNA stability highlighting the need for standardized procedures. Plasma is more suitable than serum since the latter contains much more ctDNA originating from non-malignant cells [7,8]. Thus during blood collection, it is crucial to use tubes that contain anti-coagulants that do not interfere with the stability of ctDNA. Heparin is contraindicated since it does not restrain the activity of ctDNA-degrading endonuclease and could potentially inhibit polymerase chain reaction (PCR) which is sometimes used to amplify ctDNA [9].

Special collection tubes have been developed containing fixative agents for leukocyte stabilization and ctDNA integrity preservation, like the PAXgene™ tubes (Qiagen, Germany), cell-free DNA™ blood collection tubes (Streck Inc., Omaha, USA), and ctDNA collection tubes (Roche Diagnostics, Germany). However, EDTA-coated tubes could also be used since they confer good stability of ctDNA, but plasma isolation should be performed within the day of collection to limit the release of DNA derived from leukocytes [10].

What could be evaluated using ctDNA?

Different sequencing methods have been developed ranging from the sequencing of a single gene analysis, a gene panel, the whole exome (WES), and even the whole genome (WGS) [11], [12], [13], [14].

Thus, several molecular aberrations could be detected depending on the sequencing method. Each of these aberrations has its clinical indication. We will not report everything in this review, but we will give examples of important alterations that have been validated in clinical practice including point mutations, fusions, copy number variations (CNV), and tumor fraction quantification.

Nearly all targetable mutations could be detected by ctDNA analysis with high sensitivity and specificity like EGFR mutations [15], BRAF [16], KRAS [17], BRCA 1/2 [18], IDH1 [19], PIK3CA [20], as well as oncogenic fusions like ALK, MET, NTRK, RET, ROS1 [21]. On another hand, ctDNA analysis is also an important tool to detect CNVs like amplifications or deletions [22]. HER2 amplification in breast cancer is an important example [20]. This type of targetable alteration has been usually detected through immunohistochemistry (IHC) or fluorescent in situ hybridization (FISH) which are performed on tissue biopsy. Due to tumor heterogeneity, several patients could be classified as negative for HER2 amplification in a tissue biopsy performed on one tumor site but could have HER2-positive cells at other sites. The use of liquid biopsy could help define these patients that would eventually benefit from anti-HER2 therapy [23]. This could also be extrapolated to other types of cancers [23,24], as well as other types of gene-CNV [25,26]. As for tumor fraction, it is the percentage of ctDNA among ctDNA and could be used as a marker of tumor burden. Usually, the more the cancer is advanced the allele fraction is elevated [11,27]. Its dynamics could orient clinical decisions through the detection of response or resistance even before biological and imaging tools [28]. Other types of alterations which we would not develop in this review include among others, microsatellite Instability (MSI), methylation signatures, epigenetic biomarkers, and fragment omics [29], [30], [31].

ctDNA in clincal practice

In localized cancer

One of the essential purposes for ctDNA in plasma is the identification of minimal residual disease (MRD) in cancer patients at risk of disease relapse after the curative approach. Several papers have addressed this issue to detect the feasibility of such an innovative approach to limit, on one hand, the unnecessary administration of adjuvant therapy in those at low risk of disease relapse but also, to identify those mighty who benefit from such an approach. In several subtypes of solid tumors such as breast, lung, and colorectal cancers, ctDNA has been used to detect MRD using patient-specific techniques [32], [33], [34].

Non-small cell lung cancer (NSCLC)

Gale and colleagues reported their experience with eighty-eight patients with early-stage non-small cell lung carcinoma (NSCLC) who underwent local therapy with curative intent, in whom ctDNA was extracted at baseline (before therapy) then in the postoperative setting with longer follow-up using a patient-specific assay in concordance with the tumor biopsy specimen. NSCLC patients are offered adjuvant therapy (chemotherapy, radiation therapy, or immunotherapy), based on the classical clinical, pathological, and radiological disease staging without accurate patient selection using novel biomarkers (NCCN guidelines). For instance, NSCLC who underwent concurrent chemo-radiotherapy and had detectable ctDNA, had higher outcomes with consolidation immune checkpoint inhibitors while those with the undetectable residual disease had no significant difference whether exposed to immunotherapy or not [35]. In this paper by Gale et al., ctDNA was more accurately detected at advanced stages (87% vs 24% in stages 1 and 2 respectively). Interestingly, ctDNA detection preceded clinical detection of disease recurrence by 212.5 days. Overall, of those who exhibited disease recurrence, 64.3% of patients had detectable ctDNA in plasma samples. Moreover, while early ctDNA detection within 1–3 days after local therapy (25%) was not significantly associated with disease relapse, those who had detectable fraction beyond 0.5–4 months in the postoperative setting (17% of patients) had shorter RFS (recurrence-free survival) and OS (Hazard ratio (HR)=14.8 and 5.48 respectively). Baseline ctDNA detection was also associated with shorter RFS and OD (Hazard ratio (HR) =2.97 and 3.14 respectively). Therefore, the potential detection of ctDNA after a curative therapeutic approach might guide future decision-makers to select those who require additional strategies to improve survival outcomes [36]. Other important data were reported on 466 NSCLC patients from the Impower150 study, in which immune checkpoint inhibitors-chemotherapy combination were assessed, using machine learning models with the purpose of assessing the role of ctDNA in predicting OS through different time points. These simulations and models showed that ctDNA models can predict response to therapy as well as identify high-risk groups among NSCLC patients, thus outperforming classical radiological imaging and allowing earlier differentiation between competing arms within clinical trials [37].

Colon cancer

Tie and colleagues have evaluated the role of ctDNA molecular analysis in defining the patients that can benefit from adjuvant chemotherapy in stage II colon cancer in the DYNAMIC trial. The decision was to administer adjuvant chemotherapy (5-fluoropyrimidine or oxaliplatin-based chemotherapy) to patients in whom positive detection of ctDNA is confirmed on the 4th or 7th week in the postoperative setting while those with negative findings on both tests were offered surveillance only. This was a phase 2, multicentric, 2:1 randomized controlled trial of 455 stage II colon cancer in which 302 patients were selected in the ctDNA-guided while 153 were offered a standard approach with the primary endpoint of recurrence-free survival at 2 years. Interestingly, patients in the experimental arm received less adjuvant chemotherapy (15% vs. 28%; RR=1.82 (95% CI: 1.25 to 2.65). The most significant difference was among T4 tumors (70% vs. 27% in the ctDNA-guided group) and poorly differentiated patients (24% vs. 5%). In the RFS at 2 years, this innovative approach was non-inferior to the standard approach (93.5% and 92.4% respectively, non-inferiority margin of −8.5% points). RFS at 3 years reached 86.4% in ct-DNA positive patients who received chemotherapy while it reached 92.5% in those who were exempted from chemotherapy due to negative ctDNA results. Among those who were ctDNA positive, 3-year RFS was 92.6% and 76% in oxaliplatin-based and single-agent fluoropyrimidine, respectively. Therefore, this approach led to the reduced administration of chemotherapy in a low-risk population without affecting the risk of disease relapse [38].

The same authors have previously used the same molecular approach through serial ctDNA analysis to assess the value of adjuvant chemotherapy in a multicentric phase 2 trial, which included 94 patients with stage III colon cancer. In stage III colon cancer, the addition of postoperative chemotherapy improved overall survival and is now the standard of care (Andre et al.). Among those patients with detectable postoperative ctDNA levels (21%, 4 to 10 weeks post-surgery), RFS was significantly inferior to those without detectable levels (3-year RFS: 47% vs. 76%; HR =3.8; p<0.001). Also, patients with detectable levels after adjuvant chemotherapy (17%, within 6 weeks) had a significantly lower 3-year RFS (30% vs. 77%; HR=6.8, p<0.001). After multiparametric adjustment, postoperative ctDNA levels were the only factor that remained associated with shorter RFS. Therefore, the postsurgical ctDNA value can be considered a valuable prognostic factor while the post-chemotherapy ctDNA value has defined a unique population at a higher risk of recurrence that might need further preventive therapy [39].

In NSCLC patients receiving concomitant chemoradiation, those who had undetectable levels of ctDNA had excellent outcomes regardless of ICI administration as a consolidation treatment. Nevertheless, in those with MRD, administration of ICI led to better outcomes in comparison to those who didn't thus suggesting a potential role for ctDNA analysis in personalizing treatment strategy in locally advanced NSCLC [43].

Breast cancer

In early breast cancer, identification of specific cancer mutations through TARDIS (personalized targeted digital sequencing) has suggested that after neoadjuvant chemotherapy, the detection of ctDNA is a potential biomarker for MRD detection [40]. Moss et al. also assessed the presence of breast-derived ctDNA in 235 breast cancer samples thus concluding that high levels are associated with worse prognosis and aggressive behavior but also, that their presence near the end of neoadjuvant chemotherapy can reflect the presence of residual disease [41]. Therefore, these innovative techniques represent a promising approach for the detection of driver somatic mutations but also the low levels of ctDNA which could be challenging in the early stages of breast cancer [42].

Ovarian cancer

Several studies have shown that detection of ctDNA in the postoperative setting can identify earlier disease relapse risk in comparison to radiological imaging in lung, colorectal, prostate, and breast cancer [44], [45], [46]. In ovarian cancer, the detection of TP53 gene variants in ctDNA after completion of neoadjuvant chemotherapy can also serve as a tool to determine MRD in a more sensitive way than the conventional biomarker, CA125 [47,48].

Nevertheless, with the paucity of available data and the absence of clinical utility in altering the course of treatment, the current ESMO (European society of medical oncology) does not support nor recommend the use of liquid biopsy in the detection of MRD [49].

In summary, in localized tumors across different cancer types, the detection of ctDNA levels can serve as a tool to detect MRD levels which can guide the need for treatment intensification or otherwise, avoid unnecessary adjuvant approaches in patients with low risk of disease relapse.

In advanced/metastatic cancer

Metastatic cancer progression is a complex and evolving process marked by the emergence of different subclones, which contributes to tumor genomic heterogeneity [50]. Tissue-based genomic profiling, even though of extreme importance in precision oncology, is unable to capture spatial and temporal heterogeneity, thus potentially missing important targetable drivers [51,52]. On the other hand, liquid biopsy is a noninvasive and easily repeatable method that allows the exploration of the spatial and temporal evolution of the cancer genome, as shown in different studies which compared it to tissue-based analysis [51,53,54].

Monitoring early tumor response

Liquid biopsy could also be used to monitor early tumor response. For instance, Dawson et al. reported data taken from 30 women with breast cancer, followed by radiological imaging, ca 15–3, and ctDNA. The authors showed that ctDNA is a useful, specific, and extremely sensitive biomarker of metastatic breast cancer, even superior to other biomarkers [55]. Moreover, Page et al. reported 9 cases of patients with metastatic breast cancer, in whom ctDNA variations were prognostic, valuable in detecting new mutations that could alter the treatment plan, as well as useful in assessing treatment response as they tracked the disease status whereas ca 15–3 did not match the variation [56]. Another illustration of the use of liquid biopsies for follow-up is presented in the data collected from serial blood samples from the TOPARP-A trial reported by Goodall et al. where a decrease in the level of CtDNA after 4 weeks and later after 8 weeks, correlated respectively with longer progression-free survival and longer overall survival [57].

Detection of resistance patterns

With newer specific targeted therapy emerges new resistant mutations and thus appears the need to discover and try to target them. T790M mutation is the leading example here, as 50% of patients with mNSCLC, EGFR mutant disease, acquire this mutation after the use of first-line TKI, and thus third generation TKI can be used instead of switching to chemotherapy with favorable response rate and PFS [58,59]. In the TOPARP-A trial, the clinical value of liquid biopsies in detecting genetic composition, primary mutations, and emerging subclones is shown, where in blood samples from patients with HRD or LOH mutations, who responded to olaparib, the allele frequency dropped to less than 5%, whereas it remained stable in the patient who did not respond. In addition, in patients where no response was seen, two ATM frameshift mutations were identified post-treatment [57].

Moreover, Misale et al. discovered the detection of KRAS mutant alleles in the plasma of patients treated with cetuximab as early as 10 months before radiological disease progression [60]. Another example, in mCRC (metastatic colorectal cancer), is that 38% of patients treated with EGFR blockade, convert from RAS wild type to RAS mutant, and others acquire PIK3CA and MET amplification, rendering the disease treatment-resistant. This point proved the importance of early detection of any resistance to establish a proper therapeutic strategy/approach [58,61,62]. Moving on to breast cancer, one example of mutations leading to resistant disease is estrogen receptor 1 (ESR1) which correlates with resistance to aromatase inhibitors (AIs). These tumors require the use of a different and more efficient treatment as described in PADA-1 trial, where patients with new or rising ESR-1 level who were switched to Fulvestrant, had a doubled PFS compared to those who continued same treatment with AI and CDK4/6 [63].

In addition to that, Gray et al. demonstrated that in metastatic melanoma, resistant genomic alterations to BRAF and MEK inhibitors can be detected before any evident clinical disease progression. As well as quantifying plasma, ctDNA can track the response to treatment and it precedes any radiological disease progression [64]. Wyatt et al. studied plasma samples from 65 mCRPC (metastatic castrate-resistant prostate cancer) patients treated with enzalutamide and described possible mechanisms for primary resistance including RB1 loss, AR mutations, or amplification as well as acquired resistance [65].

In summary, two main roles for liquid biopsy include the monitoring of early tumor response as well as the detection of resistance mechanisms, which can guide the therapeutic strategies in the advanced setting in numerous malignant tumors.

Limitation of ctDNA analysis

Accurate interpretation of ctDNA analysis remains a challenge for scientists as numerous and variable genomic mutations might be found, with no actual clinical correlation or benefit [51,66].

The most common finding, yet a misleading one is Clonal hematopoiesis (CH), as it is a normal cellular aging process but also a risk factor for cardiovascular diseases as well as hematologic malignancies [67], [68], [69]. Various studies have been conducted to evaluate the clinical impact of the detection of CH, [70], [71], [72], [73], [74], [75] but until today no clear clinical implication was validated. Data has shown that CH-related mutations can be detected in both the tumor and plasma, which can cause misinterpretation of data. Therefore, to avoid misinterpretation of liquid biopsy data, it is advised to integrate mutational information derived from peripheral blood cells, to discern those CH-derived alterations from tumor tissue [76]. Further studies are essential for accurate and cautious interpretation to differentiate between different CHIP (clonal hematopoiesis of indeterminate potential) detected.

Another example of randomly discovered mutations is illustrated in a Chinese randomized trial where around 6200 patients with mNSCLC were tested, and 1% of them were found to have gBRCAm. This study showed a positive correlation between gBRCA mutation and early onset of NSCLC [77]. But how do we apply this in the real clinical setup? Moreover, do patients with gBRCA mutation need to be screened for lung cancer? Questions that remain unanswered. Slavin et al. reported data from an observational case series on more than 50 cancer types, with more than 10,000 patients, concerning the prevalence of incidental germline mutations in patients with advanced solid tumors, and as expected several mutations were found without clear guidelines for counseling, screening or possible treatment guidance [78].

Several limitations of ctDNA analysis are still major obstacles in real-life practice. First, the sensitivity of ctDNA analysis, which can result in a high rate of false negative results and its specificity, because it can be affected by the release of ctDNA from non-cancer cells. Second, the variability of ctDNA release from tumor cells can significantly affect the outcomes. Third, financial and ethical limitations can also affect the availability of this innovative approach in the diagnostic and therapeutic approaches. Lastly, the interpretation of ctDNA results can be challenging, as not all mutations are clinically relevant [49,76,79].

Future perspectives

ctDNA in non-plasma body fluids

Over the last few years, blood-derived liquid biopsy has emerged as a potential alternative to tumor tissue samples in various indications. Nevertheless, many other body sources than blood are being evaluated to replace or complement this approach.

Detection of ctDNA in the saliva is an alternative mean mainly in patients with head and neck tumors. This approach has several advantages: a non-invasive screening tool, real-time monitoring of HNC patients as well as the ideal site to reflect the molecular, genomic, and pathological alterations in the head and neck region. Nevertheless, the presence of numerous non-tumorigenic components within the milieu as well as the presence of different contributors in the saliva renders the analysis and identification of ctDNA extremely challenging [80]. This source of ctDNA in the saliva can offer an option for MRD detection in oral cancers following chemoradiation.

Urine can also serve as an exciting source of ctDNA from urological cancers which is divided into 2 types: trtDNA (transrenal tumor DNA), originating from plasma, and the second from tumor shedding directly in the urine. This is a non-invasive technique without the need for healthcare professionals. However, there are several hurdles including the dilutional effect on DNA within large volumes of urine but also, preservation and transportation maneuvers with the short half-life of unpreserved DNA [81,82]. Several trials have assessed the role of trtDNA as a molecular alteration detection tool such as EGFR in NSCLC but also served as a tool for monitoring disease in early breast cancer patients following surgery and hepatocellular carcinoma after tumor ablation [83,84]. On the other hand, urinary cell-free DNA, directly from the prostate, bladder, and renal carcinoma can be used to monitor disease relapse. Zhang et al. found that in urine samples, there is a higher identification of urothelial carcinoma-associated genomic alterations in comparison to plasma with better concordance to tumor samples but also the ease of collection and storage of urine samples in preservative containers [85]. UROSEEK is a specific multiplex PCR-based assay, that has been developed for the early detection of urothelial carcinoma, and can have a sensitivity of 95% when combined with urinary cytology [86].

Malignant Pleural and peritoneal fluids, involving tumor cells infiltrating their linings, can be enriched with non-hematopoietic ctDNA cells even if the cytology was negative [87]. The advantage is that samples are easily acquired through the pleural tap with data demonstrating that more driver alterations were detected in pleural fluid in comparison to plasma but also, these involve invasive procedures [82].

Bile has been suggested to be a source of DNA sequencing in biliary tract tumors with bile ctDNA consisting of long fragments of high concordance with the molecular features detected in tissue samples [88].

Plasma ctDNA is rarely detected in primary CNS tumors or isolated brain metastasis [89]. In the CSF, ctDNA is much more abundant than in plasma in glioma patients or those with meningeal involvement. Despite their potential role as a diagnostic tool for those with equivocal histology or for monitoring of response or predicting disease relapse in CNS tumors, their implementation in clinical practice can be challenging due to the invasive nature of acquirement but also the small sample sizes which may limit the ctDNA analysis [82].

Other challenging sources of ctDNA are currently being investigated such as the seminal fluid, sputum, stools, and uterine lavage.

Conclusion

We showed in this review that ctDNA analysis is a solid non-invasive tool in precision and personalized oncology in both the adjuvant and the metastatic settings. Moreover, with their established role as a non-invasive easily repeatable mean to detect susceptibility and resistance alterations, there is an increased interest in the study of ctDNA kinetics which means the study of quantitative changes over time of ctDNA levels. ctDNA kinetics study, despite its complexity, could serve as a potential tool to adapt therapeutic strategies over time but also could become an efficacious indicator of therapeutic efficacy [90]. We have also reviewed the main technical issues related to its use as well as its limitations.

CRediT authorship contribution statement

Tarek Assi: Writing – original draft, Conceptualization. Rita Khoury: Writing – original draft. Rebecca Ibrahim: Writing – review & editing. Maria Baz: Writing – original draft, Visualization. Tony Ibrahim: Writing – original draft, Conceptualization. Axel LE Cesne: Writing – original draft, Supervision.

Declaration of Competing Interest

None

The authors confirm that they have nothing to declare.