Cell-free DNA as a biomarker in cancer

Translational research of liquid biopsy is just at the edge of routine clinical application: an emerging validity of circulating tumor DNA (ctDNA) tests suggests its use for earlier cancer detection and better monitoring of minimal residual disease (MRD) and resistance development, thus offering earlier guidance for therapy choices with the intent to cure cancer. In this review, we focus on ctDNA as an advanced and standardized validated marker in liquid biopsy. We also discuss what will be needed to reach the new milestone of personalized (precision) medicine to be used as a common standard of care. We summarize recent developments of cell-free DNA (cfDNA) and its clinical use as a biomarker in cancer. . This study is an ongoing, prospective, observational, cohort study with two major aims: (1) to validate the collaborating company’s (GRAIL, LLC) blood test on cell-free nucleic acids (cfNAs) for early detection of multiple types of cancer; and (2) to deliver low-dose CT (LDCT) screening for lung cancer. Participants will receive at least one low-dose chest CT at baseline. Another possible scan after 12 months will be randomized. The study aims to keep most scans below 1 mSv radiation dose (ultra-low dose) and all scans under 2 mSv. The liquid biopsy will evaluate the test performance, including sensitivity, specificity, and tissue of origin of detected cancers. The performance of the LDCT screening to detect lung cancers will be evaluated in comparison to established measures of risk prediction of lung cancers and other incidental findings. The study started in April 2019 with an estimated primary completion date of August 2023 and an estimated completion date of August 2030 [87] .


INTRODUCTION
Key elements of clinical routine with tumor patients include cancer detection and diagnosis as well as monitoring tumor development and treatment response. Usually, an interdisciplinary approach is necessary with complementing areas, such as medical imaging, tissue biopsy, histopathology, surgery, chemotherapy, and radiotherapy. Over decades, measurements of protein tumor markers in the blood evolved to the current standard, but translational research offers improvement: circulating cell-free DNA (cfDNA) allows easy and early access to information on tumor evolution and treatment response at low risk of injury,  [3] CTC Microscopy, case report Skin metastasis of unknown primary, "liquid autopsy" First report on tumor cells in blood; post mortem; microscopically identical cells in metastatic lesions 1889 Paget [4] CTC Autopsy Breast cancer, postulated "Seed and soil" theory of cancer metastasis 1948 Mandel and Métais [5] cfDNA Blood analysis Not related to cancer, healthy blood donors First report of (cell-free) nucleic acids in blood 1975 Fidler [6] CTC Experimental metastasis assay B16 melanoma cell lines Only a small fraction of intravenously injected tumor cells give rise to metastasis in mouse models 1977 Leon et al. [7] cfDNA Radioimmunoassay for free DNA in serum Various cancers First report on increased cfDNA levels in some cancer patients; correlation with therapy response 1989 Lo et al. [8] cfDNA PCR of Y chromosome-specific sequence; cfDNA/blood Not related to cancer Sex determination of fetus in pregnant women 2001 Reya et al. [9] CTC Applying hematopoietic stem cell knowledge to heterogeneity of cancer cells

Solid tumors and leukemia, migratory CSC
Cancer stem cell theory 2003 Balaña et al. [10] ctDNA Methylation-specific PCR of MGMT, p16, DAPK, RASSF1A GBM Detection of methylated MGMT in serum highly predictive for response to BCNU chemotherapy 2004 Allard et al. [11] CTC CellSearch™ Prostate, breast, ovarian, CCR, lung, and other cancers Detection of CTCs in 7.5 mL of blood samples 2004 Cristofanilli et al. [12] CTC CellSearch™ Amount of CTC Metastatic breast cancer Independent predictive marker: reduced PFS and reduced OS 2005 Diehl et al. [13] ctDNA dPCR, BEAMing Advanced CRC APC mutations in plasma 2008 Maheswaran et al. [14] CTC Molecular profiling EGFR mutations NSCL Monitoring therapy 2008 Diehl et al. [15] ctDNA Mutations Colorectal cancer Amount and presence/absence as an independent tumor marker, monitoring 2008 Cohen et al. [16] CTC CellSearch Clinical study Colorectal cancer Enumerating CTC 2008 De Bono et al. [17] CTC Clinical study Prostate Cancer Enumerating, the first demonstration that CTCs are the most accurate and independent predictor of OS in metastatic prostate cancer 2010 Pantel and Alix-Panabières [18] CTC Concept of analyzing tumor cells in body fluids All cancers Coined the term "liquid biopsy" 2010 Calverly et al. [19] TEP Platelet mRNA Metastatic NSCL Downregulation of gene expression in platelets 2013 Dawson et al. [20] CTC + ctDNA (comparison) Side-by-side disease monitoring Metastatic breast cancer undergoing treatment Sensitivity: ctDNA>CTC ctDNA level correlates with treatment response 2014 Bettegowda et al. [21] ctDNA Digital PCR, sequencing 14 tumor types ctDNA detectable for most tumors outside the brain 2014 Sullivan et al. [22] CTC "Negative depletion" CTC-iChip (removing leukocytes from blood) GBM (usually not metastatic) Surprising and frequent detection of CTCs in brain tumors 2015 Mazel et al. [23] CTC CellSearch Breast cancer PD-L1 detection for immune checkpoint inhibition https://dx.doi.org/10.20517/evcna.2022.20 2016 Tie et al. [24] ctDNA Sequencing Colon cancer stage II Detection of MRD after surgery; prediction of recurrence after chemotherapy 2016 Donaldson and Park [25] ctDNA Clinical studies NSCLC First FDA [29] and EMA approval to use ctDNA for EGFR-targeted therapy 2018 Cohen et al. [26] ctDNA, plus proteins from blood CancerSEEK, detecting mutations in 1933 loci of 16 genes; combined with protein tumor markers 8 cancer types Blood screening test for several common cancers 2020 Lennon et al. [27] ctDNA, protein markers plus PET-CT curative treatments, as well as the use in monitoring after therapy and a better quality of life. The vision is a quick, reliable, and easy-to-repeat test for all patients to track tumor progression in real-time, as well as monitor the response to treatment to allow earlier detection of resistance, relapse, and metastasis and adapt clinical decisions accordingly. In addition, the presence of tumor markers after initially successful therapy should confirm minimal residual disease (MRD) in the absence of other markers and lead to additional treatments before the tumor becomes incurable [33] . The current standard for most solid cancers consists of a more invasive tissue biopsy and/or surgical removal of all or most parts of the tumor with a thorough microscopic analysis by a tumor pathologist. Genetic molecular profiling of the primary tumor tissue may be included. In 2016, a new WHO classification of tumors of the nervous system introduced molecular and genetic profiling as standard practice for the analysis of some brain tumors such as medulloblastomas and glioblastomas [34] . Depending on the tumor location and the multi-morbidity of a patient, obtaining tumor tissue repeatedly can be difficult and unsuitable for long-term monitoring of tumor development. Furthermore, transthoracic lung biopsies can lead to, although rare, iatrogenic dissemination of cancer cells [35] , and fine-needle biopsies have been reported to increase CTCs in prostate cancer [36] . The snapshot of a tissue biopsy from a primary tumor may also miss minor, but progressing, tumor clones; it may also not represent metastatic clones with perhaps different molecular profiles for treatment resistances [37] . Therefore, liquid biopsy carries the potential to be the next gold standard for monitoring cancer progression. Following various research approaches to the detection and cultivation of CTCs, also leading to a better understanding of tumor progression and metastasis, hundreds of emerging clinical studies are continuing to develop new standards for monitoring melanoma, breast, prostate, colon, and head and neck cancers [38][39][40][41][42][43][44][45][46] . In this review, we focus on another part of liquid biopsy, cfDNA, which is reliably much easier to access and does not need elaborate methodologies to capture the extremely rare CTCs.

LIQUID BIOPSY MARKERS
ctDNA and CTCs are currently more established than EVs and tumor-educated platelets (TEP) as tools in liquid biopsy [ Figures 1 and 2 and Table 1]. In this review, our main focus is on ctDNA and its clinical application and potential. For completion and comparison, we mention CTCs, EVs, and TEPs only briefly.  Created/modified with SMART [1] , licensed under Creative Commons Attribution 3.0 Unported License [2] .

Figure 2.
Hypothetical ctDNA biomarker levels during tumor development and therapy. After tumor removal, the biomarker level drops significantly and will remain low in the case the patient is cured. Early detection of minimal residual disease (MRD) by ctDNA allows additional treatment (e.g., chemo-, radio-, or hormone therapy) with the intent of cure, much earlier than with clinical imaging methods.

Cell-free DNA and circulating tumor DNA
The total amount of non-cellular DNA in samples such as blood is called cfDNA, which for tumor patients contains all tumor-derived DNA in various amounts, i.e. ctDNA.
cfDNA in blood is highly fragmented with a peak of 166 bp length and many smaller peaks of 10 bp less or more in healthy individuals. Interestingly, the fragment size of the tumor-derived ctDNA appears to be shifted to a lower peak at 146 bp, which allows for selection strategies to increase the relative amount of tumor-derived ctDNA and to increase the sensitivity. Most of the cfDNA is considered to come from leukocytes, such as granulocytes, lymphocytes, and monocytes, and to a lesser extend from other cells (muscle cells, endothelial cells, epithelial cells, fibroblasts, mesenchymal stem cells, and hepatocytes) [47] ; interestingly, physical activity can trigger cfDNA release almost exclusively from granulocytes (up to 2-20fold increase) and may affect liquid biopsy [47] . Tumor-derived ctDNA can come from apoptotic or necrotic parts of the primary tumor, as well as from metastases and CTCs. ctDNA often represents only an extremely small fraction of the total cfDNA, especially at early stages and after successful therapies. Therefore, detection of low concentrations of ctDNA remains the major challenge for widespread tumor screening and treatment monitoring. Several key technologies have been developed and applied to detect potential micrometastases and local relapse before clinical and radiological manifestation. This includes several ways of targeted sequencing of known mutations as well as non-targeted, i.e., genome-wide analysis, and fragmentomics [48] .

Targeted analysis
Targeted analysis aims to identify tumor-specific mutations or methylations in ctDNA. Techniques include digital PCR (dPCR), BEAMing (beads, emulsions, amplification, magnetics), safe-sequencing system (Safe-SeqS), cancer personalized profiling by deep sequencing (CAPP-Seq), and tagged-amplicon deep sequencing (Tam-Seq). ctDNA can also be analyzed for its relative amount from total cfDNA to serve as a surrogate for tumor burden. This allows risk assessment and staging, as well as early monitoring of therapy response or treatment failure. Tight monitoring can detect mutations as targets or escape mechanisms and supports clinical decision making based on tumor evolution and resistance development. Specific mutations in the EGFR gene in ctDNA allow treatment of NSCLC with a tyrosine kinase inhibitor (TKI) [29,49] . Activating mutations of phosphatidylinositol-3-kinase catalytic subunit alpha (PIK3CA) are used to guide treatment in breast cancer. Resistant tumor cells can escape from therapies by gaining new mutations and clonal selection of the fittest tumor cells. For example, KRAS mutations can develop as a resistance mechanism during EGFR-targeted therapy for patients with CRC [50,51] .
DNA methylation is an epigenetic mechanism to regulate gene expression by adding methyl groups to the DNA [52] . In contrast to normal tissue, tumors can show a reversed hyper-and hypomethylation pattern, which can be important for tumor development and progression, as well as for clinical management of patients. Since 2003, typical methylation patterns of tumor DNA have been used to identify ctDNA in blood samples, e.g. MGMT relevant for brain tumors as an actionable target [10,53,54] . More recently, hydroxymethylation profiling with detection of 5-hydroxymethylcytosine (5hmC) as a less known molecular marker of epigenetics was applied in liquid biopsy of lung, pancreatic, and hepatocellular cancer (HCC) [55] . Lung cancer was characterized by a loss of 5hmC, whereas HCC and pancreatic cancer showed disease-specific changes, thus providing information about tumor type and stage.

Non-targeted analysis
Without prior information on specific mutations, non-targeted approaches try to investigate the entire genome, e.g., with whole exome sequencing (WES), whole genome sequencing (WGS), detection of copy number aberrations (CNA), and others [56] . This also allows the detection of subclones evolving under treatment or during natural tumor progression. Unfortunately, low amounts of ctDNA from patients without relapse or metastasis affect sensitivity and utility. Some of these techniques can detect point mutations not previously found in the primary tumor, but with specific value options for treatment and prognosis [57] .

Fragmentomics
The size of tumor-derived cfDNA from plasma tends to be shorter than normal background cfDNA [58] . These differences in fragmentation of DNA have been associated with reduced levels of a secreted DNASE1like nuclease, DNASE1L3, in many tumor types (breast, colorectal, lung, gastric, head and neck nonsquamous cell, and liver cancers) [59,60] . Future studies will have to link and translate the biology of circulating DNA and nucleases to include fragment size, end motifs, and jagged (single-stranded) ends into clinical applications in tumor biology [48] .

Other types of liquid biopsy: CTCs, EVs, TEPs
Over the last two decades, CTCs entered clinical applications to detect and count them, as well as to determine their genetic profile for prognostics and clinical decision making [ Figure 1 and Table 1]. CTCs are usually rare, but they often correlate and contribute to metastatic progression, although only a fraction of CTCs gives rise to metastasis [6] . Despite major advantages of CTCs in diagnostics, prognostics, monitoring, and as guides to therapy choice and delivery schedule, widespread use appears to be still further away. Currently, only a few highly equipped labs supported by major research funds seem to be able to collect and analyze CTCs. A common standard is not established. With CellSearch, the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) approved the first liquid biopsy in 2004 [29] . Originally, CellSearch was approved as a diagnostic tool to detect and count CTCs in blood samples to predict outcome (PFS and OS) only in metastatic breast cancer, but it later was expanded to monitor metastatic breast, colorectal, and prostate cancer patients.
miRNA are noncoding, 20-24 nucleotides long RNA molecules derived from just 1% of the whole genome. They are involved in the regulation of stability and translation of mRNA in health and disease. The potential effects of up to 1900 miRNAs are not fully understood. They can be found up-or downregulated in serum, EVs, and CTCs, as well as in urine [61] . Future studies may also include further RNA molecules, such as larger and more stable circular RNA (circRNA). circRNA can serve as an antagonistic sponge for miRNAs and can be involved in gene regulation of tumor cells. However, common standards are needed to be validated, especially since dysregulation of miRNA and circRNA may also be found in inflammatory diseases.
Tumors and normal cells can release small EVs, which contain typical proteins, DNA, and RNA. The intact cell membrane protects the enclosed compounds against degrading enzymes, such as RNAses, from outside the vesicle. Therefore, EVs can be analyzed for the potential markers. Isolation of EVs from blood is the method of choice for most solid tumors. As an exception, CSF to analyze EVs from brain tumors can give better results compared to blood samples for tumors growing close to the ventricles and due to a better signal-to-noise ratio for tumor vs. non-tumor EVs [62] .
Blood platelets are derived as anucleated cytoplasmic fragments from megakaryocytes. They can react to activation of membrane receptors and outside-in signaling and have been shown to facilitate metastasis via several mechanisms, including protecting tumor cells within the circulation from immune cells and shear stress, supporting the adhesion to endothelium through adhesion receptors, and releasing of angiogenic and mitogenic growth factors at sites of metastasis. Surprisingly, tumors can alter the RNA profile of platelets, leading to the term tumor educated platelet (TEP) [63] . The mechanism remains to be elucidated. TEPs from lung, brain, and breast cancer patients have been shown to be distinct from those with inflammatory or other diseases [19,64] .

CLINICAL STUDIES
Currently, the registry ClinicalTrials.gov from the NIH [65] finds over 900 clinical trials related to the search term "ctDNA" (911 as of 15 March 2022), most of them are ongoing (with 648 "not yet recruiting", "recruiting", "enrolling by invitation", or "active, not recruiting") and only a minority (95) listed as "completed", the remaining others being declared as "terminated", "suspended", or "unknown status". Many of these studies are not originally intended to prove the use of ctDNA as a possible measurement of treatment response for new therapy schemes, but they just routinely integrate the ctDNA analysis as an early detection arm for success or failure of the treatment, i.e., as a complement, or when data from tissue biopsy are missing. However, some of the studies can use the potential of early detection of MRD, often months before the clinical manifestation of local relapse or metastasis [ Table 3]; this lead time allows an  [15] CRC Quantification of ctDNA by patientspecific mutations Prediction/exclusion of relapse after surgery for at least a year (no long-term follow-up) 2015 Garcia-Murillas et al. [66] Early-stage breast cancer Monitoring of patient-specific mutations by dPCR from cfDNA Monitor for MRD, prediction of metastatic relapse after initial therapy (pilot study) 2015 Olsson et al. [67] Breast cancer Whole genome sequencing of primary tumor, rearrangements by dPCR in plasma ctDNA quantity predictive of poor survival 2016 Donaldson and Park [25,29] Advanced NSCL cobas EGFR Mutation Test v2 First ctDNA-based treatment selection EGFR mutation, exons 18-21 (for Erlotinib) 2017 Phallen [68] CRC, breast, lung, ovarian, cancer TEC-seq of cfDNA Detection of early stage tumors 2017 Abbosh et al. [82] Early-stage NSCLC, TRACERx ctDNA gene profiling with multiplex-PCR NGS Identify post-operative relapse 2017 Ng et al. [70] CRC PPS-ctDNA, multiplex-PCR Monitoring after surgery 2017 Schøler et al. [71] CRC Patient-specific mutations for ctDNA Postoperative detection of residual disease; prognosis for high risk of relapse; early detection of relapse and response to treatment 2017 Chaudhuri et al. [72] NSCLC stage I, II, III Gene panel with CAPPseq ctDNA Lead time 5.2 months before clinical recurrence 2018 Mehra et al. [73] mCRPC, Prostate cancer cfDNA concentration Independent prognostic factor for rPFS and OS in first-and second-line chemotherapy 2018 Annala et al. [74] mPRPC ctDNA Disruption of TP53, BRCA2 or ATM results in worse outcomes on novel AR targeting 2019 Coombes et al. [75] Early-stage breast cancer Signatera assay, Personalized, 16plex assays for patients-specific mutations in ctDNA Lead time 8.9 months (up to 2 years) ahead of clinical relapse) 2019 Garcia-Murillas et al. [76] Early-stage breast cancer Monitoring of patient-specific mutations (mutation tracking) by dPCR from plasma cfDNA Major "proof of concept" supporting the idea of clinical utility for prediction of relapse to be tested in larger studies 2020 Moding et al. [77] NSCLC ctDNA, CAPP-Seq Detection of MRD with the prediction of benefit from consolidation immunotherapy 2021 McDuff et al. [78] LARC ctDNA MRD measured by pre-and postoperative ctDNA predicts outcome for chemoradiation 2021 Taniguchi et al. [79] CRC ctDNA Testing platform combining a prospective screening registry with two phase 3 interventional studies 2022 Gale et al. [80] Early stage NSCLC Monitoring of patient-specific mutations (48 amplicons) MRD, identification of patients for further therapy 2022 Tie et al. [81] Stage II colon cancer earlier adaptation of the therapy decisions, sometimes with the intent to cure. It indicates the emerging role of ctDNA measurements in clinical settings, which may become the new gold standard, especially when such treatment studies will become new therapy regimes -and then may need or accept ctDNA as necessary control.

Breast cancer
Globally, breast cancer is the most frequent cancer and the leading cause of cancer-related death in women, although most patients can be cured at an early stage. After early surgical removal, additional treatments with chemo-, radio, and/or hormone therapy may further prevent a metastatic relapse. However, overtreatment of cancer-free patients also represents a risk of unnecessary side effects. Therefore, patients should benefit from personalized therapies adjusted for as many additional treatments as required and as few as possible. The challenge to identify MRD with ctDNA analysis from blood samples has been addressed in clinical settings.
In a pilot study in 2015, later complemented with a larger cohort in 2019, Garcia-Murillas and colleagues were able to monitor early breast cancer after initial therapy by detection of ctDNA, which was associated with a high risk of relapse in all breast cancer subtypes [66,75] . Individual somatic mutations were identified by sequencing of DNA from the primary tumors. The same mutations were later tracked by digital PCR (dPCR) in patient's blood samples. Interestingly, the lead time was 10.7 months (95%CI, 8.1-19.1 months), i.e., long before the clinical manifestation of relapse.
Metastatic breast cancer is only rarely curable at a symptomatic stage. In a retrospective study and long follow-up in 2015, Olsson and colleagues combined whole-genome sequencing of primary tumors with personalized dPCR from plasma ctDNA for quantification of tumor-specific chromosomal rearrangements [67] . With this method, they were able to detect 0.01% of tumor DNA content or one rearranged sequence per 10,000 wild-type sequences. With an average lead time of 11 months (range 0-37 months), they were able to detect metastatic recurrence accurately before its clinical manifestation. In addition, long-term disease-free survivors had undetectable ctDNA. This supported the rationale to evaluate ctDNA in larger studies to monitor early metastasis detection, adjust therapy, and avoid overtreatment.
Mutations in the tumor suppressor protein p53 (TP53) are common among most cancers, although with varying frequencies [82,83] . A high prevalence in a subgroup of breast cancers allowed for monitoring those mutations in ctDNA. Using massively parallel sequencing (MPS), Riva and colleagues detected patientspecific mutations in tumor tissue of non-metastatic triple-negative breast cancer (TNBC) [84] . They tracked these patient-specific TP53 mutations in 10 mL of plasma from different time points: (1) before neoadjuvant chemotherapy (NCT); (2) after one cycle; (3) before surgery; and (4) after surgery. With dPCR of ctDNA, a sensitivity of 75% at baseline was achieved. ctDNA levels correlated not only with tumor burden but also with tumor proliferation rate (tumor grade or mitotic index). During NCT treatment, ctDNA levels dropped in all patients. After surgery, no MRD was detectable. However, a shorter survival was correlated with a slow decrease of ctDNA during NCT [84] . Radovich and colleagues confirmed ctDNA and CTC enumeration above standard clinical parameters as independent factors for the worse outcome, i.e., distant disease-free survival (DDFS), DFS, and OS [ Table 4] [85] .

Colorectal cancer
Colorectal cancer is the second and third most common cancer in women and men, respectively, with almost two million new cases each year [90] .
In 2016, Tie and colleagues investigated a prospective cohort of patients after resection of stage II colon cancer [24] . They used massively parallel sequencing to identify patient-specific mutations from primary tumors and designed personalized Safe-SeqS assays for the identified mutations to quantify ctDNA from plasma. The presence of ctDNA in 7.9% of patients not treated with adjuvant chemotherapy was highly associated with recurrence of the disease within the median follow-up of 27 months (79% of patients). In patients who completed chemotherapy, the detection of ctDNA also correlated with a worse outcome. Therefore, ctDNA is useful to identify MRD and the high risk of recurrence in patients with stage II colon carcinoma.
Ng and colleagues developed a patient primary-tumor specific (PPS) assay to detect ctDNA in plasma samples by tracking individual mutations identified from tumor tissue [70] . ctDNA correlated well with clinical treatment outcome: ctDNA was detectable before, but not directly after, surgery. Furthermore, ctDNA was detectable ahead of clinical manifestation of recurrence, indicating its validity to detect metastasis earlier than established methods of imaging and other markers (CEA).
Schøler and colleagues also showed that ctDNA can be useful for monitoring patients with CRC [71] . The detection of ctDNA after surgery reflected MRD and identified patients with a very high risk of relapse. Follow-ups of three years allowed early detection of relapse with a lead time of 9.4 months compared to standard medical imaging, as well as monitoring additional treatment responses. In contrast to CEA levels, preoperative ctDNA detection rates correlated well with the stage of disease.
For example, CIRCULATing tumor DNA-based decision for adjuvant stage II Evaluation (CIRCULATE), as an ongoing multicenter, prospective, randomized, controlled interventional trial (NCT04089631; Table 4) [88] , evaluates the adjuvant therapy in patients with colon cancer stage II. The primary outcome of this phase 3 study is to compare the DFS in patients who are positive for ctDNA with vs. without the drug capecitabine. Patient-specific mutations previously identified from panel analysis of FFPE tumor blocks with a lack of microsatellite instability are used to detect ctDNA as inclusion criteria for the study, thus demonstrating the integration of ctDNA in clinical settings. The study started in June 2020 with expected primary completion in June 2023 and completion in June 2026 [88] .

Lung cancer
Lung cancer is the leading cause of cancer death (18.4% of total cancer deaths) and, for both sexes combined, the most common cancer (11.6%) with about 2.1 million cases each year [90] . In a seminal paper, Abbosh and colleagues used a phylogenetic approach to track the genetic dynamics of a tumor [69] . They identified predictors of ctDNA in early-stage NSCLC: non-adenocarcinoma histology, necrosis, high proliferative index, and lymphovascular invasion. Chaudhuri and colleagues used a gene panel of 128 genes for CAPP-Seq to predict recurrence with an average lead time of 5.2 months [72] . In 2016, the FDA broadened the approval of the Cobas test (Roche) to be used not only in tissue sections but also as the first liquid biopsy assay for treatment selection to identify specific EGFR mutations in cfDNA of advanced NSCLC patients [ Table 4] [29] .
The SUMMIT study at the University College London hospitals plans to enroll 25,000 participants 55-77 years of age who have no diagnosis of cancer but a high risk for lung cancer due to a significant smoking history [87] . This study is an ongoing, prospective, observational, cohort study with two major aims: (1) to validate the collaborating company's (GRAIL, LLC) blood test on cell-free nucleic acids (cfNAs) for early detection of multiple types of cancer; and (2) to deliver low-dose CT (LDCT) screening for lung cancer. Participants will receive at least one low-dose chest CT at baseline. Another possible scan after 12 months will be randomized. The study aims to keep most scans below 1 mSv radiation dose (ultra-low dose) and all scans under 2 mSv. The liquid biopsy will evaluate the test performance, including sensitivity, specificity, and tissue of origin of detected cancers. The performance of the LDCT screening to detect lung cancers will be evaluated in comparison to established measures of risk prediction of lung cancers and other incidental findings. The study started in April 2019 with an estimated primary completion date of August 2023 and an estimated completion date of August 2030 [87] .

Prostate cancer
Prostate cancer is the second most common non-cutaneous cancer in men worldwide with an estimated 1.3 million new cases each year with more than 350,000 deaths [90] . Prostate cancer contributes to 3.8% of all cancer deaths and is diagnosed in 7.1% of all cancer patients. Mehra and colleagues detected the levels of cfDNA in mCRPC patients in phase 3 clinical studies with first-and second-line chemotherapies [73] . Baseline cfDNA concentration, which is partially derived from the tumor, was an independent prognostic factor in both first-and second-line chemotherapy settings, and it correlated with known prognostic factors, shorter radiological PFS (rPFS), and OS. Higher levels of cfDNA before chemotherapy were associated with more aggressive tumors. A decrease of cfDNA within the first few weeks after initiation of chemotherapy correlated with a benefit for the patient, thus identifying a treatment response. Other studies found that cfDNA can include 15%-20% of ctDNA, depending on tumor stage and tumor burden, but localized prostate cancers typically remain below the threshold for detection of ctDNA, whereas prostate-specific antigen (PSA), a protein marker that is not uniquely an indicator for cancer, is already at a high-risk level [91] . This implies that cfDNA may currently not be feasible to detect and monitor early state or less aggressive tumors. Annala and colleagues [74] investigated cfDNA from mCRPC patients prior to novel AR therapy. Using whole-exome sequencing (WES) with capturing of coding regions of 72 selected genes, they correlated mutations of TP53, BRCA2, or ATM as predictors of worse outcomes on novel AR targeting, thus suggesting liquid biopsy as new guidance in AR-targeted therapy in general practice.

Brain tumors
Cancers of the brain and nervous system represent only 1.6% of all cancers and contribute to 2.5% of all cancer deaths. The critical location within the brain and the lack of major improvements for one of the most devastating cancers, glioblastoma, for over a century represent major challenges. Recently, Eibl and Schneemann published a review on liquid biopsy of primary brain tumors [31] . In contrast to tumors outside of the nervous system, the blood-brain barrier (BBB) may add another challenge for detecting low-level ctDNA in the blood. CSF-although also protected by the BBB-appears to be a much better source due to the normally diminished number of leukocytes as a background source for cfDNA. However, earlier studies were able to confirm specific mutations known from tissue biopsy also in the serum [10,21,92,93] , plasma [94][95][96] , or both [21] . In 1991, one of the authors described the very first TP53 mutations in primary medulloblastoma tissue biopsies [82] . This finding supports a model of histologically indistinguishable from primitive neuroectodermal tumors (PNET) [97,98] . Others were unable at that time to detect TP53 mutations in tissue biopsies or in xenografts of human medulloblastoma, except in only one cell line [99] ; however, this mutation may have been developed as a selective advantage during cell culture. Similar brain tumor models helped to elucidate and confirm several other oncogenic pathways in human brain tumors [98,[100][101][102][103][104][105][106][107] . Meanwhile, TP53 mutations in medulloblastomas are well established and can be used as a prognostic and diagnostic marker: only recently, in 2016 [34] and with an update in 2021 [108] , the World Health Organization (WHO) introduced four new diagnostic groups of this childhood brain tumor based solely on molecular genetic features. The correlation between different biological behavior and personalized risk assessment may allow preventing harmful radiation when not necessary or useful. The first two groups refer to different oncogenic signaling pathways, namely wingless/Integration-1 (WNT)-activated (Group 1) and sonic hedgehog (SHH)-activated (Group 2). WNT is a portmanteau for the Drosophila gene "wingless" (Wg), detected in mutants lacking wings, and the homologous mouse gene, integration 1 (Int-1), which was found earlier to cause tumors by insertional mutagenesis with a retrovirus; SHH refers to the hedgehog gene (hh) found in Drosophila mutants with spikes, reminiscent of a hedgehog (SHH is a vertebrate homolog and named after a character in a video game, Sonic the Hedgehog). WNT-activated medulloblastoma shows the highest five-year survival and a low prevalence of metastatic diseases. SHH-activated medulloblastoma can be further separated into two different subgroups, TP53-mutant or TP53-wildtype. SHH-activated, TP53-mutant occurs primarily in older children and has a very poor prognosis, whereas SHH-activated, TP53-wildtype, which is most common in adolescents and young children, has a good prognosis. The other two groups are non-WNT/non-SHH, Groups 3 and 4, respectively (also known as Groups C and D). Group 3 shows an increased prevalence of metastatic disease with the poorest five-year survival, whereas Group 4 has an increased prevalence of metastatic disease with a moderate five-year survival. TP53 mutations in SHH medulloblastomas are associated with poor survival and treatment failures [34] . Several subgroups have been associated with TP53 and other mutated genes: for WNT-activated, CTNNB1 and APC; for SHH-activated, TP53, PTCH1, SUFU, SMO, MYCN, and GLI2 (methylome); and for non-WNT/non-SHH, MYC, MYCN, PRDM6, and KDM6A (methylome). Since the WHO classification suggests that the diagnosis from molecular profiling of a tissue biopsy is even superior to classical histopathology, at least for brain tumors, it appears reasonable to use ctDNA-based liquid biopsy for monitoring such mutations in brain tumor patients to avoid repeated and risky neurosurgical biopsies [109] . Newer studies successfully used panels of genes. For brain tumors, CSF offers another chance to find ctDNA with a higher sensitivity than plasma or serum [110][111][112][113][114] . ctDNA from CSF even represents the genomic mutations better than plasma; CSF shows an increased sensitivity for putative actionable mutations and CNA (copy number aberrations; EGFR, PTEN, ESR1, IDH1, ERBB2, and FGFR2) [115] . Glioblastoma (glioblastoma multiforme, GBM) represents the most malignant brain tumor. Even after complete surgical removal, the tumor always relapses due to locally infiltrating cells. Chemo-and radiotherapy treatments can help temporarily, but they also trigger the evolution of the tumor to escape all current treatments. Repeated resection and biopsies are generally not indicated. Therefore, a real-time liquid biopsy with ctDNA from CSF offers an alternative at lower risk to monitor the molecular adaptations of the tumor. CSF as a source for ctDNA also provides an additional chance to investigate brain metastases from tumors from outside of the brain better, since they may contribute to the ctDNA in the blood to a lesser extent than metastases from outside of the brain.

CTDNA TESTS
Several ctDNA tests have been developed over the past few years. The spectrum of applications includes early screening, detection of targetable mutations to predict treatment response, MRD, and monitoring after treatment with early detection of resistance [ Table 5] [132] . After extraction of DNA from plasma, the obtained cfDNA is typically used with qualitative PCR or sequencing methods. Such tests can be highly personalized and target varying numbers of tumor-associated or actionable mutations; they can use a custom-built approach to profiling tumor tissue first for patient-and tumor-specific mutations and later use dPCR techniques to monitor tumor development and treatment response from repeated ctDNA probes.

Tumor screening
In a large study, the "Detecting cancers Earlier Through Elective mutation-based blood Collection and Testing" (DETECT-A) blood test demonstrated the proof-of-concept to use ctDNA for tumor screening.
The DETECT-A test represents an early version of the CancerSEEK test [ Table 5] with minor differences, such as using pre-defined thresholds for each DNA and protein biomarker instead of employing artificial intelligence (AI)/machine learning for increasing sensitivity, or for confirming or enhancing specificity [27] .
With a cohort of about 10,000 women between 65 and 75 years of age with no history of cancer and one year of follow-up, 26 cancer patients were considered with positive blood testing to be "first detected by blood testing" after confirmation by positron emission tomography-computed tomography (PET-CT): nine lung cancers, six ovarian cancers, and two colorectal cancers -most of them localized or regional. Overall, 14 of the 26 cancers had elevated ctDNA levels, 11 had elevated protein marker levels, and 1 had both elevated ctDNA and protein markers. This test was not designed for regulatory approval, but it addressed fundamental issues of feasibility and safety of multi-cancer blood tests. Indeed, the authors confirmed that a minimally invasive ctDNA-based test was able to safely detect several types of cancer in patients who had no previously known cancer. This allowed early treatment with the intent to cure. The test also produced 101 false positives and 46 false negatives, which still may allow its use as complementary to standard screening methods, although it cannot be used as a stand-alone test. With the use of AI and technology evolution, further improvements appear to be possible.

Tumor monitoring and treatment guidance
Only a few ctDNA tests are FDA approved for guiding treatment choices to identify cancer patients for mutation-specific treatments: Cobas, FoundationOne Liquid CDx, Guardant 360 CDx, and Therascreen PIK3CA [ Table 5]. As an example of detecting targetable mutations in cancer patients, the Cobas EGFR mutation test v2 (Roche) aims to detect by PCR from plasma in less than 4 h specific and actionable mutations in the EGFR gene exons 18-21 [49] . The FDA approved this test in 2016 for use in liquid biopsy to identify patients for first-line EGFR-targeted therapy (e.g., erlotinib, a tyrosine kinase inhibitor (TKI)) of metastatic NSCLC patients. A similar test was approved earlier for FFPE samples from tumor tissue. In contrast to more recently developing ctDNA tests, the first CTC test approved by the FDA, in 2004, was CellSearch (Veridex). That test uses size, density, electrical properties, and immune surface markers (EpCAM+, Cytokeratins+, and CD45-) for selection and has been cleared for breast, colorectal, and prostate cancer to predict outcome [12] .  [121] Early version of CancerSEEK [26] ctDNA plus protein markers

VISION AND CHALLENGES
As a new milestone in precision medicine, ctDNA-based liquid biopsy has shown its principal utility and safety for an emerging new era of clinical applications. Nevertheless, a few major challenges need to be addressed within the next few years to reach a widespread benefit for cancer patients: the sensitivity of ctDNA diagnostics may increase by technology improvements, thus allowing even earlier detection and, hopefully, easy-to-use multi-cancer screening platforms for at least the most common cancers. A focus will lie on those as well as developing better treatment options including with the intent to cure. Simple tests, such as detecting only actionable mutations, e.g. EGFR exon mutations, may be further developed for not yet used genes, for example, rare and tumor-specific CD44v variant exon combinations [133] . Although many studies describe a significant "lead time" to detect a relapse before any clinical methods or radiologic imaging, one should reflect that any of these clinical methods may also improve over time, e.g., a 3 or 7 Tesla (T) MRI should have a better resolution and detection capacity than a 1.5 T MRI, thus pointing to the uncertainty of comparisons over time. Hundreds of recent and ongoing clinical trials already apply ctDNA methods to monitor new treatments in cancer patients [65,134] . Large multicenter studies will further show how to best standardize and incorporate the use of ctDNA in clinical guidelines. Other arms of liquid biopsy may also be improved in parallel. Living CTCs should be further characterized by not yet applied technologies, such as functional assays with quartz-crystal microbalance (QCM) [135] or atomic force microscopy (AFM), e.g., to check for tumor-specific cell adhesion and pharmacology reactions at the single-molecule level, both applied and pioneered by one of the authors [136][137][138][139][140][141][142][143][144] . Sharing pharmacological and migratory data on CTCs, and making them findable, accessible, interoperable, and reusable (FAIR), will allow meta-analysis, data integration, and data mining to accelerate such new approaches in liquid biopsy [145] . Some of these experimental technologies may be tested in established models first to allow better reproducibility [146,147] . CellSearch with counting CTCs became clinically validated as the first CTC marker for worse outcomes in early breast cancer [148] . It appears reasonable to combine such CTC tests with ctDNA methods to improve liquid biopsy, perhaps similarly to how CancerSEEK successfully combined ctDNA mutations with protein markers.

CONCLUSION
Depending on tumor type and treatment options, ctDNA has been shown to be a valid and independent biomarker for confirming and serial monitoring of cancer. With its very high specificity, ctDNA can be used to predict the early recurrence of a wide variety of cancers after initial therapy. This enables guidance to earlier and better treatment options, including the intent to cure, but also to prevent patients from unnecessary treatments, when there is no indication for MRD or early detection of resistance development. Numerous larger and ongoing clinical studies with ctDNA as a marker for monitoring MRD or treatment response will increase our knowledge on how to apply the best strategies.
Unfortunately, sensitivity often appears to be too low to use ctDNA for general tumor screenings, and not all patients can be monitored reliably. Further improvement in technology is needed to increase sensitivity and standardization. Although ctDNA appears to be much closer to widespread routine use than the other arms of liquid biopsy (CTCs, EVs, and TEPs), perhaps, a combination of ctDNA data with functional studies on living CTCs can soon close the gap for a faster shift to the new paradigm of liquid biopsy in personalized medicine for the benefit of most cancer patients at any stage of diseases. ctDNA as a valid biomarker is ready to enter clinical routine.