Liquid biopsies: Potential and challenges

The analysis of tumor cells or tumor cell products obtained from blood or other body fluids (“liquid biopsy” [LB]) provides a broad range of opportunities in the field of oncology. Clinical application areas include early detection of cancer or tumor recurrence, individual risk assessment and therapy monitoring. LB allows to portray the entire disease as tumor cells or tumor cell products are released from all metastatic or primary tumor sites, providing comprehensive and real‐time information on tumor cell evolution, therapeutic targets and mechanisms of resistance to therapy. Here, we focus on the most prominent LB markers, circulating tumor cells (CTCs) and circulating tumor‐derived DNA (ctDNA), in the blood of patients with breast, prostate, lung and colorectal cancer, as the four most frequent tumor types in Europe. After a brief introduction of key technologies used to detect CTCs and ctDNA, we discuss recent clinical studies on these biomarkers for early detection and prognostication of cancer as well as prediction and monitoring of cancer therapies. We also point out current methodological and biological limitations that still hamper the implementation of LB into clinical practice.


| INTRODUCTION
Despite significant advances in the diagnosis and treatment of solid tumors, distant metastases remain the main cause of cancer-related deaths. However, mere analysis of the resected primary tumor alone, current standard practice in oncology, can provide misleading information regarding the characteristics of metastases. Metastases can develop unique genomic characteristics that might not be detected when examining the primary tumor. [1][2][3] In addition, metastases are often present in different organs at the same time and have one strong, partly organ-dependent heterogeneity. 1,3,4 This diagnostic dilemma led to the development of the "liquid biopsy" (LB) concept. 5 10 In this review, we will focus on CTCs and ctDNA as the most prominent LB markers, with emphasis on studies in patients with breast, prostate, lung and colorectal cancer as the most frequent solid tumors worldwide. 11 After a brief introduction into the methodology, we will discuss the current clinical applications of CTCs and ctDNA.
We analyzed scientific work published within the past 3 years and dealt with the importance of the use of CTCs and ctDNA for early detection, risk assessment and monitoring of cancer therapies.

| CTCs: methodology and technical challenges
Efficient positive enrichment of CTCs can be achieved by approaches that exploit differences between tumor cells and normal blood cells, including the differential expression of tumor-associated cell surface proteins (eg, EpCAM, mucin-1, HER2 or EGFR) or distinct physical properties (eg, larger size or educed deformability) of the tumor cells 6 ( Figure 1A). In contrast, CTCs can also be enriched by negative selection, that is, normal blood cells are removed by antibodies against CD45 or other antigens expressed on leukocytes or circulating endothelial cells. 3 Label-independent technologies based on size exclusion include microfiltration technologies, which involve passing blood through filters with small pores or microfluidic steps that are calibrated to capture CTCs. Other technologies use inertial focus strategies to separate CTCs from other blood components as Hydro-Seq, 12 which is a scalable hydrodynamic scRNA-seq barcoding technique, for high-throughput CTC analysis or dielectrophoresis (DEP), which isolates CTCs based on the different electric charges of tumor and blood cells. Most enriched CTCs are single isolated cells but enrichment of CTC clusters is possible and allows now deeper insights into the exciting biology of CTC clusters. 13 After enrichment, reliable methods are used to identify individual CTCs with specific tumor-associated biomarkers at the protein level or to a lesser degree at the messenger RNA (mRNA) level ( Figure 1A).
For patients with epithelial tumors (carcinomas), keratins have been well established as CTC markers. However, these epithelial markers can be downregulated during epithelial-to-mesenchymal transition (EMT), which might lead to false-negative findings. 14 CTCs might frequently undergo EMT, [14][15][16][17][18][19][20] indicating the need for assays based on new EMT markers. Besides EMT, metabolic markers 19 and clusters 21 also seem to define the biological potential of CTCs.
Downstream analysis of individual CTCs or clusters at the DNA, RNA or protein level have become possible over the past decade. Isolation of single CTCs can be achieved by micromanipulation or DEP-Array technology, but it usually requires sufficiently high starting CTC concentrations. 3 For DNA analysis of single cells, whole genome amplification (WGA) methods need to be employed to generate sufficient quantities of DNA for subsequent sequencing analyses. WGA may induce bias and therefore new WGA-free approaches are currently being developed. Besides RNA sequencing, multiplex reverse transcription polymerase chain reaction (RT-PCR) can provide already some insights into the heterogeneity of CTCs. 22 Analysis at the protein level usually employ immunostaining, but new approaches of multiplex proteomic approaches are on the horizon. Besides descriptive approaches, functional CTC assays exist (eg, the EPithelial Immuno-SPOT [EPISPOT] that is based on the measurement of secreted protein by viable CTCs after short-term culture). 23 In patients with extremely high CTC counts (usually >100 per mL blood), the functional properties of CTCs can be further investigated by the establishment of long-term cell culture/cell lines or the development of xenograft models 24,25 ( Figure 1A). However, the establishment of these models requires many months and the efficacy rates are very low, which makes them unsuitable as tool for clinical trials or decision-making in individual patients. Nevertheless, CTC lines and xenografts provide unique first insights into the largely unknown functional properties of CTCs.

| ctDNA: methodology and technical challenges
Cell-free DNA (cfDNA) circulating in the peripheral blood is mostly released through necrosis and apoptosis but potentially also by secretion through extracellular vesicles. 26 cfDNA consists mostly of 166 bp, which is consistent with the length of a DNA fragment wrapped around a nucleosome. Since DNA is degraded to these mononucleosomal units during apoptosis, this size indicates that programmed cell death is the primary way of release. 27 In cancer patients, only a small portion of cfDNA (usually 0.01%-5%) is ctDNA shed into the blood by tumor cells. 2 Ultrasensitive targeted approaches like droplet digital PCR (ddPCR), BEAMing or real-time PCR methods allow fast, cheap and sensitive detection of prespecified cancer-associated mutations at high sensitivity ( Figure 1B).
Next-generation sequencing (NGS)-based methods include targeted as well as untargeted approaches and stand out due to their ability of massive parallel sequencing of millions of DNA sequences. Targeted NGS methods such as TAm-Seq, Safe-SeqS and CAPP-Seq ( Figure 1B), enable to detect multiple rare mutations in ctDNA simultaneously. Even though targeted approaches show high analytical sensitivity, they are limited to mutations in a set of predefined genes, whereas untargeted approaches like whole genome sequencing or whole-exome sequencing provide the opportunity to detect novel, clinically relevant genomic aberrations without requiring information about the primary tumor ( Figure 1B).
Nevertheless, clinical use of untargeted approaches is usually hampered due to lower sensitivity, higher input sample volume requirement and higher costs. [28][29][30] Besides analytical steps, also preanalytical steps need to be considered accurately to obtain proper samples for cfDNA analysis. As the majority of cfDNA originates from normal cells and acts as a diluent for the small ctDNA fraction, circumstances that lead to an increase of nontumorous cfDNA have to be avoided. Fast processing of the sample, ambient temperatures, double plasma centrifugation and special cfDNA blood collection tubes are required to reduce the background of wild-type DNA, 31,32 and parallel sequencing of normal leukocytes is required to discriminate clonal hematopoiesis of indeterminate potential (CHIP) mutations from somatic tumor mutations. 33 Thus, ctDNA methods have to be standardized and validated before integration into clinical practice. 34,35

| Breast cancer: clinical applications
Breast cancer is the most commonly diagnosed cancer in women and the main cause of cancer-related death for women worldwide F I G U R E 1 A, Technologies for enrichment, detection and characterization of circulating tumor cells (CTCs). CTCs isolated from blood samples can be enriched using marker-dependent techniques: CTCs can be positively selected using antibodies to epithelial or tumor-associated proteins (eg, EpCAM) or negatively selected by depletion of leukocytes using anti-CD45 antibodies. Positive enrichment of CTCs can also be performed using assays based on physical CTC characteristics, including size, deformability, density and electrical charge. After enrichment, the isolated CTCs can be identified using immunocytological assays like membrane and/or intra-cytoplasmic staining with antibodies to epithelial, mesenchymal, tissue-specific or tumor-associated markers (eg, keratins). Molecular assays enable the identification of CTCs at the DNA, RNA and protein level. Functional assays can be used to detect viable CTCs based on their biological activities (eg, the fluoro-epithelial ImmunoSPOT (EPISPOT) assay for certain proteins secreted or shed by CTCs and the related EPISPOT in a drop (EPIDROP) technology that enables the detection of single CTCs in microdroplets). The functional properties of CTCs can also be investigated in vivo by the establishment of CTCderived xenografts. B, ctDNA detection technologies. ctDNA analysis is based on the identification of tumor-specific aberrations or epigenetic marks in cfDNA samples. Ultrasensitive targeted approaches like droplet digital PCR or BEAMing and NGS technologies (Tam-Seq, Safe-SeqS, CAPP-Seq) are able to detect pre-specified cancer-associated mutations at high sensitivity. Refined real-time PCR methods, like allele-specific PCR (AS-PCR), allele-specific nonextendable primer blocker PCR (AS-NEPB-PCR), coamplification at lower denaturation temperature (COLD-PCR) or peptide nuclei acid-locked nucleic acid (PNA-LNA) PCR clamp allow fast, cheap and sensitive detection of mutations. Untargeted approaches like whole genome sequencing, whole exome sequencing or FastSeqS allow the unbiased detection of genomic aberrations without requiring prespecified information about the mutation pattern of the respective primary tumor [Color figure can be viewed at wileyonlinelibrary.com] (2.1 million new cases of breast cancer and 627 000 deaths worldwide in 2018, WHO World Cancer Report, 2020). 11

| Circulating tumor cells
Many patients with primary breast cancer relapse even more than 20 years after primary tumor resection. 36 The sequential analysis of CTCs before, during and after therapy has enabled real-time monitoring of tumor evolution in individual patients with metastatic breast cancer. 37 In early-stage patients, CTC status at baseline has prognostic relevance in patients receiving adjuvant 38 or neoadjuvant 39,40 therapy (Table 1). In addition, sequential follow-up evaluation after primary therapy might be able to detect minimal residual disease (MRD). 6,36 CTC counts 2 or 5 years after completion of adjuvant chemotherapy predicted an unfavorable outcome 43,44 (Table 1). Thus, CTCs might enrich a high-risk group that can profit more from radiotherapy than patients without CTCs. 42 The molecular characterization of CTCs can provide information  (Table 1). Furthermore, patients with Ki-67 (proliferation marker)-positive CTCs had significantly reduced progression-free survival (PFS) and OS compared to patients with low proliferative CTCs. 41 Endocrine therapy is the hallmark in patients with hormone receptor-positive patients (70% of all breast cancers) but resistance to therapy occurs in a substantial fraction of patients. Estrogen receptor-1 (ESR1) mutations can result in a transcriptional profile that favors tumor progression. 50 In estrogen receptor (ER)-positive patients, failure to suppress ER signaling in CTCs predicted early progression after 3 weeks of endocrine therapy. 46 The drug-refractory ER signal transmission within CTCs only partially overlapped with the presence of ESR1 mutations, which indicates additional mechanisms of acquired endocrine drug resistance. 46 Molecular characterization of CTCs opens new avenues for a better understanding of cancer biology with potential implications for the design of new therapies. CTCs with EMT-associated (TWIST1) and cancer-stem-cell (CSC) transcripts (CD24, CD44, ALDH1) had an unfavorable survival. 47 Furthermore, chemotherapy increased the incidence of CSC+/partial EMT + CTCs that may represent a chemoresistant subpopulation of CTCs 48 (Table 1) in CTC clusters are specifically activated by hypomethylation. 13 Targeting CTC clusters 53 might be a new strategy to prevent bloodborne tumor cell dissemination.

| Circulating tumor-derived DNA
In primary breast cancer, early detection of relapse is a prime diagnostic target. Sequential patient-specific ctDNA analysis every 6 months for up to 4 years was able to predict relapse with a lead time of up to 2 years (median: 8.9 months) 54 (Table 2) The multiplicity of new therapies for breast cancer presents a challenge for treatment selection. Monitoring of ctDNA during endocrine therapy has shown that mutations in the ER genes (ESR1) are being selected during therapy 56 (Table 2). Thus, there is a need for new therapeutic approaches overcoming resistance. CDK4/6 inhibition has improved survival in advanced ER-positive breast cancer and monitoring of ctDNA levels might predict early response. 57 In the PALOMA-3 study, palbociclib plus fulvestrant induced a lower PIK3CA-ctDNA ratio (mutated copies/ml) compared to fulvestrant plus placebo, and the PIK3CA-ctDNA assessment anticipated the improved PFS seen with palbociclib ( Table 2). Combining endocrine therapy with PI3K-mTOR inhibition has also shown promise in ER-positive breast cancer in the POSEIDON study. PFS of metastatic breast cancer patients receiving everolimus plus exemestane varied depending on the properties of ctDNA; patients with low or no ctDNA exposure and with less than three mutations on ctDNA showed longer PFS and OS. 58 In patients receiving immune checkpoint inhibitors, tumor heterogeneity determined on ctDNA predicted response to treatment. 55 Overall, these are promising steps toward the use of ctDNA for monitoring tumor burden in breast cancer patients receiving new forms of therapy.

| Prostate cancer: clinical applications
Prostate cancer (PC) is the second most common cancer in men 65,66 and represents a very heterogeneous disease with aggressive and clinically indolent courses.

| Circulating tumor cells
The prostate-specific antigen (PSA) screening of blood in men is conducted for many years. After elevated PSA levels, prostate needle biopsies are necessary to allow histological assessment. Low accuracy of PSA as a biomarker and difficulties in distinguishing between indolent and aggressive PCs lead to unnecessary biopsies and overtreatment of patients. 67 An underinvestigated side effect of prostate biopsies is the potential spread of tumor cells into the blood circulation, recently suggested by an increase in CTCs after biopsy in patients with PC. 68 Moreover, potential tumor spread was also investigated in nonmetastatic high-risk PC patients before and after radiotherapy. Application of three different CTC enumeration technologies (CellSearch, fluoro-EPISPOT assay and CellCollector) disproved the hypothesis that radiotherapy leads to a release of tumor cells into the circulation 16 (Table 3).
CTC analysis has in particular driven biomarker development in patients with metastatic castration-resistant prostate cancer (mCRPC).
Here, large-scale studies have shown that the CTC count is a reliable biomarker for treatment response and prognosis in patients receiving chemotherapy or AR-targeting therapies 69,70 (Table 3) suggesting the rare occurrence of primary resistance, which may play a role if AR-targeted therapy is applied in earlier stages of PC. 72 Besides immunostaining of the ARv7 protein, other assays have focused on the detection of ARv7 mRNA using RT-PCR 22,79 or in situ hybridization with padlock probes, 80 demonstrating substantial intrapatient heterogeneity of ARv7 expression on CTCs. A large proportion of patients not responding to next-generation ADT are ARv7 negative but harbor other AR variants (in particular ARv3) that might also confer resistance. 76 Moreover, analysis of CTCs revealed a role of the proliferative Wnt signaling pathway, a downstream mediator of ARS, in antiandrogen resistance, as it was significantly upregulated in the majority of patients. 73 Finally, there is some preliminary evidence that CTC clusters might affect AR-targeted therapy 21 (Table 3).  (Table 4).
ctDNA is also a valuable target for genomic aberrations of the AR gene including mutations and amplifications or splice variants that can convey resistance to ADT; investigation of these biomarkers could identify patients that might benefit more from other therapeutic approaches such as taxane chemotherapy, poly-ADP-ribose polymerase inhibition or bone tropic radioisotopes. 71,79 For example, the high number of AR amplifications (+8 copies) on ctDNA was associated with primary resistance to the ADT. Also, mutations in the AR ligandbinding domain tended to correlate with shorter time to progression. 71 Primary resistance was also associated with genomic structural rearrangements resulting in truncated AR genes that encode AR proteins with intact N-terminal and DNA-binding domain but lack the ligandbinding domain. 71  Besides blood, cerebrospinal fluid (CSF) has been assessed for LB in patients with brain/central nervous system involvement in ADC 94 ; EGFR-activating mutations were consistent with those in primary tumors, while higher detection rates were found in CSF than plasma, indicating a blockade of free ctDNA release due to the blood-brain barrier. Interestingly, the majority of copy number variations that were detected in CSF cfDNA were unique and not identified in primary tissue, 94 suggesting that brain metastases might undergo a unique selection of viable tumor clones.
ctDNA assessment can be also used for prognostication (Table 5).
A higher mutation load in the plasma ctDNA is correlated with survival in NSCLC. 92  In SCLC, quantitative changes in ctDNA levels correlated with responses to platinum-based chemotherapy. 106 The signal-to-noise ratio was much higher in comparison with NSCLC cases, supporting future ctDNA monitoring in SCLC. 106

| Circulating tumor cells
The detection of CTCs from NSCLC patients is challenging, while SCLC patients exhibit on average more than 10 times higher CTC counts. 107,108 Interestingly, CTCs are more common in pulmonary CTC analysis is not restricted to mutations but can also reveal mechanisms of resistance to therapy that are based on transcriptional plasticity. 6 Regarding the success of immune checkpoint inhibition in NSCLC therapy, the assessment of PD-L1 on CTCs has received attention over the past 5 years. 113 Interestingly, two recent studies suggest that CTCs might provide additional information compared to analysis of tumor biopsies. 98,114 Besides technical issues such as sampling errors, this discordance might also reflect the biology of NSCLC and suggests that CTCs are selected cells with a particular phenotype different from the bulk of the primary tumor. 3 Interestingly, patients with PD-L1-positive CTCs at baseline were more frequently nonresponders to nivolumab compared to patients who had PD-L1-negative CTCs, and after developing resistance to nivolumab all patients had PD-L1-positive CTCs. 98,112 Nivolumab is an antibody that binds to PD1 on T cells and inhibits the interaction with PD-L1 on tumor cells, resulting in T-cell activation and tumor cell lysis. 115 Future studies have to explore why the emergence of PD-L1-positive CTCs is associated with the resistance to immunotherapy. 114

| Colorectal cancer: clinical applications
Colorectal cancer (CRC) is the third-most common cancer in both sexes worldwide and ranks second in terms of mortality (880.000 deaths in 2018; WHO World Cancer Report, 2020). 11

| Circulating tumor cells
In nonmetastatic CRC, the preoperative CTC detection is an independent prognostic marker. 116,117 Consistent with the high rates of liver metastasis in CRC, comparative evaluation of mesenteric and peripheral blood showed that CTCs are trapped in the liver. 118 In Stage III patients undergoing curative resection CTCs followed by mFOLFOX chemotherapy, CTC counts predicted relapse. 119 Thus, CTC detection might help to identify high-risk CRC patients. [120][121][122] Interventional studies are now needed to assess whether CRC patients with CTCs will profit from chemotherapy.
In advanced CRC, CTC enumeration before and during treatment predicts PFS and OS and provides additional information beyond CT imaging. 123 Sequential ctDNA analysis during EGFR inhibition has revealed that KRAS and NRAS mutations can rapidly emerge as a result of the selective pressure exerted by targeted therapy. 136 Interestingly, the emergent population of KRAS-mutant subclones could decline upon withdrawal of anti-EGFR therapy, 136 suggesting the potential to guide "cyclical therapy" characterized by sequential withdrawal and reintroduction of EGFR inhibitors on the basis of ctDNA analyses.
Patient-specific ctDNA assays can be developed by mutation analysis of primary tumors. 137 Furthermore, the ctDNA analyses also assisted in distinguishing recurrent CRC from a second primary cancer. 137 The evaluation of early changes in ctDNA concentration as a marker of therapeutic efficacy is another important goal 138 ; by analyzing the evolution of the ctDNA concentration at inclusion and before second or third chemotherapy cycle, patients could be clearly classified in good or bad ctDNA responder. 138 In a prospective analysis of 1046 plasma samples from 230 patients with Stage II colon cancer using NGS-based assays, 139 ctDNA was detected after surgery in 7.9% of patients who received no adjuvant chemotherapy. After a median follow-up duration of 27 months, the recurrence rate was higher in the ctDNA-positive patients than in the ctDNA-negative patients (78.7% vs 9.8%; HR 18.0, 95% CI 7.9-40.0; P < .001). 139 CtDNA detection after completion of adjuvant chemotherapy was also associated with shorter relapse-free survival (HR 11.0, 95% CI

| CONCLUSIONS
The presented studies support the clinical validity of both ctDNA and CTC for improved risk assessment (staging), monitoring of cancer therapies and early detection of relapse in cancer patients. For cancer screening, ctDNA has the advantage of higher concentrations of bioanalytes compared to the very low CTC counts in early-stage patients. However, it should be noted that the concentration of ctDNA is also low in early-stage cancer patients, which has stimulated the recent development of ultrasensitive ctDNA assays. Applying these assays has revealed a background of cancer-associated mutations in normal white blood cells, which may contaminate the ctDNA fraction. 33 Tumor heterogeneity is a hallmark of solid tumors and has an impact on the classification, diagnosis and future treatment of cancer.
Assessment of ctDNA and in particular CTCs can be also used to encompass intrapatient and interpatient tumor heterogeneity in cancer patients. 3,142 The degree of tumor heterogeneity in individual patients is an underinvestigated mechanism of resistance, which can only be targeted by combinatorial therapies. Addressing the challenge of low CTC numbers, apheresis can highly increase the number of CTCs, allowing better downstream analysis of intrapatient heterogeneity within research studies on selected patients. 143 Both ctDNA and CTCs have advantages and disadvantages as LB markers. Although isolating cfDNA from blood plasma is easy, capturing CTCs from whole blood is more demanding. The efforts needed for the subsequent downstream analyses depend on the desired read-out and the stage of disease. Detecting single mutations or CTCs in patients with advanced disease is less demanding than assessing the broad panel of mutations in early-stage patients with low amounts of ctDNA or CTCs. 3,6 The concentration of ctDNA and CTCs depends also on the localization of the tumor tissue (eg, primary or metastatic brain lesions are difficult to assess by blood analyses).