Circulating tumor cells: biology and clinical significance

Circulating tumor cells (CTCs) are tumor cells that have sloughed off the primary tumor and extravasate into and circulate in the blood. Understanding of the metastatic cascade of CTCs has tremendous potential for the identification of targets against cancer metastasis. Detecting these very rare CTCs among the massive blood cells is challenging. However, emerging technologies for CTCs detection have profoundly contributed to deepening investigation into the biology of CTCs and have facilitated their clinical application. Current technologies for the detection of CTCs are summarized herein, together with their advantages and disadvantages. The detection of CTCs is usually dependent on molecular markers, with the epithelial cell adhesion molecule being the most widely used, although molecular markers vary between different types of cancer. Properties associated with epithelial-to-mesenchymal transition and stemness have been identified in CTCs, indicating their increased metastatic capacity. Only a small proportion of CTCs can survive and eventually initiate metastases, suggesting that an interaction and modulation between CTCs and the hostile blood microenvironment is essential for CTC metastasis. Single-cell sequencing of CTCs has been extensively investigated, and has enabled researchers to reveal the genome and transcriptome of CTCs. Herein, we also review the clinical applications of CTCs, especially for monitoring response to cancer treatment and in evaluating prognosis. Hence, CTCs have and will continue to contribute to providing significant insights into metastatic processes and will open new avenues for useful clinical applications.


INTRODUCTION
Metastasis is the most lethal feature of cancer 1 . Despite significant developments in cancer diagnosis and treatment over the past centuries, metastasis remains a major obstacle to improving clinical outcomes of cancer patients 2 . Nevertheless, we have witnessed significant progress over the past two hundred years in revealing fundamental concepts underlying the development of metastasis and in creating new technologies to facilitate cancer metastasis research. Fig 1 highlights the key discoveries and milestones in the study of cancer metastasis. The 'seed and soil' hypothesis by Steven Paget 3 in the 1830s vividly clarified the progress of cancer metastasis. With the advances of science and technologies, particularly since the 2000s, a plethora of new technologies have been advanced, such as high-throughput sequencing 4,5 , transgenic mouse models 6 , CRISPER/Cas9 editing tools 7 , and single-cell sequencing 8 . With these powerful technologies, the biological phenomena underlying metastasis, such as epithelial-mesenchymal transition (EMT) of tumor cells 9 , the role of exosomes in supporting metastasis 10 , circulating tumor cells (CTCs) and CTC clusters in seeding metastatic colonies 11 , and complex interactions between tumor cells and microenvironment 12 , have gradually been unmasked, along with the discovery of numerous metastasisrelated driver genes. The 'black box' of metastasis is gradually being unveiled, and effective metastasis-targeting agents are believed to be on the horizon in the near future.
Cancer metastasis is a complex multistep process involving cancer cell invasion in the primary site, intravasation into circulation, survival in the circulation, extravasation from the circulation, and attachment to and colonization of the metastatic site (Fig. 2). CTCs are defined as tumor cells that have been sloughed from the primary tumor and are swept away by the circulatory or lymphatic systems. To date, most CTC research has focused on CTCs in the blood circulation. CTCs were first described in 1869 by Ashworth who observed "some cells" in the blood of a metastatic cancer patient with an appearance similar to tumor cells in the primary tumors 13 . CTCs have been assumed to be the substrate of metastasis. Although CTCs originate from the primary tumor, they are distinct from primary tumor cells 14 , with EMT transition properties that help them break free from the primary tumor and facilitate intravasation into the bloodstream, dissemination in clusters of CTCs to increase metastatic potential, and exhibit stemness features that enhance their ability to initiate metastasis (Fig. 2). However, most CTCs perish in the circulation, and only limited CTCs survive and infiltrate distant organs. Interactions between CTCs and the blood environment (Fig. 2), including how CTCs escape immune surveillance in the blood, have been widely implicated in the metastatic mechanisms of CTCs. It has taken more than a century for researchers to recognize the critical role of CTCs in cancer metastasis, due to the unique technical challenges required to isolate these very rare CTCs from the massive pool of circulating blood cells 15 . However, in the past two decades, emerging technologies for CTC isolation have allowed research on the biology of CTCs and have facilitated the clinical applications of CTCs in cancer screening, treatment response monitoring, and prognosis evaluation.
In this review, the biology of CTCs, as well as their interaction with the blood microenvironment, is fully reviewed. In addition, the growing number of highly sophisticated CTC enrichment and isolation technologies will be summarized. Finally, we also discuss the tremendous potential of CTCs in clinical applications.

THE BIOLOGY OF CTCS
Molecular characterization of CTCs Experimental evidence has supported the notion that tumor cells can spread even during the early stages of tumor evolution 16,17 . The molecular characterization of CTCs will add knowledge to the underlying mechanism of metastatic processes, contributing to early diagnosis and prevention of metastasis.
Molecular markers of CTCs. A panel of molecular markers has been used to detect CTCs in various cancers. CTC-associated markers used for different cancers are summarized in Table 1 18,19 . As most cancers are of epithelial origin, the most common marker used for CTCs is EpCAM, a "universal" epithelial marker of cancers 20 . EpCAM expression varies among different cancer types 21 , and EpCAM-based CTC detection technologies are widely applied for cancers that strongly express EpCAM, such as breast and prostate cancer. Many studies have shown that CTCs in breast and prostate cancer are EpCAM-positive, and have validated their prognostic value in either early or metastatic stage cases 22,23 . Other epithelial-derived cancer types, such as pancreatic 24 , colorectal 25 , and hepatocellular cancers 26 , also have a considerable detection rate of EpCAM-positive CTCs. Similarly, the presence of these EpCAM-positive CTCs predicts early distant metastasis and poorer survival of patients 25,27,28 . However, using EpCAM as a CTC marker has limitations. It cannot be used in tumors that are EpCAM-negative or with low expression, such as neurogenic cancers. CTCs can undergo EMT, and epithelial markers, including EpCAM, are down-regulated during EMT, which affects the detection rate of EpCAM-positive CTCs. Although there are doubts as to whether EpCAM-based technologies are appropriate to detect all CTCs, numerous studies have illustrated the potential value of EpCAM-positive CTCs in clinical applications 29 . To some extent, EpCAM-positive CTCs are a substantial subgroup of all CTCs, thus EpCAM-positive CTCs still could be a reliable biomarker if the cancer prognosis and therapeutic efficacy is relevant to EpCAM-positive CTCs.
Due to the EMT activity of some epithelial cancer cells, detecting only EpCAM-positive CTCs probably underestimates the actual total CTC population and misses important biological information of EpCAM-negative CTCs. In some cancer types, such as non-small-cell lung cancer (NSCLC) patients, it was even found that the quantity of EpCAM-negative CTCs was significantly larger than EpCAM-positive CTCs 30 . Nevertheless, the poor isolation of CTCs by EpCAM-based technologies can be rescued by using both epithelial and mesenchymal cancer markers, as well as by markerindependent detection methods. For example, in breast cancer, the use of fluorescent-magnetic nanoparticles consisting of a dualantibody interface targeting both EpCAM and N-cadherin has contributed to high-efficiency isolation and rapid identification of CTCs 31,32 . In biliary tract cancer, a single-cell assay for detecting CTCs allowed identification of both epithelial CTCs and nonconventional CTCs which lacked epithelial and leukocyte markers, and therefore led to an increase of CTC positivity rate 33 . The EMT program of cancer cells exhibits molecular alterations, including decreased expression of epithelial markers (E-cadherin, ZO-1, claudins, and occludins) and increased expression of mesenchymal markers (vimentin, N-cadherin, fibroblast-specific protein1, and fibronectin) 34 . EMT is executed by EMT-related transcription factors, mainly belonging to the SNAIL, TWIST, and ZEB families 34 . All these EMT-related molecules can theoretically be used for EMT-CTCs targeting methods. However, many EMT-related molecules are cytoplasmic or nuclear proteins, precluding their usage in the currently available membrane molecular-based technologies of CTC detection. Proteins such as E-cadherin, vimentin, and twist were most often used in the past 35 (Table 1), probably because of their accessibility of detection in traditional CTC detection technologies including flow cytometry sorting, immunostaining, and fluorescence in situ hybridization (FISH) staining. However, the emergence of single-cell CTC sequencing technologies 36 will make it possible to unmask the EMT status of CTCs more comprehensively, and can cover all the EMT-related molecular alternations at the RNA level.
The variety of CTCs markers indicates the heterogeneity of CTCs among different cancer types. Even in one patient, CTCs are spatio-temporally heterogenous, which may be the result of a spatially different microenvironment in the blood and temporal changes in therapy response. Thus, it is difficult to define the entire CTC population using the very limited molecular markers currently available. In addition, CTC markers should not be constant among different stages of cancer and treatment periods.
Genome analysis of CTCs. Genomic instability contributes to tumor evolution and the emergence of resistant tumor subclones. Monitoring tumor genomic instability, especially in terms of tumor resistance and metastases, greatly contributes to the evaluation of treatment response and precision medicine. The evaluation of CTCs assessment using noninvasive liquid biopsy is accessible for serial sampling to detect the genomic instability of the tumor.
Determining the status of EGFR and KRAS mutations is crucial for guiding treatment in NSCLC patients receiving EGFR tyrosine kinase inhibitors and colorectal cancer patients treated with anti-EGFR therapy respectively. The concordance of mutations between CTCs and matched primary or metastatic tumor tissue has attracted much attention. Using a microfluidic technique to capture CTC, Maheswaran et al. found that only two of 31 patients with mutations were overlooked from their detection assay 67 . They identified the EGFR activating mutation in CTCs in 92% of metastatic patients with NSCLC and detected the drug-resistant mutation T790M in CTCs of 33% of patients who responded to tyrosine kinase inhibitor therapy and in 64% of patients who exhibited clinical progression 67 . For the analysis of the KRAS gene mutation, the mutational concordance rate between CTCs and matched primary tumors ranged from 37% to 90% in colorectal cancer cases [68][69][70][71] . This difference in the concordance rate may be due to the different CTC selection protocols used in these studies. KRAS mutations are also common in pancreatic ductal adenocarcinoma (PDAC), present in 90% of PDAC cases. However, Kulemann et al. found that the discordance rate of KRAS mutation in CTCs and corresponding PDAC tumors was 42% 72 . Studies of  CTC markers mainly includes the epithelial markers, the mesenchymal markers, and the cancer specific CTC markers ARV7 androgen-receptor splice variant 7, ASGPR asialoglycoprotein receptor, CEA carcinoembryonic antigen, EGFR epidermal growth factor receptor, GPC3 glypican 3, MAGE A1-6 melanoma antigen-encoding gene family member A1-family) member A6, MRP multidrug-resistance-related proteins, PI3K α phosphatidylinositol 3-kinase α, PIM1 proviral integration site for the Moloney murine leukemia virus-1PRL3 phosphatase of regenerating liver-3, PSA prostate specific antigen, PSMA prostate-specific membrane antigen, SERPINE1/PAI1 serpin peptidase inhibitor, clade E, XAF1 XIAP-associated factor 1 Circulating tumor cells: biology and clinical significance Lin et al. gene mutation analysis in CTCs were also conducted in many cancers such as prostate cancer 73,74 , breast cancer 75,76 , hepatocellular carcinomas 77 , and a mutational discordance between CTCs and corresponding tumors were often found. The discordance rate was probably attributed to the different detection efficiency of CTC mutations, or the heterogeneity between CTCs and primary tumor cells. Genomic assessment of tumor tissue and CTCs can be complementary. So, a combination of mutational testing of CTCs and tumor specimens would guide treatment more precisely. Determining copy number alternations (CNA) of CTCs helps analyze and track cancer profiles as tumors evolve. In lung cancer, Ni et al. 78 found that CTCs exhibit reproducible CNAs patterns, similar to those of metastatic tumors, and different patients shared similar CNAs patterns. In small-cell lung cancer, a CNA-based classifier for CTCs correctly assigned 83.3% of patients as chemorefractory or chemosensitive 79 . In breast cancer, the assessment of CNAs in archived CTCs is feasible. Paoletti et al. 80 found that the CNAs of CTCs and paired metastatic tumor tissue in breast cancer patients were highly concordant, although CTCs and matched tumor tissue harbored several discordant copy number alterations, suggesting that CTCs were the subclone cells of tumor tissues. In triple-negative breast cancer, CTCs with chromosome 10 and 21q CNAs are predictive of clinical progression, and their network analysis presented connected modules including HER/ phosphatidylinositol-4,5-bisphosphate 3-kinase/RAS/JAK signaling 81 . In prostate cancer, Lambro et al. 82 revealed that CNAs of CTCs were interpatient and intercell heterogeneous, and could be missed in bulk biopsy analyses. In metastatic castration-resistant prostate cancer, whole genomic copy number analysis of CTCs showed that common genomic gains in CTCs involved genes such as androgen receptor (AR), mesenchymal-to-epithelial transition (MET), ERG, and cyclin-dependent kinase 12, while common genomic losses were observed in genes such as phosphatase and tensin homolog (PTEN), RAF1, and GATA2 83 . Similarly, Malihi et al. also observed that CNAs in genes including PTEN, RB1, TP53, and AR closely associated with genomic instability and survival in aggressive variant prostate cancer 84 .
Other genome analyses have also been conducted in CTCs. FISH testing was adopted in CTCs to detect biomarkers for treatment sensitivity, such as ALK FISH testing in CTCs of NSCLC patients 85 and HER2 FISH testing in CTCs of breast cancer patients 86,87 . Recently, based on the technique of single-cell resolution DNA methylation analysis, the DNA methylome of single CTCs and CTC clusters was revealed for breast cancer patients, and indicated that the CTC cluster hypomethylation profile obtained was associated with a poor prognosis and that treatment with Na+/K+-ATPase inhibitors to dissociate CTC clusters could revert the methylation profile of CTC clusters and suppress metastasis 88 .
Transcriptome analysis of CTCs. Single-cell sequencing has developed rapidly in recent years and has been applied to investigate the CTC transcriptomes. Single-cell expression profiles can distinguish CTCs from mesothelial cells and blood cells in lung adenocarcinoma, with representative markers including EpCAM for CTCs 89 . Single-cell sequencing-based transcriptome analysis revealed heterogeneity in the CTC subpopulation. By testing the expression of proliferation-associated genes such as the Ki-67 proliferation marker, Magbanua et al 90 . found that 65% of CTC in patients with metastatic breast cancer had low proliferation Ki-67 and that the 35% of patients with a high proliferation Ki-67 expression had a poor prognosis. Cheng et al. 91 performed a single cell transcriptome analysis of 666 CTCs in patients with metastatic breast cancer. They determined that intra-patient CTCs were heterogeneous with regard to EMT-like and MET-like states, and CTCs were enriched for the stem-like phenotype 91 . Single-cell sequencing of CTCs also greatly helped in discovering driver signaling pathways that contributed to metastasis and treatment failure. RNA-Seq of single prostate CTC indicated the activation of noncanonical Wnt signaling in antiandrogen resistant patients 92 . Further, using mouse models, ectopic expression of Wnt5a attenuated the effects of an AR inhibitor and suppression of Wnt5a could partially restore the sensitivity in drug-resistant prostate cancer cells 92 . CTCs have been associated with a poor prognosis in colorectal cancer. A study of the CTC-specific transcriptome profile 93 of six metastatic colorectal cancer patients characterized 410 CTC-specific genes, which were primarily related to cell movement and adhesion, such as VCL ITGB5, bone morphogenetic protein 6, transforming growth factor beta 1, and talin 1, and were related to cell death and proliferation, such as amyloid beta precursor protein, clusterin, and TIMP1.
Epithelial-to-mesenchymal transition of CTCs EMT is the process by which epithelial tumor cells lose their intercellular adhesion and acquire mesenchymal and invasive properties. During dissemination, tumor cells detach themselves from the basement membrane through EMT activation and directly enter the circulation, serving as CTCs traveling to distant sites. When CTCs extravasate, they then undergo a reverse process termed MET and proliferate to form macro-metastases [94][95][96] . Herein, metastatic development depends on the delicate balance of the transition between these two phenotypes. The activity of EMT-MET was also proposed to play an important role in the metastatic process of CTCs 97 . Using mouse models, it was found that epithelial-type CTCs with a restricted mesenchymal transition had the strongest lung metastases formation capacity, whereas mesenchymal-type CTCs showed limited metastatic ability 98 . In breast cancer, CTCs exhibit dynamic changes in EMT composition, and mesenchymal CTCs were found to be closely associated with cancer progression 9,98 . A plethora of studies has shown increased EMT of CTCs rather than primary tumor cells in various cancers [99][100][101][102][103] . In a study based on the bioinformatics analysis of seven sets of gene chips, Guan et al. showed that compared with primary tumors, the main changes in CTCs involved cell adhesion, EMT, and apoptosis 104 . In a prospective study including 39 patients with invasive breast cancer, Tashireva et al. observed a majority of heterogeneous CTC phenotypes (22 out of 24 detectable samples) exhibiting EMT plasticity 105 . Interestingly, it was found that fluid shear stress can induce the EMT of CTCs via JNK signaling in breast cancer, which further confirmed the relationship between augmented EMT of CTCs and poor patient survival 106 .
Clinically, combining the total CTC count and the proportion of mesenchymal CTCs 107 can be used to monitor therapeutic resistance and predict prognosis in cancer patients due to the significant survival differences of this criterion 108 . For example, since the baseline presence of total CTCs in advanced NSCLC conferred poor prognosis 109 , and the presence of over five EMT-CTCs indicated progressive disease 110 . Different numbers of total CTCs and EMT CTCs were found to play an important role in determining the prognosis of breast cancer patients. Intriguingly, it was emphasized that a better understanding of EMT-CTC subtypes and their interactions with peripheral blood mononuclear cells could help design better anti-metastatic treatments 111 . As CTC EMT-positive patients with neutrophil-tolymphocyte ratios ≥3 had an 8.6 times increased risk of disease recurrence compared with CTC EMT-negative patients with lower neutrophil levels, inflammation-based scores increased the prognostic value of CTCs in primary breast cancer 112 . Therefore, targeting the EMT pathway may prevent tumor cell spread in early-stage patients and eradicate metastatic cells in advanced stages.
The stemness of CTCs Many previous studies have indicated a subpopulation of aggressive CTCs with "stemness" traits in different cancers, which Circulating tumor cells: biology and clinical significance Lin et al. refer to their properties for self-renewal and tumor growth induction. In bladder cancer, studies have reported the high expression of OCT4, a crucial stemness maintenance protein 113 , in a subgroup of CTCs 114 . In breast cancer, the CD44 + /CD24 -/low and aldehyde dehydrogenase 1 (ALDH1) + cell phenotypes are reported to be associated with stemness. After detecting the expression of CD44, CD24, and ALDH1 of CTCs in 30 metastatic breast cancer patients, Theodoropoulos et al. found that 35.2% of 1439 CTCs were CD44 + /CD24 -/low , and 17.7% of 238 CTCs were ALDH1 high /CD24 -/low , providing evidence of CTCs stemness 96 . A CTC-3 cell line established from the peripheral blood cells of a breast cancer patient showed more aggressive growth than the widely used MCF-7 breast cancer cell line. Gene profiling revealed higher expression of the stemness markers in the CTC-3 cell line compared to MCF-7 cells 115 . In hepatocellular carcinoma (HCC), 71.4% of HCC patients were CTC positive for the cancer stem cell marker, CD44; thus, they had a significant population of CTCs with stem properties, which could contribute to tumor cell survival and dissemination 116 . In glioblastoma, RNA-seq analysis revealed Wnt activation-induced stemness and chemoresistance in CTCs 117 . In prostate cancer, the stem cell marker CD133 was observed in the majority (> 80%) of CTCs of patients with metastatic castrationresistant prostate cancer 97 and a stem-like subpopulation of the C-X-C motif chemokine receptor 4+/CD133+CTCs was more prevalent in EpCAM-negative CTCs than in EpCAM-positive CTCs 118 .
It is crucial to understand the mechanisms regulating CTCs stemness, and interfering with the CTCs subpopulation stemness properties may more efficiently suppress cancer progression and relapse. CTCs undergo considerable levels of fluid shear flow during their dissemination, and the fluid shear flow itself may have an impact on CTCs. Using a model of breast cancer with brain metastasis, it was suggested that hemodynamic shear flow could upregulate the stemness genes of CTCs in surviving under conditions of shear flow 119 . EMT-like transition of CTCs by downregulating ERK and GSK3β signaling could promote the conversion of CTCs into stem-like CTCs with high sphere-forming and tumor-initiating capacity 120 .

CTCS AND THE BLOOD MICROENVIRONMENT
When transported in the bloodstream, a major of CTCs are constrained by detrimental shear stress, or die from anoikis, a programmed cell death mechanism due to loss of cell attachment 121,122 . Only a small fraction of CTCs interact tightly with platelets, neutrophils, macrophages, myeloid-derived suppressor cells (MDSCs), or cancer-associated fibroblasts (CAFs) to escape the immune system and promote their survival 123,124 . Recently, accumulating studies suggest that the interaction and modulation between CTCs and hostile blood microenvironment is essential for adhesion to endothelial cells, tissue invasion, and tumor metastasis (Fig. 3).
Interaction of CTCs with neutrophils Neutrophils are the most abundant circulating leukocytes in humans and have recently been studied to support cancer progression 125 . An increased number of neutrophils in circulation is associated with poor prognosis in several types of cancers [126][127][128] . The formation of CTC-white blood cells (WBCs) clusters was previously reported within the bloodstream 129 . In 2019, Szczerba et al. determined that CTCs were significantly associated with neutrophils in both mouse models and breast cancer patients, exhibiting more metastatic potential with greater expression of genes that involve cell cycle progression compared to CTCs alone 130 . These observations are consistent with previous findings showing the proliferation role of neutrophils on tumor cells 131 . CTC and neutrophil binding is mediated by the cell-cell junction and possibly requires the vascular cell adhesion molecule 1 130 . Furthermore, neutrophils can directly adhere to CTCs through the Mac-1/ICAM-1 interaction and act as a bridge between tumor cells and the liver parenchyma, thus promoting extravasation and liver metastasis 132 . Thus, CTCs clusters with neutrophils anchor to the vascular endothelium for extravasation while resisting shear stress, and the process is mediated by a series of cell adhesion proteins, such as cadherin, integrin, and surface glycoprotein [133][134][135] . Intriguingly, Chen et al. found that IL-8 secreted from neutrophils was essential for neutrophil sequestration with arrested tumor cells and for the extravasation behaviors of adjacent tumor cells through the endothelial barrier 136 . Neutrophils can also promote metastasis in an indirect manner. Neutrophil extracellular traps (NETs) are web-like structures formed by DNA-histone complexes and proteins released from activated neutrophils, with the ability to impact on CTC biology 137 . Many studies have found that NETs were able to capture CTCs in the circulation and, in doing so, promoted metastatic dissemination 138,139 . In vitro and in vivo experiments have shown that CTC adhesion to NET is mediated by β1-integrin expressed on both NET and cancer cells, while this effect was abrogated following DNAse I administration 139 . In a murine model of surgical stress, the NET formation triggered the release of high-mobility group box 1, which activated TLR9-mediated pathways in CTCs and therefore accelerated the progression of liver metastases 140 . In addition, NETs can also awaken dormant cancer cells and promote metastasis. Recently, Albrengues et al. elegantly demonstrated that two NET-associated proteases, neutrophil elastase and matrix metalloproteinase 9 (MMP9), concentrate at laminin, provoke its cleavage, by generating an epitope that induced awakening of dormant cancer cells by integrin activation and FAK/ERK/MLCK/ YAP signaling 141 . In turn, tumor-expressed protein, such as protease cathepsin, has been shown to support lung metastasis of breast cancer by promoting NET formation in metastatic niches 142 . Coiled-coil domain containing protein 25, another protein expressed on the cancer cell membrane, could serve as a specific sensor for the DNA component NETs, which induces migration and adhesion of tumor cells 143 . In addition, by forming NETs, circulating neutrophils can help CTCs escape immune surveillance by suppressing the activation of peripheral leukocytes 144 , the function of natural killer (NK) cells 145 , the antitumor response of effector T cells 146 , and even cooperation with other immune cells (such as IL17-producing γδT cells) 147 . Overall, there is evidence for a pro-metastatic role of neutrophils in their interaction with CTCs, but the specific mechanism remains to be elucidated in more detail.
Interaction of CTCs with macrophages Tumor-associated macrophages (TAMs) not only contribute to metastatic progression within the primary tumor, but also promote later stages of metastasis including dissemination and extravasation of CTCs 148 . Hamilton et al. tried to investigate the CTC-macrophage interactions by co-culturing peripheral blood mononuclear cells with CTC cell lines obtained from small-cell lung cancer patients. They found that CTCs were able to induce monocyte differentiation to TAMs, which secrete a host of mediators such as osteopontin, MMP9, chitinase-3-like-1, and platelet factor to promote further leukocyte recruitment, migration, and invasion 149,150 . In another study of colorectal cancer, the feedback loop between TAMs and cancer cells is essential for the EMT program of CTCs and intravasation into the blood stream. Mechanistically, IL6 derived from TAMs regulates invasiveness through the STAT3/miR-506-3p/FoxQ1 pathway, which in turn increases CCL2 expression of TAM-educated tumor cells to help recruit macrophages 151 . TAMs seem to promote CTCs acquisition of mechanical adhesiveness and endurance, thereby helping them to form protective cell clusters and confer resistance to shear stress 152 . The fusion of macrophages and tumor cells could be a potential mechanism of immune evasion and invasion. These macrophage-tumor cell hybrids were shown to have M2-like macrophage phenotypes (CD163) as well as epithelial markers (EpCAM) 153,154 , and have been isolated from the blood of patients with multiple cancers such as PDAC 155 , melanoma 154 , breast, ovarian, and colorectal cancer 156 . Furthermore, when transplanted into mice, they spread widely and formed lesions in distant tissues 154,155 . A recent study by Gast et al. revealed that fusion hybrids can increase tumor heterogeneity and metastatic behavior, further correlating with disease stage and overall survival among several cancers 157 . Furthermore, larger hybrid sizes were also associated with worse survival among patients with non-small cell lung cancer 158 . Understanding the mechanism of direct interaction and molecular fusion between CTCs and macrophages is of great significance for identifying therapeutic targets.
Interaction of CTCs with platelets The metastasis and progression of cancer are greatly influenced by the recruitment and activation of platelets, which support the survival of CTCs as well as their seeding and outgrowth at secondary sites 159,160 . Platelets can bind to and form aggregates with CTCs in the blood stream, and CTCs expand the formation of aggregation by releasing prothrombotic and procoagulant microparticles or by expressing tissue factor 161,162 . Plateletreleased mediators, such as TGF-β, have been found to accelerate EMT in CTCs and to promote invasion and metastasis 163,164 . Xiong et al. recently determined that the expression of the heat shock protein 47 was induced during EMT, which enhanced the cancer cell-platelet interaction through its dependent collagen secretion in breast cancer cells 165 . Interestingly, platelets are believed to protect CTCs against mechanical stress 166 and induce resistance to anoikis, the latter being mediated by the activation of the YAP1 pathway 167 . Furthermore, platelets promote escape of CTCs from NK cell attack through various mechanisms including (1) platelet aggregates that produce surface shielding to defend the cytolysis effect of NK cells; 168 (2) platelet-derived normal MHC-I that are transferred to the surface of the tumor cell, thus preventing the identification of NK cells; 169 (3) downregulation of natural killer group 2, member D (NKG2D) in NK cells by platelet-derived TGF-β, as well as platelet-mediated shedding of NKG2D ligands, which contribute to impaired antitumor cytotoxicity; 170,171 and (4) plateletderived glucocorticoid-induced TNF-related ligand which activates GITR in NK cells and reduces their cytotoxicity 172 . Furthermore, NK cells and platelets can also interfere with neutrophils, T cells, and macrophages and modulate their immune function [173][174][175] . In addition to safeguard CTCs within the bloodstream, platelets are also involved in the adhesion of endothelial cells. The attachment of platelets and CTCs is mediated by platelet adhesion receptors, such as the integrin αIIbβ3 and P-selectin, thereby supporting the firm adherence of CTCs to the endothelial wall [176][177][178] . Furthermore, tumor cellactivated platelets release ATP from dense granules, which then induces the activation of the endothelial P2Y2 receptor and allows transendothelial migration of tumor cells by increasing permeability 179 . One study has also revealed that the interplay between integrin α6β1 on platelets and its receptor, a disintegrin, and metalloprotease 9 on CTCs is necessary for the extravasation process of cancer cells 180 . Platelets could also increase vascular permeability to help tumor cell extravasation. For example, a preclinical lung metastasis model showed that tumor cell-associated CD97, an G protein-coupled receptor, can initiate platelet activation thereby leading to granule secretion, including both ATP and lysophosphatidic acid release 181 . Similarly, the interplay between platelet-specific receptor glycoprotein VI and its ligand galectin-3 expressed on colon and breast cancer cells was revealed to promote platelet activation and ATP secretion 182 . Consequently, these platelet secretions favor a process of tumor metastasis by regulating vascular permeability. Recently, Xu et al. discovered that Ptx@AlbSNO can block tumor-specific platelet functions to suppress tumor EMT as well as prevent platelet adhesion around CTCs. Ptx@AlbSNO can also inhibit TGF-β secretion and enhance intratumoral immune cell infiltration to reverse the immunosuppressive TME, thereby suppressing distant metastasis 183 . Taken together, the close and complex crosstalk between CTCs and platelets might involve distinct molecule variants and signaling pathways and possibly represent a promising antitumor strategy, particularly attractive for the treatment of several cancers.

Interaction of CTCs with MDSCs
MDSCs are a heterogeneous subset of myeloid cells characterized by immunosuppressive properties that also promote metastatic dissemination. Under a standard protocol for isolating human MDSCs, Cassetta et al. found that polymorphonuclear (PMN)-MDSC were significantly expanded among most cancer types except melanoma compared with infection and inflammation 184 . CTC-MDSC clusters are thought to evade immune surveillance of the T cell response 185 . Indeed, a decrease in circulating MDSCs was associated with an increase in activated OX40+PD-1-T cells in patients with diffuse large B-cell lymphoma 186 . Furthermore, Sprouse et al. found that in vitro co-culture of CTCs derived from melanoma and breast cancer patients and PMN-MDSCs enhanced Notch activation in CTCs through direct interaction between Jagged1 (Notch1 ligands) expressed on MDSCs and the Notch1 receptor expressed on CTCs. The increased production of reactive oxygen species production of MDSCs could upregulate Notch1 receptor expression, therefore, promoting CTC proliferation 187 . The potential mechanisms underlying the interplay between CTCs and MDSCs remains to be determined.

Interaction of CTCs with CAFs
CAFs are one of the most abundant components in the TME and play a prominent role in tumor initiation, angiogenesis, metastasis, and drug resistance 188 . Mechanistically, CAFs remodel the extracellular matrix structure, which allows tumor cells to invade through the stroma and communicate with cancer cells by secreting growth factors, chemokines, and cytokines 189 . However, little is known about the interplay between CAFs and CTCs. Duda et al. first demonstrated that CTCs could carry CAFs from the primary tumor to the metastatic site in mouse models of lung cancer metastasis 190 . These host-derived CAFs directly enhance tumor cell survival and promote the formation of metastasis, while depleting CAFs from lungs significantly reduces the number of macroscopic metastasis and extends survival rate in mice. Moreover, CAFs can protect CTCs from the fluid shear forces during the dissemination process 191 . In a three-dimensional coculture model, CAFs were found to induce shear resistance to prostate tumor cells through stable intercellular contact, as well as soluble factors (such as CXCL5, CCL2, and CCL7), which are associated with cell survival, invasion, and EMT. In addition to experimental models, circulating CAFs (identified by FAP and α-SMA co-expression) have been detected in the peripheral blood of patients with metastatic breast, cancer but not in patients in the early stages 192 , and exhibit excellent precision in metastatic diagnosis (AUC-ROC, 0.975) when isolated using a novel acoustic microstreaming platform 193 .

TECHNOLOGIES FOR CTCS ENRICHMENT AND ISOLATION
Over the past few years, many methods have been proposed to capture CTCs. Due to the extremely small proportion of CTCs in patients' blood, it is still a great challenge to accurately isolate CTCs from the numerous blood cells, and especially to invent applicable methods that can efficiently detect viable CTCs for subsequent in-depth analysis. Here, we will discuss the development of CTC-related technologies over the last two decades, as they have experienced tremendous growth. We will emphasize the most innovative methods associated with nanoscale materials or novel microfluidic chips, hoping to provide a useful framework of CTC-related technologies.
Generally, there are three core strategies of CTCs technologies 194 , which include (1) capture and enrichment, (2) detection and identification, and (3) release. The first strategy of capture and enrichment involves a specific interaction between CTCs and materials through physical interactions or antibody-antigen interactions. The second strategy of detection, which means identifying the CTCs, refers to various methods, such as fluorescence microscopy, fluorescence spectrophotometry, flow cytometry, surface-enhanced Raman scattering, or electrical impedance. In the last strategy, released CTCs are mainly used for downstream analysis, such as genomics, transcriptomics, proteomics, and CTCs culture.
Classic CTC-related technologies based on physical properties The physical separation enrichment method of CTCs is based on differences between CTCs and blood cells in size, density, deformability, and electrical properties. The Isolation by size of epithelial tumor cells system 195 can filter blood samples through an 8-μm diameter polycarbonate TRACK-ETCH-type membrane, but it has low efficiency. An improved method, which consists of a pressure regulating system, the flexible micro spring array device 196 , reaches a capture efficiency of 90% with a detection of CTCs in 76% of samples. However, there are various trends of CTC counts observed from different samples, making this method not reliable for widespread use. CTCs can also be sorted using the Oncoquick system 197 , a density-dependent technique that allows red blood cells and WBCs to be filtered, or by Apostream 198 that uses dielectric electrophoresis techniques in the microfluidic chamber to capture CTCs. These systems require large volume of blood and cannot collect the CTCs of a similar size as WBCs, which are their main limitations. Overall, the methods based on physical properties are generally inefficient, poor in purity, and lack of specificity, although the vitality is good and the cost is relatively inexpensive.
Classic CTC-related technologies based on biological properties Biological property-based technology is another important method for CTC detection. Based on antibody-antigen interaction, CTCs are usually positively enriched using epithelial (EpCAM) and mesenchymal (vimentin) markers as well as negatively enriched by using CD45 to deplete unwanted leukocytes 199 . EpCAMdependent techniques are most commonly used by researchers. The CellSearch system 200 , the only FDA-approved device for clinical use, employs EpCAM antibody-coated ferromagnetic beads to enrich CK + /CD45-/DAPI + CTC and remove CK-/ CD45 + /DAPI + WBCs. However, CTCs strongly adhere to the surface of the equipment in antibody interaction-based methods, making them difficult to be released. This deficiency can be resolved in another EpCAM-dependent, MagsWeeper system 201 , which uses a magnetic rod to enrich CTCs and eliminate cells not bound to magnetic beads, allowing the release of CTCs for the following biochemical analysis. Canpatrol TM 202 is another representative of the EpCAM-dependent technique, which provides morphological, cytological, and genetic characterization of individual CTCs. In summary, techniques based EpCAM are extensively-used. However, because the CTC surface antigen has high heterogenicity, CTCs that have low expression of EpCAM may not be enriched, causing inaccurate results, while methods based on physical properties do not have this limitation. Therefore, combining the advantages of different technologies or looking for CTCs with high sensitivity and specific tumor markers have gradually won the attention of researchers.
Recently, with the development of microfluidic chips, nanomaterials, and next-generation sequencing, researchers have many advanced technologies to stimulate progress in CTC-related technologies. Importantly, researchers are trying to reach higher levels of CTC-related technologies in several key parameters: yield, purity, enrichment ratio, throughput, viability, sensitivity, specificity, release rate, accessibility for further analysis, and simplicity of equipment operation).
Recent CTC-related technologies: microfluidic-based and nanotechnology-based techniques Besides the classic CTC-related technologies discussed above, some newer technologies, such as microfluidic-based and nanotechnology-based techniques have been developed. Microfluidic-based cell sorting approaches use "intrinsic" (e.g., fluid dynamic forces) versus "extrinsic" external forces (e.g., magnetic, electric field, acoustic, and optical forces) to separate cells, and then select target cells from a sample of heterogeneous cells through different physical and biological properties 203 . The CTC-chip is a silicon microfluidic platform on which the CTCs are captured on the slides of molecular marker coated posts. The CTC-chip 204 can separate viable CTCs from whole blood without pre-labeling or processing of samples, resulting in increased cell activity and separation purity. A modified chip-based platform using gold nanoparticles on a herringbone chip (NP-HB CTC-Chip 205 ) easily detaches viable CTCs and safely releases cells for further analysis by utilizing a chemical ligand-exchange reaction with gold nanoparticles on a herringbone chip. Furthermore, the monolithic CTC-iChip 206 has high-efficiency WBC depletion and allows the characterization of CTCs with epithelial and mesenchymal characteristics. Although these microfluidic chips have greatly contributed to the development of the detection of CTCs (i.e., improved capture efficiency, viability, and depletion of WBCs), they are not widely applied for clinical use due to limitations, which include long set-up time, high initial cost, bulky instrumentation, and limited ability to perform single-cell molecular analysis.
In an attempt to capture CTCs in an automated manner ( Table  2), Zhang et al. successfully sorted MCF-7 cells from a 5 mL volume of diluted blood within 23 m with a recovery rate of 85% 207 . Even more remarkable, Jia et al. developed a less costly self-driving micro-cavity array chip to achieve cell loading, lysing, isothermal amplification, and signal read-out on a single chip 208 . This novel chip can perform genetic analysis at the single-cell level, it has great potential in personalized therapy and efficacy monitoring. Furthermore, another automated and integrated microfluidic system proposed by Wang et al. is reported to achieve CTC capture and identification within 90 m. With the advantages of automation, stability, economy, and user-friendly operation, this system provides broad prospects for cancer screening and prognosis, especially in HCC 209 . Additionally, Lee et al. invented a microfluidic-based integrated system to achieve simultaneous on-chip isolation and characterization of circulating tumors utilizing differences in magnetic field gradient and immune fluorescence. Furthermore, this novel system can differentiate onchip eight different subtypes of heterogenic CTCs, guiding the diagnosis and prognosis of breast cancer 210 . By combining the microfluidic technology and in situ molecular profiling techniques, the On-chip Post-processing Enabling chip platform has the ability to perform molecular analyses of single CTC from metastatic breast cancer and metastatic pancreatic cancer patients without any off-chip processes, suggesting its potential implementation of early molecular detection for cancer metastasis 211 . Taken together, less costly automated and integrated microfluidic systems that allow easy CTCs detection and cell analysis have great clinical value.
With the progress of nanomaterials, nanotechnology-based methods are becoming promising tools for CTC detection at an early-stage disease and for the monitoring of cancer development, as well as in vivo imaging 212 . Nanomaterials have a large surfaceto-volume ratio and allow CTC isolation at high specificity and CTC detection at high sensitivity by adsorbing numbers of targeting ligands to bind specific molecules on cancer cells. At present, studies have reported many types of nanomaterials (Table 2) for CTC detection, including magnetic nanoparticles [213][214][215][216][217] , gold nanoparticles [218][219][220] , and quantum dots [221][222][223] . For example, studies have shown that the utilization of tannic acidfunctionalized magnetic nanoparticles 214 , CoFe2O4@Ag magnetic nanohybrids 216 , and peptide-based magnetic nanoparticle 213 , enhances the capture efficiency of CTCs in breast cancer patients. Among them, peptide-based magnetic nanoparticles can distinguish epithelial and mesenchymal CTC subgroups and allow analysis at the single-cell level, the detection effects of which are supported by magnetic nanoparticles and microfluidic-based integrated systems 210 . Regarding gold nanoparticles, there have been significant developments. For example, the cytosensor proposed by Yang et al. 224 showed excellent analytical performance, with a wide linear range, satisfactory CTC release (93.7-97.4%), and good cell viability. Liu et al. reported that gold nanoparticle-modified black phosphorus nanosheets improved the stability in detecting CTC 218 . Furthermore, the combination of a microfluidic system and gold nanoparticles presents a wider range of applications. Wang et al. synthesize an interfacial zinc oxide coating with a nanostructure on the microsphere surface, which increases the specific surface area and thus leads to an improved capturing efficiency of CTCs 225 . In addition, the utilization of multicolor magnetic surface-enhanced Raman scattering nanotags and chip-based immunomagnetic separation could detect four different surface protein markers on individual tumor cells in a quantitative and simultaneous manner, thus facilitating the separation of CTC subpopulations 217 .
Although nanotechnology-based techniques can provide broad prospects for CTC research in various tumors in a cost-effective and simple manner, there are limitations and challenges. First, many factors (e.g., binding of nanoparticle probes, aggregation) can affect nanoparticle-based detections, leading to decreased reliability and reproducibility. Second, most nanoparticle-based assays are prepared for academic studies, and they are still unrealistic for widely clinical translation. Third, there is possible toxicity of nanoparticles.
In the era of precision medicine, CTCs analysis has great clinical value. Tools must be sharpened first if workers are to do their job well. Therefore, CTC-related technologies are the underlying foundation to the application of CTCs in precision medicine. Herein, we reviewed previous CTCs-related technologies based on physical and biological properties, and the most recent development of techniques associated with microfluidic-based and nanoparticle-based approaches. Although there are strengths and weakness between different methods, we believe that an effective combination of these techniques may benefit CTCs research in many ways, especially the in-depth analysis and possibility in clinical applications.

CLINICAL APPLICATIONS OF CTCS
Clinically, CTCs are now used as surrogate biomarkers for many solid cancers. Numerous studies have been carried out, mainly in breast cancer, prostate cancer, lung cancer, liver cancer, pancreatic cancer, gastric cancer, and melanoma. Although the clinical guidelines have not included the clinical use of CTCs, besides the inclusion of CTCs as part of the cM0 tumor classification (i.e., no clinical of overt metastasis but the detection of tumor cells in blood), many studies have predicted the great potential of CTCs in clinical applications. In this section, we will mainly present the role of CTCs as biomarkers for diagnosis, prognostication, and therapy monitoring in different cancers (Fig. 4).
Early diagnosis of cancer As a non-invasive method, CTC detection is attractive in assisting cancer diagnosis. Studies of CTCs used for early diagnosis of cancer in the past three years are listed in Table 3. A tumor lesion already has more than 10 9 tumor cells by the time they are detectable in patients using current imaging procedures 15 , such as computed tomography, magnetic resonance imaging, and positron emission tomography. Diagnosis of cancers as early as possible, especially for fast-progressing cancers, is the best way to defeat them. Studies have determined that CTCs are correlated with tumor stage, but the clinical utility of CTCs in cancer detection or even in early cancer diagnosis is still a matter • A need of sample dilution to avoid cell-cell interaction • Quantitative design rules are still lacking for channels Zhang et al. 207 Self-driving micro-cavity array chip • Integrate sample detection structure and vacuum driving system.
• Use the "film-polydimethylsiloxane chip-film" structure and oil sealing method during amplification reaction to minimize water loss.
• Less costly and simple • An excellent linear • Mutational gene profiling of single CTC • Achieve cell loading, lysing, isothermal amplification, and signal read-out on one chip.
• Air diffuse across the thin polydimethylsiloxane walls under some conditions.
Jia et al. 208 Automated and integrated microfluidic method • Integrate a 3D printed off-chip multisource reagent platform, a bubble retainer, and a single CTC capture microchip.
• Achieve CTC capture and identification within 90 m.
• Decrease immunostaining time and antibody consumption by 90%.
• Detect CTCs from various cancers.
• Deficiencies in patient recruitment Wang et al. 209 Integrated microfluidic method • Use magnetic field gradient and immune-fluorescence differences.
• Differentiate eight different subtypes of heterogenic CTCs • Simultaneous on-chip isolation and discrimination of CTCs. CoFe2O4@Ag magnetic nanohybrids • Gold electrodes are modified with MXene nanosheets and an HB5 aptamer is immobilized on the MXene layers • CoFe2O4@Ag magnetic nanohybrids is bonded to the HB5.
• A wide linear range of 10 2 -10 6 cells/mL, low detection limit of 47 cells/mL • Good sensitivity and selectivity in the detection of HER2-posetive cells in blood samples.
• Cost-effective and environmentally friendly • Ag has some aggregations.
• The magnetic property of magnetic probes can affect the isolation of magnetic cells.
Vajhadin et al. 216 Gold nanoparticles Two-dimensional nanozyme with gold nanoparticles • Au nanoparticles-loaded two-dimensional bimetallic PdMo nanozymes assembled with an aptamer composed of a thiolmodified EpCAM Excellent analytical performance of debate. CTCs are considered a surrogate marker of metastatic activity, but whether metastatic dissemination of CTCs in patients occurs early during tumor formation is still controversial. However, in mouse models, early dissemination seeding metastasis has been found in breast 226,227 and pancreatic 228,229 carcinogenesis, indicating that CTC circulation is likely to be a very early event in cancer progression. In Barrière et al.'s study 230 , CTCs were detected in 41% of T1 stage and 47% of axillary lymph nodenegative breast cancer patients, both of which are early-stage breast cancer. In the study by Thery et al. 231 the CTC positivity rate was 21% and 24% in lymph node-negative and positive breast cancer, respectively. Based on this hypothesis, CTCs can be detected earlier before the primary tumor is visible on imaging studies, while the biggest challenge of the application of CTC application in early cancer diagnosis is indeed their scarcity and isolation. The limited sensitivity of CTC detection methods hinders their use as an effective biomarker in early cancer diagnostics.
Evaluation of the cancer prognosis The prognostic value of CTCs has been extensively studied.
CellSearch is the only FDA-approved system for CTC detection used clinically. Based on CellSearch system 46,232-245 , CTCs represent an independent prognostic factor. Studies evaluating other CTC detection systems such as CanPatol, CTC-chip obtained the similar results. CTC enumeration is the main target of investigation, with a cut-off value of ≥5 for positivity, which usually indicates a worse prognosis. It is generally considered that increased CTC counts are correlated with higher likelihood of metastasis and cancer aggressiveness. In a meta-analysis pooling 2239 breast cancer patients including 21 studies 246 , the CTC count before neoadjuvant chemotherapy had a detrimental and decremental impact on patient survival, and patients with one, two, three to four, and five or more CTCs displayed a HR of death (95% CI) of 1.09 (0.65-1.69), 2.63 (1.42-4.54), 3.83 (2.08-6.66), and 6.25 (4.34-9.09), respectively. Furthermore, elevated baseline CTCs levels were associated with inferior survival, the presence of CTC clusters often predicted poor prognosis [247][248][249] , and increasing CTC counts or failure to clear CTCs during treatment was also a prognostic factor for worse survival 234,250,251 . Many studies have found that the molecular phenotypes of CTCs have strong prognostic value. EMT and stemness are the main molecular phenotypes of CTCs studied clinically. CTCs with expression of mesenchymal 252 or stemness 253 -related markers associated with inferior survival. The expression of other molecular markers, such as HER2 46 , CD47 254 , PD-L1 254 also have prognostic implications. Most studies investigate the prognostic value of CTCs at a single timepoint, while intriguingly, some studies have taken CTC dynamics into account. Magbanua et al. 237 developed a novel latent mixture model to stratify groups with similar CTC trajectory patterns during the treatment course, and they found that analysis of serial CTCs can further stratify the patients of poor prognosis into distinct prognostic subgroups. The dynamic changes of CTCs may act as a surrogate prognosis biomarker over the long course of cancer progression. Given the rapid advancements in the accessibility and strengthening of sequencing technologies at the single-cell level, we may expect that in the future, genomic/ transcriptional profiles of CTCs may serve as an outstanding prognostic marker, representing biological information that is more comprehensive and more closely related to prognosis. Studies of CTCs for predicting prognosis in recent three years are summarized in Table 3.
Monitoring of the therapeutic response In many clinical trials, CTCs have been used as useful biomarkers for monitoring cancer treatment responses 234,[255][256][257] , either combined with imaging examinations, serum biomarkers or alone. Researchers prefer to involve CTCs in the evaluation of therapeutic efficiency, in view of the higher sensitivity of CTCs than imaging examination in some cases 9 . As a non-invasive method, the detection of CTCs may also contribute to avoiding frequent     Table 3.
The great potential for CTCs in the clinical application of cancer diagnostics has emerged, although clinically, its use as a surrogate biomarker for cancer screening, treatment monitoring, and prognosis predicting is still limited. Once metastasis occurs, repeat biopsies of metastatic lesions are usually difficult to obtain, and different metastases are heterogenous even in the same patients. CTCs testing using peripheral blood samples is convenient, and may be more representative of the traits of metastatic cells, which are derived from different metastatic lesions in patients. Nevertheless, there is still a lack of guidelines for the clinical use of CTCs, such as a standardized CTCs detecting assay for different cancers, a combination diagnostic scheme with other clinical examinations, and indications for the appropriate timepoints for blood sampling.

DISCUSSION
Studies investigating CTCs have the great potential to reveal the fundamental processes of metastases, including the mechanisms involved in extravasation of CTCs from the primary tumor, how CTCs interact with blood cells to survive in the circulatory microenvironment, and how CTCs intravasate into the distant metastatic site to initiate new lesions. Significant molecular traits of CTCs can greatly contribute to identify targets for antimetastatic therapies. Only a small proportion of CTCs can finally generate metastases, thus studies focusing on these strongly metastatic CTCs may provide deeper insights into CTCs-related therapeutic targets.
Various CTCs detecting technologies have emerged, however, the sensitivity and specificity of these technologies still need to be further improved. Epithelial marker-based CTC detection technology, such as the CellSearch system has opened a new era for CTC analysis and clinical applications, but their drawbacks are rapidly being acknowledged and appreciated by researchers. EMT is a crucial trait of metastatic cancer cells, indicating insufficient capture efficiency of epithelial marker-based CTC detection technology. However, mesenchymal marker-based detection technologies may also be contaminated by non-CTCs, such as tumor-associated fibroblasts and endothelial clusters, which induce the risk of false positivity. Nevertheless, it is intriguing that recent studies have reported that those non-cancerous tumor-derived cells presented in cancer patients are also important surrogate biomarkers for cancer patients 264 . Cancertype-specific molecular markers for CTCs are likely another option, as CTCs of different cancer types possess different molecular markers. However, the sensitivity and specificity of known cancertype-specific CTC markers are not satisfactory. Physical-propertybased CTC detection technologies also have the problem of contamination by non-CTCs, especially for those with similar physical properties as CTCs.
Microfluidic-based and nanotechnology-based CTC detection technologies have become popular in recent years, while the efficiency of these technologies still needs further large-scale clinical validation. High cell detection efficiency and contamination removal capability are the two key strengths of a successful CTC detection technology, while substantial technical optimization of CTC detection is urgently needed to achieve these requirements.
Comprehensive characterization of CTCs is lacking. The limited amount of genomic DNA, RNA, and protein content of CTCs is a bottleneck for exploring their genome, transcriptome, epigenome, and proteome properties. Nonetheless, the emerging genome and transcriptome studies of CTCs have recently profited from the fast-evolving technology of single-cell sequencing, while the proteome studies of CTCs are still elusive due to the very limited technologies for proteome exploring at a singlecell level. However, the study of the CTC proteome is imminent, not only because it can provide a picture of the biological characterization of CTCs, but also because it can help to discover CTC-specific membrane proteins which may help optimize CTC detection.
As for the solid tumor microenvironment, the blood microenvironment around CTCs also plays a significant role in tumor survival and invasion capacity. However, knowledge of the underlying mechanisms behind the survival of CTCs is still limited, as it is a complex process that involves not only shear forces and fluid mechanics but also soluble factors and tumor-associated extracellular vesicles, which are not detailed here. Furthermore, it remains to be confirmed whether CTC clusters are more suitable for interacting with other blood components or adapting to shear forces than single migratory CTCs. If combined with specific biomarkers for strategically detecting CTCs and the interaction of CTCs with associated peripheral blood cells, we could improve the clinical practicability and monitoring power of CTCs by obtaining more comprehensive information on tumor burden and immune status of patients.
Although CTCs have shown initial promise in clinical applications, many challenges must still be overcome before CTC analysis can be widely applied in the clinic. Today, the clinical application of CTCs mainly depends on the analysis of CTC cell enumeration and molecular phenotypes. A more comprehensive characterization of CTCs based on their genome, transcriptome, and proteome with high-throughput sequencing will further benefit clinical application, but also add to the complexity and difficulty of data analysis. CTCs will be a crucial component of "Precision medicine" in the future, as the phenotypic, genotypic, and functional characterization can provide an opportunity to study drug susceptibility that is related to metastasis. The genome and transcriptome analysis of CTCs can unveil potential drug targets. Viable CTCs for drug sensitivity/resistance testing over the therapy course can guide precision medication. However, the culture of CTCs is very challenging: (1) limited methods are available to isolate viable CTCs, which also yield low numbers of CTCs and (2) a favorable circulatory microenvironment for CTC survival is difficult to mimic. Very limited CTC-derived cell lines from cancer patients have been established. Optimization of CTC culture conditions will be needed.
Furthermore, CTCs, circulating tumor DNA (ctDNA), and exosomes are all present in liquid biopsy samples. An exploration of the advantages and disadvantages of each substrate present in the liquid biopsies, and how to better incorporate them into clinical application is needed to achieve more precise diagnoses. Among liquid biopsy methods, CTCs have tremendous advantages, as isolated CTCs can be viable, which can optimize CTCderived explants or three-dimensional organoid cultures for functional testing or for drug-screening assays. The study of CTCs is attractive, and CTC detection may likely become an essential component of cancer management in the future. As the picture becomes clearer, we are fully confident about the promising potentials of CTCs.

ADDITIONAL INFORMATION
Competing interests: The authors declare no competing interests.